Circuit Specific Plasticity of Callosal Inputs Underlies Cortical Takeover.

Injury induces synaptic, circuit and systems reorganization. After unilateral amputation or stroke this functional loss disrupts the interhemispheric interaction between intact and deprived somatomotor cortices to recruit deprived cortex in response to intact limb stimulation. This recruitment has been implicated in enhanced intact sensory function. In other patients, maladaptive consequences such as phantom limb pain can occur. We used unilateral whisker denervation in male and female mice to detect circuitry alterations underlying interhemispheric cortical reorganization. Enhanced synaptic strength from the intact cortex via the corpus callosum (CC) onto deep neurons in deprived primary somatosensory barrel cortex (S1BC) has previously been detected. It was hypothesized that specificity in this plasticity may depend on to which area these neurons projected. Increased connectivity to somatomotor areas such as contralateral S1BC, primary motor cortex (M1) and secondary somatosensory cortex (S2) may underlie beneficial adaptations, while increased connectivity to pain areas like anterior cingulate cortex (ACC) might underlie maladaptive pain phenotypes. Neurons from the deprived S1BC that project to intact S1BC were hyperexcitable, had stronger responses and reduced inhibitory input to CC stimulation. M1-projecting neurons also showed increases in excitability and CC input strength that was offset with enhanced inhibition. S2 and ACC-projecting neurons showed no changes in excitability or CC input. These results demonstrate that subgroups of output neurons undergo dramatic and specific plasticity after peripheral injury. The changes in S1BC projecting neurons likely underlie enhanced reciprocal connectivity of S1BC after unilateral deprivation consistent with the model that interhemispheric takeover supports intact whisker processing.SUMMARY STATEMENTAmputation, peripheral injury and stroke patients experience widespread alterations in neural activity after sensory loss. A hallmark of this reorganization is the recruitment of deprived cortical space which likely aids processing and thus enhances performance on intact sensory systems. Conversely, this recruitment of deprived cortical space has been hypothesized to underlie phenotypes like phantom limb pain and hinder recovery. A mouse model of unilateral denervation detected remarkable specificity in alterations in the somatomotor circuit. These changes underlie increased reciprocal connectivity between intact and deprived cortical hemispheres. This increased connectivity may help explain the enhanced intact sensory processing detected in humans.


Introduction
Central or peripheral injury drives widespread cortical remodeling (Garraghty and Kaas, 1991;Navarro et al., 2007). Amputees experience a recruitment of the deprived sensory/motor cortex leading to bilateral responses to intact limb stimulation (Lotze et al., 2001;MacIver et al., 2008;Simões et al., 2012). The behavioral importance and synaptic mechanisms responsible for these changes are unknown. For example, some studies show that patients experience beneficial adaptations which enhance prosthetic use or increase intact limb sensitivity (Chen et al., 2013a;Templeton et al., 2018). Other studies indicate that remodeling may be related to phantom limb pain (Flor et al., 2006); however, this cortical remodeling is not consistently observed in phantom limb (Makin et al., 2013(Makin et al., , 2015. A rodent model of unilateral whisker denervation (infraorbital nerve transection: ION-X) mimics some of the changes detected in humans; for example there is potentiation of the pathway into intact primary somatosensory barrel cortex (S1BC) and task-evoked recruitment of deprived S1BC during intact whisker stimulation (Yu et al., 2012;Petrus et al., 2019). The recruitment of deprived S1BC is likely because of potentiation of the corpus callosum (CC) inputs from intact S1BC to deprived S1BC, which was so strong to L5 neurons that long-term potentiation (LTP) was occluded (Petrus et al., 2019). Interestingly, this plasticity was not found in L2/3 neurons, demonstrating specificity to deep layer neurons. Because deep cortical neurons often represent output cells for the sensory cortical column (Douglas and Martin, 2004), it was hypothesized that the plasticity may be specific to subsets of neurons based on the downstream brain regions they target. The CC input is believed to be relatively uniform as are cortically projecting L5 principal neurons. However, the differential plasticity of the CC restricted to the deep neurons in deprived S1BC leads to the hypothesis that the plasticity because of ION-X is highly specific.
