PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
 
Nat Neurosci. Author manuscript; available in PMC 2010 September 1.
Published in final edited form as:
Published online 2010 January 31. doi:  10.1038/nn.2490
PMCID: PMC2866453
NIHMSID: NIHMS167825

Dichotomy of functional organization in the mouse auditory cortex

Abstract

The sensory areas of the cerebral cortex possess multiple topographic representations of sensory dimensions. Gradient of frequency selectivity (tonotopy) is the dominant organizational feature in the primary auditory cortex, while other feature-based organizations are less well established. We probed the topographic organization of the mouse auditory cortex at the single cell level using in vivo two-photon Ca2+ imaging. Tonotopy was present on a large scale but was fractured on a fine scale. Intensity tuning, important in level-invariant representation, was observed in individual cells but was not topographically organized. The presence or near-absence of putative sub-threshold responses revealed a dichotomy in topographic organization. Inclusion of sub-threshold responses revealed a topographic clustering of neurons with similar response properties, while such clustering was absent in supra-threshold responses. This dichotomy indicates that groups of nearby neurons with locally shared inputs can perform independent parallel computations in ACX.

Neurons in sensory cortical areas are organized into vertical columns1, 2 such that neurons with similar stimulus selectivity are clustered together. Columns, in turn, are arranged in maps so that columns with similar response properties are nearby 13.

Encoding of stimuli in the primary auditory cortex (A1) is thought to be sparse4, 5, with a multitude of overlaid maps representing different stimulus dimensions3, 6. The dominant topographic feature in the auditory cortex (ACX) is a tonotopic map, which is best seen at moderate to near-threshold sound levels79. Different ACX regions can be distinguished based on the direction of the frequency gradient or its expression3, 6, 9. Other forms of patchy organization overlaying the tonotopic map have been described. These include organizations based on the bandwidth of frequency integration, tuning to intensity, selectivity or suppression of binaural inputs, selectivity for the direction and speed of frequency modulation, and the periodicity of the stimulus 3, 6. Such maps were revealed using techniques with limited spatial resolution, such as single unit recording (100–200 µm) 3, 610, intrinsic imaging (> 200 µm) 11, 12, and voltage sensitive dyes (> 200 µm)13, 14. Thus the fine scale architecture of ACX remains unknown.

Here, we overcome these limitations by using in vivo two-photon imaging of functional responses in mouse ACX using Ca2+ sensitive dyes1517. This allows imaging of the functional microarchitecture of cortical maps with single cell resolution. To probe for the existence of “maps” of response properties we take advantage of the ability to image the activity of many neurons simultaneously, and develop general methods for identifying maps.

Results

We bulk-loaded mouse ACX (Fig. 1a–d) with the Ca2+ indicators Fluo–4 AM (P18–P35, n = 24) or OGB–1 AM (P13–P32, n = 15) (Fig. 1d) using visually guided pressure injection. Increases in Ca2+ were observed in response to both sinusoidally amplitude modulated (SAM) tones and broadband noise sounds (Fig. 1e). Both brief and sustained fluorescence increases (Fig. 1e) were observed. In any given neuron, tone evoked responses were often maximal at a specific frequency (defined as the neurons’ preferred, or characteristic frequency, CF, as in Fig. 3a). Some neurons responded monotonically with increasing stimulus intensity. Others were nonmonotonic with intensity, achieving a maximal response at an intermediate (“best”) intensity and then decreasing beyond that as shown in Fig. 6 and Fig. 7, (see Methods for precise definition of these response measures). However, even at their CF or best intensity, neurons did not always respond to every presentation of the stimulus (Fig. 2b, S2). Thus, mean Ca2+ increases were lower (2–7%) than single trial responses (~ 10% dF/F, Fig. 1e, Fig. 2b, Fig. S2). While OGB–1 and Fluo–4 responses were similar the signal-to-noise ratio was lower with OGB–1 1820. The disparity in mean fluorescence and single trial fluorescence was quantified as reliability or fraction of responsive trials (see Methods for definition). The mean reliability to a set of stimuli varied from 0.02–0.5 with Fluo–4 (Fig. 2c) and 0.02–1 with OGB–1 and was higher with OGB–1 (median 0.28 n = 15) than Fluo–4 (median 0.18, n = 24, P < 10−10, ranksum) (Fig. 2c). Since these were mean values over a set of stimuli, some stimuli (like at CF or best intensity) produced higher reliability. For example, broadband noise stimuli typically (if monotonic) resulted in increased reliability with increasing stimulus intensity (Fig. 2b). However, as reported previously, a fraction of neurons in the imaged field did not respond to any presented auditory stimulus (Fig. 2d) 5, 21. While median responsiveness was similar for the 2 dyes (OGB–1: 75%, Fluo–4: 66%, P = 0.11, ranksum, Fig. 2d) the distributions were different (P = 0.011 One sided 2–sample KS test). The median maximum response strengths with OGB–1 and Fluo–4 was 3% (Fig. 2e) which corresponds to only a few spikes (< 10 spikes/s, based on our in vitro recordings, see Fig. 8e,f and Fig. S4), consistent with relatively low firing rates5 in the auditory cortex to tones or broadband noise.

