| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1Vanderbilt Kennedy Center for Human Development and the Department of Hearing and Speech Sciences, Vanderbilt University, Nashville, Tennessee; 2Department of Psychology, University of Connecticut, Storrs, Connecticut; and 3 Department of Otolaryngology—Head and Neck Surgery, W.M. Keck Foundation Center for Integrative Neuroscience, University of California at San Francisco, San Francisco, California
Submitted 12 December 2006; accepted in final form 11 March 2007
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
; Kelly and Masterton 1977; Talwar and Gerstein 1998), and has a central auditory system that is comparable to other mammals in its anatomical and functional organization (for review see Malmierca 2003). The rat auditory cortex, in particular, has been popular among auditory neuroscience researchers interested in experience-dependent plasticity (for recent examples see Bao et al. 2003; Engineer et al. 2004; McLin et al. 2003; Polley et al. 2004; Sun et al. 2005), pharmacological regulation of physiological and/or perceptual processes (for recent examples see Liang et al. 2006; Ling et al. 2005; Talwar et al. 2001), and the synaptic basis of receptive field organization (Kaur et al. 2004; Tan et al. 2004, 2007; Wehr and Zador 2003; Wu et al. 2006; Zhang et al. 2003). However, unlike more established animal models for auditory cortical function, such as the cat or nonhuman primate, investigations into pharmacological or experiential modifications of physiological processing in the rat auditory cortex predate a body of work that describes the functional organization of the normative rat auditory cortex in sufficient detail.
To date, tonal receptive fields, intensity-response profiles, and basic excitatory response properties were characterized for three fields of the normative rat auditory cortex: the primary auditory cortex (AI), the posterior auditory field (PAF), and the anterior auditory field (AAF) (Doron et al. 2002; Horikawa et al. 1988; Kilgard and Merzenich 1999; Phillips and Kelly 1989; Rutkowski et al. 2003; Sally and Kelly 1988). Anatomical tracer studies conducted in the rat confirmed that AI lies in cortical area TEI and receives heavy projections from the ventral division of the medial geniculate body (MGv) into the middle cortical layers (Kimura et al. 2003; Roger and Arnault 1989; Romanski and LeDoux 1993b; Winer et al. 1999). Electrophysiological recordings within AI describe well-tuned, short-latency responses that exhibit an orderly tonotopic progression of characteristic frequency (CF) from low (
1 kHz) to high (60 kHz) along a posterior-to-anterior gradient (Sally and Kelly 1988). The tonotopic gradient reverses at the posterior and anterior borders of AI to form boundaries with PAF and AAF, respectively. A recent optical imaging study confirmed the relative position and tonotopic organization of AI and AAF within a single rat and also described the existence of three additional auditory areas ventral to AI: the ventral auditory field (VAF), the ventral anterior auditory field (VAAF), and the ventral posterior auditory field (Kalatsky et al. 2005). In a recent electrophysiological mapping study, we characterized the receptive field organization within the region described as VAAF in the Kalatsky study and renamed it the suprarhinal auditory field (SRAF), explained by the fact that the low-frequency core of the tonotopic gradient is located about 0.75 mm dorsal to the rhinal fissure (Polley et al. 2006). The Kalatsky study described VAF, also referred to as the "nonprimary ventral auditory field" (e.g., Bao et al. 2003; Fig. 6 and Bao et al. 2001; Fig. 1B or "nonmonotonic auditory zone" in Wu et al. 2006), as an independent auditory field interposed between AI and SRAF. Based on the tonotopic organization of VAF visualized with the optical imaging technique, some questions have arisen as to whether VAF was simply an extension of the ventral boundary of AI or whether it was a bonafide cortical field. One of the motivations for the present study was to compare the sources of thalamic input into AI and VAF with retrograde tracer injections and compare the receptive field characteristics in these two fields for features beyond tonotopy, to more accurately determine whether VAF met the criteria for an independent auditory field.
|
|
|
|
|
|
; Schreiner 1998). Unlike the primary visual cortex, however, CF is the only receptive field feature that progresses in a smooth gradient across the entire spatial extent of an auditory cortical field. The spatial organization of other receptive field parameters, such as spectral bandwidth, binaural interaction type, minimum response latency, and intensity tuning exhibit nonrandom spatial organizations that are independent of CF but often take the form of spatially contiguous "patches" or "modules" (Heil et al. 1994; Imaizumi et al. 2004; Imig and Adrian 1977; Kelly and Sally 1988; Middlebrooks et al. 1980; Nakamoto et al. 2004; Philibert et al. 2005; Recanzone et al. 1999). With the exception of two reports describing a spatial organization for preferred intensity and binaural interaction type (Kelly and Sally 1988; Polley et al. 2006), the spatial organization of receptive fields for features other than CF were not described in any auditory cortical field for the rat. This is surprising because the auditory cortex of the rat offers two distinct advantages for investigators interested in describing the spatial organization of multiparametric receptive field characteristics: First, the rat cortex is lissencephalic, making it possible to record from all regions of the map not obscured by surface vessels without "losing" regions of the map to variations in surface topography (i.e., sulci). Second, the surface area of the rat auditory cortex is substantially smaller than the cat or monkey, making it more feasible to derive maps for multiple receptive field parameters across multiple cortical fields in the same animal while maintaining high spatial sampling resolution.