Throughout the brain, there are examples of input and output specificity, where neurons restrict changes to specific inputs, or provide altered outputs (Larsen and Sjöström, 2015;Rawson et al., 2017). Efforts to describe these changes are limited to alterations in input or output specificity between interneurons and principal cells (Trouche et al., 2013;Lu et al., 2014;Lee et al., 2016) or laminar-specific adaptations (Koester and Johnston, 2005;Sjöström and Häusser, 2006). L5 neurons have been categorized by their electrophysiological properties (Shai et al., 2015), connectivity (Hattox and Nelson, 2007), morphology (Oswald et al., 2013), location (Lefort and Petersen, 2017), and gene expression (Tasic et al., 2016;Clare et al., 2018). These cortical output neurons send long-range projections throughout the brain and are thus optimally situated to integrate columnar information (Koester and Johnston, 2005;Sjöström and Häusser, 2006;Baker et al., 2018) and effect widespread adaptations (Grillner et al., 2005;Feldmeyer, 2012;Petersen, 2019). It is not known whether a uniform input (like the CC) can produce restricted plasticity depending on output neuron connectivity. In the present work, CC input onto L5 neurons projecting to output regions involved in whisker sensation and movement [primary motor cortex (M1), intact S1BC, secondary somatosensory cortex (S2)] or pain perception [anterior cingulate cortex (ACC) and S2] were studied. Retrograde viruses (rAAVs)-labeled distinct cell populations that project from the deprived S1BC to contralateral S1BC, M1, S2, and ACC. After ION-X, reciprocally connected S1BC neurons were hyperexcitable, had stronger responses to CC stimulation and less CC mediated inhibition. Cells projecting to M1 were also hyperexcitable and had stronger CC synapses but a concurrent increase in inhibition resulted in an unchanged total callosal response. Cells projecting to S2 and ACC were unchanged in any parameters measured. It was demonstrated that L5 output neurons experience plasticity to a uniform input (the CC) depending on their output targets. The increase in bilateral connectivity supports the hypothesis that the intact S1BC colonizes the deprived S1BC after unilateral sensory loss to aid intact whisker processing.

Animals
All procedures were approved by the National Institutes of Health Animal Care and Use Committee (ACUC) under protocol 1160, and facilities are accredited by the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC). C57Bl/6 (RRID: IMSR_JAX:000664) and PV 1 cre (RRID: IMSR_JAX:008069) mice were bred in house. Both mouse lines were originally purchased from The Jackson Laboratory. Equal gender distributions were used throughout the study. No significant differences in results were detected between genders. Littermates were housed one to four per cage with food and water ad libitum on a 12/12 h light/dark cycle.

Sham/ION-X surgery
Four to six weeks after stereotaxic viral injection mice received a sham or ION-X surgery. Mice were weighed and received an intraperitoneal injection of a ketamine/xylazine cocktail (ketamine 80 mg/kg/xylazine 10 mg/kg). Once the absence of the hindpaw pinch reflex was observed, whiskers were shaved unilaterally and an incision caudal to the whisker pad was made to visualize the bundle of infraorbital nerves. The sham animals' surgery finished at this step; the incision was closed with one suture and tissue glue (Tissuemend II, Veterinary Product Laboratories). The nerve bundle was cut with scissors in ION-X experimental animals, and the incision was closed in the same manner as the sham group. Animals then received an intraperitoneal injection of Antisedan (1 mg/ kg; Atipamezole, Zoetis) to reverse the anesthesia. Animals remained on a heated surface and returned to the animal facility after they were fully ambulatory, usually after 10-30 min.

Electrophysiology
Acute slice preparation Two weeks after ION-X or sham surgeries acute slices were made for whole-cell electrophysiological recordings. Mice were anesthetized using isoflurane vapors (5% mixed with O 2 ) until the absence of the corneal reflex was observed. The brain was quickly dissected and immersed in ice-cold dissection buffer (80 mM NaCl, 3.5 mM KCl, 1.25 mM H 2 PO 4 , 25 mM NaHCO 3 , 4.5 mM MgSO 4 , 0.5 mM CaCl 2 , 10 mM glucose, and 90 mM sucrose), which was bubbled continuously with a 95% O 2 /5% CO 2 gas mixture. Brain blocks containing primary somatosensory cortex were dissected and coronally sectioned into 300-mm-thick slices using a Leica VT1000S vibratome (Leica Biosystems Inc.). Slices were incubated for 30 min at 35°C and then returned to room temperature for a minimum of 30 additional minutes before recordings began.
Slices were transferred to a submersion-style recording chamber mounted on a fixed stage (Sutter Instruments Company) with an upright Nikon Eclipse FN1 microscope (Nikon Instruments) and illuminated with oblique infrared (IR) illumination. Recordings were all performed in artificial CSF (ACSF; 124 mM NaCl, 5 mM KCl, 1.25 mM NaH 2 PO 4 ·H 2 O, 26 mM NaHCO 3 , 10 mM dextrose, 2.5 mM CaCl 2 , and 1.5 mM MgCl 2 ) bubbled with 95% O 2 /5% CO 2 at 30°C unless otherwise indicated. The ACSF was continually perfused at a rate of 2 ml/min. Voltage clamp experiments used Cs-gluconate internal solution, which contained the following: 130 mM Cs-gluconate, 8 mM KCl, 1 mM EGTA, 10 mM HEPES, 4 mM ATP, and 5 mM QX-314; pH 7.3, 285-295 mOsm. Current-clamp experiments used K-gluconate internal solution, which contained the following: 130 mM K-gluconate, 10 mM KCl, 0.2 mM EGTA, 10 mM HEPES, 4 mM MgATP, 0.5 mM NaGTP, and 10 mM Naphosphocreatine; pH 7.3, 280-290 mOsm). Cells with an access resistance higher than 25 MV and input resistance lower than 100 MV were discarded. Cells which had more than a 20% fluctuation in these values for long-term recordings were excluded from analysis. A maximum of three cells per animal, one cell per slice were recorded for each dataset.
