INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Olfactory stimuli elicit spatial patterns of activity across olfactory receptor neurons in the epithelium and in the olfactory bulb (Friedrich and Korsching 1997; Greer et al. 1981; Johnson and Leon 2000; Kent and Mozell 1992; Rubin and Katz 1999; Uchida et al. 2000). While responses of single receptor neurons to odorants have been directly studied, the strategy by which odor information is encoded in the activity of populations of receptor neurons entering olfactory bulb glomeruli is not as well understood.

The olfactory system, like other sensory systems, is capable of encoding stimulus quality independent of intensity. Psychophysical studies in humans and rodents have demonstrated that perceived odor quality can persist over a range of odorant concentration that can be more than a factor of 100 (Gross-Isserhoff and Lancet 1988; Youngentob et al. 1990). On the other hand, there are examples in humans where reported odorant quality changes with concentration (Arctander 1969; Gross-Isserhoff and Lancet 1988). A proposal about how concentration invariance of perceived odor quality might be achieved came from single-unit recordings from small populations of gustatory and olfactory receptor neurons that showed that the relative pattern of receptor neuron activity remained the same at different concentrations (Erickson 1963; Girardot and Derby 1990; O'Connell and Mozell 1969; Pfaffmann 1959).

In the present study, we sought to more fully characterize how populations of receptor neurons represent odor information at the level of input to olfactory bulb glomeruli, and to investigate the effect of concentration on these representations. We addressed these questions by imaging receptor neuron input to olfactory bulb glomeruli of the box turtle. The turtle olfactory bulb has a well-developed laminar structure similar to that of the mammalian olfactory bulb (Orrego 1961) and is an established model for studying olfactory processing (e.g., Berkowicz and Trombley 2000; Beuerman 1975; Greer et al. 1981; Jahr and Nicoll 1982; Lam et al. 2000; Mori et al. 1981; Tonosaki and Tucker 1982).

To selectively image olfactory receptor neuron activity in vivo, we loaded receptor neurons with a calcium-sensitive dye using a protocol similar to that first developed for zebrafish by Friedrich and Korsching (1997). After several days this dye appears in the olfactory neuron axon terminals in olfactory bulb glomeruli. The resulting signals from terminals in the glomeruli are a relatively fast and direct measure of action potential activity in these terminals (Friedrich and Korsching 1997; Wachowiak and Cohen 1999). We characterize the calcium signals measured from the nerve terminals as "input" to the olfactory bulb. Because the calcium indicator dye is only present in the receptor axon terminals, the neuronal source of a glomerular signal is directly determined. In contrast, even in glomerular regions, other imaging signals [intrinsic imaging, 2-deoxyglucose, and functional magnetic resonance imaging (fMRI)] represent energy utilization by the processes of all of the cell types that are active in each glomerulus.

We found that odorant representations at the input to the olfactory bulb were widely distributed, with even low odorant concentrations evoking receptor neuron input to a substantial fraction of all imaged glomeruli. However, while maps of input evoked by different odorants often involved many overlapping glomeruli, different odorants elicited distinct patterns of relative input to these glomeruli. We also found that the maps of relative input to glomeruli remained fairly constant over a concentration range of nearly three log units. Preliminary reports have appeared (Wachowiak et al. 2000a,b).

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Dye loading

Experiments were performed on three-toed box turtles (Terepene triunguis) collected from the southeastern United States. Olfactory receptor neurons were labeled with Calcium Green-1 dextran, 10 kD m.w. (Molecular Probes, Eugene, OR) by injecting 20 µl of a 2% dye solution dissolved in 0.1 M NaCl plus 0.5% Triton-X 100 into each naris. Injection was performed using a custom-shaped cannula (Cohen "Code Red" Stainer, MW Plasticworks, New Haven, CT). After staining, animals were held at room temperature for 5-18 days before recording. We observed no differences in the gross appearance of the olfactory epithelium (postmortem examination) or in the size and sensitivity of electroolfactogram (EOG) recordings between stained and unstained animals.

Optical recordings

Animals were lightly anesthetized with ketamine (30 mg/kg) and xylazine (3 mg/kg), partially immobilized with d-tubocurarine (1.5 mg/kg), and secured on a custom-made holder. The dorsal surface of both olfactory bulbs was exposed by craniotomy of the overlying skull and removal of the dura. Animals were kept chilled on ice during the surgical procedure, and local anesthetic (1% bipuvicaine) was applied around the craniotomy. Before imaging, the animal was allowed to return to room temperature. To reduce movement the olfactory bulbs were covered with 2% agarose in Ringer; a coverslip was placed on the agarose. All procedures were approved by the Yale University and Marine Biological Laboratory animal care committees.

For imaging, the preparation was illuminated with 480 ± 25 nm light using a 100-W tungsten/halogen bulb or a 150-W Xenon arc lamp (Opti-Quip, New York), and fluorescence emission above 530 nm was collected. In nine experiments, the olfactory bulb was imaged at ×4 magnification using a Macroscope (RedShirtImaging, LLC, Fairfield, CT) with a numerical aperture (n.a.) of 0.4 (Kleinfeld and Delaney 1996; Lam et al. 2000). In five experiments, an approximately 1 mm × 1 mm region of the bulb was imaged at ×15 using a Wild compound microscope objective (0.4 n.a.). Images were acquired and digitized with a 80 × 80 pixel charge-coupled device (CCD) camera (NeuroCCD; RedShirtImaging) at 100-200 Hz and time-binned to a 25-Hz frame rate before storing to disk. The light intensity reaching the CCD camera was approximately 10,000 photons/ms/pixel. In some experiments, the EOG was recorded using a Teflon-coated silver wire lead (0.003-in. OD) with exposed tip inserted into the naris.

