Dysfunction of Cortical Dendritic Integration in Neuropathic Pain Reversed by Serotoninergic Neuromodulation

Transcription

1 Article Dysfunction of Cortical Dendritic Integration in Neuropathic Pain Reversed by Serotoninergic Neuromodulation Highlights d Neuropathic pain reduces HCN function in dendrites of the anterior cingulate cortex d d d Dendritic dysfunction results in increased integration of synaptic inputs Activation of type 7 serotonin receptors (5-HT 7 ) restores normal dendritic integration Treatment of the cingulate cortex with 5-HT 7 agonist produces analgesic effects Authors Mirko Santello, Thomas Nevian Correspondence nevian@pyl.unibe.ch In Brief Chronic pain emerges from altered brain function. Santello and Nevian found that in a neuropathic pain model, cortical neurons display enhanced dendritic integration resulting from downregulation of HCN channels. Activation of serotoninergic receptors alleviates pain by reversing this dendritic dysfunction. Santello & Nevian, 215, Neuron 86, April 8, 215 ª215 Elsevier Inc.

2 Neuron Article Dysfunction of Cortical Dendritic Integration in Neuropathic Pain Reversed by Serotoninergic Neuromodulation Mirko Santello 1 and Thomas Nevian 1,2, * 1 Department of Physiology, University of Bern, Bühlplatz 5, 312 Bern, Switzerland 2 Center for Cognition, Learning and Memory, University of Bern, Fabrikstrasse 8, 312 Bern, Switzerland *Correspondence: nevian@pyl.unibe.ch SUMMARY Neuropathic pain is caused by long-term modifications of neuronal function in the peripheral nervous system, the spinal cord, and supraspinal areas. Although functional changes in the forebrain are thought to contribute to the development of persistent pain, their significance and precise subcellular nature remain unexplored. Using somatic and dendritic whole-cell patch-clamp recordings from neurons in the anterior cingulate cortex, we discovered that sciatic nerve injury caused an activity-dependent dysfunction of hyperpolarization-activated cyclic nucleotide-regulated (HCN) channels in the dendrites of layer 5 pyramidal neurons resulting in enhanced integration of excitatory postsynaptic inputs and increased neuronal firing. Specific activation of the serotonin receptor type 7 (5-HT 7 R) alleviated the lesion-induced pathology by increasing HCN channel function, restoring normal dendritic integration, and reducing mechanical pain hypersensitivity in nerve-injured animals in vivo. Thus, serotoninergic neuromodulation at the forebrain level can reverse the dendritic dysfunction induced by neuropathic pain and may represent a potential therapeutical target. INTRODUCTION Long-lasting and irreversible sensitization of the pain pathway is believed to be a key mechanism for the chronification of pain (Baron, 29; Basbaum et al., 29; Costigan et al., 29). This phenomenon comprises all levels along the pain neuraxis, from peripheral nociceptors and spinal cord to supraspinal brain areas (Apkarian et al., 29). There is growing evidence that chronic pain is also associated with increased neuronal excitability in the forebrain (Saab, 212). The anterior cingulate cortex (ACC) in this respect is of particular importance because it is consistently activated during nociception and in chronic pain states (Shackman et al., 211; Vogt, 25; Wager et al., 213). Indeed, cortical brain areas, and ACC in particular, are essential for the conscious experience of pain (Lee and Tracey, 213; Perl, 27; Treede et al., 1999). Interfering with activity in this brain area can influence mechanical hyperalgesia (Eto et al., 211; Li et al., 21) and the affective-emotional component of chronic pain (Johansen et al., 21; King et al., 29; Qu et al., 211). In human patients, it has recently been demonstrated that functional changes in medial prefrontal cortex and ACC are critical for the transition from acute to chronic pain (Baliki et al., 212). Therefore, it is of high clinical relevance to understand the plastic changes occurring on the cellular level in ACC that might contribute to the chronification of pain. The alterations in cortical neurons that have been reported after peripheral nerve injury, like long-term potentiation (LTP) of synaptic transmission (Li et al., 21) and increases in intrinsic cellular excitability (Blom et al., 214), might be caused by primary or concomitant local changes in the ion channel composition of the dendritic arborization (Frick et al., 24). Plasticity of dendritic excitability is considered as an additional mode of memory storage (Losonczy et al., 28) and might therefore be a potential mechanism for the formation of a pain memory (Sandkühler, 29). Dendritic modifications in ion channel composition might influence the input-output relation of a neuron, and they could enhance neuronal activity in the ACC, contributing to the persistent pain state. Thus, we sought to investigate sciatic nerve injury-induced neuronal changes, which lead to the development of neuropathic pain, in the main apical dendrites of layer 5 (L5) pyramidal neurons, the principal output neurons of the ACC. Furthermore, dendrites are rich in receptors for neuromodulators, which locally influence their electrical and integrative properties (Arnsten et al., 212; Dembrow et al., 21). Yet, little is known about the effect of neuromodulation on cortical brain areas involved in pain processing and the potential cellular mechanisms that mediate this effect. The serotoninergic system originating in the midbrain is activated by ascending afferent nociceptive fibers (Basbaum et al., 29) as well as several brain regions (Pollak Dorocic et al., 214; Weissbourd et al., 214). It can either facilitate or inhibit pain transmission in the spinal cord (Ossipov et al., 21; Tracey and Mantyh, 27; Zhao et al., 27). Since forebrain areas are also extensively innervated by the serotoninergic system (Porrino and Goldman-Rakic, 1982), we wondered if activation of 5-HT receptors might be able to modulate neuronal properties on the dendritic level in the ACC and may therefore represent a possible target Neuron 86, , April 8, 215 ª215 Elsevier Inc. 233

3 Figure 1. Reduction of HCN Channel Function in the Apical Dendrites of L5 Pyramidal Cells after Peripheral Nerve Injury (A) CCI surgery was performed on the sciatic nerve of the left hind paw. An electronic Von Frey filament was used to evaluate mechanical sensitization of the hind limbs as measured by the mechanical threshold necessary to evoke paw withdrawal. Seven days after surgery, brain slices containing the rostral ACC were prepared for analysis of single-cell properties. (B) Only the paw subjected to chronic constriction injury (CCI left), but not the contralateral or shamoperated hind paw, showed mechanical hyperalgesia after surgery (CCI n = 29; Sham n = 26). Dashed line indicates the withdrawal threshold averaged over all conditions before surgery. (C) Reconstruction of a biocytin-labeled L5 pyramidal neuron illustrating the experimental setup. Whole-cell patch-clamp recordings were performed at the soma and main apical dendrite. Injections of positive or negative currents at the soma evoked hyperpolarization or action potentials that back-propagated into the dendrite (inset). Notice the difference in resting membrane potential between soma and dendrite (DV). (D) Dual patch-clamp recording from the dendrite and the soma of an animal subjected to CCI showed similar electrical properties but reduced DV. (E) Summary of all Sham and CCI dual recordings (left) and pooled data (right) showing distancedependent dendritic depolarization that was significantly reduced in CCI versus Sham animals for both proximal (< 15 mm) and distal (> 15 mm) recording sites. (F and G) Hyperpolarizing current injections into the dendrite and soma evoked a characteristic voltage sag indicative of the activation of the h-current (I h ). Dendrites but not somata from CCI neurons showed reduced sag as compared to Sham. V m at dendrite was 6 mv (Sham) and 64 mv (CCI). V m at soma was 67 mv (Sham) and 68 mv (CCI). (H) Sag ratio as a function of the distance from the soma for all dendritic recordings (left). Pooled data (right) for somatic (S) and proximal and distal recording sites indicating a significant reduction in dendritic I h after CCI surgery. *p <.5, **p <.1, ***p <.1. Error bars: SEM. See also Figures S1 S3. to reverse the neuronal changes that contribute to the chronic pain state. RESULTS Peripheral Nerve Ligation Causes HCN Channel Dysfunction in Cortical Dendrites Dual whole-cell patch-clamp recordings were performed at the soma and apical dendrite of L5 pyramidal neurons in the ACC to assess changes in dendritic and somatic properties in brain slice preparations of mice subjected to chronic constriction injury of the sciatic nerve (CCI), a well-established neuropathic pain model (Bennett and Xie, 1988), and sham-operated animals (Sham) (Figures 1A 1D). All experiments were performed on brain slices from behaviorally tested animals. CCI caused mechanical hypersensitivity specifically in the injured left hindpaw, as revealed by a reduction in the mechanical withdrawal threshold upon pressure application to the plantar surface of 234 Neuron 86, , April 8, 215 ª215 Elsevier Inc.

4 the paw by an electronic von Frey filament. We recorded from 36 Sham and 34 CCI dendrites at several distances from the soma. Consistent with previous reports from prefrontal cortex (Kalmbach et al., 213) and other cortical regions (Berger et al., 21), the main apical dendrite of L5 pyramidal neurons in the ACC under control condition displayed a more depolarized resting membrane potential (V m ) as compared to the soma, and this phenomenon increased with the distance from the soma (Figure 1E). Somatic V m was not different between Sham and CCI (Sham ±.55 mv, n = 36; CCI ±.49 mv, n = 34; p =.36), but we observed that the difference between somatic and dendritic V m was less pronounced in CCI animals both at proximal (< 15 mm) and distal (> 15 mm) recording sites (proximal V dend -V soma : Sham 4.2 ±.61 mv, n = 2; CCI 2 ±.53 mv, n = 11; p =.2; distal V dend -V soma : Sham 5.75 ±.62 mv, n = 16; CCI 2.83 ±.43 mv, n = 23; p =.1). Because the V m difference is largely due to a gradient (increasing from soma to dendrite) of HCN channel density that produces a tonic depolarizing current (Berger et al., 21; Lörincz et al., 22), we wondered if there was any difference in the h-current (I h ) generated by HCN channels between the Sham and CCI condition. In line with previous studies (Lörincz et al., 22; Magee, 1998; Stuart and Spruston, 1998; Williams and Stuart, 2), we found an increase in I h from soma to dendrite in Sham animals as evaluated by the voltage sag in response to a hyperpolarizing current step (Figures 1F and 1G). Strikingly, this gradient was reduced in CCI animals (Figure 1H). Thus, although the voltage sag recorded from the soma was comparable in the two conditions (sag ratio: Sham 1.27 ±.2, CCI 1.24 ±.1; p =.11), voltage sag after CCI was significantly reduced at both proximal and distal dendritic recording sites (proximal sag ratio: Sham 1.44 ±.3, n = 2; CCI 1.29 ±.3 n = 11; p =.2; distal sag ratio: Sham 1.84 ±.13, n = 16, CCI 1.62 ±.6, n = 23; p =.41). Most Dendritic Properties in the ACC Differ from Other Cortical Regions but Are Unaltered in Neuropathic Pain Animals We then tested if other passive or active properties of these dendrites were modified in CCI animals. Interestingly, we noticed that some of the functional properties of the main apical dendrite of L5 pyramidal cells in the ACC differed substantially from those previously reported in the somatosensory cortex (Spruston, 28) and that they were more similar to the behavior of medial prefrontal cortex (prelimbic) pyramidal neurons (Gulledge and Stuart, 23; Kalmbach et al., 213). ACC dendrites were characterized by little attenuation of single back-propagating action potentials (APs) but a marked use-dependent attenuation of high-frequency trains (Figure S1), an extremely small subthreshold potential attenuation from the dendrite towards the soma, and a decrease in input resistance with distance from the soma (Figure S2). Local Na + spikes could be evoked by dendritic depolarization, but we found only very little Ca 2+ regenerativity (Figure S3). Importantly, none of the above-mentioned dendritic properties were modified in CCI animals (Figures S1 S3), indicating that a large fraction of the dendritic Na +,K +, and Ca 2+ channels that influence the active as well as the passive properties of these dendrites were not altered by the pathological state. Neuropathic Pain-Induced Malfunction of Cortical Dendrites Enhances Temporal Summation of Synaptic Inputs and Neuronal Firing Since only I h was modified in the dendrites of L5 pyramidal cells in CCI animals, we wondered which consequences this had for the functional properties of these cells. Blockade of dendritic HCN channels in other brain areas enhances synaptic summation, indicating a key role for these channels in the integration of subthreshold synaptic inputs (George et al., 29; Magee, 1999; Williams and Stuart, 2). Therefore, we hypothesized that the I h phenotype we observed in CCI animals might have altered the dendritic integrative properties. First, we tested this by injecting a train of five EPSP-shaped subthreshold currents at 5 Hz in the dendrite (Figures 2A 2C). Indeed, we observed enhanced integration that resulted in increased summation of the EPSPs at the soma (fifth/first amplitude: Sham 1.75 ±.8, n = 8; CCI 1.98 ±.9, n = 1; p =.47). Then, we investigated if EPSPs evoked by synaptic stimulation in layer 1 (L1) were also differentially integrated. We found that, although amplitude and integral of single EPSPs were comparable between the Sham and CCI condition (EPSP amplitude: Sham 8.37 ± 1.8 mv, n = 21; EPSP amplitude: CCI 1.13 ±.61 mv, n = 39; p =.12, stimulation intensity 5 V), the integration of paired EPSPs (5 Hz) was increased in animals subjected to peripheral nerve damage at the soma (Figures 2D and 2E, second EPSP amplitude: Sham ± 1.39 mv, n = 19; CCI ± 1.5 mv, n = 26; p =.3, stimulation intensity 5 V). The area under the second EPSP was significantly increased and the EPSP duration prolonged (second EPSP integral: Sham ± ms,mv, n = 19; CCI ± ms,mv, n = 26; p =.4, stimulation intensity 5 V). All these changes were consistent with a reduction of I h in CCI dendrites. In order to evaluate the effect of a complete loss of I h on dendritic integration, we blocked HCN channels in Sham animals by bath application of ZD7288 (1 mm). This treatment altered dendritic integration and enhanced the temporal summation of EPSPs at the soma to an even larger extent (Figure S4). As a functionally relevant consequence of the increased temporal summation of synaptic inputs, the probability for AP generation was increased in the CCI neurons at lower stimulation strength (Figures 2F and 2G). Since the current and voltage threshold for somatic AP generation was the same in the Sham and CCI condition (current threshold: Sham ± 8.2 pa, CCI ± 8.9 pa; p =.34), and the CCI neurons required a smaller initial EPSP during paired synaptic activity to generate an AP on the second stimulus (Figure 2H, Sham ±.84 mv, n = 16; CCI ±.62 mv, n = 18; p =.4), the increased firing was most likely due to the enhanced temporal summation of the synaptic stimuli. Origin of the Dendritic HCN Channel Phenotype HCN channels are particularly plastic channels. Both modifications of single-channel properties and the number of functional HCN channels have been reported in several neurological diseases associated with increased cortical activity (Shah et al., 24; Zhang et al., 214). In order to assess any alteration in the single-channel properties of HCN channels caused by CCI, we performed dual whole-cell recordings from the soma and Neuron 86, , April 8, 215 ª215 Elsevier Inc. 235

