Introduction

Somatosensory signals reach primary motor cortex (Goldring et al., 1970; Strick and Preston, 1982; Swadlow, 1994; Naito et al., 2002; Ferezou et al., 2007)(Hatsopoulos and Suminski, 2011; Mao et al., 2011; Murray and Keller, 2011; Alonso et al., 2023; Reddy et al., 2024). To do so, signals from mechanoreceptor afferents ascend primarily via lemniscal pathways and somatosensory thalamus to arrive in primary somatosensory cortex (S1), which sends corticocortical projections to primary motor cortex (M1). The thalamus→S1→M1 circuits are well-studied in model systems such as the rodent vibrissal pathways (Ferezou et al., 2007; Mao et al., 2011; Petersen, 2019). We recently confirmed and extended these findings for the corresponding circuits of mouse hand/forelimb S1 and M1, using slice-based optogenetic and electrophysiological methods to delineate synaptic circuit connections along the transcortical sensorimotor loop (Yamawaki et al., 2021).

A salient feature of the hand/forelimb circuits is the transition, on reaching S1 cortex, from a “streamlined” organization of the ascending subcortical circuits via thalamus to a “densely polysynaptic” organization of the local and interareal cortical circuits (Yamawaki et al., 2021). Here, we sought to characterize the timing, amplitude, and related properties of the cortical dynamics in the same S1 and M1 circuits, evoked by stimulating hand mechanoreceptors. We developed an optogenetic-electrophysiological approach guided by the desiderata of delivering brief stimuli to awake animals during simultaneous recordings of spiking activity in both areas, leveraging a mouse line enabling cortical as well as peripheral optogenetic manipulations. This approach enabled quantitative characterization of fundamental properties of the spiking dynamics of hand/forelimb S1 and M1 in response to brief peripheral stimulation.

Results

Brief stimulation of the hand evokes barrages of spiking activity in S1 and M1

To study cortical responses to hand stimulation, we aimed to develop a paradigm for delivering short-duration stimuli while electrophysiologically sampling spiking activity in hand/forelimb S1 and M1 cortical areas of awake mice. The rationale for using brief stimuli was to obtain the equivalent of an impulse response function of the system. For example, with millisecond-scale stimulation, essentially all activity, including any longer-lasting responses extending in time well after the initial pulse of input, can be ascribed to the initial input alone. We adopted an optogenetic approach for delivering “phototactile” somatosensory stimuli to the mouse’s hand with millisecond precision, combining this with cortical electrophysiological recordings of evoked spiking activity (Methods). Motivations for this approach included the ease of precisely controlling stimulus timing and intensity, and compatibility with awake animals.

Crossing a proprioceptor-specific driver line of mice (PV-Cre) with a channelrhodopsin-2 (ChR2) reporter line (Ai32) yielded expression of ChR2 in peripheral mechanoreceptor afferents, particularly around the palm (Fig. 1A) (Methods). We chose this line because it fortuitously also labels parvalbumin-type GABAergic neurons in the cortex, enabling cortical optogenetic manipulations as well. To sample stimulus-evoked cortical responses, we recorded from multi-channel linear arrays inserted in S1 and M1 of awake, head-fixed mice as they rested their hands on holding bars harboring an optical fiber through which light stimuli were delivered to the palm. As shown in an example recording, photostimulation of the hand with 5-ms pulses of blue light elicited barrages of cortical spiking activity in both areas (Fig. 1B). Although hand stimulation evoked activity in forelimb M1, it did not evoke movements of the forelimb (Supplementary Fig. S1). Following experiments, brains were sectioned and inspected to stereotaxically localize the fluorescent tracks of the dye-coated linear arrays to confirm the recording locations (Supplementary Fig. S2). This approach thus provides a way to record and analyze S1 and M1 cortical activity evoked by brief photostimulation of the hand in awake mice.

Fig. 1.

Evoked cortical responses follow a triphasic peak-suppression-rebound pattern

To characterize the cortical responses, we recorded from hand/forelimb S1 and M1 in a cohort of animals and analyzed the evoked activity; except where otherwise indicated, results reported below are all based on this sample (13 recordings from 9 animals), and analyses focused on stimulus-responsive units (Methods). Responses to hand photostimulation in multiple (generally 25) trials were analyzed based on alignment to the onset of photostimulation, to generate average peristimulus time histograms (PSTHs) for each responsive unit on each probe, as shown for one example recording (Fig. 2A-E), for each experiment as average PSTHs (Fig. 2F), and as grand average PSTHs across all recordings (Fig. 2G). In trials where the mouse’s hand was not resting on the light-delivery bar, no responses were observed (Fig. 2E, G). This also served as a control for potential visual responses to the flashing light. As seen in the individual experiments as well as in the grand averages, peripheral stimulation evoked a triphasic response in both areas, with an initial short-latency large-amplitude peak followed by an interval of suppressed activity and a subsequent late lower-amplitude rebound. We proceeded to analyze the multiple components of the evoked responses to hand stimulation in greater detail.

Fig. 2.

