Current efforts at functional mapping of multisensory neurons are hampered by the need for both cellular-level resolution and the separate visualization of activity by different sensory cues. We have used a recently developed technique that exploits the differential time course of zif268 mRNA versus protein induction in neurons after sensory stimulation. Adult male rats were visually and acoustically deprived and then exposed to one of the following stimulation sequences: (i) no sensory stimulation; (ii) 2 h visual stimulation followed by 30 min auditory stimulation; (iii) 2 h auditory stimulation followed 30 min of visual stimulation; and (iv) 2 h compound visual and auditory stimulation. The neocortex and superior colliculus (SC) were then processed for fluorescent immunocytochemistry and in situ hybridization for staining of Zif268 protein and mRNA products. We have found that activity patterns in primary visual and auditory cortices were in accord with the sequence of the compound stimulus. We also show that SC superficial layers contained a pool of exclusively unimodal neurons, similar to that of visual cortex. Activity patterns of deep SC layers contained multimodal neurons with varying degrees of visual and auditory convergence. The deep SC layers also showed that auditory processing was largely carried out by a small, bimodal group of neurons whereas visual processing was coordinated by both a large unimodal and a small bimodal pool of neurons.
It is well recognized that the capacity of an animal to respond appropriately to biologically significant stimuli requires a convergence of sensory information within the nervous system. An ability to integrate incoming neural signals across two or more sensory modalities serves to both enhance those signals and produce a unitary, coherent representation of the available sensory cues. Multisensory processing is carried out by neurons in various cortical and subcortical regions where they provide the necessary signals to guide attentive and orientation behaviors of an animal.
One of the brain regions where multisensory processing has been well characterized is the superior colliculus (SC). The SC has two key structural features — a laminar structure and a topographic layout (Kanaseki and Sprague, 1974; Stein et al., 1975; Shimozawa et al., 1984). Although the SC is anatomically divided into seven laminae and several sublaminae, it has two functional divisions — superficial (I–III) and deep (IV–VII) layers (Casagrande et al., 1972; Edwards, 1980). The neurons in the superficial layers are exclusively responsive to visual stimulation whereas the deep laminae neurons show significant modality overlap (Stein and Arigbede, 1972; Stein et al., 1976; Meredith and Stein, 1986b). Indeed, it is the deep layers that are primarily involved in multisensory processing within the SC (Stein, 1981). These neurons send their integrated signals to motor and premotor structures of the tecto-reticulo-spinal (TRS) track, which controls the movement and orientation of the eyes, pinnae or head, allowing these receptor organs to be aligned with the sensory cues of regard (Meredith and Stein, 1985; Meredith et al., 1991).
A number of unresolved issues in SC multisensory processing concern the ability to which they integrate information across diverse sensory cues. It has been shown that the responses of SC multisensory neurons to a given sensory stimulus are influenced by the presence or absence of other sensory cues (Meredith and Stein, 1983; Meredith and Stein, 1986a,b; Anastasio et al., 2000). In many SC cells, the pooling of sensory inputs seems to amplify the effects of subtle environmental cues whereas in others, responses to normally effective stimuli can be blocked. The temporal disparity among combinations of various sensory cues has also been shown to be a critical factor, influencing the fidelity of multisensory responsiveness of SC neurons (Meredith et al., 1987). And finally, the importance of the receptive field overlap from different sensory modalities has been demonstrated (Meredith and Stein, 1990,1996; Meredith et al., 1991). It is the spatial coincidence of multimodal sensory stimuli, and the resultant receptive field alignment, that is the crucial determinant for response integration to occur.
Although electrophysiological techniques have provided significant insight into multimodal processing in various brain regions, it is difficult to obtain response profiles over large spatial regions of a particular structure without considerable effort.
The elucidation of response characteristics among multi-sensory neurons should therefore benefit from imaging studies that provide information on sensory integration and spatial layout within a given neural structure. For example, a spatial profile of response integration would reveal details concerning the extent of modality convergence within a neuronal population as well as the topography of their distribution. However, this would require imaging of neural activity that provides a visual display at the cellular level. Furthermore, separate activity profiles are required for each stimulation condition so that they can reveal both modality preference as well as incidences of cross-modal responsivity.