5 Figure 2. CCI Causes Increased Dendritic Integration (A) Reconstruction of a L5 pyramidal neuron in the ACC showing the somatic and dendritic recording configuration. (B) Five EPSP-like current injections (t rise and t decay, 1 ms and 4 ms, respectively, 5 Hz) into the dendrite resulted in a larger somatic depolarization in CCI animals (red, distance 197 mm) compared to Sham (blue, distance 238 mm) even though EPSP amplitudes of the first EPSP were similar. V m at soma was 74 mv (Sham) and 68 mv (CCI). V m at dendrite was 6 mv (Sham) and 65 mv (CCI). (C) Amplitude and time integral of the last EPSP normalized to the first EPSP of the train measured at the dendritic and at the somatic location (CCI n = 1; Sham n = 8). Only dendritic recordings more than 75 mm from the soma were included. Peak amplitudes were derived from current injections that were just below the threshold for AP generation. CCI neurons displayed increased temporal summation of EPSPs at the soma. (D) Examples of single or paired EPSPs (5 Hz) recorded from the soma evoked by increasing synaptic stimulation (1 5 V) in the inner L1 for CCI and Sham. Notice that the paired EPSPs at 5 Hz show increased temporal summation at the soma in CCI as compared to Sham neurons. (E) Top, average EPSP amplitude (left) and time integral (right) plotted as function of the stimulus intensity; bottom, same quantification as above but for the second EPSP evoked by paired-pulse stimulation (5 Hz). Both amplitude and integral of the second EPSPs are significantly larger in CCI (n = 44) as compared to Sham (n = 41) for increasing stimulation intensities. The time interval over which the integral was calculated is shown in (D). The increased EPSP integration in CCI neurons is consistent with a decrease in HCN channel function in the apical dendrites. (F and G) Paired-pulse stimulation had a higher probability to evoke a somatic action potential in CCI (n = 37) versus Sham (n = 26). (H) Cumulative plot (left) and average EPSP amplitude at AP generation (right) illustrating that the amplitude of the first EPSP of a paired-pulse stimulation was lower in CCI neurons when the second generated an AP as compared to Sham. *p <.5, **p <.1, ***p <.1. Error bars: SEM. See also Figure S4. the dendrite from L5 pyramidal neurons in the ACC. Hyperpolarizing voltage steps were applied to the dendrite in voltageclamp mode (from 45 mv to 125 mv, DV = 1 mv) in order to record the HCN-mediated currents in the dendrite (Figure 3A). Taking advantage of the negligible steady-state voltage attenuation from the dendrite to the soma (Figure S2D), the somatic voltage recording in current-clamp mode could be taken as a good estimate of the actually applied dendritic voltage, thereby allowing a proper correction of the voltage clamp error (Williams and Mitchell, 28). The peak of the activated tail current after stepping back to 7 mv was taken as an estimate of HCN channel function (Figures 3B and 3C). The current-voltage relationship was identical for the Sham and CCI condition, suggesting that the single-channel properties of the dendritic HCN channels were not modified by CCI. Therefore, we tested if the number of dendritic HCN channels was changed in CCI. We performed immunohistochemical labeling of HCN channels in coronal brain slices containing the rostral ACC. Strong staining could be found in the distal tuft dendrites in layer 1 with a decreasing gradient toward deeper layers (Figure 3D). These stainings were consistent with the observed gradient in the sag ratio and immunostainings from other brain regions (Kupferman et al., 214; Lörincz et al., 22). We found a much more pronounced decay in HCN channel labeling toward deeper layers in brain slices from CCI animals. The intensity profile of HCN channel staining could be fitted with a single exponential yielding a length constant (l HCN ) for the HCN channel distribution along the midline-to-l5 axis (Figures 3D and 3E). The length constant was significantly shorter in the CCI than in the Sham condition (CCI: l HCN = 26.5 ± 2.5 mm, 236 Neuron 86, , April 8, 215 ª215 Elsevier Inc.

6 Figure 3. HCN Channel Number, but Not Single-Channel Properties, Are Altered in CCI (A) Sketch illustrating the experimental setup of a voltage clamp (VC) recording from the dendrite and a simultaneous current clamp (CC) recording from the soma. Left, example traces (blue) of a VC recording from a distal dendrite (distance 197 mm) of a Sham animal and simultaneous CC recording from the soma. Hyperpolarizing voltage steps were applied to the dendrite (from 45 mv to 125 mv, DV = 1 mv, middle trace) in order to record HCN mediated currents (upper traces). Right, example traces (red) from a CCI-operated animal (dendritic recording 182 mm from soma). (B) Expanded view of the tail currents generated by HCN channel activation as indicated by the dashed rectangles in (A). (C) Voltage dependence of the channel activation determined from tail currents (dotted line in B) in CCI (red) and Sham (blue) dendrites fitted with a Boltzmann function (solid line). Only recordings from distal dendrites are pooled in the average (Sham average distance = 175 mm, n = 3; CCI average distance = mm, n = 5). The current-voltage relationship for the CCI and Sham condition were identical. (D) Example images of immunohistochemical antibody stainings of HCN1 channels in the rostral ACC, the region where electrophysiological experiments were performed. Traces below the images show the intensity profile of HCN1 staining from the midline to layer 5. Single exponential fits to the intensity profile yielded a length constant (l HCN ) for the distribution of HCN channels (blue, Sham; red, CCI). (E) Averaged peak intensity in L1 (left) and length constants (right) for the Sham and CCI condition. **p <.1. Error bars: SEM. See also Figure S5. n = 7; Sham: l HCN = 5.6 ± 6.1 mm; n = 6; p =.3), whereas the peak intensity in the outer L1 was identical in the two conditions (CCI: 11 ± 8, n = 7; Sham: 12 ± 1; n = 6; p =.96). The analysis of the HCN channel profile performed in the same slices but in a cortical brain region 9 mm lateral to the midline (motor cortex) yielded an identical distribution of HCN channel staining in the two conditions (Figure S5). Taken together, these results suggest that CCI resulted in no modification of the single-channel properties of the HCN channels, but in a change in the HCN channel distribution with a decrease in HCN channel density in the main apical dendrite of L5 pyramidal neurons. Activity-Dependent Plasticity of HCN Channel Function Is Occluded by Peripheral Nerve Ligation Next, we investigated what cellular mechanism might induce the reduction in HCN channel function. Nociceptive stimuli result in increased synaptic activity (Shyu et al., 21) and AP discharge in pyramidal neurons in the ACC with an average frequency around 4 5 Hz (Koga et al., 21). In the case of sciatic nerve ligation that leads to the development of neuropathic pain, long-term peripheral and spinal cord sensitization (Basbaum et al., 29; Costigan et al., 29) alters the pattern of afferent activity in the nociceptive system (Terashima et al., 211) that is transmitted to supraspinal brain areas including the ACC. Hence, activity in the ACC is significantly increased after nerve damage, manifested in a discharge of several APs per second (around 4 Hz) in pyramidal neurons (Ning et al., 213). Increased excitatory synaptic activity can result in the long-term modifications of HCN channel function, as has been demonstrated in hippocampal pyramidal neurons (Brager and Johnston, 27; Fan et al., 25). We tested if such a plasticity mechanism existed in L5 pyramidal neurons in the ACC and whether it could be responsible for the downregulation of I h after nerve damage. We performed long-term whole-cell dendritic recordings (Figure 4A). After recording the baseline function of HCN channels by evaluating the sag ratio from a hyperpolarizing current step ( 5 pa, 5 ms), we mimicked the increased activity induced by the nerve damage by evoking suprathreshold EPSPs with robust extracellular stimulation in L1 at 4 Hz for 5 min (Figures 4A and 4B). Then we monitored HCN channel function for the following 3 min. After this time, we observed a significant reduction in the sag ratio in the group of naive or Sham animals from 1.73 ±.24 to 1.5 ±.15 (p =.46; n = 7; Figures 4C and 4D). This change in the sag ratio was accompanied by a significant increase in dendritic input resistance (R in : baseline 25.5 ± 5.9 MU; post-induction 31.8 ± 6.2 MU; p =.18; Figure 4G) and a slight hyperpolarization of the dendritic membrane potential (from 64.5 mv to 65.9 mv; p =.3). Consistently, the reduction in HCN channel function resulted in increased temporal summation of EPSP-like current injections into the dendrite (train of five EPSP-shaped subthreshold current injections at 5 Hz; fifth/first amplitude: baseline 1.27 ±.15; post-induction 1.65 ±.11; p =.4; Figures 4H and 4I) and a lowering of the resulting threshold for AP generation (baseline 814 ± 11 pa; post-induction 667 ± 12 pa; p =.25; Figures 4J and 4K). These results are consistent with a downregulation of HCN channel function resulting in an increase in dendritic integration and ensuing neuronal firing. We found that the induction of HCN channel plasticity required the activation of metabotropic glutamate receptors type Neuron 86, , April 8, 215 ª215 Elsevier Inc. 237