Initial peak responses in M1 are delayed and attenuated relative to S1

Focusing first on the short-latency initial peak, we calculated basic response parameters for individual responsive units for each experiment, and as averages over responsive units (Fig. 3; Table 1). Prior to stimulus onset, baseline firing rates were low in S1 and even lower in M1. Following stimulus onset, S1 activity rapidly peaked, followed by M1 activity (Fig. 3A). Only a subset of units was stimulus responsive in both areas, on average more in S1 (36.7%) than in M1 (25.6%) (Fig. 3B). Responses differed substantially in onset latency (14.7 ms in S1, 24.3 ms in M1) (Fig. 3C). Smaller peak amplitudes were found in M1 (25.8 Hz) compared to S1 (83.0 Hz) (Fig. 3D). The average ratio of M1 over S1 peak amplitude was 0.36 ± 0.16 (mean ± s.d.). Response durations were brief and slightly longer in S1 (27.0 ms) than in M1 (22.4 ms) (Fig. 3E). These patterns were consistent across experiments.

Fig. 3.

Table 1.

We explored which response features may be correlated (Fig. 3F). Perhaps unsurprisingly, both in S1 and M1, peak amplitude strongly correlated with onset latency (Fig. 3G) and peak latency (S1: ρ = −0.69, p = 0.01; M1: ρ = −0.71, p = 0.01); the faster the onset, the larger the amplitude. Few other significant correlations were found, however. Response durations were not significantly correlated with peak amplitudes in S1, where responses were generally strongest, and only relatively weakly correlated in M1 (Fig. 3H). Thus, larger responses terminated just as quickly as smaller ones. Although response amplitudes were consistently bigger in S1 than in M1 they were not significantly correlated with each other. This may reflect variability in probe placements, the numbers and types of units recorded, and other factors contributing to animal-to-animal variability.

Having measured the latencies of the evoked spiking activity reaching S1 and M1, we also estimated the corresponding effective propagation speeds. To do so, we estimated the hand-to-S1 and S1-to-M1 pathway distances and divided these by the latencies (Methods). The hand-to-S1 distance was estimated as 44.3 ± 1.1 mm, and the hand-to-S1 propagation speed was accordingly 3.0 ± 0.1 m/s based on onset latencies (2.1 ± 0.1 m/s based on peak latencies). The S1-to-M1 distance, measured as the Euclidean distance between the probes, was 0.92 ± 0.12 mm (mean ± s.d.), and the estimated S1-to-M1 propagation speed was accordingly 0.14 ± 0.11 m/s based on the difference (M1 minus S1) in onset latencies (0.14 ± 0.10 m/s based on peak latencies). Thus, the propagation speed for the entire subcortical leg of the ascending pathway was approximately 20-fold faster than for the short corticocortical leg, from S1 to M1.

These results quantify basic properties of the initial component of the evoked cortical spiking activity in S1 and M1. In S1, responses begin ∼15 ms after stimulus onset and reach peak amplitude ∼6 ms later. In M1, responses have a similar temporal profile but with a ∼10 ms lag and approximately one-third of the S1 amplitude. Thus, M1 receives a slightly delayed and substantially attenuated short-latency somatosensory response from brief hand stimulation. In both areas, responses self-terminate, extinguishing within ∼25 ms.

Post-peak activity is suppressed in S1 and rebounds in both S1 and M1

Focusing next on the later phases of the evoked responses, we noted that the initial burst of evoked activity was self-terminating, and particularly in S1 often then fell below the pre-stimulus baseline level as well (Fig. 4A). The fraction of all stimulus-responsive units that showed post-peak suppression was relatively small (22.3% in S1, 8.4% in M1), but activity of these units was greatly suppressed (to 8.2% of baseline activity level in S1, and 5.6% in M1) (Fig. 4B, C; Table 1).

Fig. 4.

In both areas, activity also tended to rebound following suppression, more prominently in S1 than in M1 and approximately simultaneously (Fig. 4D-F; Table 1). Rebound activity in S1 was observed in 30.9% of stimulus-responsive units, firing on average at 31.4 Hz, while rebound activity in M1 involved 8.1% of units firing at 20.2 Hz.

We explored correlations and scaling relationships among the components of the triphasic responses (Fig. 4G). Rebound amplitude and initial peak amplitude were strongly correlated in both S1 and M1 (Fig. 4H). In S1, the proportion of units showing suppression correlated with the amplitude of the initial peak (Fig. 4I) but not with the rebound (Fig. 4J). Rebound timing (latency) was not strongly correlated with any parameters, except for rebound amplitude in M1 only (Fig. 4G). Other correlations were weaker.

This triphasic peak-suppression-rebound pattern implies that after the initial sensory response the cortex becomes inhibited, and then disinhibited as inhibition self-terminates, as observed in other studies of sensory cortex (see Discussion). To explore this, we tested the sensitivity of the cortex to a second photostimulus, timed to coincide with the intervals of either the suppressed or rebound activity, or later (i.e., a paired-pulse paradigm, with variable lag) (Fig. 4K-N). In both S1 and M1, responses to the second stimulus fell to approximately half the amplitude of the first during the suppressed interval, returning to normal during the rebound interval and after (Fig. 4M, N).