The use of immediate–early gene (IEG) markers provides an opportunity to obtain neural activity profiles at the cellular level in response to sensory stimulation. Two markers that have been routinely used are c-fos and zif268 (Sagar et al., 1988; Dragunow and Faull, 1989; Chaudhuri and Cynader, 1993; Hughes and Dragunow, 1995; Kaczmarek and Chaudhuri, 1997). These markers can be visualized in brain tissue after neuronal activation through two robust and powerful techniques — in situ hybridization and immunocytochemistry for staining of the mRNA and protein products, respectively. However, these products by themselves reveal only a unitary map, i.e. a visual display of activated neurons in response to a single sensory stimulus. A recently developed technique exploits the differential time course of zif268 mRNA versus protein induction after neuronal stimulation (Chaudhuri, 1997; Chaudhuri et al., 1997). This dual-activity mapping technique is based on the fact that mRNA expression peaks early (20–30 min) whereas protein expression peaks late (90–120 min) after onset of stimulation. A stimulation paradigm that exploits this temporal difference in mRNA versus protein expression can be used to obtain separate immediate– early gene induction patterns after two different stimulation sequences.
We have now used this technique to obtain dual-activity immediate–early gene expression profiles of sensory neocortex and SC in response to both visual and auditory stimulation. We show here activity profiles of the SC superficial layers that contain an exclusively unimodal response, similar to that of sensory neocortex, whereas the activity profiles in deep SC contain multimodal neurons with varying degrees of sensory convergence. The activity profiles of the deep SC also showed that auditory processing was largely carried out by a small, bimodal group of neurons whereas visual processing was co-ordinated by both a large unimodal and a small bimodal pool of neurons.
Materials and Methods
Animal Treatment and Tissue Preparation
Adult male Long Evans rats (N = 16; n = 4/condition) were subjected to the following audiovisual stimulation sequences: 2 h of visual stimulation followed by 0.5 h of auditory stimulation (VA), 2 h of auditory stimulation followed by 0.5 h of visual stimulation (AV), 2 h of combined audiovisual stimulation (DS). All of these stimulation sequences followed 3 days of sensory deprivation. Control animals were deprived of sensory stimulation for 3 days (no stimulation — NS). The apparent long duration of sensory deprivation is required to reduce the baseline level of zif268 expression to absolute minimum. This practice enhances the contrast of both mRNA and protein expression between stimulated and control animals (unpublished observations). The animals were individually housed in clear cages inside a box with acoustic barrier. The box was maintained in a photographic darkroom in the absence of any light. Although one cannot preclude self-generated auditory cues from this ‘deprivation’ period, the overall expression levels of zif268 mRNA and protein in the non-stimulated animals served as a baseline of neuronal activity, to which the staining patterns in other experimental groups were compared.
The animals that received sensory stimulation after this period of deprivation did so through a single speaker at the top of the box and a single light source emitting fluorescent light at 20 W. The internal walls of the sensory deprivation chamber were plastered with various visual cues, which were readily visible in the presence of illumination provided by the lamp inside the chamber. The visual cues included sine-wave gratings with low-spatial frequency and high contrast and other geometric shapes in black and white. These cues served as the visual component of the stimulation paradigm in DS, VA and AV conditions. The auditory component of the stimulation paradigm consisted of a broadband noise segment containing a random presentation of all the frequencies audible to a rat (between 2 and 100 kHz).
All animals were decapitated under deep anesthesia with an overdose of halothane (2-bromo-2-chloro-1,1,1-trifuoroethan), their brains were removed, and occipital cortices were blocked. To preserve the mRNA content of the neurons, the cortical blocks were then rapidly frozen in isopentane (2-methylbutane) chilled in liquid nitrogen and then maintained in a –80°C freezer to be used for later cryosectioning. Frozen brains were mounted on specimen holding blocks and sectioned at –20°C into 20 μm thick sections on a Leica CM3050 cryostat. The slide-mounted tissue sections were also kept at –80°C for histological processing (in situ hybridization and immunocytochemistry experiments).
