Plasticity of local GABAergic interneurons drives olfactory habituation

Abstract

Despite its ubiquity and significance, behavioral habituation is poorly understood in terms of the underlying neural circuit mechanisms. Here, we present evidence that habituation arises from potentiation of inhibitory transmission within a circuit motif commonly repeated in the nervous system. In Drosophila, prior odorant exposure results in a selective reduction of response to this odorant. Both short-term (STH) and long-term (LTH) forms of olfactory habituation require function of the rutabaga-encoded adenylate cyclase in multiglomerular local interneurons (LNs) that mediate GABAergic inhibition in the antennal lobe; LTH additionally requires function of the cAMP response element-binding protein (CREB2) transcription factor in LNs. The odorant selectivity of STH and LTH is mirrored by requirement for NMDA receptors and GABA_A receptors in odorant-selective, glomerulus-specific projection neurons(PNs). The need for the vesicular glutamate transporter in LNs indicates that a subset of these GABAergic neurons also releases glutamate. LTH is associated with a reduction of odorant-evoked calcium fluxes in PNs as well as growth of the respective odorant-responsive glomeruli. These cellular changes use similar mechanisms to those required for behavioral habituation. Taken together with the observation that enhancement of GABAergic transmission is sufficient to attenuate olfactory behavior, these data indicate that habituation arises from glomerulus-selective potentiation of inhibitory synapses in the antennal lobe. We suggest that similar circuit mechanisms may operate in other species and sensory systems.

Habituation is a specific form of implicit learning in which repeated exposure to an unreinforced stimulus results in a decreased behavioral response (1–3). By filtering out such constant sensory input, habituation enhances an animal’s ability to focus its cognitive resources on novel or salient features of the environment. Thus, habituation serves as a building block for normal cognition (2, 4). Although it has been studied in many different contexts, causal connections between mechanisms, neuronal changes, and behavioral habituation have not been clearly identified (2). Given our current understanding, it remains unclear whether similar or distinct mechanisms underlie different forms of habituation; also unclear is how these mechanisms compare with those mechanisms used in consolidation or extinction of associative memory (1, 5).

The olfactory system provides an experimentally accessible circuit in which to analyze mechanisms that underlie different timescales of behavioral habituation (4, 6, 7). Particularly useful is the adult Drosophila olfactory system, which has organization similar to that of mammals (8, 9). Here, olfactory sensory neurons (OSNs) expressing a single type of functional odorant receptor molecules (ORs) send axons to the antennal lobe and synapse onto (i) glomerulus-specific projection neurons (PNs) that project to the mushroom body and lateral horn, (ii) multiglomerular local interneurons (LNs) that mediate both intraglomerular and transglomerular inhibition (10–12), and (iii) other PN and LN types in early stages of characterization (13–17). The antennal lobe is also innervated by neuromodulatory cells, which could regulate both olfactory responses and olfactory plasticity (18–22).

Different odor-induced behaviors show robust habituation in Drosophila (23). Larval short-term olfactory habituation (STH), which is dependent on the rutabaga (rut) -encoded type I adenylate cyclase and central N-Methyl-d-aspartate (NMDA) receptors, shows significant selectivity for the exposed odorant (24–28). In adults, transient habituation of the odor-evoked, reflexive jump response requires rut as well as other cAMP signaling components (29, 30). In addition, a long-term olfactory habituation (LTH) of the avoidance response to benzaldehyde, a chemical potentially sensed by nonolfactory mechanisms, requires rut and persists for several days (31–33). A more recent study characterized LTH of the CO₂ response, better understood in terms of the relevant, Gr21a-expressing sensory neurons and their target projection neurons (VPN) in the V glomerulus (34, 35), which is associated with a reduced olfactory startle response, selective growth of the CO₂-responsive V glomerulus, decreased CO₂-evoked calcium influx in VPN terminals in the lateral horn, and an increased CO₂-induced calcium influx into processes of a subclass of LNs within the V glomerulus (35). However, mechanisms that underlie these phenomena as well as whether they cause habituation remain unknown.

To understand the mechanisms of odorant-selective STH and LTH in terms of (i) the necessary genes, (ii) the cell types in which they function, and (iii) consequent neuronal plasticity necessary for habituation, we standardized behavioral assays for STH and LTH to selected odorants and combined them with genetic, behavioral, anatomical, and live imaging methods feasible in Drosophila. Our results indicate that olfactory habituation arises because of plasticity in the antennal lobe. In addition, they suggest a simple and theoretically generalizable model for habituation in other systems.

Results

STH and LTH in Drosophila.

After a 30-min exposure either to 15% CO₂ or 5% ethyl butyrate (EB), Drosophila show a diminished olfactory avoidance response as measured in a Y-maze test (Fig. 1 A and B and raw data in Tables S1). This STH is selective for the odorant used during training and recovers with a half-life of about 20 min (Fig. 1 B and C).

Fig. 1.

