Deafness induces complete crossmodal plasticity in a belt region of dorsal auditory cortex

Many neural areas, where patterned activity is lost following deafness, have the capacity to become activated by the remaining sensory systems. This crossmodal plasticity can be measured at perceptual/behavioural as well as physiological levels. The dorsal zone (DZ) of auditory cortex of deaf cats is involved in supranormal visual motion detection, but its physiological level of crossmodal reorganisation is not well understood. The present study of early‐deaf DZ (and hearing controls) used multiple single‐channel recording methods to examine neuronal responses to visual, auditory, somatosensory and combined stimulation. In early‐deaf DZ, no auditory activation was observed, but 100% of the neurons were responsive to visual cues of which 21% were also influenced by somatosensory stimulation. Visual and somatosensory responses were not anatomically organised as they are in hearing cats, and fewer multisensory neurons were present in the deaf condition. These crossmodal physiological results closely correspond with and support the perceptual/behavioural enhancements that occur following hearing loss.


| INTRODUCTION
It is well known that the loss or significant damage to one sensory system (e.g. blindness, deafness) leads to the substitution of the remaining, intact sensory systems in a process termed 'crossmodal plasticity'. Behavioural manifestations of crossmodal plasticity are evident in human history, with the works of blind poets such as Homer and Milton or the performances of numerous blind musicians including Louis Braille.
Such compensatory behaviors have been localized to specific brain regions by several recent neurological studies. Using behaving cats, deaf cats were shown to demonstrate enhanced performance of spatial localization tasks (especially in the visual periphery) as well as superior visual movement detection when compared with hearing controls . However, this enhanced performance was blocked when the posterior auditory field (PAF), or the dorsal zone of auditory cortex (DZ), respectively, were deactivated. Since then, the connectional basis for crossmodal plasticity in each of the areas has been examined (PAF=Butler et al., 2016;DZ=Kok et al., 2014). The connectivity of the PAF was surprisingly similar for early-deaf and hearing animals (Butler et al., 2016). For DZ, most connectivity was essentially the same among early-deaf and hearing cats, except that significant increases in inputs were seen from two adjacent visual areas (Barone et al., 2013;Kok et al., 2014). Thus, in the case of area DZ, there is both behavioral and connectional evidence for crossmodal plasticity following deafness. These observations are further supported by physiological recordings of crossmodal visual responses in DZ of congenitally deaf cats (Land et al., 2016). This study, however, focussed extensively on the maintained functional connectivity of the auditory pathway using Cochlear Implant (CI) stimulation, did not compute the proportion of neurons showing crossmodal effects, and used CI-delivered clicks to evaluate multisensory interactions. Furthermore, the participation of the other intact sensory modality-somatosensation-was not examined. Therefore, the present study sought to assess, following early hearing loss, the level of crossmodal reorganisation occurring in dorsal zone (DZ), the distribution of those responses, and the potential for multisensory processing using natural visual and somatosensory stimuli for multisensory testing.
What is known about area DZ in hearing animals is relevant to what might be expected to occur after congenital or early hearing loss. Cartographically, the region is a narrow horizontal band that caps the dorsal border of primary auditory cortex or A1. At that location, it is sandwiched between A1 (laterally) and extra-striate visual areas (Palmer et al., 1978). Auditory response features of DZ include longer latency responses and broader frequency tuning than found in A1, as well as responses to complex auditory cues (He et al., 1997;Stecker et al., 2005). In addition, a large proportion (>76%) of DZ neurons exhibit visual and/or somatosensory influences that were differentially distributed in the region (Merrikhi et al., 2022. Furthermore, approximately 57% of neurons showed multisensory convergence, of which many demonstrated multisensory integration in response to multisensory stimulation. From this high proportion of non-auditory processing in this portion of auditory cortex of hearing animals, and because crossmodal plasticity largely occurs as a re-weighting or re-distribution of existing inputs, hearing loss should result in high levels of crossmodal plasticity in DZ.

| Overview
Responses of neurons in auditory area DZ to sensory and multisensory stimuli were assessed in 10 adult cats (Felis catus; Liberty Labs, Waverly, NY). The hearing group consisted of six cats (all females), and the deaf group was composed of four cats (two males and two females) that had undergone perinatal ototoxic deafening prior to 1 month. All animals were housed in an enriched colony environment. All experimental procedures were conducted in compliance with the National Research Council's Guidelines for the Care and Use of Mammals in Neuroscience and Behavioral Research (2003), the Canadian Council on Animal Care's Guide to the Care and Use of Experimental Animals (Olfert et al., 1993) and were approved by the Animal Use Subcommittee of the University Council on Animal Care at the University of Western Ontario.

| Deafening procedures
Ototoxic deafening procedures were conducted on four animals around the time of hearing onset (14 days postnatally; Shipley et al., 1980). Deafness was induced by the co-administration of kanamycin and ethacrynic acid, which is known to destroy cochlear hair cells (Xu et al., 1993), producing rapid, profound, bilateral hearing loss. Loop diuretics such as ethacrynic acid have also been shown to minimally affect vestibular end-organ function (Elidan et al., 1986), and visual examination of the animals in the current study showed no obvious vestibular deficits. A detailed account of deafening procedures has been described in detail elsewhere (Kok et al., 2014). Briefly, animals were injected with kanamycin (300 mg/kg, s.c.) and were presented with auditory stimulation whereas ethacrynic acid was administered (35-60 mg/kg, i.v., to effect) until auditory brainstem responses (ABRs) showed no acoustically evoked activity (i.e. a flat ABR). Follow-up ABRs were conducted 3 to 6 months later to confirm deafness ( Figure 1).