These results show that both S1 and M1 excitability recovers by 250 ms after hand photostimulation. This implies that somatosensory input generates in S1, and to a lesser extent in M1, a sequence of excitation, inhibition, and disinhibition, with the two later stages scaling in proportion to the initial response amplitude. The self-terminating property of the initial large-amplitude response together with the ensuing suppression of activity suggests strong initial recruitment of inhibitory interneurons. Therefore, we next investigated the response properties of cortical interneurons, using cell-type-specific methods to identify them in the recordings and manipulate their activity.

PV neurons in S1 are strongly recruited by hand stimulation

In the same recordings, in nearly every experiment (12 recordings from 8 mice) we also used opto-tagging to identify a subset of inhibitory neurons within the population, focusing on recordings in S1, where the suppression effects were strongest, as described above. Conveniently, in the same PV-Cre x Ai32 mice in which peripheral mechanoreceptor afferents are labeled, cortical parvalbumin-expressing (PV) interneurons are also labeled. Thus, application of blue-light photostimuli to S1 while recording local population spiking activity enabled detection of activated PV neurons based on short-latency sustained firing responses (Methods; Fig. 5A, B). With this approach we parsed the population spiking activity evoked by hand photostimulation into PV versus non-PV units (Fig. 5C).

Fig. 5.

Analysis of this data set showed that peripheral stimulation activates a population comprising a larger proportion of PV (64.4% of all activated units) than non-PV units (35.6%) (Fig. 5D). The proportion of stimulus-responsive units among the PV units (64.0%) was also higher than among non-PV units (31.1%) (Fig. 5E). Furthermore, PV units collectively accounted for most of the evoked spikes (76.1%) (Fig. 5F). Further analysis showed an interesting mix of differences and similarities in key parameters across the three phases of the stimulus-evoked responses. The initial peak responses arrived with ∼5-ms shorter latency for PV than for non-PV units (Fig. 5G), consistent with the initial positive deflection in the difference PSTH (Fig. 5C, bottom). The PV units also reached ∼2-fold higher peak amplitude (Fig. 5H). During the suppression phase, the percentage of units exhibiting significant suppression was higher for PV (30.4%) and lower for non-PV units (18.0%) (Fig. 5I), and consistent with the intermediate value (22.3%) observed for all units, as noted above (Fig. 4B). The activity of these units, both PV and non-PV, was greatly suppressed (to 7.3% of baseline activity level for PV and to 9.2% for non-PV units) (Fig. 5J). In the subsequent rebound phase, a higher fraction of PV (37.6%) than non-PV (27.3%) units showed rebound activity (Fig. 5K). No differences were found in rebound amplitudes or latencies (Fig. 5L,M).

These results show that PV-type interneurons in S1 are recruited strongly and rapidly during the initial cortical response to hand stimulation. Indeed, the evoked firing of PV units even leads that of non-PV units by several milliseconds, consistent with prior observations of vibrissal S1 cortical responses to whisker deflection (Yu et al., 2016). Following the suppression phase of the response, in the rebound phase the PV units are again robustly engaged, with a temporal profile similar to the non-PV units. The prominent role of PV neurons in all phases of the responses raises the possibility that activation of PV neurons alone might emulate some or all aspects of the responses evoked by hand stimulation, which we next explored.

Selective activation of PV neurons generates suppression and rebound

Using the same mice and cortical photostimulation methods that were used in the opto-tagging experiments described above, we analyzed the effect of focally activating PV units in S1 cortex (12 recordings from 8 mice). As above, we used opto-tagging to parse PV versus non-PV units in S1, and examined their activity along with that of all units recorded simultaneously in M1 (Fig. 6A). Following the selective activation of PV units, the fraction of S1 units exhibiting significant post-peak suppression was similar for PV (18.1%) and non-PV (19.6%) units (Fig. 6B). During this interval, the firing rate diminished more for PV (to 7.3% of baseline) than for non-PV (to 10.9%) units (Fig. 6C). In the subsequent phase, 20.0% of PV and 14.9% of non-PV units exhibited rebound activity (Fig. 6D); no differences were found in rebound amplitudes and minor differences were found in the latencies (Fig. 6E, F). In M1, the overall activity also followed a pattern of suppression and rebound (Fig. 6A, G-K).

Fig. 6.

Similarly, in a subset of experiments (4 recordings from 3 mice), we selectively activated PV neurons in M1. This was also able to induce suppression and rebound locally (in M1), involving both PV and non-PV units (Supplementary Fig. S3). In both areas, focal stimulation at progressively lower stimulus intensities resulted in reduced responses, and at 25% stimulus intensity the main effect was local activation of PV units in the stimulated area, with little suppression and rebound locally and none in the other (non-stimulated) area (Supplementary Fig. S4).

These results show that the selective activation of PV neurons in both S1 and M1 cortex can generate a pattern of suppression and rebound closely resembling that observed with hand stimulation.

