Convergence of inputs from the basal ganglia with layer 5 of motor cortex and cerebellum in mouse motor thalamus

A key to motor control is the motor thalamus, where several inputs converge. One excitatory input originates from layer 5 of primary motor cortex (M1L5), while another arises from the deep cerebellar nuclei (Cb). M1L5 terminals distribute throughout the motor thalamus and overlap with GABAergic inputs from the basal ganglia output nuclei, the internal segment of the globus pallidus (GPi) and substantia nigra pars reticulata (SNr). In contrast, it is thought that Cb and basal ganglia inputs are segregated. Therefore, we hypothesized that one potential function of the GABAergic inputs from basal ganglia is to selectively inhibit, or gate, excitatory signals from M1L5 in the motor thalamus. Here, we tested this possibility and determined the circuit organization of mouse (both sexes) motor thalamus using an optogenetic strategy in acute slices. First, we demonstrated the presence of a feedforward transthalamic pathway from M1L5 through motor thalamus. Importantly, we discovered that GABAergic inputs from the GPi and SNr converge onto single motor thalamic cells with excitatory synapses from M1L5 and, unexpectedly, Cb as well. We interpret these results to indicate that a role of the basal ganglia is to gate the thalamic transmission of M1L5 and Cb information to cortex.


Introduction
The cerebral cortex, cerebellum, and basal ganglia work together to control motor behavior.Projections from these areas converge in the motor thalamus, a conglomerate of the ventral anterior (VA) and ventral lateral (VL) nuclei, often grouped together in rodents as VA/VL, as well as the ventral medial (VM) nucleus.
The GABAergic basal ganglia output nuclei, in particular the internal segment of the globus pallidus (GPi) * and substantia nigra pars reticulata (SNr), also project to the VA/VL (Sommer, 2003;Lanciego et al., 2012).Therefore, an intriguing possibility is that the basal ganglia intersect excitatory pathways through motor thalamus, providing a substrate by which the basal ganglia can gate information flow to cortex.However, a prevalent notion is that cerebellar and basal ganglia inputs are functionally segregated in the motor thalamus (Bosch-Bouju et al., 2013).Moreover, the traditional view of the basal ganglia projection to motor thalamus situates it within a cortical-basal gangliacorticothalamic loop, leaving open the possibility that direct L5 inputs innervate a subpopulation of motor thalamic relays that do not participate in this classical loop.
Here, we addressed these three outstanding questions.First, we confirm the presence of a feedforward transthalamic pathway through the VA/VL segment of mouse motor thalamus.Second, we employed a dual opsin approach and determined that inhibitory afferents from the GPi and SNr to motor thalamus frequently converge on cells receiving excitatory driving inputs from L5 of motor cortex, demonstrating the transthalamic pathway may be gated by the basal ganglia.Finally, we also find functional convergence of inputs to VA/VL from the basal ganglia and cerebellum, suggesting these circuits are not wholly segregated.Taken together, our data highlight a role for the basal ganglia in gating the relay of driving excitatory inputs through motor thalamus, determining which information streams reach cortex at any one time.
Higher order sensory thalamic nuclei, like the posterior medial nucleus or pulvinar (also known as the lateral posterior nucleus in rodents), receive robust inputs from cortical L5 and a fraction of these constitute feedforward or feedback transthalamic pathways (Theyel et al., 2010;Mo and Sherman, 2018;Blot et al., 2021;Miller-Hansen and Sherman, 2022).Physiological evidence in VM nucleus of the motor thalamic complex indicates cortico-thalamo-cortical interactions are mainly organized in reciprocal loops, demonstrating only a very limited feedforward transthalamic component from primary to secondary (i.e., anterolateral) motor cortex (Guo et al., 2018).Therefore, we investigated whether a feedforward transthalamic pathway existed in the VA/VL portion of the motor thalamus by probing for optogenetic responses from L5 of primary motor cortex (M1L5) in VA/VL relays projecting to secondary motor cortex (M2) (Figure 1A).Specifically, we injected fluororuby into all layers of M2, recorded from retrogradely-labeled neurons in VA/VL, and tested each for an M1L5 input driven by Cre-dependent expression of ChR2 in Rbp4-Cre mice (Figure 1B-C).We found that from 25 fluororuby-positive cells across 4 mice (Figure 1D), 4 cells demonstrated a clear photostimulation-dependent excitatory response (Figure 1E-G).
While it is known that the motor cortex innervates the motor thalamus (Figure 1 and (Kita and Kita, 2012;Economo et al., 2018;Winnubst et al., 2019;Prasad et al., 2020) previous anatomical studies demonstrate two zones with minimal overlap in the motor thalamus, one innervated by the basal ganglia, and the other, by the deep cerebellar nuclei (Cb) (Anderson and DeVito, 1987;Sakai et al., 1996;Kuramoto et al., 2011).Therefore, we next sought to directly compare the distribution of projections from the GPi, the Cb, and M1L5 in the same mice by employing three color anterograde labeling.We delivered red, green, and blue fluorophores via viral injection into the Cb, GPi, and M1, respectively (Extended Data Figure 2-1A-C).We used Rbp4-Cre mice, and the injection of the label into M1 was Cre-dependent, thus limiting the label to cortical L5 neurons.We then analyzed their terminal distribution of these three afferent inputs in the motor thalamus (Figure 2A-C).
To generate an input map of motor thalamus, we averaged the fluorescence arising from each input across 3 animals and overlayed them onto the coronal mouse atlas (Paxinos and Franklin, 2007) at multiple rostro-caudal planes (Figure 2C).This map corroborates previous findings using anterograde labeling in rodents in several ways.First, we find that the Cb afferents are robust throughout much of motor thalamus but particularly concentrated in the lateral portions of rostral VA/VL, as well as the medial and dorsomedial aspects of caudal VA/VL; Cb afferents also sparsely innervate the rostral VM.Further, we observe that the GPi projection to motor thalamus, at least in comparison to the Cb inputs, is more limited, concentrating in the middle of rostral VA/VL with a proscribed patch in the ventrolateral aspect of VA/VL.In contrast to both of these inputs, we find that the input from M1L5 is diffuse throughout VA/VL, appearing as coursing axons in large swaths of the region, though terminals do concentrate in VM, ventral VA/VL, and the medial boundary of VA/VL.Rostral to the sections examined, there was little representation from any input, while in more caudal sections (beyond -1.1AP), GPi terminals are virtually absent, M1L5 terminals are found primarily in VM, and Cb terminals target ventral VPL and VL (Extended Data Figure 2-1D, E).
Importantly, the simultaneous labeling of all three inputs in the same animal(s) reveals a degree of overlap between inputs from GPi and Cb, particularly in the rostral VA/VL and ventrolateral VA/VL in more caudal (approximately -1.0mm AP from bregma) sections (Figure 2C).Terminal zones from M1L5 and Cb overlapped in central and ventral aspects of the VA/VL, while overlap is also likely in the more dorsal aspects where M1L5 axons diffusely project (Figure 2C).Generally, Cb inputs overlapped with those from GPi laterally and dorsally, while M1L5 inputs tended to overlap with GPi medially.A quantitative analysis of the convergence between each set of inputs demonstrates that approximately 10% of the M1L5 and GPi input zones are overlapping (Figure 2D).