Figure 1
Functional 2-photon Ca2+ imaging in mouse ACX
Figure 2
ACX Ca2+ responses are unreliable
Figure 8
ACX cells receive shared inputs but respond differentially

Tonotopy present in coarse but not in fine spatial scales

Previous microelectrode studies showed multiple different functional areas within the mouse ACX22. These areas are defined by distinct frequency tuning properties and gradients of the tonotopic map22. All such maps (by definition) in any species were primarily delineated based on responses to tones and other simple stimuli3, 6. While ACX neurons have been additionally characterized by their responses to vocalizations, natural sounds and other complex stimuli2329, systematic mapping studies have not been done with such stimuli. Thus to describe the basic topographic organization of the mouse auditory cortical areas, and to lay the ground for future more elaborate characterizations we focused on the simpler stimuli used previously3, 6. By imaging different areas along the cortical surface we identified multiple areas of ACX based on their tuned responses to tones and direction of gradient of the CF at moderate sound levels (40–60 dB SPL) (Fig. 3 and Fig. 4) (n = 9, Fluo–4).

Figure 3
Large scale organization of ACX probed with single cell resolution
Figure 4
Tonotopy exists in A1 and AAF on large but not on small spatial scales

A1 neurons had strong responses to tones, usually with single-peaked tuning curves9, 22 (Fig. 3a, b, cells 4–8). An area containing such neurons is illustrated in Fig. 3c. Imaged areas in A1 clearly showed a coarse progression of CF from lower (~ 16 kHz, blue) to higher (~ 32 kHz, brown) frequencies (Fig. 3c and d). In addition to a part of A1, parts of two other ACX regions could be identified in this animal: the ultra frequency (UF), and the dorsal posterior (DP) region. Both UF and DP lacked clear progression of CFs and were characterized by weak responses to tones and very broad or multi-peaked tuning curves 9, 22 (Fig. 3b, cells 1–3). While neurons in the UF region were generally tuned to very high frequencies (> 45 kHz), they were sometimes also responsive to lower frequencies as shown in Fig. 3b (cells 1–3). The general location of the imaging site with respect to different functional regions of the mouse ACX22 is depicted in Fig. 3e (white dashed rectangular region) based on the responses in different fields. The imaged location in A1 was confirmed to be thalamo-recipient using retrograde labeling of thalamic projections (Fig. 3f). Figure 4 illustrates the differences in CF progression in A1 (Fig. 4a) and the anterior auditory field (AAF; Fig. 4b) regions of ACX imaged in two other animals. AAF exhibited a progression of CFs although with high variability from lower (~ 10 kHz, blue) to higher (~ 23 kHz, green) frequencies in a direction (rostral to caudal) opposite to that seen in A1 (caudal to rostral, Fig. 4a inset) (Fig. 4c). These results, therefore, confirm the existence of several fields of mouse ACX with coarsely differentiated single cell responses22, and in particular, coarse tonotopic gradients in both A1 and AAF (Fig. 4c).

Previous studies of A1 analyzed tonotopy on a large spatial scale level (> 100 µm) 3, 6. We also observed this general progression of frequency selectivity in A1 and AAF (~ 1 octave/350 µm, median 2.7 octaves/mm in A1 and −2.3 octaves mm−1 in AAF, n = 9 animals, Fig. 4c) in the rostro-caudal dimension. However, when examined closely we found a great degree of overlap of surrounding frequencies (often more than an octave apart) that locally degraded the tonotopic map (Fig. 4a, b). Furthermore, the population distribution of CF variability within a field of view was large (Fig. 4d). We used d’ analysis (Fig. 4e) to calculate the distance needed to reliably detect different tonotopic regions. With such a large variability, we calculated that the CF difference needed to discriminate A1 regions is 1/0.85 = 1.2 octaves (0.85 from Fig. 4d). Given the median slope (2.7 octaves mm−1) two locations within ~ 400 µm cannot be distinguished.

Lack of organized bandwidth maps

While a tonotopic axis is present in A1 of all species examined thus far, it is unknown what stimulus features are mapped in the orthogonal direction, the isofrequency axis. Single unit recordings in various species indicated a patchy organization along this axis based on such response properties as intensity tuning and bandwidth 3, 6. In contrast, imaging showed that sharply and broadly tuned neurons could be right next to each other (Fig. 5a and b) showing a lack of fine organization based on tuning curve bandwidths.

Figure 5
High local variability in bandwidth

Quantification of the heterogeneity by calculating the bandwidth variability showed it to be large, especially with Fluo-4 compared to OGB–1 (Fig. 5c). Thus, based on the fine scale mapping with two-photon imaging in the mouse A1, the representation of frequency tuning and across-frequency integration is far more heterogeneous than previously seen with single unit recordings and intrinsic imaging in several species, including the mouse.

Lack of organized intensity maps

Given the absence of fine scale organization based on iso-intensity frequency tuning, we next probed the organization for intensity tuning. ACX neurons are responsive to broadband stimuli, such as broadband noise. Mean responses to broadband noise stimuli of increasing intensity were monotonic in some neurons and nonmonotonic in others (Fig. 6a and Fig. 7a).

