Neural Mechanisms Underlying the Control of Dynamic Muscle Contractions in Human. Spinal Cord Injury

Transcription

1 Neural Mechanisms Underlying the Control of Dynamic Muscle Contractions in Human Spinal Cord Injury BY HYOSUB KIM B.M., The Juilliard School, 2000 D.P.T., University of Illinois at Chicago, 2012 THESIS Submitted as partial fulfillment of the requirements for the degree of Doctor of Philosophy in Neuroscience in the Graduate College of the University of Illinois at Chicago, 2015 Chicago, Illinois Defense Committee: Daniel M. Corcos, Chair and Co-Advisor T. George Hornby, Advisor Charles (CJ) Heckman, Northwestern University Sangeetha Madhavan William (Zev) Rymer, Northwestern University

2 To the Kim and Lee families, most of all my wife Christine ii

3 ACKNOWLEDGMENTS First and foremost, I would like to thank George Hornby for his excellent insights, guidance, and unwavering support throughout my training, and for emphasizing the importance of translational research. Next, I thank Daniel Corcos for always leading by example, and for teaching me to be rigorous in not only my thinking, but in my words and writing as well. Special thanks to the other three members of my committee, Charles (CJ) Heckman, Sangeetha Madhavan, and William (Zev) Rymer for their remarkable knowledge and dedication. Thanks also go to John Larson for never being too busy and always providing his best advice; Zia Hasan for his nonpareil lectures on biomechanics and motor control; Lynn Rogers for being generous with her time and expertise; past and present members of the Locomotor Recovery Laboratory, especially Chris Thompson and Kristan Leech for answering all of my questions and sharing all that they know; and, finally, to the Foundation for Physical Therapy for their advocacy and generous support. -HEK iii

4 TABLE OF CONTENTS 1. INTRODUCTION Overview Spinal cord injury: prevalence and etiology Mechanisms of injury Voluntary muscle activation impairments Quantifying voluntary activation Spinal reflexes Spasticity Dynamic maximal voluntary contractions Spinal reflex pathways and motoneuron properties Reciprocal inhibition Presynaptic inhibition Autogenic Ib inhibition Recurrent inhibition Plateau potentials Supraspinal adaptations Conclusion MUSCLE ACTIVATION VARIES WITH CONTRACTION MODE IN HUMAN SPINAL CORD INJURY Introduction Methods Subjects Experimental set-up Experimental protocols Data collection and analysis Statistical analysis Results Motor output and central activation during single MVCs Motor output during repeated anisometric contractions Washout and added bout of repeated isometric contractions Discussion Enhanced central motor drive during eccentric contractions in human SCI Differential motor output during repeated anisometric contractions Clinical significance INCREASED SPINAL REFLEX EXCITABILITY IS ASSOCIATED WITH ENHANCED CENTRAL ACTIVATION DURING VOLUNTARY LENGTHENING CONTRACTIONS IN HUMAN SPINAL CORD INJURY...53 iv

5 TABLE OF CONTENTS (continued) 3.1 Introduction Methods Subjects Experimental set-up Experimental protocols Data collection and analysis Statistical analysis Results Torque Central activation Co-activation Muscle mechanical properties Passive H-reflexes Active H-reflexes Correlations between spinal reflex excitability and central motor drive Clinical correlations Discussion Greater central activation during lengthening MVCs in human SCI Spinal reflex excitability Contribution of spinal reflex excitability to increased central activation Influence of sensory inputs on motor commands post-sci Conclusion SUPRASPINAL CHANGES FOLLOWING SPINAL CORD INJURY CONTRIUTE TO ALTERED ACTIVATION STRATEGIES DURING DYNAMIC CONTRACTIONS Introduction Methods Subjects Experimental set-up Experimental protocol Data collection and analysis Statistical analysis Results Active motor threshold MEP latency v

6 TABLE OF CONTENTS (continued) Pre-stimulus EMG Corticospinal excitability Cortical silent period Discussion Decreased corticospinal excitability MEP modulation Intracortical inhibition Limitations Conclusion CONCLUSIONS Chapter 2: Patterns of central motor drive Chapter 3: Spinal mechanisms Chapter 4: Supraspinal mechanisms Future directions Concluding remarks CITED LITERATURE APPENDIX VITA vi

7 LIST OF TABLES Table 2.1. SCI subject demographics...28 Table 2.2. SCI and control subject data during single MVCs...40 Table 3.1. SCI subject demographics...60 Table 4.1. SCI subject demographics...97 vii

8 LIST OF FIGURES Figure 1.1. Schematic representation of processes underlying voluntary activation in a healthy nervous system and one post-incomplete SCI....8 Figure 2.1. Schematic representation of experimental protocols...33 Figure 2.2. Representative single subject data during KE CAR trials...39 Figure 2.3. SCI subjects demonstrate increased KE motor output during eccentric MVCs...41 Figure 2.4. Single subject data demonstrating differences in torque development during repeated dynamic contractions...44 Figure 2.5. SCI subjects produce increasing KE torque and EMG during repeated concentric, but not eccentric, MVCs...45 Figure 2.6. Correlations between gains in KE torque and EMG...47 Figure 3.1. Representative data during PF MVCs...72 Figure 3.2. Evoked twitches and group data from CAR trials...75 Figure 3.3. Passive H-reflexes...78 Figure 3.4. H-reflexes during 75% MVC contractions...81 Figure 3.5. Correlations in SCI subjects between spinal reflex excitability during passive muscle lengthening and central activation during lengthening MVCs...83 Figure 4.1. Correlation between MEP peak-to-peak amplitudes and MEP areas Figure 4.2. Representative single subject MEP data during submaximal contractions Figure 4.3. Group data during 75% MVC and MVC dynamic contractions at 150AMT stimulation intensity Figure 4.4. Representative single subject data of CSP durations elicited at 150AMT during submaximal dynamic contractions Figure 4.5. Group CSP data during submaximal and maximal dynamic contractions viii

9 LIST OF ABBREVIATIONS 5-HT AIS AMT ANOVA ASIA CAR CNS CSP DF DTR EMG FDI GABA Hmax Hmax/Mmax HPAD H-reflex ITT KE KF LEMS Serotonin ASIA Impairment Scale Active Motor Threshold Analysis of Variance American Spinal Injury Association Central Activation Ratio Central Nervous System Cortical Silent Period Dorsiflexors Deep Tendon Reflex Electromyography (Electromyographic Activity) First Dorsal Interosseous Muscle Gamma-Aminobutyric Acid Maximal H-reflex Ratio of Maximal H-reflex to Maximal M-wave Homosynaptic Post-Activation Depression Hoffmann Reflex Interpolated Twitch Technique Knee Extensors Knee Flexors Lower Extremity Motor Scores ix

10 LIST OF ABBREVIATIONS (continued) MAS MEP MH Mmax MSO MU MVC M-wave PAD PF PIC RF RMS ROM SCATS SCI SD SEM SOL SSRI TA Modified Ashworth Scale Motor-Evoked Potential Medial Hamstrings Maximal Compound Muscle Action Potential Maximum Stimulator Output Motor Unit Maximal Voluntary Contraction Compound Muscle Action Potential Primary Afferent Depolarizing (Interneuron) Plantarflexors Persistent Inward Current Rectus Femoris Muscle Root-Mean-Square Range of Motion Spinal Cord Assessment Tool for Spastic Reflexes Spinal Cord Injury Standard Deviation Standard Error of the Mean Soleus Muscle Selective Serotonin Re-Uptake Inhibitor Tibialis Anterior Muscle x

11 LIST OF ABBREVIATIONS (continued) TMS VL VM Transcranial Magnetic Stimulation Vastus Lateralis Muscle Vastus Medialis Muscle xi

12 SUMMARY Spinal cord injury (SCI) severely diminishes an individual s ability to generate the requisite muscle force for purposeful, voluntary movements. Weakness and uncoordinated movement following SCI is caused by loss of descending inputs onto spinal neurons, reorganization of spinal circuits, and peripheral changes in muscle properties. Motor deficits are frequently accompanied by sensory impairments that likely contribute to altered patterns of voluntary muscle activation. Significantly, strength and voluntary muscle activation following incomplete SCI have been studied primarily during isometric contractions, where muscle length and joint angle are kept constant. Although isometric measurements are important, performance of functional movements, such as gait, requires combinations of lengthening (eccentric) and shortening (concentric) contractions. Hence, the focus of this dissertation was to address the gap in knowledge regarding the neural control of dynamic contractions in individuals with incomplete SCI. Three studies utilizing multiple non-invasive electrophysiological methods were conducted on a total of 14 individuals with chronic motor incomplete SCI. The first study examined central motor drive of the knee extensors during single and repeated lengthening, isometric, and shortening maximal voluntary contractions. Data collected during isolated maximal contractions suggest individuals with incomplete SCI demonstrate a markedly distinct pattern of muscle activation compared to healthy, uninjured control subjects. That is, despite overall deficits in voluntary muscle activation, SCI subjects generated markedly increased central motor drive during lengthening compared to isometric or shortening maximal contractions. In contrast, control subjects demonstrated a depression of motor drive during lengthening contractions relative to the other contraction types. Additional results also suggest xii

