Dopamine neurons that inform Drosophila olfactory memory have distinct, acute functions driving attraction and aversion

Abstract

The brain must guide immediate responses to beneficial and harmful stimuli while simultaneously writing memories for future reference. While both immediate actions and reinforcement learning are instructed by dopamine, how dopaminergic systems maintain coherence between these 2 reward functions is unknown. Through optogenetic activation experiments, we showed that the dopamine neurons that inform olfactory memory in Drosophila have a distinct, parallel function driving attraction and aversion (valence). Sensory neurons required for olfactory memory were dispensable to dopaminergic valence. A broadly projecting set of dopaminergic cells had valence that was dependent on dopamine, glutamate, and octopamine. Similarly, a more restricted dopaminergic cluster with attractive valence was reliant on dopamine and glutamate; flies avoided opto-inhibition of this narrow subset, indicating the role of this cluster in controlling ongoing behavior. Dopamine valence was distinct from output-neuron opto-valence in locomotor pattern, strength, and polarity. Overall, our data suggest that dopamine's acute effect on valence provides a mechanism by which a dopaminergic system can coherently write memories to influence future responses while guiding immediate attraction and aversion.

Fig 1. Activities in different dopaminergic cells drive valence.

(a) Schematic of the optogenetic assay showing that after an initial dark phase, half of the chamber is illuminated with 2 bands of red light. See Methods for further details. (b, c) Schematic of the 2 DAN drivers, R58E02 and R15A04, with projections to MB synaptic zones. R58E02 is expressed in nearly all PAM types, projecting to α1, β1, β2, β’1, β’2, γ4, and γ5, with weaker expression in γ1, γ2, and the peduncle. R15A04 is expressed in PAMs that project to the α1, β2, β’1, and γ5 zones. (d) Schematic of the hypothetical locomotor modes for valence. Top Flies move slowly in the favored area. Bottom Flies maneuver to remain in the favored area. Either mode increases the time spent in the preferred area. (e) R58E02>Chr flies spent more time in the light zones. The upper panel shows the preference indices (PIs) for test flies (red dots) and driver and responder controls (R58E02/+ and Chr/+, gray dots). The lower panel shows the valence effect sizes (mean differences, ΔPI) between control and test flies, with confidence intervals (black line) and the distribution of ΔPI error (blue curve). The positive ΔPI values indicate a positive valence. See S1 Table for detailed genotypes and statistics. (f) R15A04>Chr flies avoided opto-activation. The negative ΔPI values indicate avoidance. (g) Walking behavior of the subset of flies that entered the choice zone from the dark side and approached the dark–light interface. Only data from flies that approached the choice zone were included. Traces of R58E02>Chr paths (black) are aligned to choice-zone entry, i.e., locked to the time of entering the boundary area. The colored lines show the overall mean trajectory. The horizontal axis is aligned to the middle of the choice zone. Test flies slowed or stopped at the boundary, with their heads on either side of the middle of the light interface. (h) Traces of R15A04>Chr flies as they entered the choice zone from the dark side. Trajectory data were taken from epochs with 70 μW/mm illumination. Data for all panels can be found in the corresponding folder on the Zenodo data repository (https://doi.org/10.5281/zenodo.7747425). MB, mushroom body; PAM, paired-anterior-medial.

Fig 2. Inhibition of KC activity with GtACR1 prevents the formation of shock-conditioned and DAN-conditioned olfactory memories.

(a) Optogenetic olfactory conditioning protocol: 2 odors were presented in sequence, one paired with green light; the 2 odors were then presented in different arms of the chamber to test the conditioned preference (see Methods for details). (b) Silencing the KCs with ACR1 actuation decreased aversive shock memory by 75.0% [95 CI −44%, −104%]. The genetic controls were MB247-LexA/+ and LexAop-ACR1/+ (gray dots); the test animals were MB247-LexA/LexAop-ACR1 (green dots). Unilluminated flies (both control and test animals) showed robust learning with shock-paired odor (left side, 0 μW/mm²). Illumination with green light reduced the PI of test flies by ΔPI = −0.34 relative to the genetic controls (right side, 28 μW/mm²). (c) Activation of DANs with green light paired with odor, instructed an attractive olfactory memory in R58E02>Chr test flies (orange dots) relative to controls (R58E02-Gal4/+ and UAS-Chr/+, left side gray dots). The contrast between test and control animals is ΔPI = +0.32 (orange curve). Green light induced almost no memory formation in R58E02>Chr; MB247>ACR1 flies (right side, gray dots). The difference attributable to ACR1 inhibition corresponds to a performance reduction of −120% [95 CI −200%, −24%], i.e., a shift from attraction to mild aversion. Sample sizes: N _experiments = 6, 6, 6; N _flies = 144 per genotype. Data for all panels can be found in the corresponding folder on the Zenodo data repository (https://doi.org/10.5281/zenodo.7747425). KC, Kenyon cell.