An axon patch-clamp amplifier 700B (Molecular Devices) was used for voltage clamp recordings. Data were acquired through pClamp10 and analyzed with Clampfit 10.4 software (Molecular Devices).
Connection probability of callosally targeted principal cells L5 neurons labeled with rAAV-mCherry expression in deprived S1BC were visually identified and patched for whole-cell recordings. ChR2 was activated using a 455-nm light emitting diode (LED) DC2100, illuminated through the 40Â objective lens and controlled by a digital stimulator (Cygnus DG4000A), both from ThorLabs. Cells which responded to LED stimulation were counted as responding to callosal inputs, while those with no response, even to maximum LED intensity, were counted as not responding. The probability of CC response varied between groups, with reciprocally connected S1BC neurons and S2 output neurons exhibiting the greatest probability of CC response (72% and 70%, respectively), while M1 and ACC output neurons were the less likely to be targeted by the CC (60% and 35%, respectively). Less than 10% of neurons in any group were driven to spike with LED stimulation of ChR2-expressing CC fibers. Subsequent experiments consistently verified CC targeting of rAAV-labeled neurons before recordings.

Intrinsic excitability measurements
These experiments were performed as described previously . Pipettes with tip resistances of 4-to 6-MV resistance were filled with K-gluconate internal solution and after verifying CC-mediated responses, 10 mM NBQX, 100 mM APV and 50 mM bicuculline were added to ACSF to block AMPA, NMDA, and GABA A receptors, respectively. L5 principal neurons were recorded in current clamp with current injected to maintain the cells at À70 mV. Current pulses of 800-ms duration were injected at 10-s intervals with increasing amplitudes of 20-pA steps for 15 repetitions (starting with 0 pA, maximum injection of 300 pA). From the data collected, the rheobase (minimum current needed to produce a single action potential) was calculated for sham and deprived S1BC output L5 neurons. The resting membrane potential and input resistances were also measured to determine baseline excitability for these neurons. Cell properties are listed in Extended Data Figure 1-1.

Sr 21 mEPSCs
Methods are similar to those described previously . AMPAR-mediated Sr 21 mEPSCs were isolated pharmacologically with 20 mM bicuculline and 100 mM DL-2-amino-5 phosphonopentanoic acid (DL-APV). ACSF was as described previously; however, there was 0 mM CaCl 2 , 4 mM MgCl 2 , and 4 mM SrCl 2 . rAAV-labeled L5 neurons in deprived S1BC were identified visually and patched using whole-cell patch pipettes with tip resistances between 4 and 6 MV, which was filled with Cs-gluconate internal solution.
The minimum LED intensity was used to elicit reliable responses and desynchronized Sr 21 mEPSCs ChR2-mediated responses were obtained. Cells were held at À70 mV and recorded for a minimum of 10 min. Data were acquired every 10 s for a duration of 1200 ms, which included a 500-ms duration prestimulation, a 500-ms duration post-LED illumination, and a 100-ms seal test pulse. The 400-ms window preStim was used to quantify the spontaneous events (preLED) and the 400-ms window after stimulation [following a 50-ms recovery period (postLED)] was used to quantify stimulation-evoked desynchronized events (Sr 21 mEPSCs) because of the presence of SrCl 2 and absence of CaCl 2 . To calculate the amplitude of Sr 21 mEPSCs evoked by stimulation we used the following equation: [(postLED amplitude Â postLED frequency) À (preLED amplitude Â preLED frequency)]/(postLED frequency À preLED frequency). Cells were only used if the postLED frequency was at least 2 Hz higher than preLED frequency. Cell properties are listed in Extended Data Figure 3-1.
Excitatory/inhibition balance rAAV-labeled L5 neurons were patched for whole-cell recordings in voltage clamp configuration in normal ACSF with 100 mM APV added to block NMDA receptors. Callosally mediated EPSCs were elicited with LED stimulation of ChR2 transfected terminals arising from the contralateral virally transfected S1BC. LED intensity was adjusted to evoke similar sized EPSCs in sham versus ION-X neurons for each output neuron type. After stable sized events were achieved 20 repetitions were recorded from each neuron, switching between À70 mV for EPSCs and 0 mV for IPSCs.