13 SUMMARY (continued) separate control mechanisms in SCI subjects for repeated maximal efforts during lengthening versus isometric or shortening contractions. Here, SCI subjects demonstrated progressive gains in peak torque and agonist muscle activity during repeated isometric and shortening maximal contractions separated by short rests, whereas during lengthening contractions the same measures did not improve with subsequent efforts. The second study examined spinal mechanisms and provided evidence that the facilitation of motoneuron activity during lengthening contractions demonstrated by SCI subjects is partly due to increased stretch reflex excitability within agonist muscles. The third study examined supraspinal contributions to the control of dynamic contractions. In this study, specific changes in intracortical inhibitory circuits involved in the control of dynamic contractions were observed in SCI subjects. Combined, these studies suggest voluntary muscle activation is facilitated during lengthening contractions by increased stretch reflex inputs, and that supraspinal modulation of descending commands is also altered following chronic SCI. Findings from this dissertation may therefore provide the physiological basis for more targeted rehabilitation interventions. xiii

14 CHAPTER 1 INTRODUCTION 1.1 Overview Spinal cord injury (SCI) results in severe motor and sensory impairments that lead to loss of functional mobility and decreased participation in daily activities. Approximately 12,000 new cases of SCI occur annually in the United States (2014). Depending on the level, severity, and time since injury, the average yearly costs directly attributable to SCI ranges from $40,000 - $1,000,000 per individual (2014). These expenses are inversely related to the individual s level of motor function. Therefore, addressing profound strength deficits following motor incomplete SCI is a primary goal for patients, clinicians, and researchers. Weakness and paralysis following SCI is primarily due to disruption of descending inputs onto spinal neurons as well as changes to intrinsic muscle properties (Thomas et al. 2014; Thomas et al. 1997). Deficits in volitional force generation are often compounded by sensory impairments that contribute to altered patterns of voluntary muscle activation. Specifically, increased stretch reflex excitability, or spasticity, as well as other spastic motor behaviors (e.g., spasms, clonus, increased tone), often develop in the subacute and chronic stages of SCI and can result in involuntary contractions and prolonged spasms in agonist, antagonist, and/or remote muscle groups (Gorassini et al. 2004; Murray et al. 2010; Nielsen et al. 2007). While these signs of an upper motor neuron lesion are frequently treated with medications intended to decrease spinal excitability, the functional consequences of spasticity are still unclear (Nielsen et al. 2007). Furthermore, there are data which suggest spasticity may actually facilitate voluntary movements (Angeli et al. 2012; Hornby et al. 2009). 1

15 2 Voluntary muscle activation and altered reflex pathways following neurologic injury have been studied primarily during isometric contractions, where the muscle length and joint angle are kept constant. However, as functional movement, such as gait, requires shortening (concentric) and lengthening (eccentric) contractions, there has been a gap in knowledge regarding the neural control of dynamic contractions in individuals with incomplete SCI. The work presented in this dissertation is an attempt to address this gap in knowledge. The objectives of this project were to characterize patterns of muscle activation during dynamic contractions of several lower limb muscle groups and to elucidate specific spinal and supraspinal mechanisms underlying their neural control in individuals with incomplete SCI. Chapter 2 is an investigation into patterns of central motor drive during lengthening, isometric, and shortening maximal voluntary contractions (MVCs) of the knee extensors. An additional component of this initial study was to examine volitional fatigue during the different contraction types. Chapter 3 investigated spinal mechanisms which may contribute to specific patterns of central motor drive in incomplete SCI patients. Of particular interest was how the effectiveness of Ia-α motoneurons transmission is modulated across different contraction types, and how spinal reflex modulation differed between SCI and healthy subjects. Chapter 4 examined potential supraspinal changes following chronic SCI that may contribute to specific patterns of activation during lengthening and shortening contractions. The aim of this first chapter is to review chronic changes to spinal reflex pathways as well as supraspinal changes that may contribute to the specific control of dynamic muscle contractions. The chapter will begin with a brief description of SCI. Following this, methods of quantifying central motor drive to muscle and findings regarding the normal control of dynamic

16 3 contractions are described. In the discussion of spinal reflexes, special emphasis will be placed on pathways implicated in the pathophysiology of spasticity. 1.2 Spinal cord injury: prevalence and etiology Of the approximately 12,000 new cases of SCI which occur annually in the United States, 39% are caused by motor vehicle accidents (2014). Falls, violence, and sports related injuries are the next 3 most common causes of spinal lesion. SCI occurs at various anatomical and neurological levels (i.e., lowest segment where motor and sensory function is normal on both sides), and the severity of injury is also variable. This dissertation was focused on individuals with cervical or thoracic SCI who are classified according to the American Spinal Injury Association (ASIA) Impairment Scale (AIS) as motor incomplete AIS C or D. AIS grades C and D indicate preservation of some motor function below the neurological level with either less than half (AIS C) or at least half (AIS D) of the key muscles below the neurological level being able to generate movement against gravity throughout their respective joint range of motion (Maynard et al. 1997). 1.3 Mechanisms of injury The pathophysiology of SCI results from primary and secondary mechanisms of injury. The primary injury is the physical damage to the spinal cord tissue that can result from compression, traction, and/or transection (Dumont et al. 2001). Consequences of the primary injury include direct damage of neuronal cell bodies, axons, and vasculature supplying nerve cells. Secondary mechanisms of injury are often the result of the body s response to the initial mechanical injury. Neurogenic shock, vascular insults, ionic imbalances, apoptosis, and

17 4 pathologic immune responses can all contribute to an exacerbation of the initial injury (Dumont et al. 2001; Hagen 2015; Webb et al. 2010)). 1.4 Voluntary muscle activation impairments Weakness and paralysis are the hallmark features of SCI. At the muscular level, atrophy, alterations in fiber phenotypes, and increased fatigability can negatively affect force-generating capacity (Cope et al. 1986; Martin et al. 1992; Shields 1995b). These detrimental peripheral changes are compounded by deficits in excitation of lower motor pools by descending inputs due to disruption and/or demyelination of corticospinal, bulbospinal, and vestibulospinal pathways (Lin et al. 2012; Thomas et al. 2014; Thomas et al. 1997). For example, Thomas et al. showed that stimulating the radial nerve while subjects performed an isometric maximal voluntary contraction (MVC) of the elbow extensors elicited a large interpolated twitch from the triceps brachii of SCI subjects, but not in healthy control subjects. The size of the evoked twitch is commensurate with activation deficits, as it reflects the degree to which central motor drive is an insufficient source of motor pool excitation (Merton 1954). Another telling portion of the Thomas study, however, was that stimulation over the contralateral motor cortex using transcranial magnetic stimulation (TMS) evoked significantly smaller twitches than peripheral nerve stimulation did in SCI subjects. This indicated that even though transmission to the motoneurons was significantly impaired, motor cortical drive was not impaired, as the TMS pulse did not evoke substantial twitches during MVCs. 1.5 Quantifying voluntary activation

18 5 The method described above for the Thomas et al. study is referred to as the interpolated twitch technique (ITT), a valid and reliable method of assessing central motor drive (Gandevia 2001; Taylor 2009). Stimulation is most frequently provided transcutaneously over the peripheral nerve or muscle. Twitches can be interpolated by providing stimulation in the form of a single supramaximal pulse, multiple pulses separated by short interpulse intervals, or tetanic stimulation, all while the subject is performing an MVC (Shield and Zhou 2004). Two formulas are most frequently used to quantify a percentage of voluntary, or central, activation. First, one can calculate a percentage of voluntary activation by comparing the evoked twitch during the contraction (test twitch) to a twitch evoked from the resting muscle (control twitch, which is often potentiated by a previous contraction) (Allen et al. 1995; Gandevia et al. 1998): Voluntary activation (%) = (1 interpolated twitch ) x 100 control twitch (Equation 1.1) The other frequently used formula is the central activation ratio, or CAR, value (Kent-Braun and Le Blanc 1996). The advantage of using the CAR is that the number of stimulations required is reduced as there is no comparison to a control twitch. Central Activation Ratio (%) = (peak voluntary torque) (peak voluntary torque+peak stimulated torque) x 100 (Equation 1.2) Using the ITT, incomplete SCI subjects demonstrate significant deficits in voluntary activation of multiple muscles, including the triceps brachii (Thomas et al. 1997) and the flexor carpi radialis (Lin et al. 2012) in the upper limbs, and the quadriceps (Hornby et al. 2009) and plantarflexors (Jayaraman et al. 2006) in the lower limbs. Regardless of which equation was used, group means for activation of the lower extremity muscles were anywhere between 34-58% for incomplete SCI subjects, while under matched conditions it was approximately 95% for