7 Figure 4. Activity-Dependent Plasticity of HCN Channel Function Is Occluded by CCI (A) Neurolucida reconstruction of a L5 pyramidal cell showing the position of the dendritic recording pipette and the location of the extracellular synaptic stimulation electrode. (B) Depiction of the induction protocol for HCN channel plasticity. Synaptically evoked suprathreshold EPSPs were induced by extracellular stimulation in L1 (4 Hz, 5 min, 6 V) and recorded from the dendritic recording location shown in (A). (C) Left, example trace of the voltage sag induced by hyperpolarizing current steps during baseline (black, 5 pa) and 3 min post-induction (blue, 4 pa) in a Sham animal. Right, example trace of the voltage sag during baseline (black, 5 pa) and 3 min post-induction (red, 5 pa) in a CCIoperated animal. No change in the sag was observed in this case. (D) The sag ratio is significantly decreased by the 4 Hz stimulation in a group of Sham and naive animals (n = 7), but not in dendrites of animals subjected to CCI (n = 6). The distance from the soma of the recording pipettes was comparable in the two conditions. The sag was recorded at the same resting potential and with current injections, giving rise to the same initial voltage drop. (E) Example of dendritic sag induced by a 5 pa hyperpolarizing current step before and after 4 Hz synaptic stimulation in the presence of the group 1 metabotropic glutamate receptor antagonist LY (1 mm). (F) Both in the CCI preparation (n = 6) and in presence of LY (n = 5), the decrease in sag ratio was abolished. The bar graph shows the sag ratio 3 min post-induction normalized to control in the three conditions. (G) Dendritic input resistance (R in ) is increased in the control but not in the CCI condition. (H) Train of five EPSP-shaped current injections at 5 Hz in a sham- and CCI-operated animal before and after the 4 Hz induction protocol. (I) The amplitude of the fifth EPSP of the train (normalized to the first) is significantly increased in the control but not in the CCI condition. The quantification was done at the same V m. (J) Example traces of the five EPSP-shaped currents at increasing current injections, eventually eliciting APs. Left, baseline for Sham and CCI; right, 3 min after the induction protocol. Notice that the threshold for AP generation is largely reduced only in the Sham condition. V m was 63 mv and 62 mv for Sham and CCI, respectively. (K) Quantification of the reduction in the current threshold for AP generation in the control condition (blue) and in the CCI condition (red) after synaptic stimulation. *p <.5, **p <.1. Error bars: SEM. I (mglur I). Bath application of the mglur I antagonist LY (1 mm) completely blocked the modification of the sag ratio (baseline 1.38 ±.4; post-induction 1.38 ±.4; p =.94; n = 5; Figures 4E and 4F) and the corresponding changes in the other cellular properties (relative change in sag ratio: 1. ±.2; input resistance: 1.13 ±.5; EPSP summation: 1.2 ±.6; AP threshold:.94 ±.11; all p >.5; n = 5). Strikingly, this HCN channel plasticity was occluded in CCI animals. No change in the sag ratio (baseline 1.51 ±.13; post-induction 1.51 ±.1; p =.97; n = 6), input resistance (baseline 29.5 ± 8.6 MU; post-induction 29.4 ± 8.8 MU; p =.95), EPSP summation (baseline 1.33 ±.12; post-induction 1.5 ±.7; p =.7), or AP threshold (baseline 75 ± 16 pa; post-induction 717 ± 117 pa; p =.62) could be observed after the 4 Hz stimulation in the CCI condition (Figures 4C 4K). In summary, the observed downregulation of HCN channel function induced by glutamatergic activity results in similar changes in the integrative properties of the apical dendrite of L5 pyramidal neurons in the ACC as observed after nerve ligation (increased temporal summation of EPSPs leading to augmented AP firing). Furthermore, the occlusion of the HCN channel plasticity in the CCI animals suggests that the nerve ligation-induced 238 Neuron 86, , April 8, 215 ª215 Elsevier Inc.

8 downregulation of HCN channel function and the synaptically induced plasticity described above share a common induction mechanism. Thus, the observed hypofunction of dendritic HCN channels in the CCI condition is most likely induced by increased afferent glutamatergic activity resulting from the peripheral nerve damage. Serotonin 5-HT 7 Receptor Activation Rescues Cortical Dendritic Dysfunction in Neuropathic Pain The observed pathological HCN channel phenotype might contribute to the increased activity in the ACC observed after peripheral nerve injury. Therefore, we wondered if it was possible to upregulate I h pharmacologically in the ACC in the CCI condition. HCN channel function can be positively modulated by several factors (Jung et al., 21), including increased levels of camp (Wang et al., 22). Yet if this can occur locally in dendrites and whether it has a functional relevance has not been clearly demonstrated. The ACC is highly innervated by brainstem fibers releasing neuromodulators (Porrino and Goldman-Rakic, 1982). Among them, 5-HT could be of potential relevance since serotoninergic receptors have been shown to be potential targets for pain treatment (Viguier et al., 213). However, whether they have a cortical site of action is not established yet. Among all serotoninergic receptors, the 5-HT 7 is one of the least well characterized (Hedlund, 29), but it has been shown to increase cytosolic levels of camp by stimulating adenylate cyclase (Bard et al., 1993), can modulate HCN channels (Chapin and Andrade, 21), and is potentially useful for the treatment of neuropathic pain (Brenchat et al., 21). We found that this receptor was highly enriched in the apical and tuft dendrites of L5 pyramidal neurons of the ACC (Figure 5A; n = 12 slices from three animals). Its activation increased dendritic HCN channel function and reduced synaptic integration in the main apical dendrite of these cells (Figure 5). Bath application of 5- carboxamidotryptamine (, 3 nm), a 5-HT 7 agonist (Figures S6A S6E), to a CCI preparation led to a dendrite-selective depolarization and robust increases in I h (Figures 5B 5E, sag ratio dendrite: baseline 1.76 ±.6, 2.5 ±.12, n = 5, p =.15; sag ratio soma: baseline 1.27 ±.2, 1.24 ±.2, n = 1, p =.77; V dend : baseline 61.7 ± 1.7 mv, 6.7 ± 1.8 mv, n = 5, p =.45; V dend -V soma : baseline 5.5 ± 2. mv, 7.1 ± 1.8 mv, n = 5, p =.11). Accordingly, integration of EPSP-shaped current injections and synaptic excitatory postsynaptic potentials was reduced (Figures 5F 5H, fifth/first amplitude soma: baseline 2.1 ±.15, 1.52 ±.11, n = 6, p =.5; second EPSP amplitude: baseline 8.57 ± 1.83 mv, 6.1 ± 1.57 mv, n = 1, p =.28, stimulation intensity 3 V), whereas single EPSPs as well as somatic input resistance and membrane potential were not substantially modified (Figures S6F S6H). As a result, the probability to generate an AP with paired pulse stimulation was reduced to stimulation strengths comparable to the Sham condition (Figures 5I and 5J). Importantly, the effect of on I h and dendritic integration was completely blocked by the highly specific 5-HT 7 receptor antagonist SB26997, suggesting that at the used concentration acted subtype selective (Figure S6), consistent with previous reports (Vasefi et al., 213; Zhang et al., 29). Next, we investigated if the activation of 5-HT 7 receptors by indeed acted on HCN channels to modulate dendritic integration. In the presence of the HCN antagonist ZD7288, which abolished the dendritic sag, decreased dendritic resting membrane potential, and increased EPSP temporal summation (Figure S4; Figures 6A 6D), exerted no modulation of the dendritic function anymore (sag ratio dendrite: baseline 1.39 ±.14, ZD ±.3, 1.3 ±.2, n = 6, p =.31; V dend : baseline 66.5 ±.2 mv, ZD ± 2.9 mv, 5- CT 8. ± 3.7 mv, p =.79, n = 6; fifth/first amplitude soma: baseline 1.4 ±.9, ZD ±.26, 2.12 ±.14, n = 6, p =.37). Furthermore, we tested the hypothesis that the modulation of HCN channels by 5-HT 7 receptors was mediated via adenylate cyclase activity (Figures 6E 6H). We incubated slices with the adenylate cyclase inhibitor SQ In this condition, failed to modify sag ratio, the resting membrane potential, and the temporal summation of EPSPs measured at the dendrite (sag ratio dendrite: SQ ±.1, 1.35 ±.5, n = 5, p =.5; V dend : SQ ± 2.2 mv, 66.7 ± 1.5 mv, p =.39, n = 5; fifth/first amplitude soma: SQ ±.8, 1.41 ±.1, n = 5, p =.82). These results suggest that 5-HT 7 receptors selectively modulate dendritic integration by acting on HCN channels via activation of adenylate cyclase. Additionally, these results demonstrate that has no off-target effects that could influence the dendritic properties by modulating other receptors and ion channels. Finally, we found that had no significant effect on the electrical properties, synaptic transmission and synaptic integration of L2/3 pyramidal neurons and interneurons tested across all layers of the ACC (Figure S7). These results suggest that most likely specifically modulates the dendritic properties of L5 pyramidal neurons in the ACC through 5-HT 7 receptordependent upregulation of HCN channel function. Targeted Treatment of the ACC with a 5-HT 7 Receptor Agonist Reverses the Behavioral Pain Response Our results so far showed that nerve injury caused a dysfunction of HCN channels in the dendrites of L5 pyramidal cells, the primary output neurons of the cortex, increasing the input-output function and synaptically evoked AP generation. We hypothesized that if this change was in any way related to the mechanical hypersensitivity induced by CCI, interfering with dendritic integration by selective activation of 5-HT 7 receptors in these cells could change the behavioral response to mechanical stimulation of the injured hindpaw. We tested this hypothesis by injecting directly into the ACC (Figures 7A and 7B) and examining its effect on CCI-induced allodynialike behavior. Consistent with our hypothesis, this manipulation reduced mechanical hypersensitivity, whereas saline injection exerted no effect (Figures 7C 7F). A single cortical injection of increased the mechanical withdrawal threshold by 66% ± 12%. This analgesic effect decayed back to the sensitized condition within 1 hr, consistent with the diffusion of the drug from the site of action. Co-injection of with the specific 5-HT 7 receptor antagonist SB26997 had no effect on the mechanical withdrawal threshold, suggesting that indeed acted specifically on 5-HT 7 receptors in the ACC. Injection of Neuron 86, , April 8, 215 ª215 Elsevier Inc. 239

9 24 Neuron 86, , April 8, 215 ª215 Elsevier Inc. Figure 5. 5-HT 7 Receptor Activation Increases Dendritic I h and Reduces Synaptic Integration in CCI (A) Confocal image of fluorescently labeled 5-HT 7 receptors in ACC. The arrowheads indicate clearly labeled apical dendrites of L5 pyramidal neurons (B) Reconstruction of a L5 pyramidal cell from a CCI animal showing the dendritic recording configuration (distance 197 mm). (C) Time course of resting membrane potential (V m ) at the soma and dendrite of the cell in (A) during bath application of the 5-HT 7 receptor agonist 5- CT (3 nm). Summary bar graph shows that the V m difference between soma and dendrite was significantly increased after the drug application (n = 5). This effect is consistent with the influence of 5-HT 7 receptors on dendritic HCN channel function. Sham recordings gave similar results (n = 4; data not shown). (D) Sag induced by a 5 pa hyperpolarizing pulse before (red) and after (blue) at the soma were not different. At the dendrite, current injections evoking the same hyperpolarization (control: 3 pa, : 5 pa) evoked a bigger sag in the presence of. V m at soma was 66 mv () and 65 mv (). V m at dendrite was 63 mv () and 61 mv (). (E) On average the somatic sag ratio was not affected by the drug (n = 1, p =.8), whereas the average dendritic sag ratio ( 5 pa current injections) was increased by (n = 5). (F) The summation of five EPSP-shaped current waveforms (5 Hz) injected at the dendrite of the CCI neuron shown at the top was reduced after 5- CT application (example distance 177 mm). V m at soma was 64 mv () and 64 mv (). V m at dendrite was 57 mv () and 58 mv (). (G) Amplitude and time integral generated by the fifth consecutive EPSP waveform normalized to the first EPSP were significantly decreased both at the dendrite and at the soma after (n = 6). (H) L1 paired-pulse synaptic stimulation and double somatic and dendritic recording illustrating the effect of. Notice that the dendrite is more depolarized in the presence of. Right, somatic EPSP amplitude and time integral plotted as a function of the stimulus intensity at baseline and after (n = 12). (I) Example traces showing how decreased the probability of AP generation evoked by 5-Hz paired-pulse (3-4-5 V) stimulation of L1 synapses. Notice that somatic V m is similar before and after. (J) The amplitude of the first EPSP of a pairedpulse stimulation when the second generates an AP is increased in the presence of (left). Stimulus intensity versus probability of AP generation in the two conditions (right, n = 12). *p <.5, **p <.1. Error bars: SEM. See also Figures S6 and S7.