| Surgical preparation
Approximately 1-2 weeks before electrophysiological recording, animals underwent surgery to implant a head holder onto the frontal bone, perform the craniotomy and build up a recording well over DZ and the surrounding auditory, visual and somatosensory cortices using dental acrylic. The afternoon prior to surgery, animals were fasted and lightly anaesthetised with ketamine (4 mg/kg, i.m.) and Dexdomitor (.05 mg/kg, i.m.), in order to facilitate the insertion of an indwelling feline catheter into the cephalic vein for intravenous anaesthetic administration during the surgery. Each animal also received a dose of anti-inflammatory medication (dexamethasone, .05 mg/kg, i.v.) to reduce post-surgical inflammation.

| Surgical procedures
On the day of surgery, the animals were administered with atropine (.02 mg/kg., s.c.) to minimise respiratory and alimentary secretions, acepromazine (.02 mg/kg, s. c.), buprenorphine (.01 mg/kg, s.c.), Cefazolin (35 mg/kg, i.v.) and dexamethasone (.5 mg/kg, i.v.). Sodium pentobarbital (25 mg/kg to effect, i.v.) was then administered to induce general anaesthesia, followed by supplemental doses as needed. In order to inhibit the gag reflex, the mucosa of the pharynx was anaesthetised with a topical anaesthetic (Cetacaine; Cetylite Laboratories, Pennsauken, NJ, USA), and the trachea was intubated with a cuffed endotracheal tube in order to ensure adequate ventilation. Respiration was unassisted. Ophthalmic ointment (Neosporin; Kirkland, Quebec) was applied to the cornea to prevent desiccation. The animal was positioned into a stereotaxic frame (David Kopf Instruments; Tujunga, CA), and the head was fixed by palato-orbital restraints and blunt (non-rupture) ear bars, whereas the body rested on a water-filled heating pad in order to maintain core temperature at 37 C. The animal was then prepared for surgery using antiseptic procedures. Body temperature, respiration rate, heart rate, blood pressure and end tidal CO 2 were monitored continuously throughout surgery. A midline incision was made in the scalp, and the right temporalis muscle was detached medially and reflected laterally. A craniotomy was made over the right hemisphere between Horsley and Clarke (1908) coordinates A0-A15, in order to expose auditory cortex, the middle suprasylvian sulcus (MSS) and anterior somatosensory areas. Following this, an acrylic recording well was built up around the craniotomy and sealed closed with dental cement. A head holder was attached to the frontal bone of the skull using bone screws and dental acrylic. The animal was then provided with standard postoperative care (see Malhotra et al., 2004). In all cases, recovery was uneventful.

| Preparation for recording
Approximately 1-2 weeks later, electrophysiological recording procedures were initiated. Animals were administered with atropine (.02 mg/kg, s.c.), dexamethasone (.5 mg/kg, s.c.), acepromazine (.4 mg/kg, i.m.) and ketamine (35 mg/kg, i.m.). This same anaesthetic regimen has been used in numerous animal studies to record F I G U R E 1 Auditory brainstem responses (ABRs) for (a) a hearing adult animal and (b) an early-deafened animal tested at 6 months of age in response to 2000 click presentations at sound pressure levels ranging from 5 to 80 dB SPL. All responses are scaled to 1 μV. Absence of evoked responses in the early-deaf cat confirms the profound level of hearing loss induced by the early (14 days old) ototoxic treatment. and quantify multisensory responses (e.g. Allman & Meredith, 2007;Carriere et al., 2007;Perrault et al., 2005;. The trachea was intubated with a cuffed endotracheal tube in preparation for ventilation. Indwelling feline catheters were inserted into the saphenous vein bilaterally, as well as the right cephalic vein. Phenylephrine and atropine drops were administered to each eye, a clear feline contact lens with an optimal focal distance of 25 cm was inserted into the left eye (contralateral to craniotomy), and an opaque lens was inserted into the right (ipsilateral) eye. The left eye lid was sutured open to ensure the eye remained open for the duration of recording procedures. Expandable foam ear buds (sound sources) were inserted bilaterally within the ear canals near the tympanic membrane. Next, the ear canals and pinna were packed with Otoform (Betavox, Sherbrooke, QC), an expandable silicone material, to dampen/block acoustic noise exterior to the earbuds. The animal was then secured to a stereotaxic frame using the previously implanted head holder, so no wounds or pressure points were present and a continuous infusion of ketamine (8-10 mg/kg/h) and acepromazine (.04-.05 mg/kg/h) was administered. The recording well was unsealed, and the dura was reflected in preparation for recording. A layer of silicone oil was applied to the cortex to prevent desiccation. Baseline respiratory and physiological measures were recorded, and the animal was placed on a ventilator. Expired CO 2 was monitored and maintained at $4-5%. The animal was then paralysed with Nimbex (cisatracurium besylate; induction: 1.5 mg/kg, i.v.; constant infusion: 1.5 mg/kg/h, i.v.), in order to prevent ocular drift and movement of the limbs away from the somatosensory stimulators (described below). A warm water circulating pad (Gaymar, Orchard Park, NY) was used to maintain core body temperature. Animals were hydrated with constant infusions of anaesthetic and paralytic in 2.5% dextrose/half-strength lactated Ringer's solution. Dexamethasone (1.0 mg/kg, i.v.) and atropine (.03 mg/kg, s.c.) were administered on a 24-h schedule for the duration of the experiment. Finally, a digital image of the exposed cortex was taken with the aid of a surgical microscope to record the position of each electrode penetration relative to cerebral vasculature and cortical topography.

