Running Title: Striatum and Tactile Processing in the Rat Whisker System

47 48 Dorsolateral striatum (DLS) is implicated in tactile perception and receives strong 49 projections from somatosensory cortex. However the sensory representations encoded by 50 striatal projection neurons are not well understood. Here we characterized the 51 contribution of DLS to the encoding of vibrotactile information in rats by assessing 52 striatal responses to precise frequency stimuli delivered to a single vibrissa. We applied 53 stimuli in a frequency range (45-90 Hz) which evokes discriminable percepts and carries 54 most of the power of vibrissa vibration elicited by a range of complex fine textures. Both 55 medium spiny neurons and evoked potentials showed tactile responses that were 56 modulated by slow wave oscillations. Further, medium spiny neuron population 57 responses represented stimulus frequency on par with previously reported behavioural 58 benchmarks. Our results suggest that striatum encodes frequency information of 59 vibrotactile stimuli which is dynamically modulated by ongoing brain state. 60 61 62 Hawking & Gerdjikov 2 Introduction 63 Striatum provides the major input nucleus to the basal ganglia, and as such is implicated 64 in a number of both discrete and overlapping circuits critical to sensory-motor 65 transformation and appetitive behaviour (Alexander et al., 1986; Yin et al., 2004; Bolam 66 et al., 2000; DeLong and Wichmann, 2007). Dorsolateral striatum (DLS) lesions have 67 implicated this region in automatic sensory-driven behaviours (Balleine et al., 2007; Tang 68 et al., 2007; Tang et al., 2009). Some form of perceptual processing has been attributed to 69 striatum. Thus striatal dysfunction plays a role in Parkinson’s disease and this condition 70 produces impairments in somatosensory discrimination. The deficits are associated with 71 impaired caudate activity and the effect is dissociable from motor impairments (Zia et al., 72 2003; Weder et al., 1999). In rats, somatosensory cortical columns corresponding to 73 individual whiskers (the so called barrel cortex) send a strong projection to DLS, forming 74 part of a sensory-motor corticostriatal loop (Alloway et al., 2006; Parent and Hazrati, 75 1995; Wright et al., 2001). DLS responds to pulsatile air puff stimuli applied to the 76 whiskers (Pidoux et al., 2011; Syed et al., 2011). Little is know however about the role of 77 striatum in encoding specific parameters of vibrotactile stimuli in frequency ranges which 78 evoke discriminable percepts in behavioural tasks and carry most of the power of vibrissa 79 vibration elicited by complex fine textures (Hipp et al., 2006; Gerdjikov et al., 2010). In 80 cortex, sensory evoked potentials are strongly dependent on instantaneous UP/DOWN 81 states (Petersen et al., 2003; Haslinger et al., 2006). Striatal UP/DOWN states are likely 82 driven by cortex and/or oscillatory activity in thalamocortical loops, yet it is not known 83 whether slow oscillations modulate tactile representations in this structure (O'Donnell 84 and Grace, 1995; Kasanetz et al., 2008; Tseng et al., 2001; Mahon et al., 2001; Ushimaru 85 Hawking & Gerdjikov 3 et al., 2012). To begin to address these questions, here we recorded striatal responses to 86 precisely defined high frequency stimuli in urethane-anaesthetized rats. We found that 87 populations of medium spiny neurons encode vibrotactile frequency. This encoding was 88 subject to modulation by slow oscillations known to depend on cortical activity 89 suggesting interplay between these structures in sculpting DLS tactile representations. 