10 Figure 6. 5-HT 7 Receptors Interact with HCN Channels via Adenylate Cyclase (A) Reconstruction of a L5 pyramidal cell from a CCI animal showing the dendritic recording configuration (distance 15 mm). (B) Dendritic voltage responses to a hyperpolarizing current injection into the dendrite. The characteristic voltage sag observed under control conditions (red) is abolished by bath application of the HCN channel blocker ZD7288 (1 mm, yellow). Subsequent additional bath application of (3 nm, blue) had no effect on the dendritic voltage response. (C) Dendritic voltage response to five EPSP-shaped current injections (5 Hz) into the dendrite. In the presence of ZD7288 (yellow), the temporal summation was increased as compared to control (red). Subsequent bath application of had no effect on the EPSPs and did not reduce the temporal summation (blue). (D) Summary of dendritic sag ratio (left), dendritic V m (middle), and the amplitude of the fifth EPSP of the train (normalized to the first EPSP; right). ZD7288 had a significant effect on dendritic properties as compared to control, whereas in the presence of ZD7288, had no modulatory effect. (E) Reconstruction of a L5 pyramidal cell from a CCI animal showing the dendritic recording configuration (distance 16 mm). In these experiments the slices were incubated in the adenylate cyclase inhibitor SQ22536 (1 mm) for at least 2 hr before recording. (F) Dendritic voltage responses to a hyperpolarizing current injection into the dendrite in the presence of SQ22536 (purple) and after bath application of (blue). (G) Dendritic voltage response to five EPSP-shaped current injections (5 Hz) into the dendrite in the presence of SQ22536 (purple) and after bath application of (blue). (H) Summary of dendritic sag ratio (left), dendritic V m (middle), and the amplitude of the fifth EPSP of the train (normalized to the first EPSP; right). In the presence of SQ22536, had no modulatory effect on the dendritic properties. *p <.5, **p <.1. Error bars: SEM. See also Figure S4. SB26997 alone showed the tendency to further increase the behavioral response to mechanical stimulation, pointing to a bidirectional modulation of cortical pain processing by 5-HT 7 receptors. Bilateral injections of into motor cortical areas lateral to the ACC had no effect on the mechanical withdrawal threshold (Figure S8), suggesting that the modulatory effect of the drug was specific to ACC. DISCUSSION Increased neuronal activity in cortical areas has been postulated to be one of the key manifestations of chronic pain (Kuner, 21; Saab, 212). However, the molecular mechanisms that underlie the enhancement of cortical activity in this pathological condition remain largely elusive. We employed double somatic and dendritic patch-clamp recordings, in vivo pharmacology, and behavioral tests to assess cortical alterations in a mouse model of chronic pain. We found specific plastic changes in the dendrites of cortical pyramidal neurons that are associated with nerve damage-induced development of neuropathic pain. Our data show that, after peripheral nerve injury, the main apical dendrite of the L5 pyramidal neurons in the ACC displays an impairment of HCN channel function. This leads to dendritic dysfunction characterized by enhanced temporal summation of synaptic inputs and increased neuronal firing. Furthermore, we identify a serotoninergic receptor (5- HT 7 R) expressed in L5 ACC pyramidal neurons that, by selectively increasing HCN channel function in the dendrites via adenylate cyclase, restores normal dendritic integration and reduces allodynia-like behavior in vivo. Thus, our findings demonstrate for the first time that, by manipulating specifically the properties of cortical pyramidal cell dendrites, it is possible to negatively modulate their input-output relationship, thereby reducing nerve injury-induced mechanical sensitization and pain behavior in vivo. Dendrites as Sites for the Formation of a Pain Memory Dendrites receive most of the synaptic contacts, and they participate in the integration and transformation of synaptic inputs into AP output (Häusser et al., 2; London and Häusser, 25; Major et al., 213). In this respect, passive and active dendritic properties play a major role in our understanding of brain function. Dendrites are highly plastic structures (Magee and Johnston, 25; Shah et al., 21) that have been implicated in a number of pathologies of the central nervous system (Nestor and Hoffman, 212; Palmer, 214; Poolos and Johnston, Neuron 86, , April 8, 215 ª215 Elsevier Inc. 241

11 Figure 7. Injections of into the ACC Reduces Mechanical Sensitivity in CCI Animals (A) Schematic illustration of the experimental design to test mechanical hypersensitivity after CCI surgery and bilateral injection of or saline into the ACC. (B) 3D reconstruction of a brain imaged with 3D ultramicroscopy showing the volume and site of the injection in ACC after ex vivo fixation and clearing of the entire brain. The injections were performed.7 mm rostral to Bregma,.35 mm lateral to midline, and 1.75 mm below the skull surface in each hemisphere. The bottom right image shows a single coronal section, and dots indicate the center of the injection volume for (blue) and saline (white) treatments. (C) Time course showing the control group of animals first subjected to CCI surgery (red bar), then after 7 1 days bilaterally injected with saline (gray bar). Saline injection did not modify mechanical thresholds in both the injured (left) and uninjured paw (right, n = 5). (D) Local injections of in the ACC (3 mm, blue bar) alleviated mechanical hyperalgesia while having no effect on the uninjured paw. Thus, after injection, mechanical thresholds were elevated significantly compared to thresholds before treatment (p <.1, n = 6), indicating partial reduction of the mechanical hypersensitivity. (E) Withdrawal threshold of the injured paw normalized to the non-injured paw in saline (white) and (blue) treated groups. After injections, mechanical thresholds were elevated significantly compared to thresholds in the control group (p <.1). (F) Statistical analysis expressed as area under the curve (AUC) between 3 and 12 min after treatment. Bar graph shows statistical significance between and saline-treated groups, between and + SB26997 (the 5-HT 7 antagonist, n = 3), but not between + SB26997 and SB26997 (n = 4) alone. These data indicate that acts on 5-HT 7 receptors to reduce mechanical hyperalgesia in vivo. One-way or two-way ANOVA was used for statistical comparisons. *p <.5, **p <.1, ***p <.1. Error bars: SEM. See also Figure S8. 212). So far, the idea that a change in dendritic excitability could function as a mechanism for the formation of a pain memory has been completely unexplored. This is the first report providing strong evidence that changes in dendritic properties are at least partially responsible for the establishment of chronic pain. Therefore, we would like to introduce the novel concept of a cortical dendropathology of chronic pain. Not surprisingly, modulation of this dendropathology could only partially alleviate pain hypersensitivity, because additional peripheral, spinal, and central mechanisms also contribute to nerve injury-induced neuropathic pain (Basbaum et al., 29). The ACC, together with other brain areas, is essential for the conscious experience of pain (Lee and Tracey, 213; Perl, 27; Treede et al., 1999). Interfering with activity in this brain area has been repeatedly reported to modulate not only mechanical hyperalgesia (Eto et al., 211; Li et al., 21; Thibault et al., 214; Wang et al., 211; Xu et al., 28) but also the affective-emotional component of chronic pain (Johansen et al., 21; LaGraize et al., 24). It will be essential to test in the future if the malfunction of dendritic integration may also be relevant for the affective component of pain. This hypothesis can best be assessed by employing behavioral tests such as conditioned place preference (King et al., 29; Qu et al., 211), which capture the aversiveness of spontaneous neuropathic pain and depend on neuronal processing in the ACC. HCN Channels and Chronic Pain We report that HCN channels are the key molecules responsible for altered dendritic function. The finding that these specific channels are functionally downregulated in the ACC dendrites during neuropathic pain further supports the hypothesis that dendritic malfunction is essential for the formation of a pain memory in the cortex. Indeed, normal dendritic HCN channel function is important for proper learning and memory. Specifically, lack of these channels favors LTP, the synaptic correlate of memory formation (Nolan et al., 24). Hence, it is tempting to speculate that partial loss of HCN channel function found in chronic pain may promote potentiation of synaptic strength. Previously reported synaptic alterations (Li et al., 21) may therefore depend on HCN hypofunction. The exact causal relationship between these two memory mechanisms remains to be determined. HCN channels have been previously implicated in the development of chronic pain, albeit in other locations. Indeed, in nociceptive neurons HCN channel function is upregulated in response to inflammation and nerve damage, promoting the generation of AP (Emery et al., 211; Momin et al., 28). Contrary to the effect on cortical neurons as demonstrated in our study here, blockade of HCN channels in the peripheral nervous system reduces neuronal firing and exerts an analgesic effect. Thus, HCN channel activation can have opposite effects in different parts of the nociceptive system. 242 Neuron 86, , April 8, 215 ª215 Elsevier Inc.

12 Serotoninergic Modulation of Pain Processing at the Cortical Level The experience of pain is strongly influenced by neuromodulatory systems (Ossipov et al., 21). Particularly, the descending pain modulatory system can either facilitate or inhibit nociceptive processing (Tracey and Mantyh, 27). Descending inhibition is mediated by serotoninergic fibers that decrease spinal cord activity and reduce pain sensation (Porreca et al., 22). Yet, little was known about the modulatory effect of 5-HT on cortical brain areas responsible for pain sensation. We wondered if 5-HT in the ACC was able to enhance impaired I h in the CCI condition. HCN channel function can be positively modulated by enhanced levels of camp (Wang et al., 22). We found immunohistochemical evidence for the expression of 5-HT 7 receptors in the dendrites of L5 pyramidal neurons in the ACC, consistent with other reports (Duncan and Franklin, 27; Hedlund, 29). This serotonin receptor has been shown to increase cytosolic levels of camp by stimulation of adenylate cyclase (Bard et al., 1993; Chapin and Andrade, 21). Thus, we thought to take advantage of 5-HT 7 receptors to raise intracellular camp in L5 pyramidal cells, thereby enhancing I h. Other neurotransmitter systems that increase camp could have similar effects. We found that activation of 5-HT 7 receptors in these neurons activates adenylate cyclase and increases HCN channel function in the main apical dendrite to reduce dendritic integration. Interestingly, I h was unchanged at the soma, pointing to a compartmentalized interaction of 5-HT 7 receptors with HCN channels in the dendrites. This observation is consistent with the expression pattern of 5-HT 7 receptors and HCN channels revealed by our immunohistochemical analysis. Other cell types do not appear to be modulated by 5-HT 7 receptors in this brain area, suggesting that we found a specific neuromodulatory effect. Yet, we cannot fully exclude that 5-HT 7 receptors exist in other cell types or subcellular compartments in the ACC and that their activation could to some extent contribute to the behavioral effects that we observe. Knockout of 5-HT 7 receptors selectively in L5 pyramidal cells in the frontal cortex could be used in the future to test additional effects of on other cell types. In parallel, specific manipulation of camp levels in L5 pyramidal neurons with optoxr activation (Airan et al., 29) could be used to mimic the effect of 5-HT 7 activation both in the slice and in vivo. Here we give the first direct evidence that the neuromodulatory serotoninergic system, in addition to its well-established descending action, has an influence on cortical pain-processing regions. It has been previously shown that mice lacking serotoninergic neurons in the adult CNS show markedly reduced analgesia in response to opioids and antidepressants (Zhao et al., 27). Thus, the serotoninergic system is essential for the action of these drugs. Our study may therefore reveal a new site of action of some of the established chronic pain treatment strategies. Serotonin or combined serotonin-norepinephrine uptake inhibitors (Baron, 29; Marks et al., 29) might exert parts of their analgesic effect on cortical brain areas in addition to the commonly assumed action in the brainstem and spinal cord. Specific targeting of 5-HT 7 receptors might have advantages over serotonin uptake inhibitors, which increase serotonin levels globally in the brain and influence all types of 5-HT receptors with unknown consequences. Thus, our findings may open novel therapeutic venues for neuropathic pain treatment, which could exploit the specific ability of 5-HT 7 receptors to downregulate both spinal cord (Brenchat et al., 21; Viguier et al., 213) and cortical neuronal activity. EXPERIMENTAL PROCEDURES CCI and Von Frey Test All experiments were conducted in accordance with the rules of the veterinary office of the canton Bern, Switzerland. CCI of the sciatic nerve or sham operation was performed on C57BL/6 adult male mice (8 12 weeks). Mice were anesthetized with isoflurane (1% 1.5%), a small incision in the left thigh was made to expose the sciatic nerve, and a silk thread was used to make three consecutive loose ligations around it. For sham surgery the sciatic nerve was exposed but not ligated. In all animal groups, mechanical hyperalgesia was tested using an electronic Von Frey anesthesiometer by slowly applying pressure to the midplantar surface of the hind-paw with the Von Frey filament until a paw withdrawal was evoked and the corresponding stimulation strength was recorded. Patch-Clamp Electrophysiology Seven to ten days after CCI or sham surgery, coronal slices (3-mm thick) containing the ACC were prepared. All recordings were performed at 32 C 33 C. Dual somatic and dendritic whole-cell patch-clamp recordings were performed from identified L5 pyramidal neurons in the rostroventral ACC ( mm below the pial surface, mm rostral to the Bregma and contralateral to the lesion). Pyramidal neurons were selected on the clearly visible, thick apical dendrite that allowed for dual whole-cell patchclamp recordings. This selection criterion resulted in a homogeneous population of pyramidal neurons based on their firing properties and shape of the AP (i.e., all cells possessed a prominent after-hyperpolarization and a significant sag ratio at the soma). First, the soma was patched with a thick-wall borosilicate pipette (resistance 5 9 MU); the dendrite was then patched with a smaller pipette (15 22 UM), and whole-cell configuration was achieved by hyperpolarization. Focal electrical synaptic stimulation was achieved by placing a theta patch pipette located in the inner L1, 3 8 mm lateral to the main apical dendrite. For some experiments, whole-cell recordings were only performed from the apical dendrite. Interneurons and L2/L3 pyramidal cells were investigated by somatic whole-cell recordings. All cells were filled with biocytin and PFA-fixed slices were developed with the avidinbiotin-peroxidase method for Neurolucida reconstructions (Egger et al., 28). The sag ratio was calculated with the equation sag ratio = (V baseline V min )/ (V baseline V steady-state )=V max /V SS where V baseline is the resting membrane potential, V min is the minimum voltage reached soon after the hyperpolarizing current pulse, and V steady-state is the voltage recorded a few milliseconds before the end of the stimulus. For voltage-clamp recordings from the dendrites, the actual voltage was calculated at the steady-state somatic voltage generated by the dendritic voltage step, and the tail current was calculated at the peak. Immunohistochemistry and Imaging Brain slices for fluorescent immunohistochemical labeling were prepared from perfusion-fixed brains, and staining procedures described by Nolan et al. (27) were used. For immunolabeling of HCN channels, slices were incubated with a goat polyclonal anti-hcn1 (1:25) overnight and labeled with a secondary fluorescence-conjugated antiserum (Alexa 488 chicken anti-goat IgG; 1:5). Staining for 5-HT 7 receptors followed the same procedure, using primary rabbit polyclonal anti-5-ht 7 (1:25) and secondary Alexa 488 goat anti-rabbit IgG (1:25). Fluorescently labeled coronal brain slices were imaged using a confocal microscope. Imaging of labeled HCN channels was performed with a 2x objective from the rostral ACC. Imaging settings were kept identical for the Sham and CCI condition. Maximum projections of image stacks (2 frames, step size 2 mm, frame average of 4) were used for analysis in Neuron 86, , April 8, 215 ª215 Elsevier Inc. 243

