Synchronous Oscillations in the Basal-Ganglia-Cortical Network: Do They Generate Tremor and Other Symptoms of Parkinson's Disease?

Transcription

1 Synchronous Oscillations in the Basal-Ganglia-Cortical Network: Do They Generate Tremor and Other Symptoms of Parkinson's Disease? Thesis submitted for the degree of Doctor of Philosophy by Michal Rivlin-Etzion Submitted to the Senate of the Hebrew University of Jerusalem April 2009

2 This work was carried out under the supervision of Prof. Hagai Bergman Dr. Yifat Prut

3 Abstract ABSTRACT The main manifestations of Parkinson's disease (PD) are akinesia (poverty of spontaneous movements and difficulty in initiating a movement), bradykinesia (slowness of movement), muscle rigidity and a tremor of 4-7 Hz. The two major subtypes of the human disease present the akinetic/rigid symptoms with and without tremor (tremor dominant vs. akinetic/rigid dominant subtype, respectively). The major cellular event leading to PD is the death of midbrain dopaminergic neurons, resulting in dopamine depletion in the striatum, the input nuclei of the basal ganglia (BG). PD can be investigated using its primate model MPTP treated monkeys in which dopaminergic neurons are destroyed by the neurotoxin MPTP. As a result, the monkeys develop motor symptoms similar to those seen in PD patients. Interestingly, monkeys develop the Parkinsonian tremor as a function of their species African green (vervet) monkeys tend toward tremor, while macaques do not. Recordings from the globus pallidus (GP) and other BG nuclei in MPTP treated monkeys as well as in human patients have shown that Parkinsonian symptoms are accompanied by the appearance of periodic oscillations of neuronal activity. The frequency of these oscillations ranges from 4 to 15 Hz, but their relationship to tremor or other Parkinsonian symptoms is still under debate. Intuitively, tremor is considered to be the result of the low-frequency (4-7 Hz) oscillations. However, this conjecture has not been validated and has even been challenged by several studies. In my study, I attempted to elucidate the connection between Parkinsonian symptoms and neuronal activity in the MPTP primate model. To address the issue of a causal link between BG oscillations and the tremor I first developed spectral analysis tools that enable reliable detection of neuronal oscillations. Spectral analysis of neuronal spike-trains is an important tool in understanding the characteristics of neuronal activity because it provides insights into normal and pathological periodic oscillatory phenomena. However, the refractory period of the spike generation processes creates high-frequency modulations in the spike-train, since any increase in the discharge rate causes a decrease in subsequent time-bins, leading to multifaceted modifications in the structure of the spectrum. Thus, the power spectrum of the spiking activity (auto-spectrum) displays elevated energy in high frequencies relative to the energy of lower frequencies. The spectral distortion is more 1

4 Abstract dominant in neurons with high firing rates and long refractory periods (as is the case for the neurons in the GP) and can lead to reduced identification of low-frequency oscillations (such as the Parkinsonian 5-10 Hz oscillations). In the first chapter of this thesis I develop a compensation process that uses shuffling of inter-spike intervals (ISI). This ISI shuffling method enables reliable identification of oscillations in the entire frequency range. In the next step I conducted experiments in which I recorded simultaneously from the GP and the arm related primary motor cortex (MI) of two vervet monkeys using multiple electrodes. I used electrical stimulation to mimic the Parkinsonian oscillations in MI or GP while recording the evoked neuronal and muscles responses. The periodic stimulations were given at different frequencies ranging from 0.2 Hz to 15 Hz, in order to include the range of the oscillatory activity of the basal ganglia in the Parkinsonian state. The experiments were performed before and after MPTP injection and induction of Parkinsonism. The results of this set of experiments demonstrated that the functional connectivity between MI and GP is greatly enhanced following MPTP treatment, and micro-stimulation of MI or GP significantly modulates the discharge rate in the other structures in MPTP, but not in the normal state. In addition, the BG-cortical-muscle loops exhibited low-pass filter properties. MI neurons responded to the 1-2 Hz GP stimulations but not to the higher frequency stimulations, which are encountered in the Parkinsonian brain. Finally, muscle activation evoked by MI micro-stimulation was markedly attenuated at frequencies higher than 5 Hz. To further investigate the role of neuronal oscillations in generating tremor, I analyzed the spontaneous activity of six MPTP-treated monkeys, three tremulous vervets and three non-tremulous macaques (I analyzed data collected by Dr. Gali Heimer, Mr. Shlomo Elias and myself). My goal was to quantify the amount of oscillations in each monkey, and to investigate correlations with the appearance of tremor. I found that spontaneous activity in the GP in the Parkinsonian state became oscillatory in all monkeys, but that the distribution of the frequencies of oscillations varied as a function of the tremor non-tremulous monkeys tended to develop oscillations mainly at ~5 Hz, while tremulous monkeys demonstrated oscillatory activity at ~10 Hz as well. Interestingly, in the two monkeys whose cortical activity was recorded, only one developed 10 Hz oscillations in MI, although both monkeys had a similar tremor rank. 2

5 Abstract These results are consistent with the low-pass filtration properties of the BGcortical loop which were found through electrical stimulation. Based on my findings, I suggest that the Parkinsonian tremor is generated by GP 10 Hz oscillations, which are transformed via the direct BG projections to the brainstem, rather than through the BGcortical loop. The main core negative Parkinsonian symptoms, akinesia and bradykinesia, may derive from the increased synchrony in the BG-cortical loop, rather than from the oscillations. These working hypotheses could lead to a novel approach to the generation of PD symptoms which differentiates the akinesia-bradykinesia hypomotoric symptoms from the hyper-motoric tremor. Future closed loop deep brain stimulation algorithms thus should be directed toward ameliorating BG abnormal synchrony rather than GP 10 Hz oscillations. 3

6 Table of Contents TABLE OF CONTENTS ABSTRACT... 1 TABLE OF CONTENTS... 4 INTRODUCTION... 5 METHODS RESULTS I. DETECTION OF NEURAL OSCILLATIONS: THE SHUFFLING METHOD...21 II. LOW PASS FILTER PROPERTIES OF BASAL GANGLIA-CORTICAL-MUSCLE LOOPS...34 III. BASAL GANGLIA OSCILLATORY ACTIVITY AND PATHOPHYSIOLOGY OF PARKINSON'S DISEASE...52 IV. HIGH FREQUENCY OSCILLATIONS IN THE GLOBUS PALLIDUS ARE CORRELATED WITH PARKINSONIAN TREMOR...62 DISCUSSION BIBLIOGRAPHY APPENDIX I. OSCILLATORY ACTIVITY FOLLOWING DOPAMINE REPLACEMENT THERAPY...95 II. PARKINSON'S DISEASE: FIGHTING THE WILL?

7 Introduction INTRODUCTION Clinical aspects of Parkinson's disease and the MPTP primate model In 1817, almost two hundred years ago, the English physician James Parkinson wrote an Essay on the Shaking Palsy providing the first clinical description of the motor symptoms of the disease now bearing his name (Parkinson, 1817). Today, Parkinson's disease (PD) is the most common basal ganglia movement disorder, and affects from 1% to as much as 4-5% of those in the 65 and 85 year age brackets, respectively (Van Den Eeden et al., 2003). PD is characterized by poverty of movement on one hand, including the difficulty in initiating a movement (akinesia), and slowness of movement (bradykinesia); and involuntary movement on the other hand reflected as a low frequency tremor at 4-7 Hz, mostly at rest but also postural. In addition, PD patients exhibit muscle rigidity, but the classification of this symptom as hyper- or hypo- motoric remains to be established. The most consistent neuropathological finding as regards PD is the death of midbrain dopaminergic neurons, resulting in dopamine depletion in the input nuclei of the basal ganglia, the striatum (Jellinger, 1987). Although the basal ganglia are thought to be involved in processing motor actions, it is unknown how dopamine depletion leads to the contradictory symptoms of the disease. There is considerable variability among patients in the expression and extent of each symptom. The two extremities of the clinical spectrum are made up of predominantly tremulous patients and akinetic-rigid mostly non-tremulous patients. PD patients also display other symptoms such as postural instabilities, as well as depression, dementia, psychosis and other cognitive changes (Paulus and Jellinger, 1991;Fahn et al., 1998). Although there have been some reports of correlations between some clinical phenomena in patients and certain pathological findings (Paulus and Jellinger, 1991;Hirsch et al., 1992;Jellinger, 1999), the pathological and neural basis of the wide spectrum of clinical expressions of PD is still unknown. One of the most significant breakthroughs in research on PD was the discovery of 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP). This neurotoxin produces a model of the disease in primates that closely resembles the disease in humans pathologically, clinically, and pharmacologically. MPTP causes permanent degeneration 5

8 Introduction of midbrain dopaminergic neurons, mainly from the substantia nigra pars compacta (SNc), and thus is a stable and specific model of the human disease (Langston, 1987). MPTP treated monkeys exhibit most of the major motor symptoms mentioned above and also respond to conventional dopamine replacement treatment (Crossman et al., 1987). However, similar to human patients, not all primate species develop the classical resting tremor of PD. Macaques tend to be akinetic-rigid and rarely demonstrate a lowamplitude, high frequency (10 12 Hz) action-postural tremor, whereas vervet monkeys usually develop a high-amplitude, low-frequency (4 7 Hz) tremor beyond their akinesia and bradykinesia. These two variants of the primate model probably correspond to the non-tremulous and tremulous variants of human Parkinsonism. This match not only reaffirms the relevancy and accuracy of the primate model, but can also be crucial in tracing the relevant neuronal activities that are responsible for the tremor (Bergman et al., 1998b). Different therapeutic strategies in PD and their shortcomings. The wide spectrum of therapeutic strategies for PD can be divided into pharmacological versus surgical treatments. The conventional symptomatic medications used clinically are based on various dopamine precursors, agonists and factors intervening in the metabolism of dopamine. The most common and effective is levodopa - a dopamine precursor. Postsynaptic dopamine receptor agonists, although somewhat less potent, are also widely used, either as mono-therapy in the early stages or along with levodopa in progressive PD. In contrast to levodopa, the post synaptic agonists have the advantage of not requiring any residues of dopaminergic neurons and a larger time constant. Two main problems still cloud the horizon of conventional medications (Schrag and Quinn, 2000). The first is a cluster of side effects of involuntary movements referred to as dyskinesia. These are mainly constituted of chorea, involuntary jerks of the hands and legs, and dystonia, and classically appear during the on period after several years of daily dopamine replacement treatment. The second problem is the wearing off phenomenon, which is a subsiding of antiparkinsonian effects after chronic treatment and the need to constantly increase the doses, thus also usually worsening the dyskinetic side effects. 6

9 Introduction Surgical procedures for PD were developed between 1940 and With the introduction of levodopa surgical procedures were almost entirely abandoned. However, motor complications induced by levodopa and advances in surgical techniques led to a resurgence of surgical therapies in the 1990s. The most common surgical procedure today is chronic high frequency stimulation to basal ganglia nuclei, the subthalamic nucleus (STN) or the internal segment of the globus pallidus (GPi). This deep brain stimulation (DBS) can significantly reduce akinesia, rigidity and tremor, and is usually administered to advanced PD patients, as it is quite effective in reducing levodopa induced dyskinesia as well (Rodrigues et al., 2007;Valldeoriola et al., 2007;Melamed et al., 2007). Even though surgical procedures have become more common, the exact neural mechanisms that underlie the therapeutic effect of DBS are still unclear (e.g. (Perlmutter and Mink, 2006)). One of the goals of this thesis is to enhance our understanding of these mechanisms, which can improve the choices of DBS target nuclei as well as its stimulation pattern. Information processing within the basal ganglia The basal ganglia are subcortical nuclei that are related to many functions, including motor action, cognition, learning and motivation (Gerfen and Wilson, 1996;Middleton and Strick, 2000). Understanding the information processing within these nuclei in the normal state is important for defining their role in each of these processes, and is crucial for the investigation of the neuronal substrate of Parkinson s disease. The basal ganglia are comprised of the striatum, globus pallidus, subthalamic nucleus (STN) and substantia nigra (Gerfen, 2004;Haber and Gdowski, 2004). The striatum, which receives projections from all cortical areas, is the main input nucleus of the basal ganglia (Parent and Hazrati, 1995). Midbrain dopaminergic neurons, mainly from the SNc, have a modulatory effect on information flow from the cortex to the striatum (Centonze et al., 1999;Reynolds et al., 2001;Shen et al., 2008). From there the information flows either directly or indirectly through the STN and the external segment of the globus pallidus (GPe), to the output structures of the basal ganglia the internal segment of the globus pallidus (GPi) and the substantia nigra pars reticulata (SNr) (Parent and Hazrati, 1995). The GPi and SNr have two main targets: the ventral motor nuclei of the thalamus, which transmits information to the frontal cortex and closes the 7

10 Introduction Cortic-Baso-Thalamo-Cortical loop, and brainstem motor centers, which project to the spinal cord (Parent et al., 1999). The enigma surrounding the basal ganglia's function has led to the development of several models of information processing in these structures. According to the classical "box and arrow" model (Albin et al., 1989;DeLong, 1990;Mink, 1996b) different dopamine receptors (D1 or D2) are localized on different striatal populations that give rise to the direct and indirect pathways (Gerfen et al., 1990;Surmeier et al., 2007). The direct and the indirect pathways are thought to have a net effect of inhibition and excitation of basal ganglia output nuclei, respectively. Under normal conditions, striatal neurons projecting directly to GPi appear to be facilitated by dopamine actions on D1 receptors, whereas neurons projecting to GPe are inhibited by dopamine actions on D2 receptors. This model assumes that dopamine depletion in the striatum should therefore lead to both a reduction in activity of the direct inhibitory pathway and an increase in activity of the indirect excitatory pathway, synergistically leading to an increase in the activity of the output nuclei of the basal ganglia GPi and SNr (DeLong, 1990). Because the GPi/SNr thalamic projections are GABAergic and inhibitory, an increase in their discharge leads to the inhibition of thalamo-cortical neurons which results in a reduction of cortical activation that accounts for the hypokinetic manifestations of PD. The box and arrow model considers that all of the neurons in the output structures exhibit similar activity (decrease or increase their firing rate). However, experimental studies have shown that these neurons respond to behavioral events by either discharge decreases or discharge increases and even by complex discharge patterns (e.g., a biphasic modulation of discharge) (DeLong, 1971;Jaeger et al., 1995;Turner and Anderson, 1997;Handel and Glimcher, 2000;Gdowski et al., 2001;Sato and Hikosaka, 2002). Moreover, the model predicts changes in the firing rates of basal ganglia nuclei as a result of dopamine depletion. Nevertheless, a number of experimental studies have failed to demonstrate the expected rate changes, particularly within GPi neurons (Levy et al., 1997;Boraud et al., 1998;Raz et al., 2000b). In addition, the mean discharge rate of the primary motor cortex does not change following MPTP (Doudet et al., 1990;Watts and Mandir, 1992;Goldberg et al., 2002d), as opposed to the expected inhibition. Other anatomical studies have revealed that the D1 and D2 receptors are colocalized on striatal projection neurons (Aizman et al., 2000). Moreover, single neuron 8

11 Introduction labeling has failed to identify direct-pathway striatal neurons that project only to the GPi (Bolam et al., 2000). Finally, the anatomy of the basal ganglia seems to be more complex than the description in the model because of the back projections of the GPe to the striatum, and the feed-forward projections from the cortex to the GABAergic interneurons of the striatum (Bevan et al., 1998;Kita et al., 1999). These findings prompted researchers to put forward more complex models in which the interplay within the nuclei was also incorporated. According to the "action selection" model proposed by Mink (Mink, 1996a) the basal ganglia act during voluntary movements to inhibit competing motor mechanisms. Our group proposed the "reinforcement-driven dimensionality reduction" model (Bar-Gad et al., 2000a), according to which the basal ganglia compress information received from the cortex based on a reinforcement signal which is conveyed by the dopaminergic cells. The most conspicuous anatomical finding that supports this model is the strong reduction in the number of neurons through the cortico-striatal-pallidal circuit, from ~ 10 9 cells in the cortex to ~10 7 projection cells in the striatum and ~10 5 neurons in the pallidum (of the rhesus monkey). All three models mentioned above, and most other information processing models of the basal ganglia ignore the projections of basal ganglia output nuclei to brainstem motor centers and focus on the Cortic-Baso-Thalamo-Cortical loop. The existing models do not provide a comprehensive explanation either for the various changes in neural pattern that occur with the disease, primarily the appearance of synchronous oscillations (see below), or for the process that underlies these patterns to generate the antithetical symptoms of the disease. Taking into consideration both pathways could elucidate the role of basal ganglia in health and in disease. Parkinsonian neuronal patterns Periodic oscillations Recordings from different nuclei of the basal ganglia in MPTP treated monkeys as well as PD patients reveal that Parkinsonian symptoms are accompanied by several changes in neuronal activity, including the emergence of periodic burst-oscillations, often reported to be synchronized (Miller and DeLong, 1987;Filion and Tremblay, 1991;Bergman et al., 1994;Nini et al., 1995;Hutchison et al., 1997;Mandir et al., 1997;Raz et al., 2000c;Heimer et al., 2006d). The frequencies of these oscillations are 9

12 Introduction bimodally distributed around 5 (about tremor frequency) and 10 Hz. This oscillatory activity can generate the motor symptoms of the disease and particularly the tremor. However, despite the existence of neurons with tremor-frequency activity ("tremor cells") in the GP of PD patients (Hutchison et al., 1997), significant coherence between GPi activity and the peripheral tremor hardly ever occurs (Lemstra et al., 1999a). In another work where pallidal activity was recorded in a single patient undergoing pallidotomy, only one tremor cell that showed coherent activity with a peripheral muscle was found, and the coherence was intermittent (Hurtado et al., 1999b). Moreover, as opposed to previous reports (Levy et al., 2000) oscillatory activity is found in tremulous as well as non-tremulous patients (Weinberger et al., 2006a). In MPTP treated monkeys, neural oscillations are also found in species of monkeys that tend to develop infrequent, short episodes of high-frequency tremor (Miller and DeLong, 1987;Bergman et al., 1998a). Interestingly, the percentage of cells with low-frequency oscillations in the STN was reported to persist after STN lesioning or even increase in MPTP treated monkeys (Wichmann et al., 1994). Even though PD symptoms are usually thought to originate in the BG, due to dopaminergic loss, the Parkinsonian tremor may not be generated by oscillations in the basal ganglia, but rather have a cortical or spinal origin. Indeed, a magnetoencephalography study demonstrated that PD patients motor and somatosensory cortices also tend to oscillate in correlation with the tremor (Alberts et al., 1969;Volkmann et al., 1996). Another study in which single units activity were recorded in MPTP treated monkeys revealed enhanced non-oscillatory synchronization among motor cortical neurons (Goldberg et al., 2002c). Nonetheless, the local field potential of the same cortical recordings demonstrated global oscillatory activity at 10 Hz (Goldberg et al., 2004). The Parkinsonian tremor may also be generated in the thalamus, which contains "tremor cells" often recorded in PD patients. These thalamic oscillations were found to be coherent with the tremor (Lenz et al., 1988). However, it should be recalled that oscillatory activity in any neuronal structure may merely reflect the sensory ascending pathways from the periphery, or an oscillatory activity originating in another neuronal structure, and does not necessarily indicate the origin of the tremor. 10

13 Introduction Enhanced correlation The other facet of neuronal activity in Parkinsonism that has recently attracted growing attention is the study of inter-neuronal relations. In contrast to the independent neural activity of the normal monkey's brain, after induction of Parkinsonism up to 50% of recorded pallidal neurons have correlated activity (Nini et al., 1995;Bergman et al., 1998a). A high level of correlation was also found between striatal TAN and pallidal cells in MPTP monkeys (Raz et al., 2001). Similar results have also been found in Parkinsonian patients undergoing stereotaxic surgery (Hurtado et al., 1999a;Levy et al., 2000). These data support the dimensionality reduction theory (Bar-Gad et al., 2000a). The absence of adequate dopaminergic modulation causes the segregation of the circuitry to break down, thus enabling the neurons to become more synchronized with one another. According to the "action selection" model, such synchronization may damage the selection of a single action while inhibiting all other competing actions processed by the basal ganglia (Mink, 1996c), leading to the poverty of movement of the disease. However, the existence of synchronization could be an independent feature of Parkinsonism, or merely a byproduct of the tremor or self-regulating oscillators with similar frequencies. Whether or not increased inter-neuronal synchronization is a genuine independent phenomenon in Parkinsonism is an imperative issue to probe if we are to understand the pathophysiology of the disease; for this reason it is one of the goals of this thesis. Analysis tools Spectral methods are best suited for analyzing phenomena that display rhythmic behavior such as neuronal periodic oscillations (Gray, 1994;Gauss and Seifert, 2000;McCormick, 2002;Brown, 2003). The spike train of a neuron is a point process frequently modeled as a Poisson process. The power spectrum of a Poisson process exhibits equal power in all frequencies, therefore facilitating the detection of any oscillatory processes embedded within it. Nevertheless, a spike train is never a true Poisson process due to the refractory period which prevents the neuron from firing two successive spikes within a very short interval. Consequently, the distribution of the power spectrum of a typical spike train is not homogeneous: the power at high frequencies (>100Hz) is close to that of a Poisson process with the same average firing probability (Bair et al., 1994b;Franklin and Bair, 1995), but the power in the lower frequencies is significantly lower than expected. This distortion in 11

14 Introduction the spectrum of neuronal firing impedes the identification of low frequency oscillations, such as Parkinsonian pallidal oscillations in the range of 5-10 Hz. This issue has been neglected in many previous studies (e.g., (Raz et al., 2000d;Raz et al., 2001)), which have used ad-hoc techniques to identify oscillatory neurons, often depicting only a limited range of the spectrum, or choosing to work in the time domain. However, the refractoriness also distorts the auto-correlation and crosscorrelation functions (Bar-Gad et al., 2000b). The contradicting reports regarding the percentages of oscillations in single neurons as well as in pairs of neurons (Lemstra et al., 1999b;Wichmann et al., 1999;Raz et al., 2000e;Goldberg et al., 2002b;Levy et al., 2002) emphasize the need for a uniform detection method of oscillatory activity. One of the goals of this thesis is therefore to develop a robust method for the detection of neural oscillatory phenomena and to apply this method to basal ganglia activity recorded (by myself and others) in MPTP treated primates. Goals and specific aims The role of basal ganglia oscillatory activity in generating the Parkinsonian symptoms and especially the tremor has been studied intensively, but remains hotly debated. One explanation for the inconsistency in the amount of oscillatory activity in human patients as well as MPTP primates lies in the variety of symptoms in each individual. In addition, the appearance of 5 and 10 Hz oscillations could be separate phenomena, and different symptoms of the disease could be related to different ranges of frequencies. However, the establishment of a standardize method to detect the oscillations is crucial to answer these questions. This motivated my first goal: 1. To establish a standardized method to enable uniform detection of oscillations in a single neuron as well as in pairs of neurons. The lack of a uniform method for detecting oscillatory activity explains the range of techniques used in previous studies, some of which did not take the distortion in the neuronal spectrum into account. This distortion stems from the refractory period of the cells. This may have skewed the percentages of oscillatory activity reported. The technique I developed is based on spectral tools and takes the refractoriness of the neurons into consideration. 12

15 Introduction Once this technique had been developed I had a robust tool to examine the incidence of neuronal oscillations in the dopamine depleted state and their relation to PD symptoms. I used two approaches: intervention and observation. The intervention approach involved delivering electrical stimulations to the armrelated primary motor cortex (MI) or to the GP, while recording the evoked neuronal and motor responses. In the synthetic stimulation pattern I tried to mimic the oscillatory and bursty pattern often encountered in the Parkinsonian brain. The periodic stimulations were given at different frequencies, ranging from 0.2 Hz to 15 Hz, in order to include the 5 and 10 Hz oscillatory activity of the basal ganglia in the Parkinsonian state. This allowed me to trace the transformations of these patterns within the Cortico-Baso-Thalamo-Cortical loop, as well as from this circuit to the muscles, with the intent of revealing the underlying information-processing architectures and computational principles. The experiments were carried out on two vervet monkeys before and after MPTP treatment and induction of Parkinsonism. In the observation approach I analyzed the spontaneous activity recorded from the GP of six MPTP-treated monkeys in our lab, three tremulous vervets and three nontremulous macaques (the data were collected by Dr. Gali Heimer, Dr. Shlomo Elias and myself). I implemented the spectral method I had developed on the data of all six monkeys. This relatively high number of monkeys (usually studies on MPTP primates involve 2 subjects), and the fact that they demonstrated variation in their symptoms, constituted a good database to reliably examine the relationship between neuronal oscillations and Parkinsonian symptoms. In addition, in two of the tremulous monkeys I recorded the spontaneous activity in MI and GP simultaneously. These data provided important insights on information-processing in the Cortico-Baso- Thalamo-Cortical loop during spontaneous activity, particularly as regards the transfer function from GP to MI and vice versa. Using these two approaches, following goals were defined: 2. To elucidate the causal relationship between basal ganglia oscillatory activity and the clinical symptoms of PD, especially tremor. 3. To determine the role of the Cortico-Baso-Thalamo-Cortical circuit in transferring basal ganglia abnormal activity to the muscles. The three cardinal motor symptoms of PD are tremor, rigidity and akinesia. These symptoms do not necessarily originate from a single source. For instance: the 5 and 13

16 Introduction 10 Hz oscillations could be related to different symptoms of the disease; namely, low frequencies could be "translated" into tremor, whereas higher frequencies would be translated into rigidity. Moreover, different PD symptoms could result from different pathways from basal ganglia output nuclei. As mentioned above, most information processing models of the basal ganglia emphasize the role of the Cortic-Baso- Thalamo-Cortical loop in generating the symptoms of the disease. However, the projections of basal ganglia output nuclei to brainstem motor centers might have a crucial function in generating certain PD symptoms. Specifically, I attempted to answer the following questions: Can periodic oscillations delivered using electrical stimulation to MI and/or GP generate a peripheral tremor? What is the relationship between the frequency of neural oscillations in these structures and the frequency of tremor? Do all monkeys, tremulous as well as none tremulous, exhibit GP oscillations? Do the frequencies of these oscillations change as a function of the symptoms manifested in the animal? Does MI oscillate in the MPTP state? What are the frequencies of oscillations? Do GP electrical-evoked oscillations transform to MI and vice versa? What is the transfer function between the structures? Is the rigidity seen in PD actually a result of neuronal oscillations at high frequencies? Answers to these questions could suggest a specific structure that might indicate the origin of tremor. In addition, they could determine whether MI is a potential target in treating Parkinsonian symptoms, as was recently suggested (Drouot et al., 2004). 4. To characterize the functional connectivity between M1 and GP before and after induction of Parkinsonism. I tested the hypothesis that in the dopamine depleted state, GP neurons are more easily driven by MI stimulation. Such an enhancement in functional connectivity could generate the non- periodic symptoms of the disease such as akinesia. Specifically I attempted to answer the following questions: Is there synchrony between MI and GP? What are the differences between the normal and MPTP-induced state in functional connectivity? 14

17 Introduction Are there differences between the normal and MPTP-induced state in frequency transformation of oscillations? I believe that the findings reported in this dissertation provide a better understanding of the neuronal activity that underlies Parkinson's disease, and that this knowledge could guide the development of the next generation of PD treatments. 15

18 Methods METHODS Animals and behavioral paradigm My research was based on data collected from six monkeys: three vervets (African green, Cercopithecus aethiops aethiops, T, W and C; females; weighing 3, 3.5 and 3.8 kg respectively) and three macaques (one Macaca mulatta, R; female; weighing 5.7 kg, and two Macaca fascicularis, H and P; females; weighing 3.2 and 3 kg). Before any of the other procedures were carried out, the monkeys were trained to sit in a primate chair, to permit handling by the experimenter, and become familiarized with the laboratory setting. Monkeys C and R were trained to perform a simple visuomotor task (Heimer et al., 2006c). Monkey H was trained for a self-initiated probabilistic delayed visual-motor task (Arkadir et al., 2004;Morris et al., 2004). However, most of the recordings in Monkeys C, R and H were conducted during a "quiet-wakeful" state. None of the other monkeys were engaged in any behavioral task, and the animals were trained to sit quietly in the primate chair. The data from monkeys C and R were collected by Dr. Gali Heimer, Dr. Shlomo Elias recorded from monkeys H and P and I carried out the experiments on monkeys T and W, which form the basis of this dissertation. The monkeys' health was monitored by a veterinarian, and their weights and clinical status were checked daily. All experimental protocols were performed in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and the Hebrew University guidelines for the use and care of laboratory animals in research and were approved and supervised by the Institutional Animal Care and Use Committee. Surgical and recording procedures After training, a recording chamber was attached to the monkeys' skulls. The recording chamber was tilted laterally in the coronal plane in all monkeys except monkey H where it had a 0 tilt (and therefore permitted recordings from both hemispheres). All chambers were targeted by a stereotaxic device to cover most of the GP territory, and in monkeys T and W the position of the chambers allowed access to the arm- related area of the primary motor cortex (MI) as well. The exact position of the chamber was established using a magnetic resonance imaging (MRI) scan and electrophysiological mapping. Multiple extracellular recordings were carried out in all 16

19 Methods monkeys. GP activity was recorded using 8 electrodes in monkeys C and R and using 4 electrodes in monkeys H and P. In monkeys T and W, I used simultaneous recordings of the GP and MI (4 electrodes in each structure). Details of the surgery, identification of neurons, and data-recording methods are given in the Methods section of Results II and IV. After a period of recording in the normal state, Parkinsonism was induced by five intramuscular injections of 0.4 mg/kg of 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine- HCl (MPTP, Sigma, Rehovot, Israel). The MPTP injections were given under light intramuscular ketamine hydrochloride (10 mg/kg) anesthesia and over a period of 4 days (3 injections in the first 24 hours). The clinical state of the monkeys was assessed daily according to a primate scale of Parkinsonism (Benazzouz et al., 1995) and then through a subjective summary by the main experimenter. Upon termination of the recording days in the MPTP state the monkeys were treated with dopamine replacement therapy to confirm the diagnosis of Parkinsonism by the clinical improvement produced by dopamine-replacement treatment. Exact doses for each of the animal are given in the Methods section of Results IV. Monkey W died before it was given any dopaminergic treatment. At the end of the experiment, the monkeys were killed with a lethal dose of pentobarbital and perfused through the heart with saline followed by a 4% paraformaldehyde solution. The brains of the monkeys were removed, serial sections of 50 µm were cut on a freezing microtome and every 12 th section was processed for Nissl or tyrosine hydroxylase (TH) immunohistochemistry with the kind assistance of Dr. Suzanne Haber (U. of Rochester). Monkey W was perfused in a similar way within 30 min of her death. Accelerometers In monkeys C, R, T and W We used a uniaxial accelerometer (ACC) (8630C5; Kistler, Amherst, NY) to assess limb tremor and hand movements. The monkeys had the accelerometer fastened to the back of their non-restrained left wrist (contra-lateral to the stimulating hemisphere). For details regarding ACC recordings and analysis see the Methods section of Results II and IV. 17

20 Methods Data analysis of spontaneous recordings Cells were selected for recording as a function of their signal-to-noise ratio and real-time assessment of their isolation quality. Only cells with a minimum of 5 minutes of recording were included in the study. Details regarding inclusion criteria are given in the Methods section of Results II and IV. We conducted quantitative analyses of oscillatory firing patterns of all cells included in the study. The oscillatory activity of pallidal neurons was assessed using the shuffling method in order to compensate for the spectral distortion caused by the refractory period in neurons with a high discharge rate. A full description of the method, including the rationale as well as the simulations of spike trains we used for testing are given in the first chapter of the Results. Briefly, the spectrum of the original spike train was divided by the mean spectrum of the locally (T=~175 ms) shuffled spike trains (n = 20). A confidence level (p < 0.001, normalized to the total number of bins) for the compensated spectrum was constructed based on the high-frequency range of Hz, at which the spectrum was flat. A cell was considered oscillatory if its compensated spectrum contained at least two consecutive bins within the range of 4 15 Hz that crossed the p = confidence level. Due to the low discharge rate of cortical neurons (Goldberg et al., 2002a;Rivlin- Etzion et al., 2008), oscillatory activity in the cortex could be evaluated based on the original power spectral density (PSD). A confidence level (p < 0.001, normalized to the total number of bins) was constructed based on the high-frequency range of Hz, at which the spectrum was flat. A cell was considered oscillatory if its PSD contained at least two consecutive bins within the range of 4 15 Hz that crossed the p = confidence level. For the analysis of synchronous oscillations we used pairs of neurons with overlapping periods of at least 5 minutes of stable recordings. Only neuronal pairs that were recorded by different electrodes were included to avoid possible artifacts attributable to a shadowing effect of high discharge rates in cells recorded from the same electrode (Bar-Gad et al., 2001). Synchronous oscillations within GP pairs, as well as between MI and GP neurons, were assessed using the shuffling method (specified in Results I) due to the GP high discharge rate. In short, the cross-spectrum of the original spike trains was divided by the mean cross-spectrum of the globally shuffled (n = 20) spike trains. A confidence level (p < 0.001, normalized to the total number of bins) for 18

21 Methods the compensated spectrum was constructed based on the high-frequency range of Hz, at which the spectrum was flat. A correlogram was considered to have significant periodic oscillations if its compensated spectrum contained at least two consecutive bins within the range of 4 15 Hz that crossed the p < confidence level. For cortical pairs, we used as a confidence level the conventional significance L 1 criterion for the coherence function (Bloomfield, 1976;Brillinger, 1981): 1 (1 α ), where α is the level of confidence (here α=0.999), and L is the number of windows used in the calculation (length of the data divided by the window size, which in our case was 4096). The same threshold was applied when comparing the neural activity for all monkeys and for all clinical states. 1 Electrical stimulation In monkeys T and W, I recorded the neural responses to electrical stimulations delivered in MI or in GP. After recording the spontaneous neuronal activity the stimulation pattern was given through one or more electrodes (usually one, details given in the Methods section of Results II), whereas all other electrodes were used for recording. All electrodes used in this study were of the same type (glass coated tungsten micro-electrodes) and could be used for either stimulation or recording. The electrodes used for stimulation within a certain session were connected to the current source, and were not used for recording during the same session. However, the same electrode could serve for stimulation in one session and for recording in the next session. The switch between stimulation and recording mode was controlled electronically (MIP, Alpha- Omega Eng.). The amplification gain of the recording electrodes was reduced from 5000 to 5 during the stimulation pulses (0.3 ms before until 0.1 ms after the pulse) to reduce the saturation of the recording devices and to minimize the stimulus artifact. Because of possible current leakage from the GPi to GPe (or vice-versa) that can occur when stimulating at the borders of these nuclei (Ranck, 1975), I did not attempt to distinguish between stimulating sessions in the GPe and GPi. For each stimulating electrode I used current amplitudes of 40 µa in monkey T, and either 40 or 60 µa in monkey W (detailed percentages are given in the Methods section of Results II). When stimulating through 19

22 Methods more than one electrode, all the stimulating electrodes were connected to the same current source and the amplitude of the current was multiplied by the number of stimulating electrodes. I assumed an equal current distribution through the electrodes due to their similar impedance. Stimulation pattern: Each burst contained 8 biphasic symmetric pulses (each phase 0.2 ms, with negative pulse leading) given at 200 Hz, leading to bursts of 35 ms in length. The stimulation bursts were administered at different frequencies: 1, 2, 5 and 10 Hz, and in addition, 15 Hz for monkey W only. The stimulation in each of the frequencytests lasted 20 seconds, (resulting, for example, in 20 bursts for the 1 Hz stimulation and 300 bursts for the 15 Hz stimulation), with an inter-stimuli-interval of 15 seconds between the different frequency-tests. Detailed descriptions of the stimulation pattern and its analyses are given in the Methods section of Results II. 20

23 Results I RESULTS I. DETECTION OF NEURAL OSCILLATIONS: THE SHUFFLING METHOD Article information: Rivlin-Etzion M, Ritov Y, Heimer G, Bergman H, Bar-Gad I (2006). Local shuffling of spike trains boosts the accuracy of spike train spectral analysis. Journal of Neurophysiology 95:

24 J Neurophysiol 95: , First published January 11, 2006; doi: /jn Results I Innovative Methodology Local Shuffling of Spike Trains Boosts the Accuracy of Spike Train Spectral Analysis Michal Rivlin-Etzion, 1,2 Ya acov Ritov, 1,3 Gali Heimer, 2 Hagai Bergman, 1,2,4 and Izhar Bar-Gad 5 1 Center for Neural Computation, 2 Department of Physiology Hadassah Medical School; 3 Department of Statistics, 4 Roland Center for Neurodegenerative Diseases, The Hebrew University, Jerusalem; and 5 Gonda Brain Research Center, Bar-Ilan University, Ramat-Gan, Israel Submitted 18 January 2005; accepted in final form 2 January 2006 Rivlin-Etzion, Michal, Ya acov Ritov, Gali Heimer, Hagai Bergman, and Izhar Bar-Gad. Local shuffling of spike trains boosts the accuracy of spike train spectral analysis. J Neurophysiol 95: , First published January 11, 2006; doi: /jn Spectral analysis of neuronal spike trains is an important tool in understanding the characteristics of neuronal activity by providing insights into normal and pathological periodic oscillatory phenomena. However, the refractory period creates high-frequency modulations in spike-train firing rate because any rise in the discharge rate causes a descent in subsequent time bins, leading to multifaceted modifications in the structure of the spectrum. Thus the power spectrum of the spiking activity (autospectrum) displays elevated energy in high frequencies relative to the lower frequencies. The spectral distortion is more dominant in neurons with high firing rates and long refractory periods and can lead to reduced identification of low-frequency oscillations (such as the 5- to 10-Hz burst oscillations typical of Parkinsonian basal ganglia and thalamus). We propose a compensation process that uses shuffling of interspike intervals (ISIs) for reliable identification of oscillations in the entire frequency range. This compensation is further improved by local shuffling, which preserves the slow changes in the discharge rate that may be lost in global shuffling. Cross-spectra of pairs of neurons are similarly distorted regardless of their correlation level. Consequently, identification of low-frequency synchronous oscillations, even for two neurons recorded by a single electrode, is improved by ISI shuffling. The ISI local shuffling is computed with confidence limits that are based on the first-order statistics of the spike trains, thus providing a reliable estimation of auto- and cross-spectra of spike trains and making it an optimal tool for physiological studies of oscillatory neuronal phenomena. INTRODUCTION Spectral analysis is widely used in many science and engineering fields (Miller and Sigvardt 1998; Percival and Walden 1993). Spectral methods are best suited for systems that display rhythmic behavior such as the nervous system, which tends to fire in periodic oscillations in many normal and pathological states (Brown 2003; Gauss and Seifert 2000; Gray 1994; McCormick 2002). One key example is the low frequency modulation of firing activity that arises in Parkinson s disease (PD) (Hutchison et al. 1997; Nini et al. 1995). These neuronal oscillations, which can be found in the globus pallidus (GP), thalamus, and other basal-ganglia-related structures, may be important clues to the understanding of the clinical symptoms of Parkinsonism, especially the rest tremor. The spike train of a neuron is a point process frequently modeled as a Poisson process. The power spectrum of a Address for reprint requests and other correspondence: M. Rivlin-Etzion ( michriv2@alice.nc.huji.ac.il). Poisson process exhibits equal power in all frequencies. Nevertheless, the distribution of the power spectrum of a typical spike train is not homogeneous. The power at high frequencies ( 100 Hz) is close to that of a Poisson process with the same average firing probability (Bair et al. 1994; Franklin and Bair 1995). However, the power in the low frequencies is significantly smaller than expected, and a trough can be seen in the spectrum of the neuron (Fig. 1). This distortion in the spectrum of the neuronal firing makes it difficult to identify any oscillations in the low frequencies. As a result, Parkinsonian pallidal cells that oscillate in the range of 5 10 Hz may not be classified correctly as oscillatory unless the oscillation power is high enough to overcome the distortion. This issue has been neglected in many previous studies (e.g., Raz et al. 2000, 2001), which have only depicted the lower range ( 30 Hz) of the spectrum. The spectral distortion phenomenon has been described previously in cortical area MT (Bair et al. 1994; Franklin and Bair 1995) and in the auditory system (Edwards et al. 1993). It occurs because the Poissonian model does not take the refractory period, which is an important property of the neuron, into consideration. The refractoriness prevents the neuron from firing two successive spikes within a very short interval. Hence a spike train is never a true Poisson process and does not contain two spikes that occur in a range smaller than the refractory period. In this paper, we overcome these problems by applying the ISI shuffling method. We demonstrate the basic process for compensation of the refractory period and for the detection of oscillatory neurons. Then we present several refinements that enable better assessment of the frequency domain properties of spike trains. METHODS Electrophysiological recordings Electrophysiological examples were obtained from extracellular recordings made in a vervet monkey (monkey C, Cercopithecus aethiops) and a cynomolgus monkey (monkey R, Macaca fascicularis). Recordings were made in the GP before and after treatment with 1- methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP), which induced Parkinsonism. Standard recording and analysis methods were used and are detailed elsewhere (Heimer et al. 2002). Briefly, units were detected and isolated in real time using a template matching algorithm (MSD, Alpha- Omega Engineering). The isolation quality was evaluated on-line accord- The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. Downloaded from jn.physiology.org on April 29, /06 $8.00 Copyright 2006 The American Physiological Society

25 Results I Innovative Methodology 3246 RIVLIN-ETZION, RITOV, HEIMER, BERGMAN, AND BAR-GAD ing to the signal-to-noise ratio and the stability of the spike waveform. Recording quality and stability were further evaluated off-line, and neurons were included in this study only if they were well isolated according to their ISI distribution and their firing rate was stable. Spike train simulations The spike train of a simulated neuron was modeled as an orderly point process (i.e., only 1 event can occur in each bin) (Brillinger 1975; Cox and Isham 1980). It was assumed to be quasi-stationary and to satisfy a cross-mixing condition (i.e., differential changes in discharge rate that are widely spaced in time are statistically independent) (see Brillinger 1975). The simulated spike trains were modeled as a renewal process with a refractory period and a constant firing probability thereafter (Bar-Gad et al. 2001a,b). In this model, neurons had a constant firing probability (p) for each discrete time bin ( t) except that after a spike occurred, the neuron entered a refractory period (with a length of n r time bins) in which its probability of firing was smaller than the steady-state probability. We used a refractory period represented by the exponential function p n k (nr 1 n) p; n n r, where n is the number of bins after the preceding spike and n r is the length of the refractory period. An absolute refractory period is defined by k 0 leading to zero firing probability for the entire period; a relative refractory period is defined by 0 k 1. This model of the refractory period simplifies the simulations and the mathematical analysis and was thus used instead of the more realistic gamma distribution (Kuffler et al. 1957; Stein 1965). For oscillation modeling, we added a sinusoidal modulation of the firing rate with frequency f osc and amplitude p osc. In each time bin, the term p osc sin(2 f osc n t) was added to the firing probability. The sine term produced oscillatory spike trains that contained bursts that started and ended slowly, like those seen in human patients with Parkinsonian tremor (Zirh et al. 1998), as opposed to bursts with a quick start that slowly decay. To examine the spectral effects of the refractory period on the cross-spectrum, we studied a pair of neurons with a common input of strength p corr. To simulate the common input, a spike train representing the hidden common source was generated as described in the preceding text. Then two other spike trains were generated in a similar manner, but in both of them, the probability of having a spike in the nth bin was p corr higher if a spike had occurred in the spike train representing the common input in the nth bin. To simulate two neurons that were recorded from the same electrode leading to incomplete spike identification due to the shadowing effect (Bar-Gad et al. 2001b), we generated both spike trains as described in the preceding text and then deleted simultaneous spikes or spikes that occurred 1 ms apart. Throughout both the electrophysiological recordings and the simulations, the default time bin ( t) was 1 ms. The length of the simulation was in the range of the average duration of the electrophysiological recordings (10 6 bins 17 min). Spectral calculations FIG. 1. Autospectrum and interspike interval (ISI) histograms of spike trains recorded in normal and 1-methyl-4-phenyl- 1,2,3,6-tetrahydropyridine (MPTP)-treated monkeys. Examples of the spectral densities shown in logarithmic scale (A C) and the ISI histograms (D F) of globus pallidus internal segment (GPi) and external segment (GPe) neurons. A and D: GPi neuron recorded in a normal monkey (mean firing rate 53 spikes/ s). B and E: GPe (76 spikes/s) and C and F: GPi (35 spikes/s) neurons recorded in an MPTP-treated monkey. The upper Poisson confidence levels calculated from the mean and variability of the spectrum at the Hz are shown as dashed horizontal lines. Note that the peaks around 5 and 10 Hz in the spectrum of the neurons recorded in the MPTP-treated monkey (B and C) are below the Poisson confidence levels. The refractory period is evident in all 3 ISI histograms (D F). Last bin in the ISI histogram is a summation of all higher duration ISIs. The right Y scale in A C represents the autospectrum in hertz that tends to the firing rate. This scale is achieved when treating the spike train as a point process and not making the continuousprocess analogy (as is represented on the left Y scale and in the following figures). The power spectral density (PSD), cross-spectral density (CSD), and the coherence functions were estimated based on Welch s method Downloaded from jn.physiology.org on April 29, 2008 J Neurophysiol VOL 95 MAY

