Specializations of a pheromonal glomerulus in the Drosophila olfactory system

Fly lines.

Flies expressing UAS-G-CaMP1.6 (Reiff et al. 2005) on the second chromosome were crossed with GH146-Gal4 (for measurements on PNs) or OR83b-Gal4 [for measurements on olfactory receptor neurons (ORNs)]. Flies were homozygous for both Gal4 and UAS constructs. For experiments comparing pre- vs. postsynaptic activity (see Fig. 3C and Supplemental Materials, Supplemental Fig. S5; Supplemental Material for this article is available online at the J Neurophysiol website), flies expressed both GH146 and OR83b-Gal4. For measurement of presynaptic responses to single-pulse stimuli (Supplemental Fig. S10, A–C), flies expressing UAS-G-CaMP3 were used (Tian et al. 2009).

Recordings.

One- to three-day-old flies were briefly anesthetized with CO₂ and decapitated. Imaging was done on an isolated brain preparation (Wang et al. 2003). Antennae were cut with micro-scissors (FST, Foster City, CA), and the cuticle was pulled apart with a pair of forceps. Dissection was done in saline with 0.1 mM Ca²⁺ (adapted from Silbering and Galizia 2007). The brain was transferred to a glass slide with 0.5 ml of saline with 2 mM Ca²⁺ (Silbering and Galizia 2007). The anterior surface of the brain faced upward, and the brain was lightly pressed with forceps so that it adhered to the glass. Glass pipettes were pulled with a micropuller (Sutter, Novato, CA) and fire polished to match the diameter of the antennal or maxillary nerve. Pipettes were secured onto electrode holders that were attached to micromanipulators. Electrical stimuli were delivered using a Master-8 stimulator (AMPI, Jerusalem, Israel), triggered by pClamp software (Molecular Devices, Sunnyvale, CA). Acquisition was triggered on a Zeiss LIVE confocal microscope, with a 488-nm diode laser, a CCD camera, a 495-nm long-pass filter, and a ×60 objective of 0.95 NA (Zeiss, Thornwood, NY). Frame rate was 20 Hz.

Two nerves (either 2 antennal or 1 antennal and 1 maxillary nerve) were suctioned into one electrode each. Care was taken to have both antennal lobes visible at a similar cross-sectional depth; suction electrodes were slightly adjusted in the z-direction if needed to achieve this. The focal plane was roughly 12 μm beneath the anterior surface of the antennal lobe; the focal depth was roughly 5 μm (half-maximum width of the point-spread function). Stimuli were 0.1–0.3 ms, 10-V electrical pulses, sufficient to elicit excitatory postsynaptic potentials (Olsen and Wilson 2008) in PNs. Intertrial interval was 30 s; each trial consisted of a 1.5-s acquisition, stimuli being presented at 0.5 s. Pulse trains were 100 Hz for 300 ms. This duration was sufficient to evoke steady-state responses; since the bulk of the response occurred in the first 50 ms (see Fig. 7D and Supplemental Fig. S7), our results generalize to other stimulus durations. For stimulation of ventromedial glomeruli (see Fig. 6C, ii), ventromedial neuropil was suctioned into electrode and stimulated with a single 10 V, 0.4-ms pulse. This resulted in selective, bilaterally symmetric activation of ventromedial glomeruli, due to the activation of bilaterally projecting ORNs; results for ipsilateral and contralateral antennal lobes were similar and were pooled for analysis.

Pharmacology.

Excitation, mediated by nicotinic acetylcholine receptors, was blocked using 40 μM mecamylamine (Sigma, St. Louis, MO) (Kazama and Wilson 2008). Stock solution was 10 mM in 10% DMSO. GABA_A was blocked using 20 μM picrotoxin (Sigma); GABA_B was blocked using 50 μM CGP54626 (Tocris, MO, Ellisville) (Wilson and Laurent 2005). Drugs were added to preparation with a micropipette, taking care not to move the preparation. There was a 5-min wait before acquisition was resumed.

