The rILN consist of a contiguous band of cells within the internal medullary lamina that is parcellated into CL, PC, and CM nuclei (Berman et al., 1983). Although anatomical tracing suggests subtle differences in afferent and efferent connectivity among these three subregions (Van der Werf et al., 2002), it is unclear whether these nuclei represent functionally distinct cellular groups. To investigate whether rILN→DS neurons exhibited physiological differences across the three nuclear divisions, we injected a retrograde traveling tdTomato-expressing virus in the central DS of mice and assessed neurophysiological properties of rILN→DS neurons using whole-cell patch clamp electrophysiology in acute slices (Figure 1A and B). We found that neurons across the three nuclei did not significantly differ in membrane capacitance (Figure 1C; F[2,40]=1.242, p=0.30), membrane resistance (Figure 1D; F[2,40]=1.212, p=0.31), input resistance (Figure 1E; F[2,40]=0.520, p=0.60), or resting membrane potential (Figure 1F; F[2,40]=0.124, p=0.88).

To assess differences in intrinsic firing properties, we injected a current ramp to induce action potential (AP) firing. We did not observe a significant difference in the AP threshold between the three nuclei (Figure 1G; F[2,40]=2.684, p=0.08). Following a series of 0.5 s current steps, the maximum firing rate, as calculated from the first six APs, did not differ between neurons from the three rILN nuclei (Figure 1H; F[2,39]=1.411, p=0.26). AP accommodation was also not different between the nuclei (Figure 1—figure supplement 1A; F[2,35]=1.671, p=0.20). We observed that hyperpolarizing current steps induced post-hyperpolarization burst firing in all three nuclei (as shown in Figure 1C). We did not find differences in either the post-hyperpolarization burst firing inter-spike interval (Figure 1—figure supplement 1B; F[2,35]=1.003, p=0.38) or AP number (Figure 1—figure supplement 1C; F[2,40]=2.837, p=0.07).

Establishing that rILN→DS neurons display homogeneous intrinsic electrophysiological properties, we next endeavored to elucidate when this pathway is active during action learning and performance. We fiber photometrically recorded activity-dependent calcium signaling selectively from rILN→DS neurons in mice learning a lever-pressing operant task (Figure 2A). To assess whether rILN→DS neurons respond to sensory stimuli eliciting an action, action initiation, termination, kinematic properties, or reward, we trained mice to respond to the lever extension by pressing five times in a defined period of time for a sucrose pellet reward (Figure 2B). In early training stages, mice must complete one, three, or five fixed-ratio presses (FR1, FR3, and FR5, respectively) with no time limit (NTL) for completion. Intermediate schedules consist of progressively shortened response periods to complete the FR5, terminating at a 5 s time limit. We found transient increases in rILN→DS activity that accompanied the first press in the action sequence across all stages of training (Figure 2C). Comparing the average rILN→DS activity z-score at 1.5 s before and at the time of the first press, we observed significant differences from the first training protocol (fixed ratio 1; FR1), an intermediate schedule (FR5-NTL), and the final training schedule (FR5-5s time limit; time relative to press: F[1,12]=26.73, p=0.0002; schedule: F[2,24]=9.587, p=0.0009). We observed similar positive fluctuations in rILN→DS activity aligned to subsequent presses in early training protocols in which mice complete the FR lever-pressing task over an extended period (Figure 2D–F). This activity was not observed for subsequent presses in mice on the terminal training protocol, however (Figure 2D–F; schedule: F[1,12]=34.79, p<0.0001; press: F[3,36]=2.146, p=0.11).

We next examined how rILN→DS neurons activate relative to specific elements of the FR5 task. We found that rILN→DS activity peaked following the extension of the lever into the operant chamber on completed trials in intermediate and terminal stages of training and that the magnitude of this signal change increased with training (Figure 2G; time relative to lever: F[1,12]=17.27, p=0.0013; schedule: F[3,36]=14.49, p<0.0001; interaction: F[3,36]=6.776, p=0.001). The average latency to initiate the lever press for completed trials was 8.7 s for FR1 and decreased to 1.9 s (FR5-NTL) and 0.76 s (FR5-5s; Figure 2H; F[1.218,14.61]=303.0, p<0.0001). Gross rILN→DS activity at this time, as measured by area under the photometric signal, was negatively correlated with initiation latency on completed trials across all FR5 schedules (r=−0.161; –0.185 and –0.137 95% CI, p<0.0001, n=6338 trials). Thus, we examined whether this lever press-associated rILN→DS signal varied by task performance. We found that a larger signal accompanied completed trials as compared to incomplete trials (in which mice pressed one to four times; Figure 2I; time relative to first press: F[1,12]=14.73, p=0.0024; trial type: F[1,12]=2.374, p=0.15).

