Tonotopic Organization in Core Fields of the Auditory Cortex
The spatial organization of BF was assessed across the caudal-to-rostral extent of AI and AAF with ultra-high density mapping from 341 multiunit recording sites in 11 mice (N = 188 sites in 6 mice and 153 sites in 5 mice for right and left hemisphere, respectively). Recordings targeted the middle cortical layers (420-440 μm from the pial surface). We observed a clear tonotopic organization for BF with low frequency BFs around 6 kHz forming the caudal boundary of AI, a high frequency mirror-reversal around 32 kHz forming the boundary between AI and AAF, and a progressive shift to lower frequency BFs at the rostral border of AAF (). Note that mice can hear and vocalize up to 100 kHz, yet the highest frequency cortical BF recorded in our entire sample was 34.3 kHz. Our recordings were carried out at 4-7 weeks of age, several months before the onset of high frequency hearing loss in C57BL6 mice (Willott, 1986
). In fact, the paucity of BFs greater than 40 kHz has been routinely reported in the inferior colliculus and auditory cortex of several inbred and outbred mouse strains and it is thought that these acoustic frequencies may be represented in specialized fields (Stiebler et al., 1997
) or, intriguingly, according to their subharmonic cochlear distortion products (Portfors et al., 2009
Representative frequency-response areas (FRAs) from AI and AAF recording sites show well-demarcated V-shaped tuning profiles (). BF changes were plotted across the caudal-to-rostral expanse of AI and AAF with an average sampling density of 40 recordings per mm, or 1 recording site every 25 μm. We observed an orderly parabolic shift in BF that was well fit by a quadratic function, with the peak used to define the high frequency boundary separating AI and AAF (). We then combined normalized maps from individual mice to represent 0 as the AI-AAF reversal point, −1 as the caudal low frequency boundary of AI and +1 as the rostral low frequency boundary of AAF (). Using this approach, we found a significant increase in BF across the caudal-to-rostral extent of AI (one-way ANOVA, F=34.9, p < 1.10−6) and a significant decrease in BF along the caudal-to-rostral extent of AAF (F=6.88, p < 0.005).
Tonotopic Organization of MGBv Projections to AI
As a next step, we sought to determine whether tonotopy in AI arose from topographically organized MGB projections. To address this possibility, fluorescent retrograde tracers (CTB-red and CTB-green) were injected into a low and a high frequency region of AI that had been identified through microelectrode mapping as described above (). Following a waiting period of approximately 80 hours for retrograde transport, the brain was sectioned along the same plane used in auditory thalamocortical slice experiments (Cruikshank et al, 2002
), which permitted visualization of MGBv, the thalamocortical axon bundle, and AI (). In the example shown in , CTB-green was injected into a 7.0 kHz BF region and CTB-red was injected into a 22.6 kHz BF region. Qualitatively, one can appreciate that well-segregated AI injection sites were innervated by compartmentalized axon bundles that originated from spatially separable populations of MGB neurons.
Tonotopic organization of MGBv projections to AI
The projection patterns were examined in greater detail from this case () as well as a second case with injections into the low (5.7 kHz) and high (32.0 kHz) BF extremes of the AI map (). In both cases, we observed dense, spatially separable projections from MGBv to each BF region in AI accompanied by a sparser, co-mingled projection from MGBm (). Cell body locations were plotted against the anatomical boundaries of MGBv and MGBm in multiple sections across the dorsal-to-ventral extent of MGBv. A consistent pattern emerged from all sections in both cases: in MGBv, low frequency projection neurons were located caudally and laterally, and high frequency projection neurons were rostral and medial; MGBm neurons projecting to AI were not spatially separable ().
Topographic projections from MGBv, but not MGBm, to AI
Further quantification revealed that 75% of retrogradely labeled cell bodies were in MGBv, 24% from MGBm, and the remaining 1% from the posterior thalamic complex. Low frequency projection neurons in MGBv were significantly more lateral (unpaired t-test, p < 1×10−6) and caudal (p < 1 × 10−6) than high frequency projection neurons (). No significant difference in medial-lateral position was observed among low and high frequency projection neurons in MGBm (p = 0.12), although high frequency projection neurons were observed to be significantly more rostral (p < 0.005; ). Taken together, these data demonstrate that cortical projections from MGBv are topographically ordered and further suggest that MGBv may be tonotopically organized along a low-to-high, caudolateral-to-rostromedial axis ().