90 91 Method 92 Animals and surgery 93 Experiments were performed under urethane anesthesia in eight male Wistar rats (Charles 94 River Laboratories Wilmington, MA, USA) weighing between 350-500g. Rats were 95 injected with an initial dose of 1.25 mg/kg urethane which was supplemented with a second 96 0.25 mg/kg injection 20 min later. Pinch and corneal reflexes were monitored throughout the 97 experiment and the rat received 0.15 mg/kg top-ups if necessary to maintain anaesthesia. This 98 dosing regimen is similar to what was used previously to assess tactile encoding in rat barrel 99 cortex (Arabzadeh et al., 2006; Hirata and Castro-Alamancos, 2011). A similar regimen has also 100 been used to study cortical contributions to UP and DOWN states in medium spiny neurons 101 (Kasanetz et al., 2006). Thus the current approach provides results which are directly 102 comparable with these studies. Body temperature was monitored rectally and maintained at 103 37 Co using a homeothermic pad (Harvard Apparatus, Boston, MA, USA). For fluid 104 replacement, 5% glucose was continuously administered via an infusion pump (3 105 mL/hour, s.c.; Instech, K. D. Scientific, Holliston, MA, USA). Glycopyrronium bromide 106 (40μL/kg, i.m.; Anpharm, Warsaw, Poland) was given to reduce respiratory tract 107 secretions. Animals were fixed to a stereotaxic frame and the head was adjusted so that 108 lambda and bregma were on the same horizontal plane. A left-side craniotomy was 109 Hawking & Gerdjikov 4 performed to provide access to DLS, leaving dura mater intact. To prevent corneal 110 desiccation Lacri-Lube Eye Ointment (Allergan, Wesport, Ireland) was applied to the 111 eyes. Surgeries were carried out under institutional ethics approval and a project license 112 granted by the UK Home Office under the Animals (Scientific Procedures) Act 1986. 113 114 Tactile stimulation 115 The whisker stimulator was constructed from a glass capillary (1 mm o.d.) glued to a 116 piezo bender (Physik Instrumente, Karlsruhe, Germany). The tip of the capillary was 117 further thinned through heating until a whisker hair could rest snugly inside the tip 118 opening. Voltage commands were programmed in Matlab (Mathworks, Natick, MA, 119 USA) and delivered using custom-written LabVIEW software (National Instruments, 120 Austin, TX, USA). The stimuli consisted of brief pulsatile deflections (single-period 121 cosine wave, 100 Hz, duration 10 ms) presented to the right C1 whisker for 1 sec at inter122 pulse intervals of 22, 17, 13 and 11 ms corresponding to frequencies of 45, 60, 75 and 90 123 Hz (amplitude 12°, i.e. 5 mm distance from the whisker base) (Figure 1a,b). The length 124 of the glass capillary and point of attachment of the piezo element were optimized to 125 remove ringing of the stimulator. Calibration with a phototransistor (HLC1395, 126 Honeywell, Morristown, NJ, US) showed that differences in amplitude and peak velocity 127 between frequencies were smaller than three percent. The capillary tip was positioned 128 5 mm away from the skin and tilted at an angle of 155° to 175° against the whisker such 129 that the vibrissa rested against the inside wall of the capillary, ensuring that the stimulator 130 engaged the whisker immediately. 131 132 Hawking & Gerdjikov 5 Electrophysiological recordings and analysis 133 Wideband signals were acquired continuously via an op-amp based headstage amplifier 134 (HST/8o50-G1-GR, 1x gain, Plexon Inc., Dallas, TX, USA), passed through a 135 preamplifier (PBX2/16wb, 1000x gain; Plexon Inc., Dallas, TX, USA) and digitized at 40 136 kHz. Recording electrodes consisted of quartz glass-coated platinum/tungsten wires 137 pulled and ground to custom shapes in our laboratory (shank diameter 80 μm; diameter of 138 the metal core 23 μm; free tip length ~8 μm; impedance, 1-3MΩ; Thomas Recording, 139 Giessen, Germany). DLS recording electrodes pierced dura and were advanced into DLS 140 using established coordinates and covered DLS areas receiving extensive barrel cortex 141 input and known to respond to airpuffs delivered to the whisker (Hoffer and Alloway, 142 2001; Syed et al., 2011). DLS penetrations were located AP -0.8 to -1.5mm, ML 2 to 143 5mm, DV 4 to 7mm (Figure 1c). All data processing was done offline. For spike sorting 144 the raw signal was band-pass filtered 300-3,000Hz and spikes were sorted using the 145 Matlab-based Wave_Clus software to yield single-unit spike trains (Quiroga, Nadasdy, & 146 Ben-Shaul, 2004). Wave_clus performs unsupervised spike detection and sorting using 147 wavelets and super-paramagnetic clustering. All automatic detection thresholds and 148 sorting solutions were examined individually and adjusted if needed. Field potentials 149 were recorded from the same electrode and were downsampled to 5,000 Hz and evoked 150 responses extracted from the raw data using a 200 Hz low-pass Butterworth filter. 