Fluorescent In Situ Hybridization (FISH)
All solutions used for these experiments were prepared using RNase free reagents and diethylpyrocarbonate (DEPC)-treated double deionized water (ddH2O). All glassware and other instruments were RNase-decontaminated using RNase ZAP® solution or wipes (Ambion, Austin, TX). This RNase ‘alert’ approach was carried out throughout the fluorescent in situ hybridization (FISH) experiments until the end of the final post-hybridization washes.
Slide-mounted tissue sections were baked on a slide warmer at 85–95°C for ~5 min. They were then fixed in a 4% paraformaldehyde (PFA) solution made in PBS for 10 min. Following two 5 min washes in PBS, sections were permeablized in PBS solution containing 0.25% Triton X-100 (Sigma, St Louis, MO) for 5min followed by two 5 min washes in PBS. Sections were then acetylated using the following protocol: incubation for 10 min in a 0.1 M triethanolamine (TEA) solution followed by incubation for two 10 min periods in a freshly prepared 0.1 M TEA containing 0.025% acetic anhydride. Sections were then washed in 2× SSC solution for two 5 min periods, racked in a humid chamber, and incubated in 200 μl each of the pre-hybridization solution containing Hybridization Buffer (Sigma-Aldrich Co.) with 40% deionized formamide, at 47°C for 1 h.
After pre-hybridization, each section was incubated overnight at 56°C in 200 μl of the hybridization solution, which contained all the compon-ents in the pre-hybridization solution and ~0.2–0.3 μg of the DIG-labeled cRNA probe. The DIG-labeled cRNA probe was constructed through in vitro transcription using a commercially available plasmid template (mouse EGR1-antisense control template from Ambion), incorporation of hapten-modified uridine using the Digoxigenin RNA Labeling Mix (Roche Scientific, Basel, Switzerland) and the MaxiScript T7 Kit (Ambion) for in vitro transcription reactions. After overnight incubation, sections were washed in graded concentrations of SSC (2×, 1×, 0.5×, 0.2× at 56°C) twice for 10 min each. Signal amplification and visualization was achieved with the Tyramide Signal Amplification (TSA) Direct Green FISH Kit (NEN, Boston, MA) using the manufacturer's protocols.
Fluorescent Immunocytochemistry (FICC)
After the ISH processing was completed, the slide-mounted sections were incubated overnight at 4°C with primary antibody (anti-Zif268, gift from R. Bravo) at a concentration of 1:10 000 in PBS containing 3% normal horse serum (750 μl per slide). Standard immunocytochemical wash procedures were then followed. Sections were incubated with Alexa-594 conjugated anti-rabbit IgG (Molecular Probes; 1:500 in PBS/3% normal goat serum; 2 h; RT), washed, and cover-slipped in ProLong Anti-Fade mounting medium (Molecular Probes). Immunopositive cells were detected by fluorescent microscopy using a rhodamine filter (see Image and Data Analysis).
Image and Data Analysis
Digital images were captured with a Dage MTI cooled three-chip CCD camera attached to a Leica DMLB microscope and analyzed with Adobe Photoshop® 5.0. Immunofluorescent images were captured with Chroma HQ Filters 41001 and 41004 for Alexa-488 and Alexa-594 fluorochromes, respectively. All image processing was carried out using the Adobe Photoshop 5.0 software.