Partial silencing of S1 reduces M1 responses to hand stimulation

If M1 responses to hand stimulation depend on S1, then manipulations that reduce S1 activity during hand stimulation should also cause a reduction in the M1 responses. To test this, we performed focal stimulation of S1 PV neurons as above, in this case pairing it with simultaneous hand stimulation (Fig. 7). As before, hand stimulation alone evoked strong responses in S1 and weaker and delayed responses in M1 (Fig. 7A). We used low-intensity laser stimulation of the cortex (at 25% of the intensity used for the opto-tagging and cortical silencing experiments above) to isolate the effects of locally activating PV neurons in S1 without suppressing ongoing activity in M1 (Fig. 7B). When hand stimulation was paired with S1 silencing, the activity in S1 followed a hybrid pattern (reflecting the combined effect of activating the PV units and partially suppressing the sensory-evoked response), and the activity in M1 was reduced (Fig. 7C, D). A significant reduction of M1 activity was consistently observed within and across experiments (4 recordings from 4 mice; Fig. 7E; Table 2). When S1 was partially silenced during hand stimulation, responses in S1 fell to 69.7% of control levels overall, with reduced activity of both PV (to 74.2%) and non-PV units (to 58.8%). At the same time, M1 responses also fell, to 76.2% overall. These results thus indicate that the responsiveness of M1 to peripheral stimulation of the hand depends at least in part on propagation of sensory-evoked activity via S1.

Fig. 7.

Table 2.

Discussion

To characterize the propagation of somatosensory activity along the previously characterized forelimb S1 and M1 cortical circuits in the mouse (Yamawaki et al., 2021), we developed an optogenetic approach for delivering brief photostimuli to the hand while recording and in some cases manipulating population activity in the two cortical areas. The findings identify basic properties and mechanisms of population spiking dynamics in this model system for investigating neural mechanisms of sensorimotor integration in transcortical loops.

Methodological considerations

The optogenetic method for peripheral stimulation used here, though artificial, closely mimics basic features of natural somatosensory stimulation, as shown in numerous prior studies involving related approaches (O’Connor et al., 2013; Millard et al., 2015; Prsa et al., 2019; Emanuel et al., 2021; Lehnert et al., 2021; Schorscher-Petcu et al., 2021). Like electrical stimulation, which is also used in somatosensory experiments, a key advantage is the ability to deliver temporally precise (i.e., brief) stimuli to the hand. However, electrical stimulation non-specifically activates somatosensory afferents, including pain fibers, and while tolerated by human subjects usually requires general anesthesia for studies in non-human mammalian model species (Zaforas et al., 2024).

Another advantage of the optogenetic approach is the ability to restrict labeling to particular somatosensory afferents of interest. A potential disadvantage of the particular approach used here, based on using a Cre line to drive expression, is that ChR2 may be expressed in the cortex as well, potentially constraining options for cortical optogenetic manipulations. However, in this case, we specifically chose the PV-Cre line because ChR2 is also expressed in cortical PV neurons, enabling cortical opto-tagging and silencing in the same experiments. A related point is that the particular Cre line used here (PV-Cre) is one that labels proprioceptor afferents (Hippenmeyer et al., 2005; Wu et al., 2021); other lines are available for labeling diverse somatosensory subtypes of DRG neurons; e.g. (Neubarth et al., 2020; Sharma et al., 2020; Wu et al., 2021).

Our approach relied on repeated stimulation of the hand as it rested on the bar during each ∼30 s block of trials, which was unproblematic in these experiments where the goal was to broadly stimulate a large but not necessarily identical population of mechanoreceptors distributed across the palm. However, for other experimental goals such as receptive field mapping, this configuration would need to be modified.

M1 responses reflect S1→M1 corticocortical activity

To what extent do the evoked responses reflect propagation of activity along S1→M1 corticocortical circuits (Yamawaki et al., 2021)? In principle, other pathways to M1 could be involved. For example, somatosensory thalamus projects not only to S1 but also S2, which projects to M1 (Suter and Shepherd, 2015). However, thalamocortical projections to S2 arise mainly from caudal whisker-related thalamic nuclei, while the forelimb-related pathways, including those for proprioception, are concentrated in rostral nuclei projecting to S1 (Francis et al., 2008; Alonso et al., 2023; Rubio-Teves et al., 2024). As another possibility, activity in S1 might drive M1 via thalamus (Mo and Sherman, 2019). However, in the mouse hand/forelimb system, cortico-thalamo-cortical pathways are primarily organized as recurrent loops (Guo et al., 2020), which may recurrently drive corticocortical activity (Shepherd and Yamawaki, 2021). The reduction in M1 responses when S1 activity was partially silenced directly implicates S1 in driving M1 responses, and the non-PV silenced neurons in S1 likely included corticocortical neurons. Technical limitations, relating to the proximity of M1 and S1 and the spatial resolution of focal silencing, precluded even stronger silencing of S1 activity. Although residual M1 activity was most likely driven by remaining upstream activity within forelimb S1, other, albeit more circuitous, pathways may secondarily contribute. Overall, however, our findings indicate that M1 responses arise primarily from feedforward propagation of activity along S1→M1 circuits.

Response latencies reflect fast subcortical and slow corticocortical circuits

In S1, the evoked activity developed quickly following hand stimulation, with onset latencies of 15 ms on average, consistent with those found previously using related approaches (Emanuel et al., 2021). For comparison, in the rat, whisker deflection generates cortical responses with an even shorter latency of ∼10 ms (Ahissar et al., 2000), consistent with the shorter pathway distance for the whisker-barrel system. We estimated the effective propagation speed along the ascending pathway, from the periphery to S1, to be 3 m/s. Interestingly, in humans the hand-to-S1 latency is similarly short, ∼20 ms, with correspondingly much higher conduction velocities for the ascending pathways; e.g. (Desmedt and Cheron, 1980).

