All experimental procedures were approved by the Harvard Medical School Institutional Care and Use Committee and were performed in compliance with the Guide for Animal Care and Use of Laboratory Animals. Animals were housed in a temperature-controlled and humidity-controlled facility and were maintained on a 12–12 h dark–light cycle. All experiments were performed on adult animals (aged >5 weeks) of both sexes. The following mouse lines were used: C57Bl/J6; Cdx2cre (ref. 48); Calca-FlpE (ref. 49); AvilFlpO (ref. 49); Avilcre (ref. 50); Rosa26LSL-Acr1 (ref. 51); and Rosa26FSF-ReaChR (derived from ref. 52). Animals were maintained on mixed C57Bl/J6 and 129S1/SvImJ backgrounds. C57Bl/J6 mice were obtained from The Jackson Laboratory.
Experiments were not blinded because mice and treatments were easily identifiable as experiments were performed. Whenever applicable, animals were randomized to different treatment conditions. Sample sizes were not predetermined.
All recordings were performed in urethane-anaesthetized mice. Adult (aged >5 weeks) animals were anaesthetized with urethane (1.5 g kg–1) and placed on a heating pad. The head and neck were shaved and local anaesthesia (lidocaine HCl, 2%) was administered to the scalp and neck. An incision was made in the skin, and the muscle of the dorsal aspect of the neck was cut and moved aside to expose the brainstem. A head plate was attached to the skull using dental cement and the head was fixed to a custom-built frame. The dura overlying the brainstem was cut and small fragments of the occipital bone were removed. In some cases in which optical access to the DCN was required, small amounts of the posterior vermis of the cerebellum were aspirated to expose the DCN. A glass electrode filled with saline (2–3 MΩ) was put in place 200–300 µm above the gracile nucleus. The area was then flooded with low-melting point agarose dissolved in saline. After the agarose hardened, the electrode was advanced into the gracile. The electrode was guided to individual units and positioned to maximize the signal of a single unit. Recording quality and unit discrimination were continuously assessed using an audio monitor and online analysis of amplitude and spike waveforms. All recordings in the DCN were made in the hindlimb representation region of the gracile nucleus, and only units with receptive fields in the hindlimb were recorded. Recordings were targeted to the rostral–caudal level approximately where the gracile diverges bilaterally (see example in Extended Data Fig. 1) where units receiving input from the hindlimb digits and pads were most abundant. Although it has been reported that the gracile in rats, cats and primates is subdivided into ‘core’ and ‘shell’ regions, we were unable to detect clear organization in the mouse through electrophysiology experiments. Signals from single units were amplified using the ×100 AC differential amplification mode of a Multiclamp 700B instrument and sampled at 20 kHz using a Digidata 1550B controlled by Clampex 11 software (Molecular Devices). Signals were collected with an additional 20× gain, a 0.3 kHz high-pass filter and a 3 kHz Bessel filter.
In some experiments, DCN neurons were identified through antidromic activation of their axon in a target region. A craniotomy above the region of interest was performed. The head was levelled and bipolar electrodes (platinum–iridium 250 µm spacing, FHC) were lowered into the contralateral VPL of the thalamus (coordinates: 2 mm posterior to bregma, 2 mm lateral, 3.5 mm deep) or the contralateral IC (coordinates: electrode angled 45°, 4 mm posterior to bregma at tissue entry, 1 mm lateral, 1.7 mm deep). A single stimulus (60–200 µA) was applied every 3 s while searching for units in the DCN. A collision test was performed for all units that could be antidromically activated. Stimuli were triggered from spontaneous spikes, and experiments were only continued for units that passed collision testing. We found that units with reliable and precisely timed antidromic spikes (jitter < 1 ms) almost always passed a collision test. Units that failed collision testing had variable timing of activation and much longer latencies. Polysynaptic activation in the DCN was rarely detected by activating the IC, and often observed when stimulating the VPL. In a subset of experiments, once data collection was complete, the stimulus electrode was retracted from the stimulation site and coated with DiI (ThermoFisher). The electrode was then advanced back into the tissue and left in place for at least 5 min. To verify that the electrode was in the same position as before, units were antidromically activated again. Electrodes were then retracted and the animal was perfused as described below for anatomical verification of the stimulus electrode position. In another subset of experiments, following identification of a VPL-PN, the stimulus electrode was then retracted from the VPL and moved to coordinates of the posterior nucleus of the thalamus (coordinates: 2 mm posterior to bregma, 1.1 mm lateral, 3 mm deep). Units that were originally antidromically activated from the VPL failed to be activated by stimulation in the new location (number of experiments are units/animals: 3/3). Stimulation (60–200 µA) of the posterior nucleus also failed to evoke polysynaptic and multi-unit background activity in the DCN, as was often seen with VPL stimulation.