Figure 6
Intensity tuning and local heterogeneity in noise responses
Figure 7
Figure 7. Lack of organized intensity maps

Nonmonotonic coding of sound intensity can be utilized in level-invariant representation of sound objects 30 and it has been proposed from single unit recordings that organized patchy maps of intensity tuning exist 3, 6, 31. We used broadband stimuli to probe for the existence of clusters or maps based on preferred intensity. Broadband noise is useful for this purpose as in a given imaged area there were neurons of varying CFs (see Fig. 3 and and4).4). The responsiveness of each neuron within an imaging field as a function of intensity is reflected in activation plots such as those of Figs. 6b and S3. Patches of cells were activated best together at different intensities, but the numbers of neurons activated varied nonmonotonically with intensity (Fig. 6c and S3). Furthermore, individual neurons that respond to different intensities were spatially intermingled (Fig. 6d) as can be seen in the heterogeneous mix of response properties of the neuronal population when accumulating data from different intensity combinations (e.g., 3 levels in Fig. 6d and Fig. S3C, colored squares). This local heterogeneity was confirmed by imaging multiple areas of ACX and reconstructing the intensity preferences of individual neurons (Fig. 7). It was then quantified by calculating the variability of the best intensity in the imaged area on a cell-by-cell basis. This analysis showed that there was a large variability of best intensities (Fig. 7b, less with OGB–1 than with Fluo–4) across ACX. For an organized intensity map to exist there should be a gradient of best intensity across the medio-lateral (iso-frequency) axis A1. Given the range of 70 dB (based on our data) of preferred noise intensity variation and a medio-lateral extent of 0.7 mm of A1 22 a predicted slope of best intensity would be 100 dB mm−1. We investigated the existence of such a slope by analyzing the variability of best intensity distribution of our imaging sites. Based on the median variability of best intensity (Fig. 7b) d' analysis shows that to reliably detect differences in best intensity one would have to be ~ 300 µm apart resulting in a detected intensity difference of 33 dB. Given the medio-lateral extent of our imaging sites (< 300 µm, Fig. 7a), our data do not provide any evidence for such a gradient and thus for an organized intensity map. An alternative organization to a best intensity gradient is the existence of microdomains of neurons with similar best intensity. To identify such microdomains of potentially patchy intensity selectivity we quantified the average number of neurons in a local neighborhood of each neuron (5 nearest neighbors) with the same intensity preference as the central neuron. Cumulative distribution of average percentage of neighboring neurons with the same preferred intensity (Fig. 7c) shows a high degree of heterogeneity (more with Fluo–4 than with OGB–1) suggesting that while some clusters of neurons with similar intensity preferences do exist, they represent only a fraction of all neurons in ACX.

Dichotomy in spatial organization of neurons with different response properties

The response selectivity of neurons is typically calculated based on the maximum of the response (dF/F) mean during the stimulation period (for example 17, 32, 33, Fig. 3). This response measure is related to total number of spikes in the response34 and thus our analysis is equivalent to those based on average firing rate. However, this analysis ignores differences in spike timing, which can also convey different information about the stimulus35, 36 and might be spatially organized. The entire Ca2+ trace (although at a coarse temporal resolution) has information about both spike time, spike rate and possibly some subthreshold membrane voltage fluctuations that depend on the affinity of dye used.

We developed a general method (see Methods) to identify response similarity maps based on the entire significant portion of the Ca2+ (dF/F) waveform during stimulation and also in the period following stimulation (off-period, Fig. 1e, Fig. S1C). Neurons with similar response waveforms were grouped together based on k-means clustering of the significant principal components of the response waveforms to a set of intensities of broadband noise or to a set of tone frequencies. The number of clusters was varied between 2 and 15 (average 2 cells per cluster see Methods) until a preset criterion of cluster tightness was met. When the preset criterion was not met within 15 clusters we set the number of clusters to 15, which signified no significant clustering.

The center of the cluster (the centroid) represents the mean response of cells within a cluster to a particular stimulus set and thus contains information about both the stimulus selectivity and temporal response pattern. Figure 8a shows four cluster centroids obtained with responses to SAM noise at different intensities. The four different response patterns (4 colors) show distinct response classes (monotonic and weak, blue and 1; non-monotonic with different preferred intensities or time delay to response peak, other 3 clusters). Using OGB–1 as the Ca2+ indicator we often (~ 50% n = 154/290 sets) found significant spatially clustered groups of cells with clear sharp transitions between response classes within a field of view (Fig. 8b; lower right image shows the location of cells belonging to clusters described by centroids in Fig. 8a). This indicates that cells within this field of view showed similar response waveforms to the same set of stimuli. By contrast, such spatial segregation was considerably less common (Z–value 7.931, Z–test for proportions) using Fluo–4 as the Ca2+ indicator (~ 20% n = 70/321, Fig. 8b). The number of clusters formed was generally larger with Fluo–4 than with OGB–1 (Fig. 8c; P < 10−6, KS-test) while there were similar number of cells per imaging site with OGB–1 (38) and Fluo–4 (34) (P = 0.45, ranksum). Further, there were fewer cells per cluster with Fluo–4 than with OGB–1 (p < 10−5 one sided KS–test; Fig. 8c). To characterize spatial local heterogeneity of responses we quantified the number of cells in the local neighborhood (5 nearest neighbors) of a single cell that belonged to the same cluster as the central cell (Fig. 8d). The fraction was higher with OGB–1 than with Fluo–4 (median of 40% with OGB–1 versus 20% for Fluo–4. P < 10−6, ranksum). Thus cells were more similar to the neighbors with OGB–1 than Fluo–4 (Fig. 8d). Thus with Fluo–4 there is more local heterogeneity than with OGB–1. These results show a clear dichotomy in spatial arrangement of neuronal response properties seen with the 2 dyes.