19 6 controls. Interestingly, although SCI subjects in the Jayaraman et al. study demonstrated profound deficits in voluntary activation, the mechanical properties of the muscle (peak twitch torques, time to peak twitch, and half-relaxation times) did not appear altered by chronic SCI. Data derived from spinally transected cats also describe similar twitch forces compared to control animals in the triceps surae (Cope et al. 1986; Gallego et al. 1978). This finding in AIS C and D patients suggests that in certain muscles disrupted descending signaling may play a larger role in strength deficits than changes to muscle properties. These studies have all used isometric testing, as strength has traditionally been defined in experimental settings as the force, or torque, produced during an isometric MVC. However, these assessments constrain movement about the tested joint. Functional movement, on the other hand, is comprised of dynamic contractions which result in rotation about the joint. Hence, substantial information regarding voluntary movement post-sci can be gained by studying simple single-joint contractions where the muscles are allowed to shorten and lengthen due to whether the muscle torque exceeds the load (concentric), or if the load exceeds the muscle torque (eccentric) (Enoka 1996). As further discussed later, the study of dynamic contractions in SCI can also provide important information regarding the integrity of spinal reflex pathways, which are of paramount importance in the production of voluntary movement. 1.6 Spinal reflexes In 2009, Taylor succinctly wrote the following: descending drive from the motor cortex is the major determinant of the timing and strength of voluntary contractions (Taylor 2009). While true, the importance of spinal reflexes to voluntary movement was exquisitely detailed in the seminal work of Sir Charles Sherrington dating over a century ago (Sherrington

20 7 1966). Anders Lundberg in the latter half of the 20 th century further extended Sherrington s findings by focusing on the convergence of sensory afferent input and descending tracts onto interneurons interposed in spinal reflex pathways (Hultborn 2006). Since then, extensive research in humans has also shown that descending motor drive is integrated with afferent feedback signals from the periphery (i.e., muscles, joints, and skin) by spinal interneurons common to both (Nielsen 2004). In the case of SCI, afferent inputs may have a greater influence on the strength of a voluntary contraction than normal. Zijdewind and colleagues demonstrated this by providing different forms of concurrent sensory stimulation to SCI subjects performing isometric MVCs of the thenar muscles (Zijdewind et al. 2012). Heat, vibration, induced spasm, or a contralateral contraction successfully increased maximal firing rates of the tested thenar motor units (MUs), suggesting peripheral inputs can further excite the voluntarily activated motor pool. Across all subjects in this study, the stimulation that resulted in the greatest increases in MU firing rates was an induced spasm. A spasm is a sustained, involuntary, multi-joint contraction that can occur in response to various sensory inputs (Gorassini et al. 2004). The investigators either lifted the subject s knee or pushed the subject s shoulder backwards to cause a spasm, which in turn increased thenar MU activity.

21 Figure 1.1. Schematic representation of processes underlying voluntary activation in a healthy nervous system and one post-incomplete SCI. In both cases, reflex inputs synapse directly onto motoneurons and/or onto spinal interneurons. There is often convergence of spinal reflex and descending inputs onto common spinal interneurons. Hence, this illustrates the integrative actions of the central nervous system, as well as the importance of spinal modulation on motor commands. Another common theme throughout this dissertation is illustrated specifically in the incomplete SCI diagram: there is a loss of descending inputs and concomitant gain in importance of reflex inputs on shaping the motor output. 8

22 9 1.7 Spasticity Spasms are one specific type of increased motor activity that develop following an upper motor neuron lesion that are frequently referred to collectively as spastic hypertonia or spastic motor behaviors (Benz et al. 2005; Sheean and McGuire 2009). Spasticity is a component of spastic hypertonia which specifically refers to the velocity-dependent increases in stretch reflex excitability and exaggerated tendon tap responses (Lance 1980) that develop in the subacute and chronic stages of injury. The prevalence of spasticity, or stretch reflex hyperexcitability, among individuals with SCI is purportedly between 67-78% (Maynard et al. 1990). The term spasticity in this dissertation refers specifically to hyperexcitability of the stretch reflex, as originally proposed by Lance, in order to avoid conflating the central and peripheral (e.g., muscle stiffness/contracture) properties contributing to increased muscle tone. Furthermore, the following discussion of spasticity will include data from patients with various neurologic disorders, including cerebral palsy, stroke, and multiple sclerosis. Wherever findings regarding spasticity between SCI patients and those with other neurologic impairments diverge will be specified. Opinions regarding the functional consequences of spasticity vary widely. Slower and less coordinated movements have been attributed to its presence (Damiano et al. 2006; Tsao and Mirbagheri 2007), while others suggest spasticity may actually be an adaptive change to spinal circuitry post-injury. This latter viewpoint is supported by data that suggest patients may learn to use their spasticity to perform functional activities, such as standing and walking (Dietz 2001; Guttmann 1954). The physiological basis for these results, again, is that the final motor command results from integration of supraspinal and sensory inputs by inter- and motoneurons. These converging

23 10 pathways were first demonstrated using intracellular recordings in the cat (Baldissera et al. 1981) and later in humans, using non-invasive electrophysiological methods (Tanaka 1974). Based on this understanding, there are numerous sensory inputs that influence the activity levels of motoneurons, with a net reduction in motor output when sensory information is reduced. As such, spasticity may develop in order to compensate for reduced supraspinal inputs. Furthering this argument, reduction of sensory feedback by anti-spasticity medications, such as diazepam (GABA A positive allosteric modulator), baclofen (GABA B agonist), and tizanidine (α 2 adrenergic agonist) appear to not only reduce reflex hyperexcitability, but also reduce strength and may ultimately impede patients ability to perform functional movements (Hornby et al. 2004; Nielsen et al. 2007; Thomas et al. 2010). Interestingly, although stretch reflexes appear to be hyperactive at rest in spastic patients, they may also be less potentiated during voluntary movement involving low-to-moderate level submaximal voluntary contractions in comparison to able-bodied subjects (Morita et al. 2001). While it is normal for stretch reflexes in activated muscles to be potentiated, there appears to be little difference in amplitude of stretch reflexes between spastic and healthy subjects during submaximal contractions (Dietz et al. 1991; Toft et al. 1993; Woolacott and Burne 2006a). A caveat to these data is that they utilized submaximal contractions. There is recent evidence that spinal neuron excitability may be potentiated in incomplete SCI subjects during times of high volitional drive. As one example, peak knee extensor (KE) torques acutely increased with repeated isometric maximal voluntary contractions (MVCs). This behavior was attributed to the presence of increased inter- and motoneuronal excitability (Hornby et al. 2009; Thompson et al. 2011b), which was inferred from motor behaviors consistent with plateau potentials and persistent inward currents (discussed below). The relevance of these findings to

24 11 our current understanding of spasticity is that although stretch reflex gain may not be abnormally enhanced during submaximal contractions, we do not yet know how reflex gain is modulated during MVCs, and what the functional impact of abnormally increased reflex gain is. 1.8 Dynamic maximal voluntary contractions As noted earlier, important information regarding the integrity of spinal reflex pathways may be more readily extracted from studies of dynamic contractions. Certainly in cases where the functional relevance of spasticity on movement is involved, including dynamic contractions would allow for more direct translation of results to functional movements. It is therefore surprising that there has been only one study prior to the work presented in this dissertation that examined maximal activation during dynamic contractions in SCI subjects. Knutsson et al. studied dynamic lengthening and shortening MVCs of both the KEs (quadriceps) and knee flexors (KFs; i.e., hamstring muscles) in subjects with spastic paraparesis (Knutsson et al. 1997). This particular study had a heterogeneous population of spastic patients which included individuals with multiple sclerosis, hereditary spastic paraparesis, as well as incomplete SCI. Torque and electromyography (EMG) were used as measures of voluntary activation during dynamic MVCs, and there were no evoked twitches used to more directly quantify activation. The main finding from this study was that spastic patients demonstrated overactive antagonist muscles and decreased agonist activity during fast shortening MVCs, which severely limited concentric torque generation. The authors therefore posited hyperexcitable stretch reflexes of antagonist muscles and pathologically increased activity in the disynaptic Ia reciprocal inhibitory pathway (see below) from antagonist to agonist as neurophysiological mechanisms to explain their finding. Another interesting finding from this study was that agonist

25 12 EMG during lengthening MVCs of the knee flexors was significantly increased compared to during shortening MVCs. Therefore, while spastic antagonists may limit agonist activation, increased stretch reflex excitability within agonist muscles may facilitate Ia excitation of homonymous motoneurons. While these are both provocative hypotheses, they warrant further testing as these results came from a mixed population, and there are etiology-dependent differences in spastic motor behaviors (Burne et al. 2005a; Woolacott and Burne 2006b). In contrast to SCI subjects, healthy controls in this study did not demonstrate an abnormal increase in antagonist activity during fast or slow shortening MVCs. As a group, controls demonstrated significant increases in agonist EMG activity during fast shortening MVCs compared to slower shortening or lengthening MVCs, consistent with previous findings (Komi et al. 1987; Westing et al. 1991b). While the Knutsson et al. study provided an important gross comparison of motor behaviors during dynamic contractions between healthy and neurologically impaired subjects, there remain many questions regarding the neural mechanisms underlying the control of dynamic contractions in SCI. The following section will detail spinal reflex pathways that most likely contribute to specific neural activation strategies in individuals with incomplete SCI, as most if not all have been implicated in the pathophysiology of spasticity. The focus will be on how activity in these pathways is normally modulated across different contraction types. Emphasis will be placed on their activity during lengthening contractions, as in healthy adults lengthening contractions are thought to require distinct activation signals that separate them from the more consistently similar patterns of shortening and isometric muscle activation (Enoka 1996). Finally, for each pathway, we will discuss how specific alterations to them following SCI may contribute to distinct patterns of dynamic muscle activation.

