Convergence of forepaw somatosensory and motor cortical projections in the striatum, claustrum, thalamus, and pontine nuclei of cats

The basal ganglia and pontocerebellar systems regulate somesthetic-guided motor behaviors and receive prominent inputs from sensorimotor cortex. In addition, the claustrum and thalamus are forebrain subcortical structures that have connections with somatosensory and motor cortices. Our previous studies in rats have shown that primary and secondary somatosensory cortex (S1 and S2) send overlapping projections to the neostriatum and pontine nuclei, whereas, overlap of primary motor cortex (M1) and S1 was much weaker. In addition, we have shown that M1, but not S1, projects to the claustrum in rats. The goal of the current study was to compare these rodent projection patterns with connections in cats, a mammalian species that evolved in a separate phylogenetic superorder. Three different anterograde tracers were injected into the physiologically identified forepaw representations of M1, S1, and S2 in cats. Labeled fibers terminated throughout the ipsilateral striatum (caudate and putamen), claustrum, thalamus, and pontine nuclei. Digital reconstructions of tracer labeling allowed us to quantify both the normalized distribution of labeling in each subcortical area from each tracer injection, as well as the amount of tracer overlap. Surprisingly, in contrast to our previous findings in rodents, we observed M1 and S1 projections converging prominently in striatum and pons, whereas, S1 and S2 overlap was much weaker. Furthermore, whereas, rat S1 does not project to claustrum, we confirmed dense claustral inputs from S1 in cats. These findings suggest that the basal ganglia, claustrum, and pontocerebellar systems in rat and cat have evolved distinct patterns of sensorimotor cortical convergence.


Introduction
The cerebral cortex contains numerous modality-specific regions, with each area specialized for a specific type of information processing (for review see Krubitzer 2009). For example, somatosensory and motor cortical areas are thought to process aspects of touch and movement, respectively (Borich et al. 2015;Chapin and Lin 1984;Hall and Lindholm 1974;Hoffer et al. 2003;Kaas 2004; but see Matyas et al. 2010). However, information across these two modalities must be integrated in both directions to achieve appropriate somesthetic-guided movements and accurate sensory perceptions of the resulting bodily motion (Arce-McShane et al. 2016;Khateb et al. 2017;Lee et al. 2008Lee et al. , 2013Murthy and Fetz 1992;Pais-Vieira et al. 2013;Umeda et al. 2019;Zagha et al. 2013). In addition to direct corticocortical connections between sensory and motor cortices that mediate such integration, each of these cortical areas also projects to numerous subcortical structures, some of which represent additional opportunities for sensorimotor integration (Battaglia-Mayer and Caminiti 2019; Chakrabarti and Alloway 2006;Chakrabarti et al. 2008;Colechio and Alloway 2009;Fabri and Burton 1991;Ghosh et al. 1987;Hamadjida et al. 2016;Hoffer et al. 2003Hoffer et al. , 2005Huerta and Pons 1990;Izraeli and Porter 1995;Jones et al. 1978;Mao et al. 2011;Smith and Alloway 2013;Stepniewska et al. 1993Stepniewska et al. , 2020Suter and Shepherd 2015;Vogt and Pandya 1977).
Among the subcortical targets of sensorimotor cortex, the thalamus contains higher-order nuclei such as the posteromedial nucleus (POm), which receives convergent inputs from both sensory and motor cortical areas and projects back to both areas (for review see Alloway 2008;Alloway et al. 2008;Hoffer and Alloway 2001;Lu and Lin 1993;Mo and Sherman 2019). In addition, the claustrum has connections that could facilitate interactions between sensory and motor cortices (for review see Smith et al. 2019a;Jackson et al. 2020). Our previous work in rodents, for example, has shown that the claustrum receives dense inputs from motor cortex and sends bifurcating projections back to both sensory and motor cortex Alloway 2010, 2014;Smith et al. 2012). Similarly, early work in primates indicates that the dorsal region of the claustrum has reciprocal connections to somato-motor cortices (Pearson et al. 1982).
In addition to cortical projections to the thalamus and claustrum, the cerebral cortex also sends topographically organized projections to the basal ganglia and pontocerebellar systems (Brodal 1978;Dudman and Gerfen 2015;Flaherty and Graybiel 1991;Hoover and Strick 1993;Kunzle 1975Kunzle , 1977Oh et al. 2014;Proville et al. 2014;Schmahmann and Pandya 1997;Schwarz and Mock 2001;Suzuki et al. 2012;Wiesendanger and Wiesendanger 1982a, b). In turn, both the basal ganglia and the cerebellum provide feedback to sensorimotor cortex via multi-synaptic outputs to the thalamus, which are believed to remain segregated, thus forming parallel cortico-subcortico re-entrant loops (Alexander et al. 1986(Alexander et al. , 1990Aoki et al. 2019a;Brodal and Bjaalie 1992;Hoover and Strick 1993;Mercier et al. 1990;Middleton and Strick 2000;Parent and Hazrati 1995; but see Aoki et al. 2019b for evidence of convergence in basal ganglia loops). However, though largely topographic, whole-cortex mapping studies of cortico-striatal projections have demonstrated convergence of modality related cortical inputs into functional domains of striatum that are thought to have specific behavioral relevance (Hintiryan et al. 2016;Hunnicutt et al. 2016;Oh et al. 2014). In the sensorimotor domain, previous work has demonstrated convergence of sensory and motor cortices in the dorsolateral region of the striatum in rodents Charpier et al. 2020;Hoffer and Alloway 2001;Hooks et al. 2018;Hoover et al. 2003;Lee et al. 2019;Ramanathan et al. 2002) and primates Graybiel 1993, 1995), an area of the striatum that is responsible for guiding sensorimotor habits (see Alloway et al. 2017 for review). Similar anatomical tracing studies in the cortico-pontine system of rodents, cats, and primates indicate parallel themes of integration to those observed in the cortico-striatal system (Brodal 1978;Bjaalie and Brodal 1989;Leergaard et al. 2000a, b;Leergaard et al. 2004;Legg et al. 1989;Schwarz and Thier 1995;Wiesendanger et al. 1979;Wiesendanger and Wiesendanger 1982a, b).
In the current work, we sought to expand on these previous studies by investigating cortical sensorimotor integration in multiple subcortical structures in felines. The feline clade, which is part of the Laurasiatheria superorder, diverged from rodents and primates, members of the superorder Euarchontoglires, nearly 100 million years ago (Murphy et al. 2004). By employing the same physiologically guided neuroanatomical tracing paradigm used previously to map sensorimotor convergence in rodents (Alloway et al. 2000;Hoffer and Alloway 2001;Leergaard et al. 2004), we can make direct comparisons to felines to assess the hypothesis that different phylogenetic clades evolved distinct neural circuits for sensorimotor integration. Therefore, we used microstimulation mapping of primary motor cortex (M1) and receptive field mapping of primary and secondary somatosensory cortex (S1 and S2, respectively) to identify overlapping forepaw representations in each cortical area. We then injected each area with a different neuroanatomical tracer. This tripletracer paradigm allowed us to assess integration from three modality related cortical areas in the striatum, claustrum, thalamus, and basal pontine nuclei within the same animal. Our results show that, whereas, S1 and S2 projections show strong convergence in rodents, felines show significantly more S1-M1 integration, indicating that different strategies in cortico-subcortical interactions evolved across phylogenetically separate lineages.