Odorant presentation

All odorants were obtained from Sigma (>99% pure) and included: cineole, isoamyl acetate, 1-hexanal, 2-hexanone, and 2-butanone. We also screened three organic acids (propionic, butyric, and hexanoic acid), all of which failed to elicit detectable signals from the dorsal bulb even though they did elicit an EOG response. Odorants were presented as dilutions of saturated vapor. The molar concentration of saturated vapor for each odorant is as follows: cineole, 1 × 10⁴ M; isoamyl acetate, 3 × 10⁴ M; 1-hexanal, 5 × 10⁴ M; 2-hexanone, 5 × 10⁴ M; 2-butanone, 4 × 10³ M. We used dilutions of saturated vapor from 0.01 to 10%, so the odorant concentrations used in this study ranged from 1 × 10⁸ M (0.01% cineole) to 4 × 10⁴ M (10% butanone).

Fast-onset pulses of odor stimuli were delivered to the nose using a flow-dilution olfactometer described previously (Lam et al. 2000). Separate lines for each odorant were used to avoid cross-contamination. A vacuum line inserted into the pharyngeal opening of the nasal cavity controlled air and odorant access to the olfactory epithelium. Release of odorant from the olfactometer and suction through the nose were controlled by separate solenoids. The suction through the nose (250 ml/min) was initiated 2-8 s before a 2-s odor pulse was delivered. Nasal suction lasted 20 s to ensure clearance of odorant from the nose. We observed no apparent loss of response due to drying or other damage to the epithelium during the course of an experiment, which sometimes consisted of over 150 trials. We waited a minimum of 90 s between trials. Repeated presentations of the same odorant at this interstimulus interval evoked similar-amplitude signals (Fig. 2, E and F).

Data processing and analysis

Initial analyses indicated that variation in response amplitudes of repeated trials was primarily due to noise in the baseline signal (estimated from the noise in the recording prior to the stimulus) and not variability in the odorant-evoked response. The primary noise source was movement due to spontaneous respiration, causing large and slow noise events that appeared in many places in the image. For measurements with lower odorant concentrations (and smaller signals), we averaged from two to eight odor presentations to minimize the contribution of this noise, which had a frequency profile similar to the evoked signals. The individual trials were saved to disk, and trials with large breathing noise were discarded before averaging off-line. After averaging, data from each pixel were temporally filtered with a 1-Hz low-pass Gaussian filter and a 0.017-Hz high-pass digital RC filter. The signal from each pixel was divided by the resting fluorescence. Response amplitudes were measured for each pixel by taking the average signal within a 400- to 800-ms time window just preceding stimulus onset and subtracting it from the same temporal average centered around the peak of the response. The standard error of the response amplitude measurement varied with the spatial location of the measured pixel. Figures 2, E and F, 5D, and 6B show the range of error observed in a typical experiment. For display of pseudocolor maps and for measuring correlations between responses, pixels receiving light from outside the bulb and pixels overlying major blood vessels were omitted, and the resulting two-dimensional array of response amplitudes was spatially smoothed with a 2 × 2 pixel median filter. Data processing and display were performed with NeuroCCD software. Correlations (Spearman's rank correlation, ) were measured with programs written in IDL (RSI, Boulder, CO). Spearman's rank correlation was used because of the possibility of a nonnormal distribution of signal amplitudes. However, Pearson's linear correlations gave similar results.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Characteristics of the spatial patterns of afferent input to olfactory bulb glomeruli

Loading olfactory receptor neurons by perfusion of Calcium Green-1 dextran into the turtle nasal cavity resulted in specific labeling of receptor cell axons and their terminals in all olfactory bulb glomeruli (Fig. 1A). Labeling in the olfactory bulb was first evident after 4 days and continued to increase in intensity over the next several days. Labeling remained strong for at least 18 days (the longest incubation period tested). We did not detect trans-synaptic staining with Calcium Green-1 dextran, a result also reported earlier (Friedrich and Korsching 1997; Wachowiak and Cohen 1999). We observed no differences in the gross appearance of the olfactory epithelium (postmortem examination) or in the size and sensitivity of EOG recordings between stained and unstained animals.

View larger version (67K): [in this window] [in a new window]   Fig. 1. Staining and signals measured after Calcium Green-1 dextran labeling of turtle olfactory receptor neurons. A: sagittal olfactory bulb section showing pattern of dextran labeling. Left: labeling of glomeruli in all olfactory bulb regions. Right: higher magnification; labeling of olfactory nerve axons and their terminal branches within glomeruli. Absence of labeling in the external plexiform layer (epl) indicates that the dye is confined to olfactory receptor neurons. B: time course of fluorescence increase evoked by a 2-s presentation of 1.7% cineole (bottom trace), compared with the electroolfactogram response (EOG) recorded simultaneously from the olfactory epithelium (top trace). The EOG signal was band-pass filtered from 1 to 100 Hz before data acquisition. The optical signal is from 1 pixel and a single trial. Both traces were acquired at 100 Hz and then digitally filtered from 0.017 to 1 Hz. on, olfactory nerve; d, dorsal; r, rostral; onl, olfactory nerve layer; gl, glomerular layer; epl, external plexiform layer.