| Experimental design
Electrophysiological recordings were conducted within a double-walled sound chamber on an electrically shielded, vibration-free table (Technical Manufacturing Corporation, Peabody, MA, USA). Animals were exposed to electronically generated, repeatable auditory (A), visual (V) and somatosensory (S) stimuli, presented both alone and in combination (auditory-somatosensory: AS, auditory-visual: AV, visual-somatosensory: VS and auditory-visual-somatosensory; AVS) in pseudo-random order ( Figure 2a). Because this study is designed to survey as many neurons as possible using multiple singleunit recordings, such techniques also made it difficult to F I G U R E 2 The experimental paradigm and cortical recording sites. (a) Recording from cat's auditory cortex during anaesthesia. In pseudo-random order, animals were presented with different auditory, visual and somatosensory stimulations, either in isolation (i.e. A, S, V conditions) or combination (i.e. AS, AV, VS and AVS conditions). Electrode penetration sites (white circles; left photomicrograph) and tracks (black trace; right photomicrograph) in the dorsal zone of auditory cortex (DZ) for representative (b) hearing and (b) deaf cats. SMI-32 labelling profiles determined borders between different cortical areas (shown by dashed black lines). Abbreviations: A, anterior; A1, primary auditory cortex; AAF, anterior auditory field; aes, anterior ectosylvian sulcus; D, dorsal; L, lateral; mss, middle suprasylvian sulcus; pes, posterior ectosylvian sulcus; PLLS, posterolateral lateral suprasylvian area; PMLS, posteromedial lateral suprasylvian area. Scale bars: 1 mm.
adjust stimulus qualities to maximise responses of an individual neuron. Although it is well known that spatial, temporal and physical stimulus features affect multisensory responses, such complex manipulations were not attempted here and a standard stimulus set was employed to allow comparison of responses across neurons. Auditory stimuli (white noise bursts, 1-32 kHz, 500 ms duration, 65 dB SPL) were presented binaurally via the earbuds using closed-field transducers (EC1; Tucker-Davis Technologies, Alachua, FL, USA) and were digitally generated with a 24-bit digital-to-analogue converter at 156 kHz (RX6; Tucker-Davis Technologies). Acoustic signals had 5 ms rise and fall times and were cosine squared gated. Because this study is focussed on the non-auditory features of DZ, no attempt was made to characterise the specific auditory response or receptive field properties that have been detailed in other reports (Stecker et al., 2005). Visual, full-screen flashes (dark-tolight; 80 lx, 500 ms duration) were programmed in Adobe Flash and presented using a 17-in. liquid crystal monitor placed $25 cm in front of the animal. This stimulation was demonstrated to be effective at eliciting robust responses during control recordings made in the nearby visual areas: anterior/posterior lateral suprasylvian cortices (data not shown).
Somatosensory stimuli were presented using allceramic bender actuators (PL140.10; PI Ceramic, Auburn, MA, USA) with a displacement distance of 1 mm. These stimuli were demonstrated to be effective at eliciting robust responses during control recordings made in adjoining regions of somatosensory cortex (data not shown). Three stimulators were placed in contact with the animal's body and arranged to stimulate three distinct sensory nerves: (1) contralateral vibrissae (contralateral trigeminal nerve), (2) ipsilateral vibrissae (ipsilateral trigeminal nerve) and (3) contralateral forepaw (radial nerve). Furthermore, the somatosensory stimulation sites on the head and forepaw were chosen because they match the body regions reported in the only studies that have mapped somatosensory receptive fields in auditory cortex (Fu et al., 2003;. Somatosensory stimulation involved the simultaneous activation of all three stimulators. When activated, the tactile stimulators made a soft but audible 'click' noise, and the following steps were taken to resolve this issue. First, as noted above, after insertion of the expandable earbuds, the ear canals and pinna were packed with Otoform to further block external noise. Second, a fourth tactile stimulator was positioned close to, but not touching, the contralateral ear of the animal to provide a control condition for noise made by the tactile stimulators. This sham stimulation was presented alone, as well as in combination with A, V and S stimuli to determine if the 'click' noise elicited a response by itself or if it modulated neuronal responses to other sensory stimulation. When compared with levels of spontaneous activity, none of the examined neurons in hearing cats (n = 482) showed a detectable response to the sham actuator when presented alone. Furthermore, combining a sensory stimulus (A, V, or S) with the sham actuator (i.e. A/sham, V/sham, S/sham) likewise failed to reveal a statistically detectable effect. Therefore, the presence of a soft 'click' concurrent with the somatosensory stimulus neither activated nor modulated the DZ neurons in this study, nor would such a soft stimulus be expected to generate responses in deafened animals with auditory response thresholds in excess of 90 dB (see section 2.2).
When multisensory stimuli (AV, AS, VS, AVS) were presented, the auditory and somatosensory components were programmed to occur simultaneously, whereas the visual stimulus was programmed to precede each (or both) of them by $40 ms for the hearing and $65 ms for the deaf subjects (Allman & Meredith, 2007;Foxworthy et al., 2013;Kayser et al., 2008). This is because cortical response latencies for the visual system are typically longer than those for the auditory and somatosensory systems, particularly for non-primary regions (Bullier & Nowak, 1995;Carrasco & Lomber, 2011). Finally, to measure spontaneous, or baseline activity, control trials were randomly presented in which no stimulus was presented. This spontaneous activity, as well as evoked responses, were assessed within specific temporal windows, as described below.