The present study seeks to accomplish the following: 1) confirm the existence and relative position of AI, PAF, AAF, VAF, and SRAF in the rat using high-density microelectrode mapping and confirming their separate identities, when necessary, with anatomical tracer injections; 2) document spectral tuning, intensity-response functions, and excitatory response properties within each field; and 3) quantify the extent to which various receptive field parameters exhibit nonrandom spatial order within each cortical field. These data were previously presented in abstract form (Polley and Merzenich 2006).
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
Sixteen healthy adult Sprague–Dawley rats (age 14–18 wk) were used in the study. All surgical procedures were approved by the institutional care and use committee of the University of California at San Francisco and the University of Connecticut. Rats were anesthetized with pentobarbital sodium (50 mg/kg followed by 10–15 mg/kg supplements as needed). Respiratory rate, heart rate, corneal, and hind-paw withdrawal reflexes were monitored to ensure that a moderately deep anesthetic plane was maintained as uniformly as possible throughout the recording procedure. Atropine sulfate (0.1 mg/kg) and dexamethasone (0.25 mg/kg) were administered at the beginning of the surgery and every 10 h thereafter to reduce brain edema and the viscosity of bronchial secretions, respectively. A tracheotomy was performed and a cisternal drain introduced after the rat reached a surgical plane of anesthesia, to further minimize bronchial secretions and brain edema. Saline and Ringer solution were administered periodically throughout the experiment to ensure adequate hydration. Body temperature was maintained near 37.3°C with a rectal probe and homeothermic blanket system (Harvard Apparatus). Rats were placed in a head clamp manufactured in-house that allowed unimpeded access to the external auditory meati. The temporal muscle, cranium, and dura overlying the auditory cortex were removed and the exposed cortex was covered with either silicone oil for electrophysiological recording or agar and sealed with a glass coverslip for optical imaging.
Electrophysiological recording procedures
All experiments were conducted in a sound-insulated chamber (Acoustic Systems). Tone pips (20-ms duration, 5-ms raised cosine ramps) and white noise bursts (100-ms duration, 5-ms raised cosine ramps) were delivered monaurally to the contralateral ear by a calibrated STAX headphone enclosed in a metal canister that was directly coupled to the external auditory meati with a sound delivery tube. Acoustic calibration was performed with a 1/2-in. Brüel & Kjær microphone sealed to the sound delivery tube so that the speaker output could be corrected to ensure a flat frequency response (±3 dB SPL) with minimal harmonic distortion (THD <2%) over a range of frequencies spanning 1–30 kHz.
Multiunit responses were recorded with epoxylite-coated tungsten electrodes (1.5-M
impedance at 1 kHz; FHC) oriented orthogonal to the pial surface and advanced 450–575 µm into the thalamorecipient layers (layer III/IV). Two to four electrodes were simultaneously inserted into the cortex. The electrode signal was amplified, filtered (0.3–5 kHz), displayed, and stored with hardware and software manufactured by Tucker-Davis Technologies. Recording sites (about four per mm2; 100- to 200-µm spacing) were evenly distributed across the cortical region of interest while avoiding blood vessels.
Optical recording procedures
Custom software and hardware for stimulus presentation, acquisition, and analysis of optical intrinsic activity is described by Kalatsky and colleagues (2005). Intrinsic signal responses were imaged with a Dalsa 1M30 CCD camera with a 512 x 512 pixel array covering a 4.6 x 4.6-mm2 area. The surface vascular pattern was visualized under green light (546 nm). Intrinsic signals were obtained by defocusing the camera 650 µm and visualizing activity under red light (610 nm). Sounds were delivered to both ears by hollow ear tubes and were calibrated for frequency distortions in the closed system. Transient pure tones (50-ms duration, 5-ms rise and fall times) were varied in frequency (exponential steps, 2–32 kHz).
Tracer injection and histology
Retrograde tracers were injected into functionally defined auditory cortical fields in five animals. Double injections were made in three of the five rats: Cholera Toxin Beta-subunit (CTB, List Biological Laboratories) was injected into AI and CTB conjugated with gold (CTB-gold, List Biological Laboratories) was injected into VAF. In two additional rats a single injection was made into VAF with either tracer, to confirm thalamic areal patterns. Injection volumes were 10–20 nl and corresponding deposits confirmed histologically were 200 µm in diameter and >400 µm in height, on average. After a 20- to 24-h survival time under anesthesia, rats were given a lethal dose of pentobarbital and perfused transcardially with 4% paraformaldehyde. The cortex was blocked and cut into 60-µm sections with a vibratome. CTB label was immunoreacted with appropriate primary and secondary antibodies and processed with diaminobenzadine. CTB-gold label was intensified with silver for light microscopy. Alternate sections were reacted for CTB alone, CTB-gold, and for both for double-injection cases.
