Examples of convergence of visual and auditory, or visual and somatosensory, inputs onto individual neurons abound throughout the brain, but substantially fewer incidences of auditory–somatosensory neurons have been reported. The present experiments sought to examine auditory–somatosensory convergence to assess whether there is a feature of this type of convergence that might obscure it from conventional methods of multisensory detection. Auditory–somatosensory convergence was explored in cat anterior ectosylvian sulcus (AES) cortex, where higher-order somatosensory area IV (SIV) and auditory field of the anterior ectosylvian sulcus (FAES) share a common border. While neuroanatomical tracers documented a projection from FAES to SIV, physiological studies failed to reveal the bimodal neurons expected from such cross-modal connectivity. Stimulation of FAES through indwelling electrodes also failed to excite any of the SIV neurons examined. However, when stimulation of auditory FAES was combined with somatosensory stimulation, a large majority (66%) of SIV neurons showed a significant response attenuation. FAES-induced response suppression was specific to SIV, could not be elicited by activating other auditory regions and was blocked by the microiontophoretic application of the GABAergic antagonist bicuculline methiodide. Based on these data, a novel, cross-modal circuit is proposed involving projections from auditory FAES to somatosensory SIV, where local inhibitory interneurons ‘reverse the sign’ of the cross-modal signals to produce auditory–somatosensory suppression. This form of excitatory–inhibitory multisensory convergence has not been reported before and suggests that the level of interaction between auditory and somatosensory modalities has been substantially underestimated.
Even though multisensory neurons have been identified within a host of cortical areas (e.g. Jiang et al., 1994a,b; Benedek et al., 1996; Duhamel et al., 1998; Xing and Andersen, 2000; Schroeder et al., 2001), the neural circuitry underlying multisensory convergence is not well understood and broad organizational principles for multisensory processing have yet to be proposed. The overwhelming number of cortical multisensory studies has reported the convergence of visual and auditory, or visual and somatosensory inputs (e.g. Calvert, 2001). In contrast, comparatively little is known about neuronal responses to converging auditory and somatosensory information. Except for recent studies of somatosensory inputs to auditory cortex (Schroeder et al., 2001; Schroeder and Foxe, 2002; Brett-Green et al., 2003) and several attentive or cognitive studies (Jousmaki and Hari, 1998; Caclin et al., 2002; Fujiwara et al., 2002; Guest et al., 2002; Hotting et al., 2003), published accounts of auditory–somatosensory convergence or integration are comparatively rare. Furthermore, when neurons receiving auditory and somatosensory inputs have been reported, they showed the lowest incidence among the different patterns of multisensory convergence (Meredith and Stein, 1986; Wallace et al., 1992, 1993; Jiang et al., 1994a,b). Although multisensory neurons represented ∼55–60% of the neurons in the deep layers of the cat superior colliculus, the proportion of auditory–somatosensory neurons was only 4% (Meredith and Stein, 1986); auditory–somatosensory neurons in the cat anterior ectosylvian sulcal cortex comprised only 4% as well (Wallace et al., 1992); for the primate superior colliculus, the ratio was 27% multisensory to 0.9% auditory–somatosensory (Wallace et al., 1996). Despite this low level of occurrence, these reports also indicated that auditory–somatosensory convergence and integration operate under the same constraints as the other modality combinations. Therefore, further examination of this infrequent phenomenon will provide insight not only into the incidence of this particular pattern of sensory convergence, but will also contribute to our understanding of the basic principles of multisensory convergence and integration in general.
The anterior ectosylvian sulcus (AES) is one of but a few cortical locations that hosts adjoining representations of the auditory and somatosensory modalities and because the probability of connectivity increases with physical proximity (Young et al., 1995), this area seemed well suited for investigating potential cross-modal connections between these sensory modalities. The AES, as depicted in Figure 1, is found at the junction of the frontal, parietal and temporal regions of the cat cortex and is composed of a horizontal (anterior) limb and a vertical (posterior) limb. Within the dorsal bank of the horizontal limb, there is a representation of the body surface identified as somatosensory area SIV (Clemo and Stein, 1982, 1983, 1984). Receptive field reversals and cytoarchitectonic differences distinguish SIV from SII on the adjacent anterior ectosylvian gyrus (Burton et al., 1982; Clemo and Stein, 1983). To its medial/deep side, SIV transitions into an area of the AES fundus termed para-SIV, where neurons have somatosensory receptive fields that are not topographically arranged, expand to include hemi-body and whole-body representations, and exhibit multisensory responses (Clemo and Stein, 1983). Posterior to SIV the somatosensory representation transitions into the auditory field AES (FAES) (Clarey and Irvine, 1986, 1990a,b; Meredith and Clemo, 1989; Wallace et al., 1992, 1993; Korte and Rauschecker, 1993; Rauschecker and Korte, 1993; Middlebrooks et al., 1994; Rauschecker, 1995, 1996; Benedek et al., 1996). The FAES occupies the anterior and posterior banks of the vertical limb of the AES and is distinguished from both the anterior auditory field and AI by its lack of tonotopic organization, the presence of neurons broadly tuned for sound frequency (Clarey and Irvine, 1986; Korte and Rauschecker, 1993; Rauschecker and Korte, 1993; Middlebrooks et al., 1994; Rauschecker, 1996) and the sensitivity to the spatial features of auditory cues (Korte and Rauschecker, 1993; Rauschecker and Korte, 1993; Middlebrooks et al., 1994; Nelken et al., 1997). Of these two regions of the AES, SIV is perhaps the best understood because it contains a topographic map of the body surface, connects heavily to other somatosensory cortices and exhibits distinct cytoarchitectonic properties. Therefore, the somatosensory SIV cortex was selected to test the possibility that it receives cross-modal inputs from the adjacent auditory FAES and, if so, to assess the functional nature of that connection.
Materials and Methods
All procedures were performed in compliance with the Guide for Care and Use of Laboratory Animals (NIH publication 86-23) and approved by the Institutional Animal Care and Use Committee at Virginia Commonwealth University.