26 Results I Innovative Methodology SPECTRAL COMPENSATION FOR NEURAL REFRACTORY PERIOD 3247 (Halliday et al. 1995; Welch 1967). In all spectral calculations, the windows used were nonoverlapping Hanning windows with a length of 4,096; in each window, a discrete Fourier transform (DFT) was performed. The sampling frequency for spike detections was 1,000 Hz. Therefore the resolution of all spectral functions was 0.25 Hz and the maximal frequency was 500 Hz. All spectral calculations were performed using standard Matlab 6.5 (The MathWorks, Natick, MA) code. Other methods for calculating the power spectral density are available (Percival and Walden 1993) and yield qualitatively identical results. In all our neurons, simulated and real, the PSDs were flat in the high frequencies ( 100 Hz) range and converged to the expected PSD of a Poisson process with the same average firing rate. Therefore we based our initial confidence level which determines the frequencies in which the neuron tends to fire more than expected by the Poisson process on the high frequencies range. Because the spectra of all neurons in this study were flat over the last Hz, all the figures present the spectrum 300Hz, and the initial confidence levels were constructed based on the last 10% of the bins. The means SD of the PSD values (in logarithmic scale) in the range of Hz were calculated, and assuming a normal distribution (according to the central limit theorem), a confidence level of 99% (normalized to the total number of bins) was constructed. Another possible asymptotic 100 % confidence level for a Poisson process was obtained by (Halliday et al. 1995): log 10 (Pˆ 1) [norm_inv( ) log 10 e]/ L, where Pˆ 1 is the estimated firing rate, L is the number of subrecords that the data were divided into in the spectral calculation and norm_inv( ) is the inverse of the normal cumulative distribution function at the corresponding probabilities in. We tested this method on our data and the results were almost identical to those based on the last 10% bins. We only calculated confidence levels for the absolute value of the complex function of the CSD. Because the CSD was also flat in the high frequencies ( Hz), the confidence level was constructed in a manner similar to the construction of the confidence level of the PSD. In RESULTS, we show why this choice of confidence levels is not helpful for both PSD and CSD, and we propose alternative confidence levels. The coherence function is also commonly used to explore the spectral relations between two processes. It ranges from 0 to 1, and a significant value indicates linear relations between the two processes at a particular frequency. A 100 % confidence level for the coherence function is given by 1 (1 ) 1/(L 1) where L is the number of subrecords the data were divided into when calculating the coherence (Bloomfield 1976; Brillinger 1981; Rigas 1996a). The confidence limits and sensitivity of this and other methods are compared with the ISI shuffling methods in RESULTS. All the functions used for both the simulation and analysis are Matlab 6.5 (Mathworks, Natick, MA) compatible and may be found at RESULTS Distorted spectrum of spike trains The spectrum of the spiking activity of GP neurons in both normal (Fig. 1A) and Parkinsonian (Fig. 1, B and C) primates reflects an uneven distribution of power in different frequencies. The power is stereotypically lower in the range of Hz, whereas it is stable in the higher ( 100 Hz) frequencies. The low-frequency trough can be found in all pallidal neurons, including neurons that tend to fire spikes periodically. As a result, it is often difficult to identify the expected peak in the 4- to 14-Hz range in the Parkinsonian state (Fig. 1, B and C). The significance of the 10-Hz oscillations of the neurons shown in Fig. 1, B and C, is not obvious according to our initial P 0.01 confidence levels (that are based on the mean and variability at Hz, see METHODS). Moreover, the figures demonstrate that calculating the confidence levels based on a different frequency range (e.g., 3 30 Hz) (Raz et al. 2000, 2001) may lead to other errors in estimations of the oscillatory activity. The recorded neurons demonstrating the distortion in the spectrum are well isolated, and the refractory period is clearly evident in their ISI histogram (Fig. 1, D F). Indeed, the refractory period is the main source for the spectral distortion. To illustrate the effect of the refractory period, Fig. 2 shows the PSD and ISI histograms of four simulated neurons with a pure Poisson process (Fig. 2, A and E) versus three examples of simulated neurons with different refractory periods and oscillatory activity (Fig. 2, B D, F H). An intuitive explanation for the spectral distortion of spike trains can be given by examining the autocorrelation function that also suffers from the effects of refractoriness. The autocorrelation function of the high-frequency discharge GP cells reveals a short term peak followed by a small trough due to the refractory period (Bar-Gad et al. 2001a). For long offsets, the autocorrelation seems to be flat, but in fact, it contains many small peaks that are similar to the short-term peak but are less dominant. They appear because any rise in the autocorrelation function above its expected value will cause a drop from the expected value in the following time bins due to the refractory period. Therefore the autocorrelation function of a neuronal spike train with a refractory period keeps rising and falling from its expected value very rapidly. The size of these fluctuations in the autocorrelation function apparently decreases with the offset from zero lag because of the accumulation of variance in the timing of the following spikes. Nevertheless, as a result of these small fluctuations, the power of the high frequencies of the neural autocorrelation function is high compared with the low frequencies. The uneven power distribution creates a trough in the low frequencies of the auto-spectrum function that is the Fourier transform of the autocorrelation function. The mathematical analysis of this phenomenon can be found in (Bair et al. 1994; Edwards et al. 1993; Franklin and Bair 1995). A simpler explanation is given in APPENDIX A for an absolute refractory period. ISI shuffling An oscillatory spike train is a nonrenewal process because the oscillations are usually generated by periodic bursts of spikes. Therefore to detect significant oscillations, we consider as a null hypothesis a renewal process with the same distribution of ISIs. Unlike the original spike train that may be oscillatory, the ISIs of the new process do not depend on the previous ISIs or on time. The renewal process is achieved by using the ISI shuffling method (Perkel et al. 1967; Tam et al. 1988). ISI shuffling uses the first-order ISIs (i.e., the time differences between adjacent spikes) as the building blocks of the new spike train. The ISIs are therefore randomly permuted, and a new spike train is generated based on the shuffled ISIs (Fig. 3A). The PSD of the new spike train is determined solely by the first-order ISIs of the original spike train. All higherorder effects (i.e., the time difference between 2 spikes that are separated by 1 spike or more) are abolished by the shuffling process. Hence comparison of the original PSD to the shuffled one enables the detection of patterns such as periodic burst Downloaded from jn.physiology.org on April 29, 2008 J Neurophysiol VOL 95 MAY

27 Results I Innovative Methodology 3248 RIVLIN-ETZION, RITOV, HEIMER, BERGMAN, AND BAR-GAD FIG. 2. Effects of refractory period on the auto-spectrum of simulated spike trains. The spectral densities shown in logarithmic scale (A D) and the ISI histograms (E H) of simulated neurons displaying no refractory period (A and E; P 0.057, r 0 ms, k 1), an absolute refractory period (B and F; P 0.09, r 9 ms, k 0), a relative refractory period (C and G;P 0.09, r 9 ms, k 0.7), and a relative refractory period with the addition of a sinusoidal modulation (D and H; P 0.09, r 9 ms, k 0.7, f osc 10 Hz, p osc 0.007). The average firing rate of all 4 neurons was 57 spikes/s. We set the firing probability (p) to be 0.09 for the last 3 neurons to obtain this rate in spite of the refractoriness. The 10-Hz peak in the spectrum of the oscillatory neuron is below the Poissonian upper confidence level (- - -) calculated based on the 270- to 300-Hz range. All scales and conventions are as in Fig. 1. oscillations that are generated by higher-order ISIs. Repeating the shuffling process N times and averaging the results provides a less noisy estimator of the contribution of the refractory period to the PSD. The PSD of the shuffled train is shown for a simulated (Fig. 4A) and for real neurons (B). As expected, the PSD of the shuffled train does not contain the peak around 10 Hz. Yet unlike the simulated neurons, the original and the shuffled PSDs of the real neuron do not fully converge in the range of the trough. This difference was found to be correlated with higher power at low frequencies that probably arises from the persistent slow changes in the neural firing rate (Wichmann et al. 2002). We can overcome the effect of slow changes in the firing rate by performing local shuffling (see following text). Local shuffling Because slow changes are apparent in the neural firing rate of real neurons, our initial null hypothesis of a renewal process with the same distribution of ISIs is not appropriate. We therefore modified our null hypothesis into a renewal process with the same local distribution of ISIs. The ISI distribution of this renewal process does not depend on the previous ISIs, but it changes slowly over time. This hypothesis can be examined by using the local shuffling method. In the global shuffling method described in the preceding text, we permute all ISIs recorded throughout the recording period, thus abolishing temporal changes in firing rate. In the local shuffling method, the spike train is divided into segments of length T, and the ISIs in each segment are shuffled within that segment. Consequently, any changes in the firing rate that occur over a period longer than T will be preserved in the locally shuffled train. Two additional procedures further improve the outcome of the local shuffling: first, soft rather than hard division of the spike train: the actual partition is not done every time T (Fig. 3B) but rather at the time where the closest spike to time Downloaded from jn.physiology.org on April 29, 2008 J Neurophysiol VOL 95 MAY

28 Results I Innovative Methodology SPECTRAL COMPENSATION FOR NEURAL REFRACTORY PERIOD 3249 FIG. 3. The shuffling process. A: global shuffling. All ISIs are permuted randomly. B and C: local shuffling. The ISIs are locally permuted, with T 70 bins. B demonstrates a hard division: because no spike occurred in the 70th time bin, the shuffled spike train contains 10 spikes instead of 9, and the 6th ISI (with a length of 40 bins) is divided into 2. C is the soft division: although T 70, the actual division is done at the 62th bin because this is the closest bin to the 70th bin that contains a spike. There are no additional spikes in the soft division, and there is no change in the ISIs of the original spike train. T has occurred (Fig. 3C). This way, there are no additional spikes in the new spike train, and the ISI histogram is identical to the original histogram. Second, random rather than fixed segment duration: once T is set (for soft or hard division) the shuffling within the segment is done randomly, but the spikes that define the borders of the segments of the local shuffling are constant for a given spike train, and will appear in all N random repetitions. These constant spikes can lead to weak oscillations of 1/T Hz in the mean shuffled PSD. To avoid this, T should be defined as a random variable, T r, which changes randomly in a small range. In this study, we used T r [150,200] ms. Prior to shuffling of any segment, we draw T r out of a uniform distribution U(150,200). Then the closest spike to T r (soft division) is identified to set the limits of the segment, and the actual shuffling is done within that segment. An example of the original PSD and the PSD of the locally shuffled spike train is shown in Fig. 5A. The convergence was improved as compared with the PSD of the globally shuffled spike train. The improvement emerged not only in the range of 3 50 Hz but also in the range of 0 3 Hz. This is because the local shuffling maintains the high energy of slow changes in the firing rate that appear in these frequencies (Fig. 5A, bottom). Spectral compensation and confidence levels Here we present a statistical test for rejection of the null hypothesis: the confidence level for determining whether the spike train is oscillatory can be constructed for a compensated spectrum. The goal of the compensation is to overcome the distortion in the spectrum and equalize the level of detection of low-frequency oscillations to that of the high frequencies. Spectral compensation may be achieved by subtracting the shuffled PSD from the original one, or by dividing them. An advantage of compensation by division is that the variance of the compensated term is the same for all frequencies, and therefore the confidence limits are more reliable (see APPENDIX B). Hence the recommended spectral compensation for the refractoriness of the neuron in each frequency is S comp S org S shuf where S org is the original PSD and S shuf is the PSD of the shuffled train (or the mean of the shuffled trains PSDs, for N 1). Note that if S org S shuf the original PSD is merely the outcome of the first-order ISIs resulting in a compensated PSD of 1 for all frequencies. If S org ( ) S shuf ( ), then the original spike train has a power in the frequency that is beyond the expected power from its first-order ISIs, and the compensation will be 1 for that. Therefore a possible confidence level for the compensation term is constructed by its distance from the expected value 1. The distance is determined by the SD (i.e., the P value) of the compensated function in the high frequencies (here we again used the 270- to 300-Hz range), assuming a normal distribution. Figure 5B demonstrates the compensated function and its confidence level that is based on the high frequencies. Although the compensation for real neurons is less accurate than the compensation for the simulated neurons (not shown), the significance of the low-frequency oscillations that might be neglected without the shuffling compensation is often maintained with the shuffling procedure. Other potential confidence levels that use bootstrap procedure can be constructed based on a 2 assumption (Jarvis and Mitra 2001; Percival and Walden 1993). However, these confidence levels require many repetitions of the shuffling process and therefore are more costly in terms of calculation. To establish the possible differences between the suggested methods, we compare the performance of the ISI shuffling method to Halliday s method (Halliday et al. 1995). Figure 6 demonstrates the percentage of the detection of the oscillations in each method as a function of oscillation strength (p osc, see METHODS). The shuffling method is more efficient in identifying oscillations of smaller amplitude in our simulated data. Distortion and compensation of cross-spectrum and the coherence function The cross-spectrum of two neurons suffers from the same distortion described in the preceding text for the autospectrum even if the neurons are independent. The size of this distortion depends on the refractory period of both cells. (1) Downloaded from jn.physiology.org on April 29, 2008 J Neurophysiol VOL 95 MAY

29 Results I Innovative Methodology 3250 RIVLIN-ETZION, RITOV, HEIMER, BERGMAN, AND BAR-GAD FIG. 4. ISI shuffling. The power spectral densities (PSDs) and shuffled PSDs of simulated and real neurons shown in logarithmic scale. A: for the simulated neuron from Fig. 2D. B, 1 3: for the real neurons shown in Fig. 1, A C. The original spectral density is shown in black and the shuffled (N 20) spectral density is shown in gray. The confidence levels are for the original PSD and are marked by The reason is similar to the one already given regarding the PSD but relates to the cross-correlation function: the expected cross-correlation of two independent cells is flat, but again, once the cross-correlation at time t is higher than expected, the probability distribution of the value of the following bin changes accordingly and its expected value is lower than the original expected value (see APPENDIX C for two independent neurons with an absolute refractory period). Therefore the cross-correlation function has small and rapid fluctuations in it. This leads to the relatively low power in the low frequencies and therefore to a trough in the low frequencies of the cross-spectrum functions. The CSD of two simulated oscillatory cells is shown in Fig. 7A. These neurons have no common input, but they oscillate in the same frequency; therefore the CSD contains a peak in the common frequency. The CSD of two oscillatory neurons that have a nonoscillatory common input in addition to an oscillatory correlation has the same shape, but the power over the spectrum increases due to the nonoscillatory correlation (Fig. 7B). The reduction in the energy of low frequencies in the CSD is very common in pairs of recorded neurons (Fig. 8, A and B) and can be observed even if they were nonsimultaneously recorded (Fig. 8A). A reliable detection of real oscillatory correlations can be based, again, on the ISI shuffling method that was described above. The shuffling is done in both spike trains, and the compensation in each frequency is C real C org C shuf where C org is the original CSD and C shuf is the CSD of the shuffled trains (or the mean CSDs, for N 1). The significant correlation frequencies are those found above the confidence level constructed by a normal distribution of mean 1 and a SD that is based on the 270- to 300-Hz band (last 10% of the bins). The global compensation in the CSD, unlike PSD compensation, tends to be accurate and less affected by slow changes (Fig. 8, C and D). However, local shuffling may be performed to achieve better results. Note that this method can detect nonoscillatory correlations in addition to the oscillatory correlations (Fig. 7, C and D), but the former are better identified by the conventional crosscorrelation (time domain) function, which is more sensitive to correlations of this kind and is more informative about their nature. Typically, the coherence function is the cross-spectrum normalized by the autospectra. Therefore the coherence (2) Downloaded from jn.physiology.org on April 29, 2008 J Neurophysiol VOL 95 MAY

30 Results I Innovative Methodology SPECTRAL COMPENSATION FOR NEURAL REFRACTORY PERIOD 3251 FIG. 5. Compensation following local shuffling. A: local shuffled PSD of the neuron shown in Fig. 1C. The original spectral density is shown in black and the locally shuffled (N 20) spectral density is shown in red. The PSD of the global shuffling (taken from Fig. 4B3) is also presented in gray (all PSDs are shown in logarithmic scale). Note the addition of the 0- to 3-Hz range at the bottom. B: compensated spectral density based on the PSD of the locally shuffled spike train seen on the left , confidence level based on the high frequencies. function is not influenced by the refractory period, as can be seen in Figs. 7E and 8, E and F. Nonetheless, the simulated pair with the common input has nonoscillatory correlations and its coherence function tends to cross the confidence level in the high frequencies rather than in the low frequencies (except for the 10-Hz oscillatory correlations; Fig. 7F). The coherence of two neurons with a very high firing rate will cross the confidence level more vigorously. This results from the refractoriness that reduces the correlations over the FIG. 6. Comparison of performance. The percentage of detection of the oscillations of simulated neurons as a function of p osc (P 0.09, r 9 ms, k 0.7, f osc 10 Hz, p osc 0, 0.001, 0.002,..., 0.03). We generated 20 random spike trains for each p osc, and the shuffling procedure was performed on each of them. The shuffling confidence level, as well as Halliday confidence level, was calculated. The solid curve is the percentage of detection of the 10-Hz oscillations according to the shuffling method , percentage of detection of the 10-Hz oscillations according to Halliday s method. low frequencies and is also demonstrated in the compensation term (Fig. 7D). Distortion and compensation of cross-spectrum of two neurons recorded from the same electrode When two or more neurons are recorded from a single electrode, spikes that occurred within a short time interval overlap each other and therefore are hard to identify (Lewicki 1998). This identification failure is termed the shadowing effect and can produce various artifacts in the auto- and cross-correlation of the recorded neurons (Bar- Gad et al. 2001b). As a result, the cross-spectrum of two spike trains recorded from the same electrode is distorted not only by the refractory period, but from the shadowing effect as well. Figure 9A (black curve) illustrates the crossspectrum between two simulated spike trains that were recorded from the same electrode (see METHODS): the trough in the low frequencies still exists, but it becomes narrower. In addition, there is a slow decay in the power of the high frequencies. When applying the shuffling method to both spike trains, the compensation term is far from being optimal (Fig. 9, A and B, gray curve). To obtain better compensation, the characteristics of the shadowing effect must be preserved during the shuffling. First, any simultaneous spikes must be prevented (as well as adjacent spikes; depend on the shadowing duration). Second, longer ISIs should tend to occur together because each time we lose a spike in both spike trains due to the shadowing effect, an ISI that is longer than twice the refractory period (at least one refractory period before the lost spike and at least one refractory period after the lost spike) is produced in each spike train. These long ISIs tend to occur at parallel times in both spike trains and cause the enhanced correlation. ISI shuffling with the shadowing limitations is possible if the shuffling is done step by step and in parallel for both spike trains. In each Downloaded from jn.physiology.org on April 29, 2008 J Neurophysiol VOL 95 MAY

31 Results I Innovative Methodology 3252 RIVLIN-ETZION, RITOV, HEIMER, BERGMAN, AND BAR-GAD step, one ISI of one of the neurons is chosen randomly from the ISIs that have not yet been chosen. This ISI is concatenated to the end of the built train (the part of the shuffled train that was already built for that neuron). If the last spike that was added occurs in the same bin or one bin apart (the shadowing duration in our case) from a spike in the other built train, the shadowing effect must take place. We therefore remove the last ISIs out of both built trains, and we choose (of the ISIs left for each neuron) for each built train an ISI that is longer than twice the refractory period. The results of the step-by-step shuffling are shown in Fig. 9, A and B (red curve), and indicate that step-by-step shuffling should be performed for the study of synchronous oscillations of neurons recorded by the same microelectrode. DISCUSSION This manuscript discusses the distortion in the spectrum of spike trains that occurs due to the refractory period of neurons. The effect of the refractory period has been previously examined with respect to spectral shape (Bair et al. 1994; Edwards et al. 1993; Franklin and Bair 1995). However, these studies do not address the difficulties in identifying oscillatory phenomena created by the spectral shape of real spike trains. We present a method to overcome these difficulties using ISI shuffling of the spike train. The shuffling of the train may be performed in a local manner so that it overcomes the additional distortions induced by slow FIG. 7. Cross-spectrum simulated neurons. A and B: absolute CSD of simulated oscillatory neurons with a relative refractory period and the same oscillation frequency. The original CSD is shown in black and the CSD following ISI shuffling is shown in gray. C and D: compensated CSD (N 20). E and F: coherence function and its confidence level. A, C, and E were calculated for 2 oscillatory spike trains (P 0.09, r 9 ms, k 0.7, f osc 10 Hz, p osc 0.007), B, D, and F were calculated for 2 oscillatory spike trains that are also nonoscillatory correlated due to a common input (P 0.09, r 9 ms, k 0.7, f osc 10 Hz, p osc 0.007, p corr 10). The spike train of the common input: (P 0.09, r 9 ms, k 0.7). All CSDs are shown in logarithmic scale. changes in the neural firing rate. The shuffled train has exactly the same first-order properties of refractoriness as the original spike train, and therefore the spectra of both trains have the same structure. The periodic properties are revealed by division of the original by the shuffled spectra. The division process ensures equal distribution of power across all frequencies, providing reliable ways to establish confidence limits. One example of the importance of detecting oscillatory neural activity is the study of neural activity in PD, where basal ganglia and motor thalamic cells tend to fire periodically in the frequency of the tremor (Bergman et al. 1994; Filion and Tremblay 1991; Hurtado et al. 1999, 2005; Hutchison et al. 1997; Lenz et al. 1988; Levy et al. 2002; Nini et al. 1995). The method presented here enables reliable detection of auto- and cross-oscillations in the Parkinsonian state. The effect of the refractory period is crucial in neurons with high discharge rate because the distortion in their spectrum is larger in magnitude. As a result, the method is extremely useful when analyzing neurons with tonic high firing rates or neurons that tend to increase their firing rates under certain circumstances such as a stimulus or behavioral task. The method provides confidence limits based on the firstorder statistics of the spike train. Therefore our method is practical for detecting burst oscillations in all frequencies for all kinds of spike trains, regardless of the discharge rate or the length of the refractory period. This naturally includes spike Downloaded from jn.physiology.org on April 29, 2008 J Neurophysiol VOL 95 MAY

32 Results I Innovative Methodology SPECTRAL COMPENSATION FOR NEURAL REFRACTORY PERIOD 3253 trains of well-isolated neurons where the refractory period is dominant and the distortion is large as well as multi-unit recordings where there is typically little or no effect of the refractory period and thus the spectrum is not distorted. The shuffled spike train does not contain any structure of the original spike train that is inherent to second and higher order ISIs. All structures that are generated by the firstorder ISI are preserved in the shuffled train and therefore are abolished in the compensated function. Usually, periodic oscillations in the spike train of neurons with burst oscillations are caused by higher-order ISIs; therefore the compensation process suggested here is useful for detecting them. However, this is not always correct. For example, in the case of a regular neuron that fires single spikes periodically (Ahissar and Vaadia 1990), the oscillations are caused by the first-order ISIs, and the PSD of the shuffled train will be identical to the original PSD. The shuffling method suggested here is therefore not recommended for these cases. This paper sheds light on another common phenomenon in spectral studies of neuronal activity the excess energy in low (1 3 Hz) regimes. These frequencies often leak to the frequencies of interest and may mask significant oscillations in the 5- to 10-Hz region. This occurs due to the ongoing slow changes in the firing rate, which do not exist in the globally shuffled spike train. The larger distortion might be surprising because the ISIs are the same in both spike trains; therefore the expectation is to have a similar distortion in both spectra. FIG. 8. Cross-spectrum recorded neurons. A and B: absolute CSD of recorded neurons (black) and the CSD following ISI shuffling (gray). C and D: compensated CSD (N 20). E and F: coherence function and its confidence level. A, C, and E were calculated for a pair of neurons in the normal state that were recorded in different sessions. B, D, and F were calculated for a pair of simultaneously recorded neurons in the Parkinsonian state. All CSDs are shown in logarithmic scale. However, because the firing rate is not constant, the original spike train tends to have segments of short ISIs that contribute to the large distortion and other segments of longer ISIs that are less effective. Because the PSD function is the average of a nonlinear function that is calculated for each segment, a stronger distortion in the nonequally distributed segments is obtained. Unlike global shuffling, the local shuffling depends on the parameter T that is set by the researcher. Naturally, T should be smaller than the size of spectral window (4,096 ms in our case). Moreover, T should be as small as possible so the changes in the firing rate are better preserved. On the other hand, as T gets smaller, the local shuffling loses its power because the number of ISIs in each segment is close to one, and the shuffling becomes useless. The T chosen here was 200 ms and seemed to be satisfactory for most of our pallidal cells with mean discharge rate 50 Hz. Nevertheless, because the changes in the firing rate are probably variable themselves, there is no way to choose T so that the compensation in the spectrum will be perfect. To summarize, the ISI shuffling method is an important tool for compensating for the distortion in the spectrum of neuronal spike trains and therefore provides reliable confidence limits for spectral functions. It enables the detection of periodic oscillations in spike trains that could have been ignored otherwise, and thus it is recommended for the study of neural oscillations in normal and pathological states. Downloaded from jn.physiology.org on April 29, 2008 J Neurophysiol VOL 95 MAY

33 Results I Innovative Methodology 3254 RIVLIN-ETZION, RITOV, HEIMER, BERGMAN, AND BAR-GAD A log10( CSD ) B Compensated CSD C Coherence Frequency (Hz) FIG. 9. Cross-spectrum and the shadowing effect simulated neurons. A: absolute CSD of the original simulated oscillatory neurons (black), the CSD following global ISI shuffling (gray), and the CSD following the step-by-step shuffling (red). Confidence level is the dashed horizontal line. B: compensated CSD (N 20). In gray is the compensation based on the global shuffling, confidence level is the black horizontal line. In red is the compensation based on the step-by-step shuffling, confidence level is in red. All confidence levels are based on last 10% of the frequencies. C: coherence function and its confidence level. The spike trains were generated with the parameters (P 0.12, r 9 ms, k 0.7, f osc 10 Hz, p osc 0.015), and any simultaneous spikes, or spikes that occurred 1 ms apart were deleted from both spike trains. All CSDs are shown in logarithmic scale. APPENDIX A: AUTO-SPECTRUM OF A POINT-PROCESS WITH AN ABSOLUTE REFRACTORY PERIOD R x i P A x c i A x c 0 (A2) where A x c (i) is the complementary event in which a spike did occur at the ith bin. We obtain R x i p 1 p A c x i k A c x 0, i 0 (A3) k 1 Recall that the events A x (i), A x (j) are disjoint for 0 i j, therefore R x i p 1 R x i k k 1 and the last equation can be extended to R x i p R x i k p 1 p i, for all i 0 k 1 (A4) (A5) for (i) 1 for i 0, and 0 otherwise. The Fourier transform of R x (i k) iss x ( )e j k. Therefore the Fourier transform of both sides (if we take R x (i) for negative indexes), yields S x p S x e i k 1 p, 0 k 1 (A6) Because we enforced the autocorrelation function to be 0 for a negative index and because the autocorrelation is a symmetric function, the real Fourier transform is twice the real part of what we got S x 2 Re 1 p 1 p i k, 0 e k 1 (A7) Figure A1 illustrates the results for different firing probabilities p and refractory periods. Downloaded from jn.physiology.org on April 29, 2008 Let x 1, x 2,... be a process with an absolute refractory period of ms. That is P x t 1 x t 1 0,x t 2 0,...,x t 0 p P x t 1 x t 1 1 x t x t 1 0 (A1) Let R x (i) be the autocorrelation function of the process. A x (i) isthe event in which no spike occurred at the ith bin. Then FIG. A1. Auto-spectrum of a process with an absolute refractory period. The PSD function, as was calculated analytically in APPENDIX A(Eq. A7). Three curves are shown for different parameters of firing probability and absolute refractory period length. As expected: a longer refractory period causes a narrower and deeper trough. A lower firing rate causes a smaller distortion. J Neurophysiol VOL 95 MAY

34 Results I Innovative Methodology SPECTRAL COMPENSATION FOR NEURAL REFRACTORY PERIOD 3255 APPENDIX B: VARIANCE OF THE COMPENSATION If we take the compensation in each frequency to be S comp S org S shuf (B1) where S org is the original PSD and S shuf is the PSD of the shuffled train (or the mean PSDs, for N 1), the variance of the compensation term is the same for all frequencies. Proof: the variance of a quotient of two random variables can be approximated as follows (Mood et al. 1974) var X/Y E X E Y 2 var X var Y 2cov X,Y 2 2 (B2) E X E Y E X E Y Let X( ) S org ( ) and Y( ) S shuf ( ) for any. Due to the random shuffling, X( ) and Y( ) are independent and therefore cov(x, Y) 0. The estimate of the PSD has a distribution which is analogous to a 2 distribution in each with the same degrees of freedom (see Brillinger 1981, p. 164). Because the ratio of the SD to the mean for such a random variable depends solely on the degrees of freedom, it does not depend on Therefore std X E X const x var X E X 2 const x 2 and the same is true for Y (with const Y ). If E(X) E(Y), we have var X/Y const x 2 const y 2 (B3) (B4) (B5) and the term is a constant that does not depend on. Therefore the variance of the compensation for all frequencies is constant. APPENDIX C: CROSS-CORRELATION FUNCTION BETWEEN TWO PROCESSES WITH ABSOLUTE REFRACTORY PERIODS Let X and Y be two independent processes with an absolute refractory period of ms and a zero mean. The empirical cross-correlation function between them is defined as ĉ t 1 T t X s Y s t T s 1 (C1) The covariance between the empirical cross-correlation function at time t and time t m is cov ĉ t,ĉ t m E ĉ t ĉ t m E 1 T t T 2 s 1 E 1 T t T 2 1 T 2 s 1 T t m u 1 T t m s 1 u 1 T t T t m u 1 X s Y s t X u Y u t m X s X u Y s t Y u t m R x u s R y u s m (C2) where R X (u), R Y (u) are the autocorrelation functions of X and Y, respectively. Because the summation is over all values of the autocorrelation functions, we obtain for T t, m cov ĉ t,ĉ t m 1 T T R T 2 x u R y u m s 1 u 1 1 T R x u R y u m T u 1 (C3) Recall that the autocorrelation function of each of the processes is positive at u 0, negative along the refractory period, and then becomes positive again and finally converges to 0 and is symmetric around that value (Bar-Gad et al. 2001a). Therefore the covariance between the cross-correlation function at different time bins is negative if the distance between them is on the same order of magnitude as the refractory period, which yields a periodic structure of the empirical cross-correlation function. For further details and clarifications, see (Jenkins and Watts 1968; Rigas 1996b). ACKNOWLEDGMENTS We thank N. Parush, I. Nelken, and R. Kass for helpful discussions, J. A. Goldberg for useful suggestions and for sharing data with us, G. Morris and Y. Prut for comments on earlier versions of this manuscript. GRANTS This study was partly supported by a Center of Excellence grant administered by the Israel Science Foundation, by the German-Israel Binational Foundation and the BMBF Israel-Germany collaboration in medical research and by a Fighting against Parkinson grant administrated by the Netherlands Friends of the Hebrew University. Y. Ritov was supported in part by an ISF grant. REFERENCES Ahissar E and Vaadia E. Oscillatory activity of single units in a somatosensory cortex of an awake monkey and their possible role in texture analysis. Proc Natl Acad Sci USA 87: , Bair W, Koch C, Newsome W, and Britten K. Power spectrum analysis of bursting cells in area MT in the behaving monkey. J Neurosci 14: , Bar-Gad I, Ritov Y, and Bergman H. The neuronal refractory period causes a short-term peak in the autocorrelation function. J Neurosci Methods 104: , 2001a. Bar-Gad I, Ritov Y, Vaadia E, and Bergman H. Failure in identification of overlapping spikes from multiple neuron activity causes artificial correlations. J Neurosci Methods 107: 1 13, 2001b. Bergman H, Wichmann T, Karmon B, and DeLong MR. The primate subthalamic nucleus. II. Neuronal activity in the MPTP model of Parkinsonism. J Neurophysiol 72: , Bloomfield P. Fourier Analysis of Time Series: An Introduction. New York: Wiley, 1976 Brillinger DR. The identification of point processes systems. Ann Probab 3: , Brillinger DR. Time Series: Data Analysis and Theory. San Francisco: Holden-Day, Brown P. Oscillatory nature of human basal ganglia activity: relationship to the pathophysiology of Parkinson s disease. Mov Disord 18: , Cox DR and Isham V. Point Processes. London: Chapman and Hall, Edwards B, Wakefield G, and Powers NL. The spectral shaping of neural discharges by refractory effects. J Acoust Soc Am 93: , Filion M and Tremblay L. Abnormal spontaneous activity of globus pallidus neurons in monkeys with MPTP-induced Parkinsonism. Brain Res 547: , Franklin J and Bair W. The effect of a refractory period on the power spectrum of neuronal discharge. SIAM J Appl Math 55: , Gauss R and Seifert R. Pacemaker oscillations in heart and brain: a key role for hyperpolarization-activated cation channels. Chronobiol Int 17: , Downloaded from jn.physiology.org on April 29, 2008 J Neurophysiol VOL 95 MAY

35 Results I Innovative Methodology 3256 RIVLIN-ETZION, RITOV, HEIMER, BERGMAN, AND BAR-GAD Gray CM. Synchronous oscillations in neuronal systems: mechanisms and functions. J of Computational Neuroscience 1: 11 38, Halliday DM, Rosenberg JR, Amjad AM, Breeze P, Conway BA, and Farmer SF. A framework for the analysis of mixed time series/point process data theory and application to the study of physiological tremor, single motor unit discharges and electromyograms. Prog Biophys Mol Biol 64: , Heimer G, Bar-Gad I, Goldberg JA, and Bergman H. Dopamine replacement therapy reverses abnormal synchronization of pallidal neurons in the 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine primate model of Parkinsonism. J Neurosci 22: , Hurtado JM, Gray CM, Tamas LB, and Sigvardt KA. Dynamics of tremor-related oscillations in the human globus pallidus: a single case study. Proc Natl Acad Sci USA 96: , Hurtado JM, Rubchinsky LL, Sigvardt KA, Wheelock VL, and Pappas CT. Temporal evolution of oscillations and synchrony in GPi/muscle pairs in Parkinson s disease. J Neurophysiol 93: , Hutchison WD, Lozano AM, Tasker RR, Lang AE, and Dostrovsky JO. Identification and characterization of neurons with tremor- frequency activity in human globus pallidus. Exp Brain Res 113: , Jarvis MR and Mitra PP. Sampling properties of the spectrum and coherency of sequences of action potentials. Neural Comp 13: , Jenkins GM and Watts DG. Spectral Analysis and Its Applications. San Francisco,CA: Holden-Day, Kuffler SW, Fitzhugh R, and Barlow HB. Maintained activity in the cat s retina in light and darkness. J Gen Physiol 40: , Lenz FA, Tasker RR, Kwan HC, Schnider S, Kwong R, Murayama Y, Dostrovsky JO, and Murphy JT. Single unit analysis of the human ventral thalamic nuclear group: correlation of thalamic tremor cells with the 3 6 Hz component of Parkinsonian tremor. J Neurosci 8: , Levy R, Hutchison WD, Lozano AM, and Dostrovsky JO. Synchronized neuronal discharge in the basal ganglia of Parkinsonian patients is limited to oscillatory activity. J Neurosci 22: , Lewicki MS. A review of methods for spike sorting: the detection and classification of neural action potentials. Network 9: R53 R78, McCormick DA. Cortical and subcortical generators of normal and abnormal rhythmicity. Int Rev Neurobiol 49: , Miller WL and Sigvardt KA. Spectral analysis of oscillatory neural circuits. J Neurosci Methods 80: , Mood AM, Graybill FA, and Boes DC. Introduction to the Theory of Statistics (3rd ed.). New York: McGraw-Hill, Nini A, Feingold A, Slovin H, and Bergman H. Neurons in the globus pallidus do not show correlated activity in the normal monkey, but phaselocked oscillations appear in the MPTP model of Parkinsonism. J Neurophysiol 74: , Percival DB and Walden AT. Spectral analysis for physical applications: multitaper and conventional univariate techniques. Cambridge, UK: Cambridge Univ. Press, Perkel DH, Gerstein GL, and Moore GP. Neuronal spike trains and stochastic point processes. I. The single spike train. Biophys J 7: , Raz A, Frechter-Mazar V, Feingold A, Abeles M, Vaadia E, and Bergman H. Activity of pallidal and striatal tonically active neurons is correlated in MPTP-treated monkeys but not in normal monkeys. J Neurosci 21:RC 128: 1 5, Raz A, Vaadia E, and Bergman H. Firing patterns and correlations of spontaneous discharge of pallidal neurons in the normal and the tremulous 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine vervet model of Parkinsonism. J Neurosci 20: , Rigas AG. Estimation of certain parameters of a stationary hybrid process involving a time series and a point process. Math Biosci 133: , 1996a. Rigas AG. Spectral analysis of a stationary bivariate point process with applications to neurophysiological problems. J Time Series Anal 17, No. 2: , 1996b. Stein RB. A theoretical analysis of neuronal variability. Biophys J 91: , Tam DC, Ebner TJ, and Knox CK. Conditional cross-interval correlation analyses with applications to simultaneously recorded cerebellar Purkinje neurons. J Neurosci Methods 23: 23 33, Welch PD. The use of fast Fourier transform for the estimation of power spectra: a method based on time averaging over short, modified periodograms. IEEE Trans Audio Electroacoust 15: 70 73, Wichmann T, Kliem MA, and Soares J. Slow oscillatory discharge in the primate basal ganglia. J Neurophysiol 87: , Zirh TA, Lenz FA, Reich SG, and Dougherty PM. Patterns of bursting occurring in thalamic cells during Parkinsonian tremor. Neuroscience 83: , Downloaded from jn.physiology.org on April 29, 2008 J Neurophysiol VOL 95 MAY

36 Results II II. LOW PASS FILTER PROPERTIES OF BASAL GANGLIA- CORTICAL-MUSCLE LOOPS Article information: Rivlin-Etzion M, Marmor O, Saban G, Rosin B, Haber SN, Vaadia E, Prut Y, Bergman H (2008). Low-pass filter properties of basal ganglia cortical muscle loops in the normal and MPTP primate model of parkinsonism. Journal of Neuroscience 28:

37 Results II The Journal of Neuroscience, January 16, (3): Behavioral/Systems/Cognitive Low-Pass Filter Properties of Basal Ganglia Cortical Muscle Loops in the Normal and MPTP Primate Model of Parkinsonism Michal Rivlin-Etzion, 1,2 Odeya Marmor, 1 Guy Saban, 1 Boris Rosin, 1 Suzanne N. Haber, 4 Eilon Vaadia, 1,2 Yifat Prut, 1,2 and Hagai Bergman 1,2,3 1 Department of Physiology, The Hebrew University Hadassah Medical School, Jerusalem 91120, Israel, 2 The Interdisciplinary Center for Neural Computation and 3 Eric Roland Center for Neurodegenerative Diseases, The Hebrew University, Jerusalem 91904, Israel, and 4 Department of Pharmacology and Physiology, University of Rochester, Rochester, New York Oscillatory bursting activity is commonly found in the basal ganglia (BG) and the thalamus of the parkinsonian brain. The frequency of these oscillations is often similar to or higher than that of the parkinsonian tremor, but their relationship to the tremor and other parkinsonian symptoms is still under debate. We studied the frequency dependency of information transmission in the cortex BG and cortex periphery loops by recording simultaneously from multiple electrodes located in the arm-related primary motor cortex (MI) and in the globus pallidus (GP) of two vervet monkeys before and after 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP) treatment and induction of parkinsonian symptoms. We mimicked the parkinsonian bursting oscillations by stimulating with 35 ms bursts given at different frequencies through microelectrodes located in MI or GP while recording the evoked neuronal and motor responses. In the normal state, microstimulation of MI or GP does not modulate the discharge rate in the other structure. However, the functionalconnectivity between MI and GP is greatly enhanced after MPTP treatment. In the frequency domain, GP neurons usually responded equally to 1 15 Hz stimulation bursts in both states. In contrast, MI neurons demonstrated low-pass filter properties, with a cutoff frequency above 5 Hz for the MI stimulations, and below 5 Hz for the GP stimulations. Finally, muscle activation evoked by MI microstimulation was markedly attenuated at frequencies higher than 5 Hz. The low-pass properties of the pathways connecting GP to MI to muscles suggest that parkinsonian tremor is not directly driven by the BG 5 10 Hz burst oscillations despite their similar frequencies. Key words: primate; microstimulation; globus pallidus; motor cortex; frequency domain; transfer function; Parkinson s disease Introduction The BG are commonly viewed as operating in a feedforward mode. They receive projections from all cortical areas and feed this information forward via the thalamus to the frontal cortex or directly to brainstem motor nuclei (Shink et al., 1997; Mena- Segovia et al., 2004; McHaffie et al., 2005). The cortical and brainstem motor centers in turn transmit this information down to the spinal cord and muscles (Albin et al., 1989). The functionalconnectivity between the different components of this system has often been studied using electrical stimulation. Such an approach was used to map the relations between MI and muscles (Asanuma and Rosen, 1972; Palmer and Fetz, 1985) MI and BG (Ryan and Clark, 1991; Maurice et al., 1999; Nambu et al., 2000), as well as GP and muscles (Horak and Anderson, 1984). This simplified view of the BG provided profound insights Received July 26, 2007; revised Oct. 7, 2007; accepted Oct. 30, This work was partly supported by a Hebrew University Netherlands Association grant entitled Fighting against Parkinson (H.B.) and by a graduate student fellowship from the Lily Safra Interdisciplinary Center for Neural Computation (M.R.-E.). CorrespondenceshouldbeaddressedtoMichalRivlin-Etzion,DepartmentofPhysiology,TheHebrewUniversity Hadassah Medical School, P.O. Box 12272, Jerusalem 91120, Israel. michal.rivlin@mail.huji.ac.il. DOI: /JNEUROSCI Copyright 2008 Society for Neuroscience /08/ $15.00/0 35 into the pathophysiology of BG related disorders (DeLong, 1990), and led to the development of the current neurosurgical treatment of Parkinson s disease (PD) (Bergman et al., 1990; Benabid et al., 2006). However, the BG are more than a simple feedforward network, because of their reciprocal closed-loop structure (Alexander et al., 1986; Joel and Weiner, 1994; Leblois et al., 2006; Rivlin-Etzion et al., 2006). The frontal cortex, the target of BG output, is one of the main cortical areas projecting to the striatum. Moreover, the cortical projections to the brainstem and spinal cord are paralleled by ascending afferent pathways, closing the loop between the frontal cortex and the periphery. In principle, electrical stimulation could also be used to study information transfer in closed-loop neural network, although the interpretation of the obtained response might be confounded by the fact that information may be transmitted orthodromically and antidromically (i.e., from the soma to the axon terminals and vice versa). In this study, we explored the functional connectivity within the GP cortex muscle loops by stimulating through microelectrodes inserted into either MI or GP, while recording the evoked activity in both structures as well as contralateral arm movements. Because the GP and MI are indirectly connected (the striatum and the subthalamic nucleus are positioned along the pathway from the cortex to the GP, and the thalamus is similarly