Data analysis
TUNING CURVE ANALYSIS. Neural responses to pure-tone pips and white noise bursts were recorded at each penetration site. Tuning curves were reconstructed by presenting 60 pure-tone frequencies (1–30 kHz, 20-ms duration, 5-ms raised cosine ramps) at 11 sound intensities (0–70 dB SPL, 7 dB SPL increments). Onsets of sequential tone pips were separated by 400 ms. Characteristic frequencies (CFs) were defined for neurons at each sampled cortical site as the frequency that evoked a reliable response at lowest threshold. Tuning-curve bandwidths were estimated at 14 and 42 dB SPL above threshold and were expressed as a Q-factor value (CF/bandwidth). Note that tuning-curve frequency response areas were sampled at 7 dB SPL increments, necessitating the use of Q-factor measurements at sound levels evenly divisible by 7. For this reason we measured Q14 and Q42 rather than more commonly used measures such as Q10 or Q40.
ANALYSIS OF EXCITATORY RESPONSE PROPERTIES. Bursts of white noise were presented at 17 intensities ranging from 0 to 80 dB SPL in 5 dB SPL increments. Each intensity was presented 20 times in a pseudorandom order. Onsets of two sequential noise bursts were separated by 600 ms. A population peristimulus time histogram (PSTH) was created by summing the responses to all white noise stimuli occurring within a window beginning 100 ms before stimulus onset and terminating 200 ms poststimulus onset. Onset latency was defined as the intersection between the slope of the excitatory response profile and the mean prestimulus firing rate. The response duration was defined as the length of time between response onset and the point at which the excitatory response had decreased back to an amplitude equivalent to that measured at the onset latency.
RATE-LEVEL FUNCTION ANALYSIS.
Rate-level functions (RLFs) were reconstructed by summing all spikes occurring within the response duration for each intensity, expressing that value as spikes per second, subtracting the mean spontaneous (
100 ms before stimulus onset) firing rate also expressed as spikes/s from that value and then plotting the evoked firing rate for each sound intensity. The peak firing rate was defined as the maximum value within this function. Each RLF was normalized to its maximum firing rate and the following measures (defined below) were derived according to the method outlined by Schreiner et al. (1992): 1) minimum response threshold; 2) transition point; 3) best level; 4) slope of the RLF between minimum response threshold and the transition point; and 5) monotonicity. If the spike rate remained at zero for two consecutive sound levels, all sound levels less than or equal to the greater of the two levels were considered subthreshold. Minimum response threshold was defined as the first sound level in the suprathreshold region of the RLF (indicated by an arrow in Fig. 4A). In many cases, the RLF consisted of a fast-growing, low-intensity portion and a saturated or decreasing response function at higher sound intensities. The transition point was defined as the highest sound level within the fast-growing region (indicated by the diamond in Fig. 4A). Best level was defined as the sound level that evoked the greatest magnitude response. Monotonicity was defined as the slope of the RLF between the transition point and the highest sound level estimated by a linear regression analysis. In the event that the RLF increased linearly above the minimum response threshold (e.g., first example in Fig. 5A) and a transition point could not be determined, a linear regression analysis was performed on all sound levels above threshold. In both cases, the slope of the regression function was used as a quantitative measure of monotonicity, whereby a negative slope corresponded to a nonmonotonic response and a slope of 
0 corresponded to a monotonic response function. A minimum of five sound intensity values, including the transition point, had to be included in the linear regression analysis or a monotonicity value was not determined for that recording site.
CONTIGUITY ANALYSIS.
One of the primary goals of the present study was to determine whether auditory response parameters beyond CF exhibited spatially ordered representations across the cortical surface. To address this question, we examined the spatial distributions of ten auditory response parameters: 1) CF; 2) best frequency (BF); 3) tuning-curve threshold; 4) Q14; 5) Q42; 6) onset latency; 7) response duration; 8) peak firing rate; 9) best level; and 10) monotonicity. The contiguity analysis was restricted to all penetrations within a cortical field that contained data for all of these ten variables. We created an array consisting of all combinatorial possibilities of recording location pairings assessed independently for each field. The maximum difference (MaxD) was independently calculated for each variable from this array as the pair of recording penetrations that exhibited the maximum difference for the variable of interest. We then calculated the similarity index for each unique pair of recording penetrations as a fraction of MaxD based on the following calculation
 |
The distributions of CF, BF, threshold, and response duration values were fairly flat across all penetrations obtained within a single map. In other words, all values between the minimum and maximum had a roughly equal representation. This was not the case for the remaining variables. These distributions were often multimodal with a group of penetrations that, for instance, exhibited substantially higher Q-factor, longer latency, or lower best level values. For that reason, a variation of the above method was performed for analysis of Q14, Q42, onset latency, peak firing rate, best level, and monotonicity. In these cases, only the penetrations that yielded values in the top 15% (Q14, Q42, onset latency, and peak firing rate) or bottom 15% (best level and monotonicity) of the distribution in each field were used as Vreference. In so doing, we determined whether sites with the sharpest tuning, longest latency, highest firing rate, lowest best levels, or most nonmonotonic RLFs shared this similarity with neighboring penetrations beyond what would be expected by chance. We conducted the analysis using other combinations of nearest-neighbor points (5 and 7) and applying other cutoff limits to the distributions (10 and 20%), but concluded that the results obtained with three neighbor points and a 15% cutoff limit yielded the most interpretable results.