GABAergic inputs from GPi converge with glutamatergic inputs from both M1L5 and Cb (Figure 3)
Having defined the regions of overlap between GPi and M1L5, we next tested whether these inputs converge onto single thalamic cells using acute slice physiology in Rbp4 mice (Figure 3A).Our data for these experiments include 62 cells from 14 animals, though we focus on cells receiving at least one input (42/62 cells).We transduced GPi neurons with ChR2-GFP (or, in some animals, Dlx-ChR2 for inhibitory neuron selective expression) and transduced M1L5 neurons with cre-dependent ChR2-mCherry (Figure 3A,     B).Since GPi outputs are GABAergic and M1L5, glutamatergic, we readily identified their evoked activity on postsynaptic neurons recorded in the motor thalamus using a lowchloride internal pipette solution (5mM Cl -total) and clamping the cell at distinct holding potentials.Specifically, we recorded at a holding potential of -60mV to accentuate excitatory postsynaptic currents (EPSCs) from M1L5 and at -40mV for inhibitory postsynaptic currents (IPSCs) from GPi (Figure 3C).To confirm that the inputs were in fact GABAergic or glutamatergic in a subset of recorded cells, we pharmacologically inhibited GABAA receptors or ionotropic glutamate receptors (iGluRs), respectively.In cells used for representative traces, bath application of gabazine was followed by a washout period to allow recovery of the GABAA response, after which DNQX was applied.In motor thalamic relays receiving only one input, blocking the associated receptor subtype completely abolished the evoked response at all holding potentials (Figure 3C, see GPi or M1L5 only representative cells), while in relay cells receiving both inputs, the nonantagonized receptor system remained functional (Figure 3C, see Both inputs representative cell).
Cataloging the location of all recorded cells revealed that motor thalamic neurons receiving inputs from both GPi and M1L5 were spread throughout the rostral VA/VL but clustered relatively tightly around the distinct GPi terminal patch in more caudal sections (Figure 3E), as expected from our anatomical data (Figure 2).
The zones of motor thalamus innervated by the GPi and Cb are thought to be largely, if not entirely, independent (Bosch-Bouju et al., 2013;Nakamura et al., 2014).
Therefore, we next tested whether excitatory inputs from the Cb also converge with those from the GPi (Figure 3F, G).Our data for these experiments includes 50 cells from 8 animals, though, again, we analyze only those receiving an input (43 cells).
Activations of Cb inputs evoked large EPSCs, often producing an action potential or a burst of action potentials, mirroring a recent report (Schäfer et al., 2021) 3H, I).
Spatial analysis of all recorded cells showed a similar pattern to neurons receiving convergent input from GPi and M1L5-neurons innervated by GPi and Cb were primarily found in the central and ventrolateral aspects of the VA/VL (Figure 3J).
GABAergic inputs from SNr also converge with both M1L5 and Cb (Figures 4 and 5) Next, we tested whether the same organization applied to inputs from the second major output of the basal ganglia direct pathway, the SNr.We performed the same tricolor labeling experiment as above, substituting the GPi for the SNr (Figure 4A and Extended Data Figure 4-1A-C).In general, terminations from the SNr showed a similar organization to those from the GPi, with the important exception that the SNr terminal zone was shifted medially and ventrally compared to the GPi, covering the ventromedial VA/VL and VM in the sections examined (Figure 4B, C).Therefore, SNr terminals demonstrated little overlap with those from the cerebellum but exhibited substantial overlap with M1L5 terminals (Figure 4D).
Accordingly, analysis of 19 cells (35 cells total recorded, 16 with no detectable inputs) from 5 mice in the overlap zone between M1L5 and SNr terminals (Figure 5A, B), demonstrated that the SNr indeed converges with inputs from M1L5 on single cells (5/19 cells, 26.3%) (Figure 4C, D).Of the cells receiving an excitatory input (M1L5), nearly half also had convergent input from SNr (45.5%) (Figure 4D).Surprisingly, analysis of 19 cells (30 cells recorded in total, 11 with no detectable inputs) from 7 mice demonstrates that SNr inputs also converged with excitatory projections from the Cb (6/19 cells, 31.6%)(Figure
To obtain reliable paired-pulse effects, we performed laser stimulation of ChR2positive axons >300µm from the recorded cell rather than at ChR2-positive terminals, since terminal photostimulation produces unreliable paired-pulse effects (Jackman et al., 2014;Mo and Sherman, 2018).We found that both M1L5 and Cb have driver-type synaptic properties.Specifically, these inputs showed synaptic depression at both low (10Hz) and high (40Hz is shown) photostimulation frequencies, an insensitivity to antagonists of metabotropic glutamate receptors (mGluRs), and responses that were completely abolished by blockade of iGluRs (Figure 6A, B).Note that although ChR2(H134R) does not perform optimally beyond 20Hz, our previous data demonstrate optical high-frequency stimulation is sufficient to stimulate an mGluR response under these conditions (Miller-Hansen and Sherman, 2022), which we do not detect here.Compilation of the paired-pulse ratio (2 nd EPSC/1 st EPSC) for each cell tested showed a depressing phenotype for these excitatory inputs to motor thalamus (Figure 6D).
It is also unknown whether the inhibitory input from the GPi activates metabotropic GABAB receptors, one potential source of the type of sustained hyperpolarization required for the characteristic rebound spiking of thalamic relay neurons (Llinás and Jahnsen, 1982).
Therefore, we tested the synaptic properties of the GPi input to motor thalamus and, particularly, whether we could detect the presence of GABAB receptor activation.The GPi input to all cells tested was depressing at both low (10Hz) and high (40Hz) stimulation frequencies, unchanged by GABAB blockade, and completely abolished by gabazine bath application (Figure 6C).Therefore, we conclude that at least in the population of cells tested, GPi inputs to motor thalamus do not activate GABAB receptors.
Cb terminals within or outside the GPi overlap zone are not different in size (Figure 7) Next, we directly tested the notion (Deniau et al., 1992) that Cb terminals in the GPi overlap zone are smaller using dual anterograde labeling and terminal size analysis of the region in ventrolateral VA/VL where Cb and GPi overlap, versus the area immediately surrounding the overlap zone (Figure 7).Both Cb terminal size distribution (Figure 7B) and average size (Figure 7C) were the same between the overlap ("within GPi zone") and non-overlapping ("outside GPi zone") regions (within=2.53±0.1 vs. outside=3.03±0.3µm 2 ; p=0.222 by Mann-Whitney test).However, the number of terminals in a size-matched region outside the GPi overlap zone was greater (within=223.1±37.4 vs. outside=458.8±88.0terminals; *p=0.0317 by Mann-Whitney test) (Figure 7D), supporting the notion that Cb terminals that invade the GPi zone are sparser (Deniau et al., 1992).
Nevertheless, the finding that terminals outside the GPi overlap zone are no different in size to those within it agree with the functional evidence herein (i.e., Figure 5) and elsewhere (Gornati et al., 2018) that cerebellar inputs to motor thalamus are uniformly drivers.