151 Responses were averaged over 20 trials. Single unit tactile responses were calculated by 152 adding spikes in a 500 msec window after stimulus onset and subtracting the baseline 153 firing rate calculated over 100 msec. To investigate the effect of DLS slow oscillations on 154 tactile responses, LFPs were low-pass filtered < 5Hz using a second order Butterworth 155 Hawking & Gerdjikov 6 filter and a Hilbert transform was applied to obtain the instantaneous phase angle (Saleem 156 et al., 2010). Mean phase angle and resultant vector length were calculated using the 157 Matlab Circular Statistics Toolbox (Behrens, 2010; Fisher, 1993). Analyzes were 158 calculated using Neuroexplorer (Nex Technologies, Littleton, MA, USA) and custom159 written Matlab routines. Analyses of variance (ANOVA) were performed using SPSS 160 (IBM SPSS, Somers, NY, USA). 161 Results 162 DLS tactile evoked potentials are modulated by slow oscillations 163 Vibrotactile stimuli elicited evoked field potential responses in DLS (Figure 2a). Studies 164 done in cortex show a strong relationship between the phase of low frequency LFP of 165 deep cortical layers and intracellularly recorded neuronal UP/DOWN states (Saleem et 166 al., 2010). We low-pass filtered field potential traces < 5Hz and used a Hilbert 167 transformation (Figure 2b,c) to derive the instantaneous phase of spontaneous striatal 168 LFP. Instantaneous LFP phase at stimulus onset calculated in this manner was evenly 169 distributed across the 0-360 deg waveform cycle as would be expected with random 170 stimulus presentations (mean resultant vector length of the phases at onset = 0.06, p = 171 0.3). We next calculated the slope of the evoked potential curve between 0-200 ms after 172 stimulus onset and plotted it against the phase of the slow oscillations assessed at 173 stimulus onset on a trial by trial basis (compare (Wyble et al., 2000)). The magnitude of 174 the tactile response showed a clear modulation by oscillation phase (Figure 1d). The 175 response was strongest during the 0-180o portion of the slow wave oscillation 176 corresponding to a DOWN state and it was virtually absent at 270o corresponding to the 177 Hawking & Gerdjikov 7 UP state oscillation. Evoked potentials however did not appear to be modulated by 178 stimulus frequency, even when controlling for slow oscillation phase. 179 180 Populations of MSNs encode vibrotactile stimulus frequency during DOWN/UP 181 transitions 182 Medium spiny projection neurons (MSNs) represent more than 90% of rat striatal 183 neurons and unlike DLS interneurons are characterized by a relatively low firing rate 184 (Rymar et al., 2004; Berke et al., 2004). To ensure only this cell type entered our data set, 185 we recorded units with low baseline activity (< 5Hz) and the firing rates we observed 186 (1.37 spikes/s ± 1.25, mean ± s.d.) are consistent with previous studies (Mowery et al., 187 2011; Barnes et al., 2005; Schmitzer-Torbert and Redish, 2008; Sharott et al., 2009; 188 Berke et al., 2004). Further we noted that the spike waveforms of the recorded units 189 showed peak to valley intervals (mean ± SEM) of 735 ±57 μsec and peak widths of 372 ± 190 15 μsec, which are in agreement with previously reported MSN extracellular waveform 191 characteristics (Wiltschko et al., 2010). We recorded from 35 neurons. Consistent with 192 previous work showing subthreshold responses to tactile stimuli in intracellular 193 recordings but a weak spike response when single pulses were used (Pidoux et al., 2011), 194 here with the delivery of high frequency stimuli only a small percentage of cells 195 responded to a 45Hz stimulus. Seven out of the 35 recorded units (20%) showed a 196 significant change in firing rate during the 500 ms after stimulus onset (p < .05; 197 Wilcoxon signed-rank test). In all cases the response was a small increase in firing (.53 198 spikes/s ± .18). When stimulating at 60Hz, we obtained 23% responsive units (.52 199 spikes/s ± .14), at 75Hz 23% (.54 spikes/s ± .49) and at 90Hz 11% (.70 spikes/s ±.34). 