| Data acquisition
Neuronal responses to multisensory stimuli were collected using an iridium axial array microelectrode (AM-002, 200 μm diameter; FHC, Bowdoin, ME), on which 12 electrode sites are spaced linearly 150 μm apart. Impedance measures ranged from 1 to 3 MΩ. Neuronal activity was classified based on band-pass filtering as spikes (300-5000 Hz). All activity was amplified (10,000X) and digitised at 25,000 Hz (RZ2; Tucker-Davis Technologies). The recording electrodes targeted area DZ lateral to the suprasylvian sulcus, as shown in Figure 2b,c. The electrodes were inserted orthogonal to the exposed surface of cortex and lowered to a depth no further than 1800-2000 μm to avoid entering the bordering extrastriate visual areas (anterolateral and posterolateral lateral suprasylvian areas (ALLS and PLLS; see Figure 2b,c). Typically, 9-13 DZ recording penetrations were performed, and recording sessions ranged in duration from 71 to 97 h.

| Histological procedures
At the end of the experiment, animals were administered an anticoagulant (heparin, 10,000 U; 1 mL) and a vasodilator (1% sodium nitrite, 1 mL), and were overdosed with Euthanol (sodium pentobarbital, 50 mg/kg, i.v.). Animals were perfused intracardially through the ascending aorta with physiological saline (.01 M PBS), followed by fixative (4% paraformaldehyde) and 10% sucrose. The brain was stereotaxically blocked, removed, photographed, and placed in 30% sucrose until it sunk. The brain was frozen and cut in 60-μm coronal sections using a cryostat. Every second section was processed with the monoclonal antibody SMI-32 (Covance; Princeton, NJ, USA) to determine auditory and visual cortical borders (Mellott et al., 2010;van der Gucht et al., 2001). Another series of sections were stained with cresyl violet and used to visualise electrode tracks (Figure 2c). Only recording sites that could be confirmed as lying within DZ were analysed. A further confirmation of DZ recording status was performed by measuring DZ auditory response onset latency, which demonstrates longer average response latencies than in A1 (Stecker et al., 2005).

| Data analysis
All units were de-noised and waveforms were sorted manually in principal component space using Plexon offline sorter (Plexon, Dallas, TX, USA). Only units which achieved significant levels of separation in principal component space and showed a clear refractory period were classified as single units. All data analysis was performed in Matlab (Mathworks, Natick, MA, USA). Peri-stimulus time histograms (PSTHs) have a time resolution of 1 ms and, for better visual display, were smoothed using a 25 ms window. Given the relatively long duration (500 ms) of the stimuli, neurons often displayed two distinct periods of response: (1) early-response epoch (0-100 ms after stimulus onset) when there was a 'typical' large-magnitude response to stimulus onset and (2) lateresponse epoch (250-600 ms after stimulus onset) when sustained activity occurred. The two epochs were characteristically separated by a $ 150 ms period of suppression of spiking activity, which was not included in the present analysis (Merrikhi et al., 2022. Thus, a response to any stimulus or stimulus combination was measured as the average number of spikes that occurred during the early-and/or late-response epochs across different trials. These response measures are reported hereafter as spikes per second (spk/s).
Previously published methods were used to evaluate single unit data and were adapted as necessary (Allman & Meredith, 2007;Foxworthy et al., 2013;Kayser et al., 2008;Meredith et al., 1987;Sarko et al., 2013;Stanford et al., 2005;Wallace et al., 2004). To identify the stimulus modality capable of driving the neurons, first responses of neurons during either early-or late-response epochs in conditions A, V and S were compared to the average response during the corresponding period of baseline (no stimulus) using Wilcoxon sign-rank test. If this comparison reached a significance level of p < .0083 (the Bonferroni correction applied), the stimulus modality was considered capable of activating the neuron. To determine whether a multisensory stimulus elicited multisensory integration, a neuron's responses to a multisensory stimulus (i.e. AV, AS, VS and AVS) were compared to the those elicited by the most effective unisensory stimulus (termed 'UNI_BEST' response). From this comparison, if a significant ( p < .0083; the Bonferroni correction applied) difference was observed, then the response was defined as showing multisensory integration (MI). The magnitude of multisensory integration was calculated: MI = ([CM-SM UNI_BEST ]/SM UNI_BEST ) Â 100%, where CM is the response to the combined-modality stimulation and SM UNI_BEST is the response to the most effective single modality stimulus (Meredith & Stein, 1983;Stein et al., 1993;Stevenson et al., 2014). Integrated responses that showed increased activity were termed response enhancement, and those that showed decreased responsivity were identified as response depression. These quantitatively determined categorisations were confirmed by visual inspection of the PSTHs and rasters. All statistical comparisons between two auditory areas were performed using the two sample T-test. Unless otherwise stated, mean ± standard error are reported in the text.
To analyse the anatomical distribution of neurons or neuron response types within DZ, we associated each recording site with positional coordinates along two axes: dorsal-ventral (DV) and anterior-posterior (AP). Coordinates were based on the DZ dimensions of approximately 1 mm in the DV axis and $5 mm in the AP axis. Consequently, each recording site was assigned one of two possible values on DV axis and one of five values on the AP axis. The recording depth of each neuron was also logged, and these measures were used to examine feature distribution across cortical depth. The DV, AP and depth distributions of DZ neurons were statistically tested using the Spearman correlation coefficient. The anatomical distributions of different neuron response types were statistically analysed using one way analysis of variance (ANOVA), with the DV, AP or depth coordinates of neurons as the dependent variable. For illustration purposes, the distribution of neurons has been linearly interpolated in Matlab.
Ultimately, the data were tabulated in a spreadsheet that listed neuronal responses to sensory and multisensory stimulation (from which the neuronal response-type was based) during the early-response and late-response epochs and the multisensory integration (enhancement, depression and magnitude of response change) that occurred. This spreadsheet was used to calculate the incidence of different response categories of neurons as well as their multisensory properties.