Properties of STH and LTH in Drosophila. (A) Y-maze used to measure aversive odorant response of adult flies. (B) White bars, response to odorants (5% CO₂ or 10⁻³ dilution EB) before exposure; black bars, after 30 min exposure to either 15% CO₂ (Left) or 5% EB (Right). (C) Recovery profile of STH to CO₂ and EB. White bars indicate the percent response of naïve flies, and the black bars indicate the same for the 0- and 30-min recovered flies after odor exposure. ***, comparisons between preexposed and exposed animals at t = 0; +++, comparisons between exposed animals after 0 and 30 min of recovery. One-way repeated measure ANOVA shows significant difference between naive and CO₂ (F = 28.237, P < 0.001) and naive and EB treatments (F = 15.564, P < 0.001). Posthoc testing showed a significant difference between naïve and 0-min recovered flies (q = 10.507, P < 0.001; Student-Newman-Keuls (SNK) test) and 0- and 30-min recovered flies (q = 6.634, P < 0.001; SNK) for CO₂ treatment. The same test for EB treatment showed significant difference between naïve and 0-min recovered flies (q = 7.813, P < 0.001) and 0- and 30-min recovered flies (q = 4.993, P = 0.003). (D) TrpA1-mediated direct activation (29 °C) of OSN subsets for 30 min reduces behavioral responses mediated by these OSNs. White bars, control responses at 22 °C. (E) Four-day exposure to 5% CO₂ (Left) or 20% EB (Right) causes odorant selective LTH. (F) Recovery profile of LTH to CO₂ and EB. White bars indicate the percent response of naïve flies, and the black bars indicate the same for the 0- and 6-d recovered flies after odor exposure. ***, comparisons between mock-exposed and 4-d odor-exposed animals 1 d after exposure; +++, comparisons between exposed animals after 1 and 6 d of recovery. One-way repeated measure ANOVA shows significant difference between mock and CO₂ (F = 239.598, P < 0.001) and mock and EB treatments (F = 82.124, P < 0.001). Posthoc testing showed a significant difference between naïve and 0-d recovered flies (q = 30.958, P < 0.001) and 0- and 6-d recovered flies (q = 15.422, P < 0.001) for CO₂ treatment. The same test for EB treatment showed significant difference between naïve and 0-d recovered flies (q = 16.689, P < 0.001) and 0- and 6-d recovered flies (q = 14.467, P < 0.001). (G) Forced activation of TRPA1-expressing OSNs for 4 d causes odorant-selective LTH. At least eight sets of flies (n) were used for each measurement (exact n and P values are shown in Table S1). Control data showing WT behavior in various Gal4 and Upstream Activation Sequence (UAS) transgene stocks examined in isolation are shown in Fig. S2. Bars show the mean ± SEM. ⁺⁺P < 0.01; ***P < 0.001. The student's t test was used for all statistical comparisons except for Fig. 1 C and F.

Because CO₂ and EB are sensed by nonoverlapping odorant receptors expressed in distinct sets of OSNs expressing Gr21a and Or83b, respectively, odor-selective habituation could, in principle, arise from adaptation of odorant-receptor signaling pathways in distinct groups of OSNs. However, several observations suggest that STH can be induced without receptor activation and thus, cannot be explained by adaptation of the odorant-receptor signaling pathway.

Thirty minutes of direct depolarization of CO₂-sensing, Gr21a-positive OSNs expressing the heat-activated Transient Receptor Potential ion channel A1 (TRPA1) channel (36, 37) results in a reduced response to CO₂ but not to EB (Fig. 1D Left and Figs. S1 and S2). Similar 30-min activation of OSNs expressing Or83b results in selectively reduced avoidance of EB but not of CO₂ (Fig. 1D Right and Figs. S1 and S2). Thus, as reported for Drosophila larvae (28), odorant-selective STH (i) does not require odorant receptor activation and (ii) can arise from a process downstream of action potential firing in OSNs.

After a more prolonged exposure period, where Drosophila adults are maintained in the presence of an odorant (e.g., 5% CO₂) for 4 d, flies tested in a Y-maze (Fig. 1A) show a diminished avoidance response (Fig. 1E) that lasts for several days (Fig. 1F) (31, 35). LTH to CO₂ causes minimal change in response to EB (Fig. 1E Left). Similar LTH can be observed to EB after 4-d exposure to 20% EB. Like LTH to CO₂, LTH to EB is selective, and the decreased response to EB occurs with little effect on CO₂ sensitivity (Fig. 1E Right). LTH to CO₂ and EB are both reversible. When habituated animals are restored to normal rearing environment, normal olfactory sensitivity returns within a few days (Fig. 1F). As shown for STH, a “phantom” LTH to CO₂ or EB can be induced by prolonged (4-d) depolarization of appropriate OSNs. Thus, LTH can also occur without odorant receptor activation through a process downstream of action potential firing in OSNs (Fig. 1G).