Input/output (I/O) curves
rAAV-labeled L5 neurons were patched for current-clamp recordings and LED stimulation of ChR2-expressing CC terminals was used to verify CC targeting. Electrical stimulation was used to measure the I/O curve, via a bipolar electrode (FHC) controlled by a DS3 isolated constant current stimulator (Digitimer Ltd.). The stimulating electrode was placed 1-1.6 mm medial to the recorded L5 neurons to avoid thalamocortical (TC) activation into the cortical column. A reliable monosynaptic response (,2-ms onset latency) with similar kinetics to that evoked by LED activation of ChR2 expressing callosal terminals was achieved for each cell. The maximum amplitude response to intensity was measured as 100% response, and subsequent stimulus intensities were decreased to: 80%, 60%, 40%, 20%, and 10% of the 100% stimulus intensity. Rarely a spike was elicited using this method, in those cases the 100% was labeled as the stimulus intensity immediately below spike threshold. There were no significant levels of spiking detected from output neurons in sham or after ION-X. The slope was calculated for I/O curves between 20% and 80% points to avoid the plateau affecting the dynamic range window's slope. Cell properties are listed in Extended Data Figure 4-1.
Immunohistology and cell counting methods Two to three weeks after stereotaxic injection of one to four viruses, animals were anesthetized with 5% isoflurane vapor. Transcardial perfusion was performed with PBS followed by 4% paraformaldehyde and brains were removed. Cryopreserved and frozen brains expressing one to four fluorophores were sectioned at 30 mm thick on a cryostat (Leica CM3050 S) and mounted onto glass slides. Sections were coverslipped using ProLong Gold antifade reagent (Fisher Scientific). Slides were imaged on a Zeiss Axioscan.Z1. Spectral unmixing was performed using Zeiss ZEN to eliminate bleed-through between emission spectra from the four fluorescent proteins: GFP, mOrange, mAmetrine, and mCherry. Sections from animals injected with a single virus were imaged using the same parameters as the four-color samples, and a reference spectrum was selected for each chromophore. Before cell counting, each multi-injection section was spectrally unmixed using the reference spectra. Individual cell bodies for each channel were identified using the cell detection feature in NeuroInfo (MBF Bioscience). For each image, cell size and intensity filter settings were adjusted to identify rAAV-labeled cells. Individual sections were registered to the Allen Mouse Brain Atlas using NeuroInfo and labeled cells were counted in specific barrel field layers of the deprived S1BC.

Statistics
Statistical tests compared the differences between values from single cells (n). Comparison across two groups were performed with Student's t test and the Wilcoxon rank-sum test. To compare between groups of output neurons (i.e., sham S1BC vs M1, S2, and ACC) a two-way ANOVA followed by Tukey's multiple comparisons test was used. For all tests, p , 0.05 was considered statistically significant. Effect size was calculated using Cohen's d to compare the distributions of means; values accompany the extended data figures for datasets.

Results
S1BC and M1 output cells are hyperexcitable after ION-X Unilateral whisker denervation increased excitability in CC-targeted L5 neurons in deprived S1BC (Petrus et al., 2019). Our hypothesis was that this change may be restricted to neurons with specific output targets. A virus encoding for channel rhodopsin (ChR2) was injected into intact S1BC, the infraorbital (whisker) nerve was transected (ION-X) and retrograde adeno-associated viruses (rAAVs) with a red fluorophore were injected into output regions, including: intact S1BC, S2, M1, or ACC (Fig. 1A, left). Two weeks after ION-X was determined to be the maximal time point for plasticity along the intact pathway (Yu et al., 2012;Chung et al., 2017), and robust callosal alterations were also described at this time point (Yu and Koretsky, 2014;Petrus et al., 2019). Therefore, acute slices were made two weeks after ION-X, and rAAV-labeled cells were targeted for whole-cell current-clamp recordings. Responses to ChR2 were first verified to ensure only CC-targeted cells were studied. After wash-on with drugs to block synaptic activity, intrinsic neuronal properties were measured, including depolarization-induced spiking, resting membrane potential (Vm), and rheobase values (Fig. 1A, right). Deprived L5 neurons reciprocally connected to intact S1BC exhibited increased maximum spike rate, a depolarized Vm and a lower rheobase. Student's t test for sham versus ION-X p = 0.036, p = 0.044, and p = 0.038, respectively ( Fig. 1B; Extended Data Fig. 1-1). Deprived L5 neurons projecting to M1 also experienced depolarized Vm and lower rheobase values but did not exhibit altered spiking properties. Student's t test for sham versus ION-X p = 0.025, p = 0.043, and p = 0.467, respectively ( Fig. 1C; Extended Data Fig. 1-1). S2 output cells did not exhibit significant changes in any parameters measured. Student's t test for sham versus ION-X p = 0.182, p = 0.712, and p = 0.161 for spike rate, Vm, and rheobase values ( Fig. 1D; Extended Data Fig. 1-1), while ACC output cells experienced a decrease in maximum firing rate but no change in Vm or rheobase. Student's t test for sham versus ION-X p = 0.042, p = 0.203, and p = 0.136, respectively ( Fig. 1E; Extended Data Fig. 1-1). Overall, these results indicate that CCtargeted L5 output cells experience a heterogenous and output-region-specific adaptation to loss of sensory drive in deprived S1BC.