| RESULTS
A total of 315 single units from the DZ of four early-deaf cats were examined using visual, auditory, somatosensory and combined stimulation (the stimulation paradigm is illustrated in Figure 2a). Four hundred eighty-one DZ neurons were similarly tested from six hearing cats. Histologically verified recording location for both hearing conditions is shown in Figure 2b,c. These data were used to determine the response types of neurons encountered in the two different hearing conditions, assess their proportional occurrence, plot their anatomical distribution in DZ, and assess features of their multisensory processing.

| Early-deaf DZ sensory responses
Neurons in early-deaf DZ were not activated by auditory stimulation (as would be expected) but were activated by visual and/or somatosensory cues. Responses to the relatively long-duration stimuli (500 ms) characteristically occurred within an early-response epoch (0-100 ms) and a late-response epoch (250-600 ms) (see section 2). The proportion of neurons activated by visual and/or somatosensory stimulation during the different response epochs is shown in Table 1. In early-deaf animals, visual responses were most frequently (52.7% of sample) observed during both the early-and late-response epochs, but somatosensory responses largely (7.4% of sample) occurred only during the late-response epoch. These proportions of temporal activation were not significantly different to that observed (for the same modalities) in the hearing animals.
Examples of sensory responses for neurons from early-deaf DZ are shown in Figure 3, accompanied by neurons in hearing animals with similar response behaviour. Figure 3a shows a clear response of a neuron in early-deaf DZ to a visual stimulus (early: V = 3.9 ± .3 spk/s, late: V = 2.9 ± .1 spk/s) that was not significantly modulated by other sensory stimulation. In Figure 3b, a similar visual sensitivity is shown for a neuron from a hearing animal (early: V = 6.3 ± .3 spk/s, late: V = 2.8 ± .1 spk/s). These response patterns are consistent with neurons functionally identified as unisensory visual.
In other neurons (shown in Figure 3c,d), clear responses to a single sensory modality were significantly modulated by the co-presentation of a stimulus from a different modality, which are identified as subthreshold multisensory neurons. Figure 3c shows a neuron from an early-deaf cat that was activated by visual stimulus in both the early-and late-response epochs (early: V = 9.9 ± .6 spk/s, late: V = 6.2 ± .2 spk/s), and the late-period response was significantly enhanced (late: VS = 7.1 ± .2 spk/s, p = .004) by a somatosensory stimulus that was ineffective when presented alone. Similar examples of subthreshold multisensory neurons were identified in hearing animals. As illustrated in Figure 3d, a visual response (V = 16.4 ± .4 spk/s) was significantly suppressed (VS = 15.1 ± .4 spk/s, p = .006) by the presentation of an otherwise ineffective somatosensory stimulus.
Neuron responses were also identified that showed independent activation by two different sensory modalities, which were defined as bimodal neurons. The neuron shown in Figure 3e from an early-deaf animal was activated by a visual stimulus presented alone and by a somatosensory stimulus presented alone during both the early-(V = 18.2 ± .8 spk/s, S = 3.2 ± .3 spk/s) and late-