We imaged odorant-evoked receptor neuron input to dorsal glomeruli in vivo. Odorants typically evoked maximum fluorescence changes (F/F) ranging from 0.3 to 6% (Fig. 1B, bottom trace) beginning shortly after onset of the EOG response (Fig. 1B, top trace), a measure of transduction currents in the olfactory epithelium. The rise time of the fluorescence signal was 500-1,500 ms (Fig. 1B; Fig. 2, middle row; Fig. 4A, right panel). We observed only slight differences in the time course of the evoked signal for different odorants or in different olfactory bulb locations. Thus much of the information about the spatial distribution of the response can be captured by a single image representing the amplitude of the signal measured at each pixel. The absence of time course differences might, in part, be explained by the fact that the signal from each glomerulus is the population average from the terminals of hundreds or thousands of receptor cells. A second contributing factor is that the decay time of calcium signals in individual terminals is slow compared with that of membrane potential (Fig. 7A of Wachowiak and Cohen 1999), probably because this time course is, in part, determined by the time required for calcium sequestration and extrusion.

View larger version (51K): [in this window] [in a new window]   Fig. 2. Spatial patterns of odorant-evoked receptor input to the dorsal olfactory bulb imaged at ×4 magnification. Pseudocolor maps show fluorescence increases evoked by cineole (A), 2-hexanone (B), and isoamyl acetate (C). Odorant concentration is expressed as percent dilution of saturated vapor. Maps use separate normalized scaling for each odorant. D: gray-scale image of the resting fluorescence in this preparation. The structure of each odorant is shown on the bottom left of each pseudocolor map. Traces beneath each map show the time course and relative amplitudes of the evoked signals from 3 locations. Each trace is a spatial average from 4 pixels. Dashed squares in B and C indicate the region imaged at ×15 magnification in Fig. 3. E and F: stability of odorant response maps. Maps show signals evoked by 2 presentations of 0.05% cineole, 220 min apart (different preparation than in A-D). Each map is the average of 4 trials. The differences in the maps for any given pixel are generally <10%. The correlation between the 2 maps (Spearman's ) is 0.85.

We made maps of the amplitude of the odorant-evoked input to glomeruli on two spatial scales: in nine preparations from six animals, we focused on the large-scale spatial patterns of input across the entire dorsal surface of the bulb using ×4 magnification with a single pixel, ideal-case, spatial resolution of 50 µm. This resolution is not sufficient to resolve signals from individual glomeruli (average diameter ~50 µm). In five preparations from three animals, we imaged at a higher magnification (×15, 12.5 µm resolution) that was capable of resolving individual glomeruli.

At ×4 magnification, for the limited number of odorants we tested, the evoked patterns of receptor neuron input were spatially complex and odorant-specific and generally consisted of maximal signals in restricted regions (>80% of peak signal size; orange to red in pseudocolor maps) and smaller amplitude signals in much more widespread regions (20-80% of peak; blue to yellow). Figure 2, A-C, shows maps of responses evoked by the odorants cineole, 2-hexanone, and isoamyl acetate. The alkane derivatives 2-hexanone (a ketone) and isoamyl acetate (an ester) evoked maximal input to caudal-lateral glomeruli of the dorsal bulb, while cineole (a cyclic ether) evoked maximal input to rostral-medial glomeruli. However, the response maps evoked by all three odorants were spread across a very large portion of the bulb and, consequently, a large number of glomeruli. Interestingly, while the smaller amplitude signals were more widespread, they also showed spatial complexity and were not necessarily adjacent to the maximum-amplitude signals (see, for example, restricted blue and green regions in Fig. 2, A-C). Responses to repeated odorant presentations were relatively consistent from trial-to-trial, even when separated by intervals ranging from 30 to 220 min (Spearman's  = 0.76 ± 0.04; n = 9; e.g., Fig. 2, E and F). The result illustrated in Fig. 2, E and F, implies that phototoxic effects of the illuminating light are small.

The response maps evoked by 2-hexanone and isoamyl acetate in Fig. 2, B and C, show substantial overlap and appear similar at ×4 magnification. In all nine preparations imaged at this spatial scale, the odorants 2-hexanone, 2-butanone, isoamyl acetate, and 1-hexanal evoked overlapping patterns of maximal input to glomeruli in the caudal-lateral region of the dorsal bulb. To resolve differences in the response patterns evoked by these odorants, and to confirm that the optical signals could reflect input to individual glomeruli, we imaged response maps at higher magnification (×15). Figure 3 shows maps from the area indicated by the dashed boxes in Fig. 2, B and C. The images in Fig. 3, A-C, show multiple foci of peak signals with the approximate size of individual glomeruli. Comparison of the signals at the glomerulus indicated by the black arrows shows that both 2-hexanone and 2-butanone, but not isoamyl acetate, activated this glomerulus. Similarly, comparison of the signals at the glomerulus indicated by the white arrows shows that both 2-hexanone and isoamyl acetate, but not 2-butanone, activated this glomerulus. Thus each odorant tested evoked maximal levels of input to an overlapping but distinct set of glomeruli. The maps imaged at ×15 were also consistent across repeated trials of the same odorant at the same concentration ( = 0.81 ± 0.07; n = 3).