Output cells are distinct populations
Because the neurons studied here were all cortico-cortically projecting neurons, it has not been established if these are truly nonoverlapping groups of cells. Although others have described low levels of overlap between S2-projecting and M1-projecting neurons in S1 (Chakrabarti et al., 2008;Chen et al., 2013b), we sought to verify these findings and further describe the laminar profile of output neurons in S1BC. rAAVs expressing unique fluorophores were injected into the four output regions ( Fig. 2A) and efficient separation between cell labeling was detected (Fig.  2B). An Allen Brain Atlas was overlaid on the slices to detect the layers of S1BC (Fig. 2C, upper). Cells expressing more than one Figure 1. S1BC and M1 output cells are hyperexcitable after ION-X. A, left, Experimental design: unilateral injection of virus expressing channel rhodopsin (ChR2) in sham/intact S1BC, subsequent sham/ION-X surgery plus retrograde virus (rAAV) injected into various output regions; two weeks later, acute slices were made and neurons were patched for recordings. Center, Representative slice with green CC ChR2 axons and red rAAV neurons for recordings in S1BC. Right, Representative traces demonstrating depolarizing current injections prompting cell spiking. B-E, left, rAAVs were injected into different output locations (intact S1BC, M1, S2, ACC), and intrinsic neuron properties were measured. Input/firing plots, maximum firing rate, Vm, and rheobase values were measured in each output neuron type in sham/deprived S1BC. After ION-X, neurons projecting to intact S1BC have higher maximum spike rate, depolarized Vm, and lower rheobase. color were rarely seen (,6% of any group; Fig. 2C, lower; Table 1). A large component of output cells was located in L5-6 regardless of cortical target ( Fig. 2D; Table 1), further confirming that output neurons are located in deep cortical layers. Qualitatively, more cells with outputs to contralateral S1BC and M1 were detected than cells projecting to S2 and ACC (Table 1). The lack of overlap between cell groups and the location of labeling indicate that these output cells may represent an excellent mechanism by which the brain can direct both widespread and region-specific adaptations to changes in sensory experience.
Callosal input is stronger to S1BC and M1 output neurons but not S2 or ACC outputs Previously it has been shown that callosal drive onto L5 neurons likely underlies bilateral cortical recruitment of S1BC after ION-X (Petrus et al., 2019). However, in the present study, we found that hyperexcitability was restricted to output-specific subgroups of L5 neurons that project to intact S1BC and M1 (Fig. 1). Therefore, to investigate possible output-specific changes in CC inputs onto these neurons, we substituted Ca 21 with Sr 21 in the extracellular fluid and CC-mediated events were recorded by stimulating ChR2-expressing axons arising from virally transfected contralateral neurons. The amplitudes of AMPA receptor mediated Sr 21 miniature EPSCs (Sr 21 mEPSCs) were measured as a proxy for strength of the postsynaptic response to CC stimulation. After ION-X, the amplitude of postsynaptic responses to CC stimulation were larger in S1BC and M1 output cells; Student's t test p = 0.006 and p = 0.004, respectively (Fig. 3A,B; Extended Data Fig. 3-1), while S2 and ACC output cells experienced no significant changes in strength; Student's t test p = 0.423 and p = 0.698, respectively (Fig. 3C,D; Extended Data Fig. 3-1). These results demonstrate that both the intrinsic properties and postsynaptic potentiation are restricted to specific output cells after ION-X. The shift in excitability and synaptic strength may predispose CC-targeted S1BC and M1 output cells to respond to CC stimulation more readily than S2 or ACC neurons.
Callosal release and evoked inhibition adaptations are restricted to S1BC output neurons The previous experiments determined that only S1BC and M1 output neurons experienced significant changes after ION-X, thus further investigation into these types of neurons was performed. S1BC and M1 output neurons experienced a change in both intrinsic properties and postsynaptic response to CC stimulation, it is possible that these changes were accompanied by presynaptic modifications of the CC. Paired pulse ratio (PPR) was measured in CC-targeted S1BC and M1 output cells after ION-X. S1BC output cells had a small but significant shift toward paired pulse facilitation (PPF), indicating either an increase in CC release probability or, more likely, a reduction in callosally recruited inhibition ( Fig. 4A; Extended Data Fig. 4-1, Student's t test p = 0.028). The PPR was unchanged in M1 output neurons ( Fig. 4B; Extended Data Fig. 4-1, Student's t test p = 0.252). This shift to PPF may also be explained by reduced inhibition which fails to "shunt" the second stimulus's excitatory response. Thus, we further studied how callosally evoked inhibition may be altered after ION-X.
Although the CC is a glutamatergic input (Kawaguchi, 1992;Conti and Manzoni, 1994;Petreanu et al., 2007), the net result of CC activity is usually inhibitory by providing strong input onto inhibitory neurons in the contralateral hemisphere. This is referred to as interhemispheric inhibition (IHI; Daskalakis et al., 2002;Palmer et al., 2012). We hypothesized that the shift toward responsiveness in S1BC and M1 output cells may be either compensated for by an increase or further emphasized by a decrease in CC-evoked inhibition. The CC-mediated excitation/inhibition (non-overlapping) labeling was observed for most neurons in S1BC. C, upper, An atlas from Allen Brain Institute was overlaid onto slices to facilitate counting in S1BC laminae. C, lower, Example of a green S1BC and purple S2 output cell (*) overlapping. D, Cells were counted, and laminar distributions were quantified. For raw numbers and overlapping counts, see Table 1.