Fig 3. KCs are dispensable for R58E02 DAN valence.

(a) R58E02>Chr flies are attracted to green light. The left schematic illustrates the expression pattern of R58E02. Flies carrying all 3 transgenes displayed attraction to green light (green dots), resulting in positive valence (black dots and blue curves in the lower panel). Parental type control flies (R58E02-Gal4/+ or MB247-LexA/+; UAS-Chr/+, gray dots) showed a neutral preference for green light. (b) Relative to genetic controls, MB247-LexA>lexAop-ACR1 flies displayed a modest avoidance of green light at high intensities (22 and 72 μW/mm²). The schematic indicates that MB247-LexA drives ACR expression in most MB-intrinsic KCs. (c) In R58E02>Chr/MB247>ACR1 flies, preference for DAN activation mediated by R58E02-Gal4>UAS-Chr was unaffected by simultaneous opto-inhibition of the MB intrinsic cells with MB247-LexA>lexAop-ACR1. Effect sizes (blue curves) show the net effect of comparing test flies carrying all 4 transgenes (green dots), with controls (gray dots). (d) Simultaneous inhibition of ORNs using Orco>ACR1 had little impact on attraction to PAM DAN activation. At 92 μW/mm², the ΔPI was +0.40 [95 CI +0.17, +0.60], P = 0.001. Transgene abbreviations: LexAop = LOP, Gal4 = G4. Data for all panels can be found in the corresponding folder on the Zenodo data repository (https://doi.org/10.5281/zenodo.7747425). KC, Kenyon cell; MB, mushroom body; ORN, olfactory receptor neuron; PAM, paired-anterior-medial.

Fig 4. For driving DAN-mediated attraction, dopamine is a partial contributor.

(a) Schematic of the use of MB247-Gal4 to knockdown receptor expression by RNAi in the R58E02-LexA>lexAop-Chr optogenetic background. (b, c) Knocking down Dop1R1, Dop1R2, and Dop2R in KCs had minor effects on R58E02-LexA>lexAop-Chr light attraction. The gray ribbon indicates the control-valence 95% confidence interval. Experiments used 72 μW/mm² red light. (d) Schematic for the use of R58E02-Gal4 to simultaneously express Chr and knockdown TH expression. (e) Immunohistochemistry of the PAM DAN cluster stained with α-TH (red) and α-YFP (green) in flies expressing Chr-YFP in R58E02 cells. Yellow rings indicate the co-localization of α-TH and α-YFP signals in cells in the PAM cell-body cluster at 3 optical slices. (f) Immunohistochemistry images of the DANs with TH-RNAi co-expression, showing that cells with an α-YFP signal (R58E02 cells) have a greatly lower α-TH signal. (g, h) Knocking down TH expression with TH-RNAi has a moderate effect on R58E02 valence across 4 intensities. For example, at 70 μW/mm² the valence is +0.79 ΔPI in R58E02>Chr flies (g) and is reduced to +0.59 ΔPI in flies carrying the UAS-TH-RNAi knockdown transgene (h). (i) Averaging summary of the effects of reducing dopamine on R58E02-mediated valence with either gene knockdown (UAS-TH-RNAi, with or without UAS-Dicer) or a chemical inhibitor of TH activity (3-Iodo-L-tyrosine, 3IY). Each dot represents the percentage effect size of light intensity in an experiment (i.e., the R58E02>Chr; TH-RNAi experiment was replicated 3 times). Across all 3 intensities in 5 experiments, dopamine depletion resulted in an average ~46% reduction in valence. The vertical line indicates the 95% confidence interval. Data for all panels can be found in the corresponding folder on the Zenodo data repository (https://doi.org/10.5281/zenodo.7747425). KC, Kenyon cell; PAM, paired-anterior-medial.

