The Secondary Somatosensory Cortex Gates Mechanical and Thermal Sensitivity

Abstract The cerebral cortex is vital for the perception and processing of sensory stimuli. In the somatosensory axis, information is received by two distinct regions, the primary (S1) and secondary (S2) somatosensory cortices. Top-down circuits stemming from S1 can modulate mechanical and cooling but not heat stimuli such that circuit inhibition causes blunted mechanical and cooling perception. Using optogenetics and chemogenetics, we find that in contrast to S1, an inhibition of S2 output increases mechanical and heat, but not cooling sensitivity. Combining 2-photon anatomical reconstruction with chemogenetic inhibition of specific S2 circuits, we discover that S2 projections to the secondary motor cortex (M2) govern mechanical and thermal sensitivity without affecting motor or cognitive function. This suggests that while S2, like S1, encodes specific sensory information, that S2 operates through quite distinct neural substrates to modulate responsiveness to particular somatosensory stimuli and that somatosensory cortical encoding occurs in a largely parallel fashion.

Top-down control of somatosensory encoding in the spinal cord by the brain allows for 24 context-dependent modulation of behavioral responses based on changing intrinsic or 25 environmental conditions 1,2 . Somatosensory function relies on this top-down control to 26 accurately predict, evaluate, and appropriately react to mechanical, thermal, and chemical 27 stimuli 2,3 . We and others have found that the primary somatosensory cortex (S1) controls 28 somatosensory reflexive behaviors through excitatory corticospinal neurons that innervate the 29 dorsal horn 4,5 . This constitutes a feed-forward circuit whereby incoming sensory information 30 ascends to S1 for processing and through descending excitatory projections, subsequent 31 information flow in the spinal cord is amplified to facilitate accurate behavioral responses. 32 Inhibition of this S1 corticospinal circuit, therefore, produces decreased mechanical sensitivity 4 . 33 However, S1 does not appear to encode all somatosensory modalities, with a strong bias toward 34 mechanical and cooling inputs and an absence of heat encoding 4,6,7 . This suggests that distinct 35 anatomical regions may participate in the full spectrum of somatosensory top-down control. We 36 sought to identify what other cortical regions that could encode the properties that S1 does not. 37 The secondary somatosensory cortex (S2) is an adjacent cortical region that also processes 38 somatosensory stimuli 8-10 . It has been proposed, based on cortico-cortical and thalamo-cortical 39 neural circuitry, latency of responses, and homology to other sensory cortical areas, that S1 and 40 S2 exist in a hierarchy, with S2 as the higher-order cortical area, processing distinct features of 41 the somatosensory experience similar to the visual system 9-12 . The mouse whisker system has 42 provided some evidence supporting this model and subsets of neurons in S2 appear to encode 43 behavioral choice and recalling of past experience 10,13,14 . However, whether this is the case for 44 somatic stimuli from the body is unclear. S1 can also encode complex aspects of the 45 somatosensory experience questioning a hierarchical relationship and favoring the processing of information in parallel 15,16 . Evidence for parallel information processing, where differential modalities are processed within S1 and S2 is accumulating, particularly in humans [17][18][19] . 48 We hypothesized that the secondary somatosensory cortex (S2) may be a crucial substrate 49 for those somatosensory modalities that S1 does not encode. Using a combination of 50 optogenetics and chemogenetics, we now identify the hindpaw area of S2 as a cortical region 51 whose output, in contrast to S1, mitigates mechanical and heat sensitivity. Circuit mapping 52 studies combined with intersectional circuit manipulation strategies identify a higher-order 53 cortical area, the secondary motor cortex (M2), as the target of S2 that governs somatosensory 54 sensitivity. These findings reveal that S2 is a key controller of evoked somatosensory behaviors 55 in a manner quite distinct of that of S1 circuits and dependent on cortico-cortical connectivity to 56 suppress specific somatosensory responses. 57 58 Results: 59