Tessellated maps were generated by defining the Cartesian coordinates and response property of interest for each penetration site and applying the Voronoi function in Matlab (The MathWorks). Frequency-intensity recruitment functions were calculated by determining the polygonal area associated with a single penetration within the map of a cortical field and then calculating the range of sound frequencies that elicited a response from that site at each of the 11 sound intensities. The area of all polygons within the map activated by a tone of a specific frequency and intensity was then summed and expressed as a percentage of the total map area. Statistical analyses were performed using a combination of parametric and nonparametric statistical tests offered by the Systat statistical software package. The Bonferroni correction factor was applied in all instances where multiple comparisons were made using the same value. In these cases, the confidence interval (
) was reduced according to the number of comparisons (n) such that = 0.05/n.
Sample size for electrophysiological analyses
A total of 23 complete maps (AI = 6; VAF = 4; SRAF = 7; AAF = 3; PAF = 2) were reconstructed from 11 rats. In many cases, more than one cortical field was mapped in an individual rat. A map of all five fields was obtained in one rat, three fields were mapped in one rat, five rats yielded two complete maps, and a single field was mapped in four rats. Maps were defined from a total of 1,634 recording sites. Of these sites, 1,281 (AI = 406; VAF = 176; SRAF = 526; AAF = 149; PAF = 24) were sufficiently well tuned to tone pips to be included within the maps. The remainder were either unresponsive to auditory stimuli, or were excluded because they did not exhibit measurable tuning to tone frequency. Of the 1,281 sites that yielded data for tuning curve analysis, 1,224 (AI = 390; VAF = 176; SRAF = 495; AAF = 140; PAF = 23) exhibited a clear onset response to white noise bursts and were used for analysis of onset latencies, response durations, and peak firing rates. Of the 1,224 sites that yielded data for the PSTH analysis, 1,151 (AI = 376; VAF = 147; SRAF = 491; AAF = 137; PAF = 0) yielded data for all measured variables and were used in contiguity and RLF analyses. We elected not to include any recordings from PAF in contiguity or RLF analyses because 1) the sample size was small; 2) RLFs obtained from PAF recordings were extremely noisy and difficult to analyze; and 3) the areal extent of PAF was small and therefore not well suited for the contiguity analysis.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
The relative position of each auditory cortical field was determined with high-density microelectrode mapping. Multiunit activity was recorded in cortical areas TE1 and TE3 about 3–7 mm posterior of bregma and 3–6 mm lateral to bregma, from the area surrounding the middle cerebral artery to ventral regions surrounding the rhinal vein (Fig. 1, A and B). Unlike a previous report (Doron et al. 2002), we did not observe that the positions of each field assumed a precisely stereotyped position with respect to bregma or the surface vasculature. We observed that the relative orientation of each cortical field was fairly constant between individual rats, but their position in relation to bregma or major vascular landmarks could vary by as much as 0.5 mm in any direction. Individual electrode penetrations were separated by 100–200 µm (Fig. 1B) across the cortical region of interest. Multiunit responses to tone pips and white noise bursts were recorded at each site. Figure 1C illustrates the spatial distribution of CF values obtained from the recordings sites shown in Fig. 1B using a tessellated map. CF values are organized into several clearly discernible tonotopic gradients. Boundaries between contiguous cortical fields were drawn according to reversals or shifts in the tonotopic progression of CF gradients (Fig. 1D). AI, AAF, VAF, and SRAF exhibited orderly tonotopic progressions of CF, whereas PAF appeared as a thin nontonotopic "crescent" abutting the posterior edge of AI with CF values restricted to middle and high frequencies. Areal extent (mean ± SE) for each field was measured as follows: AI = 1.35 ± 0.16 mm2; VAF = 1.25 ± 0.15 mm2; SRAF = 1.51 ± 0.07 mm2; AAF = 1.21 ± 0.13 mm2; PAF = 0.41 ± 0.13 mm2. Although there was not a tonotopic reversal between AI and VAF, we provisionally define VAF as a cortical field independent from AI, based on the approximate 90° shift in the CF gradient in the ventral low-frequency border of AI and the absence of tone-evoked responses in the ventral, high-frequency-border of AI (Fig. 1, C and D). Additional electrophysiological and anatomical evidence in support of this division are provided in the subsequent sections of this report.
Spectral tuning characteristics
Analysis of pure-tone frequency-response areas from each field revealed significant differences in basic tuning-curve characteristics. Examples of tuning curves obtained from each cortical field are shown in Fig. 2A. Multiunit responses in AI, AAF, and SRAF were strongly driven by pure-tone pips and typically exhibited the classic V-shape frequency-response area that became progressively wider with increasing tone intensity. Frequency-response areas in field VAF, by contrast, often appeared as low-threshold circumscribed "islands" or "slivers" of activity with little or no tone-driven responses at higher stimulus intensities. Tuning curves in field PAF were broad and often exhibited discontinuous frequency-response areas yielding overall "patchy" appearances.
A quantitative analysis of tuning curves in each field revealed significantly lower response thresholds in field VAF (F = 18.27, P < 1 x 10–6) and SRAF (F = 20.49, P < 1 x 10–5) compared with AI (Fig. 2B). This difference was primarily attributable to an increase in tuning-curve threshold across higher CF categories in AI that was less exaggerated in VAF and SRAF. Differences in tuning-curve thresholds were not observed in AAF or PAF relative to AI (F < 1.6, P > 0.2 for both comparisons). Spectral tuning bandwidths measured 14 dB SPL above threshold were narrowest in AI and AAF (AI vs. AAF, t = 0.25, P = 0.81). Q14 values in VAF, SRAF, and PAF, by contrast, were significantly lower than AI (t > 2.0, P < 0.005 for all comparisons; Fig. 2C). Tuning curve bandwidth measured 42 dB SPL above threshold was not significantly different between AI compared with AAF, SRAF, or PAF (t < 1.0, P > 0.32 for all comparisons). Q42 values measured in VAF, however, were significantly greater than in AI (t = 3.6, P < 0.0005; Fig. 2D).