In M1, the evoked activity also developed quickly, lagging S1 by 10 ms. However, although the latencies were brief, the effective corticocortical propagation speed was slow, estimated at 0.14 m/s. This speed is comparable to the conduction velocity of thin unmyelinated cortical axons, estimated to be ∼0.2 m/s (Raastad and Shepherd, 2003), and to inter-areal propagation speeds in other corticocortical circuits in mouse cortex, estimated to be approximately 0.3 m/s (Li et al., 2018). Both the hand-to-S1 and S1-to-M1 pathways are polysynaptic, and the estimated propagation speeds thus reflect multiple biophysical properties (axonal conduction velocities, synaptic transmission delays, and more).

This 20-fold difference in subcortical versus corticocortical propagation speeds is consistent with previous characterizations of the synaptic circuit organization of the same pathways as “streamlined” for the subcortical ascending pathways to S1 and “densely polysynaptic” for the corticocortical pathway from S1 to M1 (Yamawaki et al., 2021). Thus, sensorimotor transformations in this system appear designed for signal propagation that is fast from periphery to the cortex, then slows drastically from S1 to M1 (and other areas). The slowness of corticocortical dynamics in this system accords with the idea that speed-limited sensorimotor transformations may be a general property of neural systems (Zheng and Meister, 2024).

Response amplitudes and durations reflect intrinsic mechanisms of termination and attenuation

In M1, the initial peak response amplitudes were substantially attenuated relative to S1, where amplitudes were high, far above the low baseline firing rates. In both areas, responses were self-terminating, with brief durations. Response profiles resemble the “packet”-like activity patterns previously described in diverse circuits, which have been proposed as fundamental units of information processing in neural systems (Luczak et al., 2015). They also closely resemble those of optogenetically evoked activity patterns observed in higher-order corticocortical pathways (Li et al., 2018).

Responses were not only self-terminating but also self-extending, in the sense that their durations (tens of milliseconds) were much longer than the peripheral stimulus that evoked them (5 ms). At the cortical level, this temporal dispersion of the responses did not accumulate going from S1 to M1, reflecting tight control and termination of excitation along the corticocortical pathway.

Response termination likely involves multiple mechanisms, given the importance of preventing runaway excitation. For example, cellular-level mechanisms such as spike frequency adaptation rapidly attenuate firing of mouse S1 and M1 neurons (Hattox and Nelson, 2007; Suter et al., 2013). Circuit-level mechanisms include the strong activation of cortical inhibitory interneurons by somatosensory input, considered next.

Triphasic S1 responses reflect mass activation of PV neurons

The expression of ChR2 in cortical PV neurons enabled identification of these neurons by opto-tagging during the hand-stimulation experiments, and thus quantification of their roles in the evoked responses in S1. Prior studies in S1 have shown that somatosensory stimulation potently activates cortical inhibitory interneurons, particularly PV-type neurons, with stronger and faster (shorter latency) recruitment compared to excitatory neurons (Simons, 1978; Zhu and Connors, 1999; Bruno and Simons, 2002; Swadlow, 2003). Similarly, here we found that PV neurons in S1 are massively recruited by hand stimulation, both in terms of the number (approximately equal to that of non-PV units) and the firing rate (approximately double that of non-PV units). Indeed, most (∼76%) of the evoked spikes in S1 cortex during the initial peak response were from PV neurons. PV units were moreover recruited slightly earlier (i.e., with shorter response latencies) than non-PV units. We did not study PV units in M1 in response to hand stimulation, but prior studies indicate similarly strong, short-latency recruitment of fast-spiking neurons in hand/forelimb M1 as well (Murray and Keller, 2011).

PV units remained involved in the ensuing suppression and rebound phases as well. During suppression, PV spiking fell, rather than increased; thus, the overall suppression could not be ascribed to ongoing firing of these inhibitory interneurons, but likely reflects other mechanisms, such as slow kinematics of GABAergic responses. During rebound, the contribution of PV spiking to the overall activity was again high compared to non-PV units, which in this case reflected an increase in the number of PV units rather than their firing rates. These findings are consistent with prior characterizations of peak-suppression-rebound responses in S1 (Chapin et al., 1981; Zhu and Connors, 1999).

Selective activation of PV units alone in either S1 or M1 evoked a sequence of suppression and rebound activity similar to that evoked by hand stimulation alone, indicating a prominent role of PV activity in shaping the triphasic cortical responses. Focal activation of PV neurons, used here for the purpose of mechanistically probing cortical somatosensory responses, is also commonly used for the purpose of transient focal silencing (Li et al., 2019). Thus, the rebound activity studied here as part of the triphasic somatosensory response is closely related to the rebound excitation that can occur as an unwanted confounding side effect of focal activation of interneurons (Wiegert et al., 2017; Li et al., 2019). Our findings emphasize that this is an inherent property of the cortex, elicitable not only by peripheral sensory input but also intracortically by PV activation or electrical stimulation, and likely involving intrinsic electrophysiological mechanisms as well as local cortical and cortico-thalamo-cortical (CTC) circuits (Grenier et al., 1998; Claar et al., 2023; Kumaravelu and Grill, 2024; Russo et al., 2024). Forelimb-related S1 and M1 in the mouse both form strongly recurrent CTC circuits with thalamus (Guo et al., 2018; Guo et al., 2020).