38 Results II 634 J. Neurosci., January 16, (3): Rivlin-Etzion et al. Filter Properties of Basal Ganglia Cortical Loops located between the GP and MI), microstimulation effects are probably mediated mainly through orthodromic conduction. Most connectivity studies have used a single pulse or a brief train of electrical stimulation (Asanuma and Rosen, 1972; Cheney and Fetz, 1985; Tremblay et al., 1989; Kita, 1992; Nambu et al., 2000), although an alternative approach was presented in which a prolonged pattern of electrical stimulation was applied (Graziano et al., 2002). Unlike previous studies, we used a stimulus pattern that contained 35 ms bursts delivered at different frequencies ranging from 1 to 15 Hz, in an attempt to mimic the oscillatory bursting pattern often encountered in the BG of parkinsonian patients (Weinberger et al., 2006) and MPTP-treated primates (Nini et al., 1995; Heimer et al., 2006). This pattern can be used to characterize the system transfer functions in the frequency domain (Lathi, 2005) and thus reveal the spectral properties of the BG cortex muscle network. Materials and Methods Animals. Two vervet (African green, Cercopithecus aethiops) monkeys (T and W; females; weighing 3 and 3.5 kg, respectively) were used in this study. Before any of the other procedures were performed, the monkeys were trained to sit in a primate chair, to permit handling by the experimenter, and became familiarized with the laboratory setting. The monkeys had access to standard primate chow and water ad libitum during the whole experimental period. The monkeys health was monitored by a veterinarian, and their weights and clinical status were checked daily. All experimental protocols were performed in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and the Hebrew University guidelines for the use and care of laboratory animals in research and were approved and supervised by the Institutional Animal Care and Use Committee. Surgical procedures. We attached an MRI-compatible plastic (Cilux) head holder and a 27 mm 2 recording chamber to the monkey s skull to allow access to the arm-related areas of the primary motor cortex and to the GP of the right hemisphere (Fig. 1A,B). The recording chamber was tilted 40 laterally in the coronal plane, with its center targeted at the following stereotaxic coordinates (in mm): monkey T, anterior 12, lateral 7, height 12 (above interaural line); monkey W, anterior 10, lateral 5, height 14 (Contreras et al., 1981; Martin and Bowden, 2000), and webbased digital brain atlas of the vervet monkey, currently at The chamber coordinates were verified using magnetic resonance imaging (MRI). The MRI scan [GE Healthcare (Little Chalfont, UK) 1.5 Tesla system; fast spin echo sequence; TR 5.4 s; TE ms; Echo train length 12; number of averages 4; FOV cm; pixels; coronal slices 2 mm wide] was performed with tungsten electrodes at different X-Y coordinates of the chamber (Fig. 1 A). We then aligned the two-dimensional MRI images with coronal sections of the primate atlas. Surgical procedures were performed under deep isoflurane and N 2 O inhalation general anesthesia. Analgesia and antibiotics were administered during surgery and continued for 2 d postoperatively. MRI procedures were performed under light IM Dormitor and Ketamine anesthesia. Recording began after a postoperative recovery period of6dinboth monkeys. Recording and structure identification. During recording sessions, the animals were awake and seated in a primate chair with their head and right hand restrained, but free to move their trunk and their left arm (contralateral to the recording hemisphere) and legs. Four glass-coated 150 m shaft diameter tungsten microelectrodes (impedance, M at 1 khz), confined within a cylindrical metal guide (1.36/1.65 mm inner/outer diameter) were advanced separately (EPS; Alpha Omega Engineering, Nazareth, Israel) into the arm-related area of the motor cortex. A similar set of four independently controlled microelectrodes targeted the pallidum through the same chamber (Double MT; Alpha Omega Engineering) (Fig. 1 B). Each electrode signal was amplified with a gain of 5000 and bandpass filtered with a Hz (monkey T) and a Hz (monkey W) four-pole Butterworth filter (MIP ; Alpha- 36 Omega Engineering). The signal was continuously sampled at 25 khz with 12-bits 5 V A/D converter (Alpha-Map; Alpha Omega Engineering). During the acquisition of the neuronal data, two experimenters controlled the position and spike sorting of the eight electrodes. Spikes were detected from the filtered analog data ( Hz for both monkeys) by on-line detection and sorting software using a templatematching algorithm (MSD; Alpha Omega Engineering). The quality of the detection and spike sorting was estimated and graded on-line every 3 min by the experimenters. This on-line quality estimation was based on three criteria: (1) the superimposed analog traces of the recently (20 100) sorted spikes and the waveforms of events that crossed an amplitude threshold that was set by the experimenter, (2) the cumulative distribution of the Euclidean distances between the detected events and the detection template (ASD; Alpha Omega Engineering), and (3) the stability of the discharge rate. The sampling rate of spike detection pulses was and 40 khz for monkey T and W, respectively. Mapping of the motor cortex was done by examination of the neural responses to passive movements of the contralateral limbs and by electrical stimulations through the microelectrodes using an optically coupled isolator and linear current-source stimulator (Alpha Omega Engineering). The stimulation pattern used for the mapping consisted of 50 ms of a 300 Hz burst (15 pulses) of biphasic symmetric pulses, each phase 0.2 ms, with negative pulse leading. Stimulation current amplitude during mapping ranged from 5 to 80 A. The boundaries of the arm related area of the primary motor cortex were determined as covering the area in which stimulation of 15 A or less caused a movement of the forearm, wrist or fingers. Entry into the pallidum (the lateral border of the GPe) was easily recognized in our penetrations because of the considerably different firing rate, pattern, and spike shape of pallidal versus striatal neurons. The classification of each recorded cell as belonging to either the external or internal pallidum was determined as follows: Neurons located at a depth of 1.5 mm from the striatopallidal border, or neurons that exhibited pauses in their firing pattern were categorized as GPe neurons (DeLong, 1971; Elias et al., 2007). Candidates to GPi were neurons with a highfrequency discharge rate with no pauses, 2.5 mm from the striatopallidal border or farther in the ventromedial direction. Final classification as GPi was based on other physiological identifications along the electrode trajectory (e.g., border cells), or firing pattern of the cell (which was considered only in the normal state). If the classification as GPe or GPi was in doubt, the unit was excluded from the analysis. In this study we did not focus on recording from the GP hand- or motor-related area, but rather attempted to cover most of the areas of both GP segments. Stimulation pattern. The stimulation pattern used for this study consisted of bursts of current pulses. Each burst contained eight biphasic symmetric pulses (each phase 0.2 ms, with negative pulse leading) given at 200 Hz, leading to bursts of 35 ms in length. The stimulation bursts were administered at different frequencies: 1, 2, 5 and 10 Hz, and in addition, 15 Hz for monkey W only. The stimulation in each of the frequency tests lasted 20 s (resulting, for example, in 20 bursts for the 1 Hz stimulation and 300 bursts for the 15 Hz stimulation) (Fig. 1C), with an interstimulus interval (IStimI) of 15 s before the first frequency test, between the different frequency tests, and after the last one. The order of frequency tests was determined randomly for each of the monkeys but remained constant throughout the experiment (monkey T: 5, 10, 1, 2 Hz, monkey W: 10, 1, 2, 5, 15 Hz). After situating the electrodes in positions that enabled optimal recording of single units, the stimulation pattern was given through one or more electrodes (usually one, 89.2% of all cases), whereas all other electrodes were used for recording. All electrodes used in this study were of the same type (glass coated tungsten microelectrode) and could be used for either stimulation or recording. The electrodes used for stimulation within a certain session were connected to the current source, and were not used for recording during the same session. However, the same electrode could serve for stimulation in one session and for recording in the next session. The switch between stimulation and recording mode was controlled electronically (MIP; Alpha Omega Engineering). The amplification gain of the recording electrodes was reduced from 5000 to 5 during the stimulation pulses (0.3 ms before until 0.1 ms after the pulse)

39 Results II Rivlin-Etzion et al. Filter Properties of Basal Ganglia Cortical Loops J. Neurosci., January 16, (3): Figure 1. Experimental procedure and histology. A, MRI of monkey W. An example of a coronal image. Three tungsten electrodes separated by a horizontal distance of 6 mm were inserted in the same coronal plane (ac 2, 2 mm posterior to the anterior commissure) through the recording chamber. The recording chamber is filled with 3% agar. B, Left, A schematic illustration showing the location of ac 6 coronal plane. The coronal plane crosses the central sulcus and the primary motor cortex. Right, A scheme of ac 6 coronal plane and the experimental setup: four electrodes in the GP and four electrodes in the motor cortex. Any electrode could be used as a stimulating electrode, while the other electrodes are recording. The atlas scheme is (Figure legend continues.) 37

40 Results II 636 J. Neurosci., January 16, (3): Rivlin-Etzion et al. Filter Properties of Basal Ganglia Cortical Loops to reduce the saturation of the recording devices and to minimize the stimulus artifact. In this study, we include only stimulating sessions in which all stimulating electrodes were located in the same structure MI or GP. Because of possible current leakage from the GPi to GPe (or vice-versa) that can occur when stimulating at the borders of these nuclei (Ranck, 1975), we did not attempt to distinguish between stimulating sessions in the GPe and GPi. For each stimulating electrode we used current amplitudes of 40 A in monkey T, and either 40 or 60 A in monkey W (40 A were given in 33% and 55% of the sessions in the normal and MPTP states, respectively). When stimulating through more than one electrode, all the stimulating electrodes were connected to the same current source and the amplitude of the current was multiplied by the number of stimulating electrodes. We assumed an equal current distribution through the electrodes because of their similar impedance. The percentage of sessions with a single stimulating electrode in each of the structures (MI and GP) was similar in the normal and MPTP states (89%), to avoid bias between the states. Nonetheless, we did not observe qualitative changes after stimulation with different numbers of electrodes, and therefore all stimulation sessions were pooled. Data analysis. Cells were selected for recording as a function of their signal-to-noise ratio and real-time assessment of their isolation quality. Only stable (off-line verification of the stability of the neurons firing rates throughout the stimulation session; we discarded any neuron that demonstrated a trend of decaying or increasing firing rate because this is indicative of possible neuronal injuries or unstable electrode position) and well isolated (as judged by the experimenters during real time) units were included in the analysis database of this study. Because of stimulus artifacts (see below), off-line quantification of the isolation quality (Joshua et al., 2007) could not be applied to the data set. The stimuli produced prominent artifacts in the recordings, including short-term saturations of the recording electrodes. The nonsaturating artifacts could be overcome by stimulus artifact removal techniques (Wichmann, 2000; Hashimoto et al., 2002; Bar-Gad et al., 2004). However, in this study we chose to delete all stimulation periods from the neural data, including a 5 ms epoch after the last pulse, to provide a sufficient interval for the fading of the stimulus artifact. We therefore excluded a period of 40 ms for each stimulation burst from the data analysis. We classified the neurons into different populations according to their location and the location of the stimulating electrodes. Thus, we studied the responses of MI neurons to GP stimulation (GP3 MI), responses of MI neurons to MI stimulation (MI3 MI), as well as MI3 GPe, MI3GPi, GP3GPe and GP3GPi. Each population was further divided into the normal and MPTP states. We applied two main analyses on each neural population: peristimulus time histogram (PSTH) analysis and decrease increase analysis, as detailed below. Averaged population response: PSTH analysis. Responses to stimulation bursts of each neuron (n) were first characterized in each frequency stimulation ( f) by their PSTH, constructed in 5 ms bins for a period of 600 ms around the stimulus (see examples in Figs. 2, 3): 4 (Figure legend continued.) adapted from Martin and Bowden (2000). C, The stimulation pattern includes 35 ms bursts with intraburst frequency of 200 Hz (8 pulses per burst); each pulse is symmetric biphasic, negative phase leading. The bursts were given at different frequencies for 20s: 1(atotalof20bursts), 2, 5, 10Hz, and, formonkeyw, also15hz(atotalof300bursts). The interval between the frequency tests was 15 s, and the order of frequencies was 5, 10, 1, 2 and, 10, 1, 2, 5, 15 Hz, for monkeys T and W, respectively. D, E, Photomicrographs of TH staining demonstrating the loss of dopaminergic substantia nigra pars compacta neurons in the MPTPtreatedmonkeyTcomparedwithacontrolanimal.Dwastakenfromacontrolnormalmacaque monkey; E is from the MPTP-treated monkey T (vervet). The photomicrographs illustrate the levels of rostral striatum (column 1), central striatum (column 2), and midbrain (columns 3, 4). Note the lack of TH-positive staining throughout the striatum with the exception of the ventral striatum, particularly the shell region. TH-positive cells are selectively lost in the ventral tier but sparedintheventraltegmentalarea. C, Caudate; P, putamen; VS, ventralstriatum; SN, substantia nigra; VTA, ventral tegmental area; E, GPe; I, GPi. 38 i f Stim trial n,f i i 1 response PSTH n,f, i f where Stim trial n,f (i) is the response of neuron n to the ith burst given in frequency test f (includes 100 ms before burst and 500 ms after beginning of the burst), and i f is the total number of bursts in frequency test f. When illustrating the response PSTH of single neurons (see Figs. 2, 3) we further smoothed the PSTH with a Gaussian window, 15 ms. The PSTH of each neuron was then normalized by its spontaneous firing rate, which was determined by the mean firing rate in the last 10 s of all IStimIs (a total of 50 and 60 s in monkeys T and W, respectively). Normalization was used to avoid bias of population means attributable to neurons with a high discharge rate, and was essential especially in the MI population, in which many of the neurons had a low discharge rate (see Table 3), and outlier neurons with a high discharge rate could bias the population behavior. To estimate the population response we averaged the normalized PSTH for each frequency independently over all cells and obtained the averaged normalized response. Repeating the analysis with the nonnormalized discharge rates led to qualitatively similar results. In the average PSTH analysis, as well as in all other analyses, neurons from the two monkeys usually demonstrated similar behavior. Therefore, unless specified otherwise, neurons from monkeys T and W were pooled. Modulations in firing rate: decrease increase analysis. The average PSTH characterizes the population response only if the responses of the neurons are not divergent (i.e., most or all neurons in the population tend to increase their firing rate together, or decrease the rate together). In any other case, opposing modulations may cancel each other in the average PSTH, and thus will be lost to the analysis. Therefore, in addition to the PSTH analysis, we separated the responses of each neuron into bins with increases and decreases in firing rate. Significant modulations in the firing rate of a single neuron were identified by comparing each bin in the response PSTH of the neuron with its matched IStimI mean discharge rate. To do so, we calculated the IStimI PSTH of each neuron (n) for each frequency ( f) based on the IStimI epochs: i f IStimI trial n,f i i 1 IStimI PSTH n,f, i f where IStimI trial n,f (i) is a trial of 600 ms chosen randomly from all IStimI epochs of neuron n, and i f is the total number of bursts in frequency test f. We then calculated the mean and the SD of the IStimI PSTH and compared the values of the response PSTH with these expected values. Bins in which the response of the neuron deviated from the mean of the IStimI PSTH by 2.58 SD ( p 0.01) were considered to have significant increases or decreases. Based on this single neuron analysis, we determined the fraction of neurons in the population that significantly increased or significantly decreased their firing rate in every bin. Finally, for each population, frequency, and state, we calculated the maximum fraction of neurons that responded by increasing (decreasing) their discharge rate according to the bin with the maximum response percentage. This bin also determined the mean latency of the response (we chose not to define the latency according to response onset time, because the response could start before the stimulation burst ended). Response duration was calculated as the largest continuous period in which at least 50% of the responding neurons had a significant ( p 0.01) change in their firing rate. In the decrease increase analysis we used different bin sizes for MI and GP neurons, as well as for different stimulation frequencies. For GP neurons, which had a high discharge rate (see Table 3) we used 10 ms bins, and for the MI neurons that fired at a much lower rate we used 25 ms bins for the 1, 2 and 5 Hz stimulations. Because the power of the test was higher in the high stimulation frequencies that contained many repetitions of stimulation bursts, we used 10 ms bins for the 10 Hz frequency for MI neurons as well. In the 15 Hz stimulation frequency (monkey W),

41 Results II Rivlin-Etzion et al. Filter Properties of Basal Ganglia Cortical Loops J. Neurosci., January 16, (3): the time lapse between the bursts was very short ( 66.7 ms, which left 26.7 ms after excluding the 40 ms stimulation and artifact period) and therefore the 15 Hz stimulation tests were not included in the decrease increase analysis. Diversity index. Although the decrease increase analysis reveals the prevalence of coincident modulations in the firing rate, it is not always easy to interpret. To simplify the interpretation, we defined a diversity index for each population: diversity pop,f 1 mean i I pop,f i D pop,f f I pop,f i D pop,f i, where represents the absolute value, I pop,f (i) and D pop,f (i) are the percentages of neurons that significantly increased and decreased, respectively, their firing rates in the ith bin in frequency test ( f) and population (pop). Bins included in the calculation of the diversity index start from the end of the first stimulation burst (time 40 ms) and end at time 500 ms or before the beginning of the next stimulation burst (in the high frequencies). The index ranges between 0 and 1. If I D (and both greater than 0) over all bins the index equals 1, that is to say that the population did not tend to increases or decreases alone, but rather exhibited both kinds of modulations simultaneously and therefore was highly divergent. If, for example, I 0 and D 0, the index value is 0; in other words, the population did not show opposing modulations simultaneously, and therefore was not divergent. If both I 0 and D 0 for all bins (here we defined zero as I,D 0.01 although taking a smaller value produced similar results), the population responded neither by increases nor decreases to the stimulation test, and therefore the diversity index was set to 0 for this test. Accelerometers. We used a uniaxial accelerometer (ACC) (8630C5; Kistler, Amherst, NY) to assess hand movements. Both monkeys had the accelerometer fastened to the back of their nonrestrained left wrist (contralateral to the stimulating hemisphere). The accelerometer output was amplified with a gain ranging from 4 to 20 (depending on the amplitude of the movement caused by the cortical stimulation) and bandpass filtered with a Hz four-pole Butterworth filter (MIP ; Alpha Omega Engineering). The analog output of the accelerometers was sampled at and 781 Hz in monkeys T and W, respectively. In the population analysis, the data were further digitally filtered with an eighth order Chebyshev type I 20 Hz low-pass filter and re-sampled at 200 Hz for both monkeys. Spontaneous movements, as well as other artifacts were noticeable in the accelerometer data from time to time. We therefore only included accelerometer records that had a good signal-to-noise ratio and did not include such spontaneous movements or artifacts. Because stimulation caused a different profile of movement in different recording sessions, the average population response of the arm acceleration was calculated from the normalized absolute value of the accelerometer output. Normalization of the acceleration value was obtained as for the neural data. In each recording session, an IStimI PSTH was derived for each of the frequencies based on the absolute ACC value in the IStimI periods, with a number of segments equal to the number of bursts delivered at that frequency. Then, each response PSTH of the absolute ACC output was normalized by its corresponding baseline IStimI mean value. MPTP treatment and perfusion. After a period of recording in the normal state, Parkinsonism was induced by five intramuscular injections of 0.4 mg/kg 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine-HCl (MPTP; Sigma, Rehovot, Israel). The MPTP injections were given under light intramuscular ketamine hydrochloride (10 mg/kg) anesthesia and over a period of 4d(3injections in the first 24 h). The clinical state of the monkeys was assessed daily according to a primate scale of Parkinsonism (Benazzouz et al., 1995). During termination of the recording days in the MPTP state we treated monkey T with /250 mg of Dopicar (L- 3,4-dihydroxyphenylalanine and carbidopa; Teva Pharmaceutical Industries, Petach Tikva, Israel) twice per day to verify the diagnosis of Parkinsonism by significant clinical improvement achieved with dopamine-replacement treatment. The drugs were administered orally as crushed powder dissolved in liquid. At the end of the experiment, monkey T was deeply anesthetized with a lethal dose of pentobarbital and perfused through the heart with saline, followed by a 4% paraformaldehyde fixative solution. Monkey W was perfused in a similar way within 30 Table 1. Summary of the neuronal database # of stimulation sessions #ofmi neurons #ofgpe neurons #ofgpi neurons Monkey T CTX stim. 54N, 55P 146N,166P 57N,143P 9N, 34P GP stim. 62N, 66P 238N,259P 46N,104P 3N, 42P Monkey W CTX stim. 79N, 23P 184N, 88P 99N, 86P 36N, 10P GP stim. 48N, 17P 197N,107P 47N, 48P 27N, 3P A summary of the number of stimulating sessions (left column) and the number of MI, GPe, and GPi neurons that wererecordedwhilestimulatingineachoftheareas.neuronsincludedherearethosesatisfyingtheinclusioncriteria for analysis (see Materials and Methods). Number of stimulating sessions and recorded neurons is given for normal and MPTP states. N, Normal state; P, MPTP state. Table 2. Summary of cortical stimulating sessions that produced movements # of sessions that produced movement/ # of all sessions (percentages) Normal state MPTP state Monkey T 34/54 (63%) 20/55 (36%) Monkey W 49/79 (62%) 14/23 (60%) A summary of number (and percentage) of cortical sessions that produced an arm movement that was recorded by the accelerometer out of the total number of cortical stimulating sessions, in both normal and MPTP states. We defined a session with a clear and robust response to at least a single frequency test as a session that produced movement. min of her death. Brains were removed and cryoprotected in increasing gradients of sucrose (10, 20, and finally 30%). Adjacent serial sections of 50 m, from both control animals and MPTP-treated animals, were processed for either a Nissl stain or immunocytochemistry for tyrosine hydroxylase (TH). Sections were incubated with antisera to TH (mouse anti-th, 1:20,000; Eugene Tech, Allendale, NJ) in 0.1 M phosphate buffer with 0.3% Triton X-100 and 10% normal goat serum (Incstar, Stillwater, MN) for 4 nights at 4 C and further processed using the avidin biotin method (rabbit Elite Vectastain ABC kit; Vector Laboratories, Burlingame, CA). Results Animals clinical states We recorded from the arm area of the primary motor cortex and the globus pallidus of two vervet monkeys (T and W). After 24 and 19 recording days in the normal state for monkeys T and W, respectively, the monkeys were systemically treated with the neurotoxin MPTP. The first signs of Parkinsonism appeared within 5 d from initiation of the MPTP treatment in monkey T, and included muscle rigidity, akinesia and bradykinesia. Five days later, the monkey developed a low-frequency episodic tremor, mainly in the distal limb muscles. We started the recordings of monkey T in the MPTP state on the 11th day after the first MPTP injection and the monkey remained in a stable condition of severe Parkinsonism during all recording days with an average Parkinsonism score of 19.7/25 (Benazzouz et al., 1995). Dopamine replacement therapy started 21 d after the last MPTP injection and the first response to therapy was seen within 24 h. The dopamine therapy effects included the resumption of the ability to self-feed, straightening of posture, and an increase in the amount and velocity of movement. The dense TH immunoreactivity throughout the striatum of the control monkey (Fig. 1D) was not observed in the striatum of Monkey T, with the exception of the shell of the ventral striatum (Fig. 1E). Cell loss in the midbrain was almost complete in the ventral tier and the lateral portions of the substantia nigra pars compacta. As expected (Song and Haber, 2000), the cells in the ventral tegmental area of the midbrain dopamine system remained relatively spared. 39

42 Results II 638 J. Neurosci., January 16, (3): Rivlin-Etzion et al. Filter Properties of Basal Ganglia Cortical Loops Table 3. Summary of average firing rates in the cortex and pallidum before and after MPTP treatment CTX GPe GPi Normal MPTP Normal MPTP Normal MPTP Monkey T (n 384) (n 425)** (n 103) (n 247)*** (n 12) (n 76) Monkey W (n 381) (n 195) (n 146) (n 134)*** (n 63) (n 13) Discharge rates (spikes/s) are given as mean SD for MI, GPe, and GPi neurons. The rates are calculated based on the IStimI periods between the different frequency tests. We used the 10 s epochs that start 5 s after the last burst of the stimulation. n stands for the number of neurons on which the values are based. **p 0.01, ***p significant differences of MPTP values versus corresponding normal values. Monkey W developed akinesia and bradykinesia, as well as prolonged episodes of low-frequency tremor within 5 d from the first MPTP injection. On the eighth day, monkey W was moderately rigid. Recordings from monkey W in the MPTP state started on the seventh day after the first injection. The average Parkinsonism score of monkey W was 15.1/25 during the MPTP recording days. We lost monkey W unexpectedly 11 d after the start of the MPTP injections, after 5 d of recordings. Experimental protocols During the experiment, we stimulated either in MI or in the GP in each stimulation session, while recording the evoked neural activity in those areas as well as the contralateral arm movements. The stimulation pattern consisted of 35 ms bursts (8 pulses at 200 Hz) given at different frequencies (for details, see Materials and Methods). Table 1 summarizes the number of stimulating sessions and the number of recorded neurons in each monkey, state and area. Table 2 summarizes the total number and the fraction of MI stimulation sessions that led to arm movement. Table 3 summarizes the average firing rates of the neurons calculated at the periods between the stimulations (IStimI). In line with previous studies, after MPTP a significant (Student s t test, p 0.001) decrease in the discharge frequency of GPe neurons, but not in GPi neurons, was observed (Bergman et al., 1994; Boraud et al., 1998; Raz et al., 2000; Heimer et al., 2002). We did not observe a significant difference in the discharge rate of monkey W s MI neurons [also consistent with previous studies (Doudet et al., 1990; Goldberg et al., 2002)], although the discharge rate of MI neurons of monkey T increased slightly after MPTP ( p 0.01). Responses of single neurons in the frequency domain Many of the MI and GP neurons responded to MI and/or GP stimulation. Some examples of single neuron responses to the different frequency tests are depicted in Figures 2 and 3. Figure 2A shows the response of a single neuron in the arm related area of the primary motor cortex to stimulation in that area in the normal state. Examples of the analog traces (bandpass filtered) containing the responses to single bursts in the different frequency tests appear in the upper three rows. The raster plots and the PSTHs are shown in the lower two rows for each of the frequency tests. The neuron responded to each burst by a strong inhibition that lasted 200 ms. As a result, in the 5 Hz and in higher frequency tests the cell hardly produced any action potentials in the interburst periods. Many MI neurons responded to MI stimulation with decreases in their firing rates that often resulted in zero spikes for a certain period. Naturally, the duration of this silent period determined the stimulation frequency at which MI neurons stopped firing action potentials. Another MI neuron that was recorded in the MPTP state (Fig. 2B) demonstrated a triphasic response to MI stimulation bursts: an initial short phase of elevation of the firing rate, followed by a 200 ms of suppression of discharge rate, and a rebound excitation. The rebound excitation phase was attenuated in frequencies of 5 Hz and higher. Moreover, the first phase of excitation was also reduced as the frequency of stimulation increased, despite its short latency. An example of the response of an MI neuron to GP stimulation in the MPTP state is illustrated in Figure 2D. The neuron responded by a sharp increase in firing rate that occurred 200 ms after the end of the burst. As in the case shown in Figure 2B, the response was not seen in the 5 Hz and higher frequency tests. Figure 3 depicts examples of GP responses to GP stimulation (Fig. 3A,B) and to MI stimulation (Fig. 3C,D). For example, the GPe neuron in Figure 3A tended to increase its firing rate immediately after the GP microstimulation burst. In contrast with the long duration cortical responses, the response of this GPe neuron lasted 50 ms after the burst, after which the neuron resumed its original firing rate. Furthermore, in contrast to the first excitation phase of the MI neuron (Fig. 2B) the increase in the firing rate of this GPe neuron did not change between the 1 Hz frequency test and the 10 Hz frequency test. The decrease seen in the PSTH of the 15 Hz frequency test is the result of the smoothing which was performed for purposes of illustration of the single cell PSTH (for this reason we did not use smoothing in the calculation of the population PSTH, see below). Indeed, the spike density as seen in the examples of the analog traces did not seem to decrease in the 15 Hz frequency test. This suggests that the responses of the GPe neuron in this example were not frequency dependent. MI population response to stimulation in the MI: normal and MPTP Figure 4A shows the population response of neurons in the arm related area of the primary motor cortex to microstimulation in that area in the normal and MPTP states. The normalized population PSTH in both states shows that the response is attenuated as the frequency is increased (as in the single cell examples) (Fig. 2A,B). The main difference between the normal and the MPTP states is the rebound excitation phase that occurs at the low frequencies (1 and 2 Hz stimulations) and is apparent only in the parkinsonian population. The decrease increase analysis revealed that different MI neurons did not tend to simultaneously increase or decrease their firing rate. Rather, the first bin in the decrease increase plots reveals a small fraction of MI neurons with increases in firing rate, whereas decreases become dominant from the second bin on (Fig. 4A, bottom) (see Table 5). In the 1 2 Hz tests of the MPTP state, in which the rebound excitation is apparent ( 250 ms after the burst), the fraction of decreases drops as the fraction of increases rises. The mean diversity index of the population (see Materials and Methods) was 0.15 and 0.35 in the normal and MPTP states, respectively (Table 4). These values indicate a relatively similar temporal profile of neuronal responses compared with the more divergent responses found in the GP neurons (see 40

43 Results II Rivlin-Etzion et al. Filter Properties of Basal Ganglia Cortical Loops J. Neurosci., January 16, (3): Figure 2. Responses of single primary motor cortex (MI) neurons to stimulation. A, MI neuron response to MI stimulation in the normal state. Three top rows, The bandpass filtered analog data recorded from an electrode located in MI while stimulating through another MI electrode in monkey W. First, second, and third rows show the response to the 2nd, 10th, and 20th bursts in each frequency test, respectively. All periods of stimulation bursts(40 ms per burst, to enable recovery from the last stimulus artifact) were off-line deleted and are marked by red lines. Fourth row, Raster displays of the neuron. Each dot in the raster represents an action potential of the neuron. For illustration purposes, in conditions with 20 trials (i.e., for 2 Hz and higher-frequency tests), only a subsetof20equallydistributedtrialsisshownintherasterplots. Bottom, Themeanfiringratesalignedonthestimulationbursts(PSTH). Binsize, 5ms. PSTHsaresmoothedwithaGaussianwindow, 15 ms. Note that the analog data recorded from the electrode reveal action potentials from more than a single neuron (see burst 2 of the 2 Hz frequency test), but here we focused on the cell with the highest signal-to-noise ratio. B, MI neuron response to stimulation in MI in the MPTP state of monkey T, same convention as A. C, D, Responses of MI neurons to stimulation in GP in the normal and MPTP states, monkey T, same convention as A. below). Note that the fraction of increases in the early excitation phase is underestimated because of the 25 ms binning that blends the short period of increase in the firing rate (a few milliseconds) with the decrease that follows it. However, early excitation was found only in a small fraction of MI neurons and the majority of MI neurons in the normal state responded to MI stimulation with suppression of their discharge rate after the stimulation burst (Fig. 2A). 41 MI population response to stimulation in the GP: normal and MPTP In the normal state, MI neurons were not affected by GP microstimulation (see example in Fig. 2C). This is confirmed by the population PSTH that does not change as a result of the stimulation bursts in the GP in the normal state (Fig. 4B, top, green line). The decrease increase analysis indicates very small percentages of neurons that significantly modified their firing rates as a result

44 Results II 640 J. Neurosci., January 16, (3): Rivlin-Etzion et al. Filter Properties of Basal Ganglia Cortical Loops Figure 3. Responses of single pallidal neurons to stimulation. A, GPe neuron response to stimulation in GP in the normal state of monkey W. B, GPi neuron response to stimulation in GP in the MPTP state of monkey W. C, GPi neuron response to stimulation in MI in the normal state of monkey T. D, GPe neuron response to stimulation in MI in the MPTP state of monkey W. Conventions as in Figure 2A. 42 of the stimulation bursts (Fig. 4B, bottom, green line). Nonetheless, in the MPTP state, the efficacy of the functional connectivity between GP and MI appears to increase, and the MI neuron shown in the example (Fig. 2D), as well as the population as a whole (Fig. 4B, top, black line), do respond to the GP stimulation bursts. Although the MI neurons of both monkeys elevated their discharge rate after the GP stimulation, we noted a difference in the timing of these rate modulations (Table 5). The populations from the two monkeys in the MPTP state are therefore plotted separately. In both monkeys, this cortical response was attenuated from the 5 Hz frequency test and up. The decrease increase analysis reveals that the MI population did not tend to opposing modulations in the firing rate in the MPTP state in response to the GP stimulation (Fig. 4B, bottom, black line). In accordance, the mean diversity index demonstrates low values for both monkeys (0.26 and 0.31 for monkeys T and W, respectively, vs 0.48 in the normal state) (Table 4). Despite the substantial cortical response observed in the population PSTHs of 1 and 2 Hz stimulations, the maximal fraction of neurons that responded with significant increases in the firing rate was 13%. This may indicate that a small cortical area was activated by the microstimulation, and that despite the increase in the functional connectivity between the GP and the MI, a state of all-to-all connectivity was not attained. GP population response to stimulation in the GP: normal and MPTP The GP stimulation in both states produced various responses in the neurons of the GP. When examining the populations as a

45 Results II Rivlin-Etzion et al. Filter Properties of Basal Ganglia Cortical Loops J. Neurosci., January 16, (3): Figure 4. MI population response to stimulations. A, MI population response to stimulation in MI in the normal and MPTP states. Top, The population PSTH of all recorded cortical neurons in responsetocorticalstimulation.psthswerenormalizedbythespontaneousfiringrateofthecell(seematerialsandmethods).meanpsthisshownasagreenlineforthenormalstateandasablack line for the MPTP state. Bottom, Fraction of cells with significant modulations in firing rate around the stimulation burst. The ordinate is the fraction of cells that had a significant response at each time bin (bin size was 25 ms in all frequencies except 10 Hz stimulation, in which bin size was 10 ms). The values above 0 are the fraction of cells that increased their firing rate, and the values below 0 are the fraction of cells that decreased their firing rate. The total number of recorded neurons in each of the states is given in the legend, and the numbers in parentheses detail the number of neurons recorded from monkeys T and W separately. B, MI population response to stimulation in GP in the normal and MPTP states. The green line represents the population of both monkeys in the normal state, black solid line represents the population of monkey T in the MPTP state, and dotted black line represents the population of monkey W in the MPTP state. All other conventions as in A. whole, GPe neurons responded with short increases in their firing rate (as in the example in Fig. 3A), but the decrease increase analysis also reveals decreases of discharge rate (Fig. 5A; Table 5). The values of the diversity index were indeed higher than those in the cortex (mean values 0.79 and 0.77 for the normal and MPTP states, respectively) (Table 4). The different types of responses in the nonhomogenous GPe neural population counteract, and thus lead to the relatively small elevation seen in the averaged PSTH population. There was no noticeable change between the GPe response to GP stimulation in the normal and in the MPTP states. In both states the responses of the GPe neurons, increases as well as decreases, were independent of the frequency of the stimulation burst (see also the single GPe neuron example in Fig. 3A). In contrast to the short and frequency-independent GPe responses, GPi neurons revealed a mixed frequency response in the normal state. The GPi neurons tended to have long decreases in their firing rates immediately after the stimulation burst, as well as delayed increases in the lower frequency tests (Fig. 5B, green lines; Table 5). As opposed to the GPe population, in which increases and decreases appeared simultaneously, the increases and 43 decreases in the GPi were rarely coincident, as is reflected as well by the diversity index (mean value, 0.29) (Table 4). In the parkinsonian state, the responses of GPi neurons became more divergent, and early increases were observed together with decreases (Fig. 3B) in all frequency tests (Fig. 5B, black lines), and mean diversity index rose to 0.57 (Table 4). Nevertheless, GPi neurons showed prolonged responses to GP stimulation in the MPTP state, as in the normal state. The divergent responses of the GP neurons relative to the MI neurons could be attributed to their high frequency tonic discharge rate, enabling them to respond either by an increase or a decrease of their discharge rate. Such response diversity is in line with intrinsic neuronal mechanisms directed at maximizing the information capacity of the pallidal network (Bar-Gad et al., 2003), whereas the cortical network is more redundant (Ben Shaul et al., 2003). GP population response to stimulation in the MI: normal and MPTP Figure 5, C and D, illustrates the population response of GPe and GPi neurons to stimulation in the MI. Consistent with the

46 Results II 642 J. Neurosci., January 16, (3): Rivlin-Etzion et al. Filter Properties of Basal Ganglia Cortical Loops absence of effect of GP stimulation on MI neurons in the normal state, the population PSTH confirms that in the normal state GP neurons in both segments did not change their discharge rate as a result of microstimulation in MI (see the single cell example in Fig. 3C; population analysis in Fig. 5C,D, green lines). The decrease increase analysis confirms that the flat GP responses are not attributable to balanced positive and negative responses. However, minor changes that include both increases and decreases are seen mainly in the 5 Hz test, which might reflect some resonance properties of this pathway at this frequency. In the MPTP state, cortical microstimulation did affect the discharge of GP neurons. Although the GPe population PSTH did not significantly differ from the normal state (Fig. 5C, black lines), a small fraction of the recorded GPe cells did respond to the MI stimulation (Fig. 3D). Moreover, the GPi neuronal population exhibited enhanced functional connectivity with the MI as revealed by the increases in its mean firing rates as a result of MI microstimulation at 1 and 2 Hz (Fig. 5D, black lines; Table 5). The mean diversity indexes had similar values for both segments of the GP in each of the states (Table 4), and a reduction in the diversity of GP responses to MI stimulation was observed after the MPTP treatment. We did not observe phase-locking of the discharge of pallidal neurons as a result of the cortical stimulation. This contrasts with the long-lasting oscillatory GPi activity observed after striatal stimulation (albeit with a much higher current, 500 A) in the MPTPtreated monkey (Tremblay et al., 1989). Muscle response to stimulation in MI: normal and MPTP Many of the cortical microstimulations produced forearm, wrist or finger movements (Table 2). Whether a movement occurred was a function of many factors, including the exact X-Y coordinates of the stimulation (center of the arm related motor cortex or periphery), electrode depth, the current amplitude (ranged A for each stimulating electrode), and the alertness of the animal [a stimulation that has been shown to elicit a response in an animal while alert may not elicit any response when the animal is drowsy (Alexander and DeLong, 1985)]. A few stimulations led to small thumb movements that were not recorded by the accelerometer which was located on the back of the wrist. These stimulations were categorized as no-movement stimulations. Figure 6, A and B, shows examples of the recorded acceleration of the hand in response to microstimulation in the MI in the normal and MPTP states. A reduction in the movement amplitude is seen in both examples in the frequency tests of 10 Hz and higher. The single burst examples reveal that although there was movement at the beginning of stimulation test (burst 2) in these frequencies, it rapidly decayed. The 2 and 5 Hz frequency tests, on the contrary, produced a movement that became stronger with time, as can be seen by comparing the single bursts with the normalized mean of all bursts (ACC PSTH). These examples are in line with the population average, which exhibited a substantial response for the 1 Hz stimulation, and greater response for the 2 and 5 Hz stimulations. A steep decay is seen in the 10 Hz frequency test, and there was no movement in response to the 15 Hz MI frequency test (Fig. 6C). Table 4. The diversity index Mean (index) 1 Hz 2 Hz 5 Hz 10 Hz CTX3 CTX Normal MPTP GP3 CTX Normal MPTP (T) MPTP (W) GP3 GPe Normal MPTP GP3 GPi Normal MPTP CTX3 GPe Normal MPTP CTX3 GPi Normal MPTP Diversity index values, representing the diversity of the neuronal responses to stimulation, are given as mean SD in each frequency test. The average over frequencies of all indexes is given in the Mean (index) column. Diversity index values range from 0 (nondivergent) to 1 (maximal divergence). 44 Muscle response to stimulation in GP: normal and MPTP There was no muscle response to stimulation in the GP in the normal state (in agreement with Horak and Anderson, 1984). However, in the MPTP state, a small reaction was seen in the population average in frequency tests of 1 and 2 Hz, for both monkeys (data not shown). This movement could have been the outcome of the stronger connectivity between GP and MI in the MPTP state, as derived from the cortical neural response to GP stimulation (Figs. 2D, 4B). The temporal profile of the neural and muscle responses to stimulation bursts The duration of each frequency test was 20 s, resulting in a different number of bursts for each of the frequency tests. Our findings indicate low-pass filter properties in the cortex and GPi, in which the responses decay with the increasing frequency stimulation. However, this decrease may be the outcome of the growing number of stimulations rather than the stimulation frequency. For this reason, we averaged the responses of the populations to each burst separately in each of the frequency tests, and looked at the mean normalized firing rate in the 25 ms epoch that followed the burst as a function of the number of bursts. We chose a period of 25 ms because this is the minimal common period between bursts (determined by the 15 Hz frequency test, after removing the stimulation period) for all frequency tests. The results of this analysis are shown in Figure 7. The neuronal responses to the different frequency tests of almost all pathways studied were independent of the number of bursts. An exception is the MI population whose responses to the first 20 stimulations in the MI differed for the low- (1, 2 Hz) and high- (5, 10, and 15 Hz) frequency tests. The latter show a considerable elevation of their initial responses that decayed immediately (within the first bursts), as opposed to the stable response in the low frequencies. A similar analysis of the arm acceleration revealed that the muscle responses to the high-frequency tests but not to the lowfrequency tests also decayed with the number of the bursts (Fig. 6A,B, responses to bursts 10 and 20 vs response to burst 2 in the 10 and 15 Hz frequency tests). However, the slope of the arm acceleration curve was more moderate than the slope of the MI population, especially in the 5 Hz stimulation. To summarize, MI

47 Results II Rivlin-Etzion et al. Filter Properties of Basal Ganglia Cortical Loops J. Neurosci., January 16, (3): Table 5. Main effects of microstimulation 1Hz 2Hz 5Hz 10Hz Dec Inc Dec Inc Dec Inc Dec Inc CTX 3 CTX Normal %Respond Latency (ms) Duration (ms) MPTP %Respond /10.1 a /9.7 a Latency (ms) /250 a /275 a Duration (ms) /275 a /100 a GP 3 CTX Normal MPTP %Respond 16.1/15.3 b 16.1/10.6 b Latency (ms) 250/175 b 225/175 b Duration (ms) 100/75 b 100/100 b GP 3 GPe Normal %Respond Latency (ms) Duration (ms) MPTP %Respond Latency (ms) Duration (ms) GP 3 GPi Normal %Respond Latency (ms) Duration (ms) MPTP %Respond Latency (ms) Duration (ms) CTX 3 GPe CTX 3 GPi Normal MPTP %Respond Latency (ms) Duration (ms) A summary of the main effects of microstimulation, including percentages of neurons responding by decrease (Dec) and by increase (Inc) in the firing rate, latencies (from beginning of the stimulation burst until the peak response), and durations of the responses. The values are given for each neural population and frequency stimulation, in each of the states (normal and MPTP). Details are given only for structures and stimulation frequencies with a significant number of responses, and populations that had no significant response in any of the frequencies are in bold. a First and second values (separated by /) refer to the first excitation phase and to the rebound excitation phase, respectively. b First and second values refer to monkeys T and W, respectively. In all other cases, no significant differences were found between monkeys T and W, and their results are grouped. neurons, as well as muscles, present a response that depends on the number of bursts when stimulating the MI with high-frequency bursts. This attenuation is in line with the low-pass properties of these pathways (see below). The responses of all other recorded populations were independent of the number of bursts and therefore are only dependent on the frequency of stimulation. Frequency-domain analysis of the basal ganglia cortex muscle pathways Figure 8 summarizes the frequency dependency of the responses of MI, GPe and GPi neurons to stimulation in each of the structures (MI and GP) in the frequency range tested in this study (1 15 Hz). The figure depicts the maximum value of the normalized population PSTH (Figs. 4, 5) as a function of the stimulating frequency. The maximum absolute value of the normalized arm acceleration to MI stimulation (Fig. 6) is also depicted. The responses of the GPe population to microstimulations in both MI and GP are independent of the frequency of the stimulation bursts (Figs. 5A,C, 8A). In contrast, GPi neurons changed their responses as a function of the frequency. In particular, in the parkinsonian state, their firing rate was elevated in response to 1 and 2 Hz frequency tests, but not in the higher frequency tests (Figs. 5B,D, 8A). MI neurons revealed more robust low-pass filter responses to their inputs. The response of the cortical neurons to MI stimulations decayed with frequency, such that hardly any activation was seen with burst stimulations of 10 and 15 Hz (Fig. 4A,8A). The responses of the cortical neurons to stimulations in the GP had a lower cutoff frequency and responses to stimulations were no longer observed as of the 5 Hz frequency test (Figs. 4B, 8A). Furthermore, MI stimulations of 10 Hz and higher frequencies were not transferred to the periphery, at least not at the level of movements recorded by the wrist accelerometer (Figs. 6, 8B). Discussion In this study, we explored the functional microconnectivity within the cortex BG and cortex periphery loops throughout the normal state and in MPTP-induced Parkinsonism. We recorded simultaneously the acceleration of the arm and the spik- 45