To record from inhibitory neurons in the DCN, we performed optical activation in Vgat-ChR2 animals. Units were searched while applying 200–300 ms ramps of blue light (2–4 mW mm–2) from an optic fibre (400 µm diameter, 0.39 NA) placed above the DCN. Light was delivered from a 470 nm LED (M470F3, Thorlabs). Ramps were used because pulses of light were found to generate short latency activation in most units. Most units in the DCN are glutamatergic, and we found that pulsed light drives strong synchronized GABAergic input to primary afferent terminals and stimulates them through the depolarizing action of GABA, thereby driving vesicle release onto excitatory projection neurons. When ramps were used, many units were instead silenced, as expected. Units considered optotagged were those that could be activated during ramps of light.
For optical silencing of ascending inputs in vivo, experiments were performed as described above, except dissection was performed in the dark under red-light illumination, as bright-white dissection light was capable of activating ACR1. All experiments were performed in animals homozygous for the Rosa26LSL-Acr1 allele, as light was unable to fully silence inputs in heterozygous animals. Once preparation was complete, the electrode was positioned in place above the DCN along with a 400 µm diameter 0.39 NA fibre 1 mm above the DCN. The area was flooded with low-melting point agarose to embed the fibre and electrode. The electrode was then advanced into the DCN to record from single units.
Silencing trials were pseudorandomly interleaved with control trials. For silencing trials, a 300–400 ms ramp of light (0.2–0.8 mW mm–2, 552 nm) was delivered to the surface of the DCN with mechanical stimuli delivered during the last 100–200 ms of the ramp. Light was delivered using a 552 nm LED (MINTF4, Thorlabs). We found that silencing became progressively ineffective after 500 ms of delivery of light, which was possibly due to effects of prolonged depolarization of the terminals.
Optical silencing could only be performed successfully for gentle stimuli (≤ 20–30 mN). In Cdx2-Cre;Rosa26LSL-Acr1 animals, light was unable to fully silence inputs to the DCN when delivering intense indentations. Silencing using ACR1 may substantially suppress vesicle release from ascending inputs. However, the simultaneous and ongoing activation of many inputs may still allow postsynaptic DCN neurons to reach threshold despite a large reduction in the amount of synaptic drive. Thus, experiments were limited to gentle indentation and vibratory stimuli.
Mechanical and optical stimuli
We selected units that were primarily responsive to stimulation of the hindlimb. All mechanical stimuli were generated by a DC motor with an arm connected to a blunt and smoothed acrylic probe tip that was 1 mm in diameter. The use of a large diameter probe tip with smoothed edges allowed the delivery of high forces that that did not evoke paw withdraw in awake animals (Extended Data Fig. 6). We did not observe visible damage to the skin following high-force stimuli in awake or anaesthetized animals. The motor was driven by a custom-built current supply controlled by a data acquisition board (Digidata 1550B, Molecular Devices). Step indentation forces were calibrated using a fine scale. For experiments measuring responses to various forces, the probe tip was positioned in a resting state on the skin surface, and indentations of incrementally increasing forces were applied every 3–10 s. Once the maximal force was reached, the force was reset to zero and was again incrementally increased. For experiments measuring responses to vibration, vibratory stimuli were applied at similar forces (10–20 mN) across frequencies. DCN units that did not respond to Pacinian-range vibration frequencies also did not respond when vibrations were delivered at higher forces.