To further explore this rILN→DS signal, we investigated whether rILN→DS activity correlates with general movement by recording mice freely moving in an open arena (Figure 2—figure supplement 1A). We observed increases in rILN→DS signaling aligned to movement onset (Figure 2—figure supplement 1B; t=2.817, p=0.037) and preceding the maximum velocity achieved during movement epochs (Figure 2—figure supplement 1C; t=4.116, p=0.009). Significant changes in rILN→DS activity did not accompany movement cessation (Figure 2—figure supplement 1D; t=0.341, p=0.75).

To discriminate between action initiation and the sensory cues (e.g. lever extension) that may signal action initiation, we used a Pavlovian appetitive conditioning paradigm in which sucrose pellets were administered at the termination of an auditory tone (Figure 3—figure supplement 1A–B). We did neither find tone-related changes in rILN→DS signaling nor did we observe learning-dependent changes in either the tone-paired group (Figure 3—figure supplement 1C; time relative to tone: F[1,5]=2.046, p=0.21; training stage: F[1,5]=2.466, p=0.18) or the tone non-paired control cohort (Figure 3—figure supplement 1D; time relative to tone: F[1,5]=2.312, p=0.19; training stage: F[1,5]=0.204, p=0.67).

Alternatively, our observation of training-dependent increases in action initiation-associated rILN activity may reflect the learned expectation of a rewarded action. To test this, we recorded rILN→DS activity after switching FR5 reinforcement to 0%. Mice extinguished FR5 lever pressing within several sessions (Figure 3—figure supplement 1F, t=11.69, p<0.0001), but the first press-aligned signal on completed (but unreinforced) FR5 trials did not differ significantly from reinforced sessions (Figure 3—figure supplement 1G; time relative to lever: F[1,11]=13.36, p=0.0069; reinforcement: F[1,11]=0.1628, p=0.73). In an alternative paradigm, FR5-trained mice alternated between sessions in which completed trials were reinforced at 100 or 50% probability. Lowering the probability of reinforcement did neither alter performance (Figure 3—figure supplement 1H; t=0.788, p=0.45) nor did it produce significant differences in first press-aligned rILN activity (Figure 3—figure supplement 1I; time relative to press: F[1,10]=5.571, p=0.04; reinforcement rate: F[1,10]=0.304, p=0.59).

Lastly, we observed that rILN→DS activity increases at reward acquisition. Sucrose pellet retrieval was accompanied by increased rILN activity on all FR training protocols (Figure 3A–B; time relative to retrieval: F[1,12]=57.24, p<0.0001; schedule: F[1,12]=10.42, p=0.007; interaction: F[1,12]=5.311, p=0.040). Additionally, this signal negatively correlated with reward retrieval latency across all FR5 schedules (Figure 3C; r=−0.123, –0.147, and –0.097, 95% CI, p<0.0001). Reward delivery was necessary for this activity, as unreinforced FR5 trials did not elicit changes in rILN→DS activity (Figure 3D; time relative to head entry: F[1,12]=21.01, p=0.006; reinforcement: F[1,12]=5.610, p=0.035; interaction: F[1,12]=89.14, p<0.0001). On intermediate training schedules, mice frequently check the pellet receptacle in between individual presses. We examined rILN→DS activity during these ‘premature’ receptacle head entries on trials that were ultimately completed and observed that these events were accompanied by possible negative fluctuations in rILN activity (Figure 3E; time relative to head entry: F[1,12]=3.536, p=0.085; schedule: F[2,24]=0.502, p=0.61; interaction: F[2,24]=0.019).

We next assessed the generalizability of this reward-related signal in non-operant paradigms. We recorded rILN activity from mice freely moving in an arena with strawberry milk located in two corners. rILN→DS activity increased when mice approached the strawberry milk-baited corners but not the opposing non-baited arena corners (Figure 3F; time relative to approach: F[1,5]=12.13, p=0.018; corner: F[1,5]=12.54, p=0.017; interaction: F[1,5]=7.965, p=0.037). To directly examine rILN→DS activity relative to reward consumption, we recorded from mice drinking sucrose water from bottles connected to a lickometer. rILN→DS activity significantly increased relative to lick bout onset (Figure 3G; t=3.210; p=0.005).