A biochemical similarity in the neuronal signaling cascades that underlie STH and LTH is indicated by the observation that both STH and LTH are greatly diminished in rutabaga (rut²⁰⁸⁰) mutants (Fig. 2 and Fig. S3A) deficient in an adenylate cyclase activated by the coincidence of increased cytosolic calcium and G protein-coupled receptor (GPCR) activation (38–41). Measurements of basal odor-evoked behavior in naïve flies show that the observed defects in habituation cannot be ascribed to altered odorant sensitivity in rut mutants (Fig. S3B and Tables S1 and S3).

Fig. 2.

The rut-encoded adenylate cyclase is required in GABAergic/LN1 neurons for habituation. (A) Schematic showing the major classes of olfactory neurons and Gal4 lines that mark them. GAD1 predominantly marks GABAergic neurons and 95% of LN1-expressing interneurons (45–47). The LN1 promoter is highly restricted to a subgroup of antennal lobe interneurons, is distinct from a different set of LN2-positive inhibitory interneuron class, and shows minimal overlap with kra-Gal4–marked excitatory LNs (45) (Fig. S2 F and G). (B and C) Histograms show the efficiency of STH (B) and LTH (C) to EB of rut²⁰⁸⁰ flies expressing a WT rut⁺ transgene (44) in different classes of olfactory neurons. White bars indicate olfactory response behavior of mock-treated siblings; dark bars indicate response after the relevant period of odorant exposure. (D) RNAi-mediated knockdown of rut using a validated transgenic RNAi construct (VDRC5569) (43) in LN1 or GAD1 neurons blocks STH and LTH to EB. Similar results obtained with an independent rut RNAi line, VDRC101759, are not shown. Additional Gal4 and UAS transgene alone controls are shown in Fig. S2. Bars show the mean ± SEM determined through the Student t test [n > 8 (Tables S1 and S3); ***P < 0.001].

cAMP Signaling Is Required in GABAergic LNs for Habituation.

The often unique, anatomical location where rut is required for a given learning task defines cell types in which neuronal plasticity is required for the specific form of memory (42–44). To identify neurons whose intrinsic properties change during olfactory habituation, we looked to identify the cell type(s) in which rut is required (Fig. 2, Figs. S2 and S3, and Tables S1 and S3).

Tissue-restricted expression of a WT rut⁺ transgene in either the LN1 class of antennal lobe local interneurons (45) (Fig. S2 F and G) or GABA-expressing glutamic acid decarboxylase (GAD1) -positive neurons was sufficient to rescue rut²⁰⁸⁰ defects in STH or LTH (Fig. 2 A–C and Fig. S3 C and D). Arguing against a developmental requirement in these cells, experiments using the Gal80^ts conditional expression system show that rut⁺ expression in adult LN1 neurons is sufficient to rescue the habituation defects in rut²⁰⁸⁰ (Fig. 2 A–C and Figs. S2 F and G and S3 C and D). In contrast, rut⁺ expression in OSNs, PNs, or MB247-positive mushroom-body neurons had no effect on the rut²⁰⁸⁰ mutant phenotype (Fig. 2 A–C and Fig. S3 C and D). This finding indicates that rutabaga function is necessary for both STH and LTH resides principally in adult inhibitory local interneurons in the antennal lobe.

The conclusion that rut is necessary in LN1 neurons is also supported by the observation that LN1 or GAD1 promoter-driven expression of either of two previously validated transgenic RNAi constructs against rut (43) also blocked habituation to EB (Fig. 2D and Fig. S3F). Together, these observations indicate that rut-dependent changes in the intrinsic properties of inhibitory local interneurons are necessary for behavioral habituation.

LTH Is Accompanied by a rut-Dependent Reduction in PN Responses to Odorant.

The observed requirement for rut in LNs suggests that behavioral habituation arises from rut-dependent strengthening of GABAergic LN1 synapses in the antennal lobe. This observation predicts first, that odor-evoked responses in PN dendrites will be reduced after habituation and second, that the reduction in PN response will be dependent on rut function. We tested these predictions by measuring the effects of 4-d EB exposure on EB-evoked responses in dorso-medial (DM)2 and DM5 glomeruli, which respond to EB under the conditions used here (48).

EB-evoked responses of PN dendrites in DM2 and DM5 were significantly reduced after habituation. We imaged odor-evoked calcium fluxes in GH146-Gal4, UAS GCaMP3/+ flies (also termed GH146>GCaMP3/+), which carry one copy each of GH146-Gal4 and UAS-GCaMP transgenes heterozygous to a WT + chromosome. Two-second pulses of 0.5% EB induced strong calcium fluxes measured as ∫ΔF/F, the fractional increase in GCaMP fluorescence induced by odor stimulation (Fig. 3 A and B and Table S4). However, 4-d exposure to EB caused a significant reduction in EB-evoked GCaMP responses compared with paraffin-exposed controls (DM2 P = 0.004, DM5 P < 0.001). In contrast, 4-d exposure to 5% CO₂, previously shown to reduce CO₂-evoked responses in the V projection neuron (35), did not reduce the EB response (Fig. 3B). Furthermore, EB exposure had no effect on the 3-octanol (3-Oct) response in the dorsal-central (DC)2 glomerulus (Fig. 3B Right). DC2 has previously been shown to respond to 3-Oct (49) and in our experiments, was unresponsive to 0.5% EB. These data indicate that LTH is accompanied by a selective reduction of PN responses to the habituated odor.

