Thalamic state influences timing and feature selectivity in the thalamocortical circuit

Thalamic neurons dynamically encode sensory information in a state-dependent manner as a mechanism for gating information flow to the cortex. Here, we investigated the role of thalamic state on precise feature selectivity in the thalamocortical circuit of the rat vibrissa pathway. In thalamic neurons, tonic spike triggered averaging revealed clear feature selectivity, while the feature selectivity associated with burst spikes could not be recovered with this approach. These thalamic state dependent changes propagated to cortex such that the cortical feature selectivity was diminished during the optogenetically hyperpolarized (burst biased) thalamic condition. Further analysis revealed that the perceived loss of feature selectivity was likely not due to a true loss of stimulus selectivity but instead to changes in the precision of the temporal spiking in burst firing modes. Therefore, alterations to thalamic state enable a dynamic interplay between spike timing and spike rate that shapes stimulus encoding in the thalamocortical circuit.


Introduction 14
Sensory thalamus plays a critical role in gating information flow from our sensors in 15 the periphery to sensory cortex, ultimately shaping how we perceive the world. Importantly, 16 thalamic gating properties are not static, but instead vary dynamically through a range of 17 modulatory mechanisms, including local membrane and synaptic properties (Wolfart et al., 18 2005), stimulus history , and neuromodulatory inputs from brainstem 19 and cortex (Castro-alamancos, 2002; Mease et al., 2014). Although arising from different 20 mechanisms, these modulatory inputs have the net effect of altering the baseline membrane 21 polarization level in the thalamus, which we refer to here as "thalamic state", which plays an 22 important role in determining the encoding properties of the thalamic neurons that serve as 23 primary inputs to sensory cortex. Perhaps most prominently, modulation of the baseline 24 membrane potential in thalamic neurons enables distinct tonic and burst firing modes due to 25 the selective engagement of low threshold calcium channels during prolonged 26 hyperpolarization (Suzuki and Rogawski, 1989). In addition to their roles in thalamocortical 27 oscillations (Steriade et al., 1993), it has long been posited that these two firing modes could 28 be a mechanism for dynamically controlling information processing (Sherman 2001). At the 29 thalamocortical synapse, spontaneous burst spikes are more effective at driving cortical 30 spiking (Swadlow and Gusev, 2001) and evoke larger cortical depolarizations (Bruno and 31 Sakmann, 2006) than tonic spikes. It has been proposed that both burst and tonic spikes carry 32 stimulus information (Reinagel et al., 1999), but the relationship between burst and tonic firing 33 in representing temporal stimulus information in tactile encoding remains unclear. 34 In the visual pathway, the role of thalamic state in sensory processing has identified 35 distinct sensory selectivity associated with burst versus tonic firing (Alitto et  of the bursts are characterized by a prolonged inhibitory stimulus before the depolarizing input 40 that occurs immediately prior to the spike onset. This demonstrated that bursting activity is not 41 Optogenetics surgeries: All surgical procedures followed sterile protocol. A small craniotomy 94 was made above VPm (3 mm lateral, 3 mm caudal to bregma). A 10 µL syringe (Neuros 95 Syringe, Hamilton, Inc) filled with the virus (rAAV5-CamKIIa-Jaws-KGC-GFP-ER2 or rAAV5-96 CamKIIa-eNpHR3.0-EYFP, UNC Viral Vector Core Services) was lowered to depth of 5.2 mm 97 before injecting 1 µL of virus at a rate of 0.2 µL/min (iSi system, Stoelting). The syringe 98 remained in place for five minutes after the injection was complete to allow the virus to diffuse. 99 Opsin expression was fully realized at 2-3 weeks post-surgery. and Jaws, respectively). The whisker was then stimulated without (baseline) and with 106 (hyperpolarized) light provided directly to the thalamus (50 mW/mm 2 ). Optogenetic stimulus 107 conditions (light on/hyperpolarized, light off/baseline) were interleaved to avoid long-term 108 adaptation effects. 109 Analytical Methods: Spike sorting for single channel recordings was performed online and 110 validated offline using Waveclus (Quiroga et al., 2004). Spike sorting for multichannel 111 electrodes was performed offline using the KlustaKwik software suite (Rossant et al., 2015). 112 Isolation of the unit was confirmed by the waveform amplitude (absolute and relative to the 113 background noise >3) and the interspike-interval distributions (VPm: mean of 0.22%, S1: 114 mean of 0.38% of spikes in absolute refractory period of 1ms). 115 Feature selectivity was estimated for each recorded unit using the spike triggered average 116 (STA) (Schwartz et al., 2006). 117