Fig 5. Large reductions in PAM valence require knockdown of multiple neurotransmitters.

(a) A knockdown screen for neurotransmitters that contribute to R58E02 valence. R58E02>Chr flies were crossed with RNAi transgenes targeting factors required for 5 transmitters: TH (dopamine, replicating the prior experiment), vGlut (glutamate), GAD1 (GABA), vAchT (acetylcholine), and TβH (octopamine). The vGlut and TβH knockdowns showed a reduction in valence comparable to the TH knockdown. Simultaneous knockdown of TH with either TβH or vGlut resulted in a further reduction in the effect size. (b) A combinatorial approach using 3-IY to systematically deplete dopamine along with RNAi-mediated knockdown of vGlut and TβH reveals a progressive reduction of valence with depletion of each neurotransmitter, with valence being virtually depleted when 3-IY, vGlut-RNAi and TβH-RNAi were simultaneously present (ΔPI = +0.10 [95% CI −0.16, +0.37]). Data for all panels can be found in the corresponding folder on the Zenodo data repository (https://doi.org/10.5281/zenodo.7747425). PAM, paired-anterior-medial; TH, tyrosine hydroxylase.

Fig 6. Valence mediated by dopaminergic PAM-β neurons is dependent on both dopamine and glutamate.

(a, b) An optogenetic activation screen of 22 PAM-DAN lines identified MB213B as the specifically expressing line with the strongest positive valence. Light preference was tested with 72 μW/mm² red light. The table shows the PAM cell types in which each driver expresses; “M” denotes multiple cell types; see S2 Table for further details. (c) The expression pattern of Chr-YFP with driver MB213B, showing projections to both zones of the β lobe (zones β1 and β2). (d) Replication of the MB213B>Chr screen experiment confirmed that these flies are attracted to optogenetic light at the 2 highest intensities. (e, f) Meta-analysis of 3 replicates of MB213B > TH-RNAi; Chr yielded weighted ΔPI values of +0.17 at 22 μW/mm² and +0.25 at 70 μW/mm² (orange curves). (g) Expressing vGlut-RNAi with the MB213B driver similarly resulted in reduced (but not ablated) valence. (h) MB213>ACR1 flies avoided the green-illuminated area. Data for all panels can be found in the corresponding folder on the Zenodo data repository (https://doi.org/10.5281/zenodo.7747425). PAM, paired-anterior-medial.

Fig 7. MBONs drive valence via choice effects, not speed effects.

(a) A screen of optogenetic Chr valence in 15 MBON-related lines (split-Gal4 and Gal4 drivers). Orange markers show the valence scores (black dots) and distributions (curves) of each cross, comparing test flies with controls. See S1 Table for effect sizes. The matrix key shows driver identifiers in the top row, and MBON cell types in which each driver expresses; M denotes multiple cell types. See S2 Table for further details. (b) Schematic of VT999036 projections to MBON synaptic-zone subsets. VT999036 drives expression in 2 MBON types, MBON20 and MBON21 [49]. (c) VT999036>Chr flies avoided opto-activation at the 2 highest illumination intensities (22 and 70 μW/mm²). The valence curve was produced using the same data from the screen summary. (d) When VT999036>Chr flies move through the choice point, they tend to turn away from the light. (e) In VT999036>Chr flies, a relationship between preference and speed ratios was absent. (f) In VT999036>Chr flies, choice index and preference were related. (g) Summary of regressions of DAN and MBON driver valence. Coefficients of determination for DAN lines (blue dots) and MBON lines (orange dots) are shown for 4 locomotor metrics as compared to valence (ΔPreference). The 4 metrics are Δchoice, Δspeed ratio, and the effect sizes of the dark-light and light-dark choice-point exit probabilities (ΔPEDL and ΔPELD, respectively). (h) Combined schematic of combined R58E02 and VT999036 projections to the MB. The 2 expression patterns overlap in the γ4 and γ5 regions, corresponding to MBON21. (i) Knocking down Dop1R1, Dop1R2, and Dop2R in the MBONs of VT999036-Gal4 had modest effects on R58E02>Chr light preference. The gray ribbon indicates the control-valence 95% confidence interval. Data from 70 μW/mm² red light. Data for all panels can be found in the corresponding folder on the Zenodo data repository (https://doi.org/10.5281/zenodo.7747425). MBON, mushroom body output neuron.