13 ImageJ. Intensity profiles were derived from the midline to L5. HCN intensity decreased with distance from L1 exponentially. Exponential fits to this decay yielded a characteristic length constant that was compared between the Sham and CCI condition. 5-HT 7 immunofluorescence was imaged with a 4x objective. In Vivo Drug Application For in vivo drug application, two small craniotomies were performed above the ACC (.7 mm rostral to Bregma and.35 mm lateral to midline in each hemisphere). The compounds were mixed with 1-mm Alexa 488 hydrazide and slowly injected (1.75 mm below the skull surface) with micropipettes. A digital stereotaxic system was used for targeted injection. Mice were allowed to recover for 3 min before the first test. Each injection location was evaluated and confirmed by imaging Alexa 488 fluorescence after whole-brain clearing with a 3D ultramicroscope (Dodt et al., 27). A detailed description of all experimental methods is provided in the Supplemental Experimental Procedures. SUPPLEMENTAL INFORMATION Supplemental Information includes Supplemental Experimental Procedures and eight figures and can be found with this article online at org/1.116/j.neuron ACKNOWLEDGMENTS We thank Natalie Nevian for Neurolucida reconstructions, 3D ultramicroscopy, and immunohistochemistry; Hanns-Ulrich Zeilhofer, Isabelle Decosterd, and their groups for training in CCI surgery and behavioral testing; Sigrid Marie Blom and Alberto Bisco for help with the animal preparation; Bina Santoro for help with the HCN1 immunohistochemistry protocol; and Daniela Pietrobon, Rogier Min, Florian Neubauer, Fernando Kasanetz, and Rita Zemankovics for comments on the manuscript. This work was supported by the Swiss National Science Foundation (T.N., grant PPP3_128415). Received: October 3, 214 Revised: January 27, 215 Accepted: February 2, 215 Published: March 26, 215 REFERENCES Airan, R.D., Thompson, K.R., Fenno, L.E., Bernstein, H., and Deisseroth, K. (29). Temporally precise in vivo control of intracellular signalling. Nature 458, Apkarian, A.V., Baliki, M.N., and Geha, P.Y. (29). Towards a theory of chronic pain. Prog. Neurobiol. 87, Arnsten, A.F., Wang, M.J., and Paspalas, C.D. (212). Neuromodulation of thought: flexibilities and vulnerabilities in prefrontal cortical network synapses. Neuron 76, Baliki, M.N., Petre, B., Torbey, S., Herrmann, K.M., Huang, L., Schnitzer, T.J., Fields, H.L., and Apkarian, A.V. (212). Corticostriatal functional connectivity predicts transition to chronic back pain. Nat. Neurosci. 15, Bard, J.A., Zgombick, J., Adham, N., Vaysse, P., Branchek, T.A., and Weinshank, R.L. (1993). Cloning of a novel human serotonin receptor (5- HT7) positively linked to adenylate cyclase. J. Biol. Chem. 268, Baron, R. (29). Neuropathic pain: a clinical perspective. Handbook Exp. Pharmacol Basbaum, A.I., Bautista, D.M., Scherrer, G., and Julius, D. (29). Cellular and molecular mechanisms of pain. Cell 139, Bennett, G.J., and Xie, Y.K. (1988). A peripheral mononeuropathy in rat that produces disorders of pain sensation like those seen in man. Pain 33, Berger, T., Larkum, M.E., and Lüscher, H.R. (21). High I(h) channel density in the distal apical dendrite of layer V pyramidal cells increases bidirectional attenuation of EPSPs. J. Neurophysiol. 85, Blom, S.M., Pfister, J.P., Santello, M., Senn, W., and Nevian, T. (214). Nerve injury-induced neuropathic pain causes disinhibition of the anterior cingulate cortex. J. Neurosci. 34, Brager, D.H., and Johnston, D. (27). Plasticity of intrinsic excitability during long-term depression is mediated through mglur-dependent changes in I(h) in hippocampal CA1 pyramidal neurons. J. Neurosci. 27, Brenchat, A., Nadal, X., Romero, L., Ovalle, S., Muro, A., Sánchez-Arroyos, R., Portillo-Salido, E., Pujol, M., Montero, A., Codony, X., et al. (21). Pharmacological activation of 5-HT7 receptors reduces nerve injury-induced mechanical and thermal hypersensitivity. Pain 149, Chapin, E.M., and Andrade, R. (21). A 5-HT(7) receptor-mediated depolarization in the anterodorsal thalamus. II. Involvement of the hyperpolarizationactivated current I(h). J. Pharmacol. Exp. Ther. 297, Costigan, M., Scholz, J., and Woolf, C.J. (29). Neuropathic pain: a maladaptive response of the nervous system to damage. Annu. Rev. Neurosci. 32, Dembrow, N.C., Chitwood, R.A., and Johnston, D. (21). Projection-specific neuromodulation of medial prefrontal cortex neurons. J. Neurosci. 3, Dodt, H.U., Leischner, U., Schierloh, A., Jährling, N., Mauch, C.P., Deininger, K., Deussing, J.M., Eder, M., Zieglgänsberger, W., and Becker, K. (27). Ultramicroscopy: three-dimensional visualization of neuronal networks in the whole mouse brain. Nat. Methods 4, Duncan, M.J., and Franklin, K.M. (27). Expression of 5-HT7 receptor mrna in the hamster brain: effect of aging and association with calbindin-d28k expression. Brain Res. 1143, Egger, V., Nevian, T., and Bruno, R.M. (28). Subcolumnar dendritic and axonal organization of spiny stellate and star pyramid neurons within a barrel in rat somatosensory cortex. Cereb. Cortex 18, Emery, E.C., Young, G.T., Berrocoso, E.M., Chen, L., and McNaughton, P.A. (211). HCN2 ion channels play a central role in inflammatory and neuropathic pain. Science 333, Eto, K., Wake, H., Watanabe, M., Ishibashi, H., Noda, M., Yanagawa, Y., and Nabekura, J. (211). Inter-regional contribution of enhanced activity of the primary somatosensory cortex to the anterior cingulate cortex accelerates chronic pain behavior. J. Neurosci. 31, Fan, Y., Fricker, D., Brager, D.H., Chen, X., Lu, H.C., Chitwood, R.A., and Johnston, D. (25). Activity-dependent decrease of excitability in rat hippocampal neurons through increases in I(h). Nat. Neurosci. 8, Frick, A., Magee, J., and Johnston, D. (24). LTP is accompanied by an enhanced local excitability of pyramidal neuron dendrites. Nat. Neurosci. 7, George, M.S., Abbott, L.F., and Siegelbaum, S.A. (29). HCN hyperpolarization-activated cation channels inhibit EPSPs by interactions with M-type K(+) channels. Nat. Neurosci. 12, Gulledge, A.T., and Stuart, G.J. (23). Action potential initiation and propagation in layer 5 pyramidal neurons of the rat prefrontal cortex: absence of dopamine modulation. J. Neurosci. 23, Häusser, M., Spruston, N., and Stuart, G.J. (2). Diversity and dynamics of dendritic signaling. Science 29, Hedlund, P.B. (29). The 5-HT7 receptor and disorders of the nervous system: an overview. Psychopharmacology (Berl.) 26, Johansen, J.P., Fields, H.L., and Manning, B.H. (21). The affective component of pain in rodents: direct evidence for a contribution of the anterior cingulate cortex. Proc. Natl. Acad. Sci. USA 98, Jung, S., Bullis, J.B., Lau, I.H., Jones, T.D., Warner, L.N., and Poolos, N.P. (21). Downregulation of dendritic HCN channel gating in epilepsy is mediated by altered phosphorylation signaling. J. Neurosci. 3, Kalmbach, B.E., Chitwood, R.A., Dembrow, N.C., and Johnston, D. (213). Dendritic generation of mglur-mediated slow afterdepolarization in layer 5 neurons of prefrontal cortex. J. Neurosci. 33, Neuron 86, , April 8, 215 ª215 Elsevier Inc.