26 Spinal reflex pathways and motoneuron properties Reciprocal inhibition The functional purpose of the disynaptic reciprocal Ia inhibitory pathway is to reduce activity in the antagonists when the agonists are voluntarily activated. In this pathway, Ia afferents synapse onto the Ia inhibitory interneuron which has connections to the opposing muscle group s motor pool. Supraspinal inputs also converge onto these interneurons, and there is a reciprocal organization (i.e., Ia afferents from muscles on either side of a joint can cause inhibition of the opposing muscle group). Presumably, given equal levels of central motor drive to the agonist muscles, different levels of reciprocal inhibition may affect the total force output and measures of central activation. Because the proposed project will utilize non-invasive electrophysiological techniques, and this pathway still represents the most ideal illustration of the way that noninvasive techniques may be used to study the control of spinal interneuronal pathways during movement (Nielsen 2004), we will describe in some detail the evolution of our understanding of reciprocal inhibition. Direct recordings in this pathway were first performed in the cat using intracellular electrodes (Hultborn et al. 1971). The advantage of invasive recordings from animal models is that they provide robust characterization of the input-output organization of interneuronal networks; however, their primary limitation is that they are much less powerful in characterizing the functional role of these networks during voluntary movement. Due to the experimental approach typically requiring the animal to be either anesthetized or decerebrated, recording during volitional movement becomes impossible. The technical difficulty of identifying appropriate interneurons in the intact behaving animal, where extraneous movement of the

27 14 electrodes occurs, also limits the ability of these procedures to characterize spinal network activity during movement. Therefore, the development of non-invasive methods to indirectly examine analogous pathways during movements in fully conscious humans has significantly advanced our understanding of this topic (Nielsen et al. 2007). The classical method of investigating spinal modulation in humans is through assessment of the Hoffmann reflex (H-reflex), which is the electrical analogue of the monosynaptic stretch reflex (Hoffmann 1910). By transcutaneously stimulating Ia afferent fibers coming from muscle spindles, one can assess Ia synaptic input onto motoneurons and, maintaining the stability of certain conditions, the excitability of the motor pool (Zehr 2002). Tanaka first tested the reciprocal inhibitory pathway in humans by using an inventive paired H-reflex testing paradigm (Tanaka 1974). Briefly, the Ia fibers of the agonist muscle group are stimulated first, and then immediately followed by stimulation of the antagonist Ia afferents. The size of the conditioned H-reflex in the antagonist is decreased compared to the unconditioned, or control, H-reflex amplitude if stimulations are performed at very short latencies. Both the latency and the low threshold for inhibition suggest it is mediated by the disynaptic reciprocal Ia pathway. If we examine the ankle joint, active dorsiflexion will result in inhibition of the conditioned soleus H- reflex. This inhibition is also evident prior to any EMG activity in the dorsiflexors, indicating a supraspinal influence on this pathway (Hallett 2007). To date, the role of reciprocal inhibition in the normal neural control of dynamic contractions has not been investigated using the conditioned H-reflex (Duchateau and Baudry 2014). Instead, the level of antagonist EMG has been used to infer differences in reciprocal inhibition across modes of contraction (i.e., greater antagonist EMG under one condition would indicate decreased reciprocal inhibition, resulting in greater co-activation). As co-activation

28 15 levels are frequently reported as being no different in able-bodied subjects during lengthening and shortening contractions, it was assumed that this pathway does not play a significant role in normal patterns of activation (Abbruzzese et al. 1994; Duclay and Martin 2005; Duclay et al. 2011; Nordlund et al. 2002). However, with regards to individuals with SCI, there is evidence that reciprocal inhibition is reduced in spastic SCI patients (Crone et al. 2003a). Crone et al. found that prior stimulation of the peroneal nerve inhibited the conditioned soleus H-reflex response in healthy subjects, but facilitated it in spastic SCI and stroke subjects. This was interpreted as a sign of reciprocal facilitation from ankle dorsiflexors to plantarflexors. Interestingly, though, it has been shown that there is no correlation between the degree of spasticity (as clinically assessed by the Ashworth scale) and the degree of reduced reciprocal inhibition, further pointing to issues of validity with regards to clinical tests of spasticity (Crone et al. 1994). This study provided convincing evidence of chronically altered activity within the Ia reciprocal inhibitory pathways. One limitation in extrapolating these data to dynamic contractions, especially MVCs, however, is that reflexes were evoked during passive isometric testing. Taking this into consideration, a study of dynamic knee extensor MVCs in the stroke population by Clark et al. may be particularly instructive (Clark et al. 2006). There, the authors found that in chronic stroke subjects, clinically spastic antagonists did not act as a restraint on force production or agonist activation during shortening contractions. They concluded that there was no evidence of abnormal reciprocal inhibition of agonist motor pools. Although both the Crone and Clark studies tested similar patient populations, their disparate conclusions regarding the role of reciprocal inhibition may stem from differences in passive versus active testing,

29 16 consistent with the discussion of stretch reflex gain in spastic populations from the preceding section Presynaptic Inhibition There are inhibitory synapses at both pre- and post-synaptic sides of the motoneuron which can limit its discharge rate. Reciprocal inhibition is an example of post-synaptic inhibition of the motoneuron. Pre-synaptic inhibition, on the other hand, refers to axo-axonic inhibition of Ia afferents at the Ia-α motoneuron synapse by a last order primary afferent depolarizing (PAD) interneuron (Eccles 2013; Pierrot-Deseilligny and Burke 2005). Excitatory and inhibitory inputs onto first order PAD interneurons originate from multiple sources, including Ia and Ib afferents, cutaneous afferents, as well as the vestibulospinal and corticospinal tracts (Pierrot-Deseilligny and Burke 2005). Presynaptic inhibition in humans has been investigated by measuring Ia-α motoneuron transmission using H-reflexes. The H-reflex is normally decreased during both submaximal and maximal lengthening contractions in healthy adults (Duclay and Martin 2005; Duclay et al. 2011; Duclay et al. 2014; Pinniger et al. 2001; Romanò and Schieppati 1987), compared to during isometric and muscle shortening conditions. Since H-reflex amplitude is lower during passive muscle lengthening, as well as during lengthening contractions, when muscle spindle feedback is augmented, some of the depression is attributable to presynaptic inhibition (Duchateau and Baudry 2013). Furthermore, the depression of the H-reflex suggests a reduction in the excitability (disfacilitation) of the motor pool during lengthening contractions. As Ia inputs onto motoneurons comprise the main excitation of spasticity hence, the velocity dependence we hypothesized that facilitation of the Ia-α motoneuron pathway would contribute to increased central motor drive during lengthening contractions in SCI subjects.