View larger version (50K): [in this window] [in a new window]   Fig. 3. Patterns of odorant-evoked receptor neuron input to glomeruli imaged at ×15 magnification. Pseudocolor maps showing fluorescence increases evoked by 2-hexanone (A), 2-butanone (B), and isoamyl acetate (C), imaged from the same preparation as in Fig. 2, A-D. Maps show the top 70% of the evoked signal. Note maximal levels of input to foci the approximate size of individual glomeruli. The black arrows indicate a glomerulus that was activated by 2-hexanone and 2-butanone but not isoamyl acetate. The white arrows indicate a glomerulus that was activated by 2-hexanone and isoamyl acetate but not 2-butanone.

In addition to the relatively few glomeruli receiving maximal levels of input, a striking feature of responses was the presence of widespread, moderate levels of input encompassing a much larger fraction of the dorsal bulb (Fig. 2). To confirm that these smaller signals did in fact represent input to many widespread glomeruli, as opposed to scattered light or light from out-of-focus areas of the preparation, we measured the spread of odorant-evoked signals into unlabeled or out-of-focus regions. Figure 4A shows a map of the response to 0.05% isoamyl acetate, which evoked maximal input to a region on the caudal-lateral edge of the olfactory bulb. Traces to the right of the pseudocolor map show the time course of the evoked fluorescence changes measured in the peak region (location 2), a more rostral region (location 3), the unlabeled accessory olfactory bulb (location 4), and a region over the bone lateral to the olfactory bulb (location 1). Locations 1, 3, and 4 are equidistant from the peak signal at location 2. The very small fluorescence changes seen in locations 1 and 4 must represent artifactual signal from scattered and/or out-of-focus light. However, these signals are much smaller than the signal recorded at location 3, indicating that the signal at location 3 primarily reflects afferent input to glomeruli in this region of the bulb. As a quantitative measure of the signal spread from scattered or out-of-focus light, we plotted the background fluorescence (F) and the signal size (F), along two tracks: one track progressing off of the lateral edge of the olfactory bulb (track a, Fig. 4B), and another passing through a small focus of large-amplitude signal but remaining within the olfactory bulb (track b). The plots for track a show that the evoked signal decreases to below 50% of peak levels in less than 3 pixels (approximately 150 µm), and to below 10% of peak levels in <5 pixels (250 µm). The plots for track b demonstrate that foci of evoked signals do not necessarily correspond to the regions of the most intense labeling. In a similar analysis of response maps imaged at ×15 magnification, the evoked signal decreased from maximal levels to <20% of maximum in 3-4 pixels (approximately 50 µM). These measurements, along with the observation that the moderate-amplitude signals were often distant from peak-amplitude signals, show that these smaller signals are not an artifact of scattered or out-of-focus light, but instead reflect moderate levels of receptor neuron input to glomeruli that are distributed across at least 30% and often 50-100% of the dorsal surface of the bulb (Figs. 2 and 5-7).

View larger version (54K): [in this window] [in a new window]   Fig. 4. The contribution of scattered and out-of-focus light to spatial maps of odorant-evoked input is small. A: pseudocolor map of fluorescence increase (top 70%) evoked by 0.05% isoamyl acetate, imaged at ×4, with traces showing the time course of the signal measured from 4 locations. Location 1 is lateral to the olfactory bulb, overlying a piece of skull that is out of the plane of focus, and shows almost no evoked signal. Location 2 is in the caudal-lateral bulb, in the region showing the maximum evoked signal. Location 3 is in the middle of the bulb and shows an evoked signal amplitude of slightly <30% of the maximum. Location 4 is in the accessory olfactory bulb, which is in the same plane of focus as locations 2 and 3 but is unlabeled by Calcium Green-1 dextran, and shows a slight signal. Locations 1, 3, and 4 are equidistant from location 2. Each trace is the spatial average of 6 pixels and was acquired at 100 Hz and digitally filtered from 0.017 to 1 Hz. The dashed rectangle in the pseudocolor map indicates the region shown in B. B: measurement of resting fluorescence and evoked signal along 2 tracks through the preparation. Track a passes through the region of maximum signal amplitude and off the lateral edge of the olfactory bulb. Track b passes through a small focal signal and remains on the stained part of the bulb. Right: plots of resting and evoked fluorescence along tracks a and b. For track a (Ba), note the sharp decrease in both resting and evoked fluorescence away from the edge of the bulb. For track b (Bb), the evoked signal decreases from maximal levels to 40% of maximum in about 3 pixels (150 µM) and does not parallel the resting fluorescence. Data points and error bars indicate the mean and standard error of measurements from 4 consecutive odorant presentations. The measurements indicate a minimal contribution of out-of-focus or scattered light to the appearance of odorant response maps. Unlike all other figures, the pseudocolor map and the traces have not been divided by the resting fluorescence, and so indicate absolute fluorescence changes.