Somatosensory Behavioral Responses but Does Not Produce Aversion 61
To determine if the secondary somatosensory cortex (S2) has a role in somatosensory 62 behaviors we virally targeted expression of channelrhodopsin (ChR2) into parvalbumin (PV) 63 inhibitory interneurons in the hindlimb region of mouse S2 to induce inhibition of the output 64 projection targets included the superior colliculus (SCo), the caudate putamen (CPu) (shown in 186 Fig. 3c, d, and e (arrow)), the periaqueductal gray (PAG), and cervical corticospinal tract, all of 187 which have been implicated in somatomotor circuitry 35 . S2 also made significant corticocortical 188 projections, the densest of which were local to the adjacent auditory/temporal association 189 cortices (AUD/TEa). As observed by others, S2 made significant connections to the contralateral 190 S2, as well as ipsilateral/contralateral projections to the primary somatosensory cortex (S1) (Fig.  191 3a, e) and to the ipsilateral primary motor cortex (M1) 36 (Fig. 3d). We also observed a major 192 projection to the ipsilateral prefrontal cortex, specifically the secondary motor cortex (M2) 37 193 The secondary motor cortex (M2) receives sensory inputs and exerts complex effects on 197 behavior. Indeed, during a sensory decision-making task, widefield imaging demonstrates 198 sensory to frontal waves of activity that occurs before a behavioral choice 39-41 . Further, lesioning 199 or silencing of M2 produces alterations in this behavioral choice 42-45 . We therefore hypothesized 200 that S2 connectivity with the M2 region might underlie the observed somatosensory sensitivity 201 where behavioral choice is altered to a hypersensitive state. 202 We first anatomically defined the neural identity of S2 to M2 projection neurons by 203 injecting either a retrograde dye (CTB-555) or sparse-labeling with retrograde AAV-tdTomato 204 into M2 and performed RNAscope in situ hybridization and immunofluorescence using defined 205 markers (Fig. 3g, top). The majority of S2-to-M2 neurons were located in layer V, specifically 206 Va, with scattered neurons throughout layer II/III (Fig. 3g, bottom). Using four markers of layer 207 V neurons, we found that S2-to-M2 neurons are positive for the callosal intratelencephalic (IT) excitatory neuron markers Trib2, Etv1, and Satb2, but negative for Ctip2, a marker of layer Vb 209 excitatory corticospinal neurons (Fig. 3i) [46][47][48] . Further, the scattered neurons within layer II/III 210 were also found to be excitatory as they express FoxP1, a marker of cortical excitatory neurons 211 ( Fig. 3h) 49 . This agrees with tracing data from the Allen Brain Atlas in which anterograde tracing 212 of excitatory layer V neurons in S2 contributes to the largest population of M2 projections with 213 some projections from layer II/III 50 (Supp. Fig. 4). S2 to M2 projecting neurons are, therefore, 214 excitatory intratelecephalic (IT) projection neurons, largely originating from layer Va. 215 216

Layer Specific Connectivity Reveals Hierarchical Relationship Between S2 and M2 217
The layered structure of the cortex is organized in a manner to facilitate differential 218 computations. Cortico-cortical communication pathways can stem from layer II/III to layer II/III 219 or from deeper layer IT neurons targeting superficial layers 38,51,52 . Modulatory pathways 220 typically innervate the superficial layers whereas driver pathways tend to innervate deeper 221 layers 52,53 (but see 54 ). We set out to address which layer of M2 is innervated by S2 with 222 monosynaptic rabies tracing using Cre drivers that express in layer II/III (Penk-Cre) or layer V 223 (Rbp4-Cre) (Fig. 4a). In this strategy, Cre-positive neurons act as "starter cells" that are selective 224 hosts of rabies infection and retrograde monosynaptic transport. This identifies "input cells", 225 thereby providing direct evidence of monosynaptic connectivity between two neurons (Fig. 4a). 226 Viral injections into M2 successfully targeted layer II/III and layer V neurons, respectively, 227 labelling many "starter cells" (Fig. 4b (bi-bii), c (ci-cii)). Examination of the rabies virus 228 mCherry+-labelled "input cells" in S2 compared to the total "starter cell" number in M2 revealed 229 the majority of the S2-to-M2 projection neurons in reside in deeper layers of S2 (notably layer V 230 in line with our previous retrograde tracing) and that they primarily innervate superficial layer 231 II/III Penk+ neurons in M2 (Fig. 4 biii-biv and ciii-civ). This arrangement is in line with the role 232 of these neurons playing a modulatory role on M2 activity. Recent M2 single cell sequencing 233 studies have identified layer II/III Penk+ neurons as excitatory neurons 55,56 . This suggests that 234 inhibition of S2 during somatosensory stimulation, results in decreased M2 activity, and 235 compatible with the recent findings that the net effect of S1/S2 inhibition on L2/3 neurons in M2 236 is inhibitory 57 . 237 238