14 King, T., Vera-Portocarrero, L., Gutierrez, T., Vanderah, T.W., Dussor, G., Lai, J., Fields, H.L., and Porreca, F. (29). Unmasking the tonic-aversive state in neuropathic pain. Nat. Neurosci. 12, Koga, K., Li, X., Chen, T., Steenland, H.W., Descalzi, G., and Zhuo, M. (21). In vivo whole-cell patch-clamp recording of sensory synaptic responses of cingulate pyramidal neurons to noxious mechanical stimuli in adult mice. Mol. Pain 6, 62. Kuner, R. (21). Central mechanisms of pathological pain. Nat. Med. 16, Kupferman, J.V., Basu, J., Russo, M.J., Guevarra, J., Cheung, S.K., and Siegelbaum, S.A. (214). Reelin signaling specifies the molecular identity of the pyramidal neuron distal dendritic compartment. Cell 158, LaGraize, S.C., Labuda, C.J., Rutledge, M.A., Jackson, R.L., and Fuchs, P.N. (24). Differential effect of anterior cingulate cortex lesion on mechanical hypersensitivity and escape/avoidance behavior in an animal model of neuropathic pain. Exp. Neurol. 188, Lee, M.C., and Tracey, I. (213). Imaging pain: a potent means for investigating pain mechanisms in patients. Br. J. Anaesth. 111, Li, X.Y., Ko, H.G., Chen, T., Descalzi, G., Koga, K., Wang, H., Kim, S.S., Shang, Y., Kwak, C., Park, S.W., et al. (21). Alleviating neuropathic pain hypersensitivity by inhibiting PKMzeta in the anterior cingulate cortex. Science 33, London, M., and Häusser, M. (25). Dendritic computation. Annu. Rev. Neurosci. 28, Lörincz, A., Notomi, T., Tamás, G., Shigemoto, R., and Nusser, Z. (22). Polarized and compartment-dependent distribution of HCN1 in pyramidal cell dendrites. Nat. Neurosci. 5, Losonczy, A., Makara, J.K., and Magee, J.C. (28). Compartmentalized dendritic plasticity and input feature storage in neurons. Nature 452, Magee, J.C. (1998). Dendritic hyperpolarization-activated currents modify the integrative properties of hippocampal CA1 pyramidal neurons. J. Neurosci. 18, Magee, J.C. (1999). Dendritic lh normalizes temporal summation in hippocampal CA1 neurons. Nat. Neurosci. 2, Magee, J.C., and Johnston, D. (25). Plasticity of dendritic function. Curr. Opin. Neurobiol. 15, Major, G., Larkum, M.E., and Schiller, J. (213). Active properties of neocortical pyramidal neuron dendrites. Annu. Rev. Neurosci. 36, Marks, D.M., Shah, M.J., Patkar, A.A., Masand, P.S., Park, G.Y., and Pae, C.U. (29). Serotonin-norepinephrine reuptake inhibitors for pain control: premise and promise. Curr. Neuropharmacol. 7, Momin, A., Cadiou, H., Mason, A., and McNaughton, P.A. (28). Role of the hyperpolarization-activated current Ih in somatosensory neurons. J. Physiol. 586, Nestor, M.W., and Hoffman, D.A. (212). Aberrant dendritic excitability: a common pathophysiology in CNS disorders affecting memory? Mol. Neurobiol. 45, Ning, L., Ma, L.Q., Wang, Z.R., and Wang, Y.W. (213). Chronic constriction injury induced long-term changes in spontaneous membrane-potential oscillations in anterior cingulate cortical neurons in vivo. Pain Physician 16, E577 E589. Nolan, M.F., Malleret, G., Dudman, J.T., Buhl, D.L., Santoro, B., Gibbs, E., Vronskaya, S., Buzsáki, G., Siegelbaum, S.A., Kandel, E.R., and Morozov, A. (24). A behavioral role for dendritic integration: HCN1 channels constrain spatial memory and plasticity at inputs to distal dendrites of CA1 pyramidal neurons. Cell 119, Nolan, M.F., Dudman, J.T., Dodson, P.D., and Santoro, B. (27). HCN1 channels control resting and active integrative properties of stellate cells from layer II of the entorhinal cortex. J. Neurosci. 27, Ossipov, M.H., Dussor, G.O., and Porreca, F. (21). Central modulation of pain. J. Clin. Invest. 12, Palmer, L.M. (214). Dendritic integration in pyramidal neurons during network activity and disease. Brain Res. Bull. 13, 2 1. Perl, E.R. (27). Ideas about pain, a historical view. Nat. Rev. Neurosci. 8, Pollak Dorocic, I., Fürth, D., Xuan, Y., Johansson, Y., Pozzi, L., Silberberg, G., Carlén, M., and Meletis, K. (214). A whole-brain atlas of inputs to serotonergic neurons of the dorsal and median raphe nuclei. Neuron 83, Poolos, N.P., and Johnston, D. (212). Dendritic ion channelopathy in acquired epilepsy. Epilepsia 53 (9), Porreca, F., Ossipov, M.H., and Gebhart, G.F. (22). Chronic pain and medullary descending facilitation. Trends Neurosci. 25, Porrino, L.J., and Goldman-Rakic, P.S. (1982). Brainstem innervation of prefrontal and anterior cingulate cortex in the rhesus monkey revealed by retrograde transport of HRP. J. Comp. Neurol. 25, Qu, C., King, T., Okun, A., Lai, J., Fields, H.L., and Porreca, F. (211). Lesion of the rostral anterior cingulate cortex eliminates the aversiveness of spontaneous neuropathic pain following partial or complete axotomy. Pain 152, Saab, C.Y. (212). Pain-related changes in the brain: diagnostic and therapeutic potentials. Trends Neurosci. 35, Sandkühler, J. (29). Models and mechanisms of hyperalgesia and allodynia. Physiol. Rev. 89, Shackman, A.J., Salomons, T.V., Slagter, H.A., Fox, A.S., Winter, J.J., and Davidson, R.J. (211). The integration of negative affect, pain and cognitive control in the cingulate cortex. Nat. Rev. Neurosci. 12, Shah, M.M., Anderson, A.E., Leung, V., Lin, X., and Johnston, D. (24). Seizure-induced plasticity of h channels in entorhinal cortical layer III pyramidal neurons. Neuron 44, Shah, M.M., Hammond, R.S., and Hoffman, D.A. (21). Dendritic ion channel trafficking and plasticity. Trends Neurosci. 33, Shyu, B.C., Sikes, R.W., Vogt, L.J., and Vogt, B.A. (21). Nociceptive processing by anterior cingulate pyramidal neurons. J. Neurophysiol. 13, Spruston, N. (28). Pyramidal neurons: dendritic structure and synaptic integration. Nat. Rev. Neurosci. 9, Stuart, G., and Spruston, N. (1998). Determinants of voltage attenuation in neocortical pyramidal neuron dendrites. J. Neurosci. 18, Terashima, Y., Kawamata, M., Takebayashi, T., Tanaka, S., Tanimoto, K., and Yamashita, T. (211). Changes in synaptic transmission of substantia gelatinosa neurons in a rat model of lumbar radicular pain revealed by in vivo patch-clamp recording. Pain 152, Thibault, K., Lin, W.K., Rancillac, A., Fan, M., Snollaerts, T., Sordoillet, V., Hamon, M., Smith, G.M., Lenkei, Z., and Pezet, S. (214). BDNF-dependent plasticity induced by peripheral inflammation in the primary sensory and the cingulate cortex triggers cold allodynia and reveals a major role for endogenous BDNF as a tuner of the affective aspect of pain. J. Neurosci. 34, Tracey, I., and Mantyh, P.W. (27). The cerebral signature for pain perception and its modulation. Neuron 55, Treede, R.D., Kenshalo, D.R., Gracely, R.H., and Jones, A.K. (1999). The cortical representation of pain. Pain 79, Vasefi, M.S., Yang, K., Li, J., Kruk, J.S., Heikkila, J.J., Jackson, M.F., MacDonald, J.F., and Beazely, M.A. (213). Acute 5-HT7 receptor activation increases NMDA-evoked currents and differentially alters NMDA receptor subunit phosphorylation and trafficking in hippocampal neurons. Mol. Brain 6, 24. Viguier, F., Michot, B., Hamon, M., and Bourgoin, S. (213). Multiple roles of serotonin in pain control mechanisms implications of 5 HT 7 and other 5- HT receptor types. Eur. J. Pharmacol. 716, Vogt, B.A. (25). Pain and emotion interactions in subregions of the cingulate gyrus. Nat. Rev. Neurosci. 6, Neuron 86, , April 8, 215 ª215 Elsevier Inc. 245

15 Wager, T.D., Atlas, L.Y., Lindquist, M.A., Roy, M., Woo, C.W., and Kross, E. (213). An fmri-based neurologic signature of physical pain. N. Engl. J. Med. 368, Wang, J., Chen, S., Nolan, M.F., and Siegelbaum, S.A. (22). Activity-dependent regulation of HCN pacemaker channels by cyclic AMP: signaling through dynamic allosteric coupling. Neuron 36, Wang, H., Xu, H., Wu, L.J., Kim, S.S., Chen, T., Koga, K., Descalzi, G., Gong, B., Vadakkan, K.I., Zhang, X., et al. (211). Identification of an adenylyl cyclase inhibitor for treating neuropathic and inflammatory pain. Sci. Transl. Med. 3, ra3. Weissbourd, B., Ren, J., DeLoach, K.E., Guenthner, C.J., Miyamichi, K., and Luo, L. (214). Presynaptic partners of dorsal raphe serotonergic and GABAergic neurons. Neuron 83, Williams, S.R., and Mitchell, S.J. (28). Direct measurement of somatic voltage clamp errors in central neurons. Nat. Neurosci. 11, Williams, S.R., and Stuart, G.J. (2). Site independence of EPSP time course is mediated by dendritic I(h) in neocortical pyramidal neurons. J. Neurophysiol. 83, Xu,H.,Wu,L.J.,Wang,H.,Zhang,X.,Vadakkan,K.I.,Kim,S.S.,Steenland, H.W., and Zhuo, M. (28). Presynaptic and postsynaptic amplifications of neuropathic pain in the anterior cingulate cortex. J. Neurosci. 28, Zhang, L., Renaud, L.P., and Kolaj, M. (29). Properties of a T-type Ca2+channel-activated slow afterhyperpolarization in thalamic paraventricular nucleus and other thalamic midline neurons. J. Neurophysiol. 11, Zhang, Y., Bonnan, A., Bony, G., Ferezou, I., Pietropaolo, S., Ginger, M., Sans, N., Rossier, J., Oostra, B., LeMasson, G., and Frick, A. (214). Dendritic channelopathies contribute to neocortical and sensory hyperexcitability in Fmr1(-/y) mice. Nat. Neurosci. 17, Zhao, Z.Q., Chiechio, S., Sun, Y.G., Zhang, K.H., Zhao, C.S., Scott, M., Johnson, R.L., Deneris, E.S., Renner, K.J., Gereau, R.W., 4th, and Chen, Z.F. (27). Mice lacking central serotonergic neurons show enhanced inflammatory pain and an impaired analgesic response to antidepressant drugs. J. Neurosci. 27, Neuron 86, , April 8, 215 ª215 Elsevier Inc.

16 Neuron Supplemental Information Dysfunction of Cortical Dendritic Integration in Neuropathic Pain Reversed by Serotoninergic Neuromodulation Mirko Santello and Thomas Nevian

17 A B D F H Length constant (μm) I soma Soma 1 Hz Soma 5 Hz Soma Sham 1 μm Dendrite 4 mv AP frequency (Hz) 1 na 5 ms I 1 Hz Dendrite 5 Hz Dendrite Baseline 2 mv 2 ms 5 Hz Dendrite C E G AP amplitude (mv) Latency (ms) Half width (ms) AP amplitude ratio (1 th /1 st ) AP amplitude ratio (1 th /1 st ) 4-AP (5 mm) J Sham CCI 1 Hz 5 Hz Baseline Distance (μm) Distance (μm) Distance (μm) Distance (μm) Distance (μm) 5 Hz Dendrite Cd 2+ and Ni mv 2 ms 5 ms 1 ms Figure S1 (related to Figure 1). Back-propagation of action potentials is not modified by CCI (A) Reconstruction of a L5 pyramidal neuron after simultaneous somatic (gray pipette) and dendritic (blue pipette) whole-cell recording (distance 187 μm). (B) Brief current injection at the soma (bottom) evoked a somatic AP (black) that back-propagated into the apical dendrite (blue). (C) AP amplitude (upper graph), latency between somatic and dendritic AP at peak (middle graph) and AP width at half maximum (lower graph) plotted against distance from the soma. All of these characteristic features of AP-backpropagation were not different between Sham and CCI groups (AP amplitude p =.37; latency p =.49; half width p =.37). These results suggest that voltage dependent Na + and K + channels that influence the attenuation, propagation and broadening of APs along the apical dendrite (Kampa and Stuart, 26; Spruston, 28; Stuart and Hausser, 21) were not altered by nerve injury. The exponential fit to the distance dependence of the AP amplitude yielded a length constant of λ = 691 μm for the Sham and of λ = 74 μm for the CCI group. The linear fit to the latency yielded an AP propagation velocity of 45 μm/ms for the Sham and 446 μm/ms for the CCI group. Thus, apical dendrites of L5 pyramidal neurons in ACC show slightly better back-propagation of APs than apical dendrites of L5 cells in somatosensory cortex (λ = 587 μm), but conduction velocity is a bit slower (somatosensory cortex, 58 μm/ms) (Larkum et al., 21). At distance x = μm AP amplitude and AP half-width at the soma are plotted (amplitude: CCI ± 1.81 mv, Sham ± 2.18 mv, p =.31; half width: CCI.747 ±.36 ms, Sham.751 ±.29 ms, p =.47). (D) Dendrites of ACC L5 pyramidal cells show progressive frequency and distance-dependent attenuation of the AP amplitude in a train. A train of APs at 1 Hz evoked at the soma and recorded in the dendrite (distance 187 μm). Notice that the amplitude of the AP decreased in the dendrite during the train while staying constant at the soma. (E) Amplitude ratio between the 1 th and the 1 st AP showing the distance dependency of this phenomenon. All values in the plots are averages of 5 consecutive sweeps. Distance dependency of the decrease in amplitude was fitted with an exponential function. No difference between the CCI and Sham groups was found. (F) A 5-Hz train of 1 APs showed stronger activity-dependent attenuation in the dendrite. (G) Amplitude ratio between the 1 th and the 1 st AP showing the distance dependency for a 5-Hz train. Notice that the 1th AP amplitude was 5%-6% of the 1 st for distal recordings. No difference between the CCI and Sham groups was found (p =.27). (H) Length constants derived from the exponential fits in (B) and (D) as a function of AP frequency. No difference between CCI and Sham was found for any frequency. (I) The frequency-dependent attenuation of APs in the dendrite might be due to Na + channel inactivation or to activation of K + channels. 4-aminopyridine (I A K + channel blocker, 5 mm) did not abolish the use dependent attenuation of the AP amplitude as in CA1 hippocampal pyramidal neurons (Sun et al., 211). Dendritically recorded APs (distance 172 μm) at 5 Hz before (blue) and after (cyan) drug application (n = 2). Inset shows the overlay of the first APs in both conditions illustrating the broadening of the AP in the presence of 4-AP. (J) Bath application of Cd 2+ (5 μm ) and Ni 2+ (5 μm) largely blocked the activity dependent decrease in AP amplitude (example distance 22 μm, n = 8). These results suggest that voltage gated Ca 2+ channels may be involved in the phenomenon. The inset illustrates that in agreement with previous observations (Gulledge and Stuart, 23) dendritic AP half-width was also increased by Cd 2+ and Ni 2+. This effect was presumably attributable to a block of Ca 2+ -activated K + channels (Benhassine and Berger, 29). In summary, this set of experiments suggests that the ion channels that are responsible for the frequency-dependent attenuation of APs in ACC apical dendrites are not modified by peripheral nerve injury.