Excitatory response characteristics
White noise bursts evoked a transient response from most recording locations, but a more detailed analysis of onset responses revealed significant differences in excitatory response profiles between cortical fields. PSTHs were compiled by summing the responses to all noise intensities ranging from 0 to 80 dB SPL (Fig. 3A). Onset response latency was shortest on average in AI (12.33 ± 0.26 ms) and AAF (12.18 ± 0.18 ms), which were not significantly different from one another (t = 0.45, P = 0.65). Onset responses in VAF (15.08 ± 0.4 ms), SRAF (15.73 ± 0.29 ms), and PAF (20.78 ± 2.55 ms), by contrast, were all significantly delayed relative to AI (t > 3.29, P < 0.005 for each comparison; Fig. 3B). Significantly shorter response durations were observed in AAF (15.34 ± 0.32 ms) compared with AI (18.12 ± 0.28 ms, t = 6.41, P < 1 x 10–6; Fig. 3C). Response durations were longer in VAF (19.37 ± 0.5 ms), SRAF (19.87 ± 0.43 ms), and PAF (26.3 ± 1.65 ms) compared with AI. These differences reached statistical significance for comparisons of SRAF and PAF relative to AI (t > 3.5, P < 0.001 for both tests), but failed to reach statistical significance for AI versus VAF after the Bonferroni correction for multiple comparisons (t = 2.21, P = 0.03). Last, we observed significant differences in peak firing rate between each cortical field (Fig. 3D). Peak firing rates in AAF (167.47 ± 7.84 spikes/s) were significantly higher than AI (141.71 ± 3.93 spikes/s; t = 2.9, P < 0.005). Peak firing rates in VAF (97.35 ± 5.51 spikes/s), SRAF (105.56 ± 3.06 spikes/s), and PAF (65.08 ± 9.01 spikes/s) were all significantly lower than AI (t > 7.25, P < 1 x 10–6 for all tests).
Intensity tuning characteristics
Analysis of best level and rate-level function monotonicity revealed substantial differences in neural encoding of sound intensity between cortical fields. RLFs obtained from most PAF neurons could not be analyzed reliably using our stimulus presentation and analysis protocols. It was often difficult to identify the minimum response threshold and the linear regression used to calculate montonicity was often poorly fit to the data points. RLF analysis was therefore limited to AI, AAF, VAF, and SRAF. Figure 4A presents a variety of different RLF shapes obtained from various recording locations in a single SRAF map. Best level and monotonicity become increasingly lower and more negative, respectively, from the top row to the bottom row. Distributions of best level (Fig. 4B) and monotonicity (Fig. 4C) values were compiled for AI, VAF, SRAF, and AAF. AI and AAF both exhibited best level distributions strongly skewed toward intensities 70 dB SPL, although this bias was significantly more pronounced in AAF compared with AI [Kolmogorov–Smirnov (K-S) test, P < 0.001]. The best-level distribution obtained from VAF recordings was strikingly different from AI, with the vast majority of best-level values concentrated around low-to-intermediate sound intensities (K-S test VAF vs. AI, P < 1 x 10–6). SRAF also contained a significantly greater percentage of recording sites with best-level values at low-intermediate sound intensities compared with AI (K-S test, P < 5 x 10–6). Using the RLF slope value up to –0.5 as a threshold for a nonmonotonic slope profile, we observed that the percentage of nonmonotonic RLFs varied substantially between cortical fields (AI = 17%, VAF = 76%, SRAF = 38%, AAF = 14%). A quantitative comparison of the distribution of RLF slope values between different fields revealed significant differences between AI and VAF (K-S test, P < 1 x 10–6), AI and SRAF (K-S test, P < 1 x 10–6), but not between AI and AAF (K-S test, P = 0.06; Fig. 4C).