We next tested whether reward devaluation may influence rILN→DS activity. Mice were given free access to sucrose pellets prior to 30 min test sessions in which pellets were pseudo-randomly delivered. Devaluation reduced both latency (Figure 3H; t=3.953, P=0.002) and frequency of pellet receptacle head entries (Figure 3I; t=5.068, p=0.0004) but did not significantly alter rILN→DS activity aligned to reward retrieval (Figure 3J; time relative to head entry: F[1,11]=12.23, p=0.003; reward value: F[1,11]=0.642, p=0.44).

To causally test the role of rILN→DS activity, we optogenetically inhibited dorsal striatal projecting rILN neurons in well-trained mice performing the FR5 task (Figure 4A). Delivering blue light during an epoch encompassing the action initiation and reward retrieval events (White et al., 2018; White et al., 2020) pseudo-randomly on 33% of the trials (Figure 4B) to halorhodopsin-eYFP (NpHR-eYFP) or control eYFP-expressing mice, we found that rILN→DS neuronal inhibition resulted in fewer completed FR5 trials (Figure 4C; light delivery: F[1,26]=21.90, p<0.0001; virus: F[1,26]=0.390, p=0.54) and more omissions (Figure 4D; light: F[1,26]=13.99, p=0.0009; virus: F[1,26]=0.127, p=0.72; interaction: F[1,26]=13.10, p=0.0013; Figure 4—figure supplement 1). We also observed a small but significant effect of light delivery on the percentage of incomplete trials (Figure 4E; light: F[1,26]=4.834, p=0.037; virus: F[1,26]=2.306, p=0.141).

We previously demonstrated that optogenetic activation of rILN→DS terminals reinforces actions in an intracranial self-stimulation paradigm (Cover et al., 2019). However, it is unknown whether activation of this pathway at action initiation also supports the execution of rewarded action sequences. To test this, we virally expressed channelrhodopsin (ChR2-eYFP) or a control fluorophore (eYFP) in rILN→DS neurons (Figure 4F). Blue light was delivered on 33% of trials for 2 s starting 1 s prior to lever extension (to promote the lever press action) in trained mice (Figure 4G). ChR2-eYFP mice completed more FR5s on light-paired trials (Figure 4H; light: F[1,25]=24.39, p<0.0001; virus: F[1,25]=0.002, p=0.97; interaction: F[1,25]=13.21, p=0.0013) and, correspondingly, had fewer omitted trials (Figure 4I; light: F[1,25]=29.46, p<0.0001; virus: F[1,25]=0.656, p=0.43; interaction: F[1,25]=18.85, p=0.0002), compared to both non-light delivered trials and eYFP controls (Figure 4—figure supplement 1). We did not find significant differences in the number of incomplete FR5 trials (Figure 4J; light: F[1,25]=0.358, p=0.55; virus: F[1,25]=1.441, p=0.24).

Although an extensive number of inputs to the rILN are identified (Van der Werf et al., 2002), it is largely unknown which of these afferents may relay to the striatum to possibly contribute to function. Thus, we sought to qualitatively identify afferents impinging specifically onto rILN→DS neurons. To do this, we applied a viral strategy involving injection of nuclei that are upstream of rILN with an anterograde trans-synaptic AAV1 viral construct (AAV1-hSyn-Cre-WPRE-hGH; Zingg et al., 2017) that delivers cre recombinase to rILN neurons. In addition, we injected a retrogradely traveling cre-dependent tdTomato-expressing virus (AAVrg-CAG-FLEX-tdTomato-WPRE) in the DS. This viral approach produces tdTomato labeling in all rILN→DS neurons post-synaptic to the neurons at the site of AAV1-cre injection (Figure 5—figure supplement 1). We used this method to interrogate known excitatory inputs to the rILN. We found that both the anterior cingulate cortex (ACC; Figure 5A) and orbitofrontal cortex (OFC; Figure 5B) synapse on ipsilateral rILN→DS neurons and lesser so on contralateral rILN→DS neurons. We also found that projections arising from the glutamatergic/GABAergic hypothalamic supramammillary and lateral (S/L) nuclei (Hashimotodani et al., 2018; Stuber and Wise, 2016) innervate the rILN→DS neurons residing in CM and contralateral CL (Figure 5C). Previous tracing studies indicate that the rILN receive input from the basal ganglia output projection, the substantia nigra (SN) pars reticulata (Kaufman and Rosenquist, 1985), as well as from other targets of nigral output including the superior colliculus (SC; Yamasaki et al., 1986), reticular formation (RF; Krout et al., 2002), and pedunculopontine nucleus (PPN; Hallanger et al., 1987; Huerta-Ocampo et al., 2020). We observed that projections from all four of these regions synapse on rILN→DS neurons (Figure 5D–G). These neuronal tract tracing results are quantified (Figure 5—figure supplement 2).