Sensitivity 240
To functionally manipulate S2-to-M2 IT neurons we used an intersecting viral strategy 241 with injection of a retrograde AAV encoding Cre-recombinase into the M2 region along with an 242 AAV encoding either a Cre-dependent excitatory (HM3Dq) or an inhibitory (hM4Di) 243 chemogenetic receptor (designer receptor activated exclusively by designer drugs (DREADDs)) 244 to either activate or inhibit S2-to-M2 projection neurons with the ligand clozapine n-oxide 245 (CNO) 58,59 (Fig. 5a, b). CNO application rapidly and reversibly inhibited firing in S2-to-M2 246 projection neurons in slices from the inhibitory DREADD mice (Fig. 5c). Likewise, analysis of 247 c-Fos expression as a marker of activity-dependent early immediate gene transcription 248 demonstrated that CNO injection significantly upregulated c-Fos expression in animals injected 249 with a Cre-dependent excitatory DREADD (HM3q) but not a mCherry virus (Supp. Fig. 5a-d). 250 This demonstrates our chemogenetic approach can increase or decrease S2-to-M2 neural activity 251

efficiently. 252
Examining mechanical and thermal sensitivity in these animals revealed that injection of 253 animals that express the inhibitory DREADD in the S2-to-M2 projection neurons with CNO 254 phenocopied our behavioral observations with optogenetic inhibition of S2 by activation of PV+ 255 inhibitory interneurons. Specifically, 30 minutes after CNO injection, S2-to-M2 inhibitory 256 DREADD animals showed strong tactile hypersensitivity (hM4Di post-CNO contralateral 257 (0.5534g) paw vs. ipsilateral (0.8936g)) and heat sensitivity (hM4Di contralateral paw 6.007s vs. 258 mCherry contralateral paw 16.53s) (Fig. 5d, e). Again, cold sensitivity remained unaffected 259 (mCherry contralateral 2.480s vs. hM4Di contralateral 2.867s) (Fig. 5f). Both the paw ipsilateral 260 to the virus injection and animals infected with a control Cre-dependent mCherry virus showed 261 no change in mechanical/thermal sensitivity, confirming this effect is limited to the targeted 262 hemisphere ( Fig. 5d-f). Excitation of this circuit with an excitatory DREADD produced no effect 263 on tactile sensitivity irrespective of the concentration of CNO used, suggesting that it is only the 264 inhibition of this circuit that specifically governs sensitivity to somatic stimulation ( Fig. 5d-f and 265 Supp. Fig. 5e). 266 To confirm that it is the inhibition of the input from axons projecting from S2 into the M2 267 cortical region that is responsible for the increase in mechanical and heat sensitivity, we inserted 268 cannulas into M2 in inhibitory DREADD animals, and locally microinjected CNO (as used in: 269 60,61 ) ( Fig. 5g, h). Local microinjection of CNO (300nl of 300µM) in M2 increased behavioral 270 mechanical sensitivity similar to that produced by systemic (i.p. 3mg/kg) injections (300µM 271 CNO 0.4778g vs. Saline 0.9243g) (Fig. 5i) and also had an effect on heat sensitivity (300µM 272 CNO vs. 6.031s Saline 10.48s) (Fig. 5j). This indicates that it is the output from S2 to M2 that 273 influences mechanical and thermal nociceptive threshold sensitivity. 274 The secondary motor cortex is suggested to be involved with the planning of motor 275 behavior in addition to modulating behavioral responses 44,62 . The increased withdrawal response 276 sensitivity to mechanical and thermal stimulation observed on inhibiting S2 projections to M2 could therefore, be due to an altered motor reactivity. To ascertain whether motor behavior was 278 significantly altered by changing activity of S2-to-M2 neurons, we analyzed gait and motor 279 function of mice infected with DREADDs. Examining the sciatic functional index (SFI) to 280 compare the position of one hindpaw (DREADD affected) to the other (control) during 281 locomotion, we found no significant differences in their locomotory behavior (Supp. Fig. 6a). 282 There was also no change in stride length (Supp. Fig. 6b). This suggests that the S2 to M2 circuit 283 is primarily sensory in nature and exerts its effects on an animals response to sensory stimulation 284 independent of motor planning and learning. 285 286