Since the Ca2+ affinities of OGB–1 (Kd = 170 nM) and Fluo–4 (Kd = 350 nM) are different these dyes might detect a different fraction of a putative subthreshold response. We tested the ability of the two dyes to detect subthreshold depolarizations directly by recordings in vitro in brain slices (Fig. 8e–g). Current (Im) was injected to evoke a depolarization just below spike threshold (Fig. 8e left, < Θ) or above spike threshold (Fig. 8e, right, < Θ). Below spike threshold only OGB–1 showed dF/F changes (Fig. 8e, g, P < 0.001). However, when cells spiked both Fluo–4 and OGB–1 showed dF/F changes and those changes were larger with Fluo–4 (Fig. 8e). Thus as expected 37, responses with Fluo–4 are biased towards suprathreshold responses with no detectable contribution from subthreshold membrane voltage changes (Fig. 8e, f). In contrast as expected from the higher affinity values 19, 20, Ca2+ responses with the higher affinity dye OGB–1 can detect subthreshold membrane voltage change s (Figs. 8e, f). Since subthreshold depolarizations are generated by excitatory input, we verified directly that synaptic stimulation could lead to detectable Ca2+ responses (Fig. 8g and S4). Electrical stimulation of synaptic inputs causes EPSPs and significant mean Ca2+ fluorescence signals with OGB–1 (Fig. 8g) while not with Fluo–4 (see Fig. S5). Thus, together these data suggest that OGB–1 can detect subthreshold and suprathreshold activity, while Fluo–4 only detects suprathreshold responses 37.

We observed a higher mean reliability in vivo in responses seen with OGB–1 (median 0.28) compared to Fluo–4 (median 0.18, P < 10−10 ranksum) (Fig. 2c). Since spiking responses to a given depolarization can have failures, the enhanced reliability of OGB–1 suggests detection of a larger fraction of sub–threshold activity in vivo with OGB–1. Further, the variability in CF shows a trend for lower variability with OGB–1 than Fluo–4 (although not significant, Fig. 4d) and variability in preferred intensity and bandwidth (Fig. 7b and Fig. 5c respectively) in the population was significantly lower with OGB–1 than with Fluo–4. These differences can also be accounted for by the fact that OGB–1 detects subthreshold responses while Fluo–4 does not. Thus, using these two different dyes allowed us to differentially probe for organization based on predominantly supra-threshold (Fluo–4) or combined supra-threshold and sub-threshold responses (OGB–1).

Given that OGB–1 reports a larger fraction of sub-threshold activity than Fluo–4 (Fig. 8e–g), this suggests that nearby ACX neurons have high degree of shared inputs. The lack of spatial segregation and highly spatially inhomogeneous responses with Fluo–4 suggests heterogeneity of supra-threshold responses. Since Fluo–4 responses reflect primarily suprathreshold activity this suggests that neurons perform independent computations on these shared inputs (Fig. S6).

Discussion

A key finding in our study is the presence of diverse frequency selectivity and best intensity in neurons in close proximity (as close as 20 µm). This suggests the lack of organization of precise maps based on frequency selectivity, frequency integration, or intensity selectivity. In contrast, we reveal a fine scale organization in ACX when using dyes (OGB–1) that reflect subthreshold Ca2+ changes in addition to those dependent on spiking. These maps are likely derived from shared inputs to a population of neurons. Studies in the visual cortex using OGB–117, 32, 33 use stimuli such as oriented bars that evoke responses with large number of spikes unlike the low response rates in the ACX obtained with tone and noise stimuli5. Thus it is unlikely that in the studies of visual cortex small (1%, Fig. S4) sub–threshold fluctuations were seen with large Ca2+ responses (up to 30%17) caused by high spike rates.

The presence of maps based on the temporal responses of neurons has not been reported before and suggests that inputs to ACX cells are spatially organized. The absence of such clustered organization with dyes that primarily reflect spiking activity (Fluo–4) suggests very heterogeneous processing of inputs by nearby neurons from largely shared inputs. Since cortical columns can have highly specific fine-scale embedded networks3840, they could generate heterogeneous and spatially intermingled responses. In addition, the presence of a population of neurons with diverse response properties but with shared input selectivity might imply the existence of multiple parallel ascending pathways in ACX. In addition these results also indicate that there is little neuronal redundancy in each “column” as each neuron seems to have a unique set of stimulus selectivity and response properties.

Previous studies of ACX used techniques that have spatial resolution of 100–200 µm showed that frequency selectivity and intensity tuning vary systematically across the cortical surface 3, 6, 31. The lack of a clear organization based on frequency selectivity and intensity tuning here might be due to some key technical differences. In contrast to a course sampling of neurons by repeated penetrations with a single electrode, 2–photon imaging reports cell activity with single cell resolution of all loaded neurons in an area. Thus, 2–photon imaging removes biases in cell selection 21 and crosstalk from nearby neurons.

Moreover, neuronal responses are modulated by the type and depth of anesthesia, and hence that can alter the observed organization. Early studies in unanesthetized cats had indicated weak tonotopy and a lack of organization based on spectral integration bandwidth in A14143. Similarly, studies in awake primate A1 indicated a lack of clear organization based on spectral integration bandwidth and intensity preference 44. Both of these results are similar to our study in the isofluorane (0.5–1%) anesthetized mouse. In contrast, studies showing the presence of tight tonotopy and an organization based on spectral integration and best intensity were mostly done in deeply barbiturate (or ketamine) anaesthetized cats, rodents, and primates 3, 6, 10. Thus, while (gross) tonotopy is present in all conditions, its precision and the presence of other organizational features might depend on the specific anesthetic state of the animal. Since the suppression of neuronal processing by anesthesia depends on its type and depth 45, anesthesia in varying degrees could selectively highlight organization present due to shared inputs (Fig. 8). Alternatively, the lack of intensity and bandwidth maps in our study could reflect species differences as for example orientation maps are present in cat but absent in rodent visual cortex17. In addition, our data is obtained from cells primarily in layer II/III. It is possible that, response maps could have a different organization in layer IV44, 46.