Two- to five-month-old male and female C57BL/6J (wild-type; Jackson Laboratory, #000664) mice were housed in a temperature and humidity controlled vivarium under a 12 hour light/dark cycle (lights on at 0700 hours). Mice were housed with littermates (two to five) per cage, except for those singly housed following fiber or cannula implantation. Mice performing operant tasks were weighed and fed daily to maintain 90% of ad libitum weight; all others received ad libitum food and water. All experiments were performed in accordance with the United States Public Health Service Guide for Care and Use of Laboratory Animals and were approved by the Institutional Animal Care and Use Committee (AUP 0522009) at the University of Maryland, Baltimore. Sample sizes were determined based on publication standards. Studies were completed across multiple cohorts that produced homogeneous experimental findings. Animals were randomly assigned to control and experimental groups. Group assignment was also balanced to ensure equal numbers of males and females between experimental and control cohorts. Blinding was not used in the present study as all subjects and data were treated with identical equipment, analysis methods, and exclusion criteria.

Mice were anesthetized with isoflurane (5% induction, 1–2% maintenance) and head-fixed in a mouse stereotaxic apparatus (Kopf Instruments, Tujunga, CA, USA). Bupivacaine hydrochloride (0.25%; s.c.) was applied to the scalp prior to incision. Viruses were infused at a rate of 25 nl/min using a 25 G syringe (Hamilton Company, Reno, NV, USA). All mice received carprofen (5 mg/kg; s.c.) and recovered for a minimum of 7 days to prior to behavioral testing.

For fiber photometry experiments, AAVrg-EF1a-mcherry-IRES-Cre (Addgene #55632-AAVrg) was unilaterally injected in the central DS (600 nl; distance from bregma in mm, [anterior-posterior, AP] +0.65, [medial-lateral, ML] 1.70, and [dorsal-ventral, DV] –3.00), and two unilateral injections of AAV5-Syn-Flex-GCaMP6s (Addgene #100845-AAV5) were made in the rILN (300 nl/injection; [AP] –1.25 and –1.95, [ML] 0.75, and [DV] –3.00). In a separate surgery 3–4 weeks after, a low numerical aperture (NA; 0.22) fiber (200 um core, Thorlabs, Newton, NJ, USA) was implanted in the rILN ([AP] –1.60, [ML] 0.76, and [DV] –3.00). For optogenetic manipulations, AAVrg-EF1a-mCherry-IRES-Cre was bilaterally injected in the DS (as stated above), and either AAV5-EF1a-DIO-eNpHR3.0-eYFP-WPRE (Addgene #26966-AAV5), AAV5-EF1a-DIO-hSyn-ChR2-eYFP-WPRE (Addgene #20298-AAV5), or AAV5-EF1a-DIO-eYFP-WPRE (Addgene #27056-AAV5) was injected in the rILN (350 nl/hemisphere; [AP] –1.40, [ML] ±0.75, and [DV] –3.50). Two optic fibers (NA 0.66, 200 µm core, Prizmatix, Givat-Shmuel, Israel) were later implanted in the rILN: one at ([AP] –1.60, [ML] –0.75, and [DV] –2.70) and the second at a 20° AP angle ([AP] –2.58, [ML] +0.75, and [DV] –3.03). All implants were secured with skull screws (BASi, West Lafayette, IN, USA) and dental cement.

For electrophysiology experiments, retrograde AAVrg-CAG-tdTomato (Addgene #59462-AAVrg) was injected in the DS (same coordinates and volume as previously listed) in wild-type mice to patch tdTomato-labeled rILN neurons.