For all FISH, fluorescent immunocytochemistry (FICC) and double-labeled (Double) cell counts, we digitally captured three microscope fields for a given area for each animal on a single slide. We placed considerable effort at ensuring that the captures were taken from approximately the same segments of the brain areas of interest using an established stereotaxic map of the rat brain (Paxinos and Watson, 1998). The intensity value of the background in the NS captures for all four brain regions was used as the threshold for binarizing the 8-bit digital images. Then all the cells matching the criteria of shape (for cell body and nucleus) were counted in the following manner. Since the digital images were composed of three separate channels through which the red, green and blue components of the color micrographs were captured, cell counts were performed three separate times, once in each filter mode (R for immunostaining with the red fluorochrome and G for mRNA detection with the green fluorochrome) and once with both the filters turned on, yielding a digitally superimposed image of both FISH and FICC staining. Given that the double-labeled neurons were FISH- and FICC-labeled as well, the total number of double-labeled neurons was subtracted from the total of each FISH- and FICC-labeled only count under to yield the final cell counts in all brain regions and across all experimental conditions.
The cell count data were then transferred to SuperANOVA statistical analysis software (Abacus Concepts, Inc., Cary, NC) for the analysis of variance and plotting of the bar charts. The dependent variable was expressed as the mean of labeled cells per mm3 to reflect the fact that in almost all cases more than one plane of cells was present in the span of 20 μm of tissue thickness.
The results of the FISH and FICC experiments for detection of zif268 are grouped according to the brain region from which the images were captured. These included primary visual cortex (V1M and V1B), primary auditory cortex (Au1), superficial laminae of the SC (Zo, SuG and Op), and deep laminae (InG, InWh, DpG and DpWh). The approximate coronal plane containing the brain structures used for all data captures was based on an established stereotaxic map of the rat brain (Paxinos and Watson, 1998). Captured images from primary visual cortex and auditory cortex were used as internal controls for sensory activation of the respective regions in all experimental conditions.
In all of the figures presented hereafter, the color red represents FICC staining, the color green represents FISH staining, and the color yellow represents the overlap of both. Zif268 protein staining (red) was generally localized to the nucleus while zif268 mRNA staining was limited to the cytoplasmic region. The green cytoplasmic stain often enveloped the red nuclear stain, giving rise to a region of yellow (dual labeled) within some neurons. In some neurons, however, the cyto-plasmic staining appeared as a green ring or cast around the red nucleus. In these cases the areas of overlap were negligible. However, such cells were still considered double-labeled because of the presence of both cytoplasmic and nuclear staining within the same neuron. Therefore, the criterion for double labeling was based not only on the overlap of red nuclear and green cytoplasmic stains but also on the co-occurrence of the two stains in the same neuron.
A mixed-design 3 × 4 × 4 analysis of variance (ANOVA) was performed on the cell count data pooled over four animals (three counts per animal) with Brain Region and Staining Type as within-subject variables and Condition as the between-subject variable. The variable Staining Type contained three levels (FICC, FISH and Double) while the variable Brain Region contained four levels (visual cortex, auditory cortex, SC — superficial layers, and SC — deep layers). Similarly, the variable Condition consisted of four levels (NS, DS, AV, VA). The ANOVA revealed a significant three-way interaction among the variables [F(18,264) = 41.983, P = 0.0001]. For the sake of simplicity, the bar graph illustrating this interaction was broken up by Brain Region into four separate charts and is shown in the figures in conjunction with the respective color micrographs of the same region. In all cases, the height of the bars reflects mean cell density of the sampled regions and the error bars show the standard error of mean.
Figure 1 shows the results obtained from layer IV of the primary visual cortex of rats exposed to all four experimental conditions. In all cases, the micrographs show combined FISH and FICC staining from the same section. Figure 1a displays a section of the visual cortex after 72 h of combined visual and auditory deprivation (condition NS). As can be seen here, there is very faint mRNA and protein staining in the majority of neurons. Figure 1b shows the same region of visual cortex but after 2 h of combined auditory and visual stimulation following NS (condition DS). This panel shows that a large number of neurons are activated in primary visual cortex as a result of the dual stimulation. Almost all stained cells appear to be double-labeled except for a few that are either protein or mRNA labeled only. Figure 1c shows the staining pattern in primary visual cortex in the experimental condition AV (2 h of auditory stimulation followed by 30 min of visual stimulation). With the exception of a few faint double-labeled neurons, a large majority of stained cells are mRNA-positive only. Figure 1d shows the staining pattern in response to condition VA (2 h of visual stimulation followed by 30 min of auditory stimulation). Here, a significant majority of the neurons are protein-positive only.