Discussion
Our data provides the first functional evidence for a feedforward transthalamic pathway through the mouse VA/VL.Further, we show that both GPi and SNr efferents converge with those from M1L5 and/or Cb onto single motor thalamic neurons.Note that this is quite different from previous demonstrations of convergence of Cb with layer 6 (L6) of cortex (Schäfer et al., 2021), as it is anticipated that every, or nearly every, thalamic relay receives a modulatory L6 input, whereas L5 projections are more selective and are drivers (Sherman and Guillery, 2013;Usrey and Sherman, 2019).

Experimental provisos
Whereas our data reveals that the GPi and SNr can converge with both the M1L5 and Cb pathways to motor thalamus, only a subset of thalamic neurons receives such convergent input.There are several potential reasons for the seeming sparsity of cells receiving the inputs of interest, particularly regarding M1L5.Several frontal and motor cortical areas project to the motor thalamus, concentrating in distinct subsections (Bosch-Bouju et al., 2013).Therefore, by focusing our attention (i.e., injections) on primary motor cortex, we substantially limited the number of thalamic relays potentially receiving a L5 input.Moreover, we targeted our recordings to the terminal overlap zones between any two sets of inputs.As revealed by our anatomical evidence (Figures 2 and 4), large proportions of motor thalamus are segregated to a single input (e.g., Cb versus SNr), again limiting the connection probability of recorded cells.
Within the overlap zone, however, the proportion of cells we find with convergent inputs likely reflects an underestimate of the actual numbers, because there are several additional reasons for false negatives; that is, failure to detect an existing input.For instance, not all projection neurons from basal ganglia or cortex are likely to be transfected with ChR2, especially since many L5 neurons do not express Cre in the Rbp4 mouse line (Harris et al., 2014).Also, if topographic alignment is necessary to detect such convergence, it may not have been achieved in all cases.As mentioned above, the topography of cortical projections to motor thalamus may be particularly relevant.Given these caveats, absent a specific percentage, we conclude that a sizable, functionally relevant population of neurons within the overlap zones are gated.It should be noted that our conclusions below are focused on this subset of cells receiving convergent inputs.