(E/I) ratio was measured in rAAV-labeled neurons held at À70 mV to evoke normalized excitatory AMPA events and 0 mV for inhibitory GABA events. As expected, strong inhibition was observed in both sham groups as the measured E/I ratio was ,1 (Fig. 4C,D). However, after ION-X CC-evoked inhibition was reduced to reciprocally connected S1BC output neurons with an increase in the E/I ratio to .1.5; Student's t test p = 0.043 ( Fig.  4C; Extended Data Fig. 4-1). The M1 output cells had no significant change in E/I ratio; Student's t test p = 0.175 ( Fig. 4D; Extended Data Fig. 4-1). Callosal excitation was larger as measured by larger AMPA receptor mediated Sr 21 mEPSCs (Fig. 3B), thus the CC-mediated inhibition may increase to match the excitation in M1 output neurons.
Because PV 1 cells are known to be modified by experience and exert strong perisomatic inhibition (Jiang et al., 2005;Donato et al., 2013), it was hypothesized that there may be a change in CC targeting of PV 1 interneurons. However, there was no significant change in the amplitude of CC-mediated Sr 21 mEPSCs onto PV 1 interneurons; Student's t test p = 0.957 ( Fig. 4E; Extended Data Fig. 4-1).
Overall, these results indicate that neurons undergo outputtarget-specific adaptations in response to changes in sensory experience, with the cells projecting from deprived S1BC back to the intact S1BC experiencing the largest changes. These adaptations may shift the state of reciprocally connected neurons to be more responsive to CC input after ION-X compared with other projecting cells.

S1BC-projecting neurons have increased output responses to columnar inputs
After ION-X, deprived S1BC output neurons are hyperexcitable (Fig. 1), have stronger responses to CC stimulation from the intact S1BC, and are less inhibited by the CC (Figs. 3, 4). This led to the hypothesis that the efficacy of callosal transmission to this group of neurons may be increased compared to other output neurons. rAAV-labeled neurons were targeted for whole-cell current-clamp recordings and responses to ChR2 transfected callosally projecting neurons were verified with LED stimulation. Optogenetically evoked responses are known to be unstable at lower intensity stimuli (Morales et al., 2002;Bridi et al., 2020), so I/O curves were generated with electrical stimulation. A bipolar electrode was placed on CC fibers medial to the recorded neuron to evoke events with more stable electrical stimulation. These events were monosynaptic and with similar kinetics to those evoked by LED stimulation of ChR2 expressing callosal terminals. The medial location was chosen to avoid activation of TC fibers; however, we cannot rule out the possibility that antidromic or non-callosal white matter inputs were stimulated by the electrodes. Regardless, we aimed to use the I/O curve as a rough estimate of how different output neurons respond to stimulation of callosal inputs to the cortical column. The maximum response size was found per cell and subsequent stimuli intensity were decreased by regular intervals (Fig. 5A). After ION-X, S1BC output neurons have a steeper I/O slope ( Fig. 5C; Extended Data Fig. 5-1), while no change was found after ION-X in neurons projecting to M1, S2, or ACC. Student's t test p = 0.032, p = 0.603, p = 0.618, and p = 0.341, respectively (Fig. 5D-F; Extended Data Fig. 5-1). The cumulative changes in S1BC output neurons may cooperate to produce a larger response to CC stimulation compared with other output cells. Sham (9, 2) ION-X (10, 3) L1-3 8.7 6 2.9 14.4 6 4.5 L4 0.6 6 0.2 1.5 6 0.5 L5-6 35.3 6 11.8 43.6 6 13.8 Total 44.6 6 14.9 59.5 6 18.8 ACC (mCherry) Sham (9, 2) ION-X (10, 3) L1-3 26.6 6 8.9 14.3 6 4.5 L4 1.9 6 0.6 0.3 6 0.1 L5-6 15.1 6 5.0 10.4 6 3.3 Total 43.6 6 14.5 25 6 7.9 Colocalization Sham (6, 1) ION-X (4, 1) S1BC 1 M1 11.3 6 2.4 3.8% 2.8 6 1.8 1.4% S1BC 1 S2 1.5 6 0.5 0.8% 11.5 6 6.6 6.0% S1BC 1 ACC 3.0 6 0.7 1.6% 0 6 0 0 % M1 1 S2 3.2 6 0.9 1.6% 3.8 6 3.1 2.9% M1 1 ACC 6.0 6 1.5 3.  A summary of the expected changes in circuit dynamics is depicted in Figure 5B. Note the increased intrinsic excitability and stronger CC-mediated synapses in S1BC and M1 output cells (1), and reduction in inhibition only in S1BC output neurons (-). These factors produce a stronger response to CC stimulation only in cells that project back to intact S1BC. This increased response in reciprocally connected S1BC neurons may produce increased bilateral connectivity; denoted by a larger arrow out of the cortical column projecting to contralateral S1BC.