The stimulator was attached to an articulating arm that could be moved by hand and would remain in position. For receptive field mapping, the probe tip was manually positioned over a single location of the limb where it remained in place. A trial was then initiated to deliver a brief vibration (50 Hz, 100 ms, 10–20 mN). The experiment was performed under a stereoscope equipped with a CCD (BlackFly S BFS-U3-04S2C, Flir) operated by Spinview 220.127.116.11, and was triggered to capture images for each trial. This process was repeated until enough trials were acquired (60–100) to generate a receptive field map of the entire hindlimb. For experiments measuring indentation responses at various forces, the stimulator was firmly secured to a 0.5-inch heavy post to prevent relocation or repositioning when applying high forces.
Optical stimulation was performed on the skin or DCN using a 200 µm or 400 µm diameter 0.39 NA fibre coupled to a 554 nm LED (MINTF4, Thorlabs). For skin stimulation, the fibre tip was held in place with an articulating arm and manually moved into position for each trial, as performed for mechanical stimulation. For optical receptive field mapping, five pulses (2–5 ms duration, 60 mW mm–2) were delivered to the skin at 1 Hz for each trial.
Multielectrode array recordings
Recordings in the thalamus were made in the mVPL (2 mm bregma, 2 mm lateral, 3.5 mm deep). Animals were head-plated, the DCN was exposed as described above, and a craniotomy above the VPL was performed. The dura was removed, and the area was flooded with 2% low-melting agarose dissolved in saline. A 32-channel multielectrode array (MEA; A1x32-poly2–10mm-50s-177-A32, Neuronexus) was lowered at 5 µm s–1 into the brain. Once positioned in a region where firing in many units could be evoked by brushing the hindlimb, the MEA was kept in place for 20 min to ensure stable recording. The hindlimb was embedded in modelling clay for stabilization, and the receptive fields of units were quickly assessed using a brush. The mechanical stimulator probe tip was then placed over the region of the hindlimb that could maximally activate the most units. Step indentations of 0–300 mN were applied every 3–6 s in ascending order and repeated until at least 10 trials per force were obtained. Following stimulation, the DCN was lesioned either by using a 30-gauge needle and striking through the gracile nucleus or by aspirating the gracile nucleus. Step indentations were then repeated. Throughout the experiment, the waveform of a single unit was closely monitored to assess drift, and any experiment with detectable waveform changes were discarded. In some cases, once the experiment was complete, the MEA was retracted from the brain, coated with DiI and then descended to the same coordinates and allowed to stabilize for 5 min. Animals were then anaesthetized with isoflurane and transcardially perfused with PBS followed by 4% paraformaldehyde (PFA) to assess the location of the lesion and electrode placement.
For lesioning of the dorsal column, animals were prepared as described above, and a laminectomy was performed at approximately T12 vertebrae and the dura was removed. A high-density 32-channel MEA (A1x32-poly3–5mm-25s-177-A32, Neuronexus) was inserted into the gracile nucleus and allowed to stabilize for 15 min. A baseline of 5 min of spontaneous firing was collected. The dorsal column was then lesioned at approximately T12 using a 30 gauge needle. Brief vibratory stimuli were applied to the hindlimb throughout the experiment to assess the effectiveness of the lesion.
MEA recordings were made using an Intan head-stage, recording controller (Intan Technologies RHD2132 and recording controller) and open-source acquisition software (Intan Technologies RHX data acquisition software, v.3.0.4). Data were sampled at 20 kHz and bandpassed (0.1 Hz–7.5 kHz).
Ex vivo recordings
Intracellular recordings were performed in random DCN neurons ex vivo. The brainstem was prepared as previously described13, except DCN neurons were recorded directly from the dorsal surface of a dissected brainstem. Borosilicate electrodes (2–3 MΩ) filled with internal solution (consisting of in mM: 130 potassium-gluconate, 3 KCl, 10 HEPES, 0.5 EGTA, 3 MgATP, 0.5 NaGTP, 5 phosphocreatine-Tris2 and 5 phosphocreatine-Na2, pH 7.2 with KOH) were visually guided to the DCN. Experiments were performed in recirculated artificial cerebrospinal fluid (containing in mM: 127 NaCl, 2.5 KCl, 1.25 NaH2PO4, 1.5 CaCl2, 1 MgCl2, 26 NaHCO3 and 25 glucose) oxygenated with 95% O2/5% CO2. The preparation was kept at 35 °C using an in-line heater. Recordings of random DCN cells were made using a Multiclamp 700B (Molecular Devices) and acquired using a Digidata 1550B with Clampex 11 software (Molecular Devices).