118
Where N is the number of spikes and is the stimulus segment in a window surrounding each 119 spike (-30 to +5 ms, spike-triggered ensemble, STE). The burst and tonic triggered averages 120 were computed from burst and tonic spikes, respectively. The baseline/hyperpolarized 121 condition triggered averages were computed from all spikes in a given stimulus condition. The 122 bootstrap estimate of the confidence intervals on the spike triggered average was computed 123 as the +/-2 standard deviation of this shuffled STA distribution across 500 repetitions 124 (Schwartz et al., 2006). Note that we implemented multiple techniques of estimating the 125 feature selectivity of the neurons including spike triggered covariance, generalized linear 126 models, and nonlinear-input models (McFarland et al., 2013). The results were qualitatively 127 consistent across all methods employed, so we chose to use spike triggered average 128 throughout the manuscript due to its simplicity. 129 The signal-to-noise ratio of the recovered STA was quantified as the peak-to-peak amplitude 130 of the STA within 10 milliseconds of the spike (where the significant filter activity is contained) 131 divided by the peak-to-peak amplitude of the STA from 30 to 20 milliseconds before the spike 132 (where there is no expected filter information). An SNR value of 1 means the amplitude of the 133 STA near the spike time is not different from the amplitude of the noise fluctuations. Therefore, 134 any units with an SNR value less than 2 were excluded from further analysis. 135 To make comparisons of the feature selectivity across the population of recorded neurons, we 136 computed a principle component analysis of the recovered STA (Estebanez et al., 2012). The 137 first two principle components accounted for the majority of the variance (71.8% VPm, 78.4% 138 S1). 139 The non-linearity ( | ) was estimated as the ratio of the probability of spike-trigged Where y is defined as the stimulus (s) convolved with the feature selectivity of the unit (STA) 144 (Lesica et al., 2007), referred to as filtered stimulus. For all conditions, the STA was defined 145 as the baseline or tonic spike triggered average. Throughout the manuscript, we separate the 146 firing rate (p(spike)) from the shape of the non-linearity (p(y|spike)/p(y)) to avoid confounding 147 differences in firing rate with differences in tuning. 148 The precision in the noise evoked firing was estimated for each spike classification (tonic, 149 burst, baseline, hyperpolarized). The precision was defined as: is defined for each spike as the temporal lag (tlag) for the peak correlation 153 between the STA and the stimulus segment surrounding that spike (-30 ms to +5 ms). 154

, 155
The distribution was normalized by the total number of spikes in each condition (tonic, 156 burst, baseline, hyperpolarized). 157 All pairwise statistical comparisons were computed using a Wilcoxon signed rank test unless 158 otherwise noted. 159

Results 160
Feature selectivity is conserved across the thalamus and cortex 161 We recorded thalamic and cortical extracellular spiking activity in response to sensory 162 white noise stimulation of a single whisker in the vibrissa pathway of the fentanyl-anesthetized 163 rat ( Figure 1A, see Methods). We estimated the feature selectivity for each unit as the spike 164 triggered average (STA), which captures the features of the sensory stimulus that tended to 165 precede spiking, and the static, point nonlinearity, which captures the translation into 166 suprathreshold spiking activity ( Figure 1B; see methods). Although this quantification was 167 performed on longer unique noise segments, we also recorded the response to short (4-10 168 second) frozen white noise segments to examine the response across trials. Figure 1C shows bounds on the shuffled process. This suggests that, on average, the thalamic unit is only 181 sensitive to the stimulus occurring in the previous 10-15 milliseconds. The S1 unit also 182 displays feature selectivity as evidenced by the shape and amplitude of the S1 STA 183 immediately prior to the cortical spike relative to the shuffled case. Although the VPm STA is 184 nearly ten times as large in amplitude as the S1 STA, the similarity in the temporal dynamics 185 can be visualized by shifting the VPm STA by 2 milliseconds relative to the S1 STA ( Figure  186 1C, bottom, S1 STA black, VPm STA shifted by 2 milliseconds and scaled by a factor of 0.1 187 as grey dashed line). 