Neurophysiological Dissociation of MGBv and MGBm
Findings from the tracer studies suggest three explicit predictions regarding tonotopic map orientation in the MGBv: 1) Recordings from rostral-to-caudal positions along the lateral wall of the MGBv, should yield a progressive downward shift in BF; 2) Recordings across the lateral-to-medial extent of the MGBv should yield a progressive upward shift in BF; 3) There should be no clear tonotopic organization within the MGBm. To test these predictions, we performed high-density multiunit mapping across the lateral-to-medial extent of the MGBv and MGBm using a multichannel silicon probe (50 μm separation between contacts). Penetrations were made throughout the rostral-to-caudal extent of the MGB (50-100 μm between penetrations) at the same angle as the plane of section in the thalamocortical slice (). Example FRAs shown in illustrate the drop in BF along the lateral wall of the MGBv between rostral, intermediate and caudal recording locations (left column). Furthermore, regardless of the BF at the most lateral recording position, BF increased across the first few medial recording locations. However, it was equally clear that BFs subsequently decreased as the recording location moved to deeper, more medial areas of the MGB.
Neurophysiological dissociation of MGBv from MGBm
Did this high frequency mirror reversal in BF indicate that recordings had crossed into the MGBm, or did it reveal a BF organization that was not suggested from the tracer studies? To address this question we identified positions along the rostral-to-caudal extent of the MGB that yielded clear mirror reversals with the silicon probe (, gray arrow) and then inserted a conventional tungsten microelectrode along the same trajectory (, white arrow) in order to make small electrolytic lesions lateral and/or medial to the BF reversal. An example of one such experiment is shown in , where a lateral lesion was made at a well-tuned high frequency site 0.13 mm lateral to the high frequency reversal (, top panel), and a medial lesion was made at the deepest site with tone-evoked activity, 0.21 mm medial to the high frequency reversal point (, bottom panel, , white crosses). Localization of lesions in coronal sections reacted for cytrochrome oxidase revealed that the recording site lateral to the BF reversal was in MGBv and the lesion medial to the BF reversal straddled the border between MGBm and the posterior thalamic complex (). Indeed, in 13/14 lesions, the BF reversal correctly predicted the anatomical boundary dividing MGBv from MGBm ().
Tonotopy in MGBv and MGBm
Using the high frequency reversal point as a neurophysiological marker for the boundary between the ventral and medial divisions, we were able to rigorously test each of the three predictions enumerated above. The BF reversal point was identified in each penetration (black square, ) and the slope of the linear fit lines were calculated independently from the rising and falling components of the BF functions (). Returning to the first prediction, BFs measured along the lateral wall of the MGBv were found to decrease significantly along the rostral-to-caudal extent (N=43, 1-way ANOVA F=4.67, R squared = 0.96, p < 0.05; ). For the second prediction, we found that BF increased significantly from lateral to medial positions within MGBv (F=26.1, p < 1 × 10−6; gray circles), and the slope of the increasing BF function did not differ according to where recordings were made within the rostral-to-caudal extent of MGBv (F = 1.0, p = 0.55; gray circles).
With regard to the third prediction, despite the extensive co-mingling of low and high frequency AI projection neurons within the MGBm, neurophysiological recordings indicated that BF decreased significantly across the lateral-to-medial extent of the MGBm (F = 2.75, p < 0.005; black squares) and this negative slope did not vary systematically across the rostral-to-caudal extent of the MGBm (F=0.41, p = 0.97; black squares). Therefore, tonotopy in MGBv closely matched our predictions from the retrograde tracer studies, yet we also found evidence for a mirror-reversal decreasing BF gradient within MGBm, which was not predicted by the anatomical data.