Cats (n = 9) were anesthetized with sodium pentobarbital and their heads placed in a stereotaxic frame. Under aseptic surgical conditions, a craniotomy and durectomy were performed to expose the AES cortex. In two of four cases in which retrograde tracers were injected into SIV, standard extracellular recording techniques were used to map the extent of SIV, as described elsewhere (see Clemo and Stein, 1983). To make tracer injections, a modified electrode carrier was used to support a 5 µl Hamilton syringe and its needle (31 gauge). For injections into SIV, the carrier was angled 50–60° with a 90° cant (lateral to medial) and the needle tip was inserted 2.0–2.5 mm into the dorsal bank of the AES. For injections into FAES, the carrier was angled 53–60° (from vertical), with 35–40° cant (posterior from coronal plane) and the needle tip was inserted at a point 0.8–1.5 mm anterior to the vertical limb of the AES to a depth of 5.25–5.70 mm. After a 5 min pause, horseradish peroxidase (HRP; 20% in sterile saline; 125–425 nl total volume) or biotinylated dextran amine (BDA; 10 000 mol. wt, lysine fixable, 10% in 0.1 M phosphate buffer, 345–700,nl) was pressure injected at a rate of 15 nl/min. The needle was then retracted, the cortex was covered with gel foam, the skin around the wound sutured closed and standard postoperative care was provided.
After a 24 h survival period, the animals injected with HRP were given a barbiturate overdose and perfused intracardially with heparinized saline followed by fixative (1.0% paraformaldehyde, 1.25% glutaraldehyde solution). The brain was blocked stereotaxically, removed and refrigerated overnight in 0.1 M phosphate buffer. Serial, coronal sections (50 µm thick) were cut using a vibrating microtome. One series of sections, at 200 µm intervals, was reacted using a standard diaminobenzidine (DAB) procedure to demonstrate the zone of maximal uptake around the injection site (LaVail and LaVail, 1974). A second series of sections, at 100 µm intervals, was processed using tetramethylbenzidine (TMB; Mesulam, 1978) in conjunction with nickel–cobalt intensification. Standard procedures were used to mount, counterstain and coverslip the tissue.
For animals that received BDA injections there was a 7–14 day post-injection survival period before they were given a barbiturate overdose and perfused intracardially with heparinized saline followed by fixative (4.0% paraformaldehyde, 0.5% glutaraldehyde solution). The brain was blocked stereotaxically, removed and stored in 0.1 M phosphate buffer with 25% sucrose (for cryoprotection) at 4°C until the tissue sank. Coronal sections (50 µm thick) were cut using a freezing microtome and collected serially. One series of sections, at 200–250 µm intervals, was processed for visualization of BDA using the avidin–biotin peroxidase method, according to the protocol of Veenman et al. (1992) and then intensified using silver nitrate (1.42% at 56°C) and gold chloride (0.2% at room temperature). The sections were mounted and coverslipped without counterstain. An additional series of sections, taken at 200–250 µm intervals, was counterstained using a standard cresyl violet protocol to assist in cytoarchitectonic and laminar identification.
Neuronal labeling was visualized using a light microscope (Nikon Optiphot-2) and the data plotted using a PC-driven digitizing stage controlled by Neurolucida software (MicroBrightField Inc.). A calibrated tracing of each section outline with the border between gray and white matter, the position of the injection site, labeled neurons and axon terminals was produced. Injection sites were defined as the large aggregate of densely labeled cell bodies, dendrites and axons at the terminus of the injection needle track. HRP-labeled neurons were identified by the presence of dense black beads within the cytoplasm of the soma and proximal dendrites. BDA-labeled neurons were sharply black throughout their soma which sometimes spread into the distal dendrites, although some reactive neurons revealed a lighter, reddish-brown label. BDA-labeled axon terminals appeared as sharp, black swellings at the end of thin axon stalks or as symmetrical varicosities along the course of an axon.
Tissue outlines and injection sites were plotted at 40× magnification. Labeled neurons and boutons were plotted at 200× magnification and the Neurolucida software kept a count of numbers of identified neurons and boutons. Labeled cell bodies were plotted in the region of the posterior AES cortex. The average diameter was determined by measuring the longest and shortest widths that passed through the cell center, and dividing by two. A shrinkage factor of 7.5% was then applied, as determined by the average difference in tissue area measured before and after processing. Labeled boutons were plotted within the anterior and middle AES cortex, in regions corresponding to that of SIV and para-SIV.
Once the locations of labeled neurons/boutons were plotted for a given section, the cortical laminae were traced from an adjacent, cresyl-violet-stained section and imported back to the digitized plot. The software then calculated the number of labeled neurons or boutons in supragranular (above layer IV), granular (layer IV) and infragranular (below layer IV) locations for a specific area of tissue. The number of granular, supragranular and infragranular boutons was compiled for each case and the ratios of boutons in each laminar region were calculated.
Acute (n = 5) as well as chronic preparations (n = 6) were used because, at the initiation of this portion of the study, it was not known which anesthetic would best permit recordings from these cortical areas. Both surgical procedures were essentially the same, but asepsis was maintained for animals scheduled to recover from the surgery. Animals were anesthetized (acute = isoflurane, n = 3; pentobarbital, n = 2) (chronic = sodium pentobarbital, n = 6) and their heads were placed in a stereotaxic frame. The animals were intubated through the mouth and artificially ventilated. Expiratory carbon dioxide levels were monitored and kept to within 4.0–4.5%. A heating pad was used to maintain temperature, monitored rectally, at 37°C. An intravenous line was placed in the saphenous vein to administer additional medications, fluids and, if appropriate, supplemental anesthesia.