Both excitatory pathways to motor thalamus are drivers
While previous tracing studies in cats and monkeys (Rouiller et al., 1998(Rouiller et al., , 2003;;Kakei et al., 2001;Kultas-Ilinsky et al., 2003) showed that L5 of motor cortex projects with large terminals to the motor thalamus, histological studies in monkeys and mice called into question whether corticothalamic drivers existed in much of motor thalamus (Kuramoto et al., 2011;Rovó et al., 2012).We found here that the M1L5 inputs to all motor thalamic cells displayed driver characteristics; however, EPSC amplitudes of ChR2-evoked responses from M1L5 were indeed smaller than those from Cb inputs, which often elicited action potentials (see below).One potential reason for this difference is the degree of convergence among L5 corticothalamic inputs, compared to the first order pathway from Cb, onto motor thalamic neurons.That is, whereas the total input from Cb or L5 might be equally strong, that from a single cortical area might represent only part of the driving input.
Prior anatomical data also showed Cb terminals innervating motor thalamic regions co-innervated by the basal ganglia were diffuse and small (Deniau et al., 1992), characteristics of modulators (Sherman, 2007;Sherman and Guillery, 2009).However, we find that all Cb inputs to motor thalamic neurons are drivers, including a fraction of thalamic relays that exhibited smaller evoked EPSCs from Cb.These observations agree with a recent report characterizing Cb inputs across motor thalamic nuclei (Gornati et al., 2018).Furthermore, while our analysis of synaptic terminal size confirmed sparser Cb innervation of the GPi overlap zone (Deniau et al., 1992), we were unable to detect a difference in bouton size between the GPi overlap and non-overlap zones, consistent with inputs in both regions being drivers.

Beyond the classical notion of a cortico-basal ganglia-thalamocortical loop for motor control
Our experiments reveal three related insights regarding the relationship between cortex, basal ganglia and motor thalamus that have important functional implications.
First, previous anatomical data suggest a near complete segregation of inputs from the Cb and basal ganglia to the motor thalamus (Anderson and DeVito, 1987;Deniau et al., 1992;Sakai et al., 1996;Kuramoto et al., 2011) and, while our experiments largely confirm this organization (Figures 2 and 4), we also demonstrate that these pathways in fact converge at the single cell level in a subset of motor thalamus.This finding extends prior transsynaptic labeling studies in non-human primates that highlighted an indirect interface between these systems (Hoshi et al., 2005;Bostan et al., 2010) by revealing a direct substrate for their interaction in mice.It remains unclear, however, if these particular motor thalamic relays are engaged in an aspect of motor function that depends on their receiving these convergent inputs.
Second, the textbook view of the basal ganglia situates it within an information loop from cortex to basal ganglia to thalamus and back to cortex (Kandel et al., 2020), though updates have been made (Haber and Mcfarland, 2001;Lanciego et al., 2012;Goldberg et al., 2013).However, our finding that basal ganglia inputs to motor thalamus impinge on relays receiving excitatory input from both the M1L5 and Cb pathways suggests a previously untested function of the basal ganglia-to gate information flow through motor thalamus.
Motor and frontal cortices are a major source of inputs to, as well as primary targets of, motor thalamic relays (Bosch-Bouju et al., 2013).These thalamocortical connections can be reciprocal (Bosch-Bouju et al., 2013;Collins et al., 2018;Guo et al., 2018) or nonreciprocal (Rouiller et al., 1998;McFarland and Haber, 2002).Indeed, here we directly demonstrate the presence of a feedforward transthalamic pathway from M1, through VA/VL, to M2 (Figure 1).Therefore, in areas of the motor thalamus receiving convergent input from cortical L5 and basal ganglia, some fraction of which participates in transthalamic pathways, it is plausible that the basal ganglia can gate excitatory information between cortical regions (Figure 8A).When basal ganglia outputs are active, specific thalamic relays involved in transthalamic information transfer are inhibited.Since cortical areas have both direct and transthalamic connections organized in parallel (Sherman and Guillery, 2013;Sherman, 2022), disconnecting the transthalamic excitatory stream allows only the direct, corticocortical pathway to persist.In contrast, when thalamic relays are disinhibited due to upstream inhibition of the GPi or SNr, transthalamic transmission flows (Figure 8A).Note that this would also apply to reciprocal cortico-thalamocortical connections (i.e., cortico-thalamocortical loops), which are well-represented in regions of motor thalamus (Guo et al., 2018(Guo et al., , 2019)).Thus, we reason that a potential function of basal ganglia inputs to some fraction of motor thalamic relays is to gate L5 signals, and thereby affect communication between cortical areas (Figure 8A); however, we recognize additional experiments in intact animals are required to test this speculation.Third, our data advance understanding of basal ganglia function by demonstrating a direct projection from L5 of cortex, the starting point of the classical basal ganglia loop (Figure 8B, left), to the endpoint of the loop-motor thalamic relays.In other words, some M1L5 axonal branches bypass the basal ganglia entirely (Figure 8B, right), which extends previous neuroanatomical data across species (Mooney and Konishi, 1991;Farries et al., 2005;Kita and Kita, 2012).The important aspect here is that absent this evidence, one might have concluded that M1L5 innervates a discrete subset of motor thalamic relays that do not receive a basal ganglia input, thereby preserving the traditional cortico-basal ganglia-thalamocortical loop.Together with the findings that most, if not all, thalamicprojecting M1L5 neurons branch to innervate the caudoputamen (Economo et al., 2018;Jiang et al., 2020), the intriguing possibility arises that a single M1L5 corticofugal neuron can affect motor thalamic activity indirectly through the classical basal ganglia loop (which is disinhibitory) as well as through the direct projection (excitatory).Such an organization has been anticipated in songbirds (Goldberg and Fee, 2012;Goldberg et al., 2013).Still, the precise organization of these branches (i.e., whether an M1L5 neuron can target an individual thalamic relay through both pathways) is unknown.Even more, how signals organized in this fashion ultimately contribute to motor function requires much further experimentation.
Clear next steps are: 1) to understand how motor thalamus integrates these various inputs is through causal behavioral experiments and 2) to investigate whether the inhibitory gating of transthalamic pathways occurs across higher order thalamic nuclei.