Pairing selective activation of PV units in S1 with simultaneous hand stimulation caused the evoked somatosensory responses in both S1 and M1 to be attenuated. These findings suggest a role for PV neurons in S1 in regulating the extent to which S1 activity influences M1 activity. More generally, the findings obtained by opto-tagging PV neurons in S1 indicate a prominent role for these neurons in shaping both the local S1 and downstream M1 responses to somatosensory stimulation of the hand.

Perspectives

The S1-to-M1 transcortical loop – the longest of several sensorimotor loops embedded along the neuraxis to enable somatosensory feedback control over a range of time scales – may be particularly important for mediating rapid adjustments to ongoing somatosensory information in the context of current and planned cortical motor commands (Reschechtko and Pruszynski, 2020). Our findings in the mouse provide a basis for comparisons to homologous systems, particularly the rodent vibrissal and primate hand/forelimb circuits; as alluded to above, these can exhibit notable similarities as well as differences in latencies and other dynamic properties. Our characterizations of basic response parameters in mouse hand/forelimb circuits also provide a framework for relating dynamic spiking activity to the underlying cells and their synaptic connections in this system (Yamawaki et al., 2021); as a first step, we explored the use of optogenetic manipulations of PV neurons. Additionally, building on the rapidly growing knowledge about cellular connectivity and behavior-related dynamics in “top-down” corticocortical circuits; e.g. (Petreanu et al., 2009; Petreanu et al., 2012; Xu et al., 2012; Lee et al., 2013; Zagha et al., 2013; Kinnischtzke et al., 2014; Manita et al., 2015), the stimulation and recording paradigms developed here could be adapted to investigate how inputs from higher-order motor areas such as secondary motor cortex (M2, or rostral forelimb area) interact with ascending somatosensory activity to modulate M1 output during goal-directed behaviors.

Materials and methods

Animals

Experimental studies on mice followed protocols approved by the Institutional Animal Care and Use Committee of Northwestern University and fully complied with the animal welfare guidelines of the National Institutes of Health and the Society for Neuroscience. Two mouse lines were used in this study: the PV-Cre driver line of mice (B6.129P2-Pvalb^tm1(cre)Arbr/J; RRID:IMSR_JAX:017320) (Hippenmeyer et al., 2005), which express Cre in peripheral mechanoreceptor afferents, particularly in proprioceptors, and also express Cre in cortical interneurons, particularly in parvalbumin-type cells; and, the Ai32 reporter line of mice (B6.Cg-Gt(ROSA)26Sor^{tm32(CAG-COP4*H134R/EYFP)Hze}/J; RRID:IMSR_JAX:024109) (Madisen et al., 2012), which expresses a ChR2-EYFP fusion protein in Cre-expressing cells. Homozygous mice of each line were crossed to generate PV-Cre x Ai32 offspring. Mice were bred in-house and housed in groups on a 12-hour reverse light/dark cycle, with free access to food and water. Experiments were conducted during the dark phase of the light cycle. Adult mice of both sexes were used (12 male, 7 female). Mice were not selected based on sex and were used as they became available.

Surgical procedures

Head-bar implantation was performed as previously described (Barrett et al., 2022). Briefly, mice under deep isoflurane anesthesia were placed in a stereotaxic frame. The scalp and periosteum were removed to expose the cranium, a titanium bar was cemented over lambda, and the exposed cranium was covered with dental cement. For analgesia, mice were given 0.3 mg/kg buprenorphine preoperatively, and 1.5 mg/kg meloxicam both immediately and one day postoperatively.

In a subset of experiments, corticospinal neurons were labeled by injecting a retrograde AAV (pAAVretro-CAG-GFP, 37825-AAVrg; Addgene, Watertown, MA) (Tervo et al., 2016) into the spinal cord at the same time as head-bar mounting, following previously described methods (Barrett et al., 2022).

For one experiment, EMG electrodes were fabricated and implanted as previously described (Barrett et al., 2024). Briefly, in the same surgery as the head-bar implantation, an incision was made over the upper forelimb and a knotted pair of Teflon-coated stainless steel wire (Catalog Number 793200, A-M Systems, Sequim, WA) crimped into a bent 27G needle was led under the skin, then inserted into the biceps brachii. The distal end of the wire pair was then knotted, the needle and excess wire cut away, and the incision sutured. The wires proximal to the knot were connected to a rectangular pin connector (CLP-112-02-F-D-A, Samtec, New Albany, IN), which was affixed to the head-bar with dental cement. After this surgery, the mouse received 10 mg/kg of enrofloxacin in addition to postoperative analgesia.