A craniotomy was performed to expose the cortex around the AES. A stainless steel recording well (14 mm inner diameter) was placed over the opening for recording access to SIV in the chronic experiments, as well as to support the animal’s head without obstructing the eyes, ears or face. Stimulating electrodes (parylene coated tungsten, <1 MΩ, two rows 1 mm apart, four electrodes in each row at 1 mm spacing, angled 45–50°) were implanted in FAES to a depth 6.2–7.2 mm from the surface of the middle ectosylvian gyrus (MEG). Additionally, control electrodes were positioned either in other auditory (AI, nine cases) or somatosensory (SV, two cases) cortices. Implanted electrodes were banked with gel foam and secured in place with dental acrylic. For chronically implanted animals, the scalp was sutured closed around the implant, routine postoperative analgesic and antibiotic care was provided, and ∼5–7 days elapsed before the initial recording experiment. The acute preparations were moved immediately into the recording phase.
For chronically prepared animals, recording experiments were conducted in a fashion similar to that described in detail in Meredith and Stein (1986, 1996). Briefly, recording was initiated by anesthetizing the animal (ketamine/acepromazine, n = 3; isoflurane, n = 3) and securing the implant to a supporting bar. These animals, as well as the acutely prepared ones, were intubated and ventilated. The saphenous vein was cannulated for administration of fluids and supplemental anesthetics, as needed. Each of the implanted stimulating electrodes was connected, through a stimulation isolation unit (Grass SIU-5), to a Grass stimulator (S-88). For recording, a glass insulated tungsten electrode (tip exposure < 20 µm, impedance < 1.0 MΩ) was supported by a 16 gauge steel cannula which, in turn, was guided by the X–Y slide (10 mm A-P/M-L excursion; Kopf Instruments) that covered the well. In acute preparations, no cannula was used and the recording electrode was inserted at a 40–50° lateral–medial angle directly through the craniotomy.
In both preparations, neuronal activity was amplified, displayed on an oscilloscope and played on an audiomonitor. Neurons were first identified by their responses to manually presented stimuli: somatosensory (puffs of air through a pipette, brush strokes and taps, manual pressure and joint movement); auditory (claps, clicks, whistles and hisses); visual (flashed or moving spots or bars of light from a hand held ophthalmoscope projected onto the translucent hemisphere and dark stimuli from a rectangular piece of black cardboard). In addition, a neuron’s somatosensory receptor type (hair, skin, deep or joint) was determined and its receptive field was mapped. Next, to assess whether a neuron received excitatory (orthodromic) input from the FAES, each FAES electrode (and control AI or SV electrode) was individually activated (1 anodal pulse, 0.1 ms duration, 500–600 mA current intensity, 3–5 iterations) in a fashion identical to that used in our previous studies (Clemo and Stein, 1984; Wallace et al., 1993; Meredith, 1999).
SIV Sensory Tests
To determine whether somatosensory responses of SIV neurons were influenced by auditory stimuli or auditory FAES activation, somatosensory stimuli were presented within a neuron’s receptive field alone and then in combination with a natural auditory cue or with electrical stimulation through one of the FAES electrodes. Somatosensory stimuli were produced by a narrow (3 mm diameter) metal shaft moved by an electronically driven, modified shaker (Ling 102A) as described in previous studies (Clemo and Stein, 1984; Meredith and Stein, 1986, 1996). Free-field auditory cues were electronically generated white noise bursts, 100 ms duration, 54–70 dB SPL on ‘A’ level. This stimulus was delivered through a speaker, mounted on 18″ diameter hoop [as also described in previous studies (Meredith and Stein, 1986, 1996)] and was positioned to be in spatial register with the somatosensory receptive field. During combined sensory presentations, the onsets of the stimuli were adjusted to compensate for the differences in the average latency of auditory and somatosensory responses in AES cortex [somatosensory evoked responses ∼15 ms later than auditory (Wallace et al., 1992)]. FAES activation was effected by electrical stimulation delivered through an indwelling FAES electrode (trains of pulses at 25–250 Hz, 80–100 ms duration, 0.1 ms/pulse, 37–600 µA anodal current intensity). Separate and combined modes of stimulation were interleaved to compensate for possible shifts in baseline activity. When somatosensory and electrical stimuli were combined, the electrical activation of FAES preceded the somatosensory stimulus by 85–100 ms to avoid interference between stimulation artifact and neuronal action potentials. An interstimulus interval of 10 s was employed to avoid habituation; each test was repeated 10–14 times. Neuronal activity was digitized (rate > 25 kHz) using Spike2 (Cambridge Electronic Design) software and stored for off-line analysis.
Off-line analysis consisted of using Spike2 software to examine data files and, within each file, only reliably isolated waveforms were subjected to further analysis. For a given spike waveform, the file was collated according to the testing conditions and a perstimulus-time histogram was constructed for each. From these histograms, the duration of a response was determined and the mean spike number per response (and standard deviation) was calculated over that period. Responses to the combined stimuli were statistically compared (paired t-test, P < 0.05) to that of the most effective single stimulus, and responses which showed a significant difference were defined as response interactions (Meredith and Stein, 1983, 1986, 1996; Wallace et al., 1992). In addition, the magnitude of a response interaction was determined by the following formula: C – T/T × 100 = %, where C is the response to the combined stimulus, and T is the response to the tactile stimulus alone (Meredith and Stein, 1983, 1986, 1996).
The depth of each identified neuron within a penetration was noted and recorded. Several recording penetrations were often performed in a single experiment and each successful recording penetration was marked, at its conclusion, with a small electrolytic lesion (0.5–1.0 mA for 0.5–2.0 s). At the conclusion of an acute experiment, or after a series of experiments using a chronically-prepared animal, the animal was given a barbiturate overdose and perfused with physiological saline followed by 10% formalin. The brain was removed, frozen sections (50 µm) were cut in the coronal plane through the recording and stimulation sites, processed using standard histological procedures and counterstained with cresyl violet. A projecting microscope (Bausch & Lomb) was used to trace sections and to reconstruct stimulating electrode positions, recording penetrations and the locations of examined neurons from the lesion sites. Gyral/sulcal patterns, cytoarchitectonic characteristics and observed physiological properties were used to demarcate the relevant functional cortical subdivisions. These anatomical data were then tabulated with the physiological observations (described above).