Fig. 3.

LTH to EB is accompanied by a rut-dependent decrease in the EB-evoked response of PNs. (A) Representative images of odor-evoked responses in DM2, DM5, and DC2 imaged in brains of GH146Gal4, UASGCaMP3/ + flies. Pseudocolored images show responses (ΔF) during a 2-s puff of 5 × 10⁻³ EB (DM2 and DM5) or 3 × 10⁻² dilution of 3-Oct (DC2). (Bottom) Representative traces of mean fluorescence change (ΔF/F) in the DM2 and DM5 (EB-evoked) and DC2 (3-Oct–evoked) in 4-d paraffin oil-exposed (solid lines) or 4-d EB-exposed (dashed lines) flies. (B) Quantitative analysis (expressed as percent of mock) of odor-evoked PN responses in the indicated glomeruli and genotypes. White bars indicate responses in mock-exposed flies, black bars indicate responses in 20% EB-exposed flies, and gray bars indicate responses in 5% CO₂-exposed animals. The experiment uses rut²⁰⁸⁰; UAS-rut⁺ flies in place of rut²⁰⁸⁰, because (i) like rut²⁰⁸⁰, these animals do not habituate (Fig. 2) and (ii) they show naïve GCaMP responses indistinguishable from WT controls. In the DM2 glomerulus, two-way ANOVA yielded a significant difference between EB/mock treatments [F_(1,82) = 11.227, P < 0.001]. Posthoc testing showed a significant difference between mock and EB-exposed WT flies (q = 4.167, P = 0.004; SNK) and no difference for rut²⁰⁸⁰; UAS-rut⁺ flies (q = 2.280, P = 0.111; SNK). For WT, n = 6 and n = 4 (mock and EB-exposed, respectively). For rut, n = 12 and n = 8 (mock and EB-exposed, respectively). In the DM5 glomerulus, two-way ANOVA yielded a significant difference between EB/mock treatments [F_(1,79) = 6.469, P = 0.013)]. Posthoc testing showed a significant difference between mock and EB-exposed flies in WT flies (q = 5.554, P < 0.001; SNK) and no difference for rut²⁰⁸⁰; UAS-rut⁺ flies (q = 0.804, P = 0.572; SNK). In EB-exposed flies, WT differed from the rut²⁰⁸⁰; UAS-rut⁺ genotype (q = 5.002, P < 0.001). For WT, n = 6 and n = 7 (mock and EB-exposed, respectively). For rut, n = 8 and n = 8 (mock and EB-exposed, respectively). Bars represent mean area under the curve (first 5.5 s of responses) + SEM (**P < 0.01, ***P < 0.001).

If the observed reduction in PN response causes behavioral habituation, then it should not occur in nonhabituating mutants. Consistent with this premise, 4-d EB exposure had no significant effect on EB-evoked PN responses of nonhabituating, rut²⁰⁸⁰/y; GH146 > GCaMP3/UAS rut⁺ flies (Fig. 3B). This observation tightens the relationship between the observed physiological plasticity and behavioral LTH.

GABAergic Transmission from LNs Is Necessary and Its Enhancement Is Sufficient for Habituation.

If the reduced PN response arises from enhanced GABAergic transmission onto PNs, then synaptic output from LN1 neurons should be necessary for the reduced olfactory aversion in habituated animals, which was previously observed for larval STH (28). To test this prediction for adult STH and LTH, we expressed a dominant, conditional dynamin (Shi^ts1) in local interneurons and used a rapid temperature shift, from 22 °C (permissive) to 32 °C (restrictive) to conditionally reduce transmitter release from LNs during any specified time interval (50). At permissive temperatures (22 °C), LN1 > Shi^ts1 Drosophila showed normal STH to EB and CO₂. However, unlike control flies that show robust habituation at 32 °C (Fig. S4A and Tables S2 and S3), LN1 > Shi^ts1 flies behaved like naive unhabituated animals when tested at temperatures (32 °C) for dynamin function (Fig. 4A and Fig. S4B). Therefore, synaptic output from LN1 neurons is necessary for display of STH (Fig. 4A). We also found that transmitter release from LN1 neurons was necessary during the exposure period (Fig. 4A), an observation that we consider later in this paper.

Fig. 4.