188 While this simple comparison provides an interesting observation for a single pair of 189 topographically aligned neurons, we also made comparisons of the feature selectivity across 190 Inherent in the spike triggered analysis, however, is an assumption that the average 208 filter is representative of the sensory stimulus preceding all spikes (Stanley, 2002). Yet 209 neurons in the thalamus are well known for exhibiting two fundamentally different types of 210 firing: tonic spiking and burst firing mediated through T-type calcium channels (Suzuki and 211 Rogawski, 1989). Burst spikes were classified here from the extracellular recordings as two 212 or more spikes with an inter-spike interval of less than four milliseconds with the first spike in F. Burst and tonic firing rate (p(spike)) across recorded population (p = 2.6e-5). G. Spike triggered average using different spike classifications.
In the thalamic recordings, tonic and burst spikes were interspersed throughout most 217 of the recordings. For the example thalamic unit presented in Figure 1B, we computed the 218 spike triggered average from all spikes (STA), the tonic spike triggered average from only tonic 219 spikes (tSTA), and the burst spike triggered average from only spikes that are classified as 220 being part of a burst (bSTA) ( Figure 2B). The tSTA (grey) closely resembles the STA 221 computed from all spikes (black) while the bSTA (red) is significantly degraded as evidenced 222 by the flat shape of the filter. To compare the difference between burst and tonic feature 223 selectivity across thalamic units, we quantified the signal-to-noise ratio of the STA (STASNR, 224 see methods). Across all thalamic units, the SNRSTA was higher for tonic spikes (tSTASNR) than 225 for burst spikes (bSTASNR) ( Figure 2C). 226 Given the estimated feature selectivity, we can compute the static non-linearity, or the 227 input-output function, which provides a mapping between this filtered stimulus (y) and the 228 spiking response of the neuron (p(spike|y)) by taking the ratio of the p(y|spike) to the p(y) 229 ( Figure 1B, see methods). Here, we used the tSTA as the filter for all spiking conditions when 230 estimating the non-linearity. The probability of the filtered stimulus (p(y) remains unchanged 231 when the filter is held constant. Therefore, any change in the non-linearity is then only due to 232 changes in the probability of the filtered stimulus given that a spike occurred (p(y|spike)), or 233 the spike triggered ensemble. Because the slope of the static non-linearity is determined by 234 the separation between the spike triggered ensemble and the Gaussian distributed white 235 noise, as the spike triggered ensemble distribution becomes more selective (i.e. the mean 236 moves away from the filtered stimulus distribution), the separability of the distributions 237 increases, and the slope of the non-linearity also increases. Intuitively, this means that the 238 shape of the non-linearity gives an estimate of the separability of the spike triggered ensemble 239 and the stimulus distribution, or how strongly tuned a neuron is for that particular feature, given 240 by the STA. A steeper slope in the non-linearity suggests a stronger tuning than a shallower 241 slope. Therefore, we also assessed the spiking nonlinearity as a function of the spike 242 classification. In this example unit, we found that the tonic spikes were well tuned to the STA, 243 as evidenced by the steep slope of the non-linearity while the burst spikes were not well tuned 244 to the STA, as evidenced by the relatively flat non-linearity ( Figure 2D). This trend was 245 consistent across units where the burst spikes showed reduced tuning to the STA as 246 compared to tonic spikes as assessed by the slope of the spiking nonlinearity ( Figure 2E). 