Animals
Experiments were approved by the Institutional Animal Care and Use Committee at the University of Chicago.Transgenic mice expressing cre recombinase in layer 5 of cortex were bred by crossing male Rbp4-cre KL100Gsat/Mmcd mice (GENSAT RP24-285K21; MGI:4367067) with female C57BL/6J mice.Cre positive offspring of both sexes were used in experiments.

Tissue preparation for histology and fluorescence microscopy
As described previously (Miller-Hansen and Sherman, 2022) animals were transcardially perfused with phosphate-buffered saline followed by 4% paraformaldehyde in phosphate-buffered saline, pH 7.4.The brain was extracted and postfixed in 4% paraformaldehyde overnight at 4°C before transferring to a cold 30% sucrose solution for >48 h.Brains were then cryosectioned coronally at 50μm thickness on a sliding microtome.
Brain sections were mounted on Superfrost Plus (Fisher Scientific) slides and coverslipped with Vectashield (Vector Laboratories).A microscope with a 100 W mercury lamp with fluorescence optics (Leica Microsystems) was used to image the sections and photos were taken with a Retiga 2000 monochrome CCD camera and Q Capture Pro software (Qimaging).FIJI (NIH) was used to overlay images.

Tricolor labeling analysis
Fluorescence images for 3 mice with tricolor anterograde labeling (described above) were taken under identical conditions at two coronal planes (-0.7AP and -1.0AP) according to the Paxinos and Franklin's mouse brain atlas.Each channel, corresponding to either M1L5 (blue), Cb (red), or GPi (green) for each animal was analyzed separately.To average the fluorescence across animals in order to achieve a semi-quantitative assessment of terminal zones, the fluorescence values for each pixel within a rectangular region of interest (ROI) that encapsulated the motor thalamus (as well as the area immediately surrounding it) was extracted for each channel.Then, the background fluorescence value (a region of the tissue containing no terminal fluorescence) was subtracted from all pixels.
The background-subtracted raw fluorescence values were then averaged across the three mice.An image of the ROI was then reconstructed from the raw fluorescence values by utilizing the 'text image' import function in Fiji.A binary mask of the pixels containing terminal fluorescence was generated by thresholding the image at two thresholds: 10% (more stringent) and 20% (less stringent) to generate the input maps for dense and sparse terminal labeling, respectively.This was repeated for each label/channel and overlayed onto the coronal sections displayed in Figure 2 by aligning the slices from which the fluorescence values were taken to the atlas.2D and 4D) was performed using images processed via the above method, using the less stringent threshold (examples shown in Figures 2D and 3D).The thresholded images were colored green (GPi or SNr), red (Cb), and blue (M1L5) and compiled into a composite image, for which pixels receiving multiple inputs were distinguished with a unique color.For instance, pixels that were positive for both GPi (green) and Cb (red) inputs were colored yellow (green + red = yellow).Pixels of each color within an ROI encompassing VA/VL of the motor thalamus (i.e., excluding VM) were then counted using the "color counter" plugin in Fiji.The percentage of overlapping pixels was then normalized to the sum of pixels from both inputs (e.g., the number of GPi/Cb overlapped (yellow) pixels was divided by the sum of the number of GPi and Cb pixels).This gives a percentage overlap that considers the volume of the individual projections, which is quantified for each set of inputs and displayed in Figures 2D and 4D.