Discussion
The goal of this study was to determine whether the brain can restrict experience-dependent plasticity to specific groups of output neurons. Unilateral whisker denervation (ION-X) produced robust and specific adaptations in deprived neurons targeted by the CC. Brain regions were chosen based on their connectivity with the somatomotor system (intact S1BC, M1, S2) or their function in pain perception (S2, ACC). Reciprocally connected S1BC L5 neurons in deprived S1BC experienced the most robust shift to enhance responsiveness to CC stimulation. M1 output neurons experienced a similar shift toward increased CC responsiveness but a concurrent increase in inhibition negated more dramatic change in callosal responses. No significant changes after ION-X were detected in cells projecting to S2 and ACC. These results support the hypothesis that the recruitment of deprived S1BC is to increase bilateral cortical responsiveness to intact whisker stimulation. This recruitment is thus primed to interact with the intact S1BC during whisker sensation.
Peripheral loss of sensation alters the way the brain processes sensory information, leading to beneficial or maladaptive consequences. In particular, a unilateral loss unbalances the competitive bilateral nature of sensation, which leads to reorganization in adults (Finnerty and Connors, 2000). Along the intact TC pathway, critical period-like plasticity can be reopened to increase responses after sensory deprivation (Montey and Quinlan, 2011;Chung et al., 2017;Hensch and Quinlan, 2018). Deprived TC pathways can remodel quickly (Coleman et al., 2010) or dieback after longer deprivations (Van der Loos and Woolsey, 1973;Oberlaender et al., 2012). In addition, bilateral cortical responses are observed with intact movement or sensation (Lotze et al., 2001;Pelled et al., 2007;Simões et al., 2012;Yu et al., 2012;Sammons and Keck, 2015). The purpose of this recruitment remains a puzzle: does it increase the brain processing power to enhance sensitivity to intact senses? Does it create a hyperexcitable region which by its mis-activation impairs recovery or causes pain? In the model of unilateral denervation studied in the present work, the lack of change in output cells targeting pain perception areas (S2, ACC) suggest that the adaptations are not causing a pain phenotype, but instead increasing the reciprocal connectivity between interhemispheric S1BCs.
Output S1BC neurons are distinct populations with unique characteristics as described here and by others (Hattox and Nelson, 2007;Chakrabarti et al., 2008;Chen et al., 2013b;Oswald et al., 2013;Kinnischtzke et al., 2016). Previous reports have described cortico-cortically projecting L5 neurons as a homogenous group. Here basal differences in neuronal properties were apparent indicating that this is a heterogeneous group. For example, the maximum firing rate, strength of response to the CC and I/O properties in sham animals differed significantly depending on the target of S1BC output neurons (Figs. 1, 3, 4), and after ION-X more robust alterations were observed. Outputspecific adaptations are believed to be recruited to learn new behaviors. For example, accurate whisker task performance increased depolarizations in S1 neurons projecting to S2 but not M1 (Chen et al., 2013b;Yamashita and Petersen, 2016). The cellular/synaptic mechanisms of these results are not known. These output-specific modifications may be a way the cortex restricts changes to crucial neurons rather than the entire layer and all its outputs, in response to a specific behavioral requirement (Adesnik and Naka, 2018).
The adaptations of reciprocally connected S1BC neurons measured in the present study include a shift toward hyper- Figure 4. Callosal release and evoked inhibition adaptations are restricted to S1BC output neurons. A, B, PPR was measured to detect changes in callosal recruitment of inhibition or release probability onto S1BC and M1 output neurons. After ION-X, PPR shifted to facilitation at CC synapses to S1BC output neurons. CC synapses to M1 output neurons were unchanged after ION-X. C, D, E/I (AMPA/GABA) ratio was measured in S1BC and M1 output neurons in response to CC stimulation. S1BC output neurons had significantly less inhibition after ION-X, while M1 inhibition matches excitation. E, top, Expression of green ChR2 and yellow PV 1 neurons in S1BC. Bottom, after ION-X CC-mediated Sr 21 mEPSCs have no change in amplitude to PV 1 neurons. (n) = number of cells, *p , 0.05 t test; for values, see Extended Data excitability, and a stronger postsynaptic response to CC inputs; these results are consistent with our previous study (Petrus et al., 2019). Hyperexcitability is found clinically in deprived cortex after amputation or subcortical stroke (Ziemann et al., 1998;Chen et al., 2002;Sammons and Keck, 2015). This increased activity in deprived somatomotor areas is especially pronounced two weeks after injury and predicts poor motor outcomes in stroke patients (Rehme et al., 2011). Increased excitability allows neurons to modify throughput capacity, meaning a more excitable cell is more likely to pass along presynaptic input as a postsynaptic output (Zhang and Linden, 2003). In addition, hyperexcitability may make long-term plasticity mechanisms like LTP more likely to occur at specific synapses (Moyer et al., 1996). Although the deprived cortex may globally experience less activity after removal of principal whisker input, callosal synapses onto S1BC and M1 output neurons were strengthened. Stronger postsynaptic CC responses were mediated by increased AMPA receptor activity, a hallmark of LTP (Malinow and Malenka, 2002). Reciprocally connected S1BC and M1 output neurons may combine this hyperexcitability and stronger CC responses to create a sensory processing situation such that the intact whisker pathway is more likely to evoke downstream activity only to M1 and back and forth along the CC between intact and deprived S1BCs.