The staining patterns in these panels are represented in Figure 1e by way of mean cell densities for each of the staining categories (mRNA only — FISH, protein only — FICC, double-labeled). Cell counts were obtained from four animals in each experimental group. The counts for condition NS revealed that baseline staining was quite low. Both FICC- and FISH-labeled neurons were present in roughly equal numbers, a large majority of which was double-labeled. Similar numbers of FICC- and FISH-labeled neurons were found in response to condition DS, although the vast majority of neurons were double-stained. In condition AV, the number of FICC- and dual-labeled neurons was similar to that of condition NS. The number of FISH-labeled was comparatively much higher and similar to the number of double-stained neurons in condition DS. On the contrary, the number of FICC-positive neurons in condition VA was similar to the double-stained neurons in condition DS whereas the number of FISH- and dual-labeled neurons was comparable to condition NS. The implications of this will be discussed later.
Figure 2 summarizes the results obtained from FISH and FICC staining of layer IV of the auditory cortex across all four experimental conditions. Figure 2a displays protein and mRNA staining in a section of auditory cortex in condition NS. Here again, there is very little mRNA and protein staining in most neurons. Figure 2b captures the same region of cortex but after 2 h of combined auditory and visual stimulation (condition DS). Similar to the primary visual cortex, this panel reveals a significantly larger number of neurons activated as a result of this sensory manipulation in primary auditory cortex. Almost all stained cells appear to be double-labeled except for a few single-labeled neurons. Figure 2c shows the activation pattern in the same region in response to condition AV. Almost all the stained neurons are protein-positive, with a few double-labeled neurons that are sparsely visible. Figure 2d shows the activation pattern in response to condition VA. Contrary to the activation pattern observed in response to AV, a significant majority of neurons responsive to this condition are mRNA-positive only.
Figure 2e presents mean cell counts obtained from four animals in each experimental group. As with primary visual cortex, both FISH- and FICC-stained neurons were seen at equal frequency in both conditions NS and DS. However, condition NS showed the basal expression levels in auditory cortex in the absence of stimulation whereas condition DS produced a much higher staining level that resulted from combined visual and auditory stimulation. Unlike the visual cortex, however, condition AV produced more protein-positive neurons than both mRNA- and double-labeled neurons. Similarly, condition VA produced more mRNA-labeled neurons than protein- or double-labeled neurons.
Superior Colliculus (Superficial Laminae)
FISH and FICC staining of the superficial laminae (Zo, SuG and Op) of SC is shown in Figure 3. Figure 3a shows the staining pattern after 72 h of visual and auditory deprivation (condition NS). As before, there is very faint overall mRNA and protein expression. Figure 3b captures the same region of the superficial laminae but after 2 h of combined auditory and visual stimulation (condition DS). As can be seen here, a significantly larger number of neurons are activated as a result of this sensory manipulation. The majority of stained cells appeared to be double-labeled except for a few which were either protein- or mRNA-positive only. Figure 3c shows the activation pattern in the superficial laminae in experimental condition AV. The majority of stained neurons were mRNA-positive only, with very few double-labeled neurons. Figure 3d shows the staining pattern in response to condition VA. Contrary to the activation observed in response to AV, the majority of neurons here were protein-positive only. Finally, Figure 3e presents mean cell counts obtained from four animals in each experimental group. The pattern of staining and counting profiles in SC superficial laminae in response to all four experimental conditions was similar to that found for the primary visual cortex (Fig. 1).