The heterogeneity of frequency selectivity, intensity tuning, and bandwidth in ACX does not exclude the presence of precise maps based on other stimulus features or other more complex stimuli. Two-photon imaging offers a general fine-grained unbiased approach to discovering such maps based on response properties that can also be applied to other areas of the brain. Furthermore, when coupled with the observed dichotomy of absence and presence of maps based on primarily supra-threshold or combined supra- and sub-threshold responses, our findings indicate a possibly general feature of cortical processing in that local neurons receive shared inputs but perform independent computations.

Methods Summary

Animal Preparation

Mice (C57BL6J) aged P13 – P35 were anesthetized using 2–3% isoflurane in oxygen. A small craniotomy (~ 2 mm diameter) was made over auditory cortex (see below). The intact exposed dura was covered with saline. A hollow tube was attached to the cut contralateral ear canal for sound delivery. For imaging, isoflurane anesthesia was reduced to 0.5–1% while maintaining an areflexive state of the animal with animal skin temperature usually at ~ 33–35°C. All procedures were approved by the University of Maryland IACUC.

Confirmation of imaging location in A1

The location of the craniotomy was determined stereotactically (70% of bregma-lambda and ~2 mm ventral or ~ 4.4 mm lateral). Our dye injection site was guided by vasculature22 (Figure S1A). Following imaging we inserted DiI crystals into the imaging site with a 26 gauge needle. Brain was stored in 4% paraformaldehyde at 38°C, for > 3 weeks and slices (100–200 µm) were cut to confirm labeling in the medial geniculate body (MGB) (Figure S1B). We also confirmed the imaging location by injecting anterograde tracers (choleratoxin–B) into the MGB stereotactically. After 3 days, following a craniotomy in the imaging location the presence of terminals of MGB projections in layer 3 and 4 (Fig. 1a–c) was verified. Following imaging of terminals slices were cut to confirm tracer injection in the MGB (Fig. 1a). Finally we confirmed cell responses in A1–AAF based on latency of the response. Our single unit recordings (Fig. S1C) in awake mouse A1 show latencies of ~ 20 ms. While imaging at ~ 200 ms resolution one cannot detect this latency in single cells, by analyzing the time course of responses in the stimulus onset frame we detected the onset peak of neuropil responses ~ 30 ms into the frame (Fig. S1D). Thus, cells in our imaging location had similar latencies as A1 cells.

Auditory Stimulation

Sound stimuli were digitally generated using custom software written in Matlab (Mathworks) and generated by a DA board (National Instruments), anti alias filtered (PD-AAF-18. United Electronics, Inc.), attenuated (TDT PA5), and delivered (TDT EC1) via a hollow coupler tube. The sound system was calibrated over a range of 6–70 kHz andshowed a smooth spectrum (± 10 dB). Overall sound pressure level on average was ~80 dB SPL (for tones). No correction (equalization) was done to compensate for the changes in the acoustic calibration. Most ambient noise due to the laser was below 5 kHz. Since mice have very high hearing thresholds below 5 kHz 22, this noise did not affect our experiments. Acoustic stimuli were SAM (5 Hz, modulation depth of 1) tones at different frequencies or SAM noise (usually 8–64 kHz bandwidth) of different loudness since AM stimuli elicit stronger responses27 compared to non-AM stimuli. Occasionally tone or noise pip trains (100–150 ms duration at intervals of 200–300 ms) were used. Each stimulus was repeated 6–15 times at 0.1–0.25 Hz.

Two-photon calcium imaging

AM calcium dye (Oregon Green 488 BAPTA-1 (OGB–1) or Fluo–4, Invitrogen) was prepared by dissolving 50 µg of dye in 4 µl of 20% pluronic acid in DMSO (Invitrogen) and diluted (1:5–8) with ACSF15, 17 containing either 100µM AlexaFluor 594 (for visualization) or SR-101 (for visualization and astrocyte identification). Pipettes (2–4 µm tip diameter) were pulled (Sutter P2000), filled with the dye solution and introduced into the cortex. The pipette tip was visualized under two-photon scanning mode and gradually advanced. Dye was pressure injected at a depth of 350–500 µm with 5–20 psi (PV830 Pneumatic PicoPump, WPI) pressure pulses (0.2 to 1s, 50–130 pulses total over 15 min). Once loaded cells (30–60 min post injection) were observed the pipette was withdrawn and craniotomy was covered with warm agarose and cover-slipped. Imaging was done using a two-photon laser-scanning microscope (Ultima IV, Prairie Technologies) with a Spectra Physics Mai Tai Deep See Ti-Sapphire mode-locked femto-second laser. Excitation wavelength was 800 or 810 nm for OGB1 and Fluo–4, and 870 nm for SR–101. Cells were imaged using a 20× or 40× water immersion objective (LUMPlanFI/IR Olympus 0.95 or 0.8NA) at depths of usually 150–350 µm from the cortical surface. In 2-photon experiments z-sectioning increases the SNR, unlike in epifluorescent Ca2+–measurements, where the signal is contaminated by out-of-focus fluorescence. Images were acquired simultaneously in 2 channels using a 570 nm dichroic filter. Full frame images were acquired at a resolution of 256×256 pixels at 4–10Hz. Sequences (15–30 frames) were acquired for each stimulus (onset usually at the 10th frame, duration of 5–10 frames). Cells were considered to have a significant response if the 95% confidence interval of the mean dF/F value in at least one of the frames during stimulus presentation did not encompass zero (baseline). In some experiments to acquire data at faster rates we performed linescans (50–100Hz) by drawing lines, freehand, along which the laser scanned. Scanning was repeated usually 1000 times (~ 10–20 ms each line). Analysis of dF/F was performed as in frame scans (Fig. S2).