For spinal cord silencing experiments, animals were prepared for juxtacellular DCN recordings as described above. A laminectomy was performed over L3–L5 spinal segments, the region of the spinal cord responsive to hindlimb stimulation. For control experiments, a laminectomy was performed over T10–T13 spinal segments. The exposed spinal cord and vertebral column was then flooded with low-melting point agarose. Once hardened, agarose overlying the dorsal horn was cut away to create a pool and confine drugs to the spinal cord. The dura was removed from the spinal cord using fine forceps. Before drug application, stimuli were applied to the hindlimb to evoke typical responses to ensure that the spinal cord or dorsal column had not been damaged. Drugs were then applied to the spinal cord. The non-competitive NMDA receptor antagonist MK-801 (10 mM, 10 µl, Abcam; dissolved in 90% H2O and 10% DMSO) was applied to the surface of the spinal cord and allowed to enter the cord for 3–5 min. The surface of the cord was then irrigated with saline. NBQX (10 mM, 20 µl, Abcam, dissolved in H2O) was then applied to the surface of the cord. After 5 min, the cord was covered with gelfoam, which was allowed to absorb the NBQX and remained in place for the duration of the experiment, occasionally re-wet with saline. Drugs were sequentially applied to prevent precipitates from drug mixing at high concentration. Units were then recorded, and animals were sacrificed within 2–3 h following drug application.
For controls measuring the efficacy of blockade, a laminectomy was performed over the L4 spinal cord. The vertebral column was held in place by two custom spinal clamps. The dura was removed using fine forceps or a fine needle and kept moist with saline. A 32-channel MEA was inserted into medial L4 (A1x32-poly2–10mm-50s-177-A32, Neuronexus). Step indentations (300 mN) were applied at 0.1 Hz, and the MEA was lowered such that the most dorsal channel detected minimal evoked responses. The location of the dorsal channel was assumed to be near the surface of the cord. The MEA was then kept in place for 15 min. Baseline trials were collected and then drugs were applied as described above.
Adult animals were anaesthetized with isoflurane and placed in a stereotaxic frame. The head was tilted 30° forward. The hair over the neck and caudal scalp were removed using a clipper and the skin was sanitized using isopropanol followed by betadine. Local anaesthesia (2% lidocaine HCl) was applied to the area, and an incision was made to expose neck muscles. Neck muscles were separated from the skull to expose the brainstem. A 30-gauge needle was used to cut the dura overlying the brainstem and to expose the DCN. A pipette filled with retrograde tracer (Red Retrobeads, Lumafluor or cholera toxin subunit B, 2 µg µl–1, Fisher) and 0.01% fast green was lowered into the DCN just sufficient to penetrate the surface. Once penetrating the surface, a small volume (30–50 µl) of retrograde tracer was injected. The pipette was held in place for 20 s and then removed. Care was taken to ensure tracer did not leak from the injection site following injection and that tracer labelled with fast green filled the DCN but did not extend beyond the nucleus. The pipette was then removed, and overlying muscle and skin was sutured shut. Animals were administered analgesic (Buprenex SR, 0.1 mg kg–1, ZooPharm) before surgery and monitored post-operatively. After 1–3 days, animals were transcardially perfused for tissue collection.