To isolate pre- and postsynaptic activity within a single preparation (see Fig. 3C and Supplemental Fig. S5), brains from UAS-G-CaMP1.6,GH146-Gal4;OR83b-Gal4 flies were stimulated with a single pulse to evoke a postsynaptic response. Under normal conditions, this protocol minimally evokes a change in presynaptic fluorescence (unpublished results). This was proceeded by a blockade of excitation using mecamylamine. This prevented activation of PNs as well as inhibitory circuitry, disinhibiting the presynaptic response. A 30-pulse, 100-Hz stimulus was then used to evoke presynaptic activity.

Map generation.

All visualization and analysis was performed in Matlab (r2006b; Mathworks, Natick, MA). LSM image stacks were imported into Matlab using DIPimage. To make maps of evoked activity, the average of 10 or 20 frames (depending on protocol) poststimulus was subtracted from the average 10 frames prestimulus, and divided by prestimulus average (Fig. 1D). To remove the effect of dark-state conversion of G-CaMP1.6 that occurs during illumination, the resultant map was subtracted from a control map generated for a no-stimulus condition. To correct for drift of preparation between trials, each antennal lobe was aligned to the resting fluorescence image from the initial trial using DIPimage. The stability of all proceeding analyses is discussed in Supplemental Fig. S1. All ΔF/F images shown in figures were further processed as such: 1) the region surrounding glomeruli is masked to remove low-fluorescence regions, since these suffer from low signal-to-noise ratio, and 2) image is smoothed with a Gaussian filter, σ = 0.75–1 pixels.

Analysis.

For region of interest analysis, glomeruli were manually selected. They were identifiable in ΔF/F images as contiguous, round regions of relatively uniform intensity. DA1 was identifiable by its distinct morphology and location in the dorsolateral region. To generate average maps of activity across preps, each antennal lobe was morphed into a 300-point circle consisting of 10 radial and 30 angular coordinates (Supplemental Fig. S2). Each pixel was binned according to its distance from the boundary of the antennal lobe and its angle from the centroid of the antennal lobe. Bins were averaged and plotted in polar coordinates and interpolated to generate a smoothed, circular map (using PolarToIm.m, posted on Matlab Central by P. Manandhar). The same procedure was carried out for DA1 average maps, each pixel binned according to its distance from the manually selected boundary of DA1 and its angle from the centroid of DA1. The statistical properties of each average map are described in Supplemental Fig. S3; statistical values for all comparisons are given in Table 1.

To calculate the dispersion of activity within a glomerulus, we calculated the mean distance of its pixels from its center of response, weighted by the per-pixel ΔF/F. We used this to calculate “ΔExtent,” the log ratio of ipsilateral dispersion to contralateral dispersion. To generate Fig. 2D, ΔExtent was calculated and assigned to the entire region of interest of each glomerulus. The resulting maps were then standardized (Supplemental Fig. S2) and combined to generate an average map.

To define DA1 boundaries within average maps (e.g., Fig. 2D), DA1 was manually identified in single preparations and its region of interest was assigned a value of 1. The resulting maps were standardized (Supplemental Fig. S2) and averaged, and the region whose intensity exceeded 50% of maximum was demarcated as DA1 within average maps. Regions activated by local stimulation (see Fig. 6C) were delineated by generating average maps of responses to local stimulation and thresholding the resultant map at 50% of maximum response. To visualize the medial region activated by maxillary nerve stimulation (see Fig. 6C, i), a threshold of 30% was needed.

To enhance color contrast (see Fig. 6B and Supplemental Fig. S5B, left), each pixel was assigned a z-score (#SD from the mean) based on the log ratio of intensities in the two channels. For each pixel, this ratio was exaggerated using the hyperbolic tangent function and was used to generate new values for the two channels, keeping the cumulative intensity across its two channels constant (Supplemental Fig. S4). The enhancement was done for visualization purposes alone and did not affect any quantifications.