247 Here, we have separated the difference in the slope of the non-linearity from the difference in 248 the prevalence of burst and tonic spikes (p(spike)), which is markedly higher for tonic spikes 249 than for burst spikes ( Figure 2F). Furthermore, we tested alternative burst spike classifications 250 and quantified the implication for the STA (Figure 2G, top). Across spiking classifications, 251 increased periods of silence prior to the spike (tsilence) led to decreased STASNR while bursts of 252 spikes (tisi<4 or <10) had consistently lower STASNR relative to tonic spikes (tisi>20) ( Figure 2G, 253 example unit in middle, population data in bottom). Therefore, our data do not provide 254 evidence to support a difference in feature selectivity for tonic and burst spiking in this 255 pathway, but instead suggests a reduction in stimulus selectivity in burst spiking within this 256 analytic framework. 257 258 Thalamic state dependent feature selectivity 259 The previous analysis was conducted by presenting sensory white-noise stimuli and 260 parsing measured thalamic spiking activity into tonic and burst classes, while these classes of 261 spiking were intermingled throughout the recordings. However, the thalamus was in tonic firing 262 mode, with relatively low burst firing rates ( Figure 2F). Here, we used optogenetic 263 hyperpolarization of the thalamic neurons not to silence the thalamic neurons, but instead to 264 shift the thalamus into a burst firing mode during sensory white noise stimulation ( Figure 3A). 265 Using this optogenetic manipulation, we asked whether the optogenetically manipulated firing 266 mode (baseline and hyperpolarized conditions) of the thalamus impacts feature selectivity and 267 how this relates to the classified burst/tonic modes. 268 Here, we recorded the thalamic response to sensory white noise with and without the 269 the spiking response to a frozen white noise segment without optogenetic stimulation ( Figure  273 3B, baseline condition) and with optogenetic stimulation ( Figure 3B, hyperpolarized condition). 274 We have pseudocolored the tonic spikes grey and the burst spikes red to qualitatively visualize 275 the thalamic firing mode ( Figure 3B). In the baseline condition, the response is primarily tonic 276 as evidenced by the grey raster plots ( Figure 3B, Baseline, BR = 0.10). In the hyperpolarized 277 condition (optically stimulated), the firing mode is biased towards a burst encoding scheme, 278 as evidenced by the prevalence of red burst spikes ( Figure 3B, Hyperpolarized, BR = 0.67). 279 The tonic STA showed pronounced feature selectivity for this unit while the burst STA did not 280 The similarity between the burst spike response and hyperpolarized condition can also 286 be seen in this example nonlinearity where the burst and hyperpolarized nonlinearities are 287 effectively flat while the tonic spikes and baseline condition show obvious tuning ( Figure 3D). 288 Across units, we found an overall reduction in the STASNR for the hyperpolarized condition 289 relative to the baseline condition ( Figure 3E, p = 0.037). We also found that the tuning of the 290 nonlinearity was lower for the hyperpolarized condition relative to the baseline condition as 291 reflected in the overall gain/slope ( Figure 3F). Importantly, the baseline and hyperpolarized 292 conditions both contain burst and tonic spikes. Instead of completely separating the firing 293 modes into all burst spikes or all tonic spikes, we have optogenetically altered the spiking 294 probabilities such that the baseline condition has more tonic spikes and the hyperpolarized 295 condition has more burst spikes while maintaining similar numbers of spikes ( Figure 3G). The 296 similarities between the STA and the NL properties of the burst and hyperpolarized state as 297 well as the tonic and baseline state suggest that there was no discernable difference for the 298 estimation of feature selectivity when assessed based on the state of the thalamus at the time 299 of the stimulus (hyperpolarized/baseline) versus the spike type classification (burst/tonic). 