Superior Colliculus (Deep Laminae)
Figure 4 shows the results obtained from FISH and FICC staining of the deep laminae (DpG and DpWh) of the SC in all four experimental conditions. Figure 4a displays a section of the deep laminae after 72 h of visual and auditory deprivation (condition NS). As in all other brain structures that we have examined, there was very faint overall mRNA and protein staining. Figure 4b captures the same region of the SC but after 2 h of combined auditory and visual stimulation (condition DS). A significantly larger number of neurons were activated as a result of dual sensory stimulation. Almost all stained cells appeared to be double-labeled, except for a few which were either protein- or mRNA-labeled only. Figure 4c shows the activation pattern in condition AV. Although the majority of the stained neurons were mRNA-positive, there were a few protein-positive and double-labeled neurons as well. Figure 4d shows the activation pattern in response to condition VA. Contrary to the activation pattern observed in response to AV, a significant majority of the responsive neurons were protein-positive. The size of the population of the double-labeled neurons observed in this condition was similar to that found in condition AV. However, a small number of mRNA-positive neurons were also found in this region in response to condition VA.
The cell counts obtained from all four animals in each experimental group are shown in Figure 4e. While this region of the brain behaved in a qualitatively similar fashion to superficial SC in response to conditions NS and DS, it revealed an interesting pattern of activity in response to conditions AV and VA, as noted above. For example, protein-positive and double-labeled neurons were present in greater numbers in condition AV whereas mRNA-positive and double-labeled neurons were present in relatively equal numbers in condition VA. The implications of this result will be taken up in the next section.
Figure 5 shows a dual activity map of SC deep laminae in greater detail in response to AV stimulation. Figure 5a shows FICC staining whereas Figure 5b shows FISH staining of the same area. Figure 5c shows a composite image created by digitally superimposing the Zif268 protein and mRNA staining images, giving rise to a dual activity profile of the same region. The filled arrowheads in all three panels point to two examples of a double-labeled neuron whereas the open arrowheads in Figure 5b and c point to single-labeled neurons (in this case, mRNA-positive). Although there were a few examples of Zif268 protein-positive neurons, protein expression in both conditions AV and VA was largely accompanied by mRNA expression as well and were therefore double-labeled.
Baseline and Activity-dependent Expression of zif268
In all of the brain areas examined in this study, visual and auditory silence (condition NS) revealed very sparse patterns of zif268 mRNA and protein expression. This result is not surprising, given the now well-established link between synaptic stimulation and expression of this immediate–early gene (Chaudhuri, 1997). However, the staining pattern is nevertheless informative in that it provides data on baseline expression of both gene products. This expression may be due to spontaneous transcription, in the case of the mRNA product, or subsequent translation, in the case of the protein product. That is, the presence of ongoing molecular events in these neurons may be responsible for driving spontaneous zif268 expression. Alternatively, the baseline expression may reflect stimulus-driven expression that is the consequence of spontaneous neuronal activity. Either way, the expression levels are quite low in the neocortical and SC areas that we examined and can be used to gauge the expression profiles that resulted after sensory stimulation was applied.
The application of both visual and auditory stimulation together (condition DS) for a period of 2 h showed dramatic increases in both zif268 mRNA and protein expression in all brain areas. The neocortical areas that are selective to each of these stimulus components presumably showed zif268 expression in response to that particular stimulus. We have previously shown that zif268 expression largely occurs in excitatory neuronal subtypes in visual cortex (Chaudhuri et al., 1995). As Figures 1e and 2e showed, the number of neurons expressing only the mRNA or protein products was very similar and quite low in both visual and auditory cortex. However, these same areas displayed large numbers of neurons that were double-labeled, i.e. expressing both mRNA and protein together in the same neuron. This is to be expected because the neurons in both cortical areas were subjected to continuous sensory stimulation during the 2 h. The same can be said of the expression patterns in the SC. However, given that neurons in the deep layers of the SC are known to be responsive to both visual and auditory stimulation (Meredith et al., 1987; Stein, 1988; Wallace et al., 1993), it is not possible to distinguish the specificity of those responses on the basis of simultaneous stimulation. To map sensory specificity required a stimulation regimen that could differentiate between the two types of input and therefore provide a visual display of neurons that were responsive to visual and/or auditory stimulation.