To calculate the inhibition of the response to the later stimulus in a stimulus pair (see Fig. 6E), 1) stimuli were presented at different interpulse intervals; 2) the fluorescence traces of the resulting responses were fit by applying the Matlab function nlinfit on the following formula:

where

and A₁ and A₂ are the response amplitudes, g(t) is an impulse response function, T is the time constant of decay, D isolates a slight dip of the response below baseline (as seen in Supplemental Fig. S7A), and IPI is the interpulse interval for the stimulus; and 3) the normalized response to the second pulse was determined as follows:

where A_F and A_S are the amplitudes of the responses to the first or second stimulus presented in isolation. This approach was necessary (instead of R_S = A₂/A_S) because A₁ and A₂ could not be separated at small interpulse intervals (≤50 ms).

To plot PN response in time (see Fig. 7 and Supplemental Fig. S7), the time course of the average fluorescence in a region of interest was bleach corrected by subtracting the time course of fluorescence in a no-stimulus control movie. The resultant trace was normalized and deconvolved with an exponential with a time constant τ = 363 ms, the time constant of fluorescence decay calculated in PNs in response to a single pulse. This corrected for the slow decay intrinsic to the response of G-CaMP1.6.

To test significance, the two-tailed, two-sample t-test was used; the one-sample (paired) t-test was used when the comparison utilized pairs of measures taken from the same sample (see Figs. 4D, 5B, 6D, and 7C).

DA1 encodes spatial information.

The bilateral comparison of olfactory stimuli aids source localization in many species (Porter et al. 2007; Rajan et al. 2006), including fruit flies (Borst and Heisenberg 1982; Duistermars et al. 2009). Most glomeruli in the fruit fly antennal lobe are innervated by ORNs originating from both the ipsilateral and contralateral antennae (Stocker et al. 1990). We investigated how these two ORN populations activate DA1 and other glomeruli by visualizing the response to electrical stimulation of either the left or the right antennal nerve (Fig. 2A). We imaged either presynaptic activity in ORN terminal axons (Fig. 2B) or postsynaptic activity in PN dendrites (Fig. 2C). The responses of ventrolateral glomeruli, in ORNs as well as in PNs, were largely ipsilateral (Fig. 2, B and C), matching their innervation by ORNs from the two antennae (Supplemental Fig. S5D). The rest of the glomeruli responded bilaterally: imaging the ORN terminal axons of these glomeruli revealed interlaced regions that responded to either ipsilateral or contralateral stimulation (Fig. 2B).

Unlike the segregation of ipsilateral and contralateral activity in ORN inputs, the PNs of each bilaterally responsive glomerulus exhibited virtually identical spatial patterns of activity in response to ipsilateral and contralateral stimulation (Fig. 2C, white regions, and Supplemental Fig. S6). However, this was not the case for the PN population innervating DA1. Here, ipsilateral stimulation resulted in the focal activation of a subregion within DA1, whereas contralateral stimulation triggered a more spatially distributed response (Fig. 2, C–F). The appearance of distinct spatial patterns within DA1 is not simply a by-product of DA1's large size, as this property is also apparent in DL3 (arrowheads in Fig. 2, C and D), a smaller, cVA-responsive pheromonal glomerulus (van der Goes van Naters and Carlson 2007) dorsal to DA1.

The relatively focal contralateral activation found within DA1 PNs mirrors the distribution of ORN inputs (Supplemental Fig. S5). This difference in the spatial “peakedness” of the glomerular response to ipsilateral and contralateral stimuli is quantified by the ΔExtent measure (see materials and methods). For other glomeruli, the ΔExtent was much smaller in ORNs and roughly zero in PNs (Supplemental Fig. S5E). We aligned 16 DA1s mapped to polar coordinates and plotted the average relative response of each subregion (30 angular × 10 radial bins; see Analysis) to ipsilateral and contralateral stimuli. The ventromedial region had a relatively large response to ipsilateral stimulation, while a more contralaterally responsive region was centered on a dorsolateral locus (Fig. 2F).