GABAergic transmission from LN1 interneurons is necessary and potentially sufficient for the expression of habituation memory. (A and B) UAS-Shi^ts1/+; LN1-Gal4/+ and UAS-Shi^ts1/+ flies tested for STH (A) or LTH (B) to EB at ∼23 °C (room temperature) or 32 °C either during odor exposure (blue schematic) or testing (red schematic). (C and D) Knockdown of the Rdl GABA_A receptor (through UAS-RdlRNAi; abbreviated as Rdli) in VPN or GH146 neurons selectively blocks STH/LTH to odorants that activate these respective PNs. (E) LN1Gal4/UAS-TRPA1 show reduced responses to EB and CO₂ when tested at 29 °C (dark bars) compared with 22 °C (white bars). Bars indicate mean ± SEM. ***P < 0.001 (Student t test); n of between 4 and 10 sets of flies contribute to each data point (exact n and P values in Tables S1–S3).

Transmitter release from LN1 interneurons is also required for the expression of LTH to CO₂ and EB. Thus, although both control and LN1 > Shi^ts1 adults exposed to EB or CO₂ at permissive temperatures for 4 d show habituated behavior when tested at room temperature, LN1 > Shi^ts1 flies alone behave like naive unhabituated animals when tested at temperatures nonpermissive for Shi^ts1 function (Fig. 4B and Fig. S4C).

To test if GABAergic transmission from LNs to PNs mediates the expression of habituated behavior, we asked whether habituation defects were also evident when the function of GABA receptors was knocked down in odorant-responsive PNs through the transgenic RNAi line (UAS-Rdli) previously shown to knockdown ionotropic GABA_A (Rdl) receptors (51) (Fig. S4F and Tables S1 and S3). Knockdown of GABA_A receptors in defined subsets of PNs selectively blocked habituation to their respective odorants without affecting the basal olfactory response (Fig. 4 C and D and Fig. S4 D and E). Thus, Rdl RNAi expression in GH146-positive neurons that communicate the odor of EB (48) blocked STH and LTH to EB without affecting habituation to CO₂. In contrast, Rdli RNAi expression in the CO₂-responsive projection neuron (VPN) blocks STH and LTH to CO₂ without affecting STH or LTH to EB (Fig. 4 C and D). Taken together with the observation that synaptic output from predominantly GABAergic LN1 neurons is required for display of habituation, these data point to a model in which PN inhibition by LN1 neurons is necessary for the decrease in olfactory avoidance behavior observed after habituation.

To determine whether increased transmission from LN1 neurons could be potentially sufficient to attenuate olfactory behavior, we expressed the heat-activated, cation-permeable TRPA1 channel in LN1 cells and compared olfactory response indices for these LN1 > TRPA1 flies at 22 °C, when TRPA1 channels would be silent, with responses measured just after the flies were shifted to 29 °C, when LN1 neurons show increased activity (Fig. S1). Activation of LN1 neurons was sufficient to substantially reduce aversion of naïve LN1 > TRPA1 flies to CO₂ or EB (Fig. 4E). Thus, transmitter release from LN1 neurons is necessary (Fig. 4 A and B), and the potentiation of LN1 activity is sufficient (Fig. 4E) for the attenuated behavioral response observed in habituation.

PN-Specific NMDA Receptors May Explain Odorant Selectivity in Habituation.

Forced activation of LN1 neurons results in a nonselective reduction in the behavioral response to both CO₂ and EB (Fig. 4E). This finding is easily explained by the fact that most LN1 local interneurons are multiglomerular and would be expected to inhibit most, if not all, PNs (45). However, the observation that behavioral habituation is odorant-selective posits the need for a mechanism to ensure glomerulus-selective potentiation of LN1 synapses in vivo (Fig. 5A). How may synapse-selective potentiation of GABAergic outputs be controlled?

Fig. 5.

NMDA receptors are required in PNs and VGLUT in LNs for odorant-selective habituation. (A) Diagram showing a single multiglomerular LN that forms synapses with two different PNs. (B and C) Effects of PN-specific NMDAR knockdown on STH and LTH to CO₂ and EB. VPN > dsNR1 selectively blocks CO₂ habitiuation; GH146 > dsNR1 selectively blocks EB. (D–F) Behavioral effect of GAD1 or DVGLUT knockdown in LNs and GAD1 expressing neurons. Expression of an RNAi against GAD1 (VDRC32344) or DVGLUT (VDRC104324) in LN1- (D) or GAD1-expressing neurons (E and F) blocks STH and LTH to EB and CO₂. Bars indicate mean ± SEM calculated using the Student t test. ***P < 0.001; n > 7 (Tables S1–S3 show specific n values, and Fig. S2 has additional Gal4 and UAS transgene alone controls).

Previous studies have shown that presynaptic LTP of GABAergic synapses usually arises from a heterosynaptic mechanism that requires retrograde signaling, which is often mediated through postsynaptic NMDA receptor activation (52, 53). Although glutamate is not the main excitatory transmitter in insect brains, it is likely that NMDA receptors, which typically mediate synapse plasticity rather than synaptic depolarization (54), serve evolutionarily conserved functions in synaptic plasticity (55). We, therefore, asked whether postsynaptic NMDA-type glutamate receptors on PNs could contribute to such glomerular selectivity of LN potentiation.