While Fig. 3 shows signals that appear to originate from individual glomeruli, it will not generally be the case that the signal on each pixel results from activity in a single glomerulus. First, there is some blurring from scattered or out-of-focus light (Fig. 4), which causes spread in the image plane. Second, because the glomerular layer in the turtle is more than one glomerulus thick (Fig. 1A) and the depth of focus (about 500 µm for 0.4 n.a.) (Salzberg et al. 1977) is larger than the thickness of the glomerular layer, the signals on an individual pixel will include the activity of all the glomeruli along the z-axis. If only one glomerulus is active, then glomerular resolution will be obtained; otherwise the signals will represent the activity of several glomeruli.

Effect of concentration on odorant representations

To investigate the effect of stimulus intensity on odorant representations, we imaged responses to odorant dilutions ranging from 0.017 to 10% of saturated vapor. We first measured the effect of concentration on the absolute magnitude of afferent input to glomeruli. In all cases, increasing concentration increased the amplitude of the evoked signal; this included glomeruli that received both large and small levels of afferent input.

We also compared the response patterns after normalizing each input map to its own maximum [i.e., each map was scaled so that the maximum amplitude was set to 100% (red color)]. Figure 5A shows maps of responses evoked by isoamyl acetate, imaged from the left and right olfactory bulbs in the same preparation, at three concentrations from 0.017 to 10% dilution (a 590-fold concentration range). The normalized response maps are both bilaterally symmetric and approximately identical at all three concentrations. The amplitude of the maximum fluorescence change evoked by isoamyl acetate increased by approximately 450% over this concentration range. The amplitude of input to less strongly activated glomeruli also increased. Figure 5C shows the data from the response to 0.05% isoamyl acetate using an absolute scale that was fixed to be identical to that used for 0.017%, the lowest concentration in Fig. 5A. Even this modest three-fold concentration increase resulted in a dramatic increase in the area of the bulb receiving a given level of input. Thus evaluating glomerular activation on an absolute scale resulted in odorant representations that were dramatically different and, less specific, over increasing concentrations. Figure 5D shows a plot of signal size versus concentration from three locations on the right olfactory bulb (indicated in the bottom right panel of Fig. 5A). In each location there is a nearly linear increase in signal size with the log of concentration over the entire concentration range.

View larger version (61K): [in this window] [in a new window]   Fig. 5. Normalized and absolute maps of odorant-evoked input over a large concentration range and signal vs. concentration from 3 locations. A: pseudocolor maps of normalized fluorescence signals evoked by 3 concentrations (590-fold intensity range) of isoamyl acetate in left and right olfactory bulbs of the same animal. The normalized maps at all 3 concentrations are similar. Each concentration shows a composite of 2 separate recordings (left and right bulbs). The maps are approximately bilaterally symmetric at all 3 concentrations. The correlation coefficient (Spearman's ) between the maps evoked by 0.017 and 10% dilutions is  = 0.74 and  = 0.67 for the left and right sides, respectively. The numbers 1, 2, and 3 in the bottom right panel indicate the regions used for generating the concentration-response curve in D. B: gray-scale image of the resting fluorescence in this preparation. C: map of the input signal evoked by 0.05% isoamyl acetate using an absolute scale identical to that used for the response to 0.017% isoamyl acetate in the left panel in A. Using an absolute scale results in maps that are strikingly different at different concentrations. D: plots of signal size vs. concentration for 3 locations (shown in A) showing peak (1), moderate (2), and small (3) fluorescence signals evoked by isoamyl acetate at dilutions from 0.017 to 10%. Signal amplitudes were measured from single trials and spatial averages of 9 pixels. Each data point is the mean of 2-6 trials. Error bars indicate means ± SE. The concentration-response plots for all 3 regions are approximately linear and roughly parallel over the entire concentration range.

Figure 6A illustrates response maps from a similar concentration series carried out at high (×15) magnification. As at the lower resolution, the normalized input maps were similar over a concentration range of 200, with regions receiving different levels of input showing roughly parallel concentration-response functions (Fig. 6B). However, these higher-resolution maps also revealed a somewhat greater level of complexity than was observed at lower magnification. For example, some concentration-dependent changes in the maps reflected differences in the concentration dependence of input in different regions. Nonetheless, the overall pattern of relative input to glomeruli in Fig. 6 was similar across concentration; for the data in Fig. 6A, the mean correlation coefficient, , for all across-concentration comparisons was 0.71 ± 0.06 (mean ± SE, n = 6 comparisons), and 0.14 ± 0.04 (n = 9 comparisons) for all across-odorant comparisons in the same preparation.

View larger version (53K): [in this window] [in a new window]   Fig. 6. Effect of concentration on the maps of glomerular input imaged at ×15 magnification. A: pseudocolor maps of fluorescence signals evoked by 3 concentrations (200-fold intensity range) of cineole. Separate normalized scaling for each map. The normalized maps show slight changes with concentration. Numbers on the right panel indicate locations of concentration-response function measurements in B. B: concentration-response plots measured from the locations shown in A, as described for Fig. 5. Each data point is the mean of 2-8 trials.