18 A B S D Figure S2 (related to Figure 1). Passive properties of L5 dendrites are not modified by CCI (A) Neurolucida reconstruction of a biocytin-filled L5 pyramidal cell in the ACC showing the recording locations (distance 242 μm). C D Soma/Dend H Soma/Dend Reciprocity Sham CCI Steady-state D S 2. EPSP Slope D S Dend/Soma Dend/Soma Distance (μm) G2. 5 ms EPSP Amplitude D S Soma/Dend Sham CCI Distance (μm) Dend/Soma I soma I dend D S 2 ms Steady-state S D Distance (μm) EPSP Slope S D Distance (μm) 1 mv 3 pa Distance (μm) Distance (μm) E Sham CCI <15 >15 Distance (μm) Distance (μm) R in Dend/Soma F D S I dend 1 μm 1 mv 4 pa R in Dend/Soma S D I soma EPSP Amplitude S D (B) Example traces from the cell shown in (A). Steady-state -3 pa current injections into the soma (upper traces) and dendrite (lower traces) resulted in a local membrane hyperpolarization that efficiently propagated along the apical dendrite. Whereas dendrite to soma (D S) voltage attenuation was negligible, voltage spread in the other direction from the soma to the dendrite (S D) was substantially attenuated. (C) The reciprocity of steady-state responses of signals propagating in both directions (comparing voltage traces recorded from the non-current injecting electrode in (B)) showed that the dendrites behaved as linear cables in the subthreshold range. (D) Distance dependency of steady-state attenuation from dendrite to the soma (Steady-state D S ) and vice versa (Steady-state S D ) comparing the CCI (red) and Sham (blue) conditions. The solid lines represent exponential fits to the data yielding the length constants for each condition (D S: λ Sham >> 5 μm, λ CCI = 4657 μm; S D: λ Sham = 3 μm, λ CCI = 292 μm). Attenuation is not different between Sham and CCI in both directions (D S: p =.16; S D: p =.26). (E) Normalized input resistance (R in Dend/Soma) plotted as a function of the distance from the soma. Notice that in the ACC dendritic R in decreases with the distance from the soma. This feature is contrary to the distance dependence of R in in apical dendrites of L5 pyramidal neurons in somatosensory cortex. (F) EPSP-shaped current injections (bottom, τ rise = 4 ms, τ decay = 1 ms) injected either at the dendrite (blue) and recorded at the soma (black) or injected at the soma and recorded at the dendrite. (G) Attenuation of the peak EPSP amplitude as a function of the distance from the soma. Amplitude of current injection was set to cause a maximal subthreshold EPSP. No difference in voltage attenuation was detected between the Sham and CCI condition (D S: p =.24; S D: p =.39). The solid lines represent exponential fits to the data yielding the length constants for each condition (D S: λ Sham =1225 μm, D S: λ CCI = 175 μm, S D: λ Sham = 468 μm, λ CCI = 499 μm). (H) Normalized EPSP slope as a function of the distance from the soma. In both directions a distance-dependent reduction of the slope was observed suggesting some dendritic filtering of subthreshold EPSP-like potentials. Nevertheless, there was no difference between the Sham and CCI condition (D S p =.21; S D p =.45). This set of experiments suggests that the passive membrane properties of L5 pyramidal neurons in ACC are not modified by CCI.

19 A F 7 Hz G Dendrite H Cd 2+ Ni 2+ B 1 μm Weak stimulus at AP threshold Soma Soma 1 Hz Dendrite Soma C D Latency (ms) Soma I dend At AP threshold Sham CCI 1 2 Distance (μm) Dendrite 8 pa Strong stimulus I dend I dend 15 pa 13 pa E Latency (ms) Dendrite Dendrite Soma Strong stimulus (another cell) Dendrite Soma 5 ms 5 ms Strong stimulus (15 pa) 1 2 Distance (μm) 4 mv 2 na I K n L ADP Dendrite (mv) ms Baseline 2 mv Frequency (Hz) Sham n Critical frequency (Hz) ADP Dendrite (mv) mv 1 ms Frequency (Hz) CCI Cd 2+ Ni Critical frequency (Hz) Critical Frequency (Hz) J Relative ADP (mv) pa 12 pa 18 pa Baseline Sham ** Cd 2+ Ni 2+ CCI Baseline Cd 2+ Ni 2+ 2 mv 5 ms Figure S3 (related to Figure 1). Fast local dendritic spikes can be evoked in L5 pyramidal neuron dendrites in the ACC, but they display little Ca 2+ regenerativity (A) Neurolucida reconstruction of a biocytin-filled L5 pyramidal cell in the ACC showing the experimental configuration of simultaneous somatic and dendritic whole-cell recordings (distance 163 μm). (B) Small amplitude EPSP-shaped current injection in the dendrite (τ rise = 4 ms, τ decay = 1 ms, intensity near somatic AP threshold), triggered an AP that was first detected in the soma (black) and subsequently in the dendrite (blue). A stronger stimulus (15 pa) gave rise to a fast dendritic spike that preceded the somatic AP (Schiller et al., 1997; Stuart et al., 1997). After a few milliseconds, a second AP was generated first at the soma. Notice the reduced amplitude of the 2 nd AP, which could be explained by local inactivation of the dendritic Na + channels by the local spike. (C) Occasionally (n = 2), a local spike was generated with a high stimulation intensity that did not result in somatic AP generation. In these cases the dendritic potential was largely attenuated towards the soma (75 % attenuation in this case, distance 226 μm). (D) Time latency between somatic and dendritic spike (t soma -t dendrite ) plotted as a function of the distance from the soma with a weak EPSP-like current injection near the somatic AP threshold. Red dots correspond to CCI recordings whereas blue dots are from Sham. Few distal dendrites showed a negative latency indicating a dendritic spike that preceded the somatic AP. (E) Latency between the first somatic and dendritic spike evoked with a large amplitude EPSP-like current injection (15 pa) as function of the distance from the soma. Note that distal recordings always showed negative latency values, which corresponded to dendritic spikes preceding somatic APs. These data suggest that local dendritic spikes can be evoked in ACC L5 pyramidal cell dendrites. The kinetics of the spike, its dependency on dendritic location, stimulus strength and previous firing history are reminiscent of Na + spikes recorded from dendrites of other cell types (Golding and Spruston, 1998; Larkum et al., 27; Nevian et al., 27). No significant difference was detected comparing latencies from CCI and Sham dendrites (proximal recordings p =.19; distal recordings p =.38). (F) A train of 4 APs evoked at the soma (black, evoked by 2 ms na current injections) that back-propagate into the dendrite (red) giving rise to a small after depolarization (ADP) at 1 Hz (arrow) but not at 7 Hz. The ADP results from the opening of voltage-dependent Ca 2+ channels in the distal apical dendrite evoked by the AP-induced dendritic depolarization above a characteristic critical frequency (Larkum et al., 1999a). Distance of recording was 22 μm from the soma. (G) Dendritic and somatic recordings of 4 APs evoked at increasing frequencies from 4 to 18 Hz (1 Hz steps) and aligned to the last action potential in each train. The amplitude of the dendritic ADP is calculated at the time indicated by the dashed line relative to the resting membrane potential. Notice that the somatic ADP is almost absent. None of the dendritic recordings close to the bifurcation displayed the large sustained regenerative potentials typical of Ca 2+ spikes in pyramidal neurons in somatosensory cortex (Larkum et al., 1999b). (H) After bath-application of the voltage gated Ca 2+ channel blockers Cd 2+ (5 μm) and Ni 2+ (5 μm), the dendritic ADP was almost abolished.

20 (I) Amplitude of the ADP as function of the AP frequency and fitted with a sigmoidal function (left) from the example in (B). The dashed line indicates the critical frequency of 9 Hz. In the presence of Cd 2+ and Ni 2+ no sigmoidal fit of the ADP was possible (right). (J) Summary bar graph showing that the relative dendritic ADP amplitude after bath application of Cd 2+ and Ni 2+ was significantly reduced compared to control (n = 8, 3 Sham and 5 CCI). (K) Histogram of the critical frequency distributions of 22 CCI and 24 Sham dendrites. The critical frequency was defined as the frequency at the half amplitude increase calculated from sigmoidal fits to the frequency-to-adp relationship as illustrated in (D). Bar graphs (right) showing the average critical frequencies for CCI and Sham dendrites, which were identical (p =.4). No difference was also detected by using trains of 3 or 5 APs (p =.29 and p =.5 respectively, same n). For more proximal sites, in 3 CCI dendrites (average distance 71.8 μm) and 3 Sham dendrites (average distance 97.2 μm) no critical frequency was detectable below 2 Hz. In a larger sample of somatic recordings (CCI n = 9, Sham n = 97), a critical frequency was detected in 33% of CCI and 35% of Sham neurons tested with a train of 3 APs, 61% of CCI and 72% of Sham neurons tested with a train of 4 APs and 71% of CCI and 72% of Sham neurons tested with a train of 5 APs. The average critical frequency was also not different in all groups (3 APs: CCI 77.3 ± 3.2 Hz, Sham 75.7 ± 2.4 Hz, p =.34; 4 APs: CCI 85.5 ± 2.48 Hz, Sham 82. ± 1.8 Hz, p =.13; 5 APs: CCI 86.3 ± 1.5 Hz, Sham 83 ± 1.5 Hz, p =.14). (L) Dendritic EPSP-shaped current injections at increasing intensities (6, 12 and 18 pa) did not evoke prolonged Ca 2+ spikes and Cd 2+ and Ni 2+ had little effect on the dendritic potential as illustrated in this example (n = 8). This set of experiments suggests that, similar to prefrontal cortex (Gulledge and Stuart, 23), dendritic Ca 2+ electrogenesis is significantly less pronounced in ACC L5 neurons as compared to L5 pyramidal neurons in the somatosensory cortex. Furthermore, our experiments show that the Ca 2+ regenerativity is not affected by CCI surgery. ** p <.1.

21 A 215 μm B Soma Dendrite Soma Stimulation Dendrite -73 mv -67 mv Integral ZD 7288 (1 mm) C A 2ndEPSP (mv) Integral 2ndEPSP (ms.mv) ZD 7288 Soma ZD 7288 ** Stimulation intensity (V) ** *** ** Integral 2ndEPSP (ms.mv) Stimulation intensity (V) * * A 2ndEPSP (mv) ms Dendrite * ** *** 1 mv Stimulation intensity (V) *** Stimulation intensity (V) * ** -83 mv -77 mv D E Integral 5thEPSP (ms.mv) A 5thEPSP (mv) Soma 2 1 Dendrite Soma Dendrite * 3 ** Integral ZD 7288 ** ZD 7288 I dend Integral 5thEPSP (ms.mv) A 5thEPSP (mv) ms ** 4 ms ZD 7288 ZD mv 1 na 1 mv Figure S4 (related to Figures 2 & 6). EPSP integration in L5 pyramidal neurons in the ACC is regulated by I h. (A) Top, experimental configuration showing the somatic and dendritic recording sites and the position of the extracellular electrode used for synaptic stimulation (inner L1). A 3 pa hyperpolarizing current step resulted in a hyperpolarization of the membrane and the activation of Ih resulting in a clear voltage sag at the soma and even more pronounced in the dendrite (black traces). The sag was abolished by bath application of the HCN channel blocker ZD 7288 (1 μm, 15 minutes, yellow). Notice that the rebound depolarization was also abolished and that the input resistance was increased. V m at soma was -72 mv () and -81 mv (ZD 7288). V m at dendrite was -67 mv () and -77 mv (ZD 7288). (B) Two consecutive EPSPs evoked by extracellular synaptic stimulation in L1 (5 Hz, 1-4 V) in control conditions (top) and in the presence of ZD 7288 (bottom). The example shows that, at both soma and dendrite, the evoked EPSPs were broader and their integration was enhanced. Furthermore, peak amplitudes were larger due to the increased input resistance. (C) Quantification of amplitude and time integral of the second EPSP of a paired-pulse stimulation as illustrated in (B) as a function of the stimulation intensity (n = 5). Both at the soma (left) and at the dendrite (right) blockade of HCN channels increased the amplitude and the time integral of the second EPSP. V m at soma was -69 mv () and -78 mv (ZD 7288). V m at dendrite was -59 mv () and -69 mv (ZD 7288). The integral was calculated in the time interval indicated in (B). The large increase in the amplitude was mainly due to the increase in the input resistance, while the slower decay reflected the missing influence of I h. (D) Voltage recorded at the soma and the dendrite in response to 5 EPSP-shaped current injections (4 pa) given at 5 Hz via the dendritic pipette. In the presence of ZD 7288 the summation of the EPSPs was increased. (E) Average increases in amplitude and time integral of the last EPSP-like depolarization of the 5 Hz train (n = 3). The current amplitude was the same before and after injection and was tuned to subthreshold depolarizations. Both amplitude and integral of the last EPSP were increased after I h blockade. All these results are consistent with previous reports of HCN channels shaping dendritic integration of synaptic inputs (Berger et al., 21; Day et al., 25; Magee, 1998, 1999; Williams and Stuart, 2). * p <.5, ** p <.1 *** p <.1

22 C HCN HCN 1 μm 6 NS 4 2 I λhcn = 46 μm C λhcn = 39 μm C B am CCI Sh Sham λhcn (μm) A 1 a.u. Figure S5 (related to Figure 3). HCN channel distribution in motor cortex is not modified by CCI. (A) Example images of immunohistochemical antibody stainings of HCN channels in the motor cortex 9 μm lateral from the midline in Sham and CCI conditions. (B) Intensity profile of HCN channel staining from the pia to layer 5. Single exponential fits to the intensity profile yielded a length constant (ληcν) for the distribution of HCN channels (blue, Sham; red CCI). (C) Averaged length constants for the Sham and CCI condition were not different in the motor cortex.