Discussion: 287
The central neural substrates of somatosensory behavior and how they work together to 288 orchestrate both simple and complex sensory experiences, is still largely unknown. Top-down 289 circuits, often higher order central to lower order central/peripheral circuits, are important 290 modulators of behavior. In somatosensory circuits, top-down modulation can occur from cortical 291 circuits (primary somatosensory cortex (S1), anterior cingulate cortex (ACC)) and subcortical 292 circuits (amygdala and brainstem) 4,63-65 . We previously characterized a S1 excitatory 293 corticospinal circuit in which inhibition produces decreased mechanical sensitivity 4 . However, 294 we noted that this circuit failed to respond to or alter heat-evoked thermosensory behaviors 4 . 295 Indeed, recent work has confirmed this showing that S1 primarily encodes cooling but not 296 heating 6 . We now show that the secondary somatosensory cortex (S2) can fulfill this role and 297 also modulate mechanical responses in a distinct manner from S1. Fiber photometry recordings 298 demonstrate that S2 can respond to mechanical, heat, and cooling applied to the hindpaw. Yet, 299 optogenetic inhibition of S2 produces mechanical and heat hypersensitivity without affecting 300 cooling sensitivity. Other work has identified the posterior insular cortex as a crucial substrate of 301 thermosensory behaviors 6 . While close in anatomical space and sharing some connectivity, 302 whether they interact to achieve a common goal or act on common downstream elements to exert 303 an influence on behavior is an interesting future direction. 304 Previously defined somatosensory cortical circuits which mitigate sensorimotor action 305 originate from the primary somatosensory cortex (S1) 4,66 and exert action via direct or indirect 306 spinal connections or through connections with the primary motor cortex (M1) 67 , or from S2 to 307 S1/M1 13,14,36 . However, we now find that the secondary motor cortex (M2) is an important 308 cortical substrate that S2 acts through. Indeed, chemogenetic inhibition of S2-to-M2 neurons identified S2 as important for decision making 14,71 . Specifically, S2 to S1 projections are thought 318 to be important in the encoding of choice following a stimulus 14 . It is likely that these three 319 neural substrates work together to process sensory information and produce behavioral 320 responses. However, our behavioral assays are intrinsically different in that the choice (paw 321 withdrawal or not) is not a learned behavior nor coupled to any predictable stimulus. This 322 indicates that S2 inputs to M2 may gate an animal's behavioral response to defined somatosensory stimuli, in addition to its more complex roles in sensory decision making, choice, 324 association, and learning. Our anatomical tracing supports this theory by demonstrating that layer 325 Va pyramidal neurons in S2 predominantly provide input to layer II/III neurons in M2, a 326 connectivity pattern typically associated with modulating rather than driving cortical responses 52 . 327 Taken together, this places S2/M2 circuitry as a core mediator of somatosensory 328 behavior. How this circuit works in collaboration with other defined cortical circuits is an 329 interesting future direction, but we hypothesize since distinct areas of the somatosensory cortex 330 appear to process distinct modalities that these circuits operate in a parallel fashion to provide a 331 comprehensive picture of the somatosensory environment. For cannula studies, a 0.8mm cannula was implanted in the M2 region and affixed to the skull 530 with dental cement. Skull screws were used to stabilize the implant as above. A cap that 531 consisted of a 1mm dummy cannula was used when the injector unit was not in use. 532 For rabies tracing studies, either AAV2/9-Syn-Flex-TVA-oG-GFP or AAV2/8-Syn-Flex-TVA-533 oG-GFP was injected into the secondary motor cortex of either Penk-Cre (Jax # 025112) or 534 Rbp4-Cre (MGI:4367067) animals at a depth of 200 (to target layer II/III) or 400 micrometers 535 (to target layer V), respectively. Six weeks post-injection, SADdg-EnvA-mCherry was injected 536 at four points along M2 to capture a wide breadth of starter cells. Animals were taken for 537 histology one week following. 538 Viruses used in this study include: AAV2/9.Syn.Flex.GCaMP6f.WPRE.SV40 (1E+13 Addgene 539 100833 -100nl), AAV2/1-CAG-FLEX-rev-ChR2-tdtomato (1.23E+13 gc/mL -Boston 540 Children's Hospital Viral Core -125nl), AAV2/retro-CAG-Cre-WPRE (2.89E+13 gc/mL -541 Boston Children's Hospital Viral Core -125nl), AAV2/1-Syn-DIO-hM4Di-mCherry (Boston 542 Children's Hospital Viral Core -125nl), AAV2/9-hSyn-DIO-Hm3D(Gq)-mCherry 543 (6.21432E+13 gc/mL -Boston Children's Hospital Viral Core -125nl), AAV2/9-hSyn-DIO-544 mCherry (Boston Children's Hospital Viral Core) -125nl, AAV2/9-CAG-tdTomato-WPRE 545 (1.0234E+13 gc/mL Boston Children's Hospital Viral Core -50nl), AAV2/retro-hSyn-546 tdTomato-WPRE (Boston Children's Hospital Viral Core -125 nl), SADdg-EnvA-mCherry 547 (2.26E+08 to 4.10E+10 TU/mL Boston Children's Hospital Viral Core), AAV2/9-Syn-Flex-548 TVA-oG-GFP (1e13 gc/ml Boston Children's Hospital Viral Core) AAV2/8-Syn-Flex-TVA-oG-549 GFP (1e13 gc/ml Boston Children's Hospital Viral Core). 550