300 301 Temporal precision of thalamic firing modes 302 Given the difference between the recovered estimates of burst/hyperpolarized and 303 tonic/baseline feature selectivity, we implemented a series of computational controls to identify 304 any potential shortcomings of the methodologies that could underlie these results. The first 305 issue we considered was whether or not the burst spike feature selectivity was unrecoverable 306 due to the effect of subsequent spikes in the burst. If the timing of spikes within a burst is not 307 repeatable and structured, the presence of these additional spikes will serve to destroy the 308 temporal structure in the feature selectivity as revealed by the spike triggered analysis. When 309 the bSTA was computed from only the first spike in each burst ( Figure 2B, red-dashed line), 310 there was no apparent feature selectivity for this example unit. This can also be visualized 311 across units in the STASNR where the bSTASNR is plotted when computed from all burst spikes 312 (black dot) and when computed from the first spike in each burst (red circle, Figure 2C). 313 Therefore, including all spikes in a burst (or not) does not strongly impact the ability to estimate 314 the feature selectivity from the STA. 315 The second issue we considered was the overall difference in spike rates. Spike 316 triggered analyses require a large number of spikes to effectively estimate the underlying 317 selectivity. The proportion of spikes classified as bursts was lower than the spikes classified 318 as tonic ( Figure 2F) as quantified by the burst and tonic firing rate. Therefore, it was possible 319 that we could not recover an STA for the burst spike condition due to the reduced number of 320 burst spikes relative to tonic spikes. In an example unit, we computed the tSTA using only a 321 subset of the spikes (n = 2362 of 36558 spikes corresponding to n = 2363 bursts with n = 322 7547 burst spikes) and found that the linear filter was essentially identical to the tSTA (Figure  323 2B, grey dashed line). We computed this for all thalamic units and again found that the burst-324 count matched tSTA was also significantly larger than the bSTA. Furthermore, there was no statistically significant difference in the firing rate between the baseline and hyperpolarized 326 optogenetic conditions ( Figure 3G), but still the difference in the STA persisted. This suggests 327 that simple spike counts alone were insufficient to explain the difference in the tonic/baseline 328 STA and the burst/hyperpolarized STA. 329 The third issue we considered was the inherent assumption that the feature selectivity 330 for each unit could be recovered as the STA. It was possible that the burst STA was not 331 recoverable because the burst firing mode was better estimated by a symmetric nonlinearity 332 and therefore the filter could only be recovered using spike triggered covariance (STC) 333 techniques. We therefore computed the STC for all recorded thalamic units and compared this 334 for each spiking condition. Although the dataset was more limited because the number of units 335 with a significant STC filter was lower than those with a significant STA filter (n = 13 units with 336 STC filter compared to n = 30 units with STA filter), the same trends regarding the reduction 337 in the amplitude of the filter (STCSNR) and the slope of the symmetric nonlinearity persisted 338 (data not shown). Therefore, this suggests that the method of extracting the feature selectivity 339 (STA compared to STC) was insufficient to explain the inability to estimate the feature 340 selectivity in the hyperpolarized/burst spiking conditions. 341 The fourth assumption made throughout the analysis was that burst spikes are actually 342 driven by sensory stimuli such that there is a recoverable burst spike feature selectivity. The 343 alternative explanation would be that burst spikes are not feature selective and instead occur 344 randomly due to intrinsic or other non-sensory processes. To assess this, we quantified the 345 trial-to-trial repeatability for bursts in response to frozen white noise segments. As can be seen 346 in Figure 3B, the qualitative assessment of temporally aligned bursts in response to the frozen 347 white noise segment suggests that the bursts are driven by the sensory stimulus in a 348 repeatable way. For units with a sufficient number of repeated trials, we computed the 349 reliability of the burst spiking as the correlation between the peristimulus time histogram of 350 even and odd trials in response to the frozen white noise segment. We found that all units 351 showed greater reliability than what is expected based on just the temporal correlations in the 352 burst spiking (shuffle control, p = 0.002, n = 10 thalamic units). This suggests that the bursts 353 are not randomly generated or due entirely to a non-stimulus related phenomenon. 