Materials and methods
Anatomical tracing experiments were performed on ten adult male felines (Charles River) weighing 3.1-5.0 kg. All procedures were approved by Penn State University's Institutional Animal Care and Use Committee and conformed to National Institute of Health standards.

Animal surgery
Surgical procedures were performed using sterile technique with support from veterinary staff for anesthesia and animal care (Penn State College of Medicine, Department of Comparative Medicine). In preparation for injections, animals received an injection of dexamethasone (0.1 mg/kg, intramuscular) the day before surgery. The following morning, animals were initially anesthetized with an intra-muscular injection of ketamine (10 mg/kg) and acepromazine (0.5 mg/ kg). The femoral artery was catheterized, the head was shaved, the trachea was intubated through the oral cavity, and the femoral artery was catheterized to enable intravenous injections of dexamethasone 0.1 mg/kg) and chloramphenicol (50 mg/kg). The head was then fixed into a stereotaxic device (Kopf) and the animal was ventilated with a mixture of oxygen and sevoflurane (1-2%) to maintain a stable anesthetic plane for the remainder of the procedure as determined by lack of nociceptive withdrawal reflexes and flaccid muscle tone. A sensor was attached to the tongue and vital signs (heart rate, specific oxygen, respiratory rate, and in/end tidal CO 2 ) were monitored continuously (Surgivet) and recorded every fifteen minutes. Fluids were maintained via saline drip connected to the femoral catheter. Body temperature was monitored via a rectal thermistor and was maintained by hydraulic heat pumps underneath the animal, as well as a hot air enclosure around the torso (Bear-Hugger). Ophthalmic ointment was applied to the eyes to prevent corneal drying.
The skin over the cranium was disinfected with betadine and a sterile drape was place over the entire head. Prior to the incision, the skin over the area of the craniotomies was infiltrated with numerous subcutaneous injections of lidocaine to provide local anesthesia. A single large incision was made over the right side of the skull to provide access to the frontal and parietal bones. The skin was retracted, and the underlying fascia and muscle was resected to expose bregma and the sites of the craniotomies. After measuring with surgical rulers, two craniotomies approximately 15 mm in diameter were drilled, with one spanning the primary motor (M1) and primary somatosensory (S1) cortices, and the other around the secondary somatosensory (S2) cortex.

Intracranial microstimulation (ICMS)
Following craniotomies, tungsten microelectrodes (FHC, 0.5-1.0MΩ), affixed to a stereotaxic micromanipulator were inserted in M1 coordinates to map evoked muscle twitches as used in one of our previous rodent studies (Hoffer and Alloway 2001 for M1). The electrode was inserted into the cortex via a small slit in the dura made with a 27G needle held by a pair of forceps. The electrode was lowered to a depth of ~ 1.0-1.5 mm below the dural surface to target layer 5, which contains corticospinal neurons (Suter and Shepherd 2015). A constant current source (BAK BSI-2) controlled by a digital timer (Master-8), was used to administer electrical stimulation (0.7 ms pulse, 250 Hz frequency). Current levels were initially high (~ 100 μA) and then reduced to identify the minimum current necessary for eliciting limb or forepaw movements at each cortical site (typically < 30 μA). To evoke muscle twitches, sevoflurane levels had to be reduced to 0.5-0.75%; this led to a slight increase in muscle-tone but maintained insensitivity to forepaw pinches. The electrode was moved to numerous sites to obtain a detailed map of the M1 body representation that included the center of the forepaw representation. Typical movements included dorsoflexion of the entire forepaw and brief twitches of individual digits.