48 Results II 644 J. Neurosci., January 16, (3): Rivlin-Etzion et al. Filter Properties of Basal Ganglia Cortical Loops Figure 5. GP population response to stimulations. A, GPe population response to stimulation in GP in the normal(green) and MPTP(black) states. B, GPi population response to stimulation in GP in the normal and MPTP states. C, GPe population response to stimulation in MI in the normal and MPTP states. D, GPi population response to stimulation in MI in the normal and MPTP states. Conventions as in Figure 4A, except bin size, which was 10 ms for all frequencies. 46

49 Results II Rivlin-Etzion et al. Filter Properties of Basal Ganglia Cortical Loops J. Neurosci., January 16, (3): Figure 6. Arm acceleration as a result of MI microstimulation. A, An example of the acceleration recorded from an accelerometer located on the back of the wrist while stimulating through an electrode in the contralateral motor cortex of monkey W in the normal state. First, second, and third lines show the responses to the 2nd, 10th, and 20th bursts in each frequency, respectively. All periods of stimulation bursts (40 ms per burst) are marked by red lines. Bottom, Mean acceleration values of the session aligned on the stimulation bursts. Ordinate units are arbitrary (A/D values). B, An example of arm acceleration recorded during MI stimulations in monkey T in the MPTP state. Conventions as in A. C, The normalized population PSTH of the absolute acceleration values in response to MI stimulation in the normal and MPTP states. Only sessions in which MI stimulation evoked movements are included. The average of the normalized responses is shown in green for the normal state and in black for the MPTP state. Other conventions as in Figures 4 and 5. ing activity from the contralateral MI arm related area and the GP of two vervet monkeys while microstimulating in one of the structures. To understand the role of oscillatory bursts encountered in the parkinsonian brain (Heimer et al., 2006; Weinberger et al., 2006) and the relationship between the GP, MI, and the periphery in the frequency domain, we mimicked the parkinsonian oscillatory activity using a stimulation pattern that contained 35 ms bursts delivered at different (1 15 Hz) frequencies. Because cortical neurons demonstrate bursting activity during normal movements (Georgopoulos et al., 1986; Evarts, 1966; Goldberg et al., 2002), the stimulation pattern also enabled us to characterize the transfer function between the MI and muscles. We report two main findings: first, the functional connectivity between the MI and the GP are greatly enhanced in the MPTP state. Second, in both the normal and the MPTP states, the BG MI muscle circuit demonstrates low-pass filter properties. Functional connectivity between MI and GP The functional microconnectivity between MI and both segments of the GP is weak in the normal state. A comparison of our findings and previous reports of extensive GP activation by macrostimulation of MI (Nambu et al., 2000) suggests that in the normal monkey the MI to GP functional connectivity is either 47 very specific or that converging inputs from many MI areas are needed to activate GP neurons. Had we focused on the hand related area of the GP, it is plausible that the results would have demonstrated stronger connections between the structures. Nevertheless, the MI GP reciprocal connections are strengthened in the MPTP state. These findings are not attributable to increased excitability of the parkinsonian MI, because cortical discharge rate and the fraction of stimulating sessions that produced movement did not increase after MPTP treatment. Rather, this enhancement in connectivity may have been caused by the reduction in the specificity of pallidal neurons (Filion et al., 1988), as well as by the increased synchronization of BG neurons in the dopamine depleted state (Bar-Gad et al., 2003; Strafella et al., 2005; Heimer et al., 2006), leading to widespread activation of the local effects of microstimulation. Low-pass filter properties of the BG cortical muscle network The cortex BG periphery loops demonstrate low-pass filter properties to the microstimulation pattern. The finding that MI does not follow high-frequency stimulations given in the GP is not surprising because of the multisynaptic pathway between the structures, especially because it contains inhibitory components

50 Results II 646 J. Neurosci., January 16, (3): Rivlin-Etzion et al. Filter Properties of Basal Ganglia Cortical Loops Figure 7. The neuronal response of most BG cortex pathways to the different frequency tests are not affected by the number of bursts in the test. The mean normalized firing rate [mean response (MR)] in the 25 ms that follow each burst as a function of the sequential number of burst. For illustration purposes, only up to the first 100 bursts are shown. The last row illustrates the average of the maximum amplitude in the normalized absolute arm acceleration as a function of the number of burst. All curves were smoothed using a moving average of five bursts. Green and black lines represent normal and MPTP states, respectively. (e.g., GPi to thalamus, and thalamus to inhibitory interneurons in the cortex). It has been suggested that BG - and -band oscillatory activity may arise in cortex (Hammond et al., 2007). The results of this study do not negate this possibility, because we did not find evidence for a prominent low-pass filter between MI and the two pallidal segments. Similarly, we do not expect the STN to demonstrate any substantial filtration properties. 48 The exact cutoff frequency of the different neuronal populations studied here is not known because we did not try all possible frequencies. Moreover, the cutoff frequency may change with the stimulation burst parameters [e.g., burst duration and frequency and number of the pulses/burst (McIntyre and Grill, 2002)]. Recent studies indicate a relationship between -band oscillations and akinesia (Chen et al., 2007). Because the GP oscillatory bursts

51 Results II Rivlin-Etzion et al. Filter Properties of Basal Ganglia Cortical Loops J. Neurosci., January 16, (3): Figure 8. Frequency domain analysis of the BG cortex muscle network. A, The maximum value of the normalized population PSTH of MI, GPi, and GPe neurons as a function of the stimulation frequency given in each of the structures (MI and GP). B, The maximum normalized absolute amplitude of the arm acceleration (ACC) as a function of the stimulation frequency in MI. A value of 1 stands for the mean ACC value in the no movement periods (when no stimulation occurred). Green and black bars correspond to the normal and MPTP states, respectively. are non-stereotypical, their different characteristics may determine their filtration properties as well as their role in the generation of PD symptoms. Low-pass filtering was also found between MI and the periphery level, in which MI stimulations of 10 Hz and higher did not evoke movement. Because we only recorded arm movements, we cannot rule out the possibility that this filtering is performed in muscle electrical-to-mechanical transfer. Indeed, many studies have reported significant coherence between MI and muscle EMG activity in the and even higher frequency ranges (Grosse et al., 2002). However, this cortex EMG synchronization has mainly been detected during static motor tasks, and is reduced or abolished during movement (Baker et al., 1997; Kilner et al., 1999; Salenius and Hari, 2003). The low-pass filter properties of the entire network are in line with natural frequencies of movement ( 2 Hz) (Freund and Hefter, 1993), low-pass filter properties of striato-pallidal pathway (Rav-Acha et al., 2005) and nerve to muscle transmission (Baratta et al., 1998). 49 DBS mechanisms in view of the low-pass filter properties of the network The low-pass properties of the GP MI periphery axis shed new light on the ongoing controversy regarding the mechanisms of high frequency ( 130 Hz) DBS in the treatment of advanced PD. There is considerable debate as to whether DBS mimics lesion and inactivates its targets (Wu et al., 2001; Filali et al., 2004; Maltete et al., 2007), or whether it drives them continuously at short latencies and higher frequencies (Hashimoto et al., 2003; Garcia et al., 2005). Because the bursts used in this study contained pulses given at 200 Hz, we can view the 10 and 15 Hz frequency tests as a fragmented DBS stimulation pattern, enabling us to examine the effect of DBS without the distortion of stimulation artifact. Our results suggest that the continuous DBS mechanism would similarly be filtered in the cortex or in previous levels. Although filtered, DBS indirectly alleviates the parkinsonian symptoms, probably by stopping (jamming) the abnormal oscillations of the BG (Benabid, 2003). Frequency dependency of tremor evoked by cortical stimulation A transcranial magnetic stimulation (TMS) study in which trains of repetitive stimulations at Hz were delivered to the MI of normal subjects revealed that the subjects developed an oscillatory movement (tremor) (Topka et al., 1999). Interestingly, the tremor frequency (4 7 Hz) was independent of the TMS frequency. In addition, an earlier study showed that electrical stimulation of MI during neurosurgery at a frequency of 60 Hz evoked a 5 Hz tremor, whereas Hz macrostimulation of the cortex resulted in movements of the same frequency as the stimulation (Alberts, 1972). Thus, movement evoked by 10 Hz stimulation may directly originate in the MI, whereas the tremulous movement that appears after higher frequencies of magnetic or electricalstimulation of the cortex may be attributable to the abnormal synchronous activity of the cortex. We assume that this abnormal synchronization indirectly causes tremulous movements that are the result of diminished cortical control, and probably reflect resonance properties of brainstem spinal and muscle networks. The relationship between PD tremor and BG oscillations The low-pass filter properties of the GP MI muscle networks described above have implications for our understanding of the pathophysiology of PD. Many previous studies have described oscillations in similar, or higher than, tremor frequencies in the BG of MPTP monkeys (Bergman et al., 1994; Raz et al., 1996, 2000, 2001) and in human patients (Timmermann et al., 2003; Brown, 2006; Weinberger et al., 2006). However, the correlations between the BG oscillations and the tremor are transient and intermittent (Hurtado et al., 1999; Lemstra et al., 1999; Raz et al., 2000; Hurtado et al., 2005; Heimer et al., 2006). Based on the results of the present study, we suggest that the high-frequency

52 Results II 648 J. Neurosci., January 16, (3): Rivlin-Etzion et al. Filter Properties of Basal Ganglia Cortical Loops oscillations in the BG do not directly drive the PD tremor. Rather, these BG oscillations should be considered as disrupting the normal motor processing of the MI (and motor brainstem centers) leading to the main core negative symptoms of PD, akinesia and bradykinesia. As in many other cases of motor dysfunction and weakness (Elble and Koller, 1990), PD tremor emerges, perhaps as part of the compensatory mechanisms of the nervous system. According to this hypothesis, therapeutic DBS procedures may stop the tremor because they block the abnormal neural activity of the BG. Nevertheless, we are aware that artificial, nonspecific electrical stimulation at tremor frequencies may not adequately mimic the signals generated by tremor cells that comprise only a subset of the population of BG neurons. In addition, the filter properties in the BG network of human PD patients after many years of disease evolution and therapy may differ from those observed in the vervet MPTP model within only a few days after the degeneration of dopaminergic neurons (Langston et al., 1984; Di Paolo et al., 1986; Hara et al., 1987; Sundstrom et al., 1988). Clinical studies define two extreme subtypes of PD: the akinetic-rigid and tremor-dominant (Jankovic et al., 1990). Similarly, MPTP-treated macaques develop only short episodes of high-frequency action/postural tremor (Burns et al., 1983) beyond akinesia and rigidity, whereas vervets commonly develop prolonged episodes of low-frequency tremor (Raz et al., 2000; Heimer et al., 2006). Previous studies of the pathological correlates of these clinical subtypes have failed to reach a consensus (Paulus and Jellinger, 1991; Hirsch et al., 1992). Our finding of low-pass filter properties of MI muscle pathways, and the working hypothesis of a PD tremor generation downstream to the BG network, indicates that future research on the pathological correlates of the clinical subtypes of PD should be directed toward the brainstem and spinal cord motor systems. References Alberts WW (1972) A simple view of parkinsonian tremor, electrical stimulation of cortex adjacent to the rolandic fissure in awake man. Brain Res 44: Albin RL, Young AB, Penney JB (1989) The functional anatomy of basal ganglia disorders. Trends Neurosci 12: Alexander GE, DeLong MR (1985) Microstimulation of the primate neostriatum. I. Physiological properties of striatal microexcitable zones. J Neurophysiol 53: Alexander GE, DeLong MR, Strick PL (1986) Parallel organization of functionally segregated circuits linking basal ganglia and cortex. Annu Rev Neurosci 9: Asanuma H, Rosen I (1972) Topographical organization of cortical efferent zones projecting to distal forelimb muscles in the monkey. Exp Brain Res 14: Baker SN, Olivier E, Lemon RN (1997) Coherent oscillations in monkey motor cortex and hand muscle EMG show task-dependent modulation. J Physiol (Lond) 501: Bar-Gad I, Morris G, Bergman H (2003) Information processing, dimensionality reduction and reinforcement learning in the basal ganglia. Prog Neurobiol 71: Bar-Gad I, Elias S, Vaadia E, Bergman H (2004) Complex locking rather than complete cessation of neuronal activity in the globus pallidus of a 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine-treated primate in response to pallidal microstimulation. J Neurosci 24: Baratta RV, Solomonow M, Zhou BH (1998) Frequency domain-based models of skeletal muscle. J Electromyogr Kinesiol 8: Ben Shaul Y, Stark E, Asher I, Drori R, Nadasdy Z, Abeles M (2003) Dynamical organization of directional tuning in the primate premotor and primary motor cortex. J Neurophysiol 89: Benabid AL (2003) Deep brain stimulation for Parkinson s disease. Curr Opin Neurobiol 13: Benabid AL, Deuschl G, Lang AE, Lyons KE, Rezai AR (2006) Deep brain stimulation for Parkinson s disease. Mov Disord 21 [Suppl 14]:S168 S Benazzouz A, Boraud T, Dubedat P, Boireau A, Stutzmann JM, Gross C (1995) Riluzole prevents MPTP-induced parkinsonism in the rhesus monkey: a pilot study. Eur J Pharmacol 284: Bergman H, Wichmann T, DeLong MR (1990) Reversal of experimental parkinsonism by lesions of the subthalamic nucleus. Science 249: Bergman H, Wichmann T, Karmon B, DeLong MR (1994) The primate subthalamic nucleus. II. Neuronal activity in the MPTP model of Parkinsonism. J Neurophysiol 72: Boraud T, Bezard E, Guehl D, Bioulac B, Gross C (1998) Effects of L-DOPA on neuronal activity of the globus pallidus externalis (GPe) and globus pallidus internalis (GPi) in the MPTP-treated monkey. Brain Res 787: Brown P (2006) Bad oscillations in Parkinson s disease. J Neural Transm Suppl Burns RS, Chiueh CC, Markey SP, Ebert MH, Jacobowitz DM, Kopin IJ (1983) A primate model of parkinsonism: selective destruction of dopaminergic neurons in the pars compacta of the substantia nigra by N-methyl-4-phenyl-1,2,3,6-tetrahydropyridine. Proc Natl Acad Sci USA 80: Chen CC, Litvak V, Gilbertson T, Kuhn A, Lu CS, Lee ST, Tsai CH, Tisch S, Limousin P, Hariz M, Brown P (2007) Excessive synchronization of basal ganglia neurons at 20 Hz slows movement in Parkinson s disease. Exp Neurol 205: Cheney PD, Fetz EE (1985) Comparable patterns of muscle facilitation evoked by individual corticomotoneuronal (CM) cells and by single intracortical microstimuli in primates: evidence for functional groups of CM cells. J Neurophysiol 53: Contreras CM, Mexicano G, Guzman-Flores CA (1981) Stereotaxic brain atlas of the green monkey (Cercopithecus aethiops aethiops). Bol Est Med Biol 31: DeLong MR (1971) Activity of pallidal neurons during movement. J Neurophysiol 34: DeLong MR (1990) Primate models of movement disorders of basal ganglia origin. Trends Neurosci 13: Di Paolo T, Bedard P, Daigle M, Boucher R (1986) Long-term effects of MPTP on central and peripheral catecholamine and indoleamine concentrations in monkeys. Brain Res 379: Doudet DJ, Gross C, Arluison M, Bioulac B (1990) Modifications of precentral cortex discharge and EMG activity in monkeys with MPTP-induced lesions of DA nigral neurons. Exp Brain Res 80: Elble RJ, Koller WC (1990) Tremor. Baltimore: The Johns Hopkins UP. Elias S, Joshua M, Goldberg JA, Heimer G, Arkadir D, Morris G, Bergman H (2007) Statistical properties of pauses of the high-frequency discharge neurons in the external segment of the globus pallidus. J Neurosci 27: Evarts EV (1966) Pyramidal tract activity associated with a conditioned hand movement in the monkey. J Neurophysiol 29: Filali M, Hutchison WD, Palter VN, Lozano AM, Dostrovsky JO (2004) Stimulation-induced inhibition of neuronal firing in human subthalamic nucleus. Exp Brain Res 156: Filion M, Tremblay L, Bedard PJ (1988) Abnormal influences of passive limb movement on the activity of globus pallidus neurons in parkinsonian monkeys. Brain Res 444: Freund HJ, Hefter H (1993) The role of basal ganglia in rhythmic movement. Adv Neurol 60: Garcia L, D Alessandro G, Bioulac B, Hammond C (2005) High-frequency stimulation in Parkinson s disease: more or less? Trends Neurosci 28: Georgopoulos AP, Schwartz AB, Kettner RE (1986) Neuronal population coding of movement direction. Science 233: Goldberg JA, Boraud T, Maraton S, Haber SN, Vaadia E, Bergman H (2002) Enhanced synchrony among primary motor cortex neurons in the 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine primate model of Parkinson s disease. J Neurosci 22: Graziano MS, Taylor CS, Moore T (2002) Complex movements evoked by microstimulation of precentral cortex. Neuron 34: Grosse P, Cassidy MJ, Brown P (2002) EEG-EMG, MEG-EMG and EMG- EMG frequency analysis: physiological principles and clinical applications. Clin Neurophysiol 113: Hammond C, Bergman H, Brown P (2007) Pathological synchronization in Parkinson s disease: networks, models and treatments. Trends Neurosci 30:

53 Results II Rivlin-Etzion et al. Filter Properties of Basal Ganglia Cortical Loops J. Neurosci., January 16, (3): Hara K, Tohyama I, Kimura H, Fukuda H, Nakamura S, Kameyama M (1987) Reversible serotoninergic neurotoxicity of N-methyl-4-phenyl- 1,2, 3,6-tetrahydropyridine (MPTP) in mouse striatum studied by neurochemical and immunohistochemical approaches. Brain Res 410: Hashimoto T, Elder CM, Vitek JL (2002) A template subtraction method for stimulus artifact removal in high-frequency deep brain stimulation. J Neurosci Methods 113: Hashimoto T, Elder CM, Okun MS, Patrick SK, Vitek JL (2003) Stimulation of the subthalamic nucleus changes the firing pattern of pallidal neurons. J Neurosci 23: Heimer G, Bar-Gad I, Goldberg JA, Bergman H (2002) Dopamine replacement therapy reverses abnormal synchronization of pallidal neurons in the 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine primate model of Parkinsonism. J Neurosci 22: Heimer G, Rivlin-Etzion M, Bar-Gad I, Goldberg JA, Haber SN, Bergman H (2006) Dopamine replacement therapy does not restore the full spectrum of normal pallidal activity in the 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine primate model of Parkinsonism. J Neurosci 26: Hirsch EC, Mouatt A, Faucheux B, Bonnet AM, Javoy Agid F, Graybiel AM, Agid Y (1992) Dopamine, tremor, and Parkinson s disease [letter]. Lancet 340: Horak FB, Anderson ME (1984) Influence of globus pallidus on arm movements in monkeys. II. Effects of stimulation. J Neurophysiol 52: Hurtado JM, Gray CM, Tamas LB, Sigvardt KA (1999) Dynamics of tremorrelated oscillations in the human globus pallidus: a single case study. Proc Natl Acad Sci USA 96: Hurtado JM, Rubchinsky LL, Sigvardt KA, Wheelock VL, Pappas CT (2005) Temporal evolution of oscillations and synchrony in GPi/muscle pairs in Parkinson s disease. J Neurophysiol 93: Jankovic J, McDermott M, Carter J, Gauthier S, Goetz C, Golbe L, Huber S, Koller W, Olanow C, Shoulson I (1990) Variable expression of Parkinson s disease: a base-line analysis of the DATATOP cohort. The Parkinson Study Group. Neurology 40: Joel D, Weiner I (1994) The organization of the basal gangliathalamocortical circuits: open interconnected rather than closed segregated. Neuroscience 63: Joshua M, Elias S, Levine O, Bergman H (2007) Quantifying the isolation quality of extracellularly recorded action potentials. J Neurosci Methods 163: Kilner JM, Baker SN, Salenius S, Jousmaki V, Hari R, Lemon RN (1999) Task-dependent modulation of Hz coherence between rectified EMGs from human hand and forearm muscles. J Physiol (Lond) 516: Kita H (1992) Responses of globus pallidus neurons to cortical stimulation: intracellular study in the rat. Brain Res 589: Langston JW, Irwin I, Langston EB (1984) A comparison of the acute and chronic effects of 1-methyl-4-phenyl-1,2,5,6-tetrahydropyridine (MPTP)-induced parkinsonism in humans and the squirrel monkey. Neurology 34 [Suppl 1]:268. Lathi BP (2005) Linear systems and signals. New York: Oxford UP. Leblois A, Boraud T, Meissner W, Bergman H, Hansel D (2006) Competition between feedback loops underlies normal and pathological dynamics in the basal ganglia. J Neurosci 26: Lemstra AW, Verhagen ML, Lee JI, Dougherty PM, Lenz FA (1999) Tremor-frequency (3 6 Hz) activity in the sensorimotor arm representation of the internal segment of the globus pallidus in patients with Parkinson s disease. Neurosci Lett 267: Maltete D, Jodoin N, Karachi C, Houeto JL, Navarro S, Cornu P, Agid Y, Welter ML (2007) Subthalamic stimulation and neuronal activity in the substantia nigra in Parkinson s disease. J Neurophysiol 97: Martin RF, Bowden DM (2000) Primate brain maps: structure of the macaque brain. Amsterdam: Elsevier Science. Maurice N, Deniau JM, Glowinski J, Thierry AM (1999) Relationships between the prefrontal cortex and the basal ganglia in the rat: physiology of the cortico-nigral circuits. J Neurosci 19: McHaffie JG, Stanford TR, Stein BE, Coizet V, Redgrave P (2005) Subcortical loops through the basal ganglia. Trends Neurosci 28: McIntyre CC, Grill WM (2002) Extracellular stimulation of central neurons: influence of stimulus waveform and frequency on neuronal output. J Neurophysiol 88: Mena-Segovia J, Bolam JP, Magill PJ (2004) Pedunculopontine nucleus and basal ganglia: distant relatives or part of the same family? Trends Neurosci 27: Nambu A, Tokuno H, Hamada I, Kita H, Imanishi M, Akazawa T, Ikeuchi Y, Hasegawa N (2000) Excitatory cortical inputs to pallidal neurons via the subthalamic nucleus in the monkey. J Neurophysiol 84: Nini A, Feingold A, Slovin H, Bergman H (1995) Neurons in the globus pallidus do not show correlated activity in the normal monkey, but phaselocked oscillations appear in the MPTP model of Parkinsonism. J Neurophysiol 74: Palmer SS, Fetz EE (1985) Effects of single intracortical microstimuli in motor cortex on activity of identified forearm motor units in behaving monkeys. J Neurophysiol 54: Paulus W, Jellinger K (1991) The neuropathologic basis of different clinical subgroups of Parkinson s disease. J Neuropathol Exp Neurol 50: Ranck JB (1975) Which elements are excited in electrical stimulation of mammalian central nervous system: a review. Brain Res 98: Rav-Acha M, Sagiv N, Segev I, Bergman H, Yarom Y (2005) Dynamic and spatial features of the inhibitory pallidal GABAergic synapses. Neuroscience 135: Raz A, Feingold A, Zelanskaya V, Vaadia E, Bergman H (1996) Neuronal synchronization of tonically active neurons in the striatum of normal and parkinsonian primates. J Neurophysiol 76: Raz A, Vaadia E, Bergman H (2000) Firing patterns and correlations of spontaneous discharge of pallidal neurons in the normal and the tremulous 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine vervet model of Parkinsonism. J Neurosci 20: Raz A, Frechter-Mazar V, Feingold A, Abeles M, Vaadia E, Bergman H (2001) Activity of pallidal and striatal tonically active neurons is correlated in mptp-treated monkeys but not in normal monkeys. J Neurosci 21:RC128(1 5). Rivlin-Etzion M, Marmor O, Heimer G, Raz A, Nini A, Bergman H (2006) Basal ganglia oscillations and pathophysiology of movement disorders. Curr Opin Neurobiol 16: Ryan LJ, Clark KB (1991) The role of the subthalamic nucleus in the response of globus pallidus neurons to stimulation of the prelimbic and agranular frontal cortices in rats. Exp Brain Res 86: Salenius S, Hari R (2003) Synchronous cortical oscillatory activity during motor action. Curr Opin Neurobiol 13: Shink E, Sidibe M, Smith Y (1997) Efferent connections of the internal globus pallidus in the squirrel monkey. II. Topography and synaptic organization of pallidal efferents to the pedunculopontine nucleus. J Comp Neurol 382: Song DD, Haber SN (2000) Striatal responses to partial dopaminergic lesion: evidence for compensatory sprouting. J Neurosci 20: Strafella AP, Ko JH, Grant J, Fraraccio M, Monchi O (2005) Corticostriatal functional interactions in Parkinson s disease: a rtms/[11c]raclopride PET study. Eur J Neurosci 22: Sundstrom E, Luthman J, Goldstein M, Jonsson G (1988) Time course of MPTP-induced degeneration of the nigrostriatal dopamine system in C57BL/6 mice. Brain Res Bull 21: Timmermann L, Gross J, Dirks M, Volkmann J, Freund HJ, Schnitzler A (2003) The cerebral oscillatory network of parkinsonian resting tremor. Brain 126: Topka H, Mescheriakov S, Boose A, Kuntz R, Hertrich I, Seydel L, Dichgans J, Rothwell J (1999) A cerebellar-like terminal and postural tremor induced in normal man by transcranial magnetic stimulation. Brain 122: Tremblay L, Filion M, Bedard PJ (1989) Responses of pallidal neurons to striatal stimulation in monkeys with MPTP-induced parkinsonism. Brain Res 498: Weinberger M, Mahant N, Hutchison WD, Lozano AM, Moro E, Hodaie M, Lang AE, Dostrovsky JO (2006) Beta oscillatory activity in the subthalamic nucleus and its relation to dopaminergic response in Parkinson s disease. J Neurophysiol 96: Wichmann T (2000) A digital averaging method for removal of stimulus artifacts in neurophysiologic experiments. J Neurosci Methods 98: Wu YR, Levy R, Ashby P, Tasker RR, Dostrovsky JO (2001) Does stimulation of the GPi control dyskinesia by activating inhibitory axons? Mov Disord 16:

54 Results III III. BASAL GANGLIA OSCILLATORY ACTIVITY AND PATHOPHYSIOLOGY OF PARKINSON'S DISEASE Article information: Rivlin-Etzion M, Marmor O, Heimer G, Raz A, Nini A, Bergman H (2006). Basal ganglia oscillations and pathophysiology of movement disorders. Current Opinion in Neurobiology 16:

55 Results III Basal ganglia oscillations and pathophysiology of movement disorders Michal Rivlin-Etzion 1,2, Odeya Marmor 1, Gali Heimer 1,4, Aeyal Raz 1,5, Asaph Nini 1,6 and Hagai Bergman 1,2,3 Low frequency rest tremor is one of the cardinal signs of Parkinson s disease and some of its animal models. Current physiological studies and models of the basal ganglia differ as to which aspects of neuronal activity are crucial to the pathophysiology of Parkinson s disease. There is evidence that neural oscillations and synchronization play a central role in the generation of the disease. However, parkinsonian tremor is not strictly correlated with the synchronous oscillations in the basal ganglia networks. Rather, abnormal basal ganglia output enforces abnormal thalamo-cortical processing leading to akinesia, the main negative symptom of Parkinson s disease. Parkinsonian tremor has probably evolved as a downstream compensatory mechanism. Addresses 1 Department of Physiology 2 The Interdisciplinary Center for Neural Computation 3 The Eric Roland Center for Neurodegenerative Diseases, The Hebrew University-Hadassah Medical School, Jerusalem, Israel, Department of Pediatrics, Hadassah Hebrew University Medical Center, Jerusalem, Israel, Department of Anesthesia, Rabin Medical Center-Beilinson Campus, Petach-Tikva, Israel 6 Department of intensive care, Division of Anesthesiology, Shiba Medical center, Tel-Hashomer, Israel Corresponding author: Bergman, Hagai (hagaibe@ekmd.huji.ac.il) Current Opinion in Neurobiology 2006, 16: This review comes from a themed issue on Motor systems Edited by Eve Marder and Peter L Strick Available online 3rd November /$ see front matter # 2006 Elsevier Ltd. All rights reserved. DOI /j.conb Introduction: Parkinson s disease clinical symptoms and pathology In 1817, almost two hundred years ago, the English physician James Parkinson wrote Essay on the Shaking Palsy, providing the first clinical description of the motor symptoms of the disease now bearing his name [1]. Today, Parkinson s disease (PD) is the most common basal ganglia movement disorder, and affects from 1% of those aged 65 to 4 5% of the 85 year old population [2]. Only 5% of PD cases can be attributed to specific genetic causes [3,4]. Most of the remaining cases cannot be attributed to metabolic or toxic causes either, and are, therefore, classified as idiopathic PD. PD is the result of a neurodegenerative process that causes damage to multiple neuronal circuits. The dopaminergic system is the most seriously damaged, but the noradrenergic, serotonergic and cholinergic systems are also affected [5]. On the basis of clinical observations of six patients (including two whom he met on the street and a third he observed at a distance), Parkinson described two of the most important and paradoxically related symptoms of PD: shaking now defined as a low frequency (4 7 Hz, but higher frequencies, up to 9 Hz, are encountered at early disease stages) tremor at rest (tremor amplitudes decrease during voluntary action and increase during mental stress), a hyperkinetic disorder; and palsy (or akinesia in modern terminology) characterized by a poverty of voluntary and especially involuntary movements, a hypokinetic disorder. The other cardinal motor symptoms of PD include bradykinesia (slowness of voluntary movements), rigidity (increased muscular tonus), and postural abnormalities. Cognitive and mood (emotional) deficits frequently accompany the motor symptoms. However, in this review we focus on the pathophysiology of the two main motor symptoms of PD as outlined by Parkinson: akinesia and tremor at rest. Note that we consider bradykinesia and related hypokinetic PD clinical features as akinetic symptoms. PD rigidity is characterized by a uniform resistance to passive movements owing to increased muscle response to passive stretch, and is not associated with changes of spinal alpha motor neuron excitability. Thus, and based on the clinical similarities of akinesia and rigidity as outlined below, we associate akinesia and rigidity. Finally, we limit our discussion to tremor at rest, although other non-harmonically related forms (e.g. postural and/or kinetic tremor) are very common in PD [6,7]. Akinesia versus tremor in Parkinson s disease PD is not a homogenous disease, either across patients or even within a single patient s disease course. Temporally, tremor is not a consistent feature of the disease, but rather is episodic, as opposed to akinesia. Unlike rigidity and akinesia, there is no correlation between the clinical severity of PD tremor and the severity of the dopaminergic deficit in the striatum or the clinical progression of the disease [7]. Human PD covers a broad spectrum of symptoms and can present as a predominant resting tremor (T-subtype) or Current Opinion in Neurobiology 2006, 16:

56 Results III 630 Motor Systems Glossary Direct indirect rate model of the basal ganglia: The circuitry of the basal ganglia is often divided into two major pathways, the direct pathway and the indirect pathway. The direct pathway directly connects the striatum to the GPi and SNr by GABAergic (inhibitory) projections. The indirect pathway connects the striatum to the GPi and SNr through the GPe and the STN, with net excitatory effects. Disinhibition: Removal of neuronal inhibition by inhibition. For example, cells in the striatum can inhibit neurons of the GPi and SNr, which in turn removes their tonic inhibition from the thalamus. Essential tremor: The most common movement disorder (10 20 times more prevalent than PD), characterized by a slowly progressive postural and/or kinetic tremor with no known cause. Negative symptoms: Normal behaviors or body states that are absent or diminished in a person with a mental or neurological disorder (e.g. akinesia of PD). Positive symptoms: These are the opposite of negative symptoms and refer to behaviors or body states that are practically absent in people in the general population but are present or enhanced in persons with the neurological disorder (e.g. PD tremor). Spectral harmonics: Spectral harmonics are other spectral peaks at frequencies equal to integer multiples of the fundamental frequency, usually as a result of distortions to the pure sinus generating the fundamental frequency. Spectral sidebands: In spectral analysis, a sideband is a band of frequencies higher or lower than the fundamental frequency, usually containing energy as a result of the amplitude modulation process. primarily as marked akinesia and rigidity (AR-subtype) [8], sometimes defined as the postural instability gait difficulty subtype. As early as 1877, the great French neurologist Jean-Martin Charcot noted that tremor is not always present in human PD patients, and, therefore, suggested changing the name of the disease from paralysis agitans (Latin for shaking palsy) to la maladie de Parkinson (Parkinson s disease). T-subtype PD patients have a better prognosis and slower disease progression than AR-subtype patients [8]. Interestingly, most patients with non-idiopathic PD display akinesia and rigidity but not rest tremor [9]. Anti-cholinergic agents, which were the first drugs available for the symptomatic treatment of PD, tend to have better effects on tremor than on akinetic-rigid symptoms, whereas akinesia might show better and earlier response to dopamine replacement therapy [10]. Several studies have indicated that the pathology of human T-type PD differs from that of the AR-type PD, with the retrorubral area (A8) more severely affected in the tremor-dominant form [11]. The frequency of tremor in a given PD patient is often remarkably similar in different muscles of the extremities and trunk [12]. These observations led to the assumption that a common single central oscillator controls all tremulous muscles. Coherence analysis, however, has shown that although the muscles within one body part (e.g. a limb) are mostly coherent, the tremor in different extremities, even on the same body side, is almost never coherent [13,14], indicating that different oscillators underlie parkinsonian tremor in the different extremities. This absence of tremor coherence could hint at mechanical or spinal reflex mechanisms rather than a single central oscillator. Nevertheless, several studies have failed to demonstrate any frequency reduction of the tremor as a result of load addition to the trembling limb in PD patients [6,7]. Resetting experiments, in which the tremulous limb is reset by mechanical perturbation, have been less conclusive. Initial studies indicated that resetting of tremor is much more easily achieved in essential tremor (see Glossary) than in PD tremor. However, more recent studies have shown that the resetting index varies significantly with the magnitude of the mechanical perturbation and with the tremor amplitude. When these factors were equalized, however, no significant difference was found in mean resetting indexes among PD tremor, essential tremor and normal subjects mimicking tremor. Resetting experiments with electrical stimulation of the median nerve or transcranial magnetic stimulation of the motor cortex did not show consistent resetting of the tremor rhythm when the periphery (median nerve) was stimulated, but did result in complete resetting when the cortex was stimulated [7]. In line with the central nervous system (CNS) hypothesis on the origin of PD tremor, it has long been known that different lesions within the CNS can suppress parkinsonian tremor. Early attempts to remove parts of the motor cortex or its downstream projections were successful in suppressing tremor but produced unacceptable side effects. The cerebellar receiving nuclei of the thalamus (e.g. the ventralis-intermedius, Vim) have traditionally been considered the optimal target for stereotaxic procedures for amelioration of PD and other tremors. Recently, it has been demonstrated that chronic highfrequency stimulation of these same thalamic targets, in addition to subthalamic or pallidal stimulations, are all able to efficiently suppress parkinsonian tremor and other motor symptoms [15 ]. In summary, most clinical human studies indicate that PD tremor and akinesia, although they share common origins and similarities, have significantly distinct characteristics. The role of striatal dopamine depletion and the central generators seem to be much more important in akinesia. PD tremor might be modulated by peripheral manipulation and by the activity of other central neuronal systems. It is possible that transmitter systems other than dopamine (e.g. cholinergic, serotonergic), or neural circuits other than the basal ganglia (e.g. cerebellum [16], red nucleus), play a crucial additive role in underlying this symptom. Parkinson s disease animal models Early animal models of PD were based on lesions of midbrain areas in monkeys. These anatomical lesions mainly produce rigidity and only rarely result in a spontaneous, sustained tremor. Careful analysis of the correlation between the clinical symptoms and the extent of the lesion led to the conclusion that experimental rest tremor Current Opinion in Neurobiology 2006, 16:

57 Results III Basal ganglia oscillations and pathophysiology of movement disorders Rivlin-Etzion, Marmor, Heimer, Raz, Nini and Bergman 631 is the result of damage to both the nigro-striatal dopaminergic projections and the cerebellar outflow (to the red nucleus and thalamus). Damage to only one of these neuronal systems was not sufficient for reliable generation of tremor [17]. More modern animal models of PD have shifted from anatomical to chemical lesions. Early chemical animal models of PD for example, the 6-hydroxydopamine (6- OHDA) model were limited to dopaminergic damage, and mainly reproduced the main negative symptoms of PD; namely, akinesia (see Glossary) [18]. The more recently introduced primate 1-methyl-4-phenyl-1,2,3,6- tetrahydropyridine (MPTP) model of PD [19] better mimics the clinical and the pathological picture of PD. Post-mortem examination of the brains of MPTP-treated primates reveals that the primary damage is to the dopaminergic system. However, as in human PD, other neuromodulators are also affected [20]. Monkeys treated with MPTP mainly exhibit the akinetic rigid symptoms of PD [19]. Low frequency (4 7 Hz) resting tremor is not readily replicated in MPTP-treated macaque monkeys, but other species, notably the vervet (African green) monkey, often develop a prominent lowfrequency tremor following MPTP injections [21,22]. It is important to note that the tremor usually appears several days after the development of clinical akinesia and rigidity [21,23 ]. This reversed order of presentation of clinical symptoms compared with that of the human disease could be due to the fast induction of dopamine depletion in the MPTP model that might impede the development of compensatory processes found in the slow-evolving human disease. Yet, tremor is a much more overt phenomenon than akinesia and rigidity. A human patient or his/her family might first be made aware of the slow development of PD by the more easily recognizable tremor. As in human studies, there is a low coherence between the tremors of the limbs of MPTP-treated vervet monkeys following dopamine replacement therapy [23 ]. Basal ganglia anatomy The cumulative clinical and experimental evidence outlined above strongly indicates that the major pathological event leading to the motor symptoms of PD, and especially to akinesia, is the death of midbrain dopaminergic neurons and their striatal projections. The striatum (composed of caudate, putamen and ventral striatum) is the main input stage of the basal ganglia, receiving inputs from all cortical areas, from many thalamic nuclei and even from the cerebellum [24 ]. Therefore, a good grasp of the pathophysiology of PD depends on understanding the anatomy and physiology of the basal ganglia and dopamine networks. The realization, at the turn of the 20 th century, that lesions involving the basal ganglia often result in severe disorders of motor function explains why the basal ganglia were classified as part of the extra-pyramidal system. The pyramidal system starts at the motor cortices, and through the brainstem pyramids projects to a-motoneurons, innervating the distal parts of the limb, and controlling the execution of accurate and voluntary movements. By contrast, it was assumed that the extra-pyramidal system originated at the basal ganglia and the cerebellum, descended parallel to the pyramidal system, and innervated the spinal circuits involved with more axial (postural), automatic non-voluntary movements. The revolution in anatomical methods during the second half of the 20 th century led researchers to the conclusion that the basal ganglia are part of a closed loop connecting all cortical areas sequentially through the striatum, pallidum and thalamus with the frontal cortex (Figure 1). The frontal cortex projects downstream to the spinal level. The direct projection of the basal ganglia to upper brainstem nuclei (e.g. superior colliculus and peduncolopontine nucleus [16]) will be considered here, for the sake of simplicity, as part of this descending system. The new view of the basal ganglia networks assumes that there are two segregated internal pathways that start in the striatum and converge on the output structures of the basal ganglia (the internal segment of the globus pallidus [GPi] and the substantia nigra pars reticulata [SNr]). The direct pathway is a direct GABAergic inhibitory pathway, whereas the indirect pathway is a polysynaptic disinhibitory pathway (see Glossary) through the external segment of the globus pallidus (GPe) and the subthalamic nucleus (STN). The projection striatal neurons in the direct pathway express D1 dopamine receptors, whereas those in the indirect pathway express D2 dopamine receptors [25]. Midbrain dopamine (DA) has differential effects on the two striato-pallidal pathways: it facilitates transmission along the direct pathway through the D1 receptors and inhibits transmission along the indirect pathway through the D2 receptors [25,26]. Note that in this schematic description the crucial roles of both the cholinergic and the dopaminergic innervation of the striatum on Figure 1 Schematic view of the recurrent connectivity among the cortex, basal ganglia and muscle networks. Current Opinion in Neurobiology 2006, 16:

58 Results III 632 Motor Systems plasticity and learning in the cortico-striatal synapse have been neglected [27,28 ]. Recently, single axon tracing anatomical studies have revealed an even more complex map of basal ganglia connectivity. Striatal neurons projecting to GPi and SNr send collaterals to GPe [29,30]. The physiological evidence for the importance of the direct projections from the motor cortex to the STN (the hyper-direct pathway [31]) indicate that like the striatum, the STN is an input stage of the basal ganglia [32]. Moreover, the recently described feedback projections from the GPe to the striatum [33,34], in addition to the GPe to GPi projection, strongly suggest that the GPe is a central nucleus in the basal ganglia circuitry, rather than a simple relay station in the indirect pathway. Figure 2 summarizes the current view of the complex connectivity among the basal ganglia nuclei. Figure 2 Physiological studies of the basal ganglia in normal primates Single unit recording and analysis of neuronal activity at the level of single spikes of single cells are probably the main ways to study a neuronal network. The different nuclei of the basal ganglia have a diverse background (during a quiet, awake state) spiking activity. The striatal neurons are characterized by a low frequency discharge rate (<1 spikes/s by the projection neurons and 4 10 spikes/s by the tonically active neurons [TANs], the cholinergic interneurons). This slow discharge is striking in contrast to the high (50 80 spikes/s) frequency discharge of the pallidal and SNr neurons. In all these structures the firing rate is irregular (Poisson-like), and neuronal oscillations are seldom observed in normal awake subjects. Studies exploring the relationship between spiking activity of basal ganglia neurons and body movements have revealed even more unexpected results. The akinesia associated with PD suggests that the basal ganglia play a crucial role in movement initiation. Nevertheless, most basal ganglia neurons change their firing rate after initiation of movements (Putamen: [35,36], GP: [37 39], SNr: [40]), and do not have any exclusive or consistent relationship to movement parameters such as start and/or end, velocity or amplitude [41]. Taken together with the minor impairment of motor control following focal inactivation and lesions of the output structures of the basal ganglia [42 44], the physiological results lead to the surprising conclusion that the basal ganglia do not initiate movements [42,45]. Physiological studies in the dopaminedepleted basal ganglia networks Early physiological studies of parkinsonian MPTP-treated monkeys reported changes in the discharge rate within the GPe, GPi [46,47] and STN [21]. Reversed trends of pallidal discharge rates in response to dopamine replacement therapy have been reported in both human patients [48,49] and primates [23,50,51]. The crucial role of these rate changes in the pathophysiology of PD has been verified by the subsequent findings showing that inactivation of STN and GPi could improve the motor symptoms in parkinsonian animals [52] and in human patients [15 ]. Detailed linear view of the connectivity of the cortex, basal ganglia and muscle networks. Color coding as in Figure 1. The basal ganglia inter-connectivity is shown in detail, with gray arrows representing excitatory glutamatergic projections and black round-head arrows for inhibitory GABAergic projections. Abbreviations: GPe, external segment of the globus pallidus; GPi, internal segment of the globus pallidus; STN, subthalamic nucleus. These findings contributed to the formulation and the popularity of the direct indirect model of the basal ganglia (see Glossary) [26]. Nevertheless, several studies have failed to find the expected significant changes of firing rates in the pallidum [53], thalamus [54] or motor cortical areas [55] of MPTP monkeys. This and other inconsistencies with the assumptions and the predictions of the direct indirect rate model have attracted more attention to the potential roles of other aspects of neuronal activity, such as firing patterns and neuronal synchronization, in Current Opinion in Neurobiology 2006, 16:

59 Results III Basal ganglia oscillations and pathophysiology of movement disorders Rivlin-Etzion, Marmor, Heimer, Raz, Nini and Bergman 633 the pathophysiology of PD. MPTP monkeys show an increase in the fraction of basal ganglia neurons that discharge in bursts. These bursts are either irregular or oscillatory and have been found in STN, GPe, GPi and primary motor cortex [21,22,46,47,55,56,57 ]. In most cases, the cells tend to oscillate at the tremor frequency or at double or even triple the tremor frequency [21]. Nevertheless, these studies repeatedly failed to reveal neurons with oscillations that are consistently coherent with the tremor [22,23 ]. Both STN inactivation [58] and dopamine replacement therapy [23 ] ameliorate the 4 7 Hz tremor and reduce the GPi 8 20 Hz oscillations, supporting the crucial role of double rather than the tremor frequency oscillations in tremor generation. Physiological studies and cross-correlation analysis of the activity of simultaneously recorded neurons in the basal ganglia and cortex of monkeys before and after MPTP treatment enable the assessment of the changes in neural synchronization in these areas. These studies have been conducted in the pallidum [22], as well as in the primary motor cortex [55], among striatal TANs and between TANs and pallidal neurons [59,60]. Cross correlation functions become peaked and oscillatory following MPTP treatments, suggesting that striatal dopamine depletion induces abnormal coupling of basal ganglia loops. The abnormal pallidal synchronization decreases in response to dopamine replacement therapy [23 ]. In most cases, the maximal power of the synchronous oscillations was at double the tremor frequency [22,23,59,60]. As in the MPTP primate, single unit studies of the basal ganglia of human PD patients (performed during electrophysiological mapping of the target area for therapeutic implantation of stimulating electrodes) report a high fraction of GPi cells oscillating at the tremor frequency [61]. However, as in the primate, the human studies [62,63 ] show that these oscillations are not fully coherent with the simultaneous recorded tremor. The sharp contrast between this transient, inconsistent pallidal-tremor synchronization and the high synchronicity found between thalamic Vim neurons and the tremor [64] suggests that pallidal neurons cannot be viewed as the tremor generators, or as reflecting the proprioceptive feedback of the tremor. Physiological studies of population neural activity in the dopamine-depleted basal ganglia networks Synchronization of basal ganglia neuronal activity is also evident in the local field potentials (LFPs) recorded in the subthalamic region of PD patients by the macroelectrodes used for high-frequency stimulation of these structures. These oscillations occur mainly in the b range (15 30 Hz) and following treatment with levodopa shift to higher frequencies in the gamma range [65 ]. Another study found that a significant reduction of the b range oscillations preceded the onset of the rest tremor [66 ]. In line with both the single unit and the LFP studies, magnetoencephalographic (MEG) studies [67] of T-type PD patients have revealed a strong coherence between the tremor and the activity in the motor and sensory cortices and the cerebellum at tremor frequency, and an even stronger coherency at double tremor frequency. Spectra of coherence between thalamic activity and cerebellum as well as between other brain areas have revealed additional broad peaks around 20 Hz. Studies of LFPs recorded from frontal cortex and STN of rats following 6-OHDA lesions of midbrain dopamine neurons [68] have revealed significant increases in the power and coherence of b-frequency oscillatory activity. Administration of apomorphine, the dopamine receptor agonist, to these dopamine-depleted animals suppressed the b-frequency oscillations, and increased coherent activity at gamma frequencies in the cortex and STN. Thus, the pattern of synchronization between population activity in the STN and that in the cortex in the 6- OHDA-lesioned rodent model of PD is parallel to that seen in the parkinsonian human. Recordings of both LFPs and multi-neuronal activity from microelectrodes inserted into STN in PD patients during functional neurosurgery suggest that the discharges of some of the neurons in STN are locked to b oscillations in the LFP [69]. LFP probably represents the synaptic input to a neural structure and its subthreshold slow activity. The discrepancies between LFP oscillatory activity and neuronal activity (in their frequency domain, prevalence and power) are probably due to the fact that even quite strong synchronized inputs can lead to weak neuronal correlations. Alternatively, correlations can be very low at the single-unit level, but still sum and become substantial at the population (LFP) level [70,71]. On correlation, causality and harmonics in spectral analysis Basal ganglia researchers are aware that depletion of dopamine in the striatum (the input stage of the basal ganglia) is the major event leading to clinical symptoms, including tremor, of PD. It is, thus, tempting to assume a causal relationship between the neural oscillations that are found in the STN and the globus pallidus (output stage of the basal ganglia) and the tremor. However, the closed loop structure of the cortex basal ganglia muscle networks (Figure 1), including inputs from other structures (e.g. cerebellum), suggests that basal ganglia oscillations are the result of proprioceptive feedback to the basal ganglia. The high-level energy at double the tremor frequency in many single unit correlation studies [22,59] and MEG studies [67] might suggest a 2:1 filtering mechanism downstream of the basal ganglia output [72,73]. However, there are important limitations to spectral analysis Current Opinion in Neurobiology 2006, 16:

60 Results III 634 Motor Systems [66,74 76]. In particular, any non-linear transformation (e.g. the addition of a derivative or high-pass filtered version of the original signal) might add higher harmonics (see Glossary) to the original peak. The amplitude of these higher harmonics can be larger than that of the original peak (Figure 3). Amplitude fluctuation of the tremor or neuronal analog signal (such as LFP) could add sidebands (see Glossary) both above and below the central frequency peak. Moreover, if the amplitude and the frequency fluctuation follow a complex pattern, and enough spectra are averaged (as is the common practice), the additional peaks can be distributed out over a broad frequency domain [77]. Thus, although neuronal oscillations can be detected at several levels of the basal ganglia network subsequent to dopamine depletion and the emergence of clinical PD tremor, there is not enough evidence to support the notion that the tremor follows the basal ganglia oscillations. Conclusions and future directions In this review we have explored the possible relationships between basal ganglia oscillatory activity and PD tremor. PD is the result of dopamine depletion in the striatum the input stage of the basal ganglia. Akinesia and rest tremor are two major symptoms of PD. Nevertheless, cumulative clinical and experimental evidence support the view that akinesia and rest tremor are not generated by identical neuronal mechanisms. Following striatal dopamine depletion, many basal ganglia neurons develop synchronous oscillations at the tremor frequency and at their higher harmonics in addition to in the b range. However, the PD tremor does not strictly follow the basal ganglia Figure 3 Simulated analog signals and their power spectrums. The analog signal (left column) is composed of a pure sinus wave plus its derivative (cosine function with the same frequency). K1 and K2 are scaling factors of the relative weights of the sinus and its derivate (cosine) function. The scaling factors were chosen to normalize all peak-to-peak amplitudes to the same value. Current Opinion in Neurobiology 2006, 16:

61 Results III Basal ganglia oscillations and pathophysiology of movement disorders Rivlin-Etzion, Marmor, Heimer, Raz, Nini and Bergman 635 oscillatory activity. The recent demonstration of anatomical connections between the cerebellum and the basal ganglia, at the cortical, striatal and brainstem levels might suggest that the cerebellum is associated with the movement disorders classically described as basal ganglia disorders. The crucial role of cerebellar output in the generation of PD tremor has been demonstrated by lesion studies and the efficacy of Vim intervention in treatment of PD tremor. These findings, along with the physiological studies of the normal basal ganglia indicating that the basal ganglia do not initiate movements, strengthen the supposition that the abnormal synchronous oscillations in the basal ganglia provide noisy input to the frontal cortex, and hence lead to PD akinesia. In contrast to the akinesia, we suggest that PD tremor (Parkinson s shaking ) is the result of compensation mechanisms generated downstream of the basal ganglia in order to compensate for PD akinesia ( palsy ). This hypothesis is in accordance with the fact that T-subtype PD is slower to progress than AR-subtype PD, and with the late appearance of the tremor in the MPTP-treated monkeys. Moreover, the peripheral feedback from the tremulous body segments might interrupt the noisy activity of the basal ganglia cortical networks in a similar manner to subcortical DBS or lesions. Future studies of the complex neural network of the basal ganglia and their related neuronal structures will hopefully shed more light on their role and function in health and disease. Acknowledgements This research was supported in part by a Center of Excellence grant from the Israel Science Foundation (ISF) and the Fighting against Parkinson Foundation of the Hebrew University Netherlands Association (HUNA). We thank E Singer for critical reading and language editing. References and recommended reading Papers of particular interest, published within the annual period of review, have been highlighted as: of special interest of outstanding interest 1. Parkinson J: An essay on the shaking palsy. London: Sherwood, Neely and Jones; Van Den Eeden SK, Tanner CM, Bernstein AL, Fross RD, Leimpeter A, Bloch DA, Nelson LM: Incidence of Parkinson s disease: variation by age, gender, and race/ethnicity. Am J Epidemiol 2003, 157: Farrer MJ: Genetics of Parkinson disease: paradigm shifts and future prospects. Nat Rev Genet 2006, 7: An excellent review of the current view of the complex, multi-factorial heritable basis of PD. 4. Benmoyal-Segal L, Soreq H: Gene environment interactions in sporadic Parkinson s disease. J Neurochem 2006, 97: Jellinger KA: Pathology of Parkinson s disease. Changes other than the nigrostriatal pathway. Mol Chem Neuropathol 1991, 14: Elble RJ, Koller WC: Tremor. The Johns Hopkins University Press; Deuschl G, Raethjen J, Baron R, Lindemann M, Wilms H, Krack P: The pathophysiology of parkinsonian tremor: a review. J Neurol 2000, 247(Suppl 5):V33-V Jankovic J, McDermott M, Carter J, Gauthier S, Goetz C, Golbe L, Huber S, Koller W, Olanow C, Shoulson I et al.: Variable expression of Parkinson s disease: a base-line analysis of the DATATOP cohort. The Parkinson Study Group. Neurology 1990, 40: Rajput AH: Clinical features of tremor in extrapyramidal syndromes. In Handbook of Tremor Disorders. Edited by Findley LJ, Koller WC. Marcel Dekker, Inc.; 1995: Tolosa ES, Marin C: Medical management of parkinsonian tremor. In Handbook of Tremor Disorders. Edited by Findley LJ, Koller WC. Marcel Dekker, Inc.; 1995: Paulus W, Jellinger K: The neuropathologic basis of different clinical subgroups of Parkinson s disease. J Neuropathol Exp Neurol 1991, 50: Hunker CJ, Abbs JH: Uniform frequency of parkinsonian resting tremor in the lips, jaw, tongue, and index finger. Mov Disord 1990, 5: Raethjen J, Lindemann M, Schmaljohann H, Wenzelburger R, Pfister G, Deuschl G: Multiple oscillators are causing parkinsonian and essential tremor. Mov Disord 2000, 15: Ben-Pazi H, Bergman H, Goldber JA, Giladi N, Hansel D, Reches A, Simon ES: Synchrony of rest tremor in multiple limbs in Parkinson s disease: evidence for multiple oscillators. J Neural Transm 2001, 108: Machado A, Rezai AR, Kopell BH, Gross RE, Sharan AD, Benabid AL: Deep brain stimulation for Parkinson s disease: Surgical technique and perioperative management. Mov Disord 2006, 21:S247-S258. An updated description of surgical techniques and patient management during deep brain stimulation (DBS) surgery for advanced PD. Since 1993, there have been more than DBS implants at more than 500 medical centers worldwide. This is part of a comprehensive review composed of nine manuscripts written by international experts in the field, published as Supplement 14 to volume 21 of Movement Disorders. 16. Stein JF, Aziz TZ: Does imbalance between basal ganglia and cerebellar outputs cause movement disorders? Curr Opin Neurol 1999, 12: Jenner P, Marsden CD: Neurochemical basis of parkinsonian tremor. In Movement disorders: tremor. Edited by Findley LJ, Capildeo R. Oxford University Press; 1984: Wilms H, Sievers J, Deuschl G: Animal models of tremor. Mov Disord 1999, 14: Burns RS, Chiueh CC, Markey SP, Ebert MH, Jacobowitz DM, Kopin IJ: A primate model of parkinsonism: selective destruction of dopaminergic neurons in the pars compacta of the substantia nigra by N-methyl-4-phenyl-1,2,3,6- tetrahydropyridine. Proc Natl Acad Sci USA 1983, 80: Pifl C, Schingnitz G, Hornykiewicz O: Effect of 1-methyl-4- phenyl-1,2,3,6-tetrahydropyridine on the regional distribution of brain monoamines in the rhesus monkey. Neuroscience 1991, 44: Bergman H, Wichmann T, Karmon B, DeLong MR: The primate subthalamic nucleus. II. Neuronal activity in the MPTP model of parkinsonism. J Neurophysiol 1994, 72: Raz A, Vaadia E, Bergman H: Firing patterns and correlations of spontaneous discharge of pallidal neurons in the normal and the tremulous 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine vervet model of parkinsonism. J Neurosci 2000, 20: Heimer G, Rivlin-Etzion M, Bar-Gad I, Goldberg JA, Haber SN, Bergman H: Dopamine replacement therapy does not restore the full spectrum of normal pallidal activity in the 1-methyl-4- phenyl-1,2,3,6-tetra-hydropyridine primate model of parkinsonism. J Neurosci 2006, 26: A recent study of the tremor, synchronous oscillations and bursts in the GPe and GPi of normal and MPTP-treated rhesus and vervet monkeys. 24. Hoshi E, Tremblay L, Feger J, Carras PL, Strick PL: The cerebellum communicates with the basal ganglia. Nat Neurosci 2005, 8: Using the novel technique of transneuronal transport of the rabies virus, this manuscript describes a newly discovered communication pathway Current Opinion in Neurobiology 2006, 16:

62 Results III 636 Motor Systems from the cerebellum to the input stage of the basal ganglia. Previous studies by the same group revealed interactions between the cerebellum and the basal ganglia output at the level of the frontal cortex, and other studies have suggested such an interaction at the level of the brainstem. The novel findings of this study reveal that close association between the basal ganglia and the cerebellum also exists at the level of the input stage of the basal ganglia the striatum. 25. Gerfen CR, Engber TM, Mahan LC, Susel Z, Chase TN, Monsma FJ Jr, Sibley DR: D1 and D2 dopamine receptorregulated gene expression of striatonigral and striatopallidal neurons. Science 1990, 250: Albin RL, Young AB, Penney JB: The functional anatomy of basal ganglia disorders. Trends Neurosci 1989, 12: Wang Z, Kai L, Day M, Ronesi J, Yin HH, Ding J, Tkatch T, Lovinger DM, Surmeier DJ: Dopaminergic control of corticostriatal long-term synaptic depression in medium spiny neurons is mediated by cholinergic interneurons. Neuron 2006, 50: Many previous studies have shown that activation of dopamine receptors can induce long-term potentiation (LTP) or long-term depression (LTD) in the cortico-striatal synapses. However, the role of dopamine D1 and D2 receptors in cortico-striatal plasticity was considered incompatible with their differential distribution to the direct and indirect pathway striatal neurons. Wang et al. show that the D2 dependence of LTD is not mediated by dopamine receptors of the striatal projection medium-spiny neurons. Rather, this plasticity is achieved through dopamine-induced changes in the release of acetylcholine by the cholinergic interneurons of the striatum. Besides providing a coherent hypothesis for the role of D1 and D2 dopamine receptors in cortico-striatal plasticity, this paper highlights the crucial role played by the cholinergic interneurons in the physiology of the striatum. 28. Graybiel AM: The basal ganglia: learning new tricks and loving it. Curr Opin Neurobiol 2005, 15: An excellent review of the current models and views on the role of the basal ganglia in reinforcement learning. A short review of current models of the basal ganglia is followed by a series of queries facing this field. 29. Levesque M, Parent A: The striatofugal fiber system in primates: a reevaluation of its organization based on single-axon tracing studies. Proc Natl Acad Sci USA 2005, 102: Neurons located in either striosomes or the extrastriosomal matrix in squirrel monkeys were injected with biotin dextran amine, and their labeled axons were entirely reconstructed with a camera lucida. 24 of 27 reconstructed axons arborized into the three main striatal targets (GPe, GPi and SNr), in sharp contrast with the predictions of the direct indirect model of the basal ganglia circuitry. Future studies should reveal whether or how the key concept of the direct or indirect basal ganglia, originally described in rodent studies, should be incorporated with these new primate findings. 30. Nadjar A, Brotchie JM, Guigoni C, Li Q, Zhou SB, Wang GJ, Ravenscroft P, Georges F, Crossman AR, Bezard E: Phenotype of striatofugal medium spiny neurons in parkinsonian and dyskinetic nonhuman primates: a call for a reappraisal of the functional organization of the basal ganglia. J Neurosci 2006, 26: Nambu A: A new dynamic model of the cortico-basal ganglia loop. Prog Brain Res 2004, 143: Feger J, Bevan M, Crossman AR: The projections from the parafascicular thalamic nucleus to the subthalamic nucleus and the striatum arise from separate neuronal populations: a comparison with the corticostriatal and corticosubthalamic efferents in a retrograde fluorescent double- labelling study. Neuroscience 1994, 60: Bolam JP, Hanley JJ, Booth PA, Bevan MD: Synaptic organisation of the basal ganglia. J Anat 2000, 196: Koos T, Tepper JM: Inhibitory control of neostriatal projection neurons by GABAergic interneurons. Nat Neurosci 1999, 2: Crutcher MD, DeLong MR: Single cell studies of the primate putamen. II. Relations to direction of movement and pattern of muscular activity. Exp Brain Res 1984, 53: Alexander GE, Crutcher MD: Preparation for movement: neural representations of intended direction in three motor areas of the monkey. J Neurophysiol 1990, 64: DeLong MR: Activity of pallidal neurons during movement. J Neurophysiol 1971, 34: Anderson ME, Horak FB: Influence of the globus pallidus on arm movements in monkeys. III. Timing of movement-related information. J Neurophysiol 1985, 54: Mink JW, Thach WT: Basal ganglia motor control. II. Late pallidal timing relative to movement onset and inconsistent pallidal coding of movement parameters. J Neurophysiol 1991, 65: Schultz W: Activity of pars reticulata neurons of monkey substantia nigra in relation to motor, sensory, and complex events. J Neurophysiol 1986, 55: Mink JW, Thach WT: Basal ganglia motor control. I. Nonexclusive relation of pallidal discharge to five movement modes. J Neurophysiol 1991, 65: Horak FB, Anderson ME: Influence of globus pallidus on arm movements in monkeys. I. Effects of kainic acid-induced lesions. J Neurophysiol 1984, 52: Mink JW, Thach WT: Basal ganglia motor control. III. Pallidal ablation: normal reaction time, muscle cocontraction, and slow movement. J Neurophysiol 1991, 65: Kato M, Kimura M: Effects of reversible blockade of basal ganglia on voluntary arm movement. J Neurophysiol 1992, 68: Mink JW: The basal ganglia: focused selection and inhibition of competing motor programs. Prog Neurobiol 1996, 50: Miller WC, DeLong MR: Altered tonic activity of neurons in the globus pallidus and subthalamic nucleus in the primate MPTP model of parkinsonism. In The Basal Ganglia II. Edited by Carpenter MB, Jayaraman A. Plenum Press; 1987: Filion M, Tremblay L: Abnormal spontaneous activity of globus pallidus neurons in monkeys with MPTP-induced parkinsonism. Brain Res 1991, 547: Hutchinson WD, Levy R, Dostrovsky JO, Lozano AM, Lang AE: Effects of apomorphine on globus pallidus neurons in parkinsonian patients. Ann Neurol 1997, 42: Merello M, Balej J, Delfino M, Cammarota A, Betti O, Leiguarda R: Apomorphine induces changes in GPi spontaneous outflow in patients with Parkinson s disease. Mov Disord 1999, 14: Filion M, Tremblay L, Bedard PJ: Effects of dopamine agonists on the spontaneous activity of globus pallidus neurons in monkeys with MPTP-induced parkinsonism.. Brain Res 1991, 547: Papa SM, DeSimone R, Fiorani M, Oldfield EH: Internal globus pallidus discharge is nearly suppressed during levodopainduced dyskinesias. Ann Neurol 1999, 46: Bergman H, Wichmann T, DeLong MR: Reversal of experimental parkinsonism by lesions of the subthalamic nucleus. Science 1990, 249: Boraud T, Bezard E, Bioulac B, Gross CE: From single extracellular unit recording in experimental and human Parkinsonism to the development of a functional concept of the role played by the basal ganglia in motor control. Prog Neurobiol 2002, 66: Pessiglione M, Guehl D, Rolland AS, Francois C, Hirsch EC, Feger J, Tremblay L: Thalamic neuronal activity in dopaminedepleted primates: evidence for a loss of functional segregation within basal ganglia circuits. J Neurosci 2005, 25: Goldberg JA, Boraud T, Maraton S, Haber SN, Vaadia E, Bergman H: Enhanced synchrony among primary motor cortex neurons in the 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine primate model of Parkinson s disease. J Neurosci 2002, 22: Boraud T, Bezard E, Bioulac B, Gross CE: Dopamine agonistinduced dyskinesias are correlated to both firing pattern and frequency alterations of pallidal neurones in the MPTP-treated monkey. Brain 2001, 124: Current Opinion in Neurobiology 2006, 16:

63 Results III Basal ganglia oscillations and pathophysiology of movement disorders Rivlin-Etzion, Marmor, Heimer, Raz, Nini and Bergman Wichmann T, Soares J: Neuronal firing before and after burst discharges in the monkey Basal Ganglia is predictably patterned in the normal state and altered in parkinsonism. J Neurophysiol 2006, 95: It is known that burst discharges in basal ganglia neurons are more common in PD than under normal conditions, but this study reveals significant changes in the structure of burst or peri-burst discharge in the STN, GPe and GPi of MPTP-treated macaque monkeys. Thus, complex changes in the burst structure, and in burst frequency, might contribute to abnormal information processing in PD. 58. Wichmann T, Bergman H, DeLong MR: The primate subthalamic nucleus. III. Changes in motor behavior and neuronal activity in the internal pallidum induced by subthalamic inactivation in the MPTP model of parkinsonism. J Neurophysiol 1994, 72: Raz A, Feingold A, Zelanskaya V, Vaadia E, Bergman H: Neuronal synchronization of tonically active neurons in the striatum of normal and parkinsonian primates. J Neurophysiol 1996, 76: Raz A, Frechter-Mazar V, Feingold A, Abeles M, Vaadia E, Bergman H: Activity of pallidal and striatal tonically active neurons is correlated in MPTP-treated monkeys but not in normal monkeys. J Neurosci 2001, 21:RC Hutchison WD, Lozano AM, Tasker RR, Lang AE, Dostrovsky JO: Identification and characterization of neurons with tremorfrequency activity in human globus pallidus. Exp Brain Res 1997, 113: Lemstra AW, Verhagen ML, Lee JI, Dougherty PM, Lenz FA: Tremor-frequency (3 6 Hz) activity in the sensorimotor arm representation of the internal segment of the globus pallidus in patients with Parkinson s disease. Neurosci Lett 1999, 267: Hurtado JM, Rubchinsky LL, Sigvardt KA, Wheelock VL, Pappas CT: Temporal evolution of oscillations and synchrony in GPi/muscle pairs in Parkinson s disease. J Neurophysiol 2005, 93: Advanced time-dependent phase correlation techniques were applied to 27 pairs of tremor-related GPi single units and EMG of PD patients undergoing stereotactic neurosurgery. Analysis using short (2s) sliding windows shows that oscillatory activity in both GPi oscillatory units and muscles occurs intermittently over time. There was partial overlap in the times of oscillatory activity but, in most cases, no correlation was found between the times of oscillatory episodes in the two signals. Phaselocking analysis revealed that pallidal oscillations and tremor are punctuated by phase slips, which were classified as synchronizing or desynchronizing. The results of this high-level quantitative characterization of parkinsonian tremor and pallidal oscillations can be explained by either a dynamical connectivity structure from the basal ganglia to the periphery or tremor generators downstream of the basal ganglia. 64. Lenz FA, Tasker RR, Kwan HC, Schnider S, Kwong R, Murayama Y, Dostrovsky JO, Murphy JT: Single unit analysis of the human ventral thalamic nuclear group: correlation of thalamic tremor cells with the 3-6 Hz component of parkinsonian tremor. J Neurosci 1988, 8: Kuhn AA, Kupsch A, Schneider GH, Brown P: Reduction in subthalamic 8-35 Hz oscillatory activity correlates with clinical improvement in Parkinson s disease. Eur J Neurosci 2006, 23: Previous studies by this group revealed strong 8-35 Hz LFP oscillations in the subthalamic nucleus (STN) of PD patients. In this study, they report a reduction in peak activity in the 8-35 Hz band with levodopa that is positively correlated with the improvement in the contralateral akinesia and rigidity (but not with tremor). A trend towards negative correlations was found between gamma band LFP power and PD symptoms, suggesting that positive correlations were relatively frequency-specific 66. Wang SY, Aziz TZ, Stein JF, Liu X: Time-frequency analysis of transient neuromuscular events: dynamic changes in activity of the subthalamic nucleus and forearm muscles related to the intermittent resting tremor. J Neurosci Methods 2005, 145: This study implemented a highly advanced dynamic analysis (short-time Fourier transform and continuous wavelet transform) of the functional correlation between both neural and muscular activity and the STN LFP over several episodes of transient resting tremor from a PD patient. A significant suppression in the power of the STN LFPs in the b band (10-30 Hz) preceded the onset of resting tremor. During the tremor episode the power at the tremor frequency ( Hz) in STN LFPs increased significantly. 67. Timmermann L, Gross J, Dirks M, Volkmann J, Freund HJ, Schnitzler A: The cerebral oscillatory network of parkinsonian resting tremor. Brain 2003, 126: Sharott A, Magill PJ, Harnack D, Kupsch A, Meissner W, Brown P: Dopamine depletion increases the power and coherence of beta-oscillations in the cerebral cortex and subthalamic nucleus of the awake rat. Eur J Neurosci 2005, 21: Kuhn AA, Trottenberg T, Kivi A, Kupsch A, Schneider GH, Brown P: The relationship between local field potential and neuronal discharge in the subthalamic nucleus of patients with Parkinson s disease. Exp Neurol 2005, 194: Goldberg JA, Rokni U, Boraud T, Vaadia E, Bergman H: Spike synchronization in the cortex-basal Ganglia networks of parkinsonian primates reflects global dynamics of the local field potentials. J Neurosci 2004, 24: Schneidman E, Berry MJ, Segev R, Bialek W: Weak pairwise correlations imply strongly correlated network states in a neural population. Nature 2006, 440: Pare D, Curro Dossi R, Steriade M: Neuronal basis of the parkinsonian resting tremor: a hypothesis and its implications for treatment. Neuroscience 1990, 35: Tass P, Rosenblum MG, Weule J, Kurths J, Pikovsky A, Volkmann J, Schnitzler A, Freund H-J: Detection of n:m phase locking from noisy data: application to magnetoencephalography. Phys Rev Lett 1998, 81: Halliday DM, Rosenberg JR, Amjad AM, Breeze P, Conway BA, Farmer SF: A framework for the analysis of mixed time series/ point process data theory and application to the study of physiological tremor, single motor unit discharges and electromyograms. Prog Biophys Mol Biol 1995, 64: Mehta MR, Bergman H: Loss of frequencies in autocorrelations and a procedure to recover them. J Neurosci Methods 1995, 62: Rivlin-Etzion M, Ritov Y, Heimer G, Bergman H, Bar-Gad I: Local shuffling of spike trains boosts the accuracy of spike train spectral analysis. J Neurophysiol 2006, 95: Gresty M, Buckwell D: Spectral analysis of tremor: understanding the results. J Neurol Neurosurg Psychiatry 1990, 53: Current Opinion in Neurobiology 2006, 16:

64 Results IV IV. HIGH FREQUENCY OSCILLATIONS IN THE GLOBUS PALLIDUS ARE CORRELATED WITH PARKINSONIAN TREMOR This chapter includes results which have not yet been published in the scientific literature. 62

65 Results IV 10 Hz oscillatory activity in the globus pallidus is correlated with the tremor phenomenon in the MPTP primate model of Parkinsonism Abstract Synchronous oscillations are commonly found in the basal ganglia and the thalamus of human patients and animal models of Parkinson's disease (PD). The frequency of these oscillations is often similar to that of the parkinsonian tremor, but their role in generating the tremor or other parkinsonian symptoms is still under debate. The tremor is intermittent and does not appear in all human patients. Similarly, primate models tend to develop tremor as a function of species of monkey, with the African green (vervet) monkeys usually demonstrating a high-amplitude, low-frequency (4 7 Hz) tremor beyond their akinesia and bradykinesia, whereas macaques tend to be akinetic-rigid and rarely demonstrate a low-amplitude high frequency (10 12 Hz) action-postural tremor. We took advantage of this fact and studied the appearance of the synchronicity and oscillations in six monkeys, three vervets and three macaques, before and after 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP) systemic treatment and induction of parkinsonism. Multiple extracellular recordings were conducted in the primary motor cortex (MI) of two monkeys, and in the globus pallidus (GP) of all six monkeys. All the monkeys became akinetic and bradykinetic as a result of the MPTP treatment, but only vervets demonstrated prolonged episodes of low frequency (4-6 Hz) tremor, whereas macaques were non-tremulous. The GP population exhibited ~5 Hz oscillatory activity in all six monkeys, while ~10 Hz neural oscillations were only detected in the tremulous monkeys. However, the activity of the cortical neurons became strongly oscillatory at ~10 Hz in one of these monkeys, but not the other, although both were tremulous and had comparable pallidal oscillatory activity. Finally, synchronous oscillations, when present, were centered around the higher frequencies of oscillations. These findings suggest there is a correlation between high frequency GP neural oscillations and tremor. Furthermore, based on the results, we question the involvement of MI in transferring these 10 Hz pallidal oscillations, and suggest that the brain stem could serve as a potential module to transfer GP high frequency activity to the muscles. 63

66 Results IV Introduction The motor symptoms of Parkinson's disease (PD) are characterized by two antithetical features: a poverty of movement on one hand, including reduction of spontaneous movements and difficulty in initiating a voluntary movement (akinesia) and slowness of movement (bradikinesia); and involuntary movement on the other hand which emerges as a low frequency (4-7 Hz) tremor. In addition, PD patients suffer from muscle rigidity, but the classification of this symptom as hyper- or hypo- motoric has yet to be firmly established. The major cellular event leading to the motor symptoms of PD is the death of midbrain dopaminergic neurons, resulting in dopamine (DA) depletion in the input nuclei of the basal ganglia, the striatum 1. Although the basal ganglia are related to many functions, including motor action, cognition, learning and motivation 2, 3, it is still unknown how the DA depletion leads to the conflicting hypo- and hyper- kinetic symptoms of the disease. Recordings from different nuclei of the basal ganglia in primate as well as rodent models of Parkinsonism reveal that the symptoms of the disease are accompanied by several changes in neural activity, including the emergence of an oscillatory and bursty neuronal pattern These periodic burst-oscillations can generate some or all of the motor symptoms of the disease either via the Cortico-Baso-Thalamo-Cortical (CBTC) loop 2, 12-14, or via the direct projections of BG output nuclei to brainstem motor centers 15. Despite the fact that the majority of GPi axons (estimated at 70%) branch to both the thalamus and brainstem 15, most of the information processing models of the basal ganglia ignore the latter projections and focus on the CBTC loop 13, Recently it has been hypothesized that, basal ganglia output projections to the brainstem may be involved in generating the motor deficits of the disease 19. Due to their common periodic nature, tremor is the primary symptom traditionally thought to originate from basal ganglia oscillatory activity. Yet, despite the existence of neurons with tremor-frequency activity ("tremor cells") in the GP of PD patients 20, significant correlations between pallidal activity and tremor are rare and intermittent 21, 22. Moreover, low-frequency oscillations (1-7 Hz) are present in the local field potential of the subthalamic nucleus (STN) of predominantly akinetic-rigid patients, and their power is even increased during ongoing STN-deep brain stimulation (DBS) and following dopaminergic treatment 23, 24. These findings are in line with MPTP primate studies: in addition to akinesia and rigidity, African green monkeys treated 64

67 Results IV with the neurotoxin usually develop a low-frequency (~4 7 Hz) tremor resembling the classic resting tremor of PD. Similar to trembling patients, consistent relations between tremor and oscillations have not been found in these tremulous monkeys 8. Furthermore, low frequency oscillations have been shown to persist in MPTP treated monkeys after amelioration of tremor 25. Finally, MPTP treated Macaques, which tend to be akinetic-rigid and rarely develop infrequent, short episodes of high-frequency (~10 12 Hz) action-postural tremor, also exhibit basal ganglia neuronal oscillations 4, 14. Even though PD is usually thought to originate in the BG due to its dopaminergic loss, tremor may not be generated by oscillations in the basal ganglia, but rather has a cortical or thalamic origin. Indeed, motor and somatosensory cortices of PD patients also tend to oscillate in correlation with the tremor 26-28, and "tremor cells" that oscillated in coherence with the tremor have often been found in the thalamus of PD patients It should be borne in mind that oscillatory activity in any neuronal structure may merely reflect sensory ascending pathways from the periphery, or an oscillatory activity originating in another neuronal structure, and does not necessarily indicate the specific origin of the tremor 32. Moreover, the Parkinsonian tremor may not be generated by a central oscillator, but rather have a spinal origin, as was suggested in early studies 33, 34. This paper examines oscillatory activity in the GP of six MPTP treated monkeys, tremulous as well as non-tremulous, and demonstrates that oscillatory activity appears in all of them. However, the distribution of oscillation frequencies differs across the monkeys depending on the existence of tremor. In addition, we explore the neuronal pattern of the motor cortex of two of the tremulous monkeys, and show that oscillatory activity of the motor cortex is not obligatory for tremor to emerge, indicating that tremor may not be caused by cortical oscillations, and therefore is probably not generated by BG oscillations transformed through the CBTC loop. Methods Animals. This study is based on data collected from six monkeys: three vervets (African green, Cercopithecus aethiops aethiops, T, W and C; females; weighing 3, 3.5 and 3.8 kg respectively) and three macaques (one Macaca mulatta, R; female; weighing 5.7 kg, and two Macaca fascicularis, H and P; females; weighing 3.2 and 3 kg). Before any procedures were carried out, the monkeys were trained to sit in a 65

68 Results IV primate chair, to permit handling by the experimenter, and became familiarized with the laboratory setting. Monkeys C and R were trained to perform a simple visuomotor task 9. Monkey H was trained on a self-initiated probabilistic delayed visual-motor task 35, 36. However, most of the recordings in these monkeys (C, R and H) were conducted during a "quiet-wakeful" state. None of the other monkeys (T, W, P) were engaged in any behavioral task 37, 38 and recordings were also carried in the "quietwakeful" state. The monkeys' health was monitored by a veterinarian, and their weights and clinical status were checked daily. All experimental protocols were performed in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and the Hebrew University guidelines for the use and care of laboratory animals in research and were approved and supervised by the Institutional Animal Care and Use Committee. Surgery and recording procedures. After training, a recording chamber was attached to the monkeys' skulls. The recording chamber was tilted laterally in the coronal plane in all monkeys except monkey H where it had a 0 tilt (and therefore permitted recordings from both hemispheres). All chambers were targeted by stereotaxic procedures to cover most of the globus pallidus (GP) territory. In monkeys T and W the position of the chambers allowed access to the arm- related area of the primary motor cortex (MI) as well. The exact position of the chamber was established using a magnetic resonance imaging (MRI) scan and electrophysiological mapping. Surgical and MRI procedures were carried under full (isoflurane, N 2 O) anesthesia and sedation (Dormitor, Ketamine) respectively. Extracellular simultaneous recordings using 4-8 electrodes in the GP, and 4 electrodes in MI (monkeys T and W) were performed. Details of the surgery, identification of neurons, and data-recording methods were described previously Unlike the situation in the normal monkey, the electrophysiological differences (mainly discharge pattern) between the internal and the external segments are less clear after MPTP and the induction of the PD symptoms 40. An initial study of our group 8 failed to find differences between pallidal segments, although a second study 9, 39 did find variations between GPe and GPi. In this study, therefore, we did not differentiate between neurons located in the two segments of the globus pallidus. However, we attempted to cover most of the GP area (both internal and external segments) in the recordings. 66

69 Results IV In monkey H we started recordings from the right hemisphere. After a month of recordings in the normal state the monkey developed a poverty of movement in its contralateral limbs. We stopped the recordings and started treatment with steroids for several days following which the monkey regained most of its left side motor abilities, though not fully. Recordings restarted from the left hemisphere from which we recorded throughout the rest of the experiment. Histology of monkey H revealed that it developed a medium sized hematoma in its right hemisphere. We did not observe significant anatomical changes (beyond the MPTP induced dopaminergic degeneration and the electrode tracks) in the histological examination in any of the other monkeys. MPTP treatment and perfusion. After a period of recording in the normal state, Parkinsonism was induced by five intramuscular injections of 0.4 mg/kg of 1-methyl- 4-phenyl-1,2,3,6-tetrahydropyridine-HCl (MPTP, Sigma, Rehovot, Israel). The MPTP injections were given under light intramuscular ketamine hydrochloride (10 mg/kg) anesthesia and over a period of 4 days (3 injections in the first 24 hours). The clinical state of the monkeys was assessed daily according to a primate scale of Parkinsonism 41. In addition, each monkey was given a total subjective score by its experimenter between 0 and 4 (4 being the most severe) on four cardinal motor features: akinesia and bradykinesia, flexed posture, rigidity and tremor. Upon termination of the recording days in the MPTP state the monkeys were treated with dopamine replacement therapy to verify the diagnosis of Parkinsonism by clinical improvement resulting from dopamine-replacement treatment (l-dopa and agonists). Doses for monkey T, H and P were 0.5x25/250 mg of Dopicar (L-3,4-dihydroxyphenylalanine and carbidopa; Teva Pharmaceutical Industries Ltd.) twice a day. Monkeys C and R starting doses were 0.5x25/250 mg of Dopicar (Merck Sharp and Dohme, Haarlem, The Netherlands) in the morning and 5 mg of Parlodel (Bromocryptine; Sandoz, Basel, Switzerland) divided equally between morning and evening. The drugs were administered orally as crushed powder dissolved in liquid. In monkeys T, H, C and R recordings were also conducted after treatment with dopamine replacement therapy (data not reported here). Monkey W died before it was given any dopaminergic treatment. 67

70 Results IV At the end of the experiment, the monkeys were deeply anesthetized with a lethal dose of pentobarbital and perfused through the heart with saline followed by a 4% paraformaldehyde solution. The brains of the monkeys were removed, serial sections of 50 µm were cut on a freezing microtome and every 12 th section was processed for Nissl or tyrosine hydroxylase (TH) immunohistochemistry (for details, see 9, 38 ). Monkey W was perfused in a similar way within 30 min of her death. Accelerometers. In monkeys C, R, T and W we used uniaxial accelerometers (8630C5; Kistler, Amherst, NY) to assess limb tremor. The monkeys had the accelerometer fastened to the back of their non-restrained wrist (contra-lateral to the recording hemisphere), whereas in monkeys T and C accelerometers were fastened to all four limbs. For details regarding accelerometer recordings, see 9, 38. Data analysis. Cells were selected for recording as a function of their signal-to-noise ratio and real-time assessment of their isolation quality. Only stable (minimum 5 minutes, off-line verification of the stability of the neurons' firing rates throughout the recording session; we discarded any neuron that demonstrated a trend of decaying or increasing firing rate since this is indicative of possible neuronal injuries or unstable electrode position) and isolated (as judged by the experimenters in real time) units were included in the analysis database of this study. Details regarding recording durations and total number of units in each monkey that met the above criteria and were included in this study are shown in Table 1. We conducted quantitative analyses of oscillatory firing patterns of all cells included in the study. Oscillatory activity of the cells was estimated using the power spectrum density (PSD) of the spike trains (frequency resolution, 0.25 Hz). The oscillatory activity of pallidal neurons was assessed using the shuffling method 42 in order to compensate for the spectral distortion that arises due to the refractory period of neurons with a high discharge rate. Briefly, the spectrum of the original spike train is divided by the mean spectrum of the locally (T=~175 ms) shuffled spike trains (n = 20 shuffled trains), resulting in a ratio termed 'compensated PSD'. Due to the low discharge rate of cortical neurons 38, 43, the oscillatory activity in the cortex could be evaluated based on the original PSD. A confidence level (p < 0.001, normalized to the total number of bins) was constructed based on the high-frequency range of Hz, at which the PSD was flat. A cell was considered oscillatory if its compensated 68

71 Results IV (GP) or original (cortex) PSD contained at least two consecutive bins within the range of 4 15 Hz that crossed the p = confidence level. Table 1: Mean recording durations of single and pairs of cells T W C R H P MI single GP single MI pairs GP pairs MI-GP pairs Normal MPTP Normal MPTP Normal MPTP Normal MPTP Normal MPTP 14.95±5.03 (n = 230) 11.84± 3.71 (n = 176) 12.60± 4.60 (n = 114) 12.45± 3.72 (n = 138) 14.54±4.75 (n = 568) 11.04±3.69 (n = 434) 11.44±4.56 (n = 125) 11.69±3.93 (n = 254) 12.71±4.35 (n = 695) 10.75±3.56 (n = 846) 10.5± 2.35 (n = 358) 11.45± 2.52 (n = 122) 9.63± 2.45 (n = 163) 10.34± 3.01 (n = 75) 10.04±2.41 (n = 947) 10.58±2.65 (n = 363) 8.75±2.32 (n = 202) 9.63±2.96 (n = 131) 9.03±2.38 (n = 1073) 9.80±2.79 (n = 573) 29.61± (n = 244) 28.73± 7.72 (n = 268) 25.11±11.04 (n = 786) 24.89±8.85 (n = 1165) 23.71± (n = 48) 19.82± 9.65 (n = 97) 18.98±9.22 (n = 111) 15.07±7.77 (n = 195) 19.4± (n = 115) 15.28± 6.76 (n = 157) 16.7±13.04 (n = 105) 14.15±6.15 (n = 168) Times are given in minutes as Mean±std. n stands for the number of neurons on which the values are based ± 4.69 (n = 79) 10.12± 4.21 (n = 232) 9.23±3.20 (n = 51) 9.68±4.07 (n = 312) For the analysis of synchronous oscillations we used pairs of neurons that were simultaneously recorded and showed stable and isolated overlapping activity for at least 5 minutes (Table 1). Only neuronal pairs that were recorded by different electrodes were included in this study to avoid possible artifacts attributable to a shadowing effect of high discharge rates in cells recorded from the same electrode 44. Synchronous oscillations within GP pairs as well as between MI and GP neurons were assessed using the shuffling method 42 due to GP high discharge rate. The crossspectrum of the original spike trains was divided by the mean cross-spectrum of the globally shuffled (n = 20) spike trains. A confidence level (p < 0.001, normalized to 69