Variations in spectral bandwidth and onset latency in auditory cortex and MGB
Analysis of response properties at individual recording sites proved to be another useful tool for distinguishing AI vs. AAF and MGBv vs. MGBm. FRA bandwidth measured 10 dB above threshold was significantly broader in AAF than AI (0.82 ± 0.03 vs. 0.92 ± 0.04 octaves, unpaired t-test, p < 0.05) and in MGBm compared to MGBv (1.12 ± 0.04 vs. 0.89 ± 0.03 octaves, unpaired t-test, p < 1 × 10−6
; ). Consistent with previous observations in mouse (Linden et al., 2003
) and rat (Polley et al., 2007
), we found that tone-evoked onset latency was significantly shorter in AAF than AI (12.39 ± 0.17 vs. 14.79 ± 0.3 ms, unpaired t-test, p < 0.0001). Interestingly, onset latencies decreased by 5.3 ms across the BF range in AI (1-way ANOVA, F=5.98, p < 5 × 10−5
; ), but not in AAF (F=1.32, p = 0.26). Similarly, onset latency was significantly shorter on average in MGBm than MGBv (9.75 ± 0.17 vs. 10.75 ± 0.18 ms, unpaired t-test, p < 0.0001), and decreased by 5.3 ms across the BF range in MGBv (F=24.98, p < 1 × 10−6
) without varying significantly in MGBm (F=0.59, p = 0.7; ). Note that the latency shift observed between high and low BF recording sites in AI and MGBv cannot be accounted for by the basilar membrane group delay. Tone-evoked onset spikes recorded in single nerve fibers from the basal (high frequency) versus apical (low frequency) regions of the mouse cochlea occur at 1.86 ± 0.07 and 2.89 ± 0.11 ms, respectively, which can only account for approximately 20% of the shift observed in AI and MGBv (M.C. Liberman, personal communication, , black squares).
Topographic Mapping in the Thalamocortical Slice
Using the BF gradients characterized in AI and MGBv of the intact mouse, we turned to the acute thalamocortical slice preparation to determine whether tonotopic organization corresponded to topographic connectivity. Specifically, we predicted that electrical stimulation of the MGBv along the caudolateral-to-rostromedial tonotopic gradient that emerged from tracer and mapping studies would yield a smooth shift in AI activity foci along the caudal-to-rostral low-to-high BF gradient established from our cortical mapping studies. This prediction could be effectively tested through VSD imaging in the thalamocortical slice, by applying a discrete stimulus to the MGBv and measuring subthreshold VSD signal changes across the entire caudal-to-rostral extent of AI simultaneously. VSD signal amplitude was measured across 18 ROIs (125 ×125 μm each) positioned either in layer IV, where VSD response amplitudes were greatest, or in an immediately dorsal region in layer II/III.
In support of our prediction, we observed that rostral areas of AI were most responsive to rostromedial MGBv stimulation (high BFs), whereas more caudal areas of the AI map were more effectively activated by caudolateral MGBv stimulation (low/mid BFs) (). Stimulation of MGBv at six loci along the optimized caudolateral-to-rostromedial orientation induced broad, overlapping subthreshold activation profiles that nevertheless shifted significantly across the caudal-to-rostral extent of layer IV in AI (mean ± sem ROI for site 1 vs. 6 = 7.6 ± 0.8 vs. 12.8 ± 0.8; t-test, p < 0.0001; ). Average AI layer IV responses were normalized for each stimulation site and represented as a color map, with the upward diagonal band representing the point-to-point topographic mapping between MGBv and AI (). Topographic shifts along the caudal-to-rostral axis were specific to the optimized stimulation orientation, as stimulating six loci along an orthogonal orientation did not significantly shift the activity profiles, consistent with the expected effects of stimulating the MGBv along an iso-frequency contour (9.2 ± 1.9 vs. 10.0 ± 1.1; p = 0.72; ). As a final test, we returned to the optimized orientation but shifted the ROIs to layer II/III, rather than layer IV. Topographic shifts in layer II/III were weak, but not significant (7.6 ± 0.9 vs. 10.4 ± 1.4; p = 0.1; ), in agreement with recent reports of heterogeneous tuning in the upper layers of mouse AI (Bandyopadhyay et al., 2010
; Broicher et al., 2010
; Rothschild et al., 2010
Topographic mapping in the thalamocortical slice
Linking Maps of Preferred Frequency to Spatially Distributed Frequency Representations
In one view, VSD imaging data from layer IV are in close agreement with AI and MGBv mapping studies: moving the stimulating electrode across the caudolateral-to-rostromedial tonotopic map in MGBv yields an orderly progression of activity foci across the caudal-to-rostral tonotopic map in AI. This finding thus projects a basic sensory map onto the acute auditory thalamocortical slice preparation to enable high-resolution studies of synaptic physiology. However, from another perspective, the BF maps in vivo exhibited relatively precise point-to-point mapping between a tone frequency and its preferred representation in the cortical map; whereas, stimulating a single point in MGBv evoked VSD activity across the majority of the AI map. In other words, positional shifts in peak VSD signals corroborate the in vivo mapping of AI, but the breadth and overlap of the VSD activity profiles do not. Why might this be?