Thus the activation of ORNs originating from the left and right antennae elicits unique spatial patterns of activity in DA1 PNs, a feature not seen in the PNs of generalist glomeruli.

DA1's spatial response is sexually dimorphic.

Since DA1 activation triggers sex-specific behaviors in Drosophila (Kurtovic et al. 2007), we looked for differences in the side-specific responses of male and female DA1s. Unlike male DA1, which responded more strongly to ipsilateral stimulation (Fig. 3A), female DA1 showed similar magnitudes of response to ipsilateral and contralateral stimulation (Fig. 3B) and in some cases an even stronger response to contralateral stimulation (e.g., left female DA1 in Fig. 2C). This difference could not be explained by the relative density of presynaptic innervation, since both male and female DA1 ORN responses were largely ipsilateral (Fig. 3C).

The difference between males and females in ipsilateral preference of DA1 PNs was seen across the range of stimulus durations tested. Longer trains resulted in a greater ipsilateral preference in both males and females in DA1 as well as in other glomeruli (Fig. 3D). This was the result of the ipsilateral response having a sustained component to prolonged stimuli that was absent in contralateral responses (Supplemental Fig. S7). The sexual dimorphism in ipsilateral bias seen in DA1 was small or undetectable in other glomeruli (Fig. 3D and Supplemental Fig. S8).

DA1's responsivity rapidly decreases upon activation.

The ability of pheromonal PNs to report rapid fluctuations in odor concentration is critical for localizing odor sources (Lei et al. 2009). We therefore investigated the specialization in the response dynamics of DA1 PNs relative to the PNs of other glomeruli. The activation of ORNs triggers excitatory and inhibitory interactions within and among glomeruli that together transform the output of PNs (Olsen et al. 2007; Olsen and Wilson 2008). We developed a protocol to measure the net impact of circuit interactions on PN excitability in the moments following initial activation. We stimulated one antennal nerve to activate excitatory and inhibitory neurons across the antennal lobe and then stimulated the contralateral nerve with varying delays to map time-sensitive variations in the ability of the ORNs to excite the PNs in wake of the prior stimulation (Supplemental Fig. S9 and Movie S1). Introducing a delay in the second pulse attenuated its contribution to the total evoked response, due to the predominance of inhibition evoked by the first pulse (see below). We visualized this attenuation within different glomeruli by superimposing the response to simultaneous inputs and 25-ms-staggered inputs (Fig. 4, B and C). Compared with other glomeruli, DA1 showed a decrease in its response to stimuli delayed by as little as 5 ms (Fig. 4D); this decrease was more than four times larger for a delay of 25 ms. In contrast, simultaneous inputs more linearly summed in DA1 than in other glomeruli (Fig. 4E). Plotting the per-pixel response to simultaneous vs. staggered inputs showed linear distributions of pixels belonging to DA1 or neighboring glomeruli (Fig. 4F). The linear and bimodal clustering of these pixels indicates that the decrease is uniform and confined within DA1.

GABA_A mediates rapid inhibition in DA1.

In moths, the ability to track pheromone dynamics relies on GABAergic inhibition (Lei et al. 2009), an interaction that is pronounced in the fruit fly antennal lobe (Olsen and Wilson 2008; Root et al. 2008). We pharmacologically assessed how GABA shapes the spatiotemporal response properties of DA1 PNs. Block of either GABA_A or GABA_B receptors increased the response of DA1 PNs (Fig. 5D). Normalizing these responses allowed us to assess how each pathway shaped the temporal dynamics of input integration in DA1. Block of GABA_A receptors eliminated the relatively rapid decrease in DA1's responsivity (Fig. 5, A and B), unmasking an excitatory interaction that peaked at the 10 ms interpulse interval. Interestingly, removing fast inhibition induced a complementary increase in late-phase inhibition (Fig. 5A and Supplemental Fig. S10), suggesting a greater activation of inhibitory local neurons. Block of GABA_B receptors decreased the late phase of the delay-dependent response attenuation. Only fast GABA_A inhibition shaped the spatial extent of DA1 activation, as was apparent in the distribution of activity along the proximal-distal axis (relative to PN output, see Fig. 5C) of DA1 PN dendrites. Whereas the activity profile scaled linearly upon blocking GABA_B receptors, GABA_A receptor block preferentially disinhibited the proximal dendrites of DA1 PNs (Fig. 5D). The results show that fast inhibition is evoked strongly and rapidly enough to restrict the spread of activity, as well as the influence of slow inhibition, within the PNs of DA1.