Electrophysiology: 551
To obtain brain slices containing S2, mice were anesthetized using isoflurane and decapitated 552 into oxygenated (95% O2; 5% CO2) ice-cold cutting solution (in mM): 130 K-gluconate, 15 KCl, 553 0.05 EGTA, 20 HEPES, and 25 glucose (pH 7.4 adjusted with NaOH, 310-315 mOsm). The 554 brain was then removed quickly and immersed in the ice-cold cutting solution for 60 seconds. 555 Coronal slices containing S2 were sectioned and collected. The brain was cut with a steel razor 556 blade, then sectioned into 250 μm-thick slices in the oxygenated ice-cold cutting solution using a 557  Histology 591 Mice were perfused transcardially with 4% paraformaldehyde (PFA) in PBS. Brains were 592 isolated and fixed overnight in 4% PFA before storage in 1 X PBS. Brains were sectioned with a 593 vibratome between 60-100µm or a cryostat at 30µm and mounted on slides (Fisher Permafrost). 594 Vibratome sections were permeabilized with 1 X PBS with 0.2% Triton-X100, mounted on 595 slides, and coverslipped with mounting media containing DAPI. All injection sites were aligned 596 back to the Allen Brain Atlas, injection sites with substantial off-target infection were excluded. 597 In the cannula experiments, signal from the AAV2/9-CAG-DIO-DREADD(h4Dmi) was 598 amplified using anti-mCherry (Abcam: ab167453) at 1:500 by first blocking with 1 X PBS with 599 10% goat serum and 0.3% Triton X-100 for one hour at room temperature, incubated with anti-600 mCherry overnight at 4°C, followed by incubation with goat anti-rabbit 568 (Thermo Scientific: 601 A-110011) for one hour at room temperature, and cover slipped with DAPI mounting media. For 602 identification of excitatory cortical neurons in layer II/III, anti-Foxp1 (Abcam: ab16645 1:500) 603 was used with goat anti-rabbit 488 (Thermo Scientific: A-11008). Slides were visualized and 604 imaged with a Nikon Ti-1200. 605

RNAscope
Injections of cholera toxin subunit b conjugated to Alexa Fluor 555 (CTB-555) (Thermo 607 Scientific C34776) was injected at 1X 150-200nl in the secondary motor cortex as described 608 above. Two weeks following injection, animals were processed as above but after incubation 609 with PFA were placed into 30% sucrose in 1 X PBS for 2 days. Brains were rinsed in 1 X PBS 610 and snap frozen in optimal cutting temperature compound (TissueTek 4583) and stored at -80°C 611 until processing. Brains were sectioned with a cryostat at 20 micrometers. RNAscope was 612 with Olympus VS120 SlideScanning microscope. 616