Trans-synaptic tract tracing was performed to identify inputs to rILN→DS neurons. Mice were bilaterally injected with AAVrg-CAG-FLEX-tdTomato-WPRE in the DS (Addgene #28306-AAVrg; 600 nl/hemisphere; [AP] +0.65, [ML] ±1.70, and [DV] –3.00). Afferent regions of interest were unilaterally injected with AAV1-hSyn-Cre-WPRE-hGH (Addgene #105553-AAV1) as well as Fluorogold (20 nl; Fluorochrome LLC, Denver, CO, USA) to label the injection site. The unilateral injection hemisphere was counterbalanced across animals for each input region. Stereotaxic coordinates and virus volumes are as follows: OFC ([AP] +2.50, [ML] 1.00, and [DV] –2.20; 250 nl), ACC ([AP] +1.00, [ML] 0.30, and [DV] –1.10; 200 nl), S/L ([AP] –2.40, [ML] 0.50, and [DV] –5.65; 150 nl), SN ([AP] –3.20, [ML] 1.50, and [DV] –4.80; 250 nl), SC ([AP] –3.50, [ML] 0.80, and [DV] –2.75; 240 nl), RF ([AP] –4.20, [ML] 0.75, and [DV] –4.25; 310 nl), and PPN ([AP] –4.70, [ML] 1.20, and [DV] –3.00; 250 nl).

Immunohistochemistry was performed to confirm virus expression and fiber placement; animals were excluded from experiments for poor virus expression at target regions or viral expression in non-target regions (i.e. the caudal intralaminar parafascicular nucleus).

Mice were transcardially perfused with room-temperature 0.1 M PBS, pH 7.3, followed by ice cold 4% (wt/vol) paraformaldehyde in PBS. Brains were extracted and post-fixed with 4% paraformaldehyde in PBS at 4°C. 50 µm coronal sections were cut using a Leica VT100S vibrating microtome. Goat anti-GFP (Abcam #ab6673, Waltham, MA, USA) and chicken anti-mCherry (Novus Biologicals #NBP2-25158, Littleton, CO, USA) primary antibodies were used at a 1:2000 dilution to amplify GCaMP6s and cre expression, respectively. Chicken anti-mCherry was also used to amplify virally expressed tdTomato. Donkey anti-goat conjugated to Alexa Fluor 488 (Jackson ImmunoResearch #703-545-155, West Grove, PA, USA) and donkey anti-chicken conjugated to Alexa Fluor 594 (Jackson ImmunoResearch #703-585-155) secondary antibodies were used at a 1:1500 dilution. The Brain BLAQ protocol (Kupferschmidt et al., 2015) was used for immunohistochemistry following electrophysiology.

Mice were deeply anesthetized with isoflurane before rapid decapitation and brain removal. 250 µm thick coronal slices were prepared in ice cold, 95% oxygen, 5% carbon dioxide (carbogen)-bubbled modified artificial cerebrospinal fluid (aCSF) (In mM: 194 sucrose, 30 NaCl, 4.5 KCl, 1 MgCl2, 26 NaHCO3, 1.2 NaH2PO4, and 10 D-glucose) before incubating at 32°C for 30 min in aCSF (in mM: 124 NaCl, 4.5 KCl, 2 CaCl2, 1 MgCl2, 26 NaHCO3, 1.2 NaH2PO4, and 10 D-glucose). Slices were then stored at room temperature until recording.

Thalamic brain slices were transferred to the recording chamber and perfused with carbogen-bubbled aCSF (in mM; 124 NaCl, 4.5 KCl, 2 CaCl2, 1 MgCl2, 26 NaHCO3, 1.2 NaH2PO4, and 10 D-glucose) at 29–31°C. Whole-cell recordings were made with borosilicate glass pipettes (2–5 MΩ). aCSF was perfused on slices through a gravity perfusion system. DS projecting rILN neurons were recorded with a potassium-based internal solution (in mM: 126 potassium gluconate, 4 KCl, 10 HEPES, 4 ATP-Mg, 0.3 GTP-Na, and 10 phosphocreatine; 290–295 mOsm; pH 7.3) containing a hydrazide dye conjugated with Alexa Fluor 488 (40 mM) to allow for post-hoc identification of cell location within the rILN. Membrane capacitance and resistance were collected at –60 mV voltage clamp configuration, whereas the resting membrane potential was obtained in current-clamp with a MultiClamp 700B amplifier (Molecular Devices, San Jose, CA, USA); recordings were filtered and digitized at 2 and 10 kHz, respectively. Input resistance was calculated from the mV change induced by a 140 pA hyperpolarizing current step. Data was acquired with Clampex 10.4.1.4 and analyzed with Clampfit 10.4.1.4 (Molecular Devices). Calculation of the accommodation index for rILN AP spiking was based on the entire 0.5 s current injection step that elicited maximum firing as previously calculated (White and Mathur, 2018). Cells that did not maintain sustained firing (i.e. resulted in a depolarization block) were excluded from this analysis. Cells were determined to reside in CL, PC, and CM nuclei based on the Franklin and Paxinos, 2008.