Acute slice preparation and whole cell recordings
Animals were deeply anesthetized (nonresponsive to toe pinch) and immediately transcardially perfused with 8mL of ice cold oxygenated (95% O2, 5%CO2) artificial cerebrospinal fluid, which contained the following (in mM): 125 NaCl, 25 NaHCO3, 3 KCl, 1.25 NaH2PO4, 1 MgCl2, 2 CaCl2, a 25 glucose.The brain was extracted, gluemounted on a vibratome platform (Leica) for either standard coronal slices, at a 30° AP angle from the coronal plan, or at a 55° ML angle from the coronal plane to preserve descending cortical L5 axons (Agmon and Connors, 1991) for analysis of Cb or L5 synaptic properties, respectively (Jackman et al., 2014;Mo and Sherman, 2018), and sliced in the same solution (ice-cold).Slices were cut at 385μm thickness.Brain slices were then transferred to 33 °C oxygenated artificial cerebrospinal fluid that was allowed to return to room temperature thereafter.This recovery in artificial cerebrospinal fluid occurred in the dark for 1 h before recordings began.
Slices containing terminals from the described inputs were visualized using differential interference contrast with a Axioskop 2FS microscope (Carl Zeiss).
Fluorescence from ChR2 expression was confirmed using the 5× air objective and guided recording locations.Recordings were made with a Multiclamp 700B amplifier and pCLAMP software (Molecular Devices).Recording glass pipettes with 4-6 MΩ resistance were filled with intracellular solution containing the following (in mM): 127 K-gluconate, 3 KCl, 1 MgCl2, 0.07 CaCl2, 10 Hepes, 0.1 EGTA, 2 Na2-ATP, 0.3 Na-GTP, pH 7.3, 290 mOsm.Pharmacological inactivation of iGluRs was induced by bath application of 50 μM DNQX and 100 μM AP5 (Tocris).Blockade of mGluRs was attempted by bath application of 40μM LY367385 and 30μM MPEP (combined; Tocris).Pharmacological inactivation of GABAA receptors was performed by bath application of 20µM SR 95531 (gabazine), while block of GABAB receptors was performed by application of 25µM CGP 46381 (Tocris).The locations of each patched cell were logged and displayed in the Figures.
Optogenetic stimulation was performed as previously (Mo and Sherman, 2018;Miller-Hansen and Sherman, 2022).Briefly, stimulation was delivered using a 355 nm laser (DPSS: 3505-100), controlled with galvanometer mirrors (Cambridge Technology) focused on the slice through a 5× air objective using custom software in MATLAB (MathWorks).First, focal photostimulation of the terminals was performed to test for the presence or absence of the inputs of interest.If the cell received only an excitatory input (i.e., either M1L5 or Cb, but no basal ganglia), the synaptic properties of this input were tested by distal photo stimulation of the axons (as described in Results).Four pulses of 1ms duration were delivered at a range of interstimulus intervals, including 100 ms (10Hz), 50ms (20Hz), and 25ms (40Hz) during recordings.To test for the presence of mGluR responses, 40Hz optogenetic stimulation (or, in some cases, 20 pulses of 1-ms delivered at 12 ms ISI, 83 Hz (Miller-Hansen and Sherman, 2022), was used but lack of mGluRs was equally conclusive) and responses were recorded in current clamp.In rare cases (6/177 neurons, like those used for representative traces), a single neuron underwent bath infusion of Gabazine (during which 10 sweeps of laser stimulation were performed) followed by DNQX (+APV) to conclusively demonstrate the presence of both inputs.Gabazine infusion was acute (~2 minutes) to allow washout and recovery of the inhibitory signal before applying DNQX.The whole process was typically performed in roughly 40 minutes, yet the success rate was modest.

Analysis of synaptic terminal size
Confocal Z-stacks (10-15 z-planes per image at 5µm interplane interval) from five animals injected as above for Cb and GPi were analyzed for terminal size within and just surrounding the overlap zone.Images were acquired with roughly half of the view field containing the GPi terminal zone, while the other half encompassed the surrounding region that contained Cb terminals/fibers.The z-stacks were collapsed in a maximum intensity projection (Fiji) and ROIs were manually drawn that either encapsulated the GPi terminal zone or the surrounding area with care to ensure that the ROIs were matched in terms of their areas.Images were then segmented using the 'triangle' auto threshold algorithm in Fiji.Next, the pixels that satisfied the threshold for being counted, i.e., the synaptic terminals, were analyzed separately within each ROI using the 'analyze particles' tool in Fiji.The only specific parameter used is that a particle had to represent at least three pixels to be counted.This same analysis was performed for two z-stacks (one from each tissue section) for each animal, corresponding to the locations we recorded in (AP -0.7 and AP-1.0).The data for the two images were averaged to get the value for each animal.

Data analysis and statistics
Electrophysiological data were collected using custom MATLAB software and analyzed using Graphpad Prism (v7.0).The amplitude of responses to stimulation pulses was measured by subtracting the average value for 20 ms before the delivery of a pulse (baseline) from the maximum value of the peak.The PPR was calculated by dividing the amplitude of the second pulse by that of the first pulse.Statistical tests involving two groups was performed using non-parametric Mann-Whitney U tests, where indicated, in Prism.Image analysis was conducted in FIJI (NIH) and figures were produced using Adobe Illustrator.Note the reference to the Allen Coronal Atlas (https://mouse.brain-map.org/static/atlas)depicting the GPi injection site and VA/VL terminal zone, where marked.Also note that for Cb injections, the bolus of virus was introduced without the necessity for restriction to the deep cerebellar nuclei, as these are the only outputs of the Cb.D) Representative images of coronal sections at -0.55mm AP from bregma.Injection site and terminals from GPi (green), terminals (or lack thereof) from Cb (red), and M1L5 axons (blue) coursing through the internal capsule and innervating motor thalamus are shown.Note that at this rostral pole of the motor thalamus, where a thin sliver of VA is visible, only a limited projection from GPi is visible, whereas cerebellar and M1L5 terminals are absent.Scale = 100µm for all images.E) Representative images of coronal sections at -1.4mm AP from bregma.

Figures and Figure legends
Injection site and terminals (or lack thereof) from GPi (green), terminals from Cb (red), and M1L5 axons (blue) coursing through the internal capsule and innervating motor thalamus are shown.Note that at this caudal region of the motor thalamus, Cb fibers are well represented in the VPL and VL, while only a minimal projection from GPi is visible; M1L5 terminals largely innervate VM.Scale = 500µm for all images.