The fact that VSD is sensitive to sub-threshold changes in membrane potential, while microelectrode mapping reflects spiking activity can certainly explain some of the differences between the two data sets (Berger et al., 2007
). On the other hand, the discrepancy may also be attributed to the nature of BF maps, which only represent the single preferred frequency for a given recording site, as compared to VSD activation maps, which measure the entire spatial spread of population activity evoked by a point stimulus.
To address this difference in conceptual approach and to reconcile the basic differences in the degree of tonotopic mapping precision with these two methods, we re-analyzed our in vivo recordings according to the spatial spread of tone-evoked excitation rather than preferred frequency. Returning to the same mapping experiment shown in , we surveyed which recording sites contained 8.6 and 19.7 kHz tones presented at 50 dB SPL within their FRAs (). Compared to , one can appreciate that tone representations in AI appeared less spatially focused than the orderly BF map, and tone representations in AAF were even more broad and overlapping.
To facilitate direct comparison with the VSD imaging data, we grouped recording sites from each AI and AAF map into 18 ROIs that spanned the tonotopic gradient in each field. We then selected 6 tone frequencies homologous to the 6 MGBv stimulation sites and determined the probability that neurons in each ROI would contain a given 50 dB SPL tone frequency within their FRA. Spiking probability profiles in AI were broader than the BF preference maps, but not as broad as the subthreshold VSD maps (). Tones of different frequency activated tonotopically appropriate regions within AI, yet the response profiles were quite broad, with frequencies < 20 kHz activating approximately half of the map with a 0.5 probability or greater.
Normalized color maps were created in a similar fashion to those presented in and further demonstrated that spatial representations of tone frequencies in AI were broad, but spatially organized with a narrow upward diagonal band that was fairly stable across tone levels (). Compared to AI, spatial representations of tones in AAF were degraded and level-intolerant (), which was to be expected given the greater BF scatter (), compressed range of tone frequency representation (), and broader frequency tuning (). We also investigated spatial activation profiles in the MGBv across 9 ROIs positioned along the caudolateral-to-rostromedial BF gradient. Qualitatively, the spatial code for tone frequency representation in MGBv was less organized and less level-tolerant than AI, but superior to AAF ().
Spatial activation profiles were further quantified in AI on an individual mouse basis by measuring the peak and width at half maximum of each frequency’s normalized response functions summed across 30-60 dB SPL. We observed significant spatial shifts in the caudal and rostral boundaries at half max (Friedman non-parametric ANOVA, p < 0.005 for both) as well as the peak position (p < 0.01) with tones of increasing frequency (). Overall, each tone activated one-third to one-half of AI ROIs. As a final step, we directly compared spiking half-height boundaries with the normalized VSD half-height boundaries and the average BF values across all ROIs ().
The upward diagonal slant present in all three measures demonstrates that layer IV of AI is tonotopically organized regardless of whether the input signal to the mapping function is subthreshold VSD, suprathreshold spiking or BF. However, the precision of the layer IV mapping function varies between techniques. Subthreshold activity profiles suggested a coarse mapping specificity wherein any point in the MGBv map can activate nearly any point in the AI map. Suprathreshold spiking patterns produce a much higher degree of mapping precision, wherein a given tone can recruit activity across a swath of the AI map, yet only a small cluster of neurons within this active zone will claim this tone as the BF.