Population response characteristics
High-density mapping offered the ability to measure the spatial extent of cortical activation to a broad range of tone frequencies and intensities, rather than solely focusing on the spatial distribution of preferred frequency presented at threshold. Analyses of this type were extensively used with noninvasive imaging methods and were termed the cortical-point spread or cortical recruitment function (Fischer 1973; Grinvald et al. 1994). In the auditory system, cortical recruitment functions are essentially the converse of tuning-curve bandwidth measurements. Rather than measuring the range of frequencies to which the neuron(s) is responsive, they measure the percentage of the total map area activated by a pure tone at a specific frequency and intensity. Cortical recruitment functions in AI, VAF, SRAF, and AAF appear roughly similar, with a small percentage of the map activated by mid-frequency tones at lower intensities and an increasingly greater percentage of the map responsive to tones of all frequencies at higher intensities (Fig. 5A). Closer inspection, however, revealed differences between the fields that were consistent with the aforementioned observations related to tuning-curve threshold and bandwidth. Figure 5, B–D presents the means ± SE percentage of the map responsive to three 1/4-octave-wide frequency ranges centered on low- (2.8 kHz, Fig. 5B), mid- (8 kHz, Fig. 5C), and high-frequency (22.6 kHz, Fig. 5D) tones. Low-frequency tones began to activate a small region of the map at 21–28 dB SPL which grew progressively across mid intensities before saturating at activating 40–50% of the map at intensities 56 dB SPL (Fig. 5A). The pattern of activation for low-frequency tones was not different between cortical fields (F = 0.5, P = 0.69). Middle-frequency tones, however, elicited significantly different activation patterns between cortical fields (F = 3.24, P < 0.05; Fig. 5B). AI, AAF, and SRAF exhibited a fairly linear increase in responsive area with increasing intensity, although the percentage of the SRAF map was greater at all intensities, indicating that receptive fields were more broadly tuned in this frequency range. The activation pattern in VAF was clearly nonlinear, peaking at 28 dB SPL and flattening out across all higher intensities. This is consistent with our previous observation that many recording sites in VAF were nonmonotonic. Although an ANOVA did not find a main effect across cortical fields (F = 1.2, P = 0.34), a similar overall pattern was observed across each cortical field for high-frequency tones (Fig. 5C). For example, in AI the linear increase observed to high-frequency tones of increasing intensity was significantly different from the nonlinear increase observed in VAF (ANOVA interaction term, F = 12.88, P < 1 x 10–6).
Analysis of thalamic input to AI and VAF
Although the traditional method for identifying independent cortical fields according to a mirror reversal in the tonotopic gradient did not hold for defining a border between AI and VAF, the clear differences in response characteristics, particularly to sounds of varying intensity, demonstrate that AI and VAF are functionally separable and suggest that they are independent cortical fields. To explicitly test this possibility, we compared the sources of thalamic input for AI and VAF by injecting retrograde tracers into CF-matched regions in each field. The tonotopic organization of the auditory cortex was visualized with Fourier intrinsic signal optical imaging, which was previously shown to yield a map of preferred tone frequency that is highly correlated with the CF map derived through microelectrode mapping (Kalatsky et al. 2005 and Fig. 6A). Injection pipettes containing CTB and/or CTB-gold were guided into the 8-kHz region in the AI and/or VAF map(s), and a small volume of tracer was injected (Fig. 6, B and C). The injection deposits spanned cortical layers II–V and were separated by a rostrocaudal distance of 567 ± 237 µm and a dorsoventral distance of 1,300 ± 208 µm.
Although injections were made into frequency-matched regions of the tonotopic maps, the locations of retrograde labeled neurons within the medial geniculate body (MGB) were substantially different for injections made into AI and VAF. We found that most MGB neurons projecting to AI were located in the ventral division (MGBv) as previously described for the rat (Ryugo and Killackey 1974; Winer et al. 1999). Similarly, we found that most MGB neurons projecting to VAF were located in the MGBv. However, the rostrocaudal position of labeled MGBv neurons was different and was minimally overlapping for injections into the two cortical regions. Injection of CTB into the 8-kHz region of AI and VAF clearly labeled a population of neurons in the lateral region of the MGBv in three cases, in accordance with the approximate lateral-to-medial topography for low- to high-frequency lamina in MGBv (Winer et al. 1999). However, the highest density of retrograde-labeled neurons projecting to AI were located 4,920 ± 159 µm posterior to bregma. By contrast, the highest density of retrograde-labeled neurons projecting to VAF were located 5,680 ± 150 µm posterior to bregma. The separation between coronal sections with the maximum density for each label was 760 ± 227 µm along the rostrocaudal axis (Fig. 6, D–F). Figure 7 presents a series of sequential double-reacted (Fig. 7, A–C) and single-reacted (Fig. 7, D–F) sections including the maximum label from VAF (Fig. 7B) and AI (Fig. 7F). Sections with labeled neurons projecting to AI had minimal (<2% in all cases) labeled neurons projecting to VAF and vice versa. Last, the presence of double-labeled neurons was examined under high magnification (x100): we were not able to identify any double-labeled neurons in the two cases with smaller injections and only five double-labeled neurons in the entire thalamic serial sequence with a larger injection volume. Thus there is minimal overlap in the MGBv regions projecting to A1 and VAF; most neurons projecting to A1 are in the rostral MGBv and most neurons projecting to VAF are in the caudal MGBv.
|
Spatial organization of multiple receptive field characteristics
In light of the thalamocortical connectivity differences between A1 and VAF, five auditory cortical fields can be identified according to: 1) these anatomical differences, 2) by reversals or shifts in tonotopy, and 3) on the basis of consistent differences in spectral tuning, excitatory response profiles, and responses to stimuli of varying intensity. As a next step, we investigated whether these other receptive field characteristics were spatially organized in a nonhomogeneous, nonrandom fashion. Tesselated maps representing the spatial distribution of CF, Q14, onset latency, and RLF monotonicity are presented in Fig. 8. Of the four response characteristics shown, only CF is spatially organized into a gradient in all fields. The other response characteristics, however, appear to be organized into contiguous patches of sharp tuning, long-latency responses, and nonmonotonic RLF slopes. The relative position of these patches was fairly consistent between maps. For instance, AI typically had a region of narrow spectral tuning in the high-frequency CF region and a patch of nonmonotonic RLFs in the ventral half of the mid- to high-frequency CF range. Maps of SRAF often featured a cluster of long-latency responses in the most ventral region of the map, nonmonotonic RLFs across the posterior half of the map, and sharp spectral tuning in the anterodorsal quadrant of the map. AAF consistently had a small nonmonotonic region in the most ventral region of the map and longer-latency responses in the most anterior portion of the map.