Extracellular neuronal recordings.
To identify the forepaw representation in S1 cortex and S2 cortex, a tungsten microelectrode (FHC, 1.0-1.5MΩ) was used to map both areas. Initial coordinates were identified based on previous studies (Alloway and Burton 1985a, b;Felleman et al. 1983), and detailed maps were obtained for each individual animal. Electrodes were inserted via a dural slit and lowered to a depth of 400-600 μm below the surface, corresponding to layer 4, the primary input layer from thalamus. Neuronal activity was amplified (Dagan, 1 k head-stage amp, 2 k amp, band pass filtered 300 Hz-5 kHz) and observed on an oscilloscope (Grass Digital) as well as through an audio monitor. The wooden end of a cottontipped applicator was used to probe the surface of the skin to elicit a neural response. Receptive fields were drawn, and coordinates recorded to provide a detailed map of both S1 cortex and S2 cortex, such that injections could be targeted to overlapping representations of the forepaw.

Tracer injections
Following detailed mapping of M1, S1, and S2 cortices, tracer injections were placed in overlapping representations of the forelimb. Each cortical area was targeted with a different anterograde tracer (Molecular Probes, Eugene OR) including Fluoro-Ruby (FR, D-1817), Alexa-Fluor 488 (AF, D-2290), and Biotinylated Dextran Amine (BDA, D-1956). While BDA was always injected into M1, FR and AF injections were injected into either S1 or S2 (see Table 1 for summary). Tracers were injected by means of 2.0 μL Hamilton syringes secured onto a microinjector (Kopf) attached to the stereotaxic frame as done previously (Alloway et al.  2000,2009). Each syringe had a pulled glass pipette, with a 10-50 μm tip, cemented onto the end of the Hamilton syringe needle to minimize the damage to the cortex. The tip of the pipette was inserted ~ 1.7 mm from the surface to ensure tracer deposit within layer 6 and in more superficial layers as the tracer tended to seep upward along the injection pipette. Tracers were injected slowly (100 nL/min), and total tracer volume injected at each cortical site varied from 400 to 900 nl. Each cortical area (i.e., M1, S1, or S2) received 4-7 injections to cover a wide region of the forepaw representation (see Table 1 for summary). Following mapping and tracer injections, Gel-film was placed over the cortical surface and Gel-foam inserted into the craniotomy. The incision was subsequently sutured and triple antibiotic ointment applied to the wound margins. The margins of the skin incision were injected with bupivacaine (1-2 ml) to provide local anesthesia, and the animal received a long-lasting analgesic, buprenorphine (0.01 mg/kg IM), just prior to discontinuation of the sevoflurane and removal of the head from the stereotaxic. Upon recovery of selfrespiration, the intubation tube and femoral catheter were removed, and animals were monitored until fully conscious and able to right themselves. Animals were returned to their home cage and monitored daily with veterinary-assisted post-operative care for the duration of the experiment.

Histology
Following mapping and tracer injections, Gel-film was placed over the cortical surface and Gel-foam inserted into the craniotomy. The incision was subsequently sutured and triple antibiotic ointment applied to the wound margins. At completion of all procedures, buprenorphine was administered for post-operative analgesia, the animal's head was removed from the stereotaxic and sevoflurane was discontinued. Upon recovery of self-respiration, the intubation tube and femoral catheter were removed, and animals were monitored until fully conscious and able to right themselves. Animals were returned to their home cage and monitored daily with veterinary-assisted post-operative care for the duration of the experiment.
After 10 days to allow for tracer transport, animals were deeply anesthetized with ketamine, xylazine and acepromazine. A femoral catheter was inserted, and animals were over-dosed with a lethal quantity of sodium pentobarbital (Nembutal 100 mg/kg). Animals were trans-cardially perfused with heparinized saline (3 cc heparin/L saline) until the liver blanched, and then subsequently perfused with 4% paraformaldehyde in 0.1 M PBS and 4% paraformaldehyde in 0.1 M PBS with 10% sucrose. The brain and spinal cord were removed and stored in 4% paraformaldehyde in 0.1 M PBS with 30% sucrose for cryoprotection. The day after the perfusion, the arachnoid was removed from the cortical surface with Dumont forceps and the brain and spinal cord were blocked.
After thorough fixation of the brain (~ 3 days or until it sank in the fixative solution), it was cut in 60-μm sections on a freezing microtome. Sections were stored in 0.1 M PBS and refrigerated. For processing, tissue was split into 5 serially-ordered groups for: (1) Nissl staining, (2) BDA processing, (3) fluorescence-only, and two back-up sets. Nickel-enhanced BDA processing was performed as previously described (Hoffer et al. 2005). Briefly, sections were agitated in 0.3% H 2 0 2 and then washed with 0.3% Triton X-100 (pH = 7.4) in 0.1 M PBS before a 2-h incubation in an activated avidin-biotin horseradish peroxidase solution (Vector Novocastra Laboratories, Burlingame, CA). Sections were rinsed in 0.1 M PB and then incubated in a solution of 0.05% diaminobenzidine (DAB), 0.0005% H 2 O 2 , and 0.04% NiCl2 in 0.1 M Tris buffer (pH = 7.1) for 9-12 min. The reaction was terminated with washes in a 0.1 M PB solution. After processing, sections were mounted on gel-coated glass slides, dried overnight, then dehydrated with ethanol, defatted with xylene and finally cover-slipped with Cyto-seal.