30 17 The difficulties with accurate diagnosis of spasticity and the controversy surrounding its functional significance highlight the importance of empirical investigation of specific spinal pathways during both isometric and dynamic contractions. As described earlier, H-reflexes are depressed during passive and active muscle lengthening in able-bodied subjects when compared to isometric and shortening muscle actions. Enoka, Duchateau, and Baudry have all linked this specific depression of the effectiveness of Ia-α motoneuron transmission during lengthening contractions to the difficulty healthy adults frequently demonstrate with maximally activating their muscles during lengthening MVCs (Duchateau and Baudry 2014; Duchateau and Enoka 2008). Based on this, we propose that an increase in the efficacy of transmission during lengthening could result in increased motor unit recruitment, firing frequency, or both in SCI Autogenic Ib inhibition The depression of maximal activation during lengthening MVCs in healthy adults was originally hypothesized to be due to a tension-limiting mechanism. Specifically, it was thought that due to the biomechanical properties that contribute to higher forces during muscle lengthening (Lombardi and Piazzesi 1990), autogenic Ib inhibition from Golgi tendon organs may prevent accidental injuries from excessively high forces (Aagaard et al. 2000; Westing et al. 1991b). This hypothesis is most likely inaccurate, though, as Pinniger et al. found that there was a similar relative depression of torque during both maximal and submaximal lengthening contractions (Pinniger et al. 2000). Additionally, it appears muscle length, and not the tension produced, contributes most to muscle damage during lengthening contractions (Talbot and Morgan 1998). Given these data, it appears unlikely autogenic Ib inhibition plays a significant

31 18 role in the modulation of dynamic muscle activation following SCI, especially considering the overall decrease in force-generating capacity exhibited by SCI subjects Recurrent Inhibition Maximal discharge rates of motor units (MUs) are usually reduced during lengthening contractions compared to during isometric and shortening contractions. Several studies have shown that whether sub-maximal lengthening and shortening contractions are matched for equivalent absolute or relative levels of torque, the firing rates of MUs are reduced during lengthening contractions performed by healthy adults (Del Valle and Thomas 2005; Pasquet et al. 2006). Also, in the single published study of maximal dynamic MVCs that successfully obtained MU recordings, discharge rates were significantly lower during lengthening than shortening MVCs (Del Valle and Thomas 2005). Recurrent (Renshaw cell) inhibition during muscle lengthening may contribute to this reduction in MU firing rate (Duchateau and Baudry 2014). The recurrent inhibitory pathway includes excitation of an inhibitory interneuron, or Renshaw cell, from an α motoneuron axon collateral (Hultborn et al. 1971). Renshaw cells project back to the same α motoneurons, or motor pool, they were excited by and release the inhibitory neurotransmitter glycine (Curtis et al. 1976; Eccles et al. 1954). Although recurrent inhibition is thought to be increased following SCI (Shefner et al. 1992), modulation of this pathway during different contraction types has not yet been experimentally tested in either healthy or neurologically impaired adults Plateau potentials

32 19 Finally, in addition to changes in specific spinal circuits, there is strong evidence of changes to intrinsic motoneuronal properties post-sci. Plateau properties, which refer to prolonged depolarizations (plateau potentials) that are the result of voltage-dependent persistent inward Ca 2+ and Na + currents, may develop in inter- and motoneurons during the chronic state of SCI (Murray et al. 2010). The presence of plateau potentials signifies a decrease in the synaptic current necessary to elicit action potentials. They are facilitated by the presence of the monoamines serotonin and norepinephrine and are characterized by regenerative trains of action potentials that continue beyond the time synaptic current is removed. Bennett et al. demonstrated plateau properties in rat tail motoneurons in an in vitro preparation of chronic SCI (Bennett et al. 1999; Bennett et al. 2001). For spastic SCI patients, the functional relevance of plateau potentials is that motoneurons may essentially be primed to fire action potentials even due to innocuous stimuli, such as muscle stretch. Gorassini et al. published a report in humans that demonstrated MUs of SCI subjects required significantly less synaptic drive during derecruitment at the end of a muscle spasm as compared to at the beginning of one (Gorassini et al. 2004). They used these data to support the role of plateau potentials in spasms. The contribution of plateau potentials to spasticity, as characterized by velocity-dependent resistant to stretch, however, is unclear Supraspinal adaptations In addition to spinal mechanisms contributing to altered patterns of muscle activation following SCI, there is evidence to suggest changes above the level of injury may also contribute to specific muscle activation patterns. Analogous to the use of transcutaneous nerve stimulation and H-reflexes to probe specific spinal pathways, single pulse transcranial magnetic stimulation (TMS) is used to probe cortical and corticospinal excitability. The TMS pulse elicits a motor-

33 20 evoked potential (MEP), which is the relatively synchronous evoked motor response due to stimulation (Hallett 2007; Pierrot-Deseilligny and Burke 2005). MEPs are detected by EMG electrodes placed over the muscle(s) of interest, and MEP amplitude is a measure of corticospinal function. Intracortical inhibition is an additional neurophysiological measure that can be inferred through using TMS by measuring cortical silent periods (CSPs). The CSP is a pause in the voluntary EMG activity in response to TMS. The first portion of the CSP is due to spinal refractoriness, but the later portion (>100 ms) is due to GABA-ergic cortical inhibition (Werhahn et al. 1999). CSP duration is correlated to the degree of intracortical inhibition. Duclay et. al performed an important series of experiments examining both spinal and supraspinal contributions to the control of isometric and dynamic contractions within the synergistic soleus and gastrocnemius muscles in healthy adults (Duclay et al. 2011). The TMS portion of the study analyzed MEP amplitudes during lengthening, isometric, and shortening MVCs, as well as CSPs. Interestingly, while MEP amplitudes during lengthening MVCs were smaller compared to isometric and shortening MEPs, the corresponding CSP was shorter. These results were consistent with those of Gruber et. al (Gruber et al. 2009b), and both groups postulated that increased descending drive may be a compensatory mechanism for spinal inhibitory mechanisms which diminish the motor pool s output during lengthening MVCs. This interpretation is also consistent with a previous study which demonstrated a larger cortical area is devoted to the control of lengthening over shortening contractions (Fang et al. 2004). Although there are no studies in human SCI which parallel those of Duclay et al., one study examined the effects of pairing peripheral nerve stimulation with TMS. The results suggest sensory inputs from an activated muscle are modulated differently between SCI and healthy subjects. Roy et al. demonstrated that in incomplete SCI subjects, tibialis anterior MEPs were

34 21 facilitated by prior stimulation of homonymous afferents in the deep peroneal nerve (Roy et al. 2010). In healthy subjects, if heteronymous peripheral nerve stimulation preceded the TMS pulse by ~40 ms, there was facilitation of the MEP (i.e., increase in MEP amplitude, indicative of greater corticospinal excitability), presumably due to ascending afferent signals priming cortical regions. For the less impaired incomplete SCI subjects tested, however, homonymous peripheral nerve stimulation resulted in a significant increase in MEP size when using a 50 ms versus a 30 ms interstimulus interval that was not observed in healthy subjects. This suggests homonymous afferent signaling can facilitate corticospinal excitability, at least in those SCI subjects with better preserved corticospinal transmission. Additional evidence of adaptive central nervous system changes that occur after incomplete SCI comes from a recent study that demonstrated CSP durations increased following a fatiguing protocol in healthy control subjects, but there was no increase in the CSP durations of SCI subjects following an identical protocol (Nardone et al. 2013a). This suggests that inhibitory, most likely GABA-ergic (Werhahn et al. 1999), cortical mechanisms may be downregulated after SCI in an attempt to compensate for corticospinal damage. The relationship of this finding in SCI subjects to corticospinal excitability during isolated dynamic contractions is indirect. However, the study highlights the importance of investigating possible compensatory supraspinal changes that occur following SCI which may alter patterns of muscle activation Conclusion To date, most studies of human SCI have investigated deficits in central motor drive and altered spinal reflex pathways during isometric conditions. In order to achieve a better understanding of motor control following incomplete SCI, non-invasive electrophysiological

35 22 methods can be utilized during dynamic testing. The subsequent chapters of this dissertation will describe in detail unique patterns of muscle activation, modulation of spinal reflex excitability, and potential supraspinal contributions to specific activation strategies during dynamic contractions in subjects with SCI. Note: Chapter 2 has been previously published as the following: Kim HE, Thompson CK, and Hornby TG. Muscle activation varies with contraction mode in human spinal cord injury. Muscle Nerve 51: , Chapter 3 has been accepted for publication and is now in press: Kim HE, Corcos DM, and Hornby TG. Increased spinal reflex excitability is associated with enhanced central activation during voluntary lengthening contractions in human spinal cord injury. J Neurophysiol jn , 2015.