In Vitro Slice Recordings
These procedures were essentially the same as reported previously (Meredith and Ramoa, 1998). Two animals were deeply anesthetized and, immediately prior to euthanasia, their AES cortex was exposed and surgically removed en bloc. This tissue was immediately submerged in 5°C modified (sodium free, 10 mM glucose) Ringer’s solution. Parasagittal sections 400 µm thick through the AES were taken using a Vibratome. Slices were transferred to an interface-type recording chamber where extracellular multiunit recordings were made using glass-insulated tungsten electrodes (∼1 MΩ at 1 KHz, tip exposure <20 µm). Once a neuron was isolated, its electrical activity was amplified, routed to an oscilloscope and audiomonitor, and stored on tape. Next, a site within the same column as the recording electrode, or a location anterior or posterior was stimulated electrically (concentric bipolar electrode, 250 µm diameter, 0.1 ms duration, 600 µA maximal stimulus intensity) and unit activity was recorded in response to 50 stimulus iterations (2–3 s interstimulus interval). Whether a neuronal response was observed or not, the same stimulation paradigm was repeated after the local administration (picospritzed) of bicuculline methiodide (BIC; 50 µM), the co-administration (through two separate picospritzing pipettes) of BIC and d-2-amino-5-phosphonovaleric acid (d-APV; 40 µM) and then the co-administration of BIC with 6-nitro-7-sulphamoylbeno(f)-quinoxaline-2,3-dione (NBQX; 10 µM). Following each experiment, the taped data was transferred to a computer and the activity of each neuron was analyzed quantitatively (Spike2 software) for its response to electrical stimulation of another region of the slice, whether that response was enhanced by the presence of the inhibitory antagonist BIC, and whether that enhancement could be diminished by the presence of the excitatory antagonists, d-APV and NBQX. Data from the population of neurons sampled was tabulated to assess the general pattern of response to the different treatments.
Microiontophoresis and Recording
Two animals were prepared chronically for recording in the manner described above (see Physiological Studies: Surgical Procedures). On the day of the experiment, each animal was anesthetized (ketamine/acepromazine) and, after the recording well was opened, the dura was reflected to admit the microiontophoresis pipette/electrode assembly. The assembly was constructed from a standard glass insulated recording electrode (impedance 0.6–1MΩ), to which was glued (cyanoacrylic, supported by dental acrylic) a double-barreled pipette (4″ long, 1.5 mm OD capillary glass with microfilament; A-M Systems) with tips pulled and trimmed to reveal an internal diameter opening of ∼10 µm. The pipette tips were positioned, using a dissection microscope and a pair of micropositioners, within ∼40–50 µm of the electrode tip. Approximately 30 min prior to use, a fine needle was used to fill one barrel of the pipette with BIC (10 mM, pH 3.5) while the other received normal saline. The assembly was supported by a stereotaxic carrier. Each barrel of the pipette was attached, via a deeply inserted silver wire, to a separate channel of a microiontophoresis current programmer (WPI model 260). For the barrel containing BIC, the current programmer generated a negative holding current of 15–20 nA, while a positive current of the same value was delivered through the saline channel to balance the net current flow. Once a neuron was identified, its activity was amplified and stored in the manner described above (see SIV Recording) and the following battery of tests (same as described in SIV Sensory Tests, above) was presented. Either the somatosensory probe was used to stimulate the neuron’s receptive field, or an electrical current (100–600 mA, 0.1 ms pulse duration; 100 Hz for 100 ms) was passed through one of the indwelling FAES stimulating electrodes, or the two forms of stimulation were combined in an interleaved fashion (10 repetitions) with 100 ms separating the onsets of the combined stimuli. This stimulation paradigm was presented first prior to the ejection of BIC, then immediately after its iontophoresis into the tissue (ejection current was positive 20–50 nA, continuous for 3–5 min, balanced by a negative current of the same magnitude through the saline channel) and then at regular intervals afterward until the effects of the BIC were no longer apparent. Neuronal activity was collected using Spike2 and analyzed offline using custom software (L. Keniston). This analysis calculated neuronal responses (mean and standard deviation of spikes/response) to somatosensory (alone) and somatosensory with FAES stimulation (combined) and then compared responses for pre-ejection, immediate post-BIC ejection and recovery periods to establish (i) if the somatosensory response was significantly suppressed by the FAES-stimulation, (ii) whether that suppression was blocked by the presence of BIC and (iii) when, during the recovery period, the disinhibition effect had cleared. These data for each neuron were then tabulated and graphed, so that the time course of the BIC effect could be assessed for the sample. Statistical evaluation of neuronal and population responses were conducted using the Wilcoxon signed rank test (P < 0.05).
SIV Injections: Retrograde
Injection of retrograde tracers (HRP, n = 2; BDA, n = 3) into the dorsal bank of the horizontal (anterior) limb of the AES filled the somatosensory SIV to varying degrees. Injections were directed toward the most anterior portion of SIV, where the representation of the head and forelimb reside. This was done to maximize the distance between the SIV injection site and FAES and AEV, thereby avoiding direct contamination of the posterior AES cortex by the injection site or track. In two cases, the injection was entirely confined to the dorsal bank of the AES, but in the other three (larger) injections, tracer also entered the fundus (para-SIV) and portions of the ventral bank. As expected from previous reports (Reinoso-Suarez and Roda, 1985; Barbaresi et al., 1989; Mori et al., 1991), neurons retrogradely labeled from these SIV injections were identified within the ipsilateral somatosensory cortical regions of SI–SIII and SV.
Injection of HRP (or BDA) into SIV also retrogradely labeled neurons in the posterior 3–5 mm of the AES cortex, where the sulcus becomes submerged under AI/AII auditory regions of the MEG. As depicted in Figure 1B,D,E, the dorsal and medial portion of this region corresponds to the location of the FAES, while the more ventral portion represents the AEV. FAES neurons labeled from SIV were visualized using either HRP or BDA techniques, as shown in Figure 2A. The somatodendritic labeling for these and other retrogradely marked neurons was consistent with a pyramidal-type morphology, the average diameter of which was 17.6 ± 1.67 µm (from three cats, n = 231, range = 11–27 µm).