Electrophysiological recordings

In vivo electrophysiological recordings from awake mice were performed similarly to previous studies (Barrett et al., 2022). Mice were habituated first to handling by the experimenter and then to head fixation for 3 days prior to recording. Then, under deep isoflurane anesthesia, a dental drill was used to open one or more craniotomies over the cortical areas to be recorded. Recordings were made using 64-channel silicon probes (model A1×64-Poly2-6mm-23s-160-A64, Z-coated; NeuroNexus, Ann Arbor, MI) and opto-probes (model A1×64-Poly2-6mm-23s-160-OA64LP, Z-coated, core diameter of the coupled fiber 105 μm, NA 0.22, termination 200 μm above top channel; NeuroNexus), with ∼1 MΩ impedance, and 23-μm spacing in staggered configuration (total recording length 1449 μm).

Probes were mounted on a linear translator (MTS25-Z8, ThorLabs, Inc.) on a 3-axis manipulator (1U RACK, Scientifica, Uckfield, UK). Probes were slowly (5 μm/s) inserted under software control (Kinesis, Thorlabs, Inc.) into the cortical target areas to a nominal depth of 1500 μm. Target coordinates for forelimb M1 were 0.25 mm anterior-posterior (AP), 1.35 mm medial-lateral (ML); those for forelimb S1 were 0.0 mm AP, 2.4 mm ML (Yamawaki et al., 2021). A subgroup of mice with both corticospinal and probe-track labeling was used to further confirm accurate targeting of these coordinates, based on prior characterizations of corticospinal neuron distributions extending from the hand/forelimb-related subregions of M1 medially to S1 laterally (Ueno et al., 2018; Yamawaki et al., 2021). For each mouse, recordings were made over one or two days from S1 and M1 of one or both hemispheres. After each recording session the probes were removed and the craniotomy was re-sealed with Kwik-Sil (World Precision Instruments, Sarasota, FL). Mice were transcardially perfused with 4% paraformaldehyde in phosphate buffered saline after finishing all experiments and the brains were processed to localize the probe tracks as previously described (Barrett et al., 2022).

Data acquisition hardware included RHD2132 headstages (Intan Technologies, Los Angeles, CA) and an Open Ephys data acquisition board (OEPS-6501, Open Ephys, Lisbon, Portugal). Signals were sampled at 30 KHz. Software control was with the Open Ephys GUI. To record EMG signals, the loose wires of an Intan 36-pin wire adapter were soldered to FTSH-112-03-F-DV-ND pin connectors that mate with the socket connectors of the EMG implant. This allowed EMG signals to be recorded through an RHD2216 bipolar headstage (Intan Technologies, Los Angeles, CA) connected to an Intan RHD2000 USB evaluation board

Spike sorting

Open Ephys binary files were converted to raw format using Matlab (The MathWorks, Natick, MA). As previously described (Barrett et al., 2022), spikes were detected and sorted using Kilosort (Pachitariu et al., 2016; Steinmetz et al., 2021) and verified using PHY (https://github.com/cortex-lab/phy), using standard methods to reject artifactual units (waveforms spanning more than 3 adjacent channels, or atypical waveform shapes) and identify single units (refractory period; <1% of spikes within 1 ms) and multi-units (all others). Multi-units on the same channel with similar waveform shapes were merged. Single units were merged only if they were on the same channel, displayed similar shapes, and had no spikes within ±1 ms in their cross-correlogram (Hall et al., 2021). Single units and multi-units were pooled as “active units”. Only probe recordings with at least 15 active units were analyzed.

Optogenetic photostimulation of the hand

The experimental apparatus was assembled on a vibration isolation table, and included a raised platform on which a 3D-printed hut was placed for the mouse to sit in, head-bar holders with screw clamps, and hand rest. For photostimulation of hand mechanoreceptors, the bar for the hand rest was fashioned out of a thick-walled plastic tube (diameter: 3.2 mm outside, 1.5 mm inside) by drilling a through-hole and inserting the exposed end of an optical fiber (910 μm core diameter, 0.22 NA, 1.5 m length, FG910UEC Patch cable, Thorlabs, Inc.). The fiber was coupled to a blue LED (455 nm wavelength, M455F3 source, LEDD1B driver; Thorlabs, Inc.). The bar was positioned so that the fiber tip faced the palm when mice placed their hand on the resting bar. With this configuration, light stimuli broadly illuminated the volar aspect of the hand. We visually monitored the hand placement during each block of trials, and blocks in which the mouse moved the hand were discarded.

Photostimulation parameters were controlled by Wavesurfer software (v0.945, https://wavesurfer.janelia.org/). Light intensity at the fiber tip was calibrated with a power meter (PM100D, Thorlabs, Inc.). Stimulus duration was 5 ms and intensity was 5 mW at the fiber tip, except where noted otherwise. Photostimuli were applied to the hand contralateral to the cortical recordings. For each experiment, light pulses were delivered in blocks of 25 or 27 trials with an inter-trial period of 1 s. In addition, we also recorded responses when the mouse’s hand was resting off the bar, either as a control or in combination with cortical photostimulation (described next).

Optogenetic photostimulation of cortical PV neurons

In most cases, optogenetic activation of PV neurons in S1 or M1 was performed while mice had their hands off the light-delivery bar, using a sequence of 25 pulses of blue light delivered at the tip of the opto-probe. The probe was inserted as described above and coupled via patch cord (M63L01, Thorlabs, Inc.) to the blue laser source (473 nm wavelength, MBL-III-473-100 mW, Opto Engine LLC), with an illumination intensity of 1 mW at the fiber tip. For a subset of silencing experiments, the optogenetic activation of PV neurons was interleaved with trials of hand stimulation or both hand and focal S1 stimulation while the mouse kept their hand resting on the light delivery bar. The illumination intensity was reduced to 0.25 mW at the fiber tip. There was a total of 27 trials per condition in this case. In all cases, the stimulus duration for PV activation was 20 ms.