Anatomical analysis.
Tracer labeling was analyzed with an Olympus BH-2 microscope fitted with 1 ×, 2 ×, 4 ×, 10 ×, and 20 × objectives. BDA and Nissl stains were observed in brightfield, whereas, FR and AF fluorescence was imaged using a combined fluorescein isothiocyanate/tetrarhodamine isothiocyanate (FITC/ TRITC) filter cube (51004v2, Chroma Technology, Rockingham, VT). Labeled neuronal soma and axonal varicosities of terminal arborizations were digitally reconstructed with respect to anatomical landmarks using a computer controlled AccuStage reconstruction system (St Paul, MN) and stored to hard drive for subsequent analysis. For anterogradely labeled fibers, synaptic boutons were plotted under 20 × magnification, as these structures represent putative synapses (Kincaid and Wilson 1996).
To quantify the convergence of projections from M1, S1, and S2 cortices, the plotted reconstructions of terminal labeling were analyzed using custom software (Hoffer et al. 2005), which imposed a grid of 50 μm × 50 μm squares onto the sections. This bin size was chosen to allow comparison to previous rat data obtained in our lab (Alloway et al. 2000;Hoffer and Alloway 2001;Hoffer et al. 2003Hoffer et al. , 2005Leergaard et al. 2004). For each tracer, bins containing at least two boutons were counted as labeled and colored appropriately (BDA-blue, FR-red, AF, green). Bins labeled by two tracers were colored white so that tracer overlap is expressed as a percentage of all colored bins (i.e., red, green blue, and white) that were labeled as white.
Digital photomicrographs of tracer-labeling and Nissl staining were acquired by a Cool Snap HQ CCD digital camera (Retiga) mounted on the BH-2 microscope. Images were stored as TIF files and imported into our graphic design software (Deneba Systems, Canvas X, Miami, FL) to generate figures. Quantification of tracer labeling was analyzed in Microsoft Excel, with statistics and graphing performed using Graphpad Prism. All values are reported as mean ± SEM.

Results
Neuronal tracers were administered to ten cats (Charles River). Cases were included in the final analysis only if the histological examination confirmed that all three tracer injections spanned both the supragranular and infragranular layers of cortex because virtually all cortical layers send projections to one or more of the subcortical regions analyzed in the present study (Royce 1982;Levesque et al. 1996;Mao et al. 2011). Furthermore, we only analyzed cases in which transported tracer labeling was apparent in all subcortical regions of interest: striatum, claustrum, thalamus, and the basal pontine nuclei. Five of the experiments satisfied these criteria and were included in the final analysis (see Table 1 for summary of injections).
Representative photomicrographs of tracer injections from case TI-8, shown in Fig. 1, depict biotinylated dextran amine (BDA) injected into the forepaw region of primary motor cortex (M1; Fig. 1C, C'), Fluororuby (FR) injected into the forepaw representation of primary somatosensory cortex (S1; Fig. 1D, D'), and Alexa-fluor 488 (AF) injected into the forepaw representation of secondary somatosensory cortex (S2; Fig. 1E, E'). All of these sites were identified physiologically, either via intracranial microstimulation (ICMS) mapping of M1 cortex or receptive field mapping of the S1 and S2 cortical areas, which are illustrated in Fig. 1B. Additionally, the Nissl-stained sections shown in Fig. 1 show hallmarks of motor and sensory cortex. In Fig. 1C, Nissl staining of the M1 injection site reveals characteristics of agranular cortex (i.e. motor cortex), which lacks a layer 4, but shows an expanded layer 5 with large cell bodies that likely represent corticofugal projections to the spinal cord and other subcortical structures. In contrast, Nissl staining in Fig. 1D, E of S1 and S2 reveals a prominent granular layer 4, which is characteristic of sensory cortex. The injection sites in each of these regions span all layers of cortex, including layers 5 and 6, which represent the output layers to the subcortical regions of interest. The other four cases included in the final analysis showed similar injection characteristics.
In addition to the physiologic mapping and histological verification of the injection sites, we also examined tracer labeling in thalamus to confirm the functional placement of the tracer injections. Figure 2 shows representative tracer labeling in the thalamus from the injection sites shown in Fig. 1 for case TI-8. As demonstrated by the digital reconstruction in Fig. 2B and the photomicrographs in Fig. 2D, E, anterogradely labeled terminals and retrogradely labeled neuronal cell bodies were observed in the ventrobasal complex following the S1 and S2 tracer injections. Tracer labeling also appeared in the ventromedial (VM) nucleus, a motor related region of thalamus, following tracer injections in M1 cortex. The presence of anterogradely labeled terminals further indicates that our tracer injections infiltrated layers 5 and 6 as these layers represent the origin of most corticothalamic projection neurons. Interestingly, the labeling in the thalamus was markedly segregated, with the bulk of labeling from each tracer injection forming its own topographic domain. Segregated thalamic domains were also observed in another representative case (Fig. 3 for case TI-9 shown below), in which BDA was injected in M1, AF in S1, and FR in S2.

Cortico-striatal projections
Previous studies in rodents and primates, have reported that all sensorimotor cortical areas send topographically organized projections to the striatum (Aoki et al. 2019a, b;Brodal 1978;Dudman and Gerfen 2015;Flaherty and Graybiel 1991Hintiryan et al. 2016;Hoover and Strick 1993;Hunnicutt et al. 2016;Kunzle 1975Kunzle , 1977Oh et al. 2014), Consistent with these previous findings, we observed tracer labeling from M1, S1, and S2 throughout the putamen and the dorsolateral caudate. The digital reconstruction shown in Fig. 4 from case TI-8 reveals dense intermingling of S1, S2, and M1 projections to a band of neuropil along the dorsolateral edge of the caudate that stretched 3 mm along the anterior-posterior axis. Though we did not counter-stain to reveal the location of the striosome (patch) compartments, we did observe ovoid regions in the middle of areas of dense labeling devoid of labeled terminals (see Fig. 4, bottom left section at − 3.00 mm from bregma), which may correspond to a lack of labeling in the striosomal regions. This pattern of sensorimotor cortico-striatal projections targeting the striatal matrix would be consistent with previous work in rodents, cats and primates (Flaherty and Graybiel 1993;Ragsdale and Graybiel 1981;Smith et al. 2016). Overlap analysis of the tracer labeling in a grid of 50-μm 2 bins revealed strong overlap of tracer labeling from M1 and S1 injections (shown as white bins in Fig. 5A), whereas, M1-S2 and S1-S2 overlap was minimal. In this representative case, cortico-striatal labeling from the AF injection in S2 was relatively caudal and noticeably weaker than the labeling from the M1 and S1 injections, which could account for the low overlap with projections from S2 cortex. In another case, however, we varied which tracers were injected in the cortical areas, and we still observed minimal M1-S2 and S1-S2 overlap. As shown below in Fig. 3, in case TI-9 where FR was injected 1 3 in S2, we found that the labeled projections from S2 cortex terminated in zones that were interdigitated with the overlapping terminal labeling resulting from the M1 and S1 tracer injections.