354 From these controls, we propose that the difference in the spike triggered encoding 355 properties could not be attributed to differences in the overall spike rates, the temporal 356 properties of the spikes within the burst, or the mechanism of filter estimation. Instead, we 357 propose that the burst spikes are driven by the sensory stimulus and have an underlying 358 feature selectivity, but that this cannot be recovered using spike triggered techniques due to 359 the reduced temporal precision of burst spiking relative to tonic spikes. 360 Recovering an STA relies on precise temporal spiking relative to the sensory stimulus. 361 To simulate degradation of the spike timing precision, we added independent samples of 362 normally distributed temporal jitter of varying amplitudes (standard deviation of the jitter 363 distribution) to each tonic spike for an example unit and computed the STA ( Figure 4A). 364 Across units, we quantified the degradation of the STA as the jittered-STASNR normalized by 365 the tSTASNR (0 ms jitter). The jittered-STASNR (black) is within the band expected for the 366 bSTASNR with the addition of 4 milliseconds of jitter to the spike times (red shaded, Figure 4A, 367 right). We propose that the effects of temporal jitter are particularly evident for whisker 368 selectivity, presumably due to the short temporal duration of the filters (approximately 10-15 369 milliseconds in duration, Figure 1F). 370 Given the marked effects of jitter on the ability to recover the STA, we investigated the 371 variability in the spike timing relative to the noise stimulus ( Figure 4B spiking, the raster plots would all be perfectly aligned to t0 because the similarity between the 377 stimulus and the STA would predict a spiking response at that time point. However, the timing 378 of evoked neural responses is always variable to some extent and this can be visualized for 379 this example response segment as the temporal variability of the spike times surrounding this 380 stimulus feature in the noise stimulus ( Figure 4B, as indicated by the grey stimulus bars that 381 extend from the first spike response to this particular sensory feature). For this example 382 snapshot, it is also apparent that the burst spikes in the hyperpolarized condition show greater 383 temporal variability than the tonic spikes in the baseline condition. 384 To quantify this jitter across all spikes, we developed a τjitter metric that determines the 385 time lag of the peak correlation between the STA and the stimulus segment (s(tj)) surrounding 386 each spike ( Figure 4C). Intuitively, this is a correlative method to identify the time lag between 387 when we predict a spike is most likely to occur based on the STA and the stimulus (peak 388 correlation) and when the spike actually occurred. For this analysis, we treated the tSTA as 389 the true feature selectivity of the neuron across all spiking conditions because we could not 390 recover a reliable estimate of the bSTA. 391 We computed τjitter for each spike and plotted τjitter distributions for each spike condition 392 (tonic, burst, baseline, hyperpolarized). If a neuron was infinitely precise such that when the 393 stimulus matched the spike triggered average, the neuron fired a spike without delay, this 394 distribution would be represented by a delta function at τjitter equals zero. As the variability of 395 the timing increases, the width of this distribution will also increase. For the tonic and baseline 396 condition spikes, we found a clear peak in τjitter values at τjitter equals zero ( Figure 4C, grey, 397 black). For the burst and hyperpolarized condition spikes, we observe little-to-no peak in the 398 τjitter metric at zero ( Figure 4C, red, yellow). We computed the τjitter distribution across all 399 thalamic units and found that the tonic and baseline spikes had higher peaks at τjitter equals 400 zero than the burst and hyperpolarized spike conditions ( Figure 4D). We quantified this 401 statistically by computing a precision metric ( Figure 4C) that computes the proportion of spikes 402 within ±1 millisecond of τjitter equals zero ( Figure 4E). The tonic and baseline spike condition 403 were more precise than burst and hyperpolarized conditions. 404 These data suggest that tonic spikes showed greater temporal precision in response 406 to the sensory white noise than burst spikes and that this could underlie the difference in the 407 recoverability of the feature selectivity in the thalamus between firing modes. It is well 408 established that the timing of sensory inputs is particularly important in the thalamocortical 409 circuit such that changes in thalamic spike timing could have large impacts on the downstream 410 representation of sensory information in the cortex. Next, we investigated how these changes 411 in temporal precision in optogenetically modulated thalamic states impact cortical encoding 412 properties. 413