FAES neurons retrogradely labeled from SIV were consistently found along a dorsal–ventral arc along its medial aspect. A large HRP injection that included para-SIV (Fig. 2B) as well as a smaller BDA injection that was restricted to portions of SIV (Fig. 2C) both yielded retrogradely labeled neurons sequestered along the dorso-medial aspects of the FAES. Within this region of the FAES, neurons projecting to SIV predominantly resided in lamina III (∼78% supragranular versus 19% infragranular), as depicted in Figure 2D.
FAES Injections: Anterograde
BDA (10 000 mol. wt) is also a strong anterograde tracer and injections were made in the posterior AES (n = 5) to evaluate the projection to SIV. These injections were distributed such that they collectively covered an arc of tissue from the dorsal regions of the FAES, to its medial/inferior-medial areas and then most inferiorly into the representation of AEV. The four injections that included FAES (Cases 1–4) produced terminal label observed in the adjacent auditory fields (e.g. AI, AII, AAF; see Clarey and Irvine, 1990b), while the most inferior injection (in AEV; Case 5) did not.
Following BDA injections into auditory FAES cortex, labeled axons and axon terminals were sought within somatosensory SIV. Here, networks of labeled axons were frequently observed and numerous BDA-positive boutons were clearly visible, as is evident from Figure 3. As demonstrated in Figure 4, each of the injections that encroached on the medial bank of the FAES (Cases 2–4; corresponding to where neurons retrogradely labeled from SIV were concentrated) consistently produced the densest terminal labeling in SIV. By comparison, injections placed too high (Case 1) or too low (Case 5) in the posterior AES cortex yielded reduced or no terminal labeling in SIV.
Terminal labeling from FAES was found throughout the rostro-caudal and medial–lateral extent of SIV, as also shown in Figure 4. The distribution of this terminal labeling occurred in clusters or patches that were interrupted by regions in which boutons were sparse. Some of the bouton-sparse regions occupied the full thickness of the cortical mantle, as if to segregate FAES-recipient patches. While the areas containing labeled boutons included all cortical laminae, the distribution was not homogeneous. Instead, the regions of heaviest terminal labeling were consistently found in the layers closest to the pial surface. This pattern is illustrated in representative sections from four cases in Figure 5, where FAES-labeled boutons were plotted according to their laminar location in SIV. The densest region of terminal label within SIV was consistently observed in laminae I, II and III, the supragranular layers. Lamina IV showed little terminal labeling from FAES. In the deepest layers (infragranular; V and VI), BDA-positive boutons were again evident. When the number of boutons per layer was tabulated for all cases according to their supra-, infra- or granular location, the predominance of the supragranular projection was apparent: 75% supragranular versus 21% infragranular locations.
Physiology: SIV Responses
The presence of projections from auditory FAES to somatosensory SIV was unexpected, because the wealth of physiological studies of SIV have indicated that it is a unimodal, somatosensory area. Clemo and Stein (1983) used natural stimuli from somatosensory, as well as auditory and visual modalities to probe responses of SIV neurons, and reported SIV was exclusively excited by somatosensory inputs. Subsequent reports have substantiated these observations, although some exceptions have noted examples of visual-somatosensory convergence there (Minciacchi et al., 1987; Jiang et al., 1994a,b). Therefore, electrophysiological recording techniques were used to examine the functional role of this cross-modal projection in a total of 256 neurons located in and near SIV. Upon reconstruction of the recording tracks, 144 of the neurons were identified within SIV as depicted in Figure 6. Of the SIV neurons examined, nearly all (130/144, 90.3 %) were excited exclusively by somatosensory stimulation; visual or auditory stimulation was ineffective in all but six (4.2%) neurons (see Table 1). Similarly, all neurons encountered in SII (n = 44) showed unimodal somatosensory responses, but those observed in para-SIV (n = 56) revealed a variety of sensory responses whose properties are displayed in Table 1.
All SIV neurons responsive to somatosensory stimuli were excited by low-threshold, high velocity stimuli (e.g. air puffs) consistent with activation through hair-type receptors. As illustrated in Figure 7, the different somatosensory receptive fields of the identified SIV neurons were found on all areas of the body. The most frequently encountered receptive field location was that on the forelimb/paw (40.3%; n = 102 of 253 receptive fields), while relatively fewer included locations on the face (21.3%; n = 54/253), trunk (17.8%; n = 45/253) and hindlimb/paw (19%; n = 48/253). The bias toward forelimb/paw receptive fields may have resulted from the enhanced probability of encountering the magnified representation of this body region in SIV (Clemo and Stein, 1983, 1984)
Othodromic Stimulation Tests for Excitation
Because the FAES projection to SIV arises from layer III/V pyramidal neurons that are typically glutamatergic, the possibility that this projection conveyed excitatory auditory inputs to SIV neurons was further tested using orthodromic stimulation. In 11 animals, a total of 78 stimulating electrodes were implanted within the caudal AES cortex. Of these, the tips of 58 electrodes were histologically confirmed to be located within FAES (see Fig. 6B) and 50 revealed exclusively auditory-evoked activity. Only those electrodes which exhibited auditory-evoked activity were used for subsequent combined FAES-stimulation and somatosensory tests (see Combined Somatosensory and FAES Stimulation, below). However, all implanted electrodes were used for orthodromic stimulation tests. By activating each of the indwelling FAES electrodes individually (500–600 µA, 0.1 ms duration), a total of 120 SIV neurons were examined for orthodromic input from FAES. However, none (0/120) of the SIV neurons were excited by FAES stimulation. Similarly, none of the SII (0/44) or para-SIV (0/56) neurons tested was orthodromically activated by FAES stimulation.