36 CHAPTER 2 MUSCLE ACTIVATION VARIES WITH CONTRACTION MODE IN HUMAN SPINAL CORD INJURY (Previously published: Kim HE, Thompson CK, and Hornby TG. Muscle activation varies with contraction mode in human spinal cord injury. Muscle Nerve 51: , 2015.) 2.1 Introduction Individuals with motor incomplete spinal cord injury (SCI) demonstrate impairments in the ability to generate the requisite muscle force for many purposeful, voluntary movements. In addition, these patients often present with spasticity, a term whose definition includes velocitydependent increases in tonic stretch reflexes and exaggerated tendon tap responses (Lance 1980). Impaired voluntary muscle activation contributes significantly to strength deficits (Lin et al. 2012), while adaptations in spinal neuronal circuitry contribute to spastic motor behaviors in animal models (Murray et al. 2010) and humans (Nielsen et al. 2007). In chronic SCI, central and peripheral alterations to the neuromuscular system are also thought to exacerbate fatigue, defined as the progressive decline in maximal force generating capability during repeated or sustained contractions (Gandevia 2001). Recent studies in individuals with incomplete SCI who performed 20 repeated isometric maximal voluntary contractions (MVCs) (5 s on/5 s off) of the knee extensors (KE) have not demonstrated substantial fatigue. Rather, SCI subjects generated 20-25% increases in peak torques and electromyographic (EMG) activity above single baseline MVCs performed prior to the repeated contractions, with the largest changes observed by the 3 rd - 4 th contraction (Hornby et al. 2009; Thompson et al. 2011b). In contrast, motor output declines rapidly after the 1st MVC in individuals without neurologic injury (Callahan et al. 2009; Hornby et al. 2009) as well as in individuals with SCI when contractions are elicited by electrical stimulation (Gerrits et al. 1999). 23

37 24 The term supramaximal has been used previously to describe these acute increases due to the level of volitional motor output going above what is commonly accepted as maximal torque generation (i.e., a single MVC in isolation from other contractions) (Braith et al. 1993; Ehrman and Medicine 2010). Increases in KE torque are correlated with changes in KE EMG amplitude and indirect measures of increased spinal neuron excitability, suggesting neural mechanisms underlie this phenomenon more than peripheral (i.e., muscular) changes (Thompson et al. 2011b). The findings of acute increases in volitional torques have been observed only during isometric testing, and questions remain as to whether similar behaviors are evident during anisometric (i.e., concentric and eccentric) contractions. Assessment of motor output during dynamic contractions may serve as a valuable model to better understand performance during functional tasks, which more often combine both shortening and lengthening contractions (Enoka 1996). This information is also clinically relevant, as there is increasing evidence from studies of healthy individuals that contraction mode may be a major determinant in the efficacy of resistance exercise (Kraemer et al. 1998; Norrbrand et al. 2008). Specifically, despite typically lower voluntary activation during eccentric versus concentric contractions (Babault et al. 2001; Del Valle and Thomas 2005; Pinniger et al. 2000; Westing et al. 1991b), eccentric training has been shown to increase strength more efficiently through enhanced muscular (Guilhem et al. 2010) and neural adaptations (Colliander and Tesch 1990). In patients with incomplete SCI, there are limited data detailing the efficacy of strength training protocols that employ specific concentric or eccentric activities. This may be due in part to the absence of studies in incomplete SCI that characterize baseline differences in motor output during anisometric contractions, particularly during repeated MVCs. Such characterization is

38 25 likely important, as dynamic MVCs may be affected substantially by stretch reflex sensitivity associated with spasticity. For isolated maximal concentric contractions, a study of individuals with spastic paraparesis of various etiologies, including 2 individuals with incomplete SCI, demonstrated reduced joint torque, which was attributed to spastic antagonist muscle stretch (Knutsson et al. 1997). Given these data and the previous finding of increasing motor output during repeated isometric MVCs by individuals with incomplete SCI (Hornby et al. 2009; Thompson et al. 2011b), the purpose of this study was two-fold: 1) to characterize baseline differences in volitional force generation during different modes of contraction; and, 2) to investigate volitional force generation during repeated concentric and eccentric MVCs. Neuromuscular stimulation was used to characterize baseline differences in volitional force generation across different modes of muscle contraction by quantifying central activation (i.e., activation due to processes within the central nervous system, including descending and sensory input) (Kent- Braun 1997). We hypothesized that in individuals with SCI, the magnitude of central muscle activation during single MVCs would be reduced compared to healthy individuals, but patterns of central activation observed across all contraction modes would be similar between subject groups. Also, we anticipated increases in motor output during repeated dynamic contractions would be similar to those observed previously during isometric efforts in individuals with incomplete SCI. 2.2 Methods Subjects Two separate experiments were performed in 11 individuals (1 woman) with chronic (> 1 year) motor incomplete SCI (mean age 46 years; range 27-64) recruited from the outpatient

39 26 clinics of the Rehabilitation Institute of Chicago. Eleven healthy, uninjured men (mean age 38 years; range: 25-53) were also recruited for the first experiment. All control subjects reported participating in at least a moderate level of physical activity on a regular basis. Experiments were performed on 2 separate days separated by at least 72 hours. Subjects with SCI were classified as either C or D using the American Spinal Injury Association (ASIA) Impairment Scale. Both ASIA grades indicate some preservation of motor function below the neurological level (i.e., lowest segment where motor and sensory function is normal on both sides), with either less than half (ASIA C) or at least half (ASIA D) of the key muscles below the neurological level being able to generate movement against gravity throughout their respective joint range of motion (Maynard et al. 1997). All SCI subjects demonstrated volitional knee extensor strength in at least 1 limb, as determined by the Lower Extremity Motor Score (Marino and Graves 2004). Of note, the majority of SCI subjects tested were community ambulators (see Table 2.1). Clinical assessment of spasticity was assessed utilizing a variation of the Modified Ashworth Scale (MAS) technique (Thompson et al. 2011a). As the instructions for the MAS do not include testing responses at different velocities (Bohannon and Smith 1987), in its original form it may be sensitive to muscle resistance but not velocity-dependent reflex responses. Therefore, the variation of the MAS used here involved assessment of spasticity in bilateral KEs by slowly taking the tested knee joint through its full passive range of motion prior to applying a quicker stretch through the range in order to determine MAS scores (Thompson et al. 2011a). The presence of other spastic motor behaviors (i.e., spasms) was evaluated using the Spinal Cord Assessment Tool for Spastic Reflexes (SCATS) (Benz et al. 2005), respectively. Subjects were also included if the tested limb demonstrated passive range of motion from knee flexion

40 27 (0 equals full knee extension) without pain and at least minimal amounts of voluntary KE torque throughout this range, as determined by online visual inspection of torque signals. Exclusion criteria included medical history of multiple CNS lesions, history of lower limb peripheral nerve injury, or orthopedic injury which would limit KE contractions. None of the subjects were using anti-spasticity medications at the time of this study. All subjects had prior experience using the testing apparatus. Written informed consent was obtained from all subjects, and all procedures were conducted in accordance with the Declaration of Helsinki and approved by the Institutional Review Board of Northwestern University.

41 28 Table 2.1. SCI subject demographics Subject No. Age (Yrs) Months post Injury Level AIS LEMS Ambulation Level Assistive Device MAS (KE/KF) SCATS C5-6 C4-7 C3-4 C4-6 C8-T1 C4-5 C4-6 C5-7 C6-7 C3-4 C6 D C D D D D D D C D C 40/50 39/50 40/50 44/50 50/50 37/50 46/50 47/50 41/50 43/50 30/50 Community Community Household Community Community Community Community Community Community Community Household Cane None Crutches None None Walker Cane None Walker Crutches Walker 3/1+ 0/0 0/1 1+/0 0/0 3/0 0/1+ 2/1 0/1+ 0/1+ 1+/3 F,E,C C F,E,C None None F,E,C F F,E,C F,E,C F F,E,C Abbreviations: SCI, spinal cord injury; AIS, ASIA (American Spinal Injury Association) Impairment Scale; LEMS, Lower Extremity Motor Scores (maximum possible score of 50); MAS, Modified Ashworth Scores (tested limb scores; KE = Knee Extensors, KF = Knee Flexors); SCATS, Spinal Cord Assessment Tool for Spastic Reflexes (F-Flexor; E-Extensor; C- Clonus).

42 Experimental set-up Each experimental procedure lasted approximately 1.5 hours. Subjects were seated in the adjustable height chair of the testing apparatus (System 3: Biodex Medical Systems, Shirley, NY, USA) with the hips flexed Subjects trunks were supported by the back rest of the testing chair and secured by 2 straps crossed over the chest. The lower leg was secured to the dynamometer arm, which was coupled to a 6 degrees-of-freedom load cell (ATI, Apex, NC, USA) used to assess KE torques. The axis of rotation of the dynamometer arm was aligned with the anatomical knee flexion-extension axis. All dynamic contractions were performed over an 80 range (10-90 knee flexion). Passive torques and limb inertia throughout the full range of motion were obtained during repeated 5 /s passive trials and subtracted from all testing trials using coefficients from a fitted 3rd order polynomial. The slow velocity used during passive trials was chosen to ensure spastic reflexes would not be elicited, and EMGs were inspected visually online during all trials to ensure negligible muscle activity throughout. Torque signals were low-pass filtered at 200 Hz and collected at 1000 Hz. Surface EMG was recorded using active bipolar electrodes (Delsys, Boston, MA, USA) applied over the rectus femoris (RF), vastus lateralis (VL), vastus medialis (VM), and medial hamstring (MH) muscles. EMG signals were amplified (x1000), band pass filtered ( Hz), and sampled at 1000 Hz simultaneously with the torque data Experimental Protocols A schematic representation of the 2 protocols used for these experiments is shown in Figure 2.1. The first experiment was performed to quantify muscle activation during different modes of muscle contractions and to determine velocity-dependent features in KE torque and EMG. To calculate central activation ratios (CARs), supramaximal electrical stimulation was