Cortico-claustral projections
Analysis of the tracer labeling in the claustrum revealed anterogradely labeled terminals and retrogradely labeled neuronal cell bodies throughout a 1 mm anterior-posterior extent of the dorsolateral claustrum (shown for case TI-8 in Fig. 6A). Like the striatum, we observed marked overlap between M1 and S1 projections, as shown in the tracer overlap analysis in Fig. 6B-D. Projections from S2 terminated more medially in an adjacent region to the area of M1-S1 terminals and did not overlap with any of the corticoclaustral projections from S1 and M1. A similar pattern can be seen in the digital reconstruction from case TI-9 shown in Fig. 3A.

Cortico-pontine projections
Analysis of the tracer labeling in the pontine nuclei revealed anterogradely labeled terminals throughout a ~ 3 mm anterior-posterior extent of the structure, positioned on both the Fig. 1 Representative example of cortical mapping and tracer injections from case TI-8. a Diagram illustrating location of tracer injections into the forepaw representations of primary motor cortex (M1), primary somatosensory cortex (S1), and secondary somatosensory cortex (S2). b Diagrams of cortical mapping results, showing the forelimb/forepaw movements from M1 intracortical microstimulation (ICMS) as well as receptive fields in S1 and S2 from case TI-8. Colors correspond to the receptive fields at different cortical sites within S1 and S2. c Nissl-stained coronal section showing location of M1 forepaw in TI-8. Inset shows digital reconstruction of entire coronal section and location of injection sites of biotinylated dextran amine (BDA; injection shown in blue). c' Brightfield photomicrograph showing BDA injection on the adjacent serial section processed with DAB to visualize the tracer (see Methods). d,d' Nissl d and fluorescent d' photomicrographs showing the location of Fluororuby (FR) injections into S1 forepaw. e, e' Photomicrographs of Alexafluor 488 (AF) injections into the S2 forepaw. Coordinates relative to bregma. Scale bars: 1 mm in c medial and lateral sides of the descending pyramidal tract white matter (shown for case TI-8 in Fig. 7A). As in the striatum, we observed substantial overlap between M1 and S1 projections, as shown in the tracer overlap analysis in Fig. 7B-D. In contrast to the striatum, however, we observed stronger input from S2 cortex, which showed moderate overlap with both M1 and S1 projections.

Quantitative analysis
The amount of anterograde tracer labeling can vary due to differences in injection volume, tracer efficiency and detectability. To determine whether these factors could affect our overlap analysis we performed statistical analysis on the total numbers of terminals counted across all animals used in the final analysis (n = 5). The total number of counts summed across all subcortical structures (e.g. striatum, claustrum, thalamus, and pons) was found to vary for each tracer (BDA: 11,517 ± 3534; FR: 22,057 ± 7841; AF: 2625 ± 880). However, though some tracers produced more labeled terminals than others (e.g. FR > AF), statistical analysis did not reveal any significant difference in the total number of labeled varicosities across different tracers (one-way ANOVA, F = 3.80, p = 0.053). These results suggest that differences in the amount or efficiency of the different tracers was likely not a significant factor in our overlap analysis. Furthermore, we varied which tracer was injected into what cortical site (see Table 1), further limiting the effect of tracer efficiency on the results of our overlap analysis. To quantify this measurement, we also calculated the total bouton counts produced based on cortical injection (Fig. 8a, M1: 11,517 ± 3534; S1: 19,910 ± 8598; S2: 4772 ± 2623) but found no significant differences (Kolmogorov Smirnoff test for normality, p > 0.10; one-way ANOVA, F = 1.85, p = 0.20). This second analysis further demonstrates that tracer efficiency was not a significant factor in our overlap analysis. The next quantitative analysis we performed was to identify the relative amount of labeling in each subcortical region of interest from each cortical area injected (i.e. M1, S1, and S2). This distribution was normalized for each tracer within each animal then, percentages were averaged across subjects, thus making these measurements insensitive to tracer efficiency issues. As shown in Fig. 8B, this analysis revealed significant differences in the distribution of projections from these cortical areas to these subcortical structures (2-way ANOVA, interaction F = 43.24, p < 0.0001). Post-hoc Bonferroni comparisons showed M1 projections to striatum were significantly stronger than those from S1 (t = 8.11, p < 0.001) and S2 (t = 8.93, p < 0.001), whereas, in thalamus M1 projections were significantly weaker than those from S1 (t = 9.96, p < 0.001) and S2 (t = 11.47, p < 0.001).
Analysis of tracer overlap using 50 μm 2 bins revealed significant differences in the overlap of labeled projections from M1, S1, and S2 (2-way ANOVA, interaction F = 4.00, p < 0.01). Post-hoc Bonferroni tests revealed that M1-S1 overlap was significantly greater than M1-S2 or S1-S2 overlap in both the striatum (t = 5.43, p < 0.001 and t = 5.38, p < 0.001; respectively) and in the pons (t = 3.44, p < 0.05 and t = 2.70, n.s.). We then compared these data from felines with those from our previous studies in rodents. The injection scheme showing location of anterograde tracer injections in rodents from our previous study (Leergaard et al. 2004) is shown in Fig. 9C. Quantitative analysis of the overlap analysis (bin size 50 μm 2 ) in the rodent tracing studies shows more S1-S2 overlap than M1-S1, that was significant in the pons (t = 4.87, p < 0.001). These findings are in opposition to the findings in felines, where S1-M1 overlap was significantly greater, indicating cross-species differences in sensorimotor convergence.