IHI is the inhibitory influence of bilateral cortices on each other's activity (Chiarello and Maxfield, 1996). Activity in one sensorimotor cortex inhibits activity in the contralateral hemisphere, likely through excitatory CC synapses onto interneurons, which then inhibit local circuitry Palmer et al., 2013). Consistent with this function, sham animals experienced larger inhibitory than excitatory events evoked by CC stimulation, but after ION-X, inhibition was reduced onto reciprocally connected S1BC neurons and remained strong for M1 output neurons (Fig. 4). Indeed, this reduction of inhibition onto reciprocally connected S1BC neurons may have reduced GABAergic shunting of callosal inputs, resulting in a shift toward PPF in S1BC, but not M1, output neurons (Fig. 4). Although the amplitude of CC-mediated Sr 21 mEPSCs was unchanged onto PV 1 interneurons, it is likely another group of interneurons play a more important role in IHI. It has been demonstrated that interneurons in L1, likely neurogliaform cells, produce IHI via GABA B receptors on the apical dendrites of L5 pyramidal cells (Palmer et al., 2012). Further study would be needed to determine whether this reduction in IHI between intact and deprived S1BCs is indeed mediated by activity in this subgroup of L1 interneurons. GABAergic interneurons help network pyramidal neurons decide when to fire (Pouille and Scanziani, 2001;Wehr and Zador, 2003) and increase their firing rates during tasks (Cardin et al., 2009;Sohal et al., 2009). These functions reduce pyramidal firing and fine tune their output to other brain regions (D'Souza and Burkhalter, 2017;Petersen, 2019). The reduced inhibition onto reciprocally connected S1BC L5 neurons and increase of inhibition onto MI projecting cells may be a way for the deprived cortex to enhance direct reciprocal connectivity without affecting intrahemispheric outputs.
The increased direct reciprocal connectivity after ION-X likely has a systems level function. One purpose of the CC is to integrate unilateral stimuli into a complete representation of the external world (Pietrasanta et al., 2012). To do this, it must target excitatory and inhibitory neurons (Toyama and Matsunami, 1976;Payne and Siwek, 1991;Makarov et al., 2008;Rochefort et al., 2009;Palmer et al., 2013) to shape receptive fields (Watroba et al., 2001), accurately tune to salient stimuli (Rock and Apicella, 2015) and coordinate appropriate outputs; the CC often targets L5 neurons to achieve these results (Shuler et al., 2001(Shuler et al., , 2002. What would be the purpose of increasing the direct reciprocal connectivity? There have been three proposals for the role of the takeover of deprived cortex. One is to increase cortical territory to aid in processing. The most dramatic examples of this phenomenon is language re-lateralization after damage to the dominant hemisphere (Cao et al., 1999;Finger et al., 2003) and somatosensory processes using the visual cortex for processing in blind people (Cohen et al., 1997;Merabet et al., 2008). A clear reason for the takeover after unilateral sensory loss has not been established. However, the fact that the largest effect detected in the present study underlies the cortices ability to send information between the two sensory hemispheres is consistent with an adaptive takeover hypothesis. A second proposal may be that the activation of the deprived cortex is maladaptive and plays a role in Figure 5. S1BC-projecting neurons have increased output responses to columnar inputs. A, Experimental setup with bipolar electrode stimulating CC fibers medial to the rAAV-labeled output neuron. I/O curves were generated for each output cell type, with responses recorded at regular intervals from the maximum response stimulus intensity. B, Summary diagram of detected changes in deprived S1BC after ION-X. C-F, After ION-X, S1BC output neurons have steeper I/O curve slope, while other output groups are not significantly altered. (n) = number of cells, *p , 0.05 t test; for values, see Extended Data phantom limb or prosthesis rejection. The lack of plasticity to pain pathways (S2 or ACC) argues against this in this model of denervation induced plasticity. Finally, it has been proposed that the recruitment of deprived cortex may help protect this deprived cortex from takeover from neighboring somatosensory regions like those that respond to nose or forepaw stimuli (Pluto et al., 2005;Yu and Koretsky, 2014). Such a model might require an ongoing callosal signal to report on the status of the deprived cortex. It is unknown whether these circuit level adaptations have a behavioral output, but these results are valuable to guide future studies in models of peripheral injury. Here, we have demonstrated that unilateral denervation alters cortical circuitry with a high degree of specificity to enhance bilateral responsiveness to intact senses.