43 30 provided during single isometric, concentric, and eccentric MVCs at a constant reference angle (50 knee flexion) in order to make valid comparisons across contraction modes (Babault et al. 2001; Gandevia et al. 1998; Klass et al. 2005). Dynamic contractions were performed at velocities of +20 /s and -20 /s (note: positive values refer to concentric contractions, and negative values refer to eccentric contractions here and throughout). The 20 /s speed for dynamic contractions was chosen to provide similar contraction times during isometric and dynamic MVCs (i.e., ~4 s). The stimulation was provided using a brief train (10 pulses, 600 µs duration, 100 Hz, 135 V; Grass S48, external isolation; Grass Technologies, West Warwick, RI, USA) delivered to the KEs through 3 x 5 inch self-adhesive stimulating electrodes (ConMed Corp., Utica, NY, USA) placed over the distal VM (proximal to the VM recording electrode) and the proximal VL (proximal to the VL recording electrode) (Hornby et al. 2009; Thompson et al. 2011b). The train of pulses was used in order to increase the ability of the stimulation protocol to detect deficits in muscle activation (Miller et al. 1999). For isometric contractions, stimulation was delivered manually by the experimenters during the torque plateau of a 3-5 s MVC (Miller et al. 1999). For both concentric and eccentric contractions, custom Labview software (National Instruments, Austin, TX) was used to trigger stimulation as the limb passed through 50 knee flexion. For isometric MVCs, the mean torque produced during the 100 ms preceding stimulation onset was averaged and taken to be the peak MVC torque. In the case of dynamic contractions, since the torque trace was changing due to changes in muscle length and velocity, the peak MVC torque that would have been reached at the time of the peak superimposed torque was determined by extrapolation of the slope from 50 ms of the pre-stimulus measured torque (Gandevia et al. 1998; Klass et al. 2005). CAR values were then calculated using the following

44 31 equation: CAR = Peak MVC Torque/Peak Superimposed Torque (Kent-Braun and Le Blanc 1996; Kent-Braun and Ng 1999). The ordering of contraction types for CAR determination was pseudo-randomized. Subjects were instructed to extend the knee as fast and hard as possible and were given vigorous verbal encouragement throughout each effort. For each contraction type, after 2-3 submaximal practice trials, 3 MVCs were performed with supramaximal stimulation and 2 without stimulation (5 total). If either the CAR values or peak torque during any 2 contractions differed by more than 10%, additional contractions were performed. The 2 MVC contractions without stimulation were performed to compare torque and EMG without the confounding effects of superimposed electrical stimulation. All single MVCs were separated by at least 2 min rest, with 5 minutes of rest provided between different modes of contraction. Following 10 minutes rest after determination of CAR values, subjects performed anisometric MVC contractions at 60 /s. The additional speed was chosen to examine volitional force development at faster speeds and to detect any potential velocity-dependent components to motor output. Sufficient practice with submaximal contractions was provided to familiarize all subjects with the faster speed. Three concentric and 3 eccentric MVCs were performed at 60 /s, with the order counterbalanced across all 11 subjects. In order to limit the number of trials using supramaximal stimulation and minimize fatigue, no electrical stimulation was superimposed during 60 /s MVCs. The second experiment compared motor output in the same 11 SCI subjects during repeated isometric, concentric, and eccentric contractions. Following baseline isometric contractions (at 90 flexion) and 5 minutes rest, subjects performed 5 repeated isometric MVCs (4 s on: 6 s off). A 1-minute rest followed the 5th MVC, and then a 6th and final MVC was

45 32 performed; this strategy was utilized to determine any washout of history-dependent motor behaviors. Following isometric contractions, single and repeated concentric and eccentric MVCs were performed at 20 /s. Three single baseline MVCs were performed for each dynamic contraction and separated by at least 2 minutes rest. For repeated anisometric MVCs, subjects performed 5 repeated efforts with 1 second of rest at the end position prior to the limb being returned to its starting position (90 for concentric, 10 for eccentric) at 20 /s, with an additional second of rest in the starting position used to cue the subject to prepare for the next MVC. Five repeated efforts were therefore performed (4 s on: 6 s off) before a minute of rest and a 6th washout contraction. Vigorous verbal encouragement was provided for all MVCs. Repeated eccentric contractions were always performed last during this experiment due to the potential for greater muscle damage (Enoka 1996). Additionally, 6/11 SCI subjects performed 3 repeated isometric MVCs 5 minutes after the final eccentric trial to quantify any potential decrement in motor output or motivation following the eccentric contractions.

46 Figure 2.1. Schematic representation of experimental protocols. Experiment 1 examined baseline differences in central activation of the KEs between SCI and controls (ordering of contraction modes randomized). Experiment 2 examined volitional force generation during repeated isometric, concentric, and eccentric MVCs in SCI subjects. 33

47 Data collection and analysis Data were acquired and analyzed using custom LabView software (National Instruments, Austin, TX). Torque was low-pass filtered at 10 Hz (zero-phase lag, 4th order Butterworth). Absolute peak torque was identified in each trial, and the period corresponding to ±50 ms was then averaged to represent peak torque for all dynamic MVCs without muscle stimulation. Electromyographic recordings were full-wave rectified and smoothed using a low-pass filter (10 Hz, zero-phase lag, 4th order Butterworth) to create a linear envelope for further analysis. During preliminary data collection, group data showed that peak torque during baseline anisometric contractions occurred consistently at ~66 of knee flexion (similar to previous reports) (Brughelli et al. 2010), during both concentric and eccentric contractions; data analysis of anisometric contractions therefore focused on EMG amplitude between The mean EMG amplitude present 100 ms prior to the peak torque was used for analysis of all isometric EMGs to account for the electromechanical delay (Zhou et al. 1995). Pooled knee extensor EMG activity was calculated as the average of normalized RF, VL, and VM activity. MH EMG was analyzed to assess antagonist activity during all contractions. For the first experiment (single MVCs), non-car trials with the highest peak torque at each dynamic velocity were used for analysis. Torque and EMG were normalized to slow concentric (+20) values to allow for comparisons (Knutsson et al. 1997; Westing et al. 1991b). Isometric MVCs were not included in these comparisons, as peak torques were generated at a different knee angle than during dynamic contractions. For stimulated trials, only the highest CAR values produced during each contraction mode were used for comparisons. Two-way repeated measures analyses of variance (ANOVA) with main factors of subject group (SCI and control subjects) and contraction velocity (repeated) were used to assess differences in torque,

48 35 EMG amplitudes, peak torque angles, and CAR values. Bonferroni-corrected t-tests were used to identify differences between means when ANOVAs yielded significant interactions (P < 0.05), with significance levels ranging from One-way repeated measures ANOVAs were used to assess within-group differences in baseline torque, peak torque angles, EMG, and CAR values across the different modes/velocities of contraction. Post hoc Tukey-Kramer analyses were used following significant results to determine individual differences. For the second experiment (repeated MVCs), comparisons of torque and EMG were made using data normalized to mode-matched, peak baseline contractions (i.e., all repeated isometric trials were normalized to baseline isometric efforts, all concentric trials to baseline concentric, etc.). One-way repeated measures ANOVAs were utilized to assess differences in torques, peak torque angles, and EMG across repeated contractions, followed by post hoc Tukey- Kramer analyses as appropriate. Exceptions to mode-matched comparisons in this experiment include specific paired comparisons between contraction modes matched for contraction number (ie, 1 st eccentric-1 st concentric, 2 nd eccentric- 2 nd concentric, etc.), in which case, eccentric values were normalized to baseline concentric values, and Bonferroni-corrected t-tests were utilized with α = In order to assess for fatigue, two-way repeated measures ANOVA with main factors of time (initial set versus final set) and contraction number (repeated) was used to compare increases in torque during 3 repeated isometric MVCs Statistical analysis Data in the text are shown as mean ± standard deviation, while data in the figures are presented with standard errors. All statistical analyses were performed using a statistical software package (Statview; SAS Institute, NC) with α = 0.05, except where otherwise noted. Spearman