| Early-deaf DZ sensory response proportions
When the proportion of neurons in each of the sensory response types was calculated for those sampled in DZ of early-deaf animals, unisensory visual neurons were predominant (77.5%) while comparatively few multisensory neurons were observed (bimodal = 16.8%; subthreshold = 5.7%). In terms of sensory modality F I G U R E 3 Responses of single neurons for the three classes of neurons from DZ of deaf (left) and hearing (right) cats. The coloured lines over each graph show the duration and modality of the stimulus (green = visual(V); blue = somatosensory (S); red = auditory (A)). Bar graphs shown at the right side of each PSTH plot the average response during the early-and late-response epochs for different stimulus conditions (i.e. A, V, S or VS). On these bar graphs, the dashed lines over the coloured bars (unisensory responses) represent the neuron's baseline activity. Statistically significant unisensory responses are designated with coloured asterisk. The dashed line over gray bars (multisensory responses) represents the average response of the UNI_BEST stimulus. Statistically significant enhancement is designated with grey asterisk (**p < .01, ***p < .001). All PSTHs were binned at a resolution of 1 ms and were smoothed using a 25 ms window. All error bars plotted indicate standard error of the mean. Unimodal visual neurons from DZ of (a) deaf and (b) hearing cats. These neurons were driven only by visual stimulus and their visual responses were not significantly modulated by other stimuli from different modalities. Subthreshold multisensory neurons from DZ of (c) deaf and (d) hearing cats. Both of these neurons are activated only by a visual stimulus, but the visual responses of these neurons are significantly modulated by the presence of a somatosensory stimulus during the late-response epoch. The neuron from the early-deaf DZ (c) shows response enhancement while the neuron from the hearing cat (d) demonstrates response depression. (e,f) Bimodal neurons. In Figure 3e, this neuron from an early deaf animal was activated by somatosensory and by visual stimuli presented alone, and showed significant response enhancement during the late-response epoch when the same stimuli were combined (VS). In Figure 3f, this neuron from a hearing animal was activated by visual and by auditory stimuli presented alone and showed significant multisensory response depression during the late-response epoch when visual stimulus was combined with somatosensory stimulus. DZ, dorsal zone; VS, visual-somatosensory; PSTH, peri-stimulus time histogram. responsiveness, neurons in the early-deaf DZ were predominantly visual (100%), while a smaller proportion (21. 2%) were also sensitive to somatosensory cues. In contrast, for hearing animals, the predominant response category was that of the bimodal multisensory neuron (54.5%), with unisensory auditory (24.7%), unisensory visual (7.5%), subthreshold multisensory (6.2%) and trimodal neurons (7.1%) also present. These data are summarised in Figure 4 and Table 2. When compared across the two hearing conditions, these data revealed a profound, tenfold increase in the proportion of unisensory visual neurons in the early-deaf DZ (D = 77. 5%; H = 7. 5%) and the proportion of neurons with somatosensory sensitivity more than doubled (D = 21.2%; H = 9.8%). On the other hand, bimodal neurons, which represented 54.5% of neurons in hearing DZ, were reduced in prevalence in the deaf (D = 16.8%) and only visualsomatosensory examples of bimodal neurons were observed. Thus, early-deafness resulted in a profound increase in non-auditory responsiveness that affected 100% of the sampled DZ neurons (see Figure 4b) while reducing the multisensory features of DZ.
F I G U R E 4 The proportion of different neuronal response types of DZ from early-deaf (red bars; n = 4 cats) and hearing (blue bars; n = 6 cats) animals. (a) While bimodal neurons (54.8%) were the majority of neurons in hearing group, the most common class of neurons in the early-deaf animals was unimodal visual neurons (75.7%). In deaf cats, bimodal (17.3%) and multisensory subthreshold (Subthr.; 7%) comprised a lower proportion of the DZ multisensory neurons than in hearing animals, where multisensory neurons were predominant: bimodal (54.8), subthreshold (%6.8%) and trimodal (7.3%). (b) In hearing cats, a significant portion (76.6%) of the DZ population demonstrated non-auditory influences, whereas in deaf animals, all neurons exhibits non-auditory responses. DZ, dorsal zone.
T A B L E 2 The percent of population for classes of neurons in area DZ of deaf (black numbers) and hearing (grey numbers) animals along with their baseline activity (B) and average responses to the different sensory stimuli during the early-and late-response epochs. Note: Statistically significant differences between responses of hearing and deaf to the same stimulus or stimulus combination are designated with asterisk. DZ, dorsal zone. ***p < .001, **p < .01, and *p < .05.

| Early-deaf DZ sensory response levels
The spiking activity evoked by visual, somatosensory or combined visual-somatosensory stimulation was fairly similar across the different response types in both hearing groups, as shown in Table 2. However, there was a significant exception among visual responses of bimodal neurons in deaf DZ, which were significantly (p < .01) higher than all the other visual responses in both earlydeaf as well as hearing animals (see Tables 2 and 3).

| Anatomical distribution of neuron response types early deaf DZ
We used the AP-DV coordinates for each penetration (see section 2) to plot the anatomical distribution of the different response types of neurons in DZ of early-deaf and hearing cats. The distribution of all responsive neurons showed that they were uniformly sampled across the AP-DV surface (D: r = À.1, p = .817; H: r = À.4, p = .431 and depth (D: r = .4, p = .139; H: r = .2, p = .606) in both treatment groups. When examined by neuron response type, unimodal visual, subthreshold or bimodal neurons in early-deaf cats failed to show any AP, DV or depth-related distributional trends, as depicted in Figure 5b-d, In contrast, significant distributional trends were apparent in DZ of hearing subjects (one way ANOVA; effect of AP: F = 14.1, p < .001; effect of DV: F = 4.5, p = .001; effect of depth: F = 3.4, p = .008). Specifically, in hearing cats, unimodal visual neurons were located posteriorly in DZ (one way ANOVA, multiple comparisons; unimodal visual vs bimodal: p < .001; vs subthreshold multisensory: p = .007; vs unimodal auditory: p < .001; vs trimodal: p < .001) as shown in Figure 5b. Subthreshold neurons from hearing cats also showed a distributional trend because they were segregated more superficially in DZ (one way ANOVA, multiple comparisons; subthreshold multisensory vs unimodal auditory: p = .004; vs trimodal: p = .046). Additional distributional trends in hearing DZ were observed for bimodal ( Figure 5c) and trimodal neurons (data not shown). Thus, while several distributional patterns were evident in hearing DZ, those patterns were not evident in the organisation of the early-deaf DZ.