The Drosophila NMDA receptor, a complex of NR1 and NR2 subunits, is expressed in many neurons of the adult antennal lobe (55). Levels of NR1 are substantially reduced by panneural expression of an RNAi transgene, dsNR1 (Fig. S5A). Expression of dsNR1 in CO₂-responsive VPN-positive PNs blocked STH and LTH to CO₂ without affecting habituation to EB (Fig. 5 B and C). In contrast, NMDAR knockdown in EB-responsive GH146-expressing PNs efficiently blocked STH and LTH to EB without affecting habituation to CO₂ (Fig. 5 B and C and Fig. S5C). Thus, the NMDA receptor is required in a PN-specific manner for odorant selective habituation.

In mammals and Drosophila, NMDA receptors respond to the coincidence of postsynaptic depolarization and glutamate release (54, 55). PN depolarization is driven by OSN input. To determine the source of glutamate in the Drosophila antennal lobe, we examined the distribution of the vesicular glutamate transporter (DVGLUT) in vivo. DVGLUT is expressed widely in the antennal lobe, including a significant subset of LN1 local interneurons (Fig. S5B). RNAi-based knockdown of either GAD1 or DVGLUT in LN1 blocks STH and LTH, indicating that LN1 terminals are an important source of both GABA and glutamate required for habituation (Fig. 5D). Disruption of STH and LTH after knockdown of DVGLUT in GAD1-expressing, GABAergic cells (47) further suggests that habituation requires glutamate corelease from GABAergic neurons, a phenomenon previously observed in mammalian nervous systems and recently linked to inhibitory synapse plasticity (56–58) (Fig. 5 E and F). In contrast, mushroom-body expression of the VGLUT RNAi construct in MB247-expressing neurons had no effect on STH (Fig. S5D).

The postulated need for glutamate release from inhibitory local interneurons in the lobe can also explain why STH requires shi function in LN1 cells during initial odorant exposure (Fig. 4A).

Together, the data can explain glomerulus-selective potentiation of LN–PN synapses by postulating that (i) odor stimulation causes glutamate release onto PNs in multiple glomeruli, (ii) glomerular specificity arises because glutamate activates NMDA receptors only on PNs that show coincident, odorant-induced depolarization, and (iii) NMDAR signaling in selected PNs then mediates, possibly through retrograde signaling (52, 53), the potentiation of GABAergic transmission onto these PNs.

Additional observations (Fig. 2 B and C and Fig. S3F and Fig. S5E) showing that the key signaling components (rut, NMDAR, and VGLUT) are required in adult-stage neurons provide additional data consistent with this model.

Glomerulus-Selective Structural Plasticity Requires the Same Mechanisms Required for Odorant-Selective LTH.

Previous work has shown that 4-d exposure to CO₂ or EB causes selective increase in the volume of the V and DM2 glomeruli, respectively (35). Our experiments confirmed these observations (Fig. S6) and additionally showed that 4-d EB exposure, which causes EB-selective LTH (Fig. 1), also causes particularly robust growth of the EB-responsive DM5 glomerulus that mediates the behavioral response to EB (48) (Fig. S6 A and B). We tested whether mechanisms required for odorant-selective habituation were also required for glomerulus-selective structural plasticity.

rut²⁰⁸⁰ mutants that do not show LTH to either CO₂ or EB also do not show associated increases in glomerular volume. Remarkably, expression of a WT rut⁺ transgene in LN1 interneurons or GAD1-expressing neurons not only restores normal LTH to rut mutants but also restores respective glomerulus-selective growth (Fig. 6A and Fig. S6C). Thus, rut is required in LN1/GAD1 cells for both behavioral LTH and LTH-associated structural change.

Fig. 6.

LTH is associated with glomerular-specific volume changes. (A) Effect of CO₂ or EB exposure on glomerular plasticity in rut²⁰⁸⁰; UAS-rut⁺ mutant flies expressing rut⁺ in LN1- or GAD1-expressing cells. (B) PN-specific NMDAR knockdown causes glomerulus-specific block of structural plasticity induced by CO₂ (Left) or EB (Right). Additional control data and glomeruli are shown in Fig. S6 A–D. Bars indicate mean ± SEM computed using the Student t test. n varies between 8 and 16 for each data point. ***P < 0.001; **P ≤ 0.01. White bars, mock-exposed; black bars, odor-exposed.

Similar to their role in behavioral habituation, NMDA receptors are required in odorant-responsive PNs for glomerulus-selective structural plasticity. Consistent with synapse-specific requirement for NMDARs, knockdown of NR1 in VPN neurons blocked not only LTH to CO₂ but also the CO₂-induced increase in V glomerular volume. However, these flies (which showed normal LTH to EB) showed normal EB-evoked growth of the DM5 glomerulus (Fig. 6B and Fig. S6D). In contrast, knockdown of NR1 in GH146 neurons blocked LTH to EB as well as the associated increase in DM5 glomerular volume without affecting LTH to CO₂ or associated growth of the V glomerulus. Thus, postsynaptic NMDARs in PNs contribute not only to odorant-selective habituation but also to glomerulus-selective structural plasticity.