The similarity in the normalized response maps applied to the smaller amplitude signals as well as the peak-amplitude signals (e.g., note locations of green/yellow areas for the maps in Figs. 6A and 7). With rare exceptions, we did not observe apparent saturation of the evoked signal even at the highest concentration tested (10% of saturated vapor). Similar results were observed in all preparations imaged at ×4 (n = 9) and ×15 magnification (n = 5), over a concentration range of 2-3 log units, and with as much as a 700% increase in the peak response amplitude. The continuous increase in evoked signal amplitude presumably reflects a combination of increases in individual receptor neuron action potential frequency as well as possible recruitment of additional receptor neurons with increasing odorant concentration.

The stability of these relative patterns of input was such that response maps evoked by different odorants, while overlapping, remained distinct at all concentrations tested. Figure 7 shows normalized low-magnification (×4) response maps evoked by isoamyl acetate and cineole, at concentrations ranging from 0.05 to 1.7% dilution (a 34-fold concentration range). Despite an increase in the absolute signal amplitude of approximately 300%, the normalized maps of receptor neuron input across the dorsal olfactory bulb are quite similar. For the data shown in Fig. 7, the mean correlation coefficient (Spearman's ) was 0.71 ± 0.10 (n = 6 pairwise comparisons) for different concentrations of the same odorant, and 0.17 ± 0.07 (n = 9 comparisons) for all concentrations of different odorants.

View larger version (86K): [in this window] [in a new window]   Fig. 7. Effect of concentration on normalized spatial maps of glomerular input evoked by 2 different odorants, cineole and isoamyl acetate; ×4 magnification. Pseudocolor maps of fluorescence signals (top 70%) evoked by 3 concentrations (34-fold intensity range) of isoamyl acetate and cineole in the same preparation. Separate normalized scaling for each map. The normalized maps show only small changes with concentration despite large increases in the absolute level of input. Scale bar in A, 500 µm.

We further quantified the effect of concentration on relative patterns of input by performing pairwise correlation analysis on every data set that included at least four odorant concentrations spanning a range of at least two log units. Correlation coefficients for maps imaged at ×4 and ×15 were similar, and so the data were pooled in the final analysis. The results are shown in Table 1. The correlation analysis supported the observation made in Fig. 7 that maps evoked by different odorants remained distinct at all concentrations. Correlations between odorant and air responses were near zero. The correlation coefficient for across-concentration comparisons was not statistically different from that for repeated trials of the same odorant at the same concentration; this result held true for comparisons of the two intermediate concentrations (a 6-fold concentration difference) as well as for comparisons of the highest and lowest concentrations (repeat trials vs. intermediates: P = 0.47; repeat trials vs. high/low: P = 0.06; unpaired t-tests). However, there were small but significant effects of concentration on the response maps. First, at higher concentrations, the maps often appeared less sharp than at low concentration (Figs. 5A and 6A, but not Fig. 7). Second, Table 1 shows that response maps evoked by the highest and lowest odorant concentrations (100- to 600-fold concentration difference) were slightly but significantly less correlated than those evoked by the intermediate concentrations (6-fold concentration difference; P < 0.001; paired t-test), indicating that increasing concentration did cause small changes in the relative patterns of glomerular input. These differences were due in part to a "loss" of the smallest-amplitude signals into the noise at the lowest concentrations, but also reflected increases in the size of the moderate-amplitude signals relative to that of the maximal signals. This effect can be seen in the pseudocolor maps as increases in the number of blue and green pixels (Figs. 5-7).

At the lower spatial resolution we rarely observed recruitment of input to new regions with increasing odorant concentration. Instead, with only a few exceptions, the locations with small-amplitude signals at low concentrations also showed small-amplitude signals at higher concentrations. In cases where evoked signals became detectable only at higher concentrations, the amplitudes of these signals were always the smallest-amplitude signals in the response maps. This result suggests that their appearance represented an increase in afferent input to a level above the detection threshold of our recordings, rather than an actual recruitment of input to new glomeruli. Indeed, improving the signal-to-noise ratio by extensive multi-pixel or multi-trial averaging often revealed the presence of a signal in these areas even at the lowest concentrations tested (data not shown).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We have imaged spatial patterns of receptor neuron input to glomeruli in the turtle olfactory bulb. We followed the method for selectively mapping receptor input first developed by Friedrich and Korsching (1997), and, like their study in the zebrafish, we find that different odorants are represented by input to overlapping but distinct sets of glomeruli. We report two new findings. First, even low odorant concentrations (as low as 2 × 10⁸ M for some odorants) can activate receptor neurons projecting to a substantial fraction of all imaged glomeruli. Second, the relative pattern of input across these glomeruli changes only slightly over a concentration range of up to 3 log units.

Maps of odorant-evoked input involve many glomeruli

Loading turtle olfactory receptor neurons with Calcium Green-1 dextran resulted in labeling of virtually all olfactory bulb glomeruli (Fig. 1A), and imaging the dorsal surface of the bulb allowed us to simultaneously image odorant-evoked input to approximately 30% of the total glomerular population. Even at the lowest concentrations tested, single odorants evoked input to much of the imaged area (Figs. 2, 5, and 7). Because our measurements indicated that the spread of the optical signal from scattered and out-of-focus light was limited to only a few pixels (Fig. 4), we conclude that these widespread signals reflect receptor neuron input to many glomeruli across much of the dorsal bulb. This result is consistent with previous reports that a single odorant excites a substantial fraction of tested receptor neurons (Duchamp-Viret et al. 2000; Firestein et al. 1993; Gesteland et al. 1963). Indeed, a characteristic feature of olfactory receptor neurons is their responsiveness to multiple odorants, and the ability of a single odorant to activate multiple, molecularly distinct, olfactory receptors (Duchamp-Viret et al. 1999; Malnic et al. 1999).