| Integration of visual and somatosensory responses in early deaf DZ
Although it has already been shown (above) that the proportion of multisensory neurons in early-deaf DZ is significantly reduced from that of hearing animals (D = 23.2%, H = 68.9%, p = .004), it is important to examine how multisensory information is processed in the multisensory neurons that do occur in the early-deaf condition. Figure 6a shows the response of a bimodal neuron from early-deaf DZ that was driven by visual (early: V = 18.2 ± .8 spk/s; late: V = 14.3 ± .4 spk/s) and by somatosensory (early: S = 3.2 ± .3 spk/s; late: S = 4.2 ± .2 spk/s) stimuli. When these same stimuli were combined, this neuron showed multisensory response enhancement not during the early response epoch but in the late-response (VS = 15.9 ± .4 spk/s, p = .007; MI = 11.1%) when compared to that elicited by the UNI_BEST stimulus (late: V = 14.3 ± .4 spk/s). By comparison, the bimodal neuron from the hearing DZ depicted in Figure 6b was activated by auditory stimulus during both response epochs (early: A = 15.2 ± .5 spk/s, T A B L E 3 Comparison of visual response between bimodal neurons in area DZ of deaf animals (black numbers) and other classes of neurons in hearing (grey numbers) animals. Note: Statistically significant differences are designated with asterisk. DZ, dorsal zone. ***p < .001, **p < .01, and *p < .05.
late: A = 10.8 ± .4 spk/s) and also by visual stimulus during the late-response epoch (Figure 6b; late: V = 8.4 ± .4 spk/s). Moreover, when visual and somatosensory stimuli were combined, this neuron showed multisensory response suppression during late-response epoch (Figure 6b; late: VS = 6.0 ± 0.3 spk/s, p < .001; MI = À27.7%). Thus, these and other multisensory neurons in DZ of both early-deaf and hearing animals were capable of generating multisensory integration. Although neurons in early-deaf DZ were capable of integrating multisensory stimuli, the effect was observed in a minority of neurons in the sample. Combined visual-somatosensory stimulation evoked significant response interactions in 22.2% of the multisensory neurons from early-deaf DZ during either the early-or lateresponse epoch. However, this particular stimulus combination generated integrated multisensory responses in only 3.1% of visual-somatosensory neurons in hearing DZ. These effects are visible in the scatter plots shown in Figure 7, where comparisons of the combined stimulation response are plotted against the UNI_BEST response for each DZ multisensory neuron during the early-( Figure 7a) and late-response epochs (Figure 7b Table 4 summarises the average response, incidence and magnitude of integration observed in early-deaf and hearing DZ multisensory neurons in response to VS stimulation during both response epochs. These comparisons show that although the response levels are significantly higher for multisensory neurons in early-deaf DZ than in hearing animals, the occurrence and magnitude of multisensory integration in response to visual-somatosensory stimulation is not different. In contrast, in hearing animals, integration of AV responses are frequent (44.6%) and vigorous (average MI = 29%) (Merrikhi et al., 2022 but are not included here because similar sensory modality comparisons with early-deaf responses cannot be made.