Stimulus-specific Expression Patterns in Visual and Auditory Cortex
The dual-mapping technique of zif268 expression (Chaudhuri et al., 1997) was used to assess response specificity. We reasoned that a 2 h period of visual stimulation followed by 30 min of auditory stimulation (condition VA) would be appropriate for driving protein synthesis to maximum levels in the visual cortex. However, since primary visual cortical neurons are not responsive to auditory stimulation, zif268 transcription would stop during the last 30 min of this sequence. The accumulated zif268 mRNA from the visual stimulation phase would then undergo rapid degradation and fall back to baseline levels, while Zif268 protein levels would remain high because of its longer time course of decay. With the converse stimulation sequence (condition AV), visual cortical neurons would be virtually unresponsive to the first 2-h portion of auditory stimulation and only respond during the last 30 min of visual stimulation. Therefore, neurons exposed to this condition would accumulate high levels of zif268 mRNA while Zif268 protein would remain at or near baseline levels.
This is precisely the effect that we observed. As Figure 1 showed, visual cortical neurons showed high levels of Zif268 protein and low levels of mRNA expression in condition VA and just the opposite in condition AV. Furthermore, the number of double-labeled neurons in both conditions VA and AV were quite low and slightly higher than the number of mRNA- and protein-positive neurons, respectively, in the two conditions (see Fig. 1e). Given that the number of double-labeled neurons was similar to the baseline levels (condition NS), it is likely that zif268 expression in these neurons was produced by spontaneous factors (transcription or neural activity). The very low levels of zif268 mRNA and protein in conditions VA and AV respectively also show that the time course that was used was sufficient to obtain differential staining patterns. The implications of this in terms of stimulus–transcription coupling uncertainty will be explored in the next section.
The notion of zif268 staining specificity would be strengthened if the same principle could be shown to apply to an area of the brain that is selective for a different sensory parameter. We therefore chose to analyze auditory cortex in the same manner where the predicted staining pattern would be just the opposite. That is, the response profiles to condition VA should show elevated mRNA and low protein expression whereas condition AV should show the opposite. As Figure 2 showed, this was indeed the result that we found. Thus, the specificity of this neocortical area for auditory stimuli is reflected by the selective expression of the protein product when sounds were applied for 2 h and the mRNA product when they were applied for only 30 min. Double-labeled neurons were found in somewhat larger numbers in this cortical area, though still far reduced from the primary expression product.
Stimulus–transcription Coupling Uncertainty
One of the major limitations in using immediate–early genes as a mapping tool concerns the fidelity of attributing gene expression to sensory stimulation. The link between a particular staining pattern, say that of zif268, and a particular stimulus (sensory or pharmacological) is often unclear because there are other possible factors that can contribute to non-specific gene expression. It may be that the immediate–early gene is expressed in response to a totally unrelated endogenous or exogenous factor and therefore cannot be solely attributed to the applied stimulus. This uncertainty in stimulus–transcription coupling has remained a major obstacle in molecular mapping and has prevented its widespread use in neural systems that may be especially susceptible to such concerns (Chaudhuri, 1997; Chaudhuri et al., 1997).
The use of a double-labeling technique resolves this problem in two ways. First, it can be used to assess the degree to which factors unrelated to the applied stimulus may contribute to the staining profile. This can be appreciated by examining the data obtained in this study for any of the brain areas. For example, the staining profiles in visual cortex (Fig. 1) showed that visual stimulation produced large numbers of Zif268 protein-positive neurons in condition VA and mRNA-positive neurons in condition AV. This fact illustrates the degree to which expression of the two products was linked only to visual stimulation since auditory stimulation or other factors were clearly irrelevant. Had it not been so, then the staining patterns in the two conditions would not have shown such differential effects and instead, both protein and mRNA expression would have been elevated in the two conditions. This argument is fortified by the data in Figure 2, which showed exactly the opposite profile for auditory cortex, i.e. large numbers of Zif268 protein-positive neurons in condition AV and mRNA-positive neurons in condition VA. Here, visual and other factors clearly had very little impact on gene expression because the product that would be linked to such stimulation was expressed at near baseline levels.