Sulforhodamine (SR–101) as indicator for astrocytes

SR–101 has been used as a specific marker for astrocytes32, 47 and we often found selective uptake of SR–101 by morphologically astrocyte like cells. However in a number of experiments we found SR–101 to be taken up by other cells possibly neurons, especially after 2–3 hours following dye loading. Neurons might take up SR–101 based on activity or possibly through gap junctions48. Since our observations did not indicate selective uptake of SR101 by astrocytes we could not reliably separate neuron from glia as has been reported previously12.

Negative fluorescence changes

We observed negative fluorescence changes with some stimuli in response to stimulation (on–period) or in the period following stimulus presentation (off-period) with either OGB–1 or Fluo–4. This might reflect inhibition in the Ca2+ signal33, 49. As the fluorescence trace includes sub-threshold activity (less with Fluo–4 than OGB, Fig. 8e–g) along with inhibitory responses and information about spiking we used the clustering approach of the entire fluorescence waveform (see below) as a measure of the response to stimulus to obtain a more general depiction of the activity than detecting spikes only or estimating firing rates from the fluorescence signal as done previously37, 49.

Two-photon image analysis

Images were analyzed using custom software written in Matlab (Mathworks). Cells were visualized using the average image of all frames. Cells were marked with a circle of 2 or 3 pixels radius, which usually encompassed the soma. Base line fluorescence was estimated from 3–6 frames preceding stimulus onset, as an average from the multiple stimulus repeats. Mean fluorescence of each cell was estimated for all the frames and then converted into dF/F. Bootstrap mean dF/F values are used for all calculations. 95% confidence intervals of estimates of mean dF/F were obtained using bootstrap resampling. For some of the data we employed motion correction. For X–Y plane movements we used a correlation-based correction with translation in X–Y over 5–10 pixels. If the pixel-by-pixel correlation increased significantly from the correlation at the (0,0) location the image was shifted accordingly. Such shifts did not happen often and did not cause significant changes in CF or best intensity of cells. With lack of signal improvement and this correction being time intensive it was not routinely performed. For Z movements we performed averaging over 8–15 trials to eliminate noise introduced by such fluctuations when present. Correction based on dF/F values obtained from the red (> 570nm) channel did not cause changes in CF and best intensity and hence was not performed routinely. In particular such correction is not ideal as it assumes dF/F changes in the red channel are due to Z movements exclusively, which is not true due to leakage of dye signal into the red channel.

Reliability Analysis

Mean reliability to a set of stimuli (either a set of SAM tones of different frequencies or a set of SAM broadband noise of different intensities) was calculated from responses to those stimuli in the set to which there was a significant mean response based on 95% confidence intervals. When there was a significant response in the mean, single trial responses were analyzed by thresholding the Ca2+ signal at 1.96 of the standard deviation of each frame. The number of trials in which the dF/F crossed threshold at least once in the stimulation frames were considered as responsive trials, and based on that reliability was calculated as the fraction of responsive trials. The mean of that for each cell over a set of stimuli was calculated and used in the distribution of Fig. 2c.

Response Measures

Best intensity was defined as the intensity at which the maximum of mean response (dF/F) was the highest. If this intensity was not the highest intensity used we checked if there was at least one higher intensity at which the response was significantly lower, otherwise the highest intensity was defined as the best intensity. CF for each cell was defined as the frequency producing the largest significant dF/F. Bandwidth of tuning curves was determined as the frequency extent at half-maximum response strength. Extrapolation or symmetry assumption was used for cases at the boundary.

Cluster analysis

To investigate local correlations of responses and detect classes of neurons with similar response properties with spatial segregation clustering on the Ca2+ responses was performed. A set of responses of cells in a field of view (either responses to noise at different intensities or to 3 tones spaced 1/4th octaves apart at fixed intensity) the dF/F values at every frame during the stimulus and the frames following the end of the stimulus were appended together to form a single response vector ri, i = cell number. A response matrix R = [r1 r2 … ri … rN], of all the cells (N = total number of cells) was created. Principal components analysis was performed on the matrix to obtain a reduced representation (number of components were such that the 99.5% energy was retained). Following this k-means clustering was performed on the principal component projections. k of k-means was varied from 2–15 and the clustering was stopped when the ratio of the mean intra-cluster Euclidean distances and the mean inter-cluster distances was less than 1. The corresponding k was used as the number of clusters. Tighter clusters with the criterion set to 0.5 did not change result of the differences between OGB–1 and Fluo–4. Absence of clustering was indicated if the criterion was not reached by k = 15. Maximum k (15) was chosen as median number of cells in the field of view was ~35, and k = 15 corresponds to average 2 cells per cluster following which clusters would have 1 cell indicating lack of clustering.