Weak interglomerular inhibition in DA1.

The greater inhibition that we find in DA1 could arise from intraglomerular or interglomerular interactions, both of which would be triggered by our antennal nerve stimulation protocol (Fig. 4A). To determine whether interglomerular inhibition is sufficient to elicit DA1's relatively strong and rapid inhibition, we delivered the initial stimulus to the maxillary nerve instead of the antennal nerve (Fig. 6A). Since the two nerves activate different sets of glomeruli (Rajashekar and Shamprasad 2004), any decrease in DA1's response should result exclusively from interglomerular inhibition. Comparing the response to simultaneous stimulation of the two nerves to the response when the maxillary nerve was stimulated first revealed that a dorsoventral strip of glomeruli down the middle of the antennal lobe experiences greater inhibition (Fig. 6B, magenta region) than the medial and lateral glomeruli, including DA1 (Fig. 6, B and C, i). Blockade of GABA_A receptors further increased the lateral inhibition triggered by maxillary nerve stimulation (consistent with the effect seen in Fig. 5A), but once again, DA1 was among the least inhibited (Supplemental Fig. S10D).

To make sure that this pattern of lateral inhibition does not result specifically from the activation of maxillary glomeruli, we measured the inhibition of the response to antennal nerve stimulation resulting from 100-s prior direct stimulation of ventromedial glomeruli. Maxillary nerve stimulation (Fig. 6C, i) and ventromedial neuropil stimulation (Fig. 6C, ii) evoked similar distributions of lateral inhibition: in both cases the greatest inhibition was in dorsal glomeruli, with weaker inhibition in DA1. Furthermore, the onset of inhibition following maxillary nerve stimulation (Fig. 6D) was slower than that observed following antennal nerve stimulation (Fig. 4D). Thus interglomerular inhibition failed to account for the rapid-onset inhibition that differentiates DA1 from other glomeruli (Fig. 4E). This suggests that intraglomerular interactions contribute to DA1's distinctive inhibitory response following antennal nerve stimulation.

DA1 discriminates direction of stimulus onset.

One cue that enables odor localization in several species is the sequence in which an odor plume comes into contact with spatially separated sensors (von Bekesy 1964; Gardiner 2010; Rajan et al. 2006). Responding to this cue requires that an animal make a persistent decision that far outlasts a potentially transient difference in sensor activation. We tested the ability of DA1 to report such brief temporal asymmetries in the activation of inputs originating from the left and right antennae. We stimulated each nerve with a 100-Hz train for 0.3 s (Fig. 7A), a frequency and duration within the range of ORN responses to naturally occurring odor plume filaments (van der Goes van Naters and Carlson 2007; Justus et al. 2002), and introduced varying delays in the onset of right nerve stimulation. With a relative delay of only 25 ms, less than one-tenth of the total stimulus duration, the pattern of activity in left and right DA1s resembled the pattern evoked by stimulating the left nerve alone (Fig. 7B). We quantified this with the spatial correlation between the response to paired stimulation and that of stimulating either nerve alone (Fig. 7C). For delays of 25 ms or more, the later stimulus contributed minimally to the final spatial pattern of activity. This can be seen in the time course of activity following dual-nerve stimulation; activation of DA1 by the left nerve strongly and rapidly inhibits the transient evoked by right nerve stimulation (Fig. 7D), matching the onset of GABA_A-mediated inhibition (Fig. 5A). DA1's persistent inhibition enables it to capture relative delays in stimulus onset between the two antennae (Supplemental Movie S2).