Combined Modality Tests for Subthreshold Excitation
Although the projection from FAES did not excite SIV neurons, additional tests explored the possibility that inputs from FAES might carry subthreshold excitatory or suppressive signals to SIV. These tests used a standard multisensory paradigm (see Meredith and Stein, 1986, 1996) to collide excitatory responses evoked by somatosensory stimulation with acoustically-evoked activity. A total of 12 SIV neurons (and 18 para-SIV neurons) were presented a somatosensory cue (within the cell’s receptive field), a free-field auditory cue (in spatial alignment with the tactile stimulus) and the combination of somatosensory and auditory stimuli (within 15 ms of synchrony) in an interleaved fashion. All of these neurons were responsive to somatosensory cues, but none were excited by the auditory stimulus presented alone and none revealed a significant response increase when the two cues were combined. However, in the example provided in Figure 8, an SIV neuron that was unresponsive to an auditory stimulus presented alone had its somatosensory response significantly suppressed by the presence of an auditory stimulus. That suppression was not observed in all of the neurons tested in this manner may result from the presence of anesthesia and/or the possibility that acoustically-induced suppression might be relatively weak and, therefore, might not achieve a statistically significant response change within a small number of stimulus iterations. However, it is important to note that the neuron depicted in Figure 8 also had its somatosensory response suppressed by FAES stimulation, as described below.
Combined Somatosensory and FAES Stimulation
While combined-modality tests suggested that auditory cues could suppress SIV responses to somatosensory stimuli, electrical activation of the auditory FAES documented the effect. In this series of tests, somatosensory SIV responses were examined before and in combination with FAES activation through an indwelling electrode. A representative example is shown in Figure 9, where excitation of an SIV neuron elicited by a tactile stimulus was significantly reduced (by 52%; P < 0.05, paired t-test) when combined with activation of auditory FAES. For the entire sample, combining tactile and FAES stimulation significantly suppressed responses to the tactile stimulus in 66% (51/77) of the SIV neurons and the level of suppression ranged from 24–100%, as summarized in Figure 9C and Table 2. FAES-suppressed SIV neurons were distributed throughout the anterior–posterior and medio-lateral axis of SIV (see Fig. 6, filled circles) and, accordingly, no particular body region appeared to be disproportionately affected (see Fig. 7, black bars). Furthermore, the effect was widely divergent and a single FAES stimulation site was capable of suppressing activity at different SIV locations. For example, each of the widely separated recording penetrations denoted by asterisks in Figure 6 were taken from a single animal and contained neurons whose responses were suppressed from the same FAES stimulation site. Similar examples of divergence were revealed in the four additional animals in which suppression was observed in more than one penetration. In contrast, FAES-suppression effects were not observed in neurons from the adjoining SII and para-SIV representations. As demonstrated in Figure 9D, responses in both SII and para-SIV to somatosensory stimulation were not significantly affected by concurrent FAES stimulation.
Whether the observed response suppression arose from the FAES, or was a general result of stimulation of adjacent auditory cortices, was assessed by comparing SIV responses to FAES stimulation with those evoked from sites in AI (n = 42 neurons). Figure 10A depicts such a comparison, where the tactile stimulation of an SIV neuron was paired with activation of FAES (top row) or auditory area AI (bottom row). In this example, the tactile response was significantly suppressed by the FAES stimulus, but not by the activation of AI. None of the SIV neurons whose tactile responses were suppressed by FAES stimulation were significantly influenced by activation of control electrodes in AI (see Table 2). Together with previous estimates of stimulation current spread (<1 mm; Wallace et al., 1993; Meredith, 1999), these data are consistent with the notion that the observed suppressive effects specifically involved FAES activation.
Simultaneous, dual single-unit recordings indicate that some SIV neurons are preferentially targeted over others by FAES inputs. Figure 10B illustrates records of two neurons recorded at the same time and location within SIV. The tactile response of the neuron in the top row was significantly (P < 0.05, t-test) suppressed by concurrent FAES stimulation, while that of the other neuron (bottom row) was unaffected. This same pattern of FAES suppression/non-suppression was observed in five other pairs of simultaneously recorded SIV neurons.
The nature of FAES suppression of SIV responses was characterized by systematically changing the parameters of the FAES stimulation. These tests revealed that the magnitude of suppression was directly related to the level of stimulation intensity or frequency. For example, the tactile responses of a single SIV neuron depicted in Figure 11 were suppressed by concurrent FAES stimulation and, as the rate or intensity of FAES stimulation was systematically increased, the level of response suppression likewise increased. As shown in Figure 11, this relationship was clearly monotonic for both stimulation parameters. In addition, the same relationship between increased levels of FAES stimulation intensity/frequency and increasing levels of SIV suppression was observed in all the other neurons examined in a similar fashion (n = 5, 7, respectively).
Given that FAES stimulation suppressed SIV responses, the possibility that this effect was mediated through GABAergic inhibition was initially assessed using in vitro techniques amenable to pharmacological manipulation. Parasagittal slices through the longitudinal extent of the AES were used to make extracellular multiunit recordings before and after application of BIC, the GABA antagonist. A total of 29 AES neurons were examined, of which 69% (n = 20/29) were sensitive to BIC treatment. In these cases, neurons that were either initially unaffected (n = 7/20) or minimally activated (n = 13/20) by electrical stimulation became highly excited by this stimulation in the presence of BIC, indicating that inhibitory GABAergic circuits normally suppressed the evocation of excitatory responses. In addition, in those slices whose plane of section preserved connections between FAES and SIV, 58% (n = 7/12) of the SIV neurons that were minimally activated by electrical stimulation of FAES were strongly activated by that same stimulus in the presence of BIC. An example is provided in Figure 12. In this case, the neuron responded to only 5 of 50 FAES shocks. However, when BIC was applied to the preparation, the neuron was strongly activated by the same stimulus and for a longer time period. Co-application of the excitatory antagonist d-APV served to attenuate the duration of the BIC-enhanced response, while co-administration of excitatory antagonist NBQX completely blocked all synaptic activity. These same effects of excitatory antagonists were observed in 5 of seven cells tested with d-APV and in all six neurons presented with NBQX.