Figure 1 .
Figure 1.A feedforward transthalamic pathway through VA/VL.A) Schematic of injections and experiment.B) Representative of M1 (top row) and M2 (bottom row) injection sites with the corresponding Paxinos and Franklin (Paxinos and Franklin, 2007) atlas image, demonstrating no bleed through of one injection to the next.Scale = 500µm.C) Representative low magnification (4x) DIC and fluorescence images from the recording rig (top row) and post-hoc histology of M1L5 (ChR2) and M2 (fluororuby) labeling (bottom row) demonstrating a region of overlap between the two labels in the ventral VA/VL.Scale = 500µm for top row, 100µm for bottom row.D) Representative high magnification (40x immersion) images from the recording rig demonstrating the recording electrode patched onto a fluororuby-positive neuron in VA/VL (same neuron the traces in E are derived from).Scale = 20µm.E) Representative traces in current (top) or voltage (bottom) clamp (held at -60mV) of laser-evoked responses (blue rectangles) in an M2-projecting VA/VL thalamic neuron.F) quantification of the amplitude of the first EPSP in a 10Hz pulse train for all cells in the transthalamic pathway (n=4).G) quantification of the paired-pulse ratio (PPR; amplitude of 2 nd EPSP/1 st EPSP in a 10Hz pulse train) for all cells in the transthalamic pathway (n=4).Data are mean ± SEM.

Figure 2 .
Figure 2. Anterograde labeling reveals overlap of motor thalamus inputs in ventral VA/VL.A) Schematic of injections and experiment (top) with representative low magnification image (bottom) of a section demonstrating the three terminal zones in motor thalamus and GPi injection site.White dashed line represents the entopeduncular nucleus (EPN), which shows expression of virus in addition to the globus pallidus (GP) lateral to the internal capsule (IC).Yellow dashed line represents the area of higher magnification images shown in B. Scale = 500µm.B) Representative images of coronal sections at -0.8mm and -1.0mm AP from bregma.Injection site and terminals from GPi (green), terminals from Cb (red), and M1L5 axons (blue) coursing through the internal capsule and innervating motor thalamus are visible.White dashed line represents the area imaged in higher magnification in the next row of images.Asterisk demarks the same image used for convergence analysis in D. Scale = 100µm for all images.C) Cross-animal (n=3) averaged input maps of motor thalamus inputs at -0.8mm and -1.0mm AP from bregma color coded as in A and B. Dark colors represent dense terminal fields, while lighter colors represent sparser terminal fields.In the merged image (right), only outlines of the dense terminal fields are included.D) Representative image depicting the terminal fields from each input with bit masks color coded as in A-C and quantification (right) of terminal field overlap (see Materials and Methods) for sections at -0.8mm and -1.0mm AP from bregma.See Extended Data Figure 2-1 for representative injection sites and additional images of the rostral-caudal extent of inputs to motor thalamus.Extended Data Figure 2-1.The examined inputs are segregated at the rostral and caudal poles of the motor thalamus.Representative injection sites for GPi (A), M1L5 (B) and Cb (C) for a three color-labeled animal.Viruses injected are depicted in Figure 2A.

Figure 3 .
Figure 3. GPi inputs to motor thalamus converge with those from M1L5 and Cb.A) Schematic of experiment.B) Representative images from recording apparatus of the acute coronal slice (left) showing injection sites into GPi and terminals in motor thalamus (green), M1L5 axons coursing through the internal capsule and terminals in motor thalamus (blue), and merged image.Dashed rectangle indicates approximate target of recording.Arrows in bottom left of image represent dorsal (D) and lateral (L) directions.Scale = 0.5mm C) Representative traces in voltage clamp of laser-evoked responses (blue rectangles) in three motor thalamic neurons (recorded with low Cl-internal solution) held near the resting membrane potential (Vm, -60), at -40mV, at -40mV with application of the GABAA antagonist Gabazine (20µM), and at -60mV with application of the iGluR antagonists, DNQX (50µM) and AP5 (100µM).Note that some cells only exhibit IPSCs, some only EPSCs, and some demonstrate a mixed EPSC/IPSC response, of which the excitatory or inhibitory components are abolished by iGluR or GABAA blockade, respectively.For quantification of the synaptic properties of excitatory inputs, see Figure 6D.D) Summary data depicting how many recorded cells receive only GPi input, only M1L5 input, both, or neither.E) Spatial pattern of motor thalamus neurons receiving each input type color coded to the inputs they receive.F) Schematic of experiment.G) Representative images from recording apparatus of the acute coronal slice (left) showing injection sites into GPi and terminals in motor thalamus (green), Cb terminals in motor thalamus (red), and merged image.Dashed rectangle indicates approximate target of recording.Arrows in bottom left of image represent dorsal (D) and lateral (L) directions.Scale = 0.5mm.H) Representative traces in voltage clamp of laser-evoked responses (blue rectangles) in three motor thalamic neurons (recorded with low Cl-internal solution) heldnear the resting membrane potential (Vm, -60), at -40mV, at -40mV with application of the GABAA antagonist Gabazine (20µM), and at -60mV with application of the iGluR antagonists, DNQX (50µM) and AP5 (100µM).Note that some cells only exhibit IPSCs,