Knockdown of DVGLUT in GAD1 neurons again showed effects analogous to observations in behavioral habituation. Consistent with the need for glutamate secretion from multiglomerular LNs, knockdown of DVGLUT in GAD1-expressing neurons blocked both CO₂- and EB-induced structural plasticity (Fig. S6E).

It is interesting that the GABA_A receptor (Rdl), which is essential for expression of habituated behavior (Fig. 4 C and D), is not necessary for LTH-associated structural plasticity (Fig. S6F). This finding is probably because the stable expression of GABA_A receptors is necessary for expressing the physiological consequence of plasticity in the antennal lobe, namely for the increased GABAergic transmission onto PNs. However, they are not required for the changes per se.

CREB Is Required for LTH but Not STH.

In several neurobiological preparations, structural plasticity and transcriptional gene expression mediated in part by the cAMP response element-binding protein (CREB) transcription factor is required for consolidating short-term memory into a more stably stored, long-term memory. We therefore asked if STH and LTH could be distinguished based on their need for CREB function. We found that a 1-h induction of CREB2b, an inhibitory isoform of CREB, through a heat-inducible hsCREB2b transgene blocked LTH without affecting STH (Fig. 7A and Tables S1–S3). Similar effects could also be seen when CREB2b was transiently induced in the LN1 cells using the Tub-Gal80^ts system (Fig. 7B and Tables S1 and S3). CREB2b induction not only blocked LTH but also associated structural plasticity (Fig. S7). Thus, CREB-dependent transcription in the LN1 class of local interneuron mediates specific forms of plasticity required for long-term but not short-term habituation (Fig. 7B). Taken together, these data point to a simple circuit model for habituation, which is outlined in Fig. 8.

Fig. 7.

Brief induction of an inhibitory isoform of CREB2 (CREB2b) in LNs blocks LTH. (A) Effect of a single 37°C heat pulse (1 h) applied to 0- to 10-h-old hs-dCREB2b adults on STH and LTH. (B) Effect of a single 29°C heat pulse (8 h) applied to 0- to 6-h-old LN1Gal4; TubGal80^ts > UAS dCREB2b adults on LTH. Heat pulses applied to control WT flies had no effect on STH or LTH, and the heat pulse to LN1Gal4; TubGal80^ts > UAS dCREB2b had no effect on STH. Bars indicate mean ± SEM computed using the Student t test. ***P < 0.001. White bars, mock-exposed; black bars, odor-exposed (Tables S1 and S3 show exact n and P values). CREB2b induction also blocked LTH-associated structural plasticity (Fig. S7).

Fig. 8.

A circuit model for the establishment of olfactory habituation. A simplified diagram of OSN–LN–PN connections (Fig. 2A) in a single glomerulus with labels coded black for events involved in naive olfactory behavior (cholinergic transmission from OSNs to LNs and PNs as well as GABAergic transmission from LNs to OSNs and PNs), green for shared signaling pathways initiated during the onset of STH and LTH habituation, red for hypothesized changes that have occurred after STH and LTH (increased GABAergic transmission from LNs to PNs suggested here to be presynaptic), and blue for signaling (CREB activation) and synaptic changes that occur specifically with LTH.

Discussion

Circuit Mechanism of Olfactory Habituation.

A key observation is that rut function is uniquely required in adult-stage GABAergic local interneurons for STH and LTH (Fig. 2). This observation contrasts with the rut requirement in mushroom-body neurons for olfactory aversive memory (42, 44). The demonstration of fundamentally different neural mechanisms used in olfactory habituation and olfactory-associative memory elegantly refutes a proposal of the Rescorla–Wagner model that habituation (and extinction) may be no more than associations made with an unconditioned stimulus of zero intensity (5).

The requirement for rut in inhibitory LNs also indicates that intrinsic properties of multiglomerular LNs change during habituation. However, logic, as well as anatomical and functional imaging data, indicate that glomerulus-selective plasticity must be necessary if LN changes produce odorant-selective habituation (Figs. 3 and 6). A potentially simple mechanism for glomerulus-specific potentiation of LN terminals is suggested by the specific requirement for postsynaptic NMDAR in odorant-responsive glomeruli (Figs. 5 and 8).

The observation that LTH and STH show similar dependence on rut, NMDAR, VGLUT, GABA_A receptors, and transmitter release from LN1 cells indicates a substantially shared circuit mechanism for the two timescales of habituation. The data point to a model in which transient facilitation of GABAergic synapses underlies STH; long-lasting potentiation of these synapses through CREB and synaptic growth-dependent processes underlies LTH (Fig. 8). This finding differs in three ways from synaptic facilitation that underlies Aplysia sensitization (40). First, it refers to inhibitory synapses, with potentiation that may involve a specific heterosynaptic mechanism similar to that used for inhibitory Long Term Potentiation (iLTP) in the rodent ventral tegmentum (52, 53). Second, by presenting evidence for necessary glutamate corelease from GABAergic neurons, it proposes the involvement of a relatively recently discovered synaptic mechanism for plasticity (56). Third, it posits an in vivo mechanism to enable glomerulus-specific plasticity of LN terminals.