|
|
The average real and randomized similarity indices for the complete set of response characteristics are displayed in Fig. 10 for AI, VAF, SRAF, and AAF. For all measurements, with the exception of response duration, the receptive field characteristics shared among neighboring positions in the map were significantly greater than random (paired t-test, t > 3.89, P < 0.001 for each real vs. randomized pair). A nonrandom spatial organization for response duration was not obvious, although statistically significant differences were observed in VAF (paired t-test, t = 1.99, P < 0.05) and AAF (paired t-test, t = 2.41, P < 0.05). The average differences in similarity index (real – randomized) for each variable were compared between fields (Fig. 11). The greatest difference values, indicative of the stronger spatial clustering, was observed for CF, BF, peak firing rate, best level, and RLF monotonicity. Statistically significant spatial clustering was observed for all fields, but was greater on average in AI and AAF (mean difference = 0.16 and 0.15, respectively) than in VAF or SRAF (mean difference value = 0.13 for both).
|
|
The preceding results demonstrate that receptive field parameters other than CF are distributed across the cortical map in a nonhomogeneous fashion and that these contiguous modules reflect a degree of organization that was unlikely to have arisen by chance alone. Proof for spatially organized representations of multiple receptive field parameters, however, carries a twofold requirement: 1) a demonstration that multiple response features are organized into spatial patterns (e.g., gradients, bands, patches) and 2) that the spatial organizations are mutually independent. We evaluated this second requirement by calculating the correlation matrices for each of the ten response characteristics in AI (Fig. 12A), VAF (Fig. 12B), SRAF (Fig. 12C), and AAF (Fig. 12D). Using the pseudocolor scale as an indication of the absolute value of the Pearson R correlation coefficient (the direction of the correlation, positive or negative, is irrelevant), it was clear that some response characteristics were highly correlated with one another. Spectral-tuning characteristics such as CF, BF, Q14, and Q42 were—unsurprisingly—correlated with one another, as were intensity-tuning characteristics such as tuning-curve threshold, best level, and monotonicity. Excitatory response features that exhibited spatial clustering such as onset latency and peak firing rate were not correlated with each other or with any other response feature. The degree of independence between various response parameters was statistically determined by a factor analysis. Using a standard criterion whereby eigenvalues >1.0 indicate significant factors, we found that there were at least three significant factors within each field (Fig. 12E). Spectral-tuning characteristics (CF, BF, Q14, and Q42) loaded highly onto the factor that explained the most variance (component loadings >0.5); intensity-tuning characteristics (threshold, best level, monotonicity) loaded onto the second factor (component loadings >0.5); the remaining characteristics (onset latency, response duration, peak firing rate) loaded onto the remaining one or two factors. In summary, highly correlated response characteristics covaried among themselves but not with other groupings of response characteristics. Therefore based on the contiguity analysis that demonstrated statistically significant spatial clustering and the factor analysis that demonstrated independence between several of those clusters, we can conclude that there are multiple, independent representations for response characteristics such as spectral tuning, intensity tuning, and excitatory response properties.
|
One of the principal advantages of using the rat brain is that its relatively small physical scale permits the mapping of several contiguous auditory fields with fine spatial resolution in the same animal. A complete high-resolution microelectrode map of all tonotopically organized auditory fields is difficult if not impossible to obtain from a single cat or primate cortex, given the substantially larger total area, and the fact that parts of the maps can be buried deep in the banks of sulci. Because the rat cortex is relatively small and lissencephalic, a complete map of all tonotopically organized fields can be obtained with 100- to 150-µm resolution in 24–48 h, depending on the length of stimulus sets presented at each recording location and the number of recording electrodes used simultaneously. Complete maps of several contiguous fields can be obtained in 8–24 h. The tessellated map of all five cortical fields shown in Fig. 1B is also presented in Fig. 13. In addition to CF (Fig. 13A), the spatial distributions of Q14 values (Fig. 13B), onset latency (Fig. 13C), and RLF monotonicity (Fig. 13D) are also displayed. Based on these maps, it is evident that the spatial distributions of some receptive field characteristics do not respect the tonotopic boundaries that are used to divide one cortical field from another. Longer onset latencies are observed across a continuous narrow strip spanning the posterior regions of AI, VAF, and SRAF. Similarly, nonmonotonic RLFs are organized across a broad swath of recording sites spanning the ventral region of AI, the entirety of VAF, and the posterior region of SRAF. Each distinct type of spatial organization is represented schematically in Fig. 14. Receptive field characteristics such as CF and BF are organized into topographic gradients and the location where the gradient reverses or shifts abruptly is typically used to represent a boundary for that field. Regions of sharp spectral tuning or relatively high stimulus-elicited firing rate are organized as patches, which, in the case of Q14, can span the boundary separating individual cortical fields. Response characteristics such as onset latency, best level, and RLF monotonicity are organized into continuous "meta-gradients" that span multiple cortical fields.