72 Results IV the total number of bins) for the compensated spectrum was constructed based on the high-frequency range of Hz, at which the spectrum was flat. A correlogram was considered to have significant periodic oscillations if its compensated spectrum contained at least two consecutive bins within the range of 4 15 Hz that crossed the p < confidence level. For cortical pairs, we used the conventional significance criterion for the coherence function as a confidence level 45, 46 L 1 : 1 (1 α ), where α is the level of confidence (here α=0.999), and L is the number of windows used in the calculation (length of the data divided by the window size, which in our case was 4096). All analyses were carried on custom developed software using Matlab 7.1 tools (The MathWorks, Natick, MA). The same procedures and thresholds were applied to all monkeys and all clinical states. 1 Results Animals' clinical states Table 2 summarizes the subjective scores, assigned by the researchers to the six MPTP treated monkeys whose data are included here. Three of the monkeys were macaques, and three were African green monkeys. All monkeys developed the first signs of Parkinsonism by the first day after last MPTP injection. Parkinsonian symptoms continued to evolve over the following days, and stabilized within six days or less. They remained stable until the last recording day included in this study (before any DA treatment was given, see Table 2). All monkeys developed severe akinesia and bradykinesia, as well as flexed posture. The monkeys differed in their rigidity scores. Four monkeys were severely rigid, and monkey W demonstrated the lowest rigidity score. This monkey died 7 days after last MPTP injection, and showed stable akinesia, bradykinesia and tremor by the fourth day after MPTP injections. As was previously reported 9, the tremor score differed between the two monkey species, rising in all African green monkeys at Hz, but not in the macaques (though monkey R developed high frequency (~9 Hz) tremor after dopaminergic treatment, see 9 for details). Responsiveness to DA treatment was tested to verify the diagnosis of Parkinsonism (see methods). All monkeys responded to the treatment by regaining the ability to self-feed, an increase in amount and velocity of spontaneous movements and 70

73 Results IV straightening of posture. Yet, monkeys T, C and R demonstrated a full response to the DA treatment, whereas monkeys H and P responded to a lesser extent, with a fewer voluntary movements and only partial stable posture (Table 2). Table 2: Subjective rating of the clinical scores and history of the MPTP treated monkeys* Akinesia/ First Last Flexed Monkey Species bradykine Rigidity Tremor recording recording Posture sia day day Response to DA treatment R Macaque Full P Macaque Partial H Macaque Partial C AGM Full T AGM Full W AGM NA * for all scores, 0 - normal, 4 - most severe First recording day after last MPTP injection. Last recording day after last MPTP injection (before any DA treatment was given). AGM African green monkey. Full Full response to the DA treatment including regaining of ability to self-feed, an increase in amount and velocity of movements and straightening of posture. Partial Partial response to the DA treatment including regaining of ability to self-feed, a moderate increase in amount and velocity of movements and a partial straightening of posture. NA Not available (DA treatment not given). Histology All MPTP treated monkeys had an almost complete washout of TH staining in the striatum except the shell area of the ventral striatum. Cell loss in the midbrain was almost complete in the ventral tier and the lateral portions of the substantia nigra pars compacta in all animals except monkey W, which showed a less severe loss of TH midbrain staining, probably due to its early death. As expected 47, the cells in the ventral tegmental area of the midbrain dopamine system remained relatively spared. 71

74 Doctoral Thesis, Michal Rivlin-Etzion Results IV Spontaneous activity in MI and GP: example The spontaneous activity recorded by extracellular electrodes located in the primary motor cortex and the globus pallidus of monkey T before and after MPTP injection is illustrated in Figure 1. PSDs of the spike trains revealed that cortical activity became periodic at 10 Hz in the Parkinsonian state, and coherences between cortical neurons demonstrated that these oscillations were synchronized (figure 2). There was 5 Hz activity in the pallidum in the Parkinsonian state, but the coherence analysis showed that these oscillations were synchronized to a lesser extent (note the different Y-scales in the figure). Finally, the cortical and the pallidal oscillations were not coherent. Figure 1: Example of motor cortex and pallidal activity before and after MPTP treatment. Extracellular simultaneous recordings from electrodes located in the motor cortex (blue) and globus pallidus (red) in the normal (upper traces) and parkinsonian (lower traces) states. Time scale of 0.5 second is indicated by the horizontal line. 72

75 Results IV Figure 2: Power spectral densities (PSDs) (framed) and coherences of the spike trains detected from the analog data shown in Figure 1. In cases where more than one single unit was detected by the electrode, a single spike train was chosen for the illustration. Y axes of coherences are depicted in the range of 0-0.1, except for coherences within the cortex in MPTP state where it ranges from 0 to 0.7 (indicated in the figure). Emergence of neuronal oscillations in the parkinsonian state: analysis of the six monkeys Figure 3 depicts the fraction of neuronal oscillations of single neurons (upper row) and synchronous oscillations (pair-wise correlations, lower row) before and after MPTP treatment. In the MPTP state, significant (p<0.01, χ² test) oscillatory activity of 73

76 Results IV single neurons emerged in the motor cortex of two monkeys (2/2) and in the globus pallidus of four monkeys (4/6). Pallidal oscillatory activity also emerged in the Parkinsonian state in the two other monkeys (H and P), but as they demonstrated some oscillatory activity in the normal state, the differences between the two states were not significant. Significant synchronized neuronal oscillations (p<0.01, χ² test) of the motor cortex of monkeys T and W emerged in the Parkinsonian state, although synchronous oscillations were more robust in monkey T: almost 40% of the recorded pairs. All monkeys exhibited synchronous oscillations in the Parkinsonian GP, but only monkey C demonstrated high percentages of synchronized pairs to the same extent as reported in earlier studies of vervet monkeys 8. The fraction of synchronized MI-GP pairs increased significantly in the MPTP state in both monkeys T and W (p<0.01), but did not reach the 2.5% cutoff. Figure 3: Fraction of neuronal oscillations in the motor cortex and globus pallidus. Fraction of single (upper row) and coupled (lower row) neuronal oscillations in the normal (white) and parkinsonian (black) states. Left column depicts fractions in the motor cortex of two monkeys, middle column depicts the fraction in the GP of six monkeys, and the right column presents the fraction of coupled MI-GP oscillations. 74

77 Results IV Frequency distribution of MPTP oscillatory activity The distributions of the frequencies of oscillations in the Parkinsonian state are depicted in Figure 4. Interestingly, despite the fact that both monkeys T and W were tremulous monkeys, only the MI of monkey T exhibited oscillatory activity that was centered around a single frequency 10 Hz, which was twice the tremor frequency (5 Hz, as indicated by the accelerometers, data not shown). Moreover, although all monkeys, regardless of their species or tremulous activity, had dominant ~5 Hz oscillatory activity in the GP, only tremulous vervet monkeys displayed oscillatory activity confined to ~10 Hz as well. GP synchronous oscillations, if present (mainly in monkey C), were centered around the 10 Hz - double-tremor frequency. Figure 4: Distribution of single and coupled neuronal oscillations in the motor cortex and globus pallidus in the parkinsonian state. Discussion This study was based on data collected from six MPTP treated monkeys: three tremulous African green monkeys and three non-tremulous macaques. The manifestation of 5 Hz neuronal oscillations in the GP of all monkeys suggests that these GP low frequency oscillations, although sharing a similar frequency with the Parkinsonian tremor, do not generate the tremor, and are not driven by it. In contrast, 75

78 Results IV high frequency pallidal oscillations were found only in tremulous monkeys, which might indicate a relation between these 10 Hz oscillations and the tremor. Interestingly, these oscillations are probably filtered out at the cortex or more upstream (e.g. the thalamus), since the cortical activity in only one but not the other African green monkeys became strongly oscillatory at 10 Hz. The frequencies of oscillations in the cortex of the other monkey were more uniformly distributed, despite the fact that both monkeys demonstrated a similar fraction of oscillations in the GP, which were comparably distributed bi-modally around 5 and 10 Hz. These filtration properties were also revealed using electrical stimulations delivered to the Parkinsonian GP: only bursts delivered at 1 and 2 Hz frequencies, but not those delivered at 5 Hz and higher frequencies were reflected in the motor cortex 38. We therefore hypothesize that GP high frequency oscillations are transformed via the brainstem and may generate the Parkinsonian tremor in a 2:1 or other complex relationship. We do not know why the 5 Hz pallidal oscillations do not result in a 2.5 Hz tremor as well, but a reasonable explanation could be that most of the synchronous oscillatory activity in the GP, if present, is located at the 10 Hz higher frequencies and not at the lower ones (Figure 4, as well as 8, 9. Indeed, magnetoencephalography (MEG) recordings of PD patients reveal that significant coherence activity between different cortical areas is at tremor frequency but even stronger at double tremor frequency 28. The evidence presented here does not rule out the possibility that the 10 Hz oscillatory activity in the GP is merely a reflection of the peripheral tremor and is generated by a sensory input: antagonist muscles that have opposite phases of the 5 Hz tremor can sum up at the GP level and generate a 10 Hz activity. The percentages of pallidal oscillations did not exceed 20% in most of the monkeys (figure 3). Moreover, the 10 Hz oscillations in the GP in two out of the three tremulous monkeys were found only in 5-6% of the recorded neurons. This may be due to the fact that we attempted to cover most of the pallidal area in the experiments. Had we restricted our recordings to the motor part of the GP 48, the percentages of oscillations might have significantly increased. Associating pathological neural activities and motor symptoms is one step in the search for PD treatment and its different subtypes in particular. We believe that our finding of 10 Hz pallidal activity which discriminates tremulous and non-tremulous Parkinsonian monkeys contributes to progress in finding such an association. 76

79 Results IV Reference List 1. Jellinger,K. The pathology of parkinsonism in Movement Disorders 2 (eds. Marsden,C.D. & Fahn,S.) (Butterworths, London, 1987). 2. Gerfen,C.R. & Wilson,C.J. The basal ganglia in Handbook of Chemical Neuroanatomy, Vol 12: Integrated Systems of the CNS, Part III (eds. Swanson,L.W., Bjorklund,A. & Hokfelt,T.) (Elsevier Science, 1996). 3. Middleton,F.A. & Strick,P.L. Basal ganglia output and cognition: evidence from anatomical, behavioral, and clinical studies. Brain Cogn 42, (2000). 4. Miller,W.C. & DeLong,M.R. Altered tonic activity of neurons in the globus pallidus and subthalamic nucleus in the primate MPTP model of parkinsonism. in The Basal Ganglia II (eds. Carpenter,M.B. & Jayaraman,A.) (Plenum Press, New York, 1987). 5. Filion,M. & Tremblay,L. Abnormal spontaneous activity of globus pallidus neurons in monkeys with MPTP-induced Parkinsonism. Brain Res. 547, (1991). 6. Bergman,H., Wichmann,T., Karmon,B., & DeLong,M.R. The primate subthalamic nucleus. II. Neuronal activity in the MPTP model of Parkinsonism. J. Neurophysiol. 72, (1994). 7. Nini,A., Feingold,A., Slovin,H., & Bergman,H. Neurons in the globus pallidus do not show correlated activity in the normal monkey, but phase-locked oscillations appear in the MPTP model of Parkinsonism. J. Neurophysiol. 74, (1995). 8. Raz,A., Vaadia,E., & Bergman,H. Firing patterns and correlations of spontaneous discharge of pallidal neurons in the normal and the tremulous 1-methyl- 4-phenyl-1,2,3,6-tetrahydropyridine vervet model of Parkinsonism. J Neurosci. 20, (2000). 9. Heimer,G. et al. Dopamine replacement therapy does not restore the full spectrum of normal pallidal activity in the 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine primate model of Parkinsonism. J Neurosci. 26, (2006). 10. Dejean,C., Gross,C.E., Bioulac,B., & Boraud,T. Dynamic changes in the cortex-basal ganglia network after dopamine depletion in the rat. J. Neurophysiol. 100, (2008). 11. Mallet,N. et al. Disrupted dopamine transmission and the emergence of exaggerated beta oscillations in subthalamic nucleus and cerebral cortex. J. Neurosci. 28, (2008). 12. Parent,A. & Hazrati,L.N. Functional anatomy of the basal ganglia. I. The cortico-basal ganglia-thalamo-cortical loop. Brain Res. Rev. 20, (1995). 13. Mink,J.W. The basal ganglia: focused selection and inhibition of competing motor programs. Prog. Neurobiol. 50, (1996). 14. Bergman,H. et al. Physiological aspects of information processing in the basal ganglia of normal and parkinsonian primates. Trends. Neurosci. 21, (1998). 15. Parent,M., Levesque,M., & Parent,A. The pallidofugal projection system in primates: evidence for neurons branching ipsilaterally and contralaterally to the thalamus and brainstem. J. Chem. Neuroanat. 16, (1999). 16. Albin,R.L., Young,A.B., & Penney,J.B. The functional anatomy of basal ganglia disorders. Trends Neurosci. 12, (1989). 77

80 Results IV 17. DeLong,M.R. Primate models of movement disorders of basal ganglia origin. Trends. Neurosci. 13, (1990). 18. Bar-Gad,I., Havazelet Heimer,G., Goldberg,J.A., Ruppin,E., & Bergman,H. Reinforcement driven dimensionality reduction - a model for information processing in the basal ganglia. Journal of Basic & Clinical Physiology & Pharmacology 11, (2000). 19. Stein,J.F. Akinesia, motor oscillations and the pedunculopontine nucleus in rats and men. Exp. Neurol. 215, 1-4 (2009). 20. Hutchison,W.D., Lozano,A.M., Tasker,R.R., Lang,A.E., & Dostrovsky,J.O. Identification and characterization of neurons with tremor- frequency activity in human globus pallidus. Exp. Brain Res. 113, (1997). 21. Lemstra,A.W., Verhagen,M.L., Lee,J.I., Dougherty,P.M., & Lenz,F.A. Tremorfrequency (3-6 Hz) activity in the sensorimotor arm representation of the internal segment of the globus pallidus in patients with Parkinson's disease. Neurosci. Lett. 267, (1999). 22. Hurtado,J.M., Rubchinsky,L.L., Sigvardt,K.A., Wheelock,V.L., & Pappas,C.T. Temporal evolution of oscillations and synchrony in GPi/muscle pairs in Parkinson's disease. J. Neurophysiol. 93, (2005). 23. Priori,A. et al. Rhythm-specific pharmacological modulation of subthalamic activity in Parkinson's disease. Exp. Neurol. 189, (2004). 24. Rossi,L. et al. Subthalamic local field potential oscillations during ongoing deep brain stimulation in Parkinson's disease. Brain Res. Bull. 76, (2008). 25. Wichmann,T., Bergman,H., & DeLong,M.R. The primate subthalamic nucleus. III. Changes in motor behavior and neuronal activity in the internal pallidum induced by subthalamic inactivation in the MPTP model of parkinsonism. J. Neurophysiol. 72, (1994). 26. Alberts,W.W., Wright,E.W., & Feinstein,B. Cortical potentials and Parkinsonian tremor. Nature 221, (1969). 27. Volkmann,J. et al. Central motor loop oscillations in parkinsonian resting tremor revealed by magnetoencephalography. Neurology 46, (1996). 28. Timmermann,L. et al. The cerebral oscillatory network of parkinsonian resting tremor. Brain 126, (2003). 29. Lenz,F.A. et al. Single unit analysis of the human ventral thalamic nuclear group: correlation of thalamic "tremor cells" with the 3-6 Hz component of Parkinsonian tremor. J. Neurosci. 8, (1988). 30. Hayase,N., Miyashita,N., Endo,K., & Narabayashi,H. Neuronal activity in GP and Vim of parkinsonian patients and clinical changes of tremor through surgical interventions. Stereotact. Funct. Neurosurg. 71, (1998). 31. Magnin,M., Morel,A., & Jeanmonod,D. Single-unit analysis of the pallidum, thalamus and subthalamic nucleus in parkinsonian patients. Neuroscience 96, (2000). 32. Rivlin-Etzion,M. et al. Basal ganglia oscillations and pathophysiology of movement disorders. Curr. Opin. Neurobiol. 16, (2006). 78

81 Results IV 33. Teravainen,H., Evarts,E., & Calne,D. Effects of kinesthetic input on parkinsonian tremor in Advances in Neurology, Vol. 24 (eds. Poirier,L.J., Sourkes,T.L. & Bedard,P.J.) (Raven Press, New York, 1979). 34. Teravainen,H. Oscillatory electromyographic responses to limb displacement in Parkinsonism.. J. Neurol. Neurosurg. Psychiatry 43, (1980). 35. Arkadir,D., Morris,G., Vaadia,E., & Bergman,H. Independent coding of movement direction and reward prediction by single pallidal neurons. J. Neurosci. 24, (2004). 36. Morris,G., Arkadir,D., Nevet,A., Vaadia,E., & Bergman,H. Coincident but distinct messages of midbrain dopamine and striatal tonically active neurons. Neuron 43, (2004). 37. Elias,S. et al. Statistical properties of pauses of the high-frequency discharge neurons in the external segment of the globus pallidus. J Neurosci. 27, (2007). 38. Rivlin-Etzion,M. et al. Low-pass filter properties of basal ganglia cortical muscle loops in the normal and MPTP primate model of parkinsonism. J. Neurosci. 28, (2008). 39. Heimer,G., Bar-Gad,I., Goldberg,J.A., & Bergman,H. Dopamine replacement therapy reverses abnormal synchronization of pallidal neurons in the 1-methyl-4- phenyl-1,2,3,6-tetrahydropyridine primate model of Parkinsonism. J Neurosci. 22, (2002). 40. Schiff,S.J., Dunagan,B.K., & Worth,R.M. Failure of single-unit neuronal activity to differentiate globus pallidus internus and externus in Parkinson disease. J. Neurosurg. 97, (2002). 41. Benazzouz,A. et al. Riluzole prevents MPTP-induced parkinsonism in the rhesus monkey: a pilot study.. Eur. J. Pharmacol. 284, (1995). 42. Rivlin-Etzion,M., Ritov,Y., Heimer,G., Bergman,H., & Bar-Gad,I. Local shuffling of spike trains boosts the accuracy of spike train spectral analysis. J. Neurophysiol. 95, (2006). 43. Goldberg,J.A. et al. Enhanced synchrony among primary motor cortex neurons in the 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine primate model of Parkinson's disease. J Neurosci. 22, (2002). 44. Bar-Gad,I., Ritov,Y., Vaadia,E., & Bergman,H. Failure in identification of overlapping spikes from multiple neuron activity causes artificial correlations. J Neurosci. Methods 107, 1-13 (2001). 45. Bloomfield,P. Fourier Analysis of Time Series: An Introduction(Wiley, New York, 1976). 46. Brillinger,D.R. Time Series: Data Analysis and Theory(Holden-Day, San Francisco, 1981). 47. Song,D.D. & Haber,S.N. Striatal responses to partial dopaminergic lesion: evidence for compensatory sprouting. J Neurosci. 20, (2000). 48. Francois,C. et al. Behavioural disorders induced by external globus pallidus dysfunction in primates II. Anatomical study. Brain 127, (2004). 79

82 Discussion DISCUSSION In this dissertation I explored the relationship between changes in neural activity in the Cortico-Baso-Thalamo-Cortical loop and Parkinsonian symptoms in the MPTP treated monkey. In Parkinson s disease (PD), the globus pallidus (GP) demonstrates an oscillatory and bursty pattern, but despite the fact that these oscillations often share a common frequency with the tremor, a robust correlation between the two has never been established. Another recognized phenomenon in PD is the increase in the inter-neuronal correlation in the GP. However, the effect of this synchronization on the Cortico-Baso- Thalamo-Cortical loop has not yet been studied. Oscillations and synchronization are often coupled phenomena; however, they could be independent. Thus, the role of both neural oscillations and synchronizations is essential for our understanding of the pathophysiology of PD, and can lead new therapeutic approaches to this disease. In order to address these aspects of neuronal activity, I used multiple electrodes and conducted the first experiment in which recordings of globus pallidus (GP) and arm related primary motor cortex (MI) were made simultaneously. I used electrical stimulation to mimic the Parkinsonian neuronal oscillations in both structures, while recording the evoked neuronal responses in the stimulated and the other structure, as well as arm movements (using an accelerometer). The experiments were carried on two vervet monkeys, which tend to tremor in the Parkinsonian state. In addition, I examined the spontaneous activity of several MPTP treated monkeys, tremulous as well as nontremulous. This work has two innovative characteristics: first, it is based on data collected from six MPTP treated monkeys, a much larger sample than is customarily used in experiments in the field. Second, I used a new method, the shuffling method, which I developed to obtain an accurate measure of oscillatory neuronal activity. Previous methods of neural oscillation studies have ignored the refractory period which distorts the spectrum of the spike train and therefore could lead to a biased estimation of the oscillations. Several novel discoveries were made in my work. First, in both the normal and the MPTP states, the basal ganglia (BG)-MI-muscle circuit demonstrates low-pass filter properties. This finding was based on electrical stimulations and suggests that pallidal oscillations are not transformed further in the cortex to generate Parkinsonian tremor. The spontaneous recordings supported this finding, by showing that oscillatory activity 80

83 Discussion appears in the GP of all monkeys regardless of the emergence of tremor. However, the distribution of the oscillation frequencies varied according to the appearance of tremor, with 5 Hz activity found in all monkeys, and 10 Hz oscillations detected only in tremulous monkeys. This unexpected result suggests that the higher frequency oscillations, rather than the tremor-frequency oscillatory activity, are related to the tremor. Moreover, GP 10 Hz oscillations were reflected in the MI of one tremulous monkey, but not the other, which casts doubt on the role of the Cortico-Baso-Thalamo- Cortical loop in generating the Parkinsonian tremor. Consequently, I suggest that the basal ganglia projections to brainstem motor centers, which have usually been ignored, may play a role in transferring the GP oscillations and generating the tremor. Finally, the functional connectivity between the MI and the GP are greatly enhanced in the MPTP state, with one structure (MI or GP) responding to micro stimulations in the other structure (GP or MI) in the Parkinsonian but not the normal state. These two findings can shed light on the conflicting co-existence of hypo and hyper motoric deficits in PD. Innovative methodology A prerequisite to my study was finding a reliable method to detect oscillatory neural activity. Because the spectrum of spike trains is distorted due to the refractory period of neurons, traditional methods could not be used. The effect of the refractory period is crucial in neurons with high discharge rates (e.g. pallidal neurons), since the refractoriness is more dominant in their firing pattern, resulting in a large magnitude spectral distortion. This phenomenon was previously reported (Edwards et al., 1993;Bair et al., 1994a;Franklin and Bair, 1995) but the difficulties in identifying oscillatory activity in the neuronal spectrum were never addressed. I developed a method to overcome these difficulties that uses inter spike interval (ISI) shuffling of the spike train. The shuffled train has exactly the same first-order properties of refractoriness as the original spike train, but the periodic activity is lost in the shuffling process. By division of the original by the shuffled spectra the periodic properties are revealed. The division process ensures equal distribution of power across all frequencies, providing reliable ways to establish confidence limits. The shuffling method enables the detection of auto- and cross-oscillations in spike trains that could have been ignored otherwise. 81

84 Discussion Low-pass filter properties of the BG-cortical-muscle network To understand the role of oscillatory bursts encountered in the Parkinsonian brain (Heimer et al., 2006b;Weinberger et al., 2006b) and the relationship between the GP, MI and the periphery in the frequency domain, I mimicked the Parkinsonian oscillatory activity using a stimulation pattern that contained 35 ms bursts delivered at different (1-15 Hz) frequencies in one of the neuronal structures. The cortex-bg-periphery loops exhibited low-pass filter properties to the microstimulation pattern. MI neurons attenuate their response to MI micro-stimulation as the frequency of stimulation increases. Moreover, MI neurons respond to the 1 and 2 Hz GP stimulations, but not to the 5 Hz and higher frequencies. Finally, low-pass filtering was also found between MI and the periphery level, where muscle activation evoked by MI micro-stimulation was markedly attenuated at frequencies higher than 5 Hz. I did not test the exact cut-off frequency of MI or muscle activity, and it could change with the stimulation burst parameters (e.g., burst duration and frequency and number of the pulses/burst (McIntyre and Grill, 2002)). However, the low-pass properties of the pathways connecting GP to MI to muscles suggest that Parkinsonian tremor is not driven by basal ganglia high frequency oscillations via the Cortico-Baso- Thalamo-Cortical loop. 10 Hz pallidal oscillations and the tremor Neuronal oscillations were detected using the shuffling method in data collected from the GP of six MPTP treated monkeys - three tremulous African green monkeys and three non-tremulous macaques. The appearance of the 5 Hz neuronal oscillations in the GP of all monkeys suggests that these GP low frequency oscillations, although sharing a similar frequency with the Parkinsonian tremor, do not generate the tremor, and are not driven by it. On the contrary, high frequency pallidal oscillations were found only in tremulous monkeys, which might indicate a relation between 10 Hz GP oscillations and the tremor. Interestingly, these high frequency oscillations are probably filtered out at the cortex or more upstream (e.g. the thalamus). The cortical activity was recorded in two vervet monkeys which developed tremor as a result of the MPTP treatment. However, the 10 Hz neuronal oscillations were detected in the MI of only one monkey, despite the 82

85 Discussion fact that both monkeys demonstrated a similar fraction of GP oscillations, which were comparably bimodally distributed around 5 and 10 Hz. These filtration properties are in line with the low-pass filtration revealed by the stimulation experiments, and suggest that GP high frequency oscillations are transformed via the brainstem motor centers and generate the Parkinsonian tremor in a 2:1 or other complex relationship. It is not known why the 5 Hz pallidal oscillations do not result in a 2.5 Hz tremor as well through the same pathway, but a reasonable explanation is that most of the synchronous oscillatory activity in the GP, if present, is congregated around 10 Hz and not around the lower frequencies (see Results IV, as well as (Raz et al., 2000a;Heimer et al., 2006a)). Nevertheless, the evidence presented here does not rule out the possibility that the 10 Hz oscillatory activity in the GP is merely a reflection of the peripheral tremor and is generated by a sensory input: agonist muscles that have opposite phases of 5 Hz tremor can sum up at the GP level and generate 10 Hz activity. Functional connectivity between MI and GP The functional micro-connectivity between MI and both segments of the GP is weak in the normal state. A comparison of my findings and previous reports of extensive GP activation by macro-stimulation of MI (Nambu et al., 2000) suggests that in the normal monkey the MI to GP functional connectivity is either very specific or that converging inputs from many MI areas are needed to activate GP neurons. The MI-GP reciprocal connections are strengthened in the MPTP state, with ~15% of recorded neurons in each structure responding to the micro-stimulation in the other structure. This enhancement in connectivity may be caused by the reduction in the specificity of pallidal neurons (Filion et al., 1988), or by the increased synchronization of BG neurons in the dopamine depleted state (Bar-Gad et al., 2003;Strafella et al., 2005;Heimer et al., 2006e), leading to widespread activation of the local effects of micro-stimulation. My working hypothesis was that this enhancement in functional connectivity that arises in the Parkinsonian state between the BG and the cortex plays a major role in generating the main core negative Parkinsonian symptoms akinesia and bradykinesia. The basal ganglia are related, among other functions, to motor activation. Specifically, they are thought to perform the action selection part of the movement (Mink, 1996d). In the DA depleted state the enhanced correlation in the loop prevents the selection of an action and therefore leads to a no movement situation. Moreover, the activity in the 83

86 Discussion motor cortex also becomes synchronized, making it difficult to send motor commands to specific muscles. Possibly, due to the malfunction of the Cortico-Baso-Thalamo-Cortical loop, the direct projections from the BG through the brainstem become more dominant and enable the emergence of the tremor. The functional connectivity within the Cortico-Baso-Thalamo-Cortical circuit was enhanced in both monkeys, including monkey W that died 7 days after last MPTP treatment, and before it demonstrated any significant rigidity of the muscles. This fact provides two important clues regarding synchronization. First, unlike the neuronal oscillations, which are occasionally reported to appear gradually as the disease worsens (Leblois et al., 2007;Degos et al., 2008), the synchronization in the Cortico-Baso- Thalamo-Cortical loop seems inseparable from PD. Second, rigidity is not a direct result of this synchronization. Further investigation is needed in order to determine whether the rigidity originates from GP low frequency oscillations, or is transformed like the tremor through the brainstem projections. DBS mechanisms The low-pass properties of the GP-MI-periphery axis shed new light on the ongoing controversy regarding the mechanisms underlying high frequency (~130 Hz) DBS in the treatment of advanced PD. There is considerable debate as to whether DBS mimics lesion and inactivates its targets (Wu et al., 2001;Filali et al., 2004;Maltete et al., 2007), or whether it drives them continuously at short latencies and higher frequencies (Hashimoto et al., 2003;Garcia et al., 2005). Since the bursts I used in my study contained pulses given at 200 Hz, the 10 and 15 Hz frequency stimulations can be seen as a fragmented DBS stimulation pattern, enabling us to examine the effect of DBS without the distortion of stimulation artifacts. The results suggest that the continuous DBS mechanism is similarly filtered in the cortex or in prior levels. Although filtered, DBS indirectly alleviates the Parkinsonian symptoms, probably by halting the abnormal activity of the BG (Benabid, 2003). Summary My research suggests a novel approach to the generation of PD symptoms, which differentiates akinesia-bradykinesia hypo-motoric symptoms from hyper-motoric tremor. I hypothesized that these distinct symptoms may result from different neuronal features; 84

87 Discussion in particular, the main core negative Parkinsonian symptoms may be related to the enhanced synchronization in the Cortic-Baso-Thalamo-Cortical loop, and the tremor may be related to the high frequency GP oscillations, possibly transformed via the brainstem motor centers. These neuronal oscillations, as well as periodic muscle activity, may emerge as part of a compensatory mechanism that attempts to counterbalance for the akinesia and bradykinesia. My findings indicate that future research on PD should be directed towards brainstem motor centers (e.g. the pedunculo pontine nucleus (PPN)), which may reveal the role of basal ganglia projections to the brainstem in health and disease. 85

88 Bibliography BIBLIOGRAPHY Aizman O, Brismar H, Uhlen P, Zettergren E, Levey AI, Forssberg H, Greengard P, Aperia A (2000) Anatomical and physiological evidence for D1 and D2 dopamine receptor colocalization in neostriatal neurons. Nat Neurosci 3: Alberts WW, Wright EW, Feinstein B (1969) Cortical potentials and Parkinsonian tremor. Nature 221: Albin RL, Young AB, Penney JB (1989) The functional anatomy of basal ganglia disorders. Trends Neurosci 12: Arkadir D, Morris G, Vaadia E, Bergman H (2004) Independent coding of movement direction and reward prediction by single pallidal neurons. J Neurosci 24: Bair W, Koch C, Newsome W, Britten K (1994b) Power spectrum analysis of bursting cells in area MT in the behaving monkey. J Neurosci 14: Bair W, Koch C, Newsome W, Britten K (1994a) Power spectrum analysis of bursting cells in area MT in the behaving monkey. J Neurosci 14: Bar-Gad I, Havazelet Heimer G, Goldberg JA, Ruppin E, Bergman H (2000a) Reinforcement driven dimensionality reduction - a model for information processing in the basal ganglia. Journal of Basic & Clinical Physiology & Pharmacology 11: Bar-Gad I, Morris G, Bergman H (2003) Information processing, dimensionality reduction and reinforcement learning in the basal ganglia. Prog Neurobiol 71: Bar-Gad I, Ritov Y, Bergman H (2000b) The neuronal refractory period causes a shortterm peak in the autocorrelation function. Bar-Gad I, Ritov Y, Vaadia E, Bergman H (2001) Failure in identification of overlapping spikes from multiple neuron activity causes artificial correlations. J Neurosci Methods 107:1-13. Benabid AL (2003) Deep brain stimulation for Parkinson's disease. Curr Opin Neurobiol 13: Benazzouz A, Boraud T, Dubedat P, Boireau A, Stutzmann JM, Gross C (1995) Riluzole prevents MPTP-induced parkinsonism in the rhesus monkey: a pilot study. Eur J Pharmacol 284: Bergman H, Feingold A, Nini A, Raz A, Slovin H, Abeles M, Vaadia E (1998a) Physiological aspects of information processing in the basal ganglia of normal and parkinsonian primates. Trends Neurosci 21: Bergman H, Raz A, Feingold A, Nini A, Nelken I, Hansel D, Ben-Pazi H, Reches A (1998b) Physiology of MPTP tremor. Mov Disord 13 Suppl 3:

89 Bibliography Bergman H, Wichmann T, Karmon B, DeLong MR (1994) The primate subthalamic nucleus. II. Neuronal activity in the MPTP model of Parkinsonism. J Neurophysiol 72: Bevan MD, Booth PA, Eaton SA, Bolam JP (1998) Selective innervation of neostriatal interneurons by a subclass of neuron in the globus pallidus of the rat. J Neurosci 18: Bloomfield P (1976) Fourier Analysis of Time Series: An Introduction. New York: Wiley. Bolam JP, Hanley JJ, Booth PA, Bevan MD (2000) Synaptic organisation of the basal ganglia. J Anat 196: Boraud T, Bezard E, Guehl D, Bioulac B, Gross C (1998) Effects of L-DOPA on neuronal activity of the globus pallidus externalis (GPe) and globus pallidus internalis (GPi) in the MPTP-treated monkey. Brain Res 787: Brillinger DR (1981) Time Series: Data Analysis and Theory. San Francisco: Holden- Day. Brown P (2003) Oscillatory nature of human basal ganglia activity: relationship to the pathophysiology of Parkinson's disease. Mov Disord 18: Centonze D, Gubellini P, Picconi B, Calabresi P, Giacomini P, Bernardi G (1999) Unilateral dopamine denervation blocks corticostriatal LTP. J Neurophysiol 82: Crossman AR, Clarke CE, Boyce S, Robertson RG, Sambrook MA (1987) MPTPinduced parkinsonism in the monkey: neurochemical pathology, complications of treatment and pathophysiological mechanisms. Can J Neurol Sci 14: Degos B, Deniau JM, Chavez M, Maurice N (2008) Chronic but not Acute Dopaminergic Transmission Interruption Promotes a Progressive Increase in Cortical Beta Frequency Synchronization: Relationships to Vigilance State and Akinesia. Cereb Cortex. DeLong MR (1971) Activity of pallidal neurons during movement. J Neurophysiol 34: DeLong MR (1990) Primate models of movement disorders of basal ganglia origin. Trends Neurosci 13: Doudet DJ, Gross C, Arluison M, Bioulac B (1990) Modifications of precentral cortex discharge and EMG activity in monkeys with MPTP-induced lesions of DA nigral neurons. Exp Brain Res 80: Drouot X, Oshino S, Jarraya B, Besret L, Kishima H, Remy P, Dauguet J, Lefaucheur JP, Dolle F, Conde F, Bottlaender M, Peschanski M, Keravel Y, Hantraye P, Palfi S (2004) Functional recovery in a primate model of Parkinson's disease following motor cortex stimulation. Neuron 44:

90 Bibliography Edwards B, Wakefield G, Powers NL (1993) The spectral shaping of neural discharges by refractory effects. J Acoust Soc Am 93(6): Fahn S, Greene PE, Ford B, Bressman SB (1998) Handbook of Movement Disorders. Philadelphia: Blackwell Science. Filali M, Hutchison WD, Palter VN, Lozano AM, Dostrovsky JO (2004) Stimulationinduced inhibition of neuronal firing in human subthalamic nucleus. Exp Brain Res. Filion M, Tremblay L (1991) Abnormal spontaneous activity of globus pallidus neurons in monkeys with MPTP-induced Parkinsonism. Brain Res 547: Filion M, Tremblay L, Bedard PJ (1988) Abnormal influences of passive limb movement on the activity of globus pallidus neurons in parkinsonian monkeys. Brain Res 444: Franklin J, Bair W (1995) The Effect of a Refractory Period on the Power Spectrum of Neuronal Discharge. SIAM Journal on Applied Mathematics 55, No. 4: Garcia L, D'Alessandro G, Bioulac B, Hammond C (2005) High-frequency stimulation in Parkinson's disease: more or less? Trends Neurosci 28: Gauss R, Seifert R (2000) Pacemaker oscillations in heart and brain: a key role for hyperpolarization-activated cation channels. Chronobiol Int 17: Gdowski MJ, Miller LE, Parrish T, Nenonene EK, Houk JC (2001) Context dependency in the globus pallidus internal segment during targeted arm movements. J Neurophysiol 85: Gerfen CR (2004) Basal Ganglia. In: The Rat Nervous System (Paxinos G, ed), pp London: Elsevier Academic Press. Gerfen CR, Engber TM, Mahan LC, Susel Z, Chase TN, Monsma Jr. FJ, Sibley DR (1990) D1 and D2 dopamine receptor-regulated gene expression of striatonigral and striatopallidal neurons. Science 250: Gerfen CR, Wilson CJ (1996) The basal ganglia. In: Handbook of Chemical Neuroanatomy, Vol 12: Integrated Systems of the CNS, Part III (Swanson LW, Bjorklund A, Hokfelt T, eds), pp Elsevier Science. Goldberg JA, Boraud T, Maraton S, Haber SN, Vaadia E, Bergman H (2002d) Enhanced synchrony among primary motor cortex neurons in the 1-methyl-4-phenyl-1,2,3,6- tetrahydropyridine primate model of Parkinson's disease. J Neurosci 22: Goldberg JA, Boraud T, Maraton S, Haber SN, Vaadia E, Bergman H (2002c) Enhanced synchrony among primary motor cortex neurons in the 1-methyl-4-phenyl-1,2,3,6- tetrahydropyridine primate model of Parkinson's disease. J Neurosci 22: Goldberg JA, Boraud T, Maraton S, Haber SN, Vaadia E, Bergman H (2002b) Enhanced synchrony among primary motor cortex neurons in the 1-methyl-4-phenyl-1,2,3,6- tetrahydropyridine primate model of Parkinson's disease. J Neurosci 22:

91 Bibliography Goldberg JA, Boraud T, Maraton S, Haber SN, Vaadia E, Bergman H (2002a) Enhanced synchrony among primary motor cortex neurons in the 1-methyl-4-phenyl-1,2,3,6- tetrahydropyridine primate model of Parkinson's disease. J Neurosci 22: Goldberg JA, Rokni U, Boraud T, Vaadia E, Bergman H (2004) Spike synchronization in the cortex/basal-ganglia networks of Parkinsonian primates reflects global dynamics of the local field potentials. J Neurosci 24: Gray CM (1994) Synchronous oscillations in neuronal systems: Mechanisms and functions. J of Computational Neuroscience 1: Haber SN, Gdowski MJ (2004) The Basal Ganglia. In: The Human Nervous System (Paxinos G, Mai LM, eds), pp London : Elsevier Academic Press. Handel A, Glimcher PW (2000) Contextual modulation of substantia nigra pars reticulata neurons. J Neurophysiol 83: Hashimoto T, Elder CM, Okun MS, Patrick SK, Vitek JL (2003) Stimulation of the subthalamic nucleus changes the firing pattern of pallidal neurons. J Neurosci 23: Heimer G, Rivlin-Etzion M, Bar-Gad I, Goldberg JA, Haber SN, Bergman H (2006e) Dopamine replacement therapy does not restore the full spectrum of normal pallidal activity in the 1-methyl-4-phenyl-1,2,3,6-tetra-hydropyridine primate model of Parkinsonism. J Neurosci 26: Heimer G, Rivlin-Etzion M, Bar-Gad I, Goldberg JA, Haber SN, Bergman H (2006a) Dopamine replacement therapy does not restore the full spectrum of normal pallidal activity in the 1-methyl-4-phenyl-1,2,3,6-tetra-hydropyridine primate model of Parkinsonism. J Neurosci 26: Heimer G, Rivlin-Etzion M, Bar-Gad I, Goldberg JA, Haber SN, Bergman H (2006b) Dopamine replacement therapy does not restore the full spectrum of normal pallidal activity in the 1-methyl-4-phenyl-1,2,3,6-tetra-hydropyridine primate model of Parkinsonism. J Neurosci 26: Heimer G, Rivlin-Etzion M, Bar-Gad I, Goldberg JA, Haber SN, Bergman H (2006c) Dopamine replacement therapy does not restore the full spectrum of normal pallidal activity in the 1-methyl-4-phenyl-1,2,3,6-tetra-hydropyridine primate model of Parkinsonism. J Neurosci 26: Heimer G, Rivlin-Etzion M, Bar-Gad I, Goldberg JA, Haber SN, Bergman H (2006d) Dopamine replacement therapy does not restore the full spectrum of normal pallidal activity in the 1-methyl-4-phenyl-1,2,3,6-tetra-hydropyridine primate model of Parkinsonism. J Neurosci 26: Hirsch EC, Mouatt A, Faucheux B, Bonnet AM, Javoy Agid F, Graybiel AM, Agid Y (1992) Dopamine, tremor, and Parkinson's disease [letter]. Lancet 340:

92 Bibliography Hurtado JM, Gray CM, Tamas LB, Sigvardt KA (1999a) Dynamics of tremor-related oscillations in the human globus pallidus: a single case study. Proc Natl Acad Sci U S A 96: Hurtado JM, Gray CM, Tamas LB, Sigvardt KA (1999b) Dynamics of tremor-related oscillations in the human globus pallidus: a single case study. Proc Natl Acad Sci U S A 96: Hutchison WD, Lozano AM, Tasker RR, Lang AE, Dostrovsky JO (1997) Identification and characterization of neurons with tremor- frequency activity in human globus pallidus. Exp Brain Res 113: Jaeger D, Gilman S, Aldridge JW (1995) Neuronal activity in the striatum and pallidum of primates related to the execution of externally cued reaching movements. Brain Res 694: Jellinger K (1987) The pathology of parkinsonism. In: Movement Disorders 2 (Marsden CD, Fahn S, eds), pp London: Butterworths. Jellinger KA (1999) Post mortem studies in Parkinson's disease--is it possible to detect brain areas for specific symptoms? J Neural Transm Suppl 56:1-29. Kita H, Tokuno H, Nambu A (1999) Monkey globus pallidus external segment neurons projecting to the neostriatum. Neuroreport 10: Langston JW (1987) MPTP: The promise of a new neurotoxin. In: Movement Disorders 2 (Marsden CD, Fahn S, eds), pp London: Butterworths & Co. Leblois A, Meissner W, Bioulac B, Gross CE, Hansel D, Boraud T (2007) Late emergence of synchronized oscillatory activity in the pallidum during progressive parkinsonism. Eur J Neurosci 26: Lemstra AW, Verhagen ML, Lee JI, Dougherty PM, Lenz FA (1999a) Tremor-frequency (3-6 Hz) activity in the sensorimotor arm representation of the internal segment of the globus pallidus in patients with Parkinson's disease. Neurosci Lett 267: Lemstra AW, Verhagen ML, Lee JI, Dougherty PM, Lenz FA (1999b) Tremorfrequency (3-6 Hz) activity in the sensorimotor arm representation of the internal segment of the globus pallidus in patients with Parkinson's disease. Neurosci Lett 267: Lenz FA, Tasker RR, Kwan HC, Schnider S, Kwong R, Murayama Y, Dostrovsky JO, Murphy JT (1988) Single unit analysis of the human ventral thalamic nuclear group: correlation of thalamic "tremor cells" with the 3-6 Hz component of Parkinsonian tremor. J Neurosci 8: Levy R, Hazrati LN, Herrero MT, Vila M, Hassani OK, Mouroux M, Ruberg M, ASENSI H, Agid Y, Feger J, Obeso JA, Parent A, Hirsch EC (1997) Re-evaluation of the functional anatomy of the basal ganglia in normal and Parkinsonian states. Neuroscience 76:

93 Bibliography Levy R, Hutchison WD, Lozano AM, Dostrovsky JO (2000) High-frequency synchronization of neuronal activity in the subthalamic nucleus of parkinsonian patients with limb tremor. J Neurosci 20: Levy R, Hutchison WD, Lozano AM, Dostrovsky JO (2002) Synchronized neuronal discharge in the basal ganglia of Parkinsonian patients is limited to oscillatory activity. J Neurosci 22: Maltete D, Jodoin N, Karachi C, Houeto JL, Navarro S, Cornu P, Agid Y, Welter ML (2007) Subthalamic stimulation and neuronal activity in the substantia nigra in Parkinson's disease. J Neurophysiol 97: Mandir AS, Rowland LH, Dougherty PM, Lenz FA (1997) Microelectrode recording and stimulation techniques during stereotactic procedures in the thalamus and pallidum. Adv Neurol 74: McCormick DA (2002) Cortical and subcortical generators of normal and abnormal rhythmicity. Int Rev Neurobiol 49: McIntyre CC, Grill WM (2002) Extracellular stimulation of central neurons: influence of stimulus waveform and frequency on neuronal output. J Neurophysiol 88: Melamed E, Ziv I, Djaldetti R (2007) Management of motor complications in advanced Parkinson's disease. Mov Disord 22 Suppl 17:S379-S384. Middleton FA, Strick PL (2000) Basal ganglia output and cognition: evidence from anatomical, behavioral, and clinical studies. Brain Cogn 42: Miller WC, DeLong MR (1987) Altered tonic activity of neurons in the globus pallidus and subthalamic nucleus in the primate MPTP model of parkinsonism. In: The Basal Ganglia II (Carpenter MB, Jayaraman A, eds), pp New York: Plenum Press. Mink JW (1996b) The basal ganglia: focused selection and inhibition of competing motor programs. Prog Neurobiol 50: Mink JW (1996c) The basal ganglia: focused selection and inhibition of competing motor programs. Prog Neurobiol 50: Mink JW (1996d) The basal ganglia: focused selection and inhibition of competing motor programs. Prog Neurobiol 50: Mink JW (1996a) The basal ganglia: focused selection and inhibition of competing motor programs. Prog Neurobiol 50: Morris G, Arkadir D, Nevet A, Vaadia E, Bergman H (2004) Coincident but distinct messages of midbrain dopamine and striatal tonically active neurons. Neuron 43: Nambu A, Tokuno H, Hamada I, Kita H, Imanishi M, Akazawa T, Ikeuchi Y, Hasegawa N (2000) Excitatory cortical inputs to pallidal neurons via the subthalamic nucleus in the monkey. J Neurophysiol 84:

94 Bibliography Nini A, Feingold A, Slovin H, Bergman H (1995) Neurons in the globus pallidus do not show correlated activity in the normal monkey, but phase-locked oscillations appear in the MPTP model of Parkinsonism. J Neurophysiol 74: Parent A, Hazrati LN (1995) Functional anatomy of the basal ganglia. I. The corticobasal ganglia-thalamo-cortical loop. Brain Res Rev 20: Parent M, Levesque M, Parent A (1999) The pallidofugal projection system in primates: evidence for neurons branching ipsilaterally and contralaterally to the thalamus and brainstem. J Chem Neuroanat 16: Parkinson J (1817) An essay on the shaking palsy. London: Sherwood, Neely and Jones. Paulus W, Jellinger K (1991) The neuropathologic basis of different clinical subgroups of Parkinson's disease. J Neuropathol Exp Neurol 50: Perlmutter JS, Mink JW (2006) Deep Brain Stimulation. Annu Rev Neurosci. Ranck JB (1975) Which elements are excited in electrical stimulation of mammalian central nervous system: A review. Brain Res 98: Raz A, Frechter-Mazar V, Feingold A, Abeles M, Vaadia E, Bergman H (2001) Activity of pallidal and striatal tonically active neurons is correlated in MPTP-treated monkeys but not in normal monkeys. J Neurosci 21:RC128:1-5. Raz A, Vaadia E, Bergman H (2000c) Firing patterns and correlations of spontaneous discharge of pallidal neurons in the normal and the tremulous 1-methyl-4-phenyl-1,2,3,6- tetrahydropyridine vervet model of Parkinsonism. J Neurosci 20: Raz A, Vaadia E, Bergman H (2000b) Firing patterns and correlations of spontaneous discharge of pallidal neurons in the normal and the tremulous 1-methyl-4-phenyl-1,2,3,6- tetrahydropyridine vervet model of Parkinsonism. J Neurosci 20: Raz A, Vaadia E, Bergman H (2000a) Firing patterns and correlations of spontaneous discharge of pallidal neurons in the normal and the tremulous 1-methyl-4-phenyl-1,2,3,6- tetrahydropyridine vervet model of Parkinsonism. J Neurosci 20: Raz A, Vaadia E, Bergman H (2000e) Firing patterns and correlations of spontaneous discharge of pallidal neurons in the normal and the tremulous 1-methyl-4-phenyl-1,2,3,6- tetrahydropyridine vervet model of Parkinsonism. J Neurosci 20: Raz A, Vaadia E, Bergman H (2000d) Firing patterns and correlations of spontaneous discharge of pallidal neurons in the normal and the tremulous 1-methyl-4-phenyl-1,2,3,6- tetrahydropyridine vervet model of Parkinsonism. J Neurosci 20: Reynolds JN, Hyland BI, Wickens JR (2001) A cellular mechanism of reward-related learning. Nature 413: Rivlin-Etzion M, Marmor O, Saban G, Rosin B, Haber SN, Vaadia E, Prut Y, Bergman H (2008) Low-pass filter properties of basal ganglia cortical muscle loops in the normal and MPTP primate model of parkinsonism. J Neurosci 28:

95 Bibliography Rodrigues JP, Walters SE, Watson P, Stell R, Mastaglia FL (2007) Globus pallidus stimulation improves both motor and nonmotor aspects of quality of life in advanced Parkinson's disease. Mov Disord 22: Sato M, Hikosaka O (2002) Role of primate substantia nigra pars reticulata in rewardoriented saccadic eye movement. J Neurosci 22: Schrag A, Quinn N (2000) Dyskinesias and motor fluctuations in Parkinson's disease. A community-based study. Brain 123 ( Pt 11): Shen W, Flajolet M, Greengard P, Surmeier DJ (2008) Dichotomous dopaminergic control of striatal synaptic plasticity. Science 321: Strafella AP, Ko JH, Grant J, Fraraccio M, Monchi O (2005) Corticostriatal functional interactions in Parkinson's disease: a rtms/[11c]raclopride PET study. Eur J Neurosci 22: Surmeier DJ, Ding J, Day M, Wang Z, Shen W (2007) D1 and D2 dopamine-receptor modulation of striatal glutamatergic signaling in striatal medium spiny neurons. Trends Neurosci 30: Turner RS, Anderson ME (1997) Pallidal discharge related to the kinematics of reaching movements in two dimensions. J Neurophysiol 77: Valldeoriola F, Morsi O, Tolosa E, Rumia J, Marti MJ, Martinez-Martin P (2007) Prospective comparative study on cost-effectiveness of subthalamic stimulation and best medical treatment in advanced Parkinson's disease. Mov Disord 22: Van Den Eeden SK, Tanner CM, Bernstein AL, Fross RD, Leimpeter A, Bloch DA, Nelson LM (2003) Incidence of Parkinson's disease: variation by age, gender, and race/ethnicity. Am J Epidemiol 157: Volkmann J, Joliot M, Mogilner A, Ioannides AA, Lado F, Fazzini E, Ribary U, Llinas R (1996) Central motor loop oscillations in parkinsonian resting tremor revealed by magnetoencephalography. Neurology 46: Watts RL, Mandir AS (1992) The role of motor cortex in the pathophysiology of voluntary movement deficits associated with parkinsonism. Neurologic clinics 10: Weinberger M, Mahant N, Hutchison WD, Lozano AM, Moro E, Hodaie M, Lang AE, Dostrovsky JO (2006b) Beta oscillatory activity in the subthalamic nucleus and its relation to dopaminergic response in Parkinson's disease. J Neurophysiol 96: Weinberger M, Mahant N, Hutchison WD, Lozano AM, Moro E, Hodaie M, Lang AE, Dostrovsky JO (2006a) Beta oscillatory activity in the subthalamic nucleus and its relation to dopaminergic response in Parkinson's disease. J Neurophysiol 96: Wichmann T, Bergman H, DeLong MR (1994) The primate subthalamic nucleus. III. Changes in motor behavior and neuronal activity in the internal pallidum induced by 93

96 Bibliography subthalamic inactivation in the MPTP model of parkinsonism. J Neurophysiol 72: Wichmann T, Bergman H, Starr PA, Subramanian T, Watts RL, DeLong MR (1999) Comparison of MPTP-induced changes in spontaneous neuronal discharge in the internal pallidal segment and in the substantia nigra pars reticulata in primates. Exp Brain Res 125: Wu YR, Levy R, Ashby P, Tasker RR, Dostrovsky JO (2001) Does stimulation of the GPi control dyskinesia by activating inhibitory axons? Mov Disord 16:

97 Appendix I APPENDIX I. OSCILLATORY ACTIVITY FOLLOWING DOPAMINE REPLACEMENT THERAPY 95

98 Appendix I The Journal of Neuroscience, August 2, (31): Neurobiology of Disease Dopamine Replacement Therapy Does Not Restore the Full Spectrum of Normal Pallidal Activity in the 1-Methyl-4- Phenyl-1,2,3,6-Tetra-Hydropyridine Primate Model of Parkinsonism Gali Heimer, 1,4 Michal Rivlin-Etzion, 1,2 Izhar Bar-Gad, 5 Joshua A. Goldberg, 6 Suzanne N. Haber, 7 and Hagai Bergman 1,2,3 1 Department of Physiology, 2 Interdisciplinary Center for Neural Computation, and 3 Eric Roland Center for Neurodegenerative Diseases, The Hebrew University Hadassah Medical School, Jerusalem, Israel, 91120, 4 Department of Pediatrics, Hadassah Hebrew University Medical Center, Jerusalem, Israel, 91120, 5 Gonda Brain Research Center, Bar-Ilan University, Ramat-Gan, Israel, 52900, 6 Department of Biology, University of Texas at San Antonio, San Antonio, Texas 78249, and 7 Department of Pharmacology and Physiology, University of Rochester, Rochester, New York Current physiological studies emphasize the role of neuronal oscillations and synchronization in the pathophysiology of Parkinson s disease; however, little is known about their specific roles in the neuronal substrate of dopamine replacement therapy (DRT). We investigated oscillatory activity and correlations throughout the different states of levodopa-naive parkinsonism as well as Off On and dyskinetic states of DRT in the external globus pallidum (GPe) of tremulous (vervet) and rigid-akinetic (macaque) monkeys and in the internal globus pallidum (GPi) of the vervet monkey. We found that, although oscillatory activity of cells and interneuronal correlation in both pallidal segments increases after induction of parkinsonism with 1-methyl-4-phenyl-1,2,3,6-tetra-hydropyridine (MPTP) and decreases in response to DRT, important differences exist between the two pallidal segments. In the GPi, the fraction of oscillatory cells and relative power of oscillations were significantly higher than in the GPe, and the dominant frequency was within the range of Hz compared with a range of Hz within the GPe. The interneuronal correlations were mostly oscillatory in the GPi, whereas at least half are non-oscillatory in the GPe. We demonstrate that the tremor characteristics after exposure to DRT do not resemble those of the normal or the levodopa-naive state. Moreover, although DRT reverses the MPTP-induced neuronal changes (rate, pattern, and pairwise correlations), the balance between GPe and GPi fails to restore. We therefore suggest that this imbalance reflects additional abnormal organization of the basal ganglia networks in response to dopamine replacement and may constitute the physiological substrate of the limitations and side effects of chronic DRT. Key words: Parkinson s disease; basal ganglia; MPTP; tremor; cross-correlations; levodopa Introduction Early physiological studies of parkinsonian 1-methyl-4-phenyl- 1,2,3,6-tetra-hydropyridine (MPTP)-treated monkeys reported changes in the discharge rate within the external globus pallidum (GPe), internal globus pallidum (GPi) (Miller and DeLong, 1987; Filion and Tremblay, 1991), and the subthalamic nucleus (STN) (Bergman et al., 1994). Subsequent findings showed that inactivation of STN and GPi could improve the motor symptoms in parkinsonian animals (Bergman et al., 1990; Aziz et al., 1991) and human patients (Lang et al., 1997; Kumar et al., 2000; Krack et al., Received May 2, 2005; revised June 9, 2006; accepted June 10, This work was supported in part by a Center of Excellence Grant from the Israel Science Foundation and the Fighting against Parkinson Foundation of the Hebrew University Netherlands Association (H.B.) and National Institutes of Health Grant MH (S.N.H.). V. Sharkansky provided technical support. We thank G. Goelman (Hadassah Hospital, Jerusalem, Israel) for assistance with the MRI imaging. We thank E. Vaadia for critical reading and fruitful discussions. Correspondence should be addressed to Gali Heimer, Department of Physiology, The Hebrew University Hadassah Medical School, P.O. Box 12272, Jerusalem, Israel galih@md.huji.ac.il. DOI: /JNEUROSCI Copyright 2006 Society for Neuroscience /06/ $15.00/ ; Walter and Vitek, 2004). Finally, reversed trends of pallidal discharge rates in response to dopamine replacement therapy (DRT) have been reported in both human patients (Hutchinson et al., 1997a; Merello et al., 1999; Levy et al., 2001) and primates (Filion et al., 1991; Papa et al., 1999; Heimer et al., 2002). These findings contributed to the formulation and the popularity of the rate model of the basal ganglia (Albin et al., 1989; DeLong, 1990). Nevertheless, subsequent studies have challenged the basic tenets of this model. Several studies have failed to find the expected significant changes of firing rates in the pallidum (Boraud et al., 1998; Raz et al., 2000), thalamus (Pessiglione et al., 2005), or motor cortical areas (Doudet et al., 1990; Watts and Mandir, 1992; Goldberg et al., 2002) of MPTP monkeys. Similarly, biochemical and metabolic studies indicate that GPe activity does not change in parkinsonism (Levy et al., 1997). Moreover, the rate model fails to explain the success of inactivation of the STN and GPi in the treatment of dyskinesia (Marsden and Obeso, 1994). The inconsistencies with the rate model have brought more attention to the potential role of other aspects of neuronal activity

99 Appendix I 8102 J. Neurosci., August 2, (31): Heimer et al. Pallidal Activity in MPTP and Dopamine Replacement such as firing patterns (Boraud et al., 2001; Wichmann and Soares, 2006) and neuronal synchronization (Bergman et al., 1998) in the pathophysiology of Parkinson s disease (PD). Recent studies have reported an increase in both oscillatory activity and correlation of pallidal cells in MPTP primates (Nini et al., 1995; Raz et al., 2000) and parkinsonian patients (Hurtado et al., 1999; Levy et al., 2000). Although abnormal pallidal synchronization has been shown to decrease in response to DRT (Heimer et al., 2002), detailed studies of pattern and synchronization during different DRT stages are still lacking. Moreover, recent human studies (Levy et al., 2002) have only found oscillatory neuronal correlation in tremulous patients, which leads to the question of whether the increased neuronal synchronization is not merely a byproduct of the tremor or of the activity of independent neural oscillators with similar frequencies. Human studies are limited by constraints of recording duration, selected anatomical targets, and the clinical state of the patients (e.g., most operated patients have already developed dyskinesia). In this study, we combined multielectrode recordings in the pallidum of control and MPTP-treated monkeys with a newly improved tool for spectral analysis of spike trains (Rivlin-Etzion et al., 2006). Using these tools, we investigated the role of pallidal oscillatory and non-oscillatory correlation throughout the different clinical states of MPTP-induced parkinsonism and DRT (levodopa-naive parkinsonian state and optimal and dyskinesiainducing DRT states). Materials and Methods Animals and behavioral paradigm. Two monkeys, a vervet (African green monkey, Cercopithecus aethiops aethiops, female, weight of 3.8 kg, monkey Q) and a rhesus (Macaca mulatta, female, weight of 5.7 kg, monkey R) were trained to perform a simple visuomotor task. The monkeys health was monitored by a veterinarian, and their fluid consumption, diet, and weight were assessed daily. All procedures were in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals (1996) and with the Hebrew University guidelines for the use and care of laboratory animals in research, approved and supervised by the Institutional Committee for Animal Care and Use. Surgical procedures. After training, an 18 mm Cilux recording chamber was attached to the skull over a trephine hole to allow access to the pallidum. The recording chamber was tilted 50 laterally in the coronal plan, with its center targeted at the following stereotaxic coordinates (in mm): monkey Q, anterior 13, lateral 7, height 3 (Contreras et al., 1981); monkey R, anterior 12, lateral 7, height 5 (Paxinos et al., 2000). The chamber coordinates were adjusted and then verified using magnetic resonance imaging (MRI) [BioSpec 4.7 tesla animal system (Bruker, Ettlingen, Germany), fast-spin echo sequence; effective echo time, 80 ms; repetition time, 2.5 s; 13 coronal slices, 2 mm wide]. All surgical and MRI procedures were performed under deep general anesthesia. Neural activity: recording and analysis. During recording sessions, the monkeys heads were immobilized, and eight glass-coated tungsten microelectrodes (impedance of M at 1000 Hz), confined within a cylindrical guide (1.65 mm inner diameter), were advanced separately (EPS; Alpha-Omega Engineering, Nazareth, Israel) into the pallidum. Each electrode signal was amplified with a gain of ,000 and bandpass filtered with a Hz four-pole Butterworth filter (MCP 2.8; Alpha-Omega Engineering). This electrical activity was sorted and classified on-line using a template-matching algorithm (MSD 3.21; Alpha-Omega Engineering). The sampling rate of spike detection pulses and behavioral events was 12 khz (AlphaMap 5.0; Alpha-Omega Engineering). In the vervet monkey Q, we also recorded the continuous analog output of the electrodes, which was sampled at 24 khz. Entry into the pallidum (the functional lateral border of the GPe) was easily recognized in our penetrations because of the considerably different firing rate, pattern, and spike shape of pallidal versus striatal neurons. The classification of each recorded cell as belonging to either the external or internal pallidum was determined by several criteria: the depth of the electrode [depth from the striatal pallidal functional border of all GPe and GPi cells included in the study was and mm (mean SD, respectively)], other anatomical/physiological structures identified along the electrode trajectory (e.g., border cells), and the firing pattern of the cell (which only served as a criterion in the normal state). If the classification as GPe or GPi was in doubt, the unit was excluded from the analysis. In monkey R, no GPi cells corresponded to all of the above criteria and we report only on the GPe activity of this monkey. Cells were selected for recording as a function of their signal-to-noise ratio and real-time assessment of their isolation quality. Only stable and well isolated (as judged by stable spike waveforms and off-line verification of stable firing rates) units were included in this study. The stable recording time was and min (mean SD) in monkeys Q and R, respectively. Minimum stable recording time was 4 and 3.3 min for monkeys Q and R, respectively. The total number of units that met the above criteria and were included in this study was as follows: 623 and 102 GPe units from monkeys Q and R, respectively, and 174 GPi units from monkey Q. We attempted to cover most of the area of the GPe and GPi. We estimated the spatial location of our recordings using the alignment of the x y coordinates of the chamber with the MR images and atlases (Contreras et al., 1981; Paxinos et al., 2000). Our penetrations span the stereotaxic range of A10 to A15 (GPe) and A11 to A14 (GPi) in monkey Q and A9 to A15 (GPe) in monkey R. We conducted quantitative analyses of firing rates and oscillatory firing patterns of all pallidal cells included in the study. Cross-correlations were calculated for pairs of neurons with overlapping periods of stable recording. Only neuronal pairs that were recorded by different electrodes were included to avoid possible artifacts in the cross-correlograms attributable to a shadowing effect of high discharge rates in cells recorded from the same electrode (Bar-Gad et al., 2001). Although the shadowing effect may be compensated for in neurons that exhibit a near Poissonian firing pattern (Bar-Gad et al., 2001), estimation of oscillatory correlations between neighboring units may be overbiased because of the shadowing effect. Statistical tests were accepted as significant at p 0.01 unless specified otherwise. The same threshold was applied when comparing the neural activity for both monkeys and for all clinical states. Oscillatory activity of the cells was estimated using the power spectrum density of the spike trains (frequency resolution, 0.25 Hz). To compensate for artifacts caused by the refractory period, we used the shuffling method (Rivlin-Etzion et al., 2006) in which the spectrum of the original spike train is divided by the mean spectrum of the locally (T 175 ms) shuffled (n 20) spike trains. A confidence level ( p 0.01, normalized to the total number of bins) for the compensated spectrum was constructed based on the high-frequency range of Hz, at which the spectrum was flat. A cell was considered oscillatory if its compensated spectrum contained at least two consecutive bins within the range of Hz that crossed the p 0.01 confidence level. For the analysis of neuronal correlation, we tested the null hypothesis of independent activity in both the time and frequency domains. We therefore searched for either significant peaks and troughs in the crosscorrelogram (a nonflat correlogram) or a peak in the cross-spectra (indicative of a significant periodic oscillatory correlation). The crosscorrelograms were calculated for 5000 ms offset, using 1 ms bins and recording edge corrections. Baseline firing rate and SD were estimated using the first and last 500 ms of the 5000 ms cross-correlogram. A peak Table 1. Order of appearance of clinical symptoms in the MPTP monkeys Days after first MPTP injection Symptom Monkey Q (vervet) Monkey R (macaque) Lower limb dystonia 3 3 Flexed posture 3 3 Bradykinesia 3 4 Akinesia 4 4 Freezing 4 5 Rigidity 5 7 Tremor

100 Appendix I Heimer et al. Pallidal Activity in MPTP and Dopamine Replacement J. Neurosci., August 2, (31): Figure1. PhotomicrographsofTHstainingdemonstratingthelossofdopaminergicsubstantianigraparscompactaneuronsin the MPTP-treated monkeys compared with a control animal. A was taken from a control normal macaque monkey; B and C are from the MPTP-treated monkeys R (macaque) and Q (vervet), respectively. The photomicrographs illustrate the levels of rostral striatum (row 1), central striatum (row 2), and midbrain (row 3). Note the lack of TH-positive staining throughout the striatum with the exception of the ventral striatum, particularly the shell region. TH-positive cells are selectively lost in the ventral tier (see arrows) but selectively spared in the ventral tegmental area. C, Caudate; P, putamen; VS, ventral striatum; SN, substantia nigra; VTA, ventral tegmental area. or trough was considered significant if it was made up of at least two consecutive bins that crossed the threshold of p 0.01 (normalized to the total number of bins) and was within an offset of 250 ms from 0. The significance of the oscillatory correlations was assessed using the cross-spectral density of the spike trains. As in the analysis of single-cell oscillations, we used the shuffling method (Rivlin-Etzion et al., 2006). The cross-spectrum of the original spike trains was divided by the mean cross-spectrum of the globally shuffled (n 20) spike trains. A confidence level ( p 0.01, normalized to the total number of bins) for the Table 2. Accelerometer recording data compensated spectrum was constructed based on the high-frequency range of Hz, at which the spectrum was flat. A correlogram was considered to have significant periodic oscillations if its compensated spectrum contained at least two consecutive bins within the range of Hz that crossed the p 0.01 confidence level. A noncorrelated pair was defined as a pair that had neither a significant peak in the crossspectra nor a significant peak or trough in the cross-correlogram. An oscillatory correlated pair was defined as a neuronal pair that had a significant peak in the cross-spectra. A nonoscillatory correlated pair was defined as a neuronal pair that only had a significant peak or trough in the cross-correlogram but did not have a significant peak in the cross-spectra. Comparison between the mean neuronal firing rate throughout the different clinical states was done using the Student s t test, with a significance of p 0.01 unless otherwise stated. In the analysis of rate changes in the continuously recorded cells, a significant change was defined as a decrease or increase of the firing rate of the cell by more than 10% of the baseline (before treatment) firing rate. The comparison of the fraction of neuronal oscillations and pairwise correlations in the different states was done using a 2 test with a significance of p Tremor: recording and analysis. We used uniaxial accelerometers (8630C5; Kistler, Amherst, NY) to assess limb tremor. The analog output of the accelerometers was sampled at 712 and 521 Hz in monkeys Q and R, respectively. Monkey R had one accelerometer fastened to its right hand (contralateral to the recorded hemisphere), and monkey Q had four accelerometers fastened to each of its four limbs. The accelerometers were attached distally on either the back of the hand or the foot. In the vervet monkey Q, we encountered many recording artifacts attributable to the strong tremor and collision of the monkey s limbs against solid surfaces, resulting in saturation of the recording apparatus and truncation of the data. We therefore only included accelerometer records that had a good signal-to-noise ratio and did not include such artifacts. Visual inspection of the raw traces revealed that the tremor episodes tended to be short; hence, calculating the power spectra over long periods would lead to a misrepresentation of the spectral content of these episodes. We therefore cut the data into 10 s fragments and performed NOR PNT POT-on POT-off PDT-on PDT-off Monkey Q (vervet) Recording days Accelerometers a Total segments 13,124 15, Tremulous segments (%) 4.5% 41.0% 50.9% 43.2% 42.3% 45% Coherent pairs 19.2% (5 of 26) 69.6% (39 of 56) 10.0% (3 of 30) 15.2% (5 of 33) 16.7% (5 of 30) 17.9% (5 of 28) Monkey R (macaque) Recording days Accelerometers Total segments Tremulous segments (%) 1.3% 4.5% 14.8% 0.7% POT-on, On periods of parkinsonian monkey undergoing optimal DRT on a daily basis; POT-off, Off periods of parkinsonian monkey undergoing optimal DRT; PDT-on, On periods of parkinsonian monkey undergoing daily DRT after the development of dyskinesia; PDT-off, Off periods of parkinsonian monkey undergoing daily DRT treatment after the development of dyskinesia. a Accelerometers refer to the number of accelerometer recordings within each state. In monkey Q, this number is higher than the number of recording days because of the multiple limb recordings and is lower than four times the number of recording days because of the discarding of sessions with artifacts. 98

101 Appendix I 8104 J. Neurosci., August 2, (31): Heimer et al. Pallidal Activity in MPTP and Dopamine Replacement power spectra analysis for every segment with a 1024 bin fast Fourier transform, yielding a spectral resolution of 0.7 and 0.5 Hz in monkeys Q and R, respectively. The SD of the spectrum noise was calculated from the tail of the spectrum ( Hz). We considered all peaks that were over 7 SD above a p 0.01 threshold to be significant (this combination of thresholds yielded the most similar results to visual judgment). Because of recording artifacts in low ( 3 Hz) frequencies (presumed to reflect limb movements) and 30 Hz (presumed to result from the resonance frequency of the specific accelerometers used), we only included the peaks that were between the frequencies of 3.5 and 28 Hz. If a segment had at least one significant peak within that range, it was considered tremulous. We calculated for each limb and each state the percentage of all the tremulous segments and the frequency distribution of the tremor frequencies of all the identified peaks. In monkey Q, the coherence between limb movements was calculated on the data of the simultaneously recorded accelerometers. For each pair of simultaneously recorded accelerometers, a coherence function was calculated for the same frequency range applied in the tremor analysis, and significant peaks were searched for in the range of Hz. We checked the mean of the coherence function between 25 and 30 Hz as an indicator of the noise level of the coherence function and excluded the functions with mean noise (a threshold chosen according to visual inspection of the data). In addition, we excluded the coherence function if the two accelerometers had an unequivocal noise artifact in the same frequency. We first used a standard significance criterion for the coherence function (Bloomfield, 1976; Brillinger, 1981): 1 1 a 1 L 1, where a is the level of confidence (here a 0.999), and L is the number of windows used in the calculation (length of the data divided by the window size, which in our case was 4096). We then applied a second threshold and considered as significant only those functions in which the integral between the two points in which the function crossed the significance line exceeded 0.04 (a threshold chosen according to visual inspection of the data). The overall coherence for each state was calculated as the percentage of accelerometer pairs that had at least one significant peak in their coherence function of the total number of possible pairs. MPTP and dopamine replacement therapy. Parkinsonism was induced by five intramuscular injections of 0.4 mg/kg of the MPTP-HCl neurotoxin (Aldrich, Milwaukee, WI) over a period of 4 d (two injections on the first day). Both monkeys were clinically assessed on a regular basis using a modified primate clinical staging scale (Hoehn and Yahr, 1967; Imbert et al., 2000). In both monkeys, severe parkinsonism developed within 5 d from initiation of the four day MPTP treatment, and recordings were resumed 4 d after the last injection. After 14 d of recordings in the parkinsonian state and 18 d after the last Figure2. Examplesofaccelerometerrecordingsandanalysis. A, B, Foursimultaneousaccelerometerrecordingsoflimbtremor inmonkeyqinthelevodopa-naiveparkinsonianstate(a)andtheonperiodofoptimallytreatedstate(b).thetremorisshownin two different timescales of 100 and 2 s. The tremor in the levodopa-naive parkinsonian state is quite regular, and coherence between the limbs is evident in both between- and within-tremor episodes. In contrast, in the treatment state, the tremor traces are erratic and there is no apparent coherence between the limbs. C, The spectrograms and power spectra of the same four simultaneously recorded accelerometer traces shown in the top panel of A. The relative power ordinate is expressed as a logarithmic scale. LH, Left hand; RH, right hand; LF, left foot; RF, right foot. 99 MPTP injections (in both monkeys), we initiated dopamine replacement therapy on a daily basis. Starting doses for monkey Q were /250 mg of Dopicar [L-3,4-dihydroxyphenylalanine and carbidopa; Merck Sharp and Dohme, Haarlem, The Netherlands] in the morning and 5 mg of Parlodel (Bromocryptine; Sandoz, Basel, Switzerland) divided equally

102 Appendix I Heimer et al. Pallidal Activity in MPTP and Dopamine Replacement J. Neurosci., August 2, (31): Figure 3. Results of the tremor analysis in both monkeys. A and B illustrate the percentage of segments that exhibited significant Hz tremor, in monkeys R and Q, respectively, throughout the clinical states. C illustrates the percentage of coherence among all pairs of simultaneously recorded accelerometers in monkey Q throughout the clinical states. D illustrates the distribution of the tremor frequencies in both monkeys, from the normal state (top) to the dyskinetic Off state (bottom) in the same order as they appear in A C. All significant Hz tremor peaks (possibly more than 1 for a single segment) are shown. NOR, Normal state; PNT, parkinsonian levodopa-naive (not treated) monkey; POT-on, on periods of parkinsonian monkey undergoing optimal DRT on a daily basis; POT-off, Off periods of parkinsonian monkey undergoing optimal DRT; PDT-on, On periods of parkinsonian monkey undergoing daily DRT after the development of dyskinesia. PDT-off, Off periods of parkinsonian monkey undergoing daily DRT treatment after the development of dyskinesia. between morning and evening. Starting doses for monkey R were /250 mg of Dopicar and 5 mg of Parlodel twice daily, in the morning and in the evening. The drugs were administered orally as crushed powder dissolved in liquid. The doses were slowly increased and adjusted to achieve optimal clinical response, and then the recordings were resumed. After a period of recording in the optimal treatment state, we gradually increased the doses until the development of dyskinesia and resumed the recordings once more. Maximal doses attained in monkey Q were /250 mg of Dopicar with 2.5 mg of Parlodel in the morning and 1 25/250 mg of Dopicar with 2.5 mg Parlodel in the evening. Maximal doses attained in monkey R were /250 mg of Dopicar with 5 mg of Parlodel in the morning and /250 mg of Dopicar with 5 mg of Parlodel in the evening. The clinical state was assessed daily by human observation. During the recordings, Off periods were defined either as the periods before the morning dose or periods of over 5 h from the last dose providing there were clear symptoms of severe parkinsonism. The clinical definition of the Off On transition was based on observation of limbs and tail 100 movements, resumption of task performance, or appearance of involuntary dyskinetic movements (at the stage these had already developed). On periods were defined as the periods after the Off On transition and up to 3 h from the administration of drugs. We use the following abbreviations for the clinical states: NOR refers to the normal state, PNT refers to the parkinsonian levodopa-naive ( no treatment ) state, POT refers to the state of optimal treatment in the parkinsonian monkey, and PDT refers to the state of dyskinesia-inducing treatment in the parkinsonian monkey. When using the phrase all parkinsonian states, we refer to the levodopa-naive state along with the Off states of both the optimal treatment (POT) and the dyskinetic treatment (PDT) periods. When referring to Off or On states without mentioning POT or PDT, we mean the Off or On states of both optimal and dyskinetic treatment. When referring to all treatment states, we mean both the optimal and dyskinetic treatment states in monkey Q and the dyskinetic state in monkey R. In the recordings during the DRT state, we recorded each day for min in the Off period before the morning dose. We then administered the medications while keeping the electrodes in position and subsequently resumed the recordings. In many cases, this protocol enabled recording of a given unit before, during, and after administration of DRT. In other cases, when units were lost during the oral administration of the drugs, it still enabled us to record the activity of other cells in the immediate vicinity of the cells studied before DRT. The protocol also made it possible in some instances to record the ongoing discharge changes of cells in response to the medication. In the data analysis, however, we only included the stable segments before and after these transients. Periods of complete cessation of GPi discharge after DRT were excluded from the analysis, because it was impossible to characterize the firing patterns and neuronal synchronization of cells that were virtually inactive. Because in monkey R we started using the above protocol only after the appearance of dyskinesia, the recordings of the optimal treatment state in this monkey did not match our Off/On criteria and were omitted from the study. Histology. After the last recording session (98 d from last MPTP injection in monkey Q and 151 d in monkey R), the monkeys were deeply anesthetized with a lethal dose of pentobarbital and perfused through the heart with saline, followed by a 4% paraformaldehyde fixative solution. Brains were removed and cryoprotected in increasing gradients of sucrose (10, 20, and finally 30%). Adjacent serial sections of 50 m, from both control animals and MPTP-treated animals, were processed for either a Nissl stain or immunocytochemistry for tyrosine hydroxylase (TH). Sections were incubated with antisera to TH (mouse anti-th, 1:20,000; Eugene Tech, Allendale, NJ) in 0.1 M phosphate buffer with 0.3% Triton X-100 and 10% normal goat serum (Incstar, Stillwater, MN) for 4 nights at 4 C and further processed using the avidin biotin method (rabbit Elite Vectastain ABC kit; Vector Laboratories, Burlingame, CA). Nissl histology was used to verify the recording location. However, the use of multiple-electrode recording, with a guide of outer diameter 2.2 mm, does not enable exact reconstruction of all penetration tracks, and therefore we only verified the recording boundaries to be inside the

103 Appendix I 8106 J. Neurosci., August 2, (31): Heimer et al. Pallidal Activity in MPTP and Dopamine Replacement pallidum. To estimate the dopaminergic fiber loss, the striatum was divided into three regions based on cortical inputs (limbic, associative, and motor) and analyzed for optical density of TH-positive fibers (NIH Image version 1.63). Mean density measurements were corrected for background staining by subtracting the average density sampled from white matter areas in each section. To standardize intensity across scans, an autoradiographic [ 14 C] microscale, multilevel reference strip (Amersham Biosciences, Arlington Heights, IL) was used to calibrate NIH Image. Results Clinical states The first signs of parkinsonism appeared on the third day of the MPTP injections in both monkeys and continued to evolve over the following 2 7 d (Table 1). Many prolonged episodes of low-frequency distal tremor were observed in monkey Q (vervet), whereas the short-lived tremor episodes of monkey R (macaque) appeared mostly in response to agitation and involved mainly the axial and proximal muscles. Both monkeys remained in a stable condition of severe parkinsonism during all recording days in the parkinsonian levodopa-naive state. Dopamine replacement therapy commenced 18 d after the last MPTP injection in both monkeys. The first response to therapy was seen after the third dose of medication in monkey Q and the fourth dose in monkey R ( 24 h after initiation of treatment). The Figure 4. Discharge rates of five GPe neurons recorded simultaneously and continuously throughout the clinical Off On transition in the dyskinetic treatment state (PDT). A, Continuous 28 min recording illustrates changes in the neuronal discharge rates, pattern, andsynchronizationinresponsetodrt. Medicationwasadministeredorally33minbeforeonsetofrecording(black arrow); clinical Off On transition was noted after 20 min of recording (open arrow). B D, Ten second traces of the raw analog recording of the same cells shown in A. Note the repeated long (several seconds) synchronous bursts in the Off period (B, C) that disappear in the consecutive On period (D). clinical effects of DRT included regaining of ability to self-feed, an increase in amount and velocity of movements, straightening of posture, and also a resumption (although suboptimal) of performance of the behavioral paradigm. The effects of DRT did not include a reduction of clinically observed tremor (for similar observations in human patients, see Vidailhet et al., 1999). In the macaque monkey R, which was relatively nontremulous in the parkinsonian levodopanaive state, the On state was accompanied by the appearance of tremor episodes. In the tremulous vervet monkey, there was an increase in apparent amplitude of the tremor rather than a change in its incidence. The first signs of peak-dose dyskinesia appeared after 6 and 7 weeks of daily DRT in monkeys Q and R, respectively. The dyskinesia manifested as overall hyperactivity, involuntary jerks of the limbs, torticolis, and episodes of circling. Table 3. Pallidal firing rates, oscillatory activity, and interneuronal correlations NOR PNT POT-on POT-off PDT-on PDT-off A. Neuronal firing rates (mean SEM) GPe monkey R (n 17) (n 47) (n 23) (n 15) GPe monkey Q (n 173) (n 181) (n 32) (n 56) (n 85) (n 96) GPi monkey Q (n 34) (n 50) (n 12) (n 35) (n 28) (n 17) B. Percentage of oscillatory cells GPe monkey R 0% (0 of 17) 44.7% (21 of 47) 17.4% (4 of 23) 13.3% (2 of 15) GPe monkey Q 3.5% (6 of 173) 33.7% (61 of 181) 6.3% (2 of 32) 12.5% (7 of 56) 3.5% (3 of 85) 14.6% (14 of 96) Gpi monkey Q 2.9% (1 of 34) 82.0% (41 of 50) 25.0% (3 of 12) 71.4% (25 of 35) 10.7% (3 of 28) 64.7% (11 of 17) C. Percentage of correlated pairs GPe-GPe monkey R 8.3% (1 of 12) 32.3% (20 of 62) 30.8% (8 of 26) 70.0% (21 of 30) GPe-GPe monkey Q 16.2% (53 of 327) 38.3% (164 of 428) 17.5% (7 of 40) 54.8% (57 of 104) 21.3% (29 of 136) 67.9% (161 of 237) GPi-GPi monkey Q 16.3% (7 of 43) 91.8% (56 of 61) 55.6% (5 of 9) 82.6% (19 of 23) 22.7% (5 of 22) 85.0% (17 of 20) D. Percentage of oscillatory out of all correlated pairs GPe-GPe monkey R 0% (0 of 1) 5.0% (1 of 20) 0% (0 of 8) 4.8% (1 of 21) GPe-GPe monkey Q 3.8% (2 of 53) 43.3% (71 of 164) 0% (0 of 7) 14.0% (0 of 7) 0% (0 of 29) 5.6% (9 of 161) GPi-GPi monkey Q 0% (0 of 7) 100.0% (56 of 56) 100.0% (5 of 5) 89.5% (17 of 19) 0% (0 of 5) 82.4% (14 of 17) Definitions are as in Table

104 Appendix I Heimer et al. Pallidal Activity in MPTP and Dopamine Replacement J. Neurosci., August 2, (31): Figure 5. A C, Summary of mean pallidal firing rates, fraction of oscillatory cells, and fraction of correlatedneuronalpairsthroughouttheclinicalstates. A, GPefiringratesdecreaseinthedopaminedepleted states and increase in response to dopamine replacement. The opposite occurs in the neuronsofthegpi.theerrorbarsrepresentthesem.clinicalstatedefinitionsareasinfigure3.b,fraction of Hz oscillatory cells is increased in both nuclei after induction of parkinsonism but is more pronounced in the GPi. During Off periods of treatment, oscillatory level in the GPi remains high, whereasinthegpeitissignificantlylowerthanthelevodopa-naivestate. C, Neuronalnon-oscillatory (filledbars)andoscillatory(stripedbars)correlationinthepallidum.neuronalcorrelationinthegpeis mainly non-oscillatory. It is increased in all parkinsonian states and reversed to near normal levels in response to dopamine replacement. Non-oscillatory correlation level is gradually increased from the levodopa-naive parkinsonian state throughout the Off states of treatment. In the GPi, neuronal correlationismostlyoscillatoryandreacheshigherlevelsthaninthegpe. Thecorrelationleveldecreases only partially in the On state of optimal treatment and profoundly in the dyskinetic On state. D, The balanceofgpe/gpineuronalactivitythatexistsinthenormalstateisdisruptedaftermptpandfailsto be restored by DRT. The GPe/GPi rate ratio decreases in the parkinsonian states and overshoots in responsetodrt.inallparkinsonianandtreatedstates,single-celloscillatoryactivityisstrongerwithin the GPi. Whereas in the PNT and POT states the correlated activity within the GPi is considerably strongerthanwithingpe,inthepdtstatesthegpe/gpicorrelationratioapproximates1.theordinate is given as a logarithmic scale. 102 Histology In the control animals, there was dense TH immunoreactivity throughout the striatum (Fig. 1 A). In contrast, both MPTPtreated animals had optical density measures close to 0 throughout the striatum, with the exception of the limbic region (Fig. 1B, C). At the microscopic level, few fibers remained in the dorsal striatum. The two MPTP animals did not differ with respect to TH fiber loss. Cell loss in the midbrain was almost complete in the ventral tier of the substantia nigra pars compacta. However, there were a few remaining cells in both animals. As expected (Song and Haber, 2000), the cells in the ventral tegmental area of the midbrain dopamine system remained relatively spared. Accelerometers and tremor Table 2 summarizes the total number of accelerometer recording days and the final number of accelerometers and segments used in the analysis. In the vervet monkey Q, the final number of accelerometry recordings used for the analysis was relatively low compared with the number of recording days attributable to the discarding of sessions with artifacts. Figure 2 presents an example of raw data and spectrograms of simultaneous recordings of four accelerometers in the vervet monkey Q during the levodopanaive parkinsonian state (Fig. 2A,C) and the On state of optimal treatment (Fig. 2B). In the normal state, the percentage of tremulous segments was relatively low in both monkeys (Fig. 3A,B). There was only a mild increase in the fraction of tremulous segments after MPTP in the parkinsonian macaque monkey R, whereas in the vervet monkey Q, it increased drastically to 40% of the segments. Throughout the optimal and dyskinetic treatment states in monkey Q, the percentage of tremulous segments remained high during both the Off and On periods. In monkey R, there was a clear increase in percentage of tremor in response to DRT in the dyskinetic state, reaching an even higher level than in the levodopa-naive parkinsonian state. The level of coherence between pairs of simultaneously recorded accelerometers in monkey Q is shown in Figure 3C. Along with the increase in the fraction of tremor segments, the interlimb coherence level also increased dramatically in the levodopanaive parkinsonian state. Nevertheless, after introduction of DRT, the coherence level decreased to the same level as in the normal state despite the sustained high percentage of tremor episodes. The high interlimb coherence in the levodopa-naive parkinsonian monkey and its absence in the treated monkey are illustrated in Figure 2, A and B, respectively, both between discrete tremor episodes and in the tremor waveforms within the episodes. Moreover, in the treated state, there was irregularity not only between accelerometers but also within the trace of each limb itself. These differences in the characteristics of the tremor were also reflected in the incidence of the different tremor frequencies in the six clinical states (Fig. 3D). Although in the levodopa-naive parkinsonian state of the vervet monkey Q there were two distinct tremor peaks at 6 and 11 Hz (Raz et al., 2000), the tremor frequencies in all treatment states were widely distributed. Neuronal firing rates An example of a continuous recording of five GPe units during the Off On transition in the dyskinetic state is shown in Figure 4. The continuous recording displays the rate changes as well as modifications of discharge patterns in response to DRT. Consistent with previous reports, the GPe cells increased their firing rate drastically in response to the medication. In addition, whereas in the Off period the GPe cells discharged in a synchronized burst-