It is pleasing that, in all instances tested, physiological and structural plasticity induced by 4-d odorant exposure requires the same mechanisms required for behavioral LTH (Figs. 3–6 and Figs. S6 and S7). When taken together, these different lines of experimental evidence come close to establishing a causal connection between behavioral habituation and accompanying synaptic plasticity in the antennal lobe.

It is important to acknowledge that, although our experiments show that plasticity of LN–PN synapses contributes substantially to the process of behavioral habituation, it remains possible that plasticity of other synapses, such as of recently identified excitatory inputs made onto inhibitory LNs (16, 17, 59), also accompany and contribute to olfactory habituation (60).

Potentially General Mechanism for Habituation?

The conserved organization of olfactory systems suggests that mechanisms of olfactory STH and LTH could be conserved across species (8, 9). Although this prediction remains poorly tested, early observations indicate that a form of pheromonal habituation in rodents, termed the Bruce effect, may arise from enhanced inhibitory feedback onto mitral cells in the vomeronasal organ (61–63).

Less obviously, two features of the circuit mechanism that we describe suggest that it is scalable and generalizable. First, selective strengthening of inhibitory transmission onto active glomeruli can be used to selectively dampen either uniglomerular (CO₂) or multiglomerular (EB) responses; thus, the mechanism is scalable. Second, the antennal lobe/olfactory bulb uses a circuit motif commonly repeated throughout the brain, in which an excitatory principal cell activates not only a downstream neuron but also local inhibitory interneurons, which among other things, limit principal cell excitation (64–66).

It is possible that, in nonolfactory regions of the brains, a sustained pattern of principal neuron activity induced by a prolonged, unreinforced stimulus could similarly result in the specific potentiation of local inhibition onto these principal neurons. Subsequently, the pattern of principal cell activity induced by a second exposure to a now familiar stimulus would be selectively gated such that it would create only weak activation of downstream neurons. In this manner, the circuit model that we propose for olfactory habituation could be theoretically generalized. We expect more studies to test the biological validity of this observation.

Experimental Procedures

Drosophila Stocks.

Unless otherwise stated, all flies were raised at 25 °C on standard cornmeal agar medium. The stocks were obtained from stock centers or through the generosity of Drosophila colleagues as listed in SI Experimental Procedures.

Meauring Olfactory Responses and Habituation.

Olfactory avoidance was measured using an upright Y-Maze apparatus (Fig. 1). Odorant was drawn through one arm of the maze, and control air was drawn through the other arm. Flies starved overnight were allowed into the entry tube, and their preference for the arm with the odorant (O) vs. the control (C) arm with air was quantified as a response index [RI; the difference in the number of flies in the odorant and control arms as a fraction of the total flies RI = (O − C)/(O + C)]. At least eight batches were assayed for each data point, RI values were normalized to the appropriate control (e.g., response of naïve flies for STH, paraffin-exposed flies for LTH to EB, and air exposed for LTH to CO₂), and the data were plotted as percent control response. Except where otherwise indicated, comparisons were carried out by the unpaired Student t test. All batches were coded, and the experimenter was blind to the genotype being tested.

STH.

Two-day-old flies (unless otherwise stated) starved overnight were pretested to 5% CO₂ or 10⁻³ dilution EB to measure the naïve RI. RIs were again determined after the same set of flies was exposed to either 15% CO₂ or 5% EB for 30 min.

LTH.

The protocol was adapted from refs. 31 and 35. Flies aged between 0 and 12 h were collected; 5% CO₂ habituation involved 4-d exposure to CO₂ (air control). EB habituation involved 4-d incubation with a perforated tube with odorant 1:5 dilution of EB in light liquid paraffin oil (paraffin control). For TrpA1-mediated “phantom” habituation, control flies were maintained and tested at room temperature (22 °C); experimental flies were tested at room temperature after exposure to 29 °C for 30 min or 4 d.

Brain Dissection, Immunohistochemistry, and Volume Measurements.

Brains were dissected and stained using standard conditions (67). For measurements of glomerular volume, 3D confocal stacks (FV1000; Olympus) were analyzed using Amira 5.2.0. Comparisons were made using the unpaired Student t test. Except for the rut rescue analysis where male flies were used, 10–15 female flies from each genotype, and at least 10 glomeruli were analyzed to determine the mean and SEM. Investigators performing the measurements were blind with respect to the genotype and experimental treatment.

Imaging Calcium Dynamics in Vivo.

The procedure was as previously described with minor modifications (12) (SI Experimental Procedures). Analysis of changes in fluorescence was performed (with user blinded to genotype) using a custom Matlab script. WT and rutabaga flies (mock or EB-exposed) were compared using a two-way ANOVA. A Student Newman–Keuls posthoc test was used where appropriate. A Student t test was used when only comparing two groups of data.