Fiber Photometry 617
Mice were habituated to the fiber optic cable for one hour each on two separate days. Photometry 618 signals were acquired by alternative illumination with 470nm (GCaMP) and 410nm light 619 (isobestic control) (Neurophotometrics FP3001). Any mice with calcium responses that occurred 620 during whisker deflection (suggesting fiber was placed in barrel cortex) or sound (suggesting 621 fiber was place in adjacent auditory cortex) were excluded and only animals with responses to 622 hindpaw stimulation were used. Tactile stimuli, heat stimuli, and cold stimuli were presented 10, 623 5, and 10 times per mouse separated by at least 30 seconds in the case of tactile and 1 minute in 624 the case of thermal. Analysis of Z-scored delta F/F was performed in Matlab as described in 72 . 625 For von Frey and acetone, averaged traces across animals were aligned to the point when the 626 stimulus was applied. For heat stimuli via Hargreave's (see below), traces were aligned to the 627 paw withdrawal event, as a ramping heat stimulus until a paw withdrawal in a freely-moving 628 mouse can produce have variable times of heat exposure (see Fig. 1g for example of variation) .
Motor behaviors in chemogenetic manipulated mice were recorded via Digigait. In brief, mice 676 were placed on an illuminated treadmill with a camera below to capture paw placement. The 677 treadmill was set to 20cm and mice allowed to walk for a 5 minute period. Sciatic functional 678 index and stride length were calculated with the Digitgait Noldus software. 679

Two Photon and One Photon Serial Tomography Mapping of S2 Projection Targets 680
Two mice with 50nl injections of AAV2/8-CAG-tdTomato in S2 were used for serial two photon 681 tomography mapping. Mice were perfused transcardically with 4% paraformaldehyde (PFA), 682 brains isolated and incubated in 4% PFA overnight at 4°C. Brains were then embedded in 4.5% 683 oxidized agarose and coronal sections of 1.38um^2 resolution at 50um optical section were taken 684 using the TissueCyte (Tissue Vision Inc.). Images were aligned back to the Allen Brain Atlas 685 using Neuroinfo software (MBF Bioscience) and projections were manually annotated and 686 quantified using ImageJ. One mouse with a 100nl injection of AAV2/8-CAG-tdTomato in S2 687 was processed traditionally with a vibratome and imaged with an IXM Confocal microscope to 688 confirm the two photon findings. 689

Clozapine-N-Oxide (CNO) administration during behavioral tests 690
To activate DREADD chemogenetic receptors in behavioral tests, 3 mg/kg clozapine-n-oxide 691 (CNO -Enzo Biosciences BML-NS105) (first dissolved in DMSO and saline added until 0.02% 692 DMSO final solution) was injected intraperitoneally 30 minutes prior to behavioral 693 measurement. All behavioral measurements were separated by at least 24 hours to ensure 694 metabolism of CNO. 695

Clozapine-N-Oxide (CNO) local administration during behavioral tests
For local administration of CNO in M2 via cannula, a 33 gauge injector cannula was attached to 697 the cannula pedestal (P1 Technologies C315GS). Mice were habituated to the injector cannula 698 for 2 days prior to behavioral assessment for 30 minutes each. On the day of testing, mice were 699 habituated to the cannula injector for 5 minutes, 300 nanoliters of saline or different 700 concentrations of CNO injected at 100nl/min, and diffusion allowed to proceed for 2 minutes 701 before behavioral assessment 702

Validation of Excitatory DREADD via cFos Staining 703
Three mice with S2-to-M2 neurons labeled with either mCherry or Excitatory DREADD were 704 injected with 3mg/kg CNO. Animals were placed in the dark for one hour before perfusion. 705 Histology proceeded as detailed above. 706

Zymosan Injection into the Hindpaw for Peripheral Inflammation 707
Zymosan (20µl 5mg/mL Sigma: Z4250) was injected into the hindpaw and GCaMP fiber 708 photometry was performed as above during von Frey and Hargreave's assays at baseline and the 709 peak of sensitivity (4 hours post injection).