Nonetheless, odorant-evoked patterns of receptor neuron input showed a clear odorant specificity and spatial organization (Figs. 2, 3, and 7). On a large spatial scale, odorants evoked maximal input to relatively restricted domains of multiple glomeruli and moderate, but still spatially organized, input to more widespread areas. These domains had a regional chemotopy, with ketones, aldehydes, and esters maximally activating glomeruli in the caudal-lateral bulb, and cineole (a cyclic ether) maximally activating rostral glomeruli. At a higher magnification, the different odorants we tested maximally activated specific sets of glomeruli (Fig. 3).

The fact that many glomeruli are activated and that regions with small-amplitude signals also have complex maps (e.g., Fig. 2A) is also consistent with studies in rats showing that the ability to discriminate odorants can persist even after ablation of a large fraction of the olfactory bulb (Lu and Slotnick 1998). Even a small portion of a complex map may contain enough information to identify the odorant that elicited that map.

Concentration dependence of odorant representations

Our recordings of the input to the olfactory bulb showed a continuous increase in the absolute amount of receptor neuron input to glomeruli with increasing odorant concentration (e.g., Fig. 5, A, C, and D). However, the maps of relative (normalized) input changed only slightly over a large concentration range (Figs. 5-7). Response maps evoked by different odorant concentrations were statistically just as well correlated as maps evoked by repeated presentations of the same odorant at the same concentration (Table 1). This observation held true for both the large-scale regional maps and small-scale glomerular maps of receptor neuron input, and for signals of maximal as well as moderate amplitude.

One concentration-dependent change in the normalized maps was a slight (<10% per log unit of concentration) increase in the relative amplitude of the smaller signals. In addition, odorant representations often become less sharply defined at higher odorant concentrations (Figs. 5 and 6 but not Fig. 7). These effects on the relative maps might arise from a partial saturation of the most sensitive glomeruli.

Single vertebrate olfactory receptor neurons are reported to have a dynamic range of only 1-2 log units (Firestein et al. 1993; Reisert and Matthews 2001; Trotier 1994), yet our recordings often showed continuous increases in receptor neuron input spanning nearly three log units (Figs. 5 and 6). One explanation for these results is that individual receptor neurons expressing the same receptor protein (and converging onto the same glomerulus) may have identical odorant response profiles but differing thresholds, thus increasing the dynamic range of receptor neuron input to glomeruli and increasing the stability of odorant representations across concentration (Cleland and Linster 1999; Firestein et al. 1993). Such an effect would only be evident at the population level.

Maps of receptor neuron input to the zebrafish and mouse olfactory bulbs

Friedrich and Korsching (1997) were the first to characterize maps of input to the olfactory bulb in experiments they carried out in zebrafish. Consistent with our results in the turtle, they reported increases in absolute levels of receptor neuron input with concentration, with individual glomeruli showing a dynamic range of two or more log units. Individual glomeruli could have different thresholds and different dynamic ranges, similar to the results observed in the turtle at high magnification (Fig. 6B). However, Friedrich and Korsching (1997) reported concentration-dependent changes in the patterns of input to zebrafish glomeruli. These changes were significant enough to suggest that, in contrast to our results, the patterns of receptor neuron input do not encode odorant identity independent of concentration.

We have recently carried out similar measurements in C57/Bl6 mice (Wachowiak and Cohen 2001). In contrast with the results from the turtle, patterns of receptor neuron input to the mouse olfactory bulb are highly concentration dependent. At low concentrations (<0.3% of saturated vapor), odorants evoke input to only a few glomeruli, while at slightly higher concentrations more than one-half of the imaged glomeruli are activated. The reasons for the difference in results between turtle and mouse is unclear. One hypothesis is that the concentration-dependent recruitment of input to new glomeruli in the mouse, as compared with the turtle, might reflect differences in the specificity of receptor neurons to a range of odorants, with more broadly tuned receptor neurons in the turtle showing less recruitment and thereby more concentration invariance. It is also unclear whether the apparent differences in the concentration dependence of odorant representations implies that these animals use different strategies for processing olfactory information. One possibility is that the mouse also forms a concentration-invariant map but at a later stage of olfactory processing.

The significant overlap in maps of glomeruli activated by different odorants, along with the stability of these maps across concentration, is consistent with an "across-fiber" odor coding strategy first suggested by Pfaffmann (1959) and Erickson (1963). Such a coding scheme, which takes into account different relative amounts of activation of a population of broadly tuned receptor neurons, has been demonstrated to encode stimulus quality independent of concentration in recordings from populations of gustatory as well as olfactory receptor neurons (Girardot and Derby 1990; O'Connell and Mozell 1969). Our study integrates this across-fiber coding strategy with the spatial organization of receptor cell input to the olfactory bulb and suggests the hypothesis that, at least in the turtle, concentration-invariant odorant recognition might be achieved by comparing relative levels of receptor neuron input to olfactory bulb glomeruli.