Discussion
In this study we analyzed the convergence of sensory and motor cortical projections in subcortical structures; specifically, the striatum, claustrum, thalamus and pons. Using a triple anterograde tracer approach in cats, we injected the forepaw representation in M1, S1, and S2 cortices (which are equidistant) and analyzed the distribution and overlap of anterogradely labeled terminals. These new findings were subsequently compared to our previous tracing experiments in rodents (Hoffer and Alloway 2001;Leergaard et al. 2004), to determine how sensorimotor cortical projections to the striatum, claustrum, thalamus and pons vary across evolutionarily distinct lineages (Euarchontoglires and Laurasiatheria).
These experiments yielded several key findings. First, in cats we observed the greatest tracer overlap between M1 and S1 terminals in the striatum and pons, whereas, in rodents the overlap in these target regions was stronger for the projections from S1 and S2. Second, whereas, S1 and S2 cortices in the cat sent the majority of their projections to the thalamus (~ 70%), M1 projections to the thalamus were weak (< 20%); the most abundant M1 projections terminated in the striatum (~ 60%). This finding contrasts with our previous rodent data in which thalamus was consistently the strongest target of M1 ). Third, in the cat claustrum we observed clear input from both M1 and S1 cortex, in contrast to our previous observations in rats where we observed weak input from M1 forepaw and no input from S1 (Smith and Alloway 2010;Smith et al. 2012). Together these findings suggest evolutionary differences in the principles guiding the distribution and integration of sensorimotor cortical projections to the subcortex.

Relative contribution of sensorimotor projections to subcortical structures
One of the primary findings of the present study, shown in Fig. 8B, is that while M1 contributed most of its projections to the striatum, S1 and S2 sent the majority of their inputs to Fig. 4 Representative digital reconstruction of terminal labeling in striatum (caudate nucleus) of case TI-8. Colors correspond to different tracers (BDA, blue; FR, red; AF, green). Note the dense inter-mingling of the M1 (blue) and S1 (red) projections in the dorsolateral region. Coordinates relative to bregma. Scale bar: 1 mm the thalamus with equivalently minor contributions to striatum and pons. Furthermore, M1, S1 and S2 projections were almost equally represented within the pons. Finally, all three cortical areas sent relatively few projections to the claustrum compared to striatum, thalamus or pons. These findings may give insight into the relative neural processes that occur in these subcortical structures to guide behavior.
In the case of the claustrum, our findings show relatively few connections with sensorimotor cortex compared to the other subcortical structures. Our previous tracing studies in rodents yielded similar findings Smith and Alloway 2010), Specifically, in rodents we observed that the M1 forepaw region had very sparse connections to the dorsal region of the ipsilateral claustrum, and no input from S1 cortex was detectable (Smith et al. 2012). This suggests a limited role of forelimb-related somato-motor inputs in dictating the functional output of the claustrum as discussed in more detail below.
Interestingly, motor cortex in felines contributes very little input to the thalamus compared to S1 and S2 cortices, which sent the overwhelming majority of their inputs to VPL. Cortico-thalamic feedback is thought to play a role in attention by enhancing sensory signals from the periphery and optimizing sensory acuity through receptive field modulation (Alitto and Usrey 2003;Alloway 2008;Briggs and Usrey 2008). Our data suggest that corticothalamic projections play a stronger role in processing sensory information than motor information, at least in felines. Interestingly, in rodents, M1 forepaw projections targeted thalamus (45%) more strongly than striatum (30%) (Alloway et al. 2009), Fig. 5 Representative overlap analysis of anterograde tracer labeling in striatum (caudate nucleus) from case TI-8. Colors correspond to different tracers (BDA, blue; FR, red; AF, green). White bins represent overlap. Bin size = 50 μm 2 . Each panel shows only two tracers, with M1 (blue) and S1 (red) in panel a, M1 (blue) and S2 (green) in b, and S1 (red) and S2 (green) in panel c. Percentages represent proportion of total bins containing overlap. Note the moderate overlap of M1 and S1 projections in the dorsolateral region shown in panel a. Scale bars: 1 mm in a providing evidence for species specific patterns in the distribution of cortico-subcortical projections.
The cortico-cerebellar system is thought to be involved in motor planning as well as monitoring and adjusting motor output based on sensory feedback (Chen et al. 2016;Doyon et al. 2003;Gao et al. 2018;Ramnani 2006), whereas, striatum is thought to be involved in action selection (Friend and Kravitz 2014;Klaus et al. 2019;Redgrave et al. 1999). By virtue of its dominant inputs to the striatum, our results indicate M1 plays a stronger role in the action selection process than the action refinement process (though it is involved in both). Furthermore, our results reveal substantially more inputs from M1 to the striatum than from S1 and S2, suggesting that motor cortex plays a more prominent role in action selection than sensory cortex. With regards to the pons, as indicated by the largely equivalent inputs from M1, S1, and S2, our data indicate an equally important role for motor and sensory information in sensory-guided correction of motor behavior via the cortico-cerebellar system.