414
Optogenetic modulation of thalamic firing modes directly impacts cortical representation of 415 sensory information. 416 Cortical neurons that receive direct thalamic input are integrating information over a 417 population of thalamocortical neurons that can be operating in different firing modes. This 418 makes it difficult to determine the impact of a single burst from a single neuron on information 419 representation in the pathway. Instead, we used the optogenetic manipulation of thalamic state 420 as presented in Figure 3 to bias the activity of the thalamic population towards burst firing 421 (hyperpolarized condition) and investigated the effects on the cortex. Here, we transfected the 422 thalamus with a hyperpolarizing opsin and lowered an optic fiber into the thalamus while 423 recording the cortical activity extracellularly ( Figure 5A) in response to sensory white noise 424 with and without optogenetic manipulation of the thalamus (hyperpolarized VPm, baseline 425 VPm). 426 For an example unit, we have plotted the cortical STA in the baseline and 427 hyperpolarized VPm conditions ( Figure 5B). Here, the amplitude of the cortical STA was 428 smaller when the thalamus is hyperpolarized compared to when it is not ( Figure 5B). This 429 cortical unit also shows a reduced tuning to the STA when the thalamus was hyperpolarized 430 ( Figure 5C). Across the population of recorded cortical neurons, we saw the same effect seen 431 in this example neuron of a reduced STASNR when the VPm was hyperpolarized compared to 432 when it was not ( Figure 5D) and a reduction in the tuning across all cortical units as quantified 433 by the spiking nonlinearity ( Figure 5E). These findings mirror what was seen for thalamic 434 neurons when comparing the baseline and the optogenetically manipulated conditions 435 demonstrating that the changes in thalamic encoding properties are propagated to cortex. 436 Interestingly, there was no significant difference in the noise-evoked firing rate in the 437 cortex as a function of the VPm condition ( Figure 5F). This suggests that it was not overall 438 spike counts influencing the cortical feature selectivity. Instead, we propose the temporal jitter 439 in the thalamic spiking patterns propagated to cortex. We investigated the temporal precision 440 of the cortical spiking in response to the sensory white noise using the same methodology 441 employed for the thalamus. As we saw for the thalamus, the cortical spikes from this example 442 unit also showed greater temporal precision in the baseline VPm condition compared to the 443 hyperpolarized VPm condition ( Figure 5G) as evidenced by the peak in the τjitter distribution 444 around τjitter equals zero. This effect was consistent across the population of recorded cortical 445 units ( Figure 5H) and showed significant differences in the precision of the cortical firing 446 ( Figure 5I). This suggests that the temporal jitter present in the thalamus is transmitted to the 447 cortex where it also impacts the representation of sensory information. 448

449
We have primarily considered thalamic state-dependent encoding as a feedforward 504 representation from thalamus to cortex, but the highly interconnected thalamocortical circuitry 505 is a dynamic interaction that shapes coding properties in both feedforward and feedback 506 manner. Changes in thalamic activity impact cortical activity which then provides feedback to 507 thalamus to further alter activity (Crandall et  to shift the temporal precision of the thalamic spike timing through changes to the thalamic 513 state, or the baseline membrane potential, provides a biophysical mechanism for the thalamus 514 to gate information flow to cortex. Furthermore, this mechanism could be under both 515 feedforward and feedback control. This sets the stage for a dynamic interaction between 516 thalamic and cortical states to drive highly interactive patterns of neural activity. 517