Figure 4 .
Figure 4. Anterograde labeling reveals stronger overlap between SNr and M1L5 compared to Cb and SNr in VA/VL.A) Schematic of injections and experiment (top) with representative low magnification image (bottom) of a section demonstrating the three terminal zones in motor thalamus.Yellow dashed line represents the area of higher magnification images shown in B. Scale = 500µm.B) Representative images of coronal sections at -0.8mm and -1.0mm AP from bregma.Terminals from SNr (yellow), terminals from Cb (red), and M1L5 axons (blue) coursing through the internal capsule and innervating motor thalamus are visible.White dashed line represents the area imaged in higher magnification in the next row of images.Asterisk demarks the same image used for convergence analysis in D. Scale = 100µm for all images.C) Cross-animal (n=2) averaged input maps of motor thalamus inputs at -0.8mm and -1.0mm AP from bregma color coded as in A and B. Dark colors represent dense terminal fields, while lighter colors represent sparser terminal fields.In the merged image (right), only outlines of the dense terminal fields are included.D) Representative image depicting the terminal fields from each input with bit masks color coded as in A-C and quantification (right) of terminal field overlap (see Materials and Methods) for sections at -0.8mm and -1.0mm AP from bregma.See Extended Data Figure 4-1 for representative injection sites.Extended Data Figure 4-1.Representative injection sites for tricolor labeled SNr animal.Representative injection sites for SNr (A), M1L5 (B) and Cb (C) for a three colorlabeled animal.Viruses injected are depicted in Figure 4A.Note that for Cb injections, the bolus of virus was introduced without the necessity for strict restriction to the deep cerebellar nuclei, as these are the only outputs of the Cb.Scale = 500µm for all images.

Figure 5 .
Figure 5. SNr inputs to motor thalamus converge with those from M1L5 and Cb.A) Schematic of experiment.B) Representative images from recording apparatus of the acute coronal slice showing SNr synaptic terminals in the motor thalamus (yellow), and M1L5 axons invading the motor thalamus through the internal capsule (blue), and merged image.Arrows in bottom left of image represent dorsal (D) and lateral (L) directions.Scale = 0.5mm C) Representative traces in voltage clamp of laser-evoked responses (blue rectangles) in three motor thalamic neurons (recorded with low Cl-internal solution) heldnear the resting membrane potential (Vm, -60), at -40mV, at -40mV with application of the GABAA antagonist Gabazine (20µM), and at -60mV with application of the iGluR antagonists, DNQX (50µM) and AP5 (100µM).Note that some cells only exhibit IPSCs, some only EPSCs, and some demonstrate a mixed EPSC/IPSC response, of which the excitatory or inhibitory components are abolished by iGluR or GABAA blockade, respectively.For quantification of the synaptic properties of excitatory inputs, see Figure

Figure 6 .
Figure 6.All inputs examined are drivers.A) Representative current clamp recordings of motor thalamic neuron demonstrating distal activation of M1L5 axons (>300µm from recorded cell) gives a depressing response at low (10Hz, left) and high frequencies (40Hz, right) that is insensitive to mGluR antagonists (40μM LY367385 and 30μM MPEP) but abolished by DNQX.B) Representative current clamp recordings of motor thalamic neuron demonstrating distal activation of Cb axons (>300µm from recorded cell) gives a depressing response at low (10Hz, left) and high frequencies (40Hz, right) that is insensitive to mGluR antagonists (40μM LY367385 and 30μM MPEP) but abolished by DNQX.C) Representative current clamp recordings of motor thalamic neuron demonstrating distal activation of GPi axons (>300µm from recorded cell) gives a depressing response at low (10Hz, left) and high frequencies (40Hz, right) that is insensitive to GABAB antagonists (CGP 46381, 25µM) but abolished by gabazine (20µM).D) Compiled paired-pulse ratio (2 nd EPSC/1 st EPSC) data for all cells recorded receiving only the input of interest (not mixed inputs) plotted according to the amplitude of the first EPSC.PPRs below 1.0 (dotted line) are considered depressing, while those above 1.0 are considered facilitating.For Cb, n=30 cells; for M1L5, n=15 cells.

Figure 7 .
Figure 7. Cb terminal sizes within and outside the GPi overlap zone are equal.A) Representative maximum intensity projections of terminal overlap fields in low magnification (left) and terminals from motor thalamic inputs with GPi and Cb terminal zones distinguished.Thresholded image demonstrates binary masking of terminals for high-throughput estimation of terminal size.Scale = 100µm B) Population distribution of Cb terminal size averaged across five animals (two sections per animal, averaged).C) Average Cb terminal sizes within and outside the GPi overlap area.Ns=not significantly different; n=5 mice.Data are represented as mean ± SEM D) Number of terminals

Figure 8 .
Figure 8.The basal ganglia gate information flow between cortical regions.A) Schematic of the circuit organization of motor thalamus with respect to cortex and basal ganglia, showing the direct (orange) and transthalamic (blue) pathways through motor thalamus.When basal ganglia projections (red) to motor thalamic relays are active, they potently inhibit (red filled motor thalamic cells) and thereby gate transthalamic information flow between cortical regions.In contrast, when basal ganglia inputs to motor thalamus are themselves inhibited (open fill motor thalamic cells), they allow transthalamic information flow between cortical regions, where it may encounter signals from the direct corticocortical pathways.Via this gating function, the basal ganglia can regulate which cortical areas are connected by only direct corticocortical pathways, only transthalamic pathways, both, or neither.Importantly, this regulation is dynamic, such that the pattern of connected areas can change over short time scales (see left, time T0, versus right, time Tn).B) Box diagram of the basal ganglia circuitry according to the traditional textbook view (left) and with the novel direct connection between motor cortex L5 and motor thalamus shown (right; asterisk).That is, our data demonstrate that M1L5 drivers innervate cells receiving GPi and SNr inputs.Given this organization, M1L5 neurons could simultaneously disinhibit (via the basal ganglia loop, black arrows) and drive (via the direct corticothalamic projection, blue arrow) motor thalamic relays.