While the pharmacologic disinhibition of inputs to SIV neurons in vitro was consistent with inhibition of SIV neurons by GABAergic circuits controlled by FAES, the general results of these tests might be expected from most neocortical regions within which ∼30% of the neurons are GABAergic. To examine the involvement of GABAergic inhibition directly in FAES stimulation-induced suppression of SIV responses, two animals were used to make electrophysiological recordings before and after the microiontophoretic application of BIC. These methods were successful in reducing or blocking FAES stimulation-induced suppression of SIV responses. A representative example is provided in Figure 13, where the response of an SIV neuron to a tactile stimulus was significantly suppressed by concurrent stimulation of the FAES (compare black histogram of FAES suppression superimposed on the grey, tactile-alone histogram). This suppression, however, was almost completely blocked by the microiontophoresis of BIC, as evidenced by the nearly equal responses to the BIC + FAES +tactile combination (black histogram) overlying the BIC + tactile-alone tests (grey histogram). Ultimately, when the tissue had recovered from the effects of the BIC, FAES stimulation induced suppression of SIV responses returned, as evidenced by the widely separated responses to the FAES + tactile (black histogram) versus the tactile-alone (grey histogram) tests. A total of 29 neurons were examined in this fashion, of which 10 completed the entire series of tests, as summarized in Figure 13D.
These experiments have demonstrated a novel form of multisensory convergence, whereby inputs from one modality act to suppress, rather than enhance, activity in another. This effect was specific for auditory–somatosensory convergence resulting from the FAES projection to SIV: FAES activation did not suppress responses in adjacent SII and para-SIV regions nor did stimulation of non-FAES auditory areas have an effect on SIV activity. However, its effect was broadly distributed within SIV rather than topographic. Also, cross-modal suppression, which occurred in 66% of SIV neurons, was directly (not inversely) related to the efficacy of the FAES stimulation. Furthermore, the cross-modal effect was mediated through GABAergic interneurons, since the FAES-induced suppression was initiated through pyramidal neurons in FAES but blocked by iontophoretic application of BIC (the GABA antagonist) in SIV.
Collectively, these observations do not support the likelihood that the FAES projection onto SIV neurons results in the conventional, well-known form of multisensory convergence. At present, the neuronal basis for multisensory convergence and integration has largely been explored in the deep layers of the superior colliculus (e.g. Meredith and Stein, 1983, 1986, 1996; King and Palmer, 1985) and the principles revealed there have been upheld by additional examples identified throughout the neuroaxis and across phylogeny (for a review, see Stein and Meredith, 1993). In these reports, the multisensory neurons that were encountered were excited by a stimulus presented in one modality and excited by an independent cue presented in another (i.e. excitatory–excitatory convergence). Therefore, without combining stimuli for concurrent presentation, it has been possible to establish the multisensory nature of many neurons by their independent, excitatory responses to the different modalities. However, in the present case (and previous studies as well; see Clemo and Stein, 1982, 1983, 1984; Jiang et al., 1994a,b; Wallace et al., 1992; Rauschecker and Korte, 1993), the multisensory nature of SIV neurons was not apparent when stimuli from different sensory modalities were presented alone. In the present study, the multisensory properties of SIV neurons became evident only during combined tests, where excitation elicited through one modality was suppressed by the effects of stimuli from another. Because the nature of the processing controlled by FAES is GABAergic, this ‘new’ arrangement revealed for SIV neurons might best be regarded as an excitatory–inhibitory form of multisensory convergence.
That inhibition plays a role in multisensory processing, however, is not new. Multisensory response depression occurs as a result of inhibition of responses to one modality by stimulation of another (Meredith and Stein, 1983, 1986, 1996; King and Palmer, 1985). Of the reported examples of multisensory response depression, these effects occurred in neurons that were excitable by stimuli from more than one modality. Thus, some excitatory–excitatory inputs apparently carry complex signals whereby one or both excitatory channels are accompanied by inhibitory receptive field surrounds (Meredith and Stein, 1996) or by potent post-excitatory inhibitory periods (Meredith et al., 1987). In this way, a stimulus in one modality that was out of spatial and/or temporal alignment could depress activity initiated by another. These mechanisms for multisensory inhibition do not appear to apply to the present observations. The involvement of spatial inhibition seems unlikely here, because all natural stimulation tests were conducted with stimulus combinations that were presented in spatial and temporal alignment. On the other hand, the suppression between FAES and SIV may represent a cross-modal adaptation of GABAergic circuitry which has been well-documented in modality-specific cortices to underlie the generation of a variety of response properties. For example, local GABA-mediated circuits have been demonstrated to modulate the sharpness of frequency tuning in auditory cortex (e.g. Fuzessery and Hall, 1996; Richter et al., 1999; Wang et al., 2000), orientation tuning in visual areas (Sillito, 1975; Sillito et al., 1985) and spatial inhibition in barrel field/somatosensory cortex (e.g. Chowdhury and Rasmusson, 2002; Li et al., 2002). Therefore, it seems reasonable to speculate that the cross-modal GABAergic suppression observed between FAES and SIV may represent a recruitment of these well-established within-modality connectivity patterns for multisensory purposes. Furthermore, this excitatory–inhibitory form of convergence may underlie a variety of cross-modal inhibitory phenomena, such as that induced by visual motion on vestibular activity in parieto-insular cortex (Brandt et al., 1998).