200 Hawking & Gerdjikov 8 Thus tactile responses of individual MSNs to stimuli at different frequencies were sparse, 201 similar to responses to pulses (Pidoux et al., 2011). To investigate a possible contribution 202 of UP/DOWN state transitions, instantaneous LFP phase was used to construct phase 203 histograms of MSN spike times (Figure 3a,b) (Saleem et al., 2010). Spontaneous MSN 204 firing was highest during the negative component of the LFP wave consistent with recent 205 combined intracellular and LFP recordings in cortex showing a mean phase distribution 206 of the low pass-filtered LFP trace of 200-225° during UP states (Figure 3c) (Saleem et 207 al., 2010). Thus our DLS results are consistent with the previously observed phase 208 relationship between cortical membrane states and slow oscillation phase. 209 210 We next tested if slow oscillation phase affects DLS tactile responses. To do this, we 211 extracted the instantaneous phase of striatal LFP at the onset of each stimulus. Within 212 each unit (i.e., individual trials serving as replications), phase did not appear to modulate 213 the relationship between stimulus frequency and firing rate (ps > .10 for the phase x 214 stimulus interaction term in individual unit multiple regressions). However in the 215 population (i.e., individual cell averages serving replications), the tactile response was 216 related to LFP phase at stimulus onset (Figure 4). This was supported by a phase (4 217 quadrants) x stimulus (4 frequencies) repeated measures ANOVA [main effect of phase, 218 F(3, 102) = 10.31, p < .01; interaction, F(9, 306) = 1.86, p = .056]. We tested different 219 amplitude (2 to 12°) x frequency (45-90Hz) combinations on a subset of cells (n = 17). 220 This did not reveal any amplitude effects, but confirmed the phase x frequency 221 interaction [F(9, 27) = 2.34, p = .017]. The interaction was followed up with simple one222 way ANOVAs at each phase interval, which showed a clear frequency response for 223 Hawking & Gerdjikov 9 stimuli arriving between instantaneous phases of 90 and 180o [F(3, 102) = 3.24, p =.025]. 224 In previous work, cortical DOWN states showed a mean phase distribution of 290-320° 225 (compared to 110-135° during UP states; here 0° corresponds to the positive peak in the 226 slow oscillation, cf. Figure 1c) thus the 90-180o quadrant roughly corresponds to the 227 transition from Down and Up states. Specifically firing rate was significantly higher for 228 45Hz compared to 75 and 90 Hz and for 60 Hz compared to 90 Hz (pair-wise 229 comparisons, p < .05). The effect of frequency was not significant in other phase 230 intervals. We conclude that tactile responding is sparse and shows no stimulus frequency 231 modulation at the level of individual MSNs. However firing rate was related to stimulus 232 frequency in the population and this effect was dynamically modulated by slow 233 oscillation phase: vibrotactile frequency was encoded during the transition from DOWN 234 to UP states. 235 236 Spike-phase coding refers to the encoding of information in the relative timing of 237 neuronal activity to slow background rhythms (Kayser et al., 2009). To assess whether 238 the timing of spikes relative to the ongoing slow oscillation is related to stimulation 239 frequency, we calculated the average LFP slow oscillation phase of all recorded neurons 240 in a 2 sec window prior to tactile stimulation as well as during the 1 sec of tactile 241 stimulation separately for each of the 4 stimulation frequencies. However a 4 x 2 242 (frequency x window) within-subjects ANOVA across neurons showed no effect of 243 frequency or analysis window (ps > .5). Thus we found no evidence for spike-phase 244 coding of frequency information in this system. 245