In vitro recording

Slice physiology methods are as published previously 50. Mice are deeply anesthetized with isofluorane (Halocarbon). A block of brain containing ACX was removed and slices (350–400 µm thick) were cut on a vibrating microtome (Leica) in ice-cold ACSF containing (in mM): 130 NaCl, 3 KCl, 1.25 KH2PO4, 20 NaHCO3, 10 glucose, 1.3 MgSO4, 2.5 CaCl2 (pH 7.35–7.4, in 95%O2–5%CO2), incubated for 1 hour in ACSF at 30C and then at room temperature. For recording, slices were held in a chamber under the 2-photon microscope and superfused (2–4 ml/min) with ACSF at room temperature. Electrodes were filled with (in mM) 110 K-gluconate, 4 KCl, 4 NaCl, 0.2 CaCl2, 10 HEPES, 2 Mg-ATP, 1 MgCl2 and 5 glutathione (pH 7.2, 300 mOsm). 40uM OGB–1 or Fluo–4 (K-salt, Invitrogen) was added to the pipette solution at the day of the experiment. Whole-cell recordings were performed with a patch clamp amplifier (Multiclamp 700B, Axon Instruments). Data were acquired with an AD board (National Instruments) using custom software written in MATLAB (Mathworks). Electrical stimulation was applied with a stimulus isolator (Cygnus). Stimuli were applied every 30s and responses to 10–20 repeats were averaged. Sequences of images were acquired during whole cell recordings at ~ 10 Hz, similar to in vivo experiments. Imaging frame duration, dwell times and analysis were similar to in vivo experiments.

Supplementary Material

Acknowledgements

Supported by NIDCD R21 DC009454 (POK), NIDCD R01DC009607 (POK), RO1DC005779 (SAS), AFOSR DURIP (SAS & POK), and ISR Seed Grant (SAS & POK). The authors also thank D. Winkowski for many helpful comments and help with imaging and J. Zemskova and A. Sheikh for histological help.

Footnotes

Author contributions: SB performed in vivo studies. SB and POK performed in vitro studies. POK planned and supervised the project. SB, SAS, and POK contributed to experimental design, discussed the results and wrote the manuscript.