While the precise organization of the cross-modal excitatory–inhibitory convergence onto SIV neurons is not yet known, numerous clues exist in the present data that are consistent with the connectivity schematized in Figure 14. First, the size, morphology and laminar location of the FAES–SIV projection neurons are consistent with excitatory cortico-cortical connections described for numerous other regions. Next, upon reaching SIV, FAES terminals excite inhibitory interneurons which, in turn, powerfully inhibit principal SIV neurons perhaps by terminations on their soma, proximal dendrites or axon hillock [as documented for different varieties of inhibitory interneurons (see Kawaguchi and Kubota, 1997)]. These arrangements would be consistent with the present in vivo and in vitro experiments demonstrating GABAergic involvement in the FAES suppression of excitatory SIV responses. Furthermore, the results from in vitro slices also revealed excitatory SIV responses to FAES stimulation which appear to be held at subthreshold levels by local GABAergic circuits. Therefore, there is a possibility that principal SIV neurons also receive excitatory inputs from FAES, but do so on their distal dendrites in a manner that is easily regulated by more proximally effected inhibition.
In addition to illustrating the possible neuronal connectivity between FAES and SIV, Figure 14 also summarizes their areal relationships. FAES projection neurons were primarily located in lamina III, although some were also found in lamina V. FAES axon terminals in SIV primarily targeted the supragranular layers, with a smaller proportion of boutons also found in the infragranular layers. Projections that exhibit these primarily supragranular-to-supragranular characteristics do not fall neatly within classical hierarchical connectional schemes (Felleman and Van Essen, 1991). However, projections considered to be ‘feedback’ in nature preferentially target the supragranular layers can also have a bilaminar origin, as was observed here. Therefore, the projection between the FAES and SIV regions most likely represents a feedback hierarchical relationship.
The functional or behavioral role of SIV is unknown. Only one study, published as an abstract (Jiang and Guitton, 1995), has examined SIV in behaving animals and showed that SIV stimulation evokes goal-directed movements of the head and body. Given the lack of behavioral data regarding SIV, the possible function of FAES modulation of SIV is rather speculative. Because FAES neurons encode the spatial components of auditory cues (Korte and Rauschecker, 1993; Rauschecker and Korte, 1993; Jiang et al., 1994a,b; Middlebrooks et al., 1994; Rauschecker, 1996) and SIV contains a map of the body surface (Clemo and Stein, 1983), connections between the two areas could serve to provide updates on the spatial alignments among the two modalities. However, such an effect would argue for a point-to-point or block-to-block connectional relationship between the two areas, which is not consistent with the divergent projection pattern observed here. Alternatively, these connections may play a role in cross-modal selective attention, whereby attention to one modality (e.g. auditory) broadly suppresses activity in another (e.g. somatosensory). Attention to stimuli from one modality is well known to suppress event-related potentials evoked by another (Hillyard et al., 1984; Teder-Salejarvi et al., 1999; Hotting et al., 2003) and the cross-modal connections identified here may provide the neuronal basis for such an effect. This notion is supported further by the anatomical divergence of the projection from FAES to SIV, whereby broad areas of tactile activity in SIV might be modulated by the auditory attentional state of the animal. Along this line of reasoning, because both SIV and FAES are heavily connected with the superior colliculus, it is possible that shifts in attention, for example, toward an auditory stimulus, could reduce activity levels (via FAES–SIV suppression) of cortical somatosensory afferents to the colliculus while elevating those for auditory afferents, thereby enhancing the likelihood of attending to an auditory stimulus. These effects would be consistent with a role in selective attention.
These results underscore some of the fundamental differences between excitatory–excitatory and excitatory–inhibitory forms of multisensory convergence. Unlike the well-known excitatory–excitatory forms of multisensory convergence, the present excitatory–inhibitory multisensory neurons are virtually undetectable as being multisensory using standard extracellular recording and search techniques. Afferents from the inhibitory modality, such as those to SIV neurons from auditory FAES, become apparent only when directly paired against effective stimulation in the other modality (or during high levels spontaneous activity). In the present experiments, sequential single-modality tests revealed a nearly uniform unimodal sample (only 2.3% were multisensory), while combined tests revealed a surprisingly high proportion of auditory FAES-affected SIV neurons (66% were multisensory). Ultimately, the revelation of these ‘silently multisensory’ neurons suggests that the incidence of multisensory processing in the brain, or at least that between auditory and somatosensory modalities, has been vastly underestimated. Consequently, these observations strongly support the assertion, based on large numbers of behavioral and brain imaging studies, that ‘cross-modal interactions are the rule and not the exception . . . and that the cortical pathways previously thought to be sensory-specific are modulated by signals from other modalities’ (Shimojo and Shams, 2001).
We are indebted to Dr A. Ramoa for the use of his equipment to conduct the in vitro slice experiments and to Dr T. Salt for his expertise in microiontophoresis which he generously shared with us. Supported by NIH NS39460 and by the Human Frontiers Science Program RGO174.
Address correspondence to Dr M. Alex Meredith, Department of Anatomy and Neurobiology, Virginia Commonwealth University School of Medicine, Richmond, VA 23298-0709 USA. Email: email@example.com.
|V||1 (0.7%)||0||4 (7%)|
|A||2 (1.4%)||0||3 (5%)|
|S||130 (90.3%)||44 (100%)||38 (68%)|
|AS||3 (2.1%)||0||2 (4%)|
|Unresponsive||8 (5.5%)||0||5 (9%)|
|V||1 (0.7%)||0||4 (7%)|
|A||2 (1.4%)||0||3 (5%)|
|S||130 (90.3%)||44 (100%)||38 (68%)|
|AS||3 (2.1%)||0||2 (4%)|
|Unresponsive||8 (5.5%)||0||5 (9%)|
|Suppression||51 (66%)||0 (0%)||0 (0%)||0 (0%)|
|Facilitation||4 (5%)||0 (0%)||2 (6%)||0 (0%)|
|No effect||22 (29%)||8 (100%)||34 (94%)||42 (100%)|
|Total||77 (100%)||8 (100%)||36 (100%)||42 (100%)|
|Suppression||51 (66%)||0 (0%)||0 (0%)||0 (0%)|
|Facilitation||4 (5%)||0 (0%)||2 (6%)||0 (0%)|
|No effect||22 (29%)||8 (100%)||34 (94%)||42 (100%)|
|Total||77 (100%)||8 (100%)||36 (100%)||42 (100%)|