Sensorimotor cortico-claustral projections
The claustrum is a unique subcortical structure whose function remains undefined (for review see Jackson et al. 2020). Many studies have sought to identify its function based on its anatomical connectivity, which is largely organized into reciprocal connections with the neocortex (Smith et al. 2019a;Jackson et al. 2020). One striking finding from our current studies is the minimal representation of sensorimotor inputs to the claustrum, compared to striatum, thalamus and Fig. 6 Representative digital reconstruction of tracer labeling and overlap analysis of anterogradely labeled terminals in claustrum of case TI-8. a Digital reconstruction of tracer labeling throughout anterior-posterior extent of claustrum. Colors correspond to different tracers (BDA, blue; FR, red; AF, green). Small dots represent terminal synaptic boutons, large dots represent neuronal cell bodies. Note the strong presence of S1 projections. Coordinates relative to bregma. Scale bar: 1 mm. b-d Overlap analysis of anterogradely labeled ter-minals. Colors correspond to different tracers (BDA, blue; FR, red; AF, green). White bins represent overlap. Bin size = 50 μm 2 . Each panel shows only two tracers, with M1 (blue) and S1 (red) in panel b, M1 (blue) and S2 (green) in c, and S1 (red) and S2 (green) in panel d. Percentages represent proportion of total bins containing overlap. Note the moderate overlap of M1 and S1 projections shown in panel b. Scale bars: 1 mm in a 1 3 pons. Consistent with other species, these projections were restricted to the dorsolateral region (Smith et al. 2019a), which represents only a small region of the total claustrum. Previous tracing data in cats has suggested that auditory and visual inputs are more prominent in cats (LeVay and Sherk 1981;Macchi et al. 1981). More recent findings in rodents have found the strongest inputs to the claustrum arise from limbic structures like the amygdala, MD thalamus as well as prefrontal cortex (Jackson et al. 2020). Furthermore, the claustrum in rodents has been shown to be a critical node in the rodent homolog of the salience network ). This potential role in sensory salience is thought to Fig. 7 Representative digital reconstruction of tracer labeling and overlap analysis of anterogradely labeled terminals in the pontine nuclei of case TI-8. a Digital reconstruction of tracer labeling throughout anterior-posterior extent of the pons. Colors correspond to different tracers (BDA, blue; FR, red; AF, green). Small dots represent terminal synaptic boutons. Note the strong intermingling of M1, S1, and S2 projections. Coordinates relative to bregma. b-d Overlap analysis of anterogradely labeled terminals. Colors correspond to different tracers (BDA, blue; FR, red; AF, green). White bins represent overlap. Bin size = 50 μm 2 . Each panel shows only two tracers, with M1 (blue) and S1 (red) in panel b, M1 (blue) and S2 (green) in c, and S1 (red) and S2 (green) in panel d. Percentages represent proportion of total bins containing overlap. Note the moderate overlap of all tracers. Scale bar: 1 mm in a 1 3 be funneled towards motor cortical areas to direct attention by guiding behavioral output towards the most relevant sensory stimuli. The limited role of somato-motor processing of the forelimb in the cat claustrum, with more expansive auditory and visual inputs, would fit with a role in predation and tracking salient visual and auditory stimuli related to prey. However, additional studies directly comparing these differences is needed to make any definitive conclusions.
In our previous studies mapping the connections of sensorimotor cortex with the claustrum in rodents Alloway 2010, 2014;Smith et al. 2012), we observed weak projections from the forepaw representation in M1 cortex to the claustrum and no projections from any region of S1 cortex. In contrast, our current findings in cats has clearly revealed inputs from both S1 and M1 cortices, which terminated in the same overlapping region of the dorsal claustrum, which is the sensorimotor domain in the claustrum of rats and monkeys as well (Smith et al. 2019a;Pearson et al. 1982). Interestingly, we saw almost no connections between S2 cortex and claustrum in cat, which aligns with an earlier finding observed while mapping claustro-cortical connections in felines using retrograde tracers (Macchi et al. 1981).
These findings further align with other tracing studies across different primate species that have shown differences in cortico-claustral connectivity (Sherk 1986). Specifically, in some primates the primary visual cortex has strong projections to the claustrum, whereas, it is absent in other species. Indeed, we have also shown that V1 has no projections to the claustrum in rodents (Smith and Alloway 2014). Interestingly, LeVay and Sherk found that in cats, numerous visual areas, including area 17 contributed profuse projections to the claustrum (LeVay and Sherk 1981). Thus, no clear pattern has yet emerged as to why some sensory cortical areas project to the claustrum and others do not. But, though the functional meaning of these differences is unclear, further systematic study of the cortico-claustral connectome across species, and their functional relation to behavioral specializations within those species, will undoubtedly reveal crucial inferences as to the function of the claustrum.
However, the functional significance of integrating S1 and M1 information, but not the inputs from S2 is unclear. Additionally, this preference for S1-M1 integration is contrary to our previous findings in rodents (Leergaard et al. 2004), which suggest species-specific patterns of cortico-subcortical sensorimotor integration.
One possibility for explaining this finding could be if S2 projections are processed in parallel to S1 and thus target different subregions within M1 and subcortical structures in cats. In our previous tracing studies in rodents (Hoffer and Alloway 2001;Hoffer et al. 2003Hoffer et al. , 2005Leergaard Fig. 9 Quantification of anterograde tracer overlap from sensorimotor cortex injections in cat and rat. a Injection scheme from this study employing a triple tracer paradigm injecting M1, S1, and S2 in the same animal with three different tracers. b Summary of overlap analysis using 50 μm 2 bins, with a threshold requiring two counts per bin revealed a significant effect (2-way ANOVA, interaction F = 4.00, p < 0.01). Post-hoc Bonferroni tests revealed that M1-S1 overlap was significantly greater than M1-S2 and S1-S2 in the striatum (t = 5.43, p < 0.001 and t = 5.38, p < 0.001; respectively) and in the pons (t = 3.44, p < 0.05 and t = 2.70, n.s.). c Injection scheme showing location of anterograde tracer injections in rodents for previous studies (see Table 1 in Leergaard et al. 2004 for information about injections and Table 2 for overlap data). d Graph summarizing findings from overlap analysis (bin size 50 μm 2 ) in rodent tracing studies shows more S1-S2 overlap than M1-S1, that was significant in the pons (t = 4.87, p < 0.001) Smith and Alloway 2013), homotypic S1 and S2 whisker representations were found to have overlapping projections to M1 as well as to subcortical structures such as striatum and pons. While it has been shown that S2 projects to M1 in cats (Mori 1997), no study has characterized whether S1 and S2 projections overlap in M1 of cats. In the more elaborated sensory cortex of the cat, parallel processing streams from each sensory area may target the same general region of a downstream structure, while targeting separate neurons within that region, accounting for our limited ability to detect S1-S2 convergence compared to S1-M1 projections.
A related observation from our current dataset reveals that S1 and S2 forepaw projections were distributed into three distinct clusters in the ventrobasal thalamus (specifically VPL). As shown in Figs. 2 and 3, a core region of VPL contained almost exclusively S1 labeling, while a shell region located more dorsally in VPL contained both S1 and S2 labeled terminal inputs and retrogradely labeled neurons. A third area on the very dorsal edge of VPL contained almost exclusively S2 connections. These findings nicely align with previous dual-retrograde tracing experiments from S1 and S2 forelimb in cats (Spreafico et al. 1981). When analyzed with high resolution using our overlap analysis, these thalamic projections from S1 and S2 yielded very weak overlap despite occupying similar regional domains in VPL and having overlapping receptive fields of the forepaw. In contrast, S1-M1 overlap in thalamus was equally as weak as S1-S2, despite significantly robust S1-M1 overlap in striatum and pons. Together these data indicate that (1) S1-M1 integration is a critical feature within the feline striatum and pons, though not in thalamus and (2) S1 and S2 may represent parallel processing streams in each of these subcortical structures.