References

1. Mountcastle VB. The columnar organization of the neocortex. Brain. 1997;120(Pt 4):701–722. [PubMed]
2. Hubel DH, Wiesel TN. Ferrier lecture. Functional architecture of macaque monkey visual cortex. Proc R Soc Lond B Biol Sci. 1977;198:1–59. [PubMed]
3. Schreiner CE, Winer JA. Auditory cortex mapmaking: principles, projections, and plasticity. Neuron. 2007;56:356–365. [PMC free article] [PubMed]
4. Koulakov AA, Hromadka T, Zador AM. Correlated connectivity and the distribution of firing rates in the neocortex. J Neurosci. 2009;29:3685–3694. [PMC free article] [PubMed]
5. Hromadka T, Deweese MR, Zador AM. Sparse representation of sounds in the unanesthetized auditory cortex. PLoS Biol. 2008;6:e16. [PMC free article] [PubMed]
6. Schreiner CE, Read HL, Sutter ML. Modular organization of frequency integration in primary auditory cortex. Annu Rev Neurosci. 2000;23:501–529. [PubMed]
7. Bizley JK, Nodal FR, Nelken I, King AJ. Functional Organization of Ferret Auditory Cortex. Cereb Cortex. 2005 [PubMed]
8. Merzenich MM, Knight PL, Roth GL. Cochleotopic organization of primary auditory cortex in the cat. Brain Res. 1973;63:343–346. [PubMed]
9. Ehret G. The auditory cortex. J Comp Physiol [A] 1997;181:547–557. [PubMed]
10. Recanzone GH, Schreiner CE, Sutter ML, Beitel RE, Merzenich MM. Functional organization of spectral receptive fields in the primary auditory cortex of the owl monkey. J Comp Neurol. 1999;415:460–481. [PubMed]
11. Nelken I, et al. Large-scale organization of ferret auditory cortex revealed using continuous acquisition of intrinsic optical signals. J Neurophysiol. 2004;92:2574–2588. [PubMed]
12. Kalatsky VA, Polley DB, Merzenich MM, Schreiner CE, Stryker MP. Fine functional organization of auditory cortex revealed by Fourier optical imaging. Proc Natl Acad Sci U S A. 2005;102:13325–13330. [PubMed]
13. Blasdel GG, Salama G. Voltage-sensitive dyes reveal a modular organization in monkey striate cortex. Nature. 1986;321:579–585. [PubMed]
14. Tsytsarev V, Fukuyama H, Pope D, Pumbo E, Kimura M. Optical imaging of interaural time difference representation in rat auditory cortex. Front Neuroengineering. 2009;2:2. [PMC free article] [PubMed]
15. Stosiek C, Garaschuk O, Holthoff K, Konnerth A. In vivo two-photon calcium imaging of neuronal networks. Proc Natl Acad Sci U S A. 2003;100:7319–7324. [PubMed]
16. Svoboda K, Yasuda R. Principles of two-photon excitation microscopy and its applications to neuroscience. Neuron. 2006;50:823–839. [PubMed]
17. Ohki K, Chung S, Ch'ng YH, Kara P, Reid RC. Functional imaging with cellular resolution reveals precise micro-architecture in visual cortex. Nature. 2005;433:597–603. [PubMed]
18. Yasuda R, et al. Imaging calcium concentration dynamics in small neuronal compartments. Sci STKE. 2004;2004:15. [PubMed]
19. Thomas D, et al. A comparison of fluorescent Ca2+ indicator properties and their use in measuring elementary and global Ca2+ signals. Cell Calcium. 2000;28:213–223. [PubMed]
20. Paredes RM, Etzler JC, Watts LT, Zheng W, Lechleiter JD. Chemical calcium indicators. Methods. 2008;46:143–151. [PMC free article] [PubMed]
21. Shoham S, O'Connor DH, Segev R. How silent is the brain: is there a "dark matter" problem in neuroscience? J Comp Physiol A Neuroethol Sens Neural Behav Physiol. 2006;192:777–784. [PubMed]
22. Stiebler I, Neulist R, Fichtel I, Ehret G. The auditory cortex of the house mouse: left-right differences, tonotopic organization and quantitative analysis of frequency representation. J Comp Physiol [A] 1997;181:559–571. [PubMed]
23. Wang X, Kadia SC. Differential representation of species-specific primate vocalizations in the auditory cortices of marmoset and cat. J Neurophysiol. 2001;86:2616–2620. [PubMed]
24. Galindo-Leon EE, Lin FG, Liu RC. Inhibitory plasticity in a lateral band improves cortical detection of natural vocalizations. Neuron. 2009;62:705–716. [PMC free article] [PubMed]
25. Barbour DL, Wang X. Contrast tuning in auditory cortex. Science. 2003;299:1073–1075. [PMC free article] [PubMed]
26. Kowalski N, Depireux DA, Shamma SA. Analysis of dynamic spectra in ferret primary auditory cortex. I. Characteristics of single-unit responses to moving ripple spectra. J Neurophysiol. 1996;76:3503–3523. [PubMed]
27. Wang X, Lu T, Snider RK, Liang L. Sustained firing in auditory cortex evoked by preferred stimuli. Nature. 2005;435:341–346. [PubMed]
28. deCharms RC, Blake DT, Merzenich MM. Optimizing sound features for cortical neurons. Science. 1998;280:1439–1443. [PubMed]
29. Nelken I, Rotman Y, Bar Yosef O. Responses of auditory-cortex neurons to structural features of natural sounds. Nature. 1999;397:154–157. [PubMed]
30. Sadagopan S, Wang X. Level invariant representation of sounds by populations of neurons in primary auditory cortex. J Neurosci. 2008;28:3415–3426. [PubMed]
31. Polley DB, Read HL, Storace DA, Merzenich MM. Multiparametric auditory receptive field organization across five cortical fields in the albino rat. J Neurophysiol. 2007;97:3621–3638. [PubMed]
32. Schummers J, Yu H, Sur M. Tuned responses of astrocytes and their influence on hemodynamic signals in the visual cortex. Science. 2008;320:1638–1643. [PubMed]
33. Li Y, Van Hooser SD, Mazurek M, White LE, Fitzpatrick D. Experience with moving visual stimuli drives the early development of cortical direction selectivity. Nature. 2008;456:952–956. [PMC free article] [PubMed]
34. Smetters D, Majewska A, Yuste R. Detecting action potentials in neuronal populations with calcium imaging. Methods. 1999;18:215–221. [PubMed]
35. Young ED, Sachs MB. Representation of steady-state vowels in the temporal aspects of the discharge patterns of populations of auditory-nerve fibers. J Acoust Soc Am. 1979;66:1381–1403. [PubMed]
36. Bandyopadhyay S, Young ED. Discrimination of voiced stop consonants based on auditory nerve discharges. J Neurosci. 2004;24:531–541. [PubMed]
37. Sato TR, Gray NW, Mainen ZF, Svoboda K. The Functional Microarchitecture of the Mouse Barrel Cortex. PLoS Biol. 2007;5:e189. [PMC free article] [PubMed]
38. Song S, Sjostrom PJ, Reigl M, Nelson S, Chklovskii DB. Highly nonrandom features of synaptic connectivity in local cortical circuits. PLoS Biol. 2005;3:e68. [PubMed]
39. Yoshimura Y, Dantzker JL, Callaway EM. Excitatory cortical neurons form fine-scale functional networks. Nature. 2005;433:868–873. [PubMed]
40. Yoshimura Y, Callaway EM. Fine-scale specificity of cortical networks depends on inhibitory cell type and connectivity. Nat Neurosci. 2005;8:1552–1559. [PubMed]
41. Evans EF, Ross HF, Whitfield IC. The spatial distribution of unit characteristic frequency in the primary auditory cortex of the cat. J Physiol. 1965;179:238–247. [PubMed]
42. Goldstein MH, Jr., Abeles M. Note on tonotopic organization of primary auditory cortex in the cat. Brain Res. 1975;100:188–191. [PubMed]
43. Goldstein MH, Jr., Abeles M, Daly RL, McIntosh J. Functional architecture in cat primary auditory cortex: tonotopic organization. J Neurophysiol. 1970;33:188–197. [PubMed]
44. Recanzone GH, Guard DC, Phan ML. Frequency and intensity response properties of single neurons in the auditory cortex of the behaving macaque monkey. J Neurophysiol. 2000;83:2315–2331. [PubMed]
45. Christiaan Stronks H, Aarts MC, Klis SF. Effects of isoflurane on auditory evoked potentials in the cochlea and brainstem of guinea pigs. Hear Res. 2009 [PubMed]
46. Wallace MN, Palmer AR. Laminar differences in the response properties of cells in the primary auditory cortex. Exp Brain Res. 2008;184:179–191. [PubMed]
47. Nimmerjahn A, Kirchhoff F, Kerr JN, Helmchen F. Sulforhodamine 101 as a specific marker of astroglia in the neocortex in vivo. Nat Methods. 2004;1:31–37. [PubMed]
48. Keifer J, Vyas D, Houk JC. Sulforhodamine labeling of neural circuits engaged in motor pattern generation in the in vitro turtle brainstem-cerebellum. J Neurosci. 1992;12:3187–3199. [PubMed]
49. Moreaux L, Laurent G. A simple method to reconstruct firing rates from dendritic calcium signals. Front Neurosci. 2008;2:176–185. [PMC free article] [PubMed]
50. Zhao C, Kao JPY, Kanold PO. Functional Excitatory Microcircuits in Neonatal Cortex Connect Thalamus and Layer 4. Journal of Neuroscience. 2009;29:15479–11488. [PMC free article] [PubMed]