Animals were anaesthetized with isoflurane and transcardially perfused with PBS followed by 4% PFA in PBS. Brains were removed and post-fixed in PFA overnight. The brain, spinal cord and dorsal root ganglia (DRGs) were dissected free. The isolated spinal cord or brain was mounted in low-melting point agarose, and sections of the thoracic spinal cord (60 µm thick) were made on a Leica VT1000S vibratome. For immunohistochemistry, free-floating sections were first permeabilized with 0.1% Triton-X100 in PBS for 30 min at room temperature. Sections were then incubated with 0.1% Triton-X100 and 4% normal goat serum (Abcam) for 30 min at room temperature. Primary antibodieswere then added (mouse anti-NeuN, 1:1,000, MAB377 clone A60, Millipore; guinea-pig anti-VGLUT1, 1:2,000, Synaptic Systems 135302) and incubated overnight at 4 °C. Sections were then washed three times with PBS with 0.1% Triton-X100 and 4% normal goat serum for 10 min at room temperature. Secondary antibodies were then applied (IB4-Alexa 647, 1:300, ThermoFisher I32450; goat anti-mouse Alexa-488, 1:500, Abcam ab150113; FITC goat anti-GFP, 1:500, Abcam ab6662) for 2 h at room temperature and washed with PBS three times for 10 min. Sections were mounted onto glass slides using Fluoromount aqueous mounting medium (Sigma). Sections were imaged with a Zeiss LSM 700 confocal microscope using a ×20, 0.8 NA oil-immersion objective and Zen software. DRGs were whole-mounted and imaged using a Zeiss LSM 700 confocal microscope using a ×10, 0.45 NA air objective.
PSDC neurons were counted in z-stacks (3–4 µm z-spacing) of 60 µm thick thoracic spinal cord sections. A total of 10–20 sections were analysed per animal, and the average number of PSDC neurons per section was measured and multiplied by 16.67 to estimate the number PSDC neurons in one segment (1,000 µm) of spinal cord for that animal. Aβ-LTMRs were counted in whole-mounted thoracic DRGs and compared to spinal cords of the same animals. Labelled DRG neurons were counted in z-stacks of the entire DRG (5–6 µm z-spacing). If more than one DRG was analysed per animal, the number of counted cells was averaged between the two.
Behaviour experiments were performed in two C57Bl/J6 males and four Calca-FlpE;Rosa26FSF-ReaChR animals (2 male, 2 female). Animals were anaesthetized with 2% isoflurane. Hair over the scalp was clipped using a shaver and skin was disinfected using isopropanol followed by betadine. Local anaesthesia (2% lidocaine HCl) was applied to the scalp. An incision was made and the dorsal skull was cleared of skin and muscle using a scalpel. A head plate was rested on top of the skull and fixed in place using dental cement. Animals were administered analgesic (Buprenex SR, 0.1 mg kg–1, ZooPharm) before surgery and monitored post-operatively.
Animals were allowed to recover for 2–3 weeks. Head-plated animals were then transferred to a rig consisting of an acrylic platform. Animals were head-fixed to a suspended post, but were otherwise free to move. Mice were allowed to habituate to the rig for 5–10 min. Gentle mechanical stimuli were then delivered to prevent startle responses. First a brush was used to gently stroke the trunk and hindlimbs. After 5–10 strokes with a brush, the indenter was introduced and 300 ms indentations of 10 mN were delivered using the same smoothed 1 mm probe tip used for the electrophysiology experiments. The indenter and probe tip were the same as those used for experiments in anaesthetized animals. Indentations were delivered to the middle of the dorsal hindlimb as the ventral side was inaccessible in awake animals with unrestrained paws. When the animal was still, the indenter tip was positioned in place above the hindlimb, a trial was triggered and indentation was delivered after 1 s of baseline. Stimuli were delivered approximately every 20 s. After a few trials using 10 mN stimuli, animals no longer were startled by indentation, and the force was increased to 300 mN. Trials were recorded using a CCD camera (BlackFly S BFS-U3-04S2C, Flir) attached to a stereoscope with frames captured at 100 Hz using Spinview software. The first trial of 300 mN typically evoked a startle response and was therefore discarded. The subsequent trials were then collected for analysis (5–12 trials). Several seconds following 300 mN indentation, a small force (<10 mN) was briefly applied to the probe to bring it into position near the skin again for the subsequent trials.
In four Calca-FlpE;Rosa26FSF-ReaChR animals, once mechanical stimulation was completed, a set of trials were performed in which the skin was optically stimulated. A 400 µm 0.39 NA fibre coupled to a 554 nm LED was held in place above the centre of the dorsal hindlimb. A 50 ms pulse of light (about 30 mW mm–2) was delivered to the skin. The first trial was discarded, and subsequently 4–7 trials were collected for analysis. Experiments were ended after fewer trials because animals showed clear signs of pain.