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
The primary auditory cortex
We found that AI is composed of neurons with well-tuned tonal receptive fields and firing rates that most often increased monotonically with the intensity of tonal or broadband stimuli as previously described in the rat (Doron et al. 2002; Horikawa et al. 1988; Kilgard and Merzenich 1999; Phillips and Kelly 1989; Polley et al. 2004, 2006; Sally and Kelly 1988; Wu et al. 2006). The AI map featured small, spatially independent modules of sharp spectral tuning or nonmonotonic rate-level functions. This is consistent with previous reports in the cat and primate, although in those species the AI map typically features multiple modules of sharp spectral tuning or nonmonotonic intensity tuning, whereas AI maps in the rat typically featured a single module for each feature (Cheung et al. 2001; Philibert et al. 2005; Phillips et al. 1994; Schreiner and Mendelson 1990; Sutter and Schreiner 1995). Injection of retrograde tracer into the middle-frequency representation of AI yielded a heavy deposit of label in the rostral region of the MGBv and a sparse, scattered distribution of label in the MGBm, as was previously reported in the rat (Kimura et al. 2003; Roger and Arnault 1989; Romanski and LeDoux 1993b; Ryugo and Killackey 1974; Winer et al. 1999).
The ventral auditory field
This is the first study that fully characterizes VAF in the rat. Earlier optical and electrophysiological mapping experiments conducted in our laboratory indicated the presence of an auditory region that shared a low-frequency boundary with AI, but was separated from the high-frequency CF representation in AI by a cortical region that was either unresponsive or poorly driven by tonal stimuli (e.g., Bao et al. 2001, Fig. 1B; Bao et al. 2003, Fig. 6; Kalatsky et al. 2005). The current data provide additional anatomical and physiological evidence to support the claim that this region is an independent field within the rat auditory cortex. We found that thalamic inputs into VAF originated from regions of the MGBv that were about 0.75 mm caudal from the AI projection zone. This result is not unexpected given the known topographic cortical projection pattern along the rostrocaudal extent of an isofrequency region of the MGv, but nonetheless confirms that frequency-matched regions of VAF and AI receive inputs from nonoverlapping thalamic neuron populations (Cetas et al. 2001; Huang and Winer 2000; Redies et al. 1989; Velenovsky et al. 2003). A comparable study in the guinea pig found that rostrodorsal regions of a single isofrequency contour in AI receives projections from the rostral pole of an isofrequency zone of MGv, whereas the caudoventral region of the same isofrequency contour received inputs from the caudal pole of the MGv (see Redies et al. 1989; Fig. 4). However, in light of findings from more recent, higher-resolution mapping of the guinea pig auditory cortex, these injections were likely made into an isofrequency contour that spanned AI and the immediately adjacent ventral field (Horikawa et al. 2001; Nishimura et al. 2007) in a fashion similar to that in our study. Based on this interpretation, the present results from VAF and AI in the rat are in perfect concordance with the thalamocortical connectivity patterns in the guinea pig.
Importantly, the thalamic inputs in VAF and AI could also be distinguished based on projections from other divisions of the medial geniculate complex. VAF was found to receive a comparatively minor but consistently present input from MGBd, in agreement with an elegant recent study that described a projection from the MGBd to a circumscribed region on the ventral margin of cortical area TE1 in the rat, which very likely corresponds to VAF in our study (Donishi et al. 2006). Finally, we found that VAF could be distinguished from AI based on neurophysiological markers. Whereas the boundary between AI and VAF could not be delineated according to a tonotopic reversal, the relationship between sound intensity and firing rate served as a clear physiological marker between the fields. Approximately 80% of multiunit responses in VAF exhibited low response thresholds and nonmonotonic intensity-response profiles, with best-level values commonly falling within a range of 20–40 dB SPL, in contrast to AI recording sites in which approximately 80% had best-level values in excess of 70 dB SPL and monotonic intensity-response profiles.
Despite the differences in the source of thalamic input, it seems unlikely that the marked differences in intensity-response characteristics between AI and VAF reflect a simple recapitulation of the intensity-tuning characteristics of the thalamic input. Although the excitatory synaptic inputs recorded in the middle cortical layers can exhibit intensity tuning (Wu et al. 2006), intracortical inhibition was previously shown to either enhance or entirely account for nonmonotonic intensity-response functions (Tan et al. 2007; Wu et al. 2006). Moreover, extracellular recordings conducted in the thalamus showed that similar percentages of monotonic and nonmonotonic intensity-response functions are found in the MGBv and MGBd (Edeline et al. 1999).
The suprarhinal auditory field
We have also characterized a second ventral auditory field, SRAF, which is situated about 2 mm ventral to AI in cortical area TE3. SRAF exhibited a well-organized low-to-high ventral-to-dorsal tonotopic gradient and multiunit responses with longer onset latencies and broader spectral bandwidths than AI. Like AI, SRAF neurons exhibited a spatial organization for features beyond CF. We invariably found a cluster of recording sites with long-onset latencies (often >20 ms) in the posteroventral quadrant of the field, a cluster of recording sites with narrow spectral bandwidths in the anterodorsal quadrant, and a swath of recordings sites with nonmonotonic intensity-response profiles in the posterior half of the map. The spatial contiguity analysis revealed that none of these clusters were likely to