| DISCUSSION
These results show that the pattern of sensory responsiveness found in neurons for DZ of hearing cats (>92% are responsive to auditory stimulation) is profoundly changed by early-deafness (0% responsive to auditory; 100% F I G U R E 5 Anatomical distribution of the different response types of area DZ in early deaf (left) and hearing (right) cats. (a) Each recording site (white circles) was assigned a coordinate location according to anterior-posterior (AP) /dorso-ventral (DV) grid (white grid). The heatmaps show the distribution of (b) unimodal visual, (c) bimodal (c) and (d) subthreshold neurons for early-deaf (left heatmaps) and hearing (right heatmaps) animals. Bar plots (far right) compare the average AP coordinates of each class of neurons between deaf (D) and hearing (H) animals. By comparing the left and right heatmaps for the different response types of neurons, no trends in anatomical organisation was apparent for the early-deaf cases, although significant distributional differences were observed in the hearing animals. DZ, dorsal zone. responsive to visual stimulation of which 21% are also sensitive to tactile cues). Such crossmodal plasticity following early hearing loss has been functionally observed in a variety of other animals and brain areas: humans (Auer et al., 2007;Finney et al., 2001), cat anterior auditory field (Meredith & Lomber, 2011), ferret A1 and anterior auditory field , and mouse A1 (Hunt et al., 2006). Crossmodal plasticity has also been examined in DZ of congenitally deaf cats, where a small proportion of visual responses were observed, which were not segregated within the DZ (Land et al., 2016). However, somatosensory effects were not examined, and sampling and testing methodologies were dissimilar from the present study such that further comparisons cannot be made. Studies of crossmodal plasticity that used comparable methods and animal models are compared in Table 5. Together, these studies show that early hearing loss results in total replacement by the remaining sensory modalities. Curiously, that replacement in some areas is dominated by vision (DZ, FAES) while others by somatosensory (AAF). However, the dominant replacement modality seems to correspond with its proportional presence in the same areas in hearing animals. For example, more visual responses (than somatosensory) were observed in DZ and FAES of hearing animals, and those areas were dominated by visual responsivity in early-deaf animals. Likewise, the only non-auditory responses observed in AAF of hearing animals were somatosensory, and that modality dominated activity of the early-deaf AAF. These observations provide further support for the hypothesis that crossmodal plasticity following early hearing loss is largely the result of 'unmasking' or 'upregulation' of existing nonauditory inputs to an auditory region (Bavelier & Neville, 2002), rather than the ingrowth of novel connections (Meredith et al., 2016) or territorial expansion (Meredith & Lomber, 2017).
Although the result of crossmodal plasticity is strongly evident in the present study, there was a substantial loss of multisensory neurons and multisensory processing in early-deaf DZ. Multisensory neurons (bimodal and subthreshold) were predominant (68.9%) in hearing DZ, but represented only 23.2% of the neurons in the early-deaf. This underscores the fact, despite their frequent use, that the terms 'crossmodal' and 'multisensory' are not synonyms and are not interchangeable. The F I G U R E 6 Bimodal neurons from (a) early-deaf and (b) hearing DZ showing their responses to separate auditory (A, red), visual (V, green), and somatosensory (S, blue) stimulation as well as to combined visual-somatosensory (VS. grey) cues. The bar graphs plot the average response of the neurons to different sensory stimulations during the early-(left) and late-response (right) epochs. Dashed lines over coloured bars show the neuron's baseline activity. Statistically significant responses by unisensory stimulations are designated with coloured asterisks. Dashed lines over grey bars show the response of neurons to UNI_BEST stimulus. Statistically significant enhancement or depression is designated with grey asterisk (**p < .01, ***p < .001). The bimodal neuron from the (a) deaf animal was driven by both visual and somatosensory stimuli presented alone and showed significant response enhancement to VS stimulation during the late-response epoch. The bimodal neuron from the (b) hearing cat was activated by the auditory stimulus during both the early-and late-response epochs, whereas the visual stimulus drove the neuron during the late-response epoch. This neuron demonstrated significant response suppression in response to the combined VS stimulus during the late-response epoch. DZ, dorsal zone. loss of multisensory neurons has important implications. Most higher-order cortical areas show a predominance of multisensory neurons (reviewed in Meredith et al., 2020), and area DZ in hearing animals is regarded as a higherorder area (Land et al., 2016). However, the threefold loss in multisensory neurons following deafness indicates that the general organisational principles of cortex may not be applicable for cortical systems reared with abnormal sensory experience. Reduced multisensory neurons also suggests that the role of multisensory processing is reduced in crossmodally reorganised regions. However, although the incidence of multisensory neurons in early-deaf DZ was substantially reduced, proportionally more of them showed significant response changes after multisensory stimulation than did their hearing counterparts (D = 26% showed multisensory integration; H = 4.2%). That said, the magnitude of multisensory integration was not different between early-deaf and hearing DZ neurons. For visual-somatosensory neurons in DZ, multisensory integration ranged from 33.4 to À28.2% in early-deaf DZ and 27.3 to À32.5% in the hearing. For visual-auditory interactions in hearing DZ, multisensory integration averaged from 6.3 to 10% (Merrikhi et al., 2022). A possible explanation for these low levels of multisensory integration F I G U R E 7 Integration of visual and somatosensory responses in multisensory DZ neurons during different response epochs. Responses of these neurons to VS stimulation are plotted against their UNI_BEST responses (x-axis) in scatter plots for early-deaf (left column) and hearing DZ (middle column). The inset histograms in each scatter plot show the percent of population plotted against the response change (VS -UNI_BEST). The histograms in right column demonstrate the distribution of magnitude of response change (%) for early-deaf (red) and hearing DZ (blue) for integrating (filled histograms) and non-integrating (hollow histograms) neurons. (a) During the early-response epoch, early-deaf DZ neurons show non-significant response change to VS stimulation whereas hearing DZ neurons exhibit response depression ('**'; p < .01). Although a few percent of DZ multisensory neurons in early-deaf and hearing DZ show response enhancement (right red histograms) and depression (right blue histogram), respectively, during the early-response epoch; however, these values are not statistically significant. (b) During the late-response epoch, both early-deaf and hearing DZ multisensory neurons exhibit non-significant response changes to VS stimulation. Similar to Figure 7a, during the late-response epoch, the deafness did not significantly change the average magnitude of multisensory integration of DZ neurons despite the observation of a small portion of neurons showing response enhancement in early-deaf (right red histogram) and depression in hearing DZ (right blue histogram). DZ, dorsal zone. may be derived from their comparison with integration levels generated in other areas. For example, in hearing animals, multisensory integration can achieve response changes of 325% in area FAES , while those in the superior colliculus (SC) can exceed 1000% (Meredith & Stein, 1983). The FAES and SC are strongly connected (Meredith & Clemo, 1989), and high levels of integration in those structures are clearly consistent with their roles in detection and localisation behaviours. On the other hand, it has been postulated that low levels of multisensory integration are consistent with a role in the perceptual binding of multisensory objects (Bizley et al., 2016). The present observations would suggest that multisensory integration in earlydeaf DZ, limited as it is, could be consistent with the latter.
The present study also examined the anatomical distribution of crossmodal responses in early deaf DZ. However, the organisational patterns observed for sensory responses in hearing animals was not observed. Specifically, for hearing DZ, visual responses were segregated posteriorly in the structure, presumably near the sources of visual cortical inputs, whereas somatosensory responses were clustered anteriorly, near presumed sources of somatosensory cortical inputs (Merrikhi et al., 2022. In contrast, no distributional trends were observed for visual or somatosensory responses in the early-deaf DZ that essentially filled the entire extent of the region. These observations are consistent with the effects of late hearing loss in A1 and AAF of ferrets, which showed no crossmodal somatotopy nor any residual effects of tonotopic organisation by the somatosensory replacement modality (Allman et al., 2009). Thus, regional patterns of neuronal distribution laid down during normal development do not appear to influence the organisation of crossmodal plasticity, either before or after the closure of critical developmental periods. What remains to be determined is how the nervous system derives adaptive information from these disorganised arrangements. T A B L E 4 The average response, incidence (Inci.; as a % of sample) and magnitude (% response change) of multisensory integration (MI) observed for multisensory enhancement (Enh.) or multisensory depression (Depr.) in early-deaf (black numbers) and hearing (grey numbers) DZ multisensory neurons. Note: All values are determined from responses to VS stimulation during either the early-or late-response epochs. Statistically significant differences are designated with asterisk. DZ, dorsal zone; VS, visual-somatosensory. ***p < .001, **p < .01, and *p < .05.
T A B L E 5 Proportion (percentage) of neurons showing auditory, visual or somatosensory responsiveness (includes unisensory and bimodal neurons, so total exceeds 100%) in hearing and early-deaf animals by area.