Photometry data were collected using a customized in vivo fiber photometry system. Two single-wavelength laser modules were used: a 473 nm laser for optimal GCaMP6s excitation and a 405 nm laser to excite GCaMP6s at its isosbestic wavelength (Opto Engine, Midvale, UT, USA). Emission from 405 nm excitation was used to control for signal artifacts due to photometry cable motion, background fluorescence, and other sources of noise (Kim et al., 2016). The two lasers were multiplexed at 10 or 15 Hz, resulting in a continuous 20 or 30 Hz pulse train. Both laser beams were bounced into a dichroic filter cube designed for 473 and 405 nm excitation as well as for 510 nm emission (Chroma Technology Corp., Bellow Falls, VT, USA). The two excitation wavelengths were focused through a ×4 fluorite objective (Olympus, Tokyo, Japan) onto a multimode fiber bundle (Thorlabs). One fiber was connected to the mouse through a chronic unilateral multimode fiber implant, and another fiber was placed inside a tube of Alexa Fluor 488 to control for variability in laser energy. Emissions from GCaMP6s and Alexa Fluor 488 were detected as an image of the fiber bundle using an ORCA-Flash4.0 LT high-resolution CMOS camera (Hamamatsu Photonics, Hamamatsu City, Japan). Laser multiplexing and image acquisition were synchronized using an Arduino Leonardo microcontroller. Trial or time-dependent recordings were initiated through MedPC or Ethovision systems. Camera image acquisition parameters were controlled through HCImage software for Hamamatsu cameras.

Electrophysiological data were analyzed using Clampex software. Statistical analyses were performed in Prism (version 6.01; GraphPad Software, San Diego, CA, USA) or MATLAB (R2019a; The MathWorks, Inc, Natick, MA, USA).

Photometry data were analyzed using a combination of custom MATLAB code (see Source code 1) and Prism. Pixel intensities imaged from the fiber implanted in the rILN and the fluorophore control fiber were first averaged. The background signal in the absence of laser transmission was then subtracted from the averaged signals. Two separate regressions were performed to minimize any noise sources. First, 473 and 405 nm photometry signals were regressed with the corresponding control fluorophore signal, and the residuals of the regression were then used for further processing. The 473 nm signal was then regressed with the 405 nm signal as a covariate, and the residuals of the regression were extracted as the fully processed photometry signal (z-score) from the rILN (White et al., 2020). The photometry signal was then statistically analyzed based on performance or operant task event (i.e. movement-aligned activity or FR trial outcome). For each analysis comparing two or more categories of rILN signal, the number of averaged events or trials were standardized across each compared category for each animal. The average z-score for three consecutive time points was statistically analyzed with the appropriate t-test or ANOVA. Signal AUC was computed in MATLAB.

To quantify relative strength of afferent innervation to rILN→DS neurons identified in the di-synaptic circuit tracing study, tdTomato-labeled cells were counted in CM, ipsilateral CL, and contralateral CL. Separate ratios of cell counts for CM: ipsilateral CL and ipsilateral CL: contralateral CL were derived, correcting for area, within each coronal slice and averaged across slices and cases for a given assessed afferent. Ratios indicating a difference ≥40% of labeled neurons between the compared rILN nuclei are denoted by line thickness in Figure 5 summary diagrams. This criterion was also used to assess relative strength of ipsilateral and contralateral afferent→afferent→DS di-synaptic circuits, when applicable.

All statistical analyses are reported in the Results. The specific statistical test as well as value and description of n are listed in the figure legends. Averaged data is expressed as mean ± SE. Correlations are expressed as Pearson’s r and 95% CIs. All statistical tests were conducted two-sided, when applicable. Significant post-hoc Holm-Ŝidàk tests for ANOVAs are indicated in the figures. All asterisks in figures indicate: * p<0.05, ** p<0.01, *** p<0.001, and **** p<0.0001.