105 Appendix I 8108 J. Neurosci., August 2, (31): Heimer et al. Pallidal Activity in MPTP and Dopamine Replacement Table 4. Changes in firing rates, oscillatory firing patterns, and synchronization in pallidal neurons continuously recorded over the Off On transition Optimal treatment state Dyskinetic treatment state A. Firing rate n Net change a Inc b NC b Dec b n Net change a Inc b NC b Dec b GPe 12 1 *1.6 ( 36 spk/s) 9 (75%) 3 (25%) 0 (0%) 44 1 *2.2 ( 38 spk/s) 37 (84%) 3 (9%) 4 (7%) GPi 4 2 *1.2 ( 10 spk/s) 1 (25%) 1 (25%) 2 (50%) 4 2 *1.8 ( 27 spk/s) 0 (0%) 0 (0%) B. Oscillatory firing pattern c n Osc Osc NoOsc NoOsc n Osc Osc NoOsc NoOsc Osc NoOsc Osc NoOsc Osc NoOsc GPe 12 17% 8% 75% 0% 44 12% 2% 84% 2% GPi 4 0% 75% 25% 0% 4 0% 50% 50% 0% C. Pairwise neuronal synchronization d n Cor Cor NoCor NoCor n Cor Cor NoCor NoCor Cor NoCor Cor NoCor Cor NoCor GPe 13 15% 15% 70% 0% % 26% 25% 3% GPi 5 40% 40% 20% 0% 2 100% 0% 0% 0% GPe continuously recorded cells are from both monkeys. a The average change of firing rate for all continuously recorded cells. The change is expressed as the factor by which the average rate increased or decreased and by spikes per second (spk/s). b The number of cells that decreased (Dec), increased (Inc), or did not change (NC) their discharge rate significantly after the medication. An increase or decrease of 10% from the baseline rate of the cell was considered significant. c The fraction of cells that changed their discharge pattern from oscillatory (Osc) to non-oscillatory (NoOsc), remained oscillatory, remained non-oscillatory, or changed their pattern from non-oscillatory to oscillatory in response to the medication. d The fraction of neuronal pairs that changed their correlation mode from correlated (Cor) to noncorrelated (NoCor), remained correlated, remained noncorrelated, or changed their mode from noncorrelated to correlated in response to the medication. 4 (100%) NoOsc 2 Osc NoCor 2 Cor ing manner, under the influence of medication, this bursting activity subsided considerably. Table 3A summarizes the mean firing rates of all GPe and GPi cells recorded in each clinical state. After MPTP treatment, GPe firing rates decreased only slightly in the macaque monkey R but significantly in the vervet monkey Q (Fig. 5A). During treatment, the GPe rates in the Off states continued to decrease in both monkeys so that, in the PDT-Off state, the GPe firing rates were significantly lower than those in the normal state in monkey R as well. In response to DRT (On states), the GPe rates increased significantly compared with the Off state in both monkeys and exceeded the normal state rates. The increase in GPi rates after MPTP was not significant; however, the GPi rates in the POT-Off state increased further and were marginally ( p 0.05) significantly higher than in the normal state (Fig. 5A). In response to DRT, in both optimal and dyskinetic states, neuronal rates in the GPi decreased significantly and were also significantly lower than in the normal state. While in the normal state, the ratio of GPe to GPi mean neuronal firing rates approached 1; in all parkinsonian states, it decreased by nearly twofold (Fig. 5D). Conversely, in the optimal treatment On state, the mean GPe rates were almost twice the mean GPi rates, and this disparity was even larger in the dyskinetic On state. The results of the continuously (before and after the clinical influence of DRT) recorded neurons verify the population averages given above (Table 4 A). Oscillatory activity of single neurons Figure 6 presents examples of raw analog traces, autocorrelation functions, power spectra, and spectrograms of oscillatory GPi (Fig. 6A, B) and GPe (Fig. 6C,D) cells during the levodopa-naive parkinsonian state (Fig. 6A, C) and Off state of optimal treatment (Fig. 6B,D). Table 3B summarizes the percentage of all Hz oscillatory GPe and GPi cells in each clinical state. Figure 7 depicts two examples of the complex and dynamic relationships between the pallidal neuronal oscillations and the tremor. It should be noted that, although this analysis was conducted in the 103 levodopa-naive period when the tremor of different body segments is highly coherent (Fig. 3C), there is only a partial overlap in the times of oscillatory neuronal activity and tremor. In line with previous reports (Lemstra et al., 1999; Raz et al., 2000; Hurtado et al., 2005), these results support the hypothesis of dynamical functional connection between the basal ganglia networks involved in tremor generation and the skeleton motor periphery. We therefore limited this paper to population level analysis of the tremor and the pallidal oscillations (see below). After induction of parkinsonism, there was a significant increase in the percentage of oscillatory GPe cells in both monkeys (Fig. 5B). During Off states of the treatment phases, however, the fraction of oscillatory GPe cells was significantly lower than in the levodopa-naive parkinsonian state ( p 0.01 for monkey Q and p 0.05 for monkey R). Although there was a small decrease in the fraction of oscillatory cells in response to DRT in the GPe of the vervet monkey Q, it was not significant in the optimal treatment and significant only at p 0.05 in the dyskinetic treatment. In the GPe of monkey R, there was no significant change in the fraction of oscillatory cells in response to DRT. The fraction of oscillatory GPe cells in the On states of treatment in both monkeys was not significantly different than in the normal state. In the GPi of monkey Q, there was a vast increase in oscillatory activity after induction of parkinsonism (Fig. 5B). Nonetheless, in contrast to the GPe, the fraction of oscillatory GPi cells remained high in the Off periods of the treatment states and was not significantly different than in the levodopanaive parkinsonian state. Moreover, again in contrast to the GPe, in the GPi, there was a significant decrease in the fraction of oscillatory cells in response to DRT in both the optimal and dyskinetic treatment. The fact that rate changes in the two pallidal segments went in opposite directions, whereas the changes in oscillatory activity were in the same direction, rules out the possibility that detection of oscillations was merely an artifact of the rate changes. In all parkinsonian states, the fraction of oscillatory cells in the

106 Appendix I Heimer et al. Pallidal Activity in MPTP and Dopamine Replacement J. Neurosci., August 2, (31): Figure 6. Examples of oscillatory activity of two GPi cells (A, B) and two GPe cells (C, D) of the vervet monkey Q during the levodopa-naive parkinsonian state (A, C) and optimal treatment Off state (B, D). For each unit, the top shows 2sofrawanalog signal. The bottom shows (from left to right) the autocorrelation, compensated (by shuffling) power spectrum, and spectrogram of the discharge of the cell. Ordinate of the autocorrelations, power spectra, and color bar of spectrograms appear at the same range for both GPi cells and both GPe cells. The examples illustrate the tendency of many GPi cells to exhibit strong 10 Hz oscillations compared with GPe cells that oscillate to a lesser extent and mostly at 6 Hz. Clinical state definitions are as in Figure 3. GPi was more than twice that of the GPe (Fig. 5D). In the normal state, conversely, there was no significant difference between the oscillatory fractions within the GPe compared with the GPi. Examining the Off/On ratio of the fraction of oscillatory cells shows that the extent of the Off On changes in oscillations was consistently larger in the GPi compared with the GPe. In addition, the magnitude of decrease in oscillatory percentage in response to DRT was larger in the dyskinetic treatment than in the optimal treatment in both nuclei. 104 The results of the cells recorded continuously over the Off On transition are summarized in Table 4B. Most cells that were oscillatory in the Off state became non-oscillatory in the On state. Among all of the continuously recorded cells that were not oscillatory in the Off state, only one (a GPe cell in the dyskinetic state) became oscillatory in the On state. The frequencies at which the pallidal cells oscillate are illustrated in Figure 8. The different frequency groups (Fig. 8B) were defined according to the clusters that are visible in Figure 8A. Whereas the dominant frequency of oscillations of GPe cells in both monkeys was between 4.5 and 7.5 Hz, in GPi cells the Hz oscillations were more than twice as common as the Hz group (see also Fig. 6). The tendency of GPe and GPi cells toward 6 and 10 Hz oscillations, respectively, was maintained throughout all of the parkinsonian states. Of all the oscillatory ( Hz) GPe cells in monkeys Q and R, 63.4 and 70.4%, respectively, only oscillated at Hz, 22.6 and 11.1% only oscillated at Hz, and 11.8 and 14.8% oscillated at both frequency domains. The mean frequency of all GPe cells that oscillated between 4.5 and 7.5 Hz was 5.8 and 6.1 Hz for monkeys Q and R, respectively. Of all oscillatory ( Hz) GPi cells in monkey Q, 69.5% only oscillated at Hz, 13.4% only oscillated at Hz, and 13.4% oscillated at both and Hz. The mean frequency of all oscillations in the GPi of monkey Q was 10.4 Hz. The only notable difference in the frequency distribution between the different clinical states was a relatively higher fraction of Hz oscillations in the GPi during On compared with Off states. Another qualitative difference between the neuronal oscillations of the GPe and GPi was the significantly ( p 0.01, Student s t test) higher relative power of all GPi oscillations in general and particularly the Hz group (Fig. 6, compare A,B with C,D and Fig. 8). The high relative power peaks at 10 Hz in the GPi (Fig. 8A) originated exclusively from the levodopa-naive parkinsonian state and the Off state of optimal treatment. We failed to find neuronal oscillations at Hz in the autocorrelograms of either monkey. No gamma frequency oscillations ( Hz) were seen in monkey R. In monkey Q, a small fraction of cells oscillated at this range in the normal state (1.2 and 5.9% of the GPe and GPi, respectively). Gamma oscillations were not detected in our spike trains in any other state, except in 2% of the GPi cells in the levodopa-naive parkinsonian state.

107 Appendix I 8110 J. Neurosci., August 2, (31): Heimer et al. Pallidal Activity in MPTP and Dopamine Replacement Neuronal correlation Figure 9, A and C, illustrates oscillatory and non-oscillatory, respectively, peaked cross-correlogram and crossspectra matrices, indicating that both types of interneuronal correlation exist in the pallidum of the parkinsonian monkey. Figure 9, B and D, which shows the flattening of the cross-correlogram and cross-spectra matrices of the same cells after administration of medication, is an example of the decreased level of neuronal oscillatory and non-oscillatory correlation in response to DRT. Table 3, C and D, summarizes the percentage of the correlated neuronal pairs within the GPe and GPi during each clinical state. Whereas in the normal state most neuronal pairs within the GPe of both monkeys were uncorrelated, after induction of parkinsonism, there was more than a twofold increase in the fraction of correlated pairs (Fig. 5C). Throughout the Off states of treatment, the percentage of correlated GPe pairs continued to increase, and, in the Off state of dyskinetic treatment, it was significantly higher than in the levodopa-naive parkinsonian state in both monkeys. In both monkeys, this increase in GPe correlation was attributable solely to a rise in non-oscillatory correlations. In response to DRT, the fraction of Figure 7. Examples of the complex temporal and spectral relationship between cell oscillations and tremor. A illustrates the same GPe cell shown in Figure 6C, and B illustrates the same GPi cell shown in Figure 6A (both are from monkey Q during the PNT state). On the left of each panel are the spectrogram of the neuronal activity, the spectrogram of a simultaneously recorded accelerometer, and their cross-spectrogram (from top to bottom). On the right of each panel are the corresponding auto- and cross-spectra. 105 correlated GPe pairs decreased significantly in both monkeys and was not significantly different than in the normal state. Although overall the level of correlation in the GPe during the levodopanaive parkinsonian state was similar for both monkeys, its characterization varied. In the tremulous vervet monkey Q, nearly half of all significant correlations were oscillatory, whereas in the nontremulous macaque monkey R, the oscillatory cross-correlations amounted to merely 5% of all significant correlations. The correlation level in the normal GPi was low and nonoscillatory in nature, similar to that of the normal GPe (Fig. 5C). After induction of parkinsonism, the percentage of correlated GPi pairs increased dramatically by more than fivefold and consisted of oscillatory correlations alone. In contrast to the GPe, in the GPi there was no significant difference in the overall level of correlation between the levodopa-naive parkinsonian state and the Off states of treatment. However, like the GPe, there was a significant decrease in the relative fraction of oscillatory cross-correlations during the treatment Off states and emergence of non-oscillatory correlations that were nonexistent in the levodopa-naive parkinsonian state. Whereas in the GPe the response to DRT was similar in both optimal and dyskinetic treatment, in the GPi there was a clear difference. The decrease in correlations in the optimal On state was minimal and the correlation level remained significantly higher than in the normal state. Furthermore, all correlated pairs in this state exhibited oscillatory correlations. Conversely, in the dyskinetic On state, the GPi correlation level decreased to a near normal level and comprised nonoscillatory correlation exclusively. Stable continuous recordings before and after the clinical influence of DRT revealed similar effects to the population results. As with the population results, most pairs that were correlated in the Off state became uncorrelated in the On state. Only a few neuronal pairs that were not correlated in the Off state synchronized their activity after the Off On transition (Table 4C). The frequencies of the oscillatory cross-correlations are shown in Figure 8. In the GPi, consistent with single-cell oscillations, the main frequency was between 7.5 and 13.5 Hz, and oscillatory cross-correlations exceeded those of the GPe in both their incidence and strength. In the GPe, conversely, the main frequency of oscillatory cross-correlations was also between 7.5 and 13.5 Hz, in contrast to the Hz oscillation that dominated the single-cell autocorrelations. Moreover, unlike the single-cell oscillations, Hz oscillatory cross-correlations were identified. Higher-frequency ( 30 Hz) oscillations were not found in our data. Discussion In this study, we explored the role of oscillatory and nonoscillatory pallidal activity at the level of both single-cell and interneuronal correlation, throughout the clinical states of parkinsonism and DRT. We recorded the extracellular spiking activity from the GPe (of macaque and vervet monkeys) and GPi (of the vervet monkey only). We combined tremor accelerometry, multiple-electrode recordings, population correlation, and spectral analysis (Rivlin-Etzion et al., 2006) with continuous recording of the same cells through the Off On transition. We produced and recorded from multiple clinical states similar to those observed in parkinsonian patients (akinetic-rigid vs tremordominant PD and optimal vs dyskinetic DRT) and compared these with normal control and levodopa-naive parkinsonian

108 Appendix I Heimer et al. Pallidal Activity in MPTP and Dopamine Replacement J. Neurosci., August 2, (31): Figure 8. Relative power and frequency distribution of single-cell oscillations and pairwise oscillatory correlations. A, Plot of the relative power of the power spectra peaks of single-cell oscillations (top row) and pairwise oscillatory correlations (bottom row) as a function of their frequency. All significant 1 30 Hz peaks (possibly more than 1 peak for a single cell or a single neuronal pair) from all clinical states were included. The plot demonstrates the different frequency clusters, the higher relative power of GPi oscillations, and the appearance of the Hz oscillatory correlations compared with the absence of such frequencies in the single-cell oscillations. B, Frequency distribution of single-cell oscillations (top row) and oscillatory pairwise correlations (bottom row). The different frequency groups were defined according to the clusters seen in A. Units or neuronal pairs that oscillated in more than one frequency group were counted more than once. Despite the different dominant frequencies of single-cell oscillations within the GPe and GPi ( 6 and 10 Hz, respectively), in both nuclei the chief frequency of oscillatory correlations is 10 Hz. Clinical state definitions are as in Figure 3. states, which are lacking in human studies. Our fast-induction MPTP model might not exactly reflect the slow and progressive anatomical/biochemical compensatory changes in the basal ganglia that occur in idiopathic PD or in the slowprogressive MPTP models (Russ et al., 1991; Perez et al., 1994). However, it enabled prolonged recording over the multiple PD states in the same animal, including the state of levodopa-induced dyskinesia. Using these methods, we demonstrate that the tremor characteristics after exposure to DRT do not resemble those of the normal or the levodopa-naive state. Moreover, 106 the balance between the neural activity (rate, pattern, and pairwise correlations) between the GPe and GPi is disrupted after MPTP treatment. These results therefore call for reappraisal of our current models of basal ganglia and Parkinson s disease pathophysiology (see below). Tremor analysis The tremor analysis confirms previous reports of an abundance of tremor episodes in the MPTP-treated vervet monkey compared with the relatively nontremulous macaque (Redmond et al., 1985; Bergman et al., 1994). In both monkeys, there was either no change or even an increase in the extent of the tremor in response to DRT. This phenomenon is occasionally observed in human patients (Vidailhet et al., 1999) and may be attributed to tremor masking by severe akinesia and rigidity. As in human studies (Hurtado et al., 2000; Raethjen et al., 2000; Ben-Pazi et al., 2001), the DRT states were characterized by a low coherence level between the tremor of the limbs even in Off states; however, in the levodopa-naive state, limb tremor was highly coherent. These results indicate (in line with the electrophysiological studies discussed below) that DRT causes major changes in the functional organization of the basal ganglia. The Off state of human parkinsonian patients, after many years of DRT, therefore may not represent a pure dopamine-depleted state; rather, it may be the result of complex interactions between natural compensatory processes for dopamine depletion, as well as the neuronal responses to chronic DRT. Neuronal firing rates In the analysis of neuronal firing rates in the levodopa-naive parkinsonian state, we only found a significant change in the discharge rate of GPe cells in the vervet monkey. These inconclusive results are in line with other primate studies (Boraud et al., 2002). Nevertheless, as in primate (Filion et al., 1991; Papa et al., 1999; Boraud et al., 2001) and human (Levy et al., 2001; Stefani et al., 2002) studies, the increase in GPe and decrease in GPi rates in response to DRT were robust. As a result, during the Off periods, GPi rates were significantly higher than the GPe rates, whereas the opposite occurred in On periods. Single-cell oscillations The low fraction of oscillatory cells in both nuclei significantly increased after induction of parkinsonism, and different degrees of decrease were observed in response to DRT. Nevertheless, notable differences were found between the neuronal oscillations of the two pallidal nuclei. In the GPi, the high oscillatory level was maintained throughout the DRT-Off states, whereas in the GPe

109 Appendix I 8112 J. Neurosci., August 2, (31): Heimer et al. Pallidal Activity in MPTP and Dopamine Replacement of both monkeys, it was significantly lower during the DRT-Off states compared with the levodopa-naive parkinsonian state. Throughout all MPTP and DRT states, the fraction and relative power of oscillatory cells within the GPi was higher than that of the GPe, and the dominant frequency of single-cell oscillations in the GPi was Hz compared with Hz in the GPe. The small fraction of cells oscillating in more than one frequency and the uneven distribution of the two main frequencies between the GPe and GPi may indicate that each frequency results from genuine physiological characteristics of the cells or the network rather than from a harmonic artifact of the spectral analysis. Human studies have reported a high fraction of GPi cells oscillating at the tremor frequency (Hutchinson et al., 1997b); however, recent primate (Raz et al., 2000) and human (Hurtado et al., 1999, 2005; Lemstra et al., 1999) studies show that these oscillations are not fully coherent with the simultaneous recorded tremor. Our results also reveal several discrepancies between the pallidal oscillations and the tremor. Moreover, comparison of the GPe activity in the tremulous and nontremulous animals implies that the formation of oscillatory correlations rather than single-cell oscillations plays a major role in tremor generation. We demonstrate that, in contrast to similar fractions of oscillatory cells and correlated pairs in the GPe of both animals in the levodopa-naive state, most pairwise correlations in the nontremulous macaque are non-oscillatory, whereas in the tremulous vervet, nearly half are oscillatory. However, these findings are population based and circumstantial; therefore, additional studies are needed to evaluate the specific temporal relationships between single or assembly neuronal oscillations and the tremor phenomenon. Interneuronal correlations In line with previous MPTP studies (Nini et al., 1995; Raz et al., 2000), we found an increased level of pairwise neuronal correlations in both pallidal segments after MPTP treatment. This abnormal synchronization decreased by a variable extent in response to DRT. The disparity between the two pallidal nuclei was also present in terms of interneuronal correlations. In contrast to the similarly low interneuronal synchronization in both GPe and GPi in the normal state, after MPTP, the level of neuronal synchronization in the GPi significantly exceeded that of the GPe. Moreover, whereas neuronal correlations in the GPe were mostly non-oscillatory, the vast majority of correlated GPi pairs exhibited 10 Hz oscillatory correlations. A recent study in human patients only found oscillatory single-cell activity and interneuronal synchronization in tremulous patients and failed to find any non-oscillatory correlations (Levy et al., 2002). These results may imply that the oscillatory Figure 9. Example of the decrease in neuronal synchronization in response to DRT. In each panel, the left bottom triangle is the cross-correlogram matrix, and the right top triangle is the cross-spectra matrix. A, B, Prominent oscillatory synchronization in the Off period of optimal treatment state (A) is significantly decreased after response to the medication (B). The matrices were constructed based on the continuous simultaneous recordings of two GPe cells and three GPi cells during Off period and the subsequentonperiod22minlater. C, D, Widepeaksofnon-oscillatorysynchronizationintheOffperiod(C) ofdyskinetictreatment state are flattened after response to the medication (D). The matrices were constructed based on the continuous simultaneous recordings of the same five GPe cells shown in Figure 4, during the Off period and the subsequent On period 26 min later. The ordinate of the conditional firing rate is expressed in the same range of 12 spikes/s around the average firing rate. The ordinate of the relative power is also the same for all cross-spectra in A and B (0 35 Hz) and for all cross-spectra in C and D (0 20 Hz). 107 correlation is merely a byproduct of the tremor or independent oscillators with similar frequencies. We found a significant fraction of non-oscillatory synchronization in our recordings. Additional findings presented here indicating different frequency regimens for the autocorrelation and cross-correlation functions, as well as distinct modulation of the tremor, single-cell, and interneuronal oscillations by the DRT further suggest that the increased neuronal synchronization within the pallidum is not simply a reflection of tremor or single-cell oscillations. Neuronal substrates of levodopa-induced dyskinesia A major aim of our study was to characterize the physiological differences in neuronal activity during optimal and dyskinetic treatment. Our results are in line with a previous study that found lower neuronal rates in the GPi during dyskinetic On compared with optimal On (Papa et al., 1999). In addition, we show that the imbalance of GPe/GPi firing rates in response to DRT is further increased in the dyskinetic On state. We found a more pronounced Off On decrease in oscillatory activity in the dyskinetic state in both pallidal nuclei. Similarly, the relative proportion of oscillatory correlations was lower in the dyskinetic On state. After

110 Appendix I Heimer et al. Pallidal Activity in MPTP and Dopamine Replacement J. Neurosci., August 2, (31): introduction of DRT, prolonged synchronous bursts of GPe neurons appeared in the Off states and were considerably more prominent after the development of dyskinesia. We propose that the weakening of network oscillations within the GPi in the dyskinetic state enables the emerging of bursting activity in the GPe or other structures such as the thalamus to manifest in the form of dyskinesia. The occurrence of dyskinesia during On rather than Off periods can be attributed to the concurrent decrease of GPi rates, which further reduces its inhibitory effect. We conclude that dopamine depletion and replacement therapy result in changes in many aspects of neural activity (e.g., rate, pattern, and synchronization) over the whole basal ganglia cortical networks; consequently, the GPe/GPi balance that exists in the normal state (in these three aspects of neuronal activity) is disturbed and fails to reinstate thereafter. Recent studies (Bolam et al., 2000; Lee et al., 2004; Levesque and Parent, 2005) have indicated that the GPe should not be considered as a simple relay structure in the indirect striatopallidal pathway. Rather, the GPe plays a major role in the control of the entire basal ganglia circuitry. Nevertheless, despite the strong inhibitory projections from the GPe to the GPi (Hazrati et al., 1990; Bolam et al., 2000), the GPe activity does not impose mirror changes in the GPi neuronal activity. Our results, demonstrating disruption of the GPe/ GPi balance after MPTP and DRT, suggest that this imbalance is a hallmark of the pathophysiology of parkinsonism and DRT. They further indicate that chronic DRT may cause additional abnormal organization of the basal ganglia networks beyond the effects of dopamine depletion. Future attempts at improving invasive therapies of human patients with advanced PD and DRT side effects should therefore aim at restoring the full spectrum of normal GPe and GPi activity. 108 References Albin RL, Young AB, Penney JB (1989) The functional anatomy of basal ganglia disorders. Trends Neurosci 12: Aziz TZ, Peggs D, Sambrook MA, Crossman AR (1991) Lesion of the subthalamic nucleus for the alleviation of 1-methyl-4-phenyl-1,2,3,6- tetrahydropyridine (MPTP)-induced parkinsonism in the primate. Mov Disord 6: Bar-Gad I, Ritov Y, Bergman H (2001) Failure in identification of overlapping spikes from multiple neuron activity causes artificial correlations. J Neurosci Methods 107:1 13. Ben-Pazi H, Bergman H, Goldber JA, Giladi N, Hansel D, Reches A, Simon ES (2001) Synchrony of rest tremor in multiple limbs in parkinson s disease: evidence for multiple oscillators. J Neural Transm 108: Bergman H, Wichmann T, DeLong MR (1990) Reversal of experimental parkinsonism by lesions of the subthalamic nucleus. Science 249: Bergman H, Wichmann T, Karmon B, DeLong MR (1994) The primate subthalamic nucleus. II. Neuronal activity in the MPTP model of parkinsonism. J Neurophysiol 72: Bergman H, Feingold A, Nini A, Raz A, Slovin H, Abeles M, Vaadia E (1998) Physiological aspects of information processing in the basal ganglia of normal and parkinsonian primates. Trends Neurosci 21: Bloomfield P (1976) Fourier analysis of time series: an introduction. New York: Wiley. Bolam JP, Hanley JJ, Booth PA, Bevan MD (2000) Synaptic organisation of the basal ganglia. J Anat 196: Boraud T, Bezard E, Guehl D, Bioulac B, Gross C (1998) Effects of L-DOPA on neuronal activity of the globus pallidus externalis (GPe) and globus pallidus internalis (GPi) in the MPTP-treated monkey. Brain Res 787: Boraud T, Bezard E, Bioulac B, Gross CE (2001) Dopamine agonist-induced dyskinesias are correlated to both firing pattern and frequency alterations of pallidal neurones in the MPTP-treated monkey. Brain 124: Boraud T, Bezard E, Bioulac B, Gross CE (2002) From single extracellular unit recording in experimental and human Parkinsonism to the development of a functional concept of the role played by the basal ganglia in motor control. Prog Neurobiol 66: Brillinger DR (1981) Time series data analysis and theory. San Francisco: Holden-Day. Contreras CM, Mexicano G, Guzman-Flores CA (1981) Stereotaxic brain atlas of the green monkey (Cercopithecus Aethiops Aethiops). Bol Est Med Biol 31: DeLong MR (1990) Primate models of movement disorders of basal ganglia origin. Trends Neurosci 13: Doudet DJ, Gross C, Arluison M, Bioulac B (1990) Modifications of precentral cortex discharge and EMG activity in monkeys with MPTP-induced lesions of DA nigral neurons. Exp Brain Res 80: Filion M, Tremblay L (1991) Abnormal spontaneous activity of globus pallidus neurons in monkeys with MPTP-induced parkinsonism. Brain Res 547: Filion M, Tremblay L, Bedard PJ (1991) Effects of dopamine agonists on the spontaneous activity of globus pallidus neurons in monkeys with MPTPinduced parkinsonism. Brain Res 547: Goldberg JA, Boraud T, Maraton S, Haber SN, Vaadia E, Bergman H (2002) Enhanced synchrony among primary motor cortex neurons in the 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine primate model of Parkinson s disease. J Neurosci 22: Hazrati LN, Parent A, Mitchell S, Haber SN (1990) Evidence for interconnections between the two segments of the globus pallidus in primates: a PHA-L anterograde tracing study. Brain Res 533: Heimer G, Bar-Gad I, Goldberg JA, Bergman H (2002) Dopamine replacement therapy reverses abnormal synchronization of pallidal neurons in the 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine primate model of parkinsonism. J Neurosci 22: Hoehn MM, Yahr MD (1967) Parkinsonism: onset, progression and mortality. Neurology 17: Hurtado JM, Gray CM, Tamas LB, Sigvardt KA (1999) Dynamics of tremorrelated oscillations in the human globus pallidus: a single case study. Proc Natl Acad Sci USA 96: Hurtado JM, Lachaux JP, Beckley DJ, Gray CM, Sigvardt KA (2000) Interand intralimb oscillator coupling in parkinsonian tremor. Mov Disord 15: Hurtado JM, Rubchinsky LL, Sigvardt KA, Wheelock VL, Pappas CT (2005) Temporal evolution of oscillations and synchrony in GPi/muscle pairs in Parkinson s disease. J Neurophysiol 93: Hutchinson WD, Levy R, Dostrovsky JO, Lozano AM, Lang AE (1997a) Effects of apomorphine on globus pallidus neurons in parkinsonian patients. Ann Neurol 42: Hutchinson WD, Lozano AM, Tasker RR, Lang AE, Dostrovsky JO (1997b) Identification and characterization of neurons with tremor- frequency activity in human globus pallidus. Exp Brain Res 113: Imbert C, Bezard E, Guitraud S, Boraud T, Gross CE (2000) Comparison of eight clinical rating scales used for the assessment of MPTP-induced parkinsonism in the Macaque monkey. J Neurosci Methods 96: Krack P, Batir A, Van Blercom N, Chabardes S, Fraix V, Ardouin C, Koudsie A, Limousin PD, Benazzouz A, Lebas JF, Benabid AL, Pollak P (2003) Five-year follow-up of bilateral stimulation of the subthalamic nucleus in advanced Parkinson s disease. N Engl J Med 349: Kumar R, Lang AE, Rodriguez-Oroz MC, Lozano AM, Limousin P, Pollak P, Benabid AL, Guridi J, Ramos E, van der Linden C, Vandewalle A, Caemaert J, Lannoo E, van den Abbeele D, Vingerhoets G, Wolters M, Obeso JA (2000) Deep brain stimulation of the globus pallidus pars interna in advanced Parkinson s disease. Neurology 55:S34 S39. Lang AE, Lozano AM, Montgomery E, Duff J, Tasker R, Hutchinson W (1997) Posteroventral medial pallidotomy in advanced Parkinson s disease. N Engl J Med 337: Lee CR, Abercrombie ED, Tepper JM (2004) Pallidal control of substantia nigra dopaminergic neuron firing pattern and its relation to extracellular neostriatal dopamine levels. Neuroscience 129: Lemstra AW, Verhagen ML, Lee JI, Dougherty PM, Lenz FA (1999) Tremor-frequency (3 6 Hz) activity in the sensorimotor arm representation of the internal segment of the globus pallidus in patients with Parkinson s disease. Neurosci Lett 267: Levesque M, Parent A (2005) The striatofugal fiber system in primates: a reevaluation of its organization based on single-axon tracing studies. Proc Natl Acad Sci USA 102: Levy R, Hazrati LN, Herrero MT, Vila M, Hassani OK, Mouroux M, Ruberg

111 Appendix I 8114 J. Neurosci., August 2, (31): Heimer et al. Pallidal Activity in MPTP and Dopamine Replacement M, ASENSI H, Agid Y, Feger J, Obeso JA, Parent A, Hirsch EC (1997) Re-evaluation of the functional anatomy of the basal ganglia in normal and Parkinsonian states. Neuroscience 76: Levy R, Hutchison WD, Lozano AM, Dostrovsky JO (2000) High-frequency synchronization of neuronal activity in the subthalamic nucleus of parkinsonian patients with limb tremor. J Neurosci 20: Levy R, Dostrovsky JO, Lang AE, Sime E, Hutchison WD, Lozano AM (2001) Effects of apomorphine on subthalamic nucleus and globus pallidus internus neurons in patients with Parkinson s disease. J Neurophysiol 86: Levy R, Hutchison WD, Lozano AM, Dostrovsky JO (2002) Synchronized neuronal discharge in the basal ganglia of parkinsonian patients is limited to oscillatory activity. J Neurosci 22: Marsden CD, Obeso JA (1994) The functions of the basal ganglia and the paradox of stereotaxic surgery in Parkinson s disease. Brain 117: Merello M, Balej J, Delfino M, Cammarota A, Betti O, Leiguarda R (1999) Apomorphine induces changes in GPi spontaneous outflow in patients with Parkinson s disease. Mov Disord 14: Miller WC, DeLong MR (1987) Altered tonic activity of neurons in the globus pallidus and subthalamic nucleus in the primate MPTP model of parkinsonism. In: The basal ganglia II (Carpenter MB, Jayaraman A, eds), pp New York: Plenum. Nini A, Feingold A, Slovin H, Bergman H (1995) Neurons in the globus pallidus do not show correlated activity in the normal monkey, but phaselocked oscillations appear in the MPTP model of parkinsonism. J Neurophysiol 74: Papa SM, DeSimone R, Fiorani M, Oldfield EH (1999) Internal globus pallidus discharge is nearly suppressed during levodopa-induced dyskinesias. Ann Neurol 46: Paxinos G, Huang X-F, Toga AW (2000) The rhesus monkey brain in stereotaxic coordinates. San Diego: Academic. Perez OI, Oset C, Luquin MR, Herrero MT, Obeso JA, Del Rio J (1994) MPTP-induced parkinsonism in primates: pattern of striatal dopamine loss following acute and chronic administration. Neurosci Lett 175: Pessiglione M, Guehl D, Rolland AS, Francois C, Hirsch EC, Feger J, Tremblay L (2005) Thalamic neuronal activity in dopamine-depleted primates: evidence for a loss of functional segregation within basal ganglia circuits. J Neurosci 25: Raethjen J, Lindemann M, Schmaljohann H, Wenzelburger R, Pfister G, Deuschl G (2000) Multiple oscillators are causing parkinsonian and essential tremor. Mov Disord 15: Raz A, Vaadia E, Bergman H (2000) Firing patterns and correlations of spontaneous discharge of pallidal neurons in the normal and the tremulous 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine vervet model of parkinsonism. J Neurosci 20: Redmond DE, Roth RH, Sladek JR (1985) MPTP produces classic parkinsonian syndrome in African green monkeys. Soc Neurosci Abstr 11:54.1. Rivlin-Etzion M, Ritov Y, Heimer G, Bergman H, Bar-Gad I (2006) Local shuffling of spike trains boosts the accuracy of spike train spectral analysis. J Neurophysiol 95: Russ H, Mihatsch W, Gerlach M, Riederer P, Przuntek H (1991) Neurochemical and behavioural features induced by chronic low dose treatment with 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP) in the common marmoset: implications for Parkinson s disease? Neurosci Lett 123: Song DD, Haber SN (2000) Striatal responses to partial dopaminergic lesion: evidence for compensatory sprouting. J Neurosci 20: Stefani A, Bassi A, Mazzone P, Pierantozzi M, Gattoni G, Altibrandi MG, Giacomini P, Peppe A, Bernardi G, Stanzione P (2002) Subdyskinetic apomorphine responses in globus pallidus and subthalamus of parkinsonian patients: lack of clear evidence for the indirect pathway. Clin Neurophysiol 113: Vidailhet M, Bonnet AM, Marconi R, Durif F, Agid Y (1999) The phenomenology of L-dopa-induced dyskinesias in Parkinson s disease. Mov Disord 14 [Suppl 1]: Walter BL, Vitek JL (2004) Surgical treatment for Parkinson s disease. Lancet Neurol 3: Watts RL, Mandir AS (1992) The role of motor cortex in the pathophysiology of voluntary movement deficits associated with parkinsonism. Neurol Clin 10: Wichmann T, Soares J (2006) Neuronal firing before and after burst discharges in the monkey basal ganglia is predictably patterned in the normal state and altered in parkinsonism. J Neurophysiol 95:

112 Appendix II II. PARKINSON'S DISEASE: FIGHTING THE WILL? 110

113 Appendix II The Journal of Neuroscience, October 31, (44): Journal Club Editor s Note: These short reviews of a recent paper in the Journal, written exclusively by graduate students or postdoctoral fellows, are intended to mimic the journal clubs that exist in your own departments or institutions. For more information on the format and purpose of the Journal Club, please see Parkinson s Disease: Fighting the Will? Yael Niv 1,2 and Michal Rivlin-Etzion 2,3 1 Center for the Study of Brain, Mind and Behavior, Princeton University, Princeton, New Jersey 08544, 2 Interdisciplinary Center for Neural Computation, The Hebrew University of Jerusalem, Jerusalem, Israel, and 3 Department of Physiology, The Hebrew University-Hadassah Medical School, Jerusalem, Israel Review of Mazzoni et al. ( A phenomenon familiar to clinicians treating patients with Parkinson s disease (PD) is kinesia paradoxica: astonishing displays of sudden mobility and agility by otherwise akinetic PD patients in instances of emergency (for instance, the case, perhaps apocryphal, in which an immobile patient suddenly leapt from his wheelchair to save a drowning man). Recently, Mazzoni et al. (2007) in their paper in The Journal of Neuroscience showed that such normal motor behavior amid the general motor symptoms of Parkinson s disease is not confined to extreme cases. Mazzoni et al. (2007) asked PD patients and healthy controls to move their arm to a prespecified target at different speeds [Mazzoni et al. (2007), their Fig. 1a ( full/27/27/7105/f1)]. The accuracy and peak velocity of the movements were monitored and reported to the subjects after each trial [Mazzoni et al. (2007), their Fig. 1b,c ( cgi/content/full/27/27/7105/f1)], with target distance and required speed range varying between experimental blocks. To complete a block subjects had to perform Received Sept. 2, 2007; revised Sept. 11, 2007; accepted Sept. 16, This work was supported by the Human Frontiers Science Program (Y.N.) and by a Ms. Lily Safra Interdisciplinary Center for Neural Computationgraduatestudentfellowship(M.R.).WearegratefultoPietroMazzoni, Peter Dayan, Nathaniel Daw, and Hagai Bergman for helpful comments. Correspondence should be addressed to Yael Niv at the above address. yael@princeton.edu. DOI: /JNEUROSCI Copyright 2007SocietyforNeuroscience /07/ $15.00/0 20 movements within the required range. The question that interested Mazzoni et al. (2007) was whether the profound slowing of movement (bradykinesia) observed in PD might result from impaired decision-making and action selection, rather than from compensation for reduced accuracy of rapid movements. This simple and elegant design revealed a multitude of interesting results. First, like healthy controls, PD patients were able to make movements at the required velocity in each condition. Moreover, when the accuracy and the kinematics of the 20 movements that met criteria were examined, PD patients were no different from healthy controls [Mazzoni et al. (2007), their Figs. 2 ( jneurosci.org/cgi/content/full/27/27/ 7105/F2), 3 ( cgi/content/full/27/27/7105/f3), 4 ( full/27/27/7105/f4)]. In fact, a comparison of the accuracies of all movements of similar velocities (valid and invalid) from different blocks confirmed that patients were, in general, as accurate as controls [Mazzoni et al. (2007), their Fig. 5 ( 27/7105/F5)]. This result demonstrates convincingly that Parkinson s patients can display motor behavior that matches that of healthy controls in both speed and accuracy, even in mundane situations such as moving an arm to reach an arbitrary target. 111 However, Mazzoni et al. (2007) did find one major difference: PD patients made significantly more slow movements before reaching the criterion of 20 movements within the required speed range. This difference was most apparent when the task was difficult, that is, in blocks that required high speed relative to a small distance (rapid acceleration and deceleration of movement). For example, for a 6 cm target and a velocity requirement of cm/s, PD patients made, on average, over 20 invalid (too slow) movements, whereas controls made only 10 [Mazzoni et al. (2007), their Fig. 6 org/cgi/content/full/27/27/7105/f6)]. Thus, although PD patients were capable of performing high-velocity movements, they performed these actions with lower probability. Figure 1 shows, for each block, the probability distributions from which movement velocities were chosen, as estimated from the peak velocities of all movements (valid and invalid). Although spanning the same range of velocities, the probability distribution used by PD patients was shifted to the left (i.e., it assigned higher probability to slower speeds) compared with that of controls. If the slowing of behavior is not a result of compensation in terms of a speedaccuracy tradeoff, and if Parkinson s patients are physically able to perform the required actions at no loss of accuracy, why do they seem to choose to behave more slowly?

114 Appendix II J. Neurosci., October 31, (44): Niv and Rivlin-Etzion Journal Club Figure 1. Velocity probability distributions of Parkinson s patients are shifted toward slower movements in all but the easiest conditions. Probability distribution functions for each condition were estimated using kernel density estimation based on all movements (valid and invalid) (see Mazzoni et al., 2007), and normalized [for non-normalized histograms, see Mazzoni et al. (2007), their Fig. 7 ( In red are the estimated distributions for PD patients, in black are those for healthy controls. Shaded areas mark the required range of velocities in each condition [slow (S), 17 37cm/s; medium(m), 37 57cm/s; fast(f), 57 77cm/s; veryfast(vf), 77 97cm/s], andnumbersinbold(6,12, or16) denote the distance to target in centimeters. Dashed line, Empirical mean. Data are courtesy of Dr. Pietro Mazzoni (Mazzoni et al., 2007) 112 Parkinson s disease has long been associated with the progressive death of midbrain dopaminergic neurons and a consequent lack of dopamine in the basal ganglia, specifically in areas of the striatum that are implicated in motor control and movement initiation. More recently, dopamine in the basal ganglia has been tightly linked to reinforcement learning via reward prediction error signals (Montague et al., 1996). Can a reinforcement learning deficiency cause motor slowing? Based on their experimental results, and building on a previous reinforcement learning model of response speed (Niv et al., 2007), Mazzoni et al. (2007) hypothesized that the slowness of movement in PD may be attributable to higher sensitivity to movement energy costs as a result of loss of dopamine. However, according to the model, optimal choice of movement time (or action latency) should take into account not only motor and energetic costs, but also the opportunity cost of devoting time to one action and not to others. This suggests another possible source of bradykinesia: because the steady-state net rate of rewards per unit time quantifies the cost of time, the model shows that decreasing the net rate of rewards will influence the pace of all actions such that they will be performed more slowly. Importantly, it has been suggested that the net rate of rewards is represented by tonic levels of dopamine in the striatum (Niv et al., 2007).Thus, loss of dopamine may cause bradykinesia not through a speed/accuracy tradeoff, but rather by affecting decision making through an enhancement of the costs of movement or through a distorted value of time itself. Mazzoni et al. s (2007) study highlights an important characteristic of dopamine-dependent decision making, namely, its implicit nature. Subjects were given full instructions and feedback, allowing them to explicitly (or consciously) appreciate that faster responding would, overall, reduce both time and motor efforts. Despite this, the dopaminedependent cost/benefit computation chose slower responses, which had the detrimental effect of necessitating more movements. Interestingly, the tendency to perform actions that are too slow was more apparent in difficult conditions [Mazzoni et al. (2007), their Fig. 9 ( org/cgi/content/full/27/27/7105/f9)], in line with results from dopamine-depleted rats, which showed more prominent response-rate impairments for more difficult tasks (Salamone and Correa, 2002). This may reflect stronger reliance on implicit ( habitual ) dopamine-dependent mechanisms of response choice when tasks are more difficult, whereas simple tasks may be more amenable to explicit ( goal-directed ) mechanisms that are relatively dopamine-independent. It may also reflect a learning deficiency: daily movements typically fall within a certain comfortable range of accelerations and velocities, with which subjects are likely to have had much experience. The difficult conditions were less similar to highly practiced movements, perhaps necessitating online learning through feedback. Because this learning presumably relies on dopaminergic signaling in the basal ganglia (Montague et al., 1996), impaired learning in PD patients may explain why they performed worse in these conditions. To shed additional light on the relationship between dopamine depletion and bradykinesia, it would be interesting to study how behavior in this paradigm differs between patients in their on and off therapy states, and in different stages of the disease. This is especially pressing given that the patients in this study were in the early stages of the disease [Mazzoni et al. (2007), their Table 1 ( jneurosci.org/cgi/content/full/27/27/ 7105/T1)], and half were taking dopamine agonists at the time of testing. Of course, abnormal neural activity is observed in the basal ganglia even with dopamine replacement treatment (Heimer et al., 2006), complicating interpretations regarding the role of dopamine. Another question for future research is whether PD also affects the profile of motor actions such that there is a more limited repertoire of movements available to PD patients [as suggested by Figs. 3 (