Technical limitations
In our current study we observed different patterns in sensorimotor convergence which varied from rodents. Primarily, in cats, we saw greater S1-M1 convergence, similar to what has been observed in the putamen for primates (Flaherty and Graybiel 1993), whereas, in rodents we observed greater S1-S2 convergence. Our findings here have two main caveats that limit making overly definitive interpretations about their functional significance. First, whereas, we injected the forepaw representation in cats, our previous tracing studies largely targeted the whisker representations in rodents. These body parts have very different behavioral roles, and thus may have evolved unique neural circuit organization to facilitate their specific needs. Second, the distance between these cortical representations varies between the two species. In rodents, the whisker regions of S1 and S2 are spatially adjacent, whereas, M1 is many millimeters anterior. Thus, the higher S1-S2 convergence in striatum and pons observed in rodents may be due to a rule of "cortical proximity" (e.g. cortical areas that are closer together are more likely to have overlapping projections in subcortical structures compared to cortical areas that are far apart). In contrast to our rodent experiments on cortical whisker representations, we specifically chose the forepaw representation in cats, as the M1, S1 and S2 representations are equidistant, to mitigate effects of cortical proximity. In rodents, however, the M1 and S1 forepaw representations are immediately adjacent with S2 significantly separated, which inherently prevents an equivalent comparison of rodents to cats. Nonetheless, while more mapping studies for different body regions in other species could reveal additional support for variations in cortico-subcortico convergence patterns across different evolutionary lineages, our results indicate distinct differences in the rules that govern sensorimotor corticofugal projection patterns in rodents and cats.

Technical considerations of tracer localization to cortical areas in cat
There is a remote possibility that some of the tracer injected into S1 or M1 cortex may have extended into the edges of cortical area 3a, but the amount of labeling would have been minimal. In cats, area 3a occupies a 2-mm wide strip of cortex located between cortical areas 3b and 4 (Dykes et al. 1980). Studies have reported that area 3a is unresponsive to cutaneous stimulation but can be activated by vagal nerve stimulation or by vigorous stimulation of deep receptors (Felleman et al. 1983;Ito and Craig 2003). Although deep receptor stimulation was not tested in our experiments, efforts were made to identify S1 injection sites in the medial bank of the coronal sulcus that responded to low-threshold cutaneous stimulation of individual digits (Felleman et al. 1983). Similarly, tracer injections in M1 were identified by locating sites in the caudal bank of the cruciate sulcus in which small amounts of electrical current, typically 30 μA or less, were sufficient to evoke muscular twitches in the forepaw. Although it is conceivable that some tracer may have diffused into areas 3a from the S1 or M1 injection sites, it should be noted that these injections sites were widely separated from each other (see Fig. 1a) and it seems likely that most, if not all, of area 3a was located between the S1 and M1 tracer injections.
With regards to the appropriate targeting of our S2 injection, several facts indicate that our S2 tracer injections did not involve other somatosensory representations. Consistent with prior studies of the somatotopic organization of S2 (Burton et al. 1982;Alloway and Burton 1985a, b) our mapping procedures identified similar topographically organized regions in the rostral ectosylvian gyrus that responded to low-threshold forepaw stimulation. Following this preliminary mapping, injections were placed at digitsensitive sites that were separated from the surrounding suprasylvian or anterior ectosylvian sulci, which are known to contain other somesthetic representations (Burton et al. 1982). In fact, our histology confirmed that our racer deposits, which were approximately 1.0 mm in diameter, were located near the center of the ectosylvian gyrus (see Fig. 1E, E'), consistent with the known location of S2 in cats.