To determine whether mechanical stimulation generated enough force to prevent the animals from withdrawing their paw, optical stimulation was delivered during mechanical stimulation in two Calca-FlpE;Rosa26FSF-ReaChR animals. A 300 mN, 300 ms indentation was delivered as described above. At 100 ms after the onset of indentation, a 50 ms pulse of light was delivered at or near the indentation as described above. Animals readily withdrew their hindlimb within 30 ms of light onset when 300 mN of force was still being applied.
Juxtacellular recordings in the DCN were analysed offline using custom written scripts in Matlab (Mathworks). Spikes were detected using an amplitude threshold, and recordings in which the unit could not be isolated by amplitude were discarded.
Receptive field maps were generated using custom written scripts in Matlab. An image of the hindlimb from the experiment was used as a template. Images capturing the probe or optic fibre for each trial were cycled through, and the location of the probe or fibre tip was manually marked on the template image. The coordinates of each stimulus location and the template image were then combined with electrophysiology data to identify the number of evoked spikes for each stimulus location. The receptive field map for each unit is displayed with pointsindicating the location of the stimulus and normalized change in firing rate from baseline. The maximal change in firing rate is indicated in text within the image. The maximal firing rate was measured as the number of spikes over the course of the 100 ms vibratory stimulus or the number of spikes within a 20 ms window following optical stimulation. The spontaneous firing rate was measured as the average firing rate before mechanical stimulation across trials. To measure the receptive field area, a threshold was taken at the half-maximal evoked firing rate. Stimulus locations that evoked the half-maximal firing rate or more were included in the receptive field. The area bound by these points was used as a measure of the receptive field area.
Correlations of optical and mechanical receptive fields were determined by comparing regions of the hindlimb. First, the optical and mechanical receptive fields were mapped as described above. The paw was then subdivided into segments corresponding to digits and pads. The stimulus trials falling within each segment were averaged to obtain an average evoked firing rate for each segment. The optically evoked firing rate for each segment was plotted against its mechanically evoked firing rate for that same region. This generated a plot of 17 points, each representing the optical and mechanical response of one region of the hindlimb. A linear fit was performed, and the Pearson’s correlation coefficient (R2) was used as a measure of receptive field correlation. As a control, a similar analysis was performed but optical and mechanical receptive fields were shuffled between units.
Spike sorting (MEAs)
MEA recordings underwent initial analysis using Kilosort 2.0 (ref. 53). Before analysis, baseline and lesions trials were interleaved to prevent artefactual drift corrections by Kilosort. Drift monitoring was performed during acquisition, and experiments with detectable changes in spike waveforms were discarded. Default detection settings were used for analysis, except that template amplitude thresholds were set to [5 2]. Clusters were manually curated using Phy54 by examining spike waveforms and autocorrelations to identify putative single units. As receptive fields were not mapped for MEA experiments, units underwent selection for inclusion. Units that had low thresholds at baseline (<10 mN) or units that had a mechanically evoked firing that was unaffected by lesioning the DCN were included. Units that had high thresholds (>10 mN) at baseline and became mechanically insensitive following DCN lesion were not included for analysis because the stimulus may have been off the centre of the receptive field.
Z-scores were computed by measuring the average baseline firing rate within a 0.5–1 s window before stimulation. For optical silencing, pharmacology and lesion experiments in which the same units were monitored before and after manipulation, the Z-scores in experimental conditions were computed from the mean and standard deviations of baseline trials.
Videos of mechanical and optical stimuli delivered to the hindlimb were collected in trials lasting 3 s. The paw position was semiautomatically tracked using the Video Labeler application in Matlab. Baseline paw position was collected 0.5 s before stimulus delivery for the trial. The baseline position was set as the origin, and the position of the paw relative to the starting position was measured for the entire trial. For average plots, the position was averaged across trials for each animal. For total movement, the total change in position was summed over the average trial for each animal.
Statistical analyses were performed with the significance threshold set at P < 0.05. All summary data are presented as the mean ± s.e.m. unless otherwise noted. All t-tests were two-tailed.
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.