Analysis of stimulus-aligned data

For each active unit, spiking activity was binned in 5 ms bins, aligned to stimulus onset. Peristimulus time histograms (PSTHs) spanning 0.5 s before to 0.5 s after stimulation were constructed by averaging the stimulus-aligned binned firing rates over trials (generally 25). The PSTHs were z-scored, and stimulus-responsive units were identified as those for which the maximum z-scored firing rate exceeded 2.5 s.d. above the baseline mean value (250 ms preceding the stimulus onset).

Analyses focused on stimulus-responsive units, except where otherwise indicated. Dynamic properties of each responsive unit’s PSTH were extracted as follows. To analyze properties associated with the early main response, for each trial, we first baseline-subtracted the raw PSTHs using their baseline mean value. Peak amplitude was calculated as the maximum firing rate in the average baseline-subtracted PSTH, over a time window of 100 ms post-stimulus. The time bin of the maximum post-stimulus firing rate was taken as the peak latency. The time bin when the firing rate first exceeded 2.5 s.d. was considered as the onset latency. Response duration was computed as the time window from when the firing rate first exceeded 2.5 s.d. above baseline to when it first fell below 1 s.d above baseline.

To analyze properties associated with the post-peak suppression phase of the responses, we identified units with significant suppression: for each unit, we calculated the mean firing rate during this interval, between 110 to 170 ms (based on visual inspection of average PSTHs), and statistically compared this to the mean of the pre-stimulus baseline (Wilcoxon’s signed rank test). As the response tended to rebound after this period, we identified units with significant rebound: for each unit, we calculated the rebound amplitude by finding the maximum of the PSTH during the rebound window, between 190 to 400 ms, and averaging from 20 ms before to 20 ms after this value, and statistically compared this with the mean of the pre-stimulus baseline.

For each experiment, once the parameters and PSTHs were computed for all responsive units, values were pooled and averaged. These averages were then pooled to obtain grand averages across experiments.

Probe track localization

As previously described (Barrett et al., 2022), recording sites were confirmed by coating the probes with a fluorescent dye (DiI, Vybrant Multicolor Cell Labeling Kit, Invitrogen, Carlsbad, CA) prior to recordings, harvesting and histologically processing the brains after experiments, and localizing the probe tracks in 100 μm brain sections under epifluorescence microscopy. In two cases, because probe tracks could not be recovered histologically, areal assignment was based on the targeted coordinates.

Pathway morphometry

S1-to-M1 distance was calculated as the Euclidean distance between the S1 and M1 recording sites, localized based on probe tracks as described above. To estimate the hand-to-S1 distance we separately estimated the hand-to-cord and cord-to-cortex distances and summed these. The distance from the palm of the hand to the sixth cervical (C-6) spinal cord segment was estimated from morphometric measurements of cadaveric specimens (N = 5 mice). The distance from C-6 cord to S1 cortex was estimated from published MRI images (Harrison et al., 2013).

Statistical analysis

Group data are presented as mean ± s.d. unless otherwise indicated. For group comparisons, we primarily used nonparametric tests. Except where indicated otherwise, for paired samples Wilcoxon’s signed rank was calculated with a minimum sample size of n = 6. Where the number of experiments was smaller than this, paired t-Student tests were calculated instead (Fig. 7E, S3), or within-animal Wilcoxon’s signed rank tests were performed (Table 2), using the Benjamini-Hochberg procedure applied to the resulting p-values to control for the false discovery rate (Benjamini and Hochberg, 1995). The Friedman test was used for one-way repeated measures analysis of variance, with a minimum sample size of n = 4, and Dunn-Sidak as a post hoc multiple comparison test. For correlations, Spearman’s coefficients were calculated and t-tested afterwards. In all cases significance was set at p < 0.05.

Supplementary figures

Supplementary Fig. S1.

Supplementary Fig. S2.

Supplementary Fig. S3.

Supplementary Fig. S4.

Acknowledgements

We thank Megan Martin and Miraya Baid for technical assistance.

Additional information

Data availability

Data and code will be made available at Zenodo.

Funding

National Institute of Neurological Disorders and Stroke (NS125594, NS061963) - Gordon Shepherd.

Ethics

Animal experimentation was performed in accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health, following guidelines and procedures approved by the institutional animal care and use committee (IACUC) of Northwestern University.

Author contributions

• Daniela Piña Novo: Conceptualization, data curation, formal analysis, investigation, methodology, software, resources, visualization, writing original draft, writing review and editing.

• Mang Gao: Conceptualization, resources, writing review and editing.

• Jianing Yu: Conceptualization, writing review and editing.

• John M. Barrett: Conceptualization, formal analysis, investigation, methodology, project administration, software, resources, supervision, writing review and editing.

• Gordon M. G. Shepherd: Conceptualization, funding acquisition, project administration, resources, writing original draft, writing review and editing.