A second factor related to use of the double-labeling technique concerns the actual link between the applied stimulus and gene expression. That link is strengthened if one looks only at the mRNA expression profile in the map. The argument here is that if a selective mRNA response is found in the absence of an accompanying protein response, then such staining can be used to delineate neurons that specifically respond to that stimulus because it is highly unlikely that an unrelated endogenous factor was suddenly present during the last 30 min and not before. Therefore, mRNA-positive neurons can be taken to reflect activation that is stimulus-specific because otherwise those neurons would also contain Zif268 protein. This property of the double-labeling approach is especially important in interpreting staining patterns in brain areas that are susceptible to multimodal activation, such as the SC.
Neural Activity Profiles of the SC
We found the zif268 mRNA and protein staining patterns in the superficial laminae of the SC to be very similar to that of the visual cortex. This is not surprising given that the superficial laminae of the SC are visually driven and receive direct input from the retina (Lugo-Garcia and Kicliter, 1988; Illing, 1996). Neural activity in visual cortex is mirrored to a large degree in the SC superficial laminae, thereby producing similar zif268 mRNA and protein staining patterns in response to the experimental conditions of this study. The fact that Zif268 protein expression in condition AV and mRNA expression in condition VA is negligible shows the unimodal (visual) response characteristic of the superficial laminae neurons. The near-baseline response of these two products also shows that factors unrelated to visual stimulation, such as somatosensory, olfactory or endocrine factors were either negligible or did not have a significant impact upon zif268 expression.
The deep laminae of the SC displayed a somewhat different overall staining pattern from that seen in the superficial laminae or the primary sensory areas. The major difference was that the deep laminae neurons displayed responsiveness to both auditory and visual stimulation. Although both conditions VA and AV revealed a larger number of neurons that responded only to the visual component of the stimulus, a smaller but distinct number of neurons were found to be responsive to the auditory component. This is evident from the profile of Zif268 protein-positive neurons in condition AV and mRNA-positive neurons in condition VA. There are two facts that emerge from this staining data. First, the number of auditory neurons is far smaller than visual neurons in the SC deep layers. This is evident from the large number of mRNA-positive neurons that appeared in condition AV in comparison to condition VA. And second, most of the auditory neurons are also responsive to visual stimulation. This is apparent from the double-labeling profile that showed the majority of auditory neurons in condition VA to also contain Zif268 protein. Thus, although there are very few auditory neurons in SC deep layers, they appear to be largely multisensory in nature and therefore also responsive to visual stimulation.
The only caveat to this interpretation is the fact that a subset of SC neurons also responds to other sensory factors, such as tactile (Meredith et al., 1991, 1992; Stein et al., 1993; Wallace and Stein, 1996), nociceptive (Wang et al., 2000), and perhaps chemosensory stimulation. Although it appears from our staining profiles that clusters of neurons exist with differential and overlapping visual and auditory sensitivity, it may be that these other factors may also play a role in guiding gene expression. However, that role appears to be minimal, given the low baseline response in condition NS, where only visual and auditory silence was applied. Nevertheless, it cannot be ruled out entirely. To assess the separate role of uncontrolled factors in this context would require at least a third marker, one that has a different time course of expression in relation to zif268 mRNA and protein. Although there are several potential candidates, such as phosphorylated products of certain second and third messengers, none have yet been successfully applied in molecular mapping studies in conjunction with either c-fos or zif268. Nevertheless, continuing developments in molecular mapping techniques should provide further insights into the response characteristics of multisensory neurons, their integrative capacities across visual, auditory and somatosensory modalities, and the topographic layout of these neurons within a given neural structure.
We are grateful to Fariborz R-Dehghan for technical assistance, especially with the histological processing. We also thank Dr Leszek Kaczmarek for advice and critical comments. We thank Martine Turgeon for creating the auditory stimulus. This work was supported by grants from the Canadian Institute of Health Research (CIHR) and Natural Sciences and Engineering Research Council (NSERC) of Canada to A.C. and NSERC Post-Graduate Scholarship (PGSB) to S.Z.