While our results in the turtle are consistent with a relative, across-fiber, coding hypothesis, the same is not true for the results in the zebrafish (Friedrich and Korsching 1997) and our results in the mouse. In these preparations the input to the olfactory bulb is consistent with a more general combinatorial coding scheme where an additional dimension, concentration, is also encoded by different combinations of receptor neuron input to glomeruli (e.g., Ma and Shepherd 2000; Malnic et al. 1999).

Although there are differences in the concentration dependence of input to the zebrafish, turtle, and mouse bulbs, other features of odorant representations are similar in these animals. First, at all but the lowest concentrations, response maps involve input to many glomeruli, with a high degree of overlap in the identity of glomeruli that respond to different odorants. In the turtle and in the mouse, odorant representations often involve a large fraction of all imaged glomeruli. Second, odorant representations in these animals show a broad chemotopic organization. However, this chemotopy is apparent on a regional rather than a glomerular scale, and, as already mentioned, can be highly overlapping for different odorants. Thus the role of chemotopy in discriminating odorants, especially structurally similar odorants, may be small. Finally, despite their highly distributed and overlapping nature, the odorant representations in the zebrafish, turtle, and mouse are apparently odorant-specific and remain so over a wide range of concentrations. Thus while these animals may use different processing strategies to recognize odorant quality independent of intensity, it seems likely that a common, fundamental step in this processing involves the comparison of relative amounts of receptor neuron input across large numbers of glomeruli.

Comparison with other types of glomerular maps

Previous studies using other, less specific, imaging methods have reported conflicting results concerning the effect of concentration on odorant representations. In the insect antennal lobe, imaging glomerular activity with bath-applied calcium-sensitive dyes, which presumably stain both afferent terminals and postsynaptic neurons, resulted in maps that changed qualitatively over an odorant concentration range of a factor of 100 (Galizia et al. 2000; Sachse et al. 1999), a result different from our result in the turtle. In contrast, maps of energy utilization in rodents based on fMRI recordings (Xu et al. 2000) and 2-deoxyglucose imaging (Johnson and Leon 2000; Schaefer et al. 2000) have reported concentration-invariant patterns of glomerular activity for some odorants. However, other 2-deoxyglucose imaging studies have reported concentration-dependent increases in the number of activated glomeruli (Johnson and Leon 2000; Stewart et al. 1979), as has imaging of intrinsic optical signals (Rubin and Katz 1999). Resolving these different findings is difficult because the signal type, the signal time course, and the species studied are different. Additional experiments directly comparing these different signals in the same animal are needed.

Comparison with spatial and temporal patterns of postsynaptic activity in the vertebrate olfactory bulb

Earlier studies have used voltage-sensitive dyes to image patterns of odorant-evoked activity in the salamander and turtle olfactory bulbs (Cinelli et al. 1995; Kauer et al. 1987; Lam et al. 2000). In these experiments, the dye nonselectively stains all olfactory bulb neurons, and the dye signal is thought to primarily reflect postsynaptic neuronal activity (Lam et al. 2000). In contrast to the signals reported here, odorants evoked spatial patterns of voltage-sensitive dye signals that were very widespread and not localized to individual glomeruli (Cinelli et al. 1995; Kauer et al. 1987; Lam et al. 2000). In fact, in the turtle, isoamyl acetate and cineole elicit voltage-sensitive dye signals with apparently identical spatial (and temporal) patterns (Y. W. Lam, L. B. Cohen, and M. Zochowski, unpublished observations), while these odorants provide maximal receptor neuron input to different olfactory bulb regions (present study). The widespread nature of the voltage-sensitive dye signals, measured in both turtle and salamander, may reflect the extensive lateral branching of mitral cell secondary dendrites, and further supports the hypothesis that processing of olfactory input involves distributed activity across much, if not all, of the olfactory bulb. The olfactory bulb oscillations recorded with both electrodes and voltage-sensitive dyes (for review, see Laurent et al. 2001) suggest that temporal features of olfactory bulb activity may play an important role at this postsynaptic level. The relationship between the spatially organized, temporally simple patterns of receptor neuron input and the spatially widespread, temporally complex patterns of postsynaptic activity, remains to be determined.

How are maps of afferent input to glomeruli transformed in the olfactory bulb?

Postsynaptic processing might sharpen odorant representations (Yokoi et al. 1995). This hypothesis could be tested by comparing input maps with output maps of mitral/tufted cell activity. Odorant representations may also be modulated presynaptically: several recent studies have demonstrated that inhibitory interneurons can mediate presynaptic inhibition of olfactory receptor axon terminals (Aroniadou-Anderjaska et al. 2000; Hsia et al. 1999; Wachowiak and Cohen 1999). Thus the maps of receptor cell input that we measured may already reflect some degree of processing by presynaptic inhibition. A testable hypothesis is that presynaptic inhibition plays a role in maintaining the concentration-invariant representations of olfactory input that we have observed in the present study. Such a function could be analogous to the "gain control" mechanism proposed for presynaptic inhibition of mechanosensory afferents in insects (Burrows and Matheson 1994), allowing for sensory information to be reliably encoded over a wide dynamic range.