49 36 rank correlation coefficients were used to test for potential correlations between clinical measures of spasticity and electrophysiological measures of muscle activity and voluntary activation. Correlations between KE torque and EMG during repeated contractions were determined using Pearson product moments. 2.3 Results Clinical and demographic data for SCI subjects are listed in Table 2.1. The average age of SCI subjects was 46 ± 11 years, and the average duration post-injury was 115 ± 94 months. Ten out of 11 SCI subjects demonstrated clinical signs of spastic motor behaviors (i.e., scores > 0 on either MAS or SCATS). Specifically, 5/11 demonstrated positive scores on the MAS in the tested KEs, 9/11 tested positive on the SCATS, and only 1 subject did not have scores on either Motor output and central activation during single MVCs Analysis of absolute peak torque across contraction velocities demonstrated substantial strength deficits in the SCI group, with a significant main effect for subject group [F (3,60) = 65.98; P < ; see Table 2.2]. Markedly different patterns of motor output between SCI subjects and controls were observed following normalization of peak torques to slow concentric (+20) values (Fig. 2.3). Significant main effects for subject group [F (3,60) = 33.80; P < ] and contraction velocity [F (3,60) = 62.49; P < ] were observed, with the primary finding of significant interaction of these factors [F (3,60) = 28.49, P < ]. More directly, substantially greater torques during eccentric contractions in patients with SCI appear to be the primary determinant of the main effects, as comparisons at each velocity between subject groups revealed differences in normalized torque at -20 and -60 (both P < ; Fig. 2.3A), but not at +60 (P = 0.53). A one-way ANOVA and posthoc assessments of torque values for SCI subjects revealed

50 37 significant increases during eccentric MVCs of 97 ± 44% and 117 ± 44% during -20 and -60 velocities [F (3,30) = 65.07; P < ]. Conversely, one-way ANOVA of control subject torque values revealed smaller increases during eccentric MVCs [13 ± 38% and 11 ± 29% during -20 and -60 MVCs; F (3,30) = 6.11; P = ]. There were no significant effects across contraction velocity or group for peak torque knee angles (all P > 0.05). Normalized EMG amplitudes across contraction velocities revealed markedly different patterns between SCI and control subjects consistent with the torque data. There were significant main effects for subject group [F (3,60) = 13.16; P = ] and velocity [F (3,60) = 9.608; P < ] as well as a significant interaction [F (3,60) = 8.34; P = ]. The largest differences observed with group comparisons at each velocity revealed differences only during eccentric velocities (both P < ; see Fig. 2.3B), with no differences in +60 KE EMG (P = 0.39). Using a one-way ANOVA, normalized knee extensor EMGs were increased significantly in SCI subjects by 49 ± 44% and 73 ± 74% at -20 and -60 MVCs, respectively, while control subjects demonstrated a non-significant reduction in KE EMG during eccentric contractions (Fig. 2.3B). For SCI subjects, analysis of individual muscles during eccentric velocities revealed increased activity in all recorded agonists (RF, VL, and VM; all P < 0.05; Fig. 2.3C). Conversely, MH EMG in subjects with SCI showed no significant difference when comparing across all tested velocities [F (3,30) = 0.74, P = 0.54; see Figure 2.3C for comparisons to +20]. As depicted in Fig. 2.2, central activation deficits were observed in the SCI subjects compared to control subjects, regardless of contraction velocity (all P < 0.001; Fig. 2.3D), although the subject groups demonstrated contrasting patterns of central activation. Controls produced different CAR values across contraction velocities [F (2,20) = 4.07; P = 0.033], and post hoc tests revealed a significant reduction during eccentric MVCs (0.81 ± 0.15), as compared to

51 38 during isometric (0.91 ± 0.08) and concentric MVCs (0.86 ± 0.09), consistent with previous reports of healthy adults (Babault et al. 2001; Beltman et al. 2004). SCI subjects also produced different CAR values across contraction modes [F (2,20) = 31.71; P < ]. In contrast to controls, however, SCI subjects demonstrated greater CAR values during eccentric MVCs (0.62 ± 0.16) than both isometric and concentric CARs (0.50 ± 0.15; 0.36 ± 0.15, respectively). Isometric CARs were also greater than concentric CARs (Fig. 2.3D). Analyses of SCI subjects KE Modified Ashworth Scores and eccentric CAR values and peak eccentric torque values using Spearman rho yielded no significant correlations (P = 0.46 and 0.40, respectively).

52 Figure 2.2. Representative single subject data during KE CAR trials. Supramaximal stimulation of the KEs results in a larger evoked twitch during the eccentric MVC than during the isometric or concentric MVC in the control subject. During matched conditions, muscle stimulation evokes a relatively smaller twitch during an eccentric versus isometric or concentric MVC in the SCI subject, resulting in an increased eccentric CAR value. 39

53 40 Table 2.2. SCI and control subject data during single MVCs -60 /s -20 /s /s +60 /s Torque (Nm) SCI 109 ± ± ± ± ± 19 Control 232 ± ± ± ± ± 48 Peak Torque Angle ( ) SCI 64 ± ± ± ± 6.6 Control 74 ± ± ± ± 12.1 CAR SCI ± ± ± Control ± ± ± Values are means ± SD. Central Activation Ratio = peak volitional torque/peak superimposed torque.

54 Figure 2.3. SCI subjects demonstrate increased KE motor output during eccentric MVCs. After normalization to slow concentric trials (i.e., +20), significant gains in both KE torque (A) and EMG (B) are observed at eccentric versus concentric speeds in SCI subjects (filled circles), as compared to control subjects (open squares). (C) Individual KE muscle EMG amplitudes in SCI subjects increased during -20 MVCs (RF=159±15%, VL=147±21%, and VM=141±17% of +20 values), with differences between subject groups for each KE muscle, but not MH. (D) Differences in CAR values across contraction types between SCI and controls. Significant difference between SCI and control subjects: *P < 0.05, **P < 0.001, ***P <

55 Motor output during repeated anisometric contractions Peak torques and EMG activity were assessed during 5 repeated isometric, concentric, or eccentric MVCs. Consistent with previous data (Hornby et al. 2009; Thompson et al. 2011b), the SCI subjects demonstrated enhanced KE torque and pooled KE EMG during repeated isometric MVCs, with significant increases in torque [F (4,40) = 14.77, P < ] and EMG [F (4,40) = 8.384, P < ] above baseline values during the third to fifth contractions. Average maximal gains in torque and corresponding KE EMG for all 11 subjects were 27 ± 16% and 29 ± 31%, respectively. During repeated anisometric contractions at 20 /s, substantial differences in motor output were observed between concentric and eccentric MVCs. Figure 2.4 shows representative singlesubject data illustrating increases in torque and EMG from baseline MVCs only during concentric MVCs. Pooled data shown in Fig. 2.5 indicate significant increases in torque and KE EMG during concentric efforts, with specific differences between the 1 st versus 2 nd -5th contractions and no differences between the 2nd-5th efforts. Examination of group means at each contraction number revealed peak gains at the 4th MVC for both torque (131 ± 29% of baseline MVC) and pooled KE EMG (135 ± 33% of baseline MVC). The average of maximal increases in torque across all 11 subjects was 136 ± 26% of baseline, with pooled KE EMG at matched contractions being 133 ± 32% of baseline. Significant increases in EMG were observed for all individual KE muscles, with the largest changes in VL EMG (151 ± 56% of baseline). MH EMG demonstrated no change in magnitude during repeated concentric MVCs [F (4, 40) = 1.231, P = 0.32]. The average peak torque angles also did not vary significantly with repeated contractions [mean range of 66-72º across repeated MVCs; F (4,40) = 0.98, P = 0.43].

56 43 In contrast to repeated concentric contractions, no significant changes in torque or EMG were observed with repeated eccentric MVCs (Fig. 2.5). Eccentric torques, pooled KE EMG, and MH EMG remained close to baseline values [mean ranges of %, 95-99%, %, respectively; F (4,40) =2.210, 0.139, 1.409; P = 0.09, 0.99, and 0.25, respectively]. Joint angles at which peak torques were observed during repeated eccentric contractions were also consistent [mean range of ; F (4,40) = 0.70, P = 0.60]. Importantly, however, KE torques generated during eccentric MVCs were larger than concentric efforts. Specific paired comparisons using Bonferroni corrected t-tests (α = 0.01) of averaged torques for each contraction mode matched for contraction number (i.e., 1st eccentric 1st concentric, etc.) revealed significant differences across all 5 contractions. Similar comparisons of KE EMGs also revealed significant differences in the 1st of 5 repeated contractions, with the second and third contractions approaching statistical significance (P = 0.02 and 0.05, respectively) and no differences between the fourth and fifth contractions (P = 0.27 and 0.07, respectively; Fig. 2.5).

57 Figure 2.4. Single subject data demonstrating differences in torque development during repeated dynamic contractions. Representative data from a single subject reveal increasing KE torque and EMG (only VL is shown) during repeated concentric (A) but not eccentric (B) MVCs. Dashed horizontal lines represent peak torque during baseline contractions. 44

58 Figure 2.5. SCI subjects produce increasing KE torque and EMG during repeated concentric, but not eccentric, MVCs. SCI subjects produced increasing KE torque (A) and EMG (B) during the second through fifth repeated concentric MVCs. During repeated eccentric MVCs, SCI subjects maintained similar levels of torque (A) and EMG (B) across all 5 contractions (mean ranges of % and 95-99% of baseline values, respectively). No changes in MH EMG were observed with either type of repeated contractions (not shown). (* indicates differences from eccentric to concentric contractions; indicates differences from mode-matched first contraction). 45