23 A F Figure S6 (related to Figure 5). The 5-HT 7 antagonist SB occludes the effect of on I h and synaptic integration in CCI neurons, while alone does not affect single EPSP amplitude, integral, resting membrane potential and input resistance at the soma in CCI L5 pyramidal neurons. B C V dend -V soma D -62 mv -65 mv -73 mv Integral 4 ms E A 2ndEPSP (mv) Soma Soma +SB SB Integral Sag ratio (V max /V SS ) mv + SB μm Soma Dendrite Integral 2ndEPSP (ms.mv) Stimulation intensity (V) 2 1 Dendrite Sag ratio (V max /V SS ) + SB mv 2 ms Dendrite + SB mv Stimulation intensity (V) G H A EPSP (mv) Integral EPSP (ms.mv) V m (mv) R i (MΩ) μm Integral 1 Stimulation 1 mv 2 ms Time (s) In order to show the specificity of for 5-HT 7 receptors, we tested the residual effects of after blocking the 5-HT 7 receptors with the highly specific antagonist SB (1 μm). (A) Neurolucida reconstruction of a L5 pyramidal cell with somatic and dendritic pipettes showing the recording locations. (B) A 5 pa hyperpolarizing current injection at the soma or the dendrite resulted in the characteristic voltage sag by the activation of I h during control recordings (black traces). In the presence of SB 26997, (3 nm, at least 15 minutes) did not cause any change in the voltage response to the hyperpolarizing current step. V m at soma was -61 mv () and -66 mv (+SB26997). V m at dendrite was -6 mv () and -61 mv (+SB26997). (C) Summary of the effect of in the presence of the specific 5-HT 7 antagonist SB No dendrite-specific depolarization in ACC neurons from CCI animals was observed (p =.68, n = 5, average distance 152 μm). The corresponding somatic membrane potential was also not modified ( 71.9 ± 2.56 mv, 73.1 ± 2.31 mv, p =.16, n = 5). Therefore, no change in the difference in the resting membrane potential between soma and dendrite (left graph) was observed. The sag ratio under this condition was also not modulated in the soma or the dendrite (middle graph, soma p =.14, n = 7; right graph, dendrite p =.39, n = 5). (D) Paired-pulse synaptic stimulation (5 Hz) of the inner L1 at increasing stimulation intensities (1-5 V). Recordings were performed in the presence of SB Additional bath application of (blue) had only minor effects on EPSP integration as compared to control (black). (E) Amplitude and integral of the second somatic EPSP as a function of synaptic stimulation intensity. SB was present during the entire experiment blocking the effect of bath application (p >.5 for every stimulation intensity, n = 5). Single EPSP amplitude and integral were also not changed (p =.13 and p =.11 respectively). These results suggest that exerted its effect on I h specifically by the activation of 5-HT 7 receptors. (F) Biocytin-filled L5 pyramidal neuron illustrating the location of the stimulation electrode in L1. (G) Example trace of EPSPs before (control, black) and after (3 nm, blue). Notice that the EPSP slightly decayed faster after. (H) Average peak EPSP amplitudes and integral plotted over time in L5 pyramidal neurons before and during bath application of (3 nm; blue shaded area; n = 1). The peak EPSP amplitude was not significantly modified (control 5.6 ±.8 mv, 5. ±.9 mv, p =.9, n = 1) as well as the AP firing threshold (control ± 13.8 pa, ± 1.42 pa, p =.8, n = 9). The EPSP time integral showed a slight tendency to decrease (p =.14) that was due to the faster EPSP decay in the presence of. The integral was calculated in the time interval indicated in (G). concentration was chosen to activate mainly 5-HT 7 receptors and to minimize contributions from other 5-HT receptor subtypes (e.g. 5-HT 1 activation (Zhang et al., 29). This was reflected by the fact that somatic resting membrane potential (V m : control ± 1.21 mv, ± 1.17 mv, n = 1, p =.8) and input resistance (R i : control 38.3 ± 2.2, 36.2 ± 2.3, p =.27) were not significantly affected by. Recordings from Sham preparations gave similar results (n = 7, data not shown). These results suggest that activation of 5-HT 7 receptors mainly affects dendritic I h but has little or no effect on the basic somatic or single EPSP properties.

24 A D F H J Layer 2/3 pyramidal cell 1 μm Integral Layer 2/3 Layer 5 1 μm Stimulation -86 mv -86 mv -73 mv -73 mv -69 mv -69 mv 4 ms -72 mv -72 mv 4 ms 1 mv 5 mv B E K A EPSP (mv) A 2ndEPSP (mv) -86 mv -86 mv 2 ms A EPSP (mv) A 2ndEPSP (mv) I 2 ms 4 mv 4 mv 4 pa 2 ms Stimulation intensity (V) Stimulation intensity (V) 4 mv G Layer 1 AP threshold (pa) Stimulation intensity (V) Stimulation intensity (V) Integral EPSP (ms.mv) Integral 2ndEPSP (ms.mv) Integral EPSP (ms.mv) Integral 2ndEPSP (ms.mv) C AP threshold (pa) Sag ratio (V max /V SS ) Stimulation intensity (V) Stimulation intensity (V) -68 mv -7 mv Sag ratio (V max /V SS ) Stimulation intensity (V) Stimulation intensity (V) Figure S7 (related to Figure 5). L2/3 pyramidal cells and interneurons in the ACC are not affected by. (A) Reconstruction of a L2/3 pyramidal cell showing the location of the electrodes used for somatic whole-cell recordings and extracellular synaptic stimulation. (B) Examples of evoked responses elicited by 6 ms current injections of 3 and 45 pa in the CCI neuron in (A) before (black) and after bath application of (blue, 3 nm). The insets show the initial voltage drop at an extended scale illustrating that no hyperpolarization induced voltage sag was present in these cells, consistent with previous reports (Larkum et al., 27). (C) Summary of the average AP threshold and sag ratio under control conditions and in the presence of. Neither of these parameters were affected by the drug (AP threshold p =.31; sag p =.5347, n = 7 for both, 3 cells were in the L3 and 4 cells in L2). The Sag was quantified from voltage responses to a 5 pa hyperpolarizing current injection. (D) Single and paired-pulse EPSPs (5 Hz) at increasing stimulation intensities (1-3 V) under control conditions and in the presence of. Stimulations above 3 V evoked APs and were therefore not considered. (E) Amplitude and integral of a single EPSP (upper graphs) and of the second EPSP from a 5 Hz paired pulse stimulation as a function of the stimulation intensity. did not modify any of these values (p >.5 for all, n = 7). The integral was calculated as for L5 pyramidal neurons in a 6-ms time window (indicated in (D)). These results show that, in L2/3 pyramidal cells, the 5-HT 7 receptor agonist does not change the firing properties nor excitatory synaptic transmission and EPSP integration. Moreover, resting membrane potential and input resistance are also not modified by the drug (V m : control -8.4 ± 1.5 mv, ± 1.2 mv, p =.26; n = 7; firing current threshold: control ± 5.7 pa, 28.3 ± 44. pa, p =.31), suggesting that 5-HT 7 receptors may not be present in these cells or do not directly influence their electrical properties. (F to H) Reconstruction of biocytin-filled interneurons located in L2/3 (F), L1 (G) and L5 (H) of the ACC and corresponding firing patterns before and after bath application of (3 nm). Dendrites are represented in red and axons in blue. Traces correspond to 6 ms depolarizing current injections of 5, 15 and 3 pa for the different interneurons respectively and a 3 pa hyperpolarizing current step. All recordings were performed in CCI animals. Individual populations of interneurons showed no specific modulation after bath application, therefore averages were pooled over all interneuron subtypes. Cells were first identified based on the absence of a clear main apical dendrite. Subsequently, somatic and dendritic morphology were determined by neurolucida reconstruction. Somata were polygonal, round, triangular and fusiform whereas dendritic morphologies were multipolar or bitufted. The interneurons had several types of firing patterns: continuous and shuttering fast spiking as well as adapting and non-adapting non-fast spiking (Petilla Interneuron Nomenclature et al., 28). (I) AP thresholds were not modified by (p =.61, n = 1) and sag (extremely small or absent in these interneurons) was also unchanged (p =.93, n = 1). (J) Example EPSP responses to synaptic stimulation (intensity 1-3 V). Single or paired-pulse (5 Hz) extracellular stimulation was used. No effect of bath application was observed. The stimulation electrode was placed in the inner L1 (same as for L5 and L2/3 pyramidal cells), or in L2/3 for interneurons of L5. (K) Amplitude and integral of a single EPSP (upper graphs) and of the second EPSP from a 5 Hz paired pulse stimulation (lower graphs) as a function of the stimulation intensity. Extracellular stimulations above 3 V triggered action potentials and were not taken into consideration. No statistical differences were found for the EPSP parameters (p >.5 for all, n = 7). Input resistance and membrane potential were also not modified by (p =.53 and p =.58 respectively, n = 1). Although we cannot exclude that a subpopulation of interneurons might express 5-HT 7 receptors, our results support the idea that 5-HT 7 activation does not substantially modify the electrical and synaptic properties of interneurons in the ACC.

25 CCI Relative threshold 1 mm Before surgery 6 12 Time relative to injection (min) 1. 2 Right Left Inj E Saline Before surgery 6 12 Time relative to injection (min) 1 Right Left Before surgery 6 12 Time relative to injection (min) NS in 5- e C T 2 D CCI Mechanical threshold (g) 6 C CCI Saline AUC (g*h) B Sa l Bilateral motor cortex injections Mechanical threshold (g) A Figure S8 (related to Figure 7). Injection of in the motor cortex does not alter neuropathic pain behavior in nerve-injured animals (A) Location and volume of the injection sites in the primary motor cortex revealed by 3D-ultramicropy reconstruction of the brain. Alexa 488 was co-injected with saline or to estimate the exact location of the injection site. Fluorescence was detected after ex vivo fixation and clearing of the brain. Motor cortex was stereotactically identified and bilateral injections were performed.1 mm posterior to the Bregma, 1.6 mm lateral to the midline and.65 mm below the skull surface corresponding to the mouse hind limb motor cortex (Ayling et al., 29). The coronal section shows the center of the injection volume for all (blue) and saline (white) treatments. (B) Time course of the mechanical threshold of the left and right hind paw in animals subjected to chronic constriction injury (CCI, red bar) of the left sciatic nerve. 7-8 days after ligation the animals showed allodynia-like behavior on the injured side. This behavior was not modified by the acute injection of saline solution in the motor cortex (gray bar) over the time course of 2 hours (n = 5). (C) Injection of the 5-HT7 agonist (3 μm) in the primary motor cortex did not influence mechanical hypersensitivity in CCI animals (n = 4), contrary to its effect on the ACC. (D) Time course of the withdrawal threshold of the left hind paw normalized to the response of the non-injured right hind paw. Blue circles represent treated animals and white circles correspond to the control group (saline). (E) Bar graphs showing the statistical analysis expressed as area under the curve (AUC) between 3 and 12 min after or saline injections in the motor cortex. The two experimental groups were not different (p =.89, unpaired t-test). Error bars: S.E.M.