Driver lines for studying associative learning in Drosophila

Abstract

The mushroom body (MB) is the center for associative learning in insects. In Drosophila, intersectional split-GAL4 drivers and electron microscopy (EM) connectomes have laid the foundation for precise interrogation of the MB neural circuits. However, investigation of many cell types upstream and downstream of the MB has been hindered due to lack of specific driver lines. Here we describe a new collection of over 800 split-GAL4 and split-LexA drivers that cover approximately 300 cell types, including sugar sensory neurons, putative nociceptive ascending neurons, olfactory and thermo-/hygro-sensory projection neurons, interneurons connected with the MB-extrinsic neurons, and various other cell types. We characterized activation phenotypes for a subset of these lines and identified a sugar sensory neuron line most suitable for reward substitution. Leveraging the thousands of confocal microscopy images associated with the collection, we analyzed neuronal morphological stereotypy and discovered that one set of mushroom body output neurons, MBON08/MBON09, exhibits striking individuality and asymmetry across animals. In conjunction with the EM connectome maps, the driver lines reported here offer a powerful resource for functional dissection of neural circuits for associative learning in adult Drosophila.

Keywords: D. melanogaster; dopamine; drivers; individuality; learning; mushroom body; neuroscience.

Figure 1.. Generation and annotation of split-GAL4 lines.

(A) In associative learning, flies adjust their behavioral responses to conditioned stimuli (CS), such as odors and colors, based on the contingency with innately rewarding or punishing unconditioned stimuli (US), such as sugar, bitter, shock and heat. A schematic of Drosophila melanogaster is from Namiki et al., 2018. (B) An image of a standard fly brain with a rendering of the mushroom bodies (MB) and the antennal lobes (AL). Projection neurons (PN) convey information from the AL to the calyx of the MB and the lateral horn (LH). (C) A simplified diagram of the mushroom body circuit. The identity of sensory stimuli is represented by sparse activity patterns of Kenyon cells (KCs). A subset of dopaminergic neurons (DANs) respond to punishment/reward. Dopamine modulates weights of synapses between KCs and MB output neurons (MBONs) depending on the activity status of KCs. The skewed activity patterns of MBONs across compartments in response to the learned stimulus drive memory-based actions and feedback pathways to DANs. (D) A summary of the workflow to generate split-GAL4 lines. (E) Coverage of the collection. The crepine (CRE), the superior medial protocerebrum (SMP), the superior intermediate protocerebrum (SIP) and the superior lateral protocerebrum (SLP) are MB adjacent brain areas where MBONs and DANs most often have arborizations. CX, central complex. LAL, lateral accessory lobes. (F) Examples of cell types covered by the collection. Expression patterns of CsChrimson-mVenus (green) are shown with neuropil counterstaining of Bruchpilot (Brp) with nc82 antibody (magenta). The whole body image of Gr64f-Gal4 line at the left middle panel is shown with muscle counterstaining (magenta). Light gray labels indicate EM-identified neurons labeled by each line (see Supplementary file 1 for details). Putative cell types are bracketed.

Figure 1—figure supplement 1.. Examples of cell types covered by the split-GAL4 lines in this collection.

Additional examples of cell types covered by the collection. Expression patterns of CsChrimson-mVenus (green) are shown with neuropil counterstaining of Brp (magenta). Only one brain hemisphere is shown. Light gray labels indicate EM-identified neurons labeled by each line (see Supplementary file 1 for details).

Figure 2.. LM-EM Match of the CRE011-specific driver SS45245.

(A) Expression pattern of SS45245-split-GAL4 in the brain. (B) MCFO image of SS45245 showing individual labeled neurons. (C) Frontal (top) and ventral (bottom) views of segmented light microscopy (LM) images of an exemplary split-Gal4 line (SS45245) visualized with a membrane reporter (myr-smGFP-FLAG) that was aligned to the JRC2018 standard brain. Projections are shown with outline of relevant neuropils. (D) The skeleton reconstructed from electron microscopy (EM) data of the matched cell type CRE011 in the hemibrain connectome. The CRE011 cell on the right hemisphere is shown. (E) Synaptic connectivity of CRE011 with MBONs and DANs in the MB derived from the hemibrain connectome.

Figure 2—figure supplement 1.. LM-EM match of SS00460.

(A) Frontal and ventral views of segmented light microscopy (LM) images of SS00460 in the JRC2018 standard brain. Projections are shown with the outline of relevant neuropils. A’ (ventral) is a rotated view of A (frontal) with the reference axis indicated. (B) The skeleton reconstructed from electron microscopy (EM) data in the hemibrain volume of the matched cell type v2LN44. (C) EM images of neurons with the most input connections (upstream) to v2LN44 in the hemibrain connectome. The number after ‘#’ indicates the number of connections between the matched cell type (i.e. v2LN44) and its synaptic partners. The color code and EM BodyId for each EM-reconstructed neuron are also listed. (D) EM images of neurons with the most output connections (downstream) from v2LN44 in the hemibrain connectome. In Figure 2—figure supplement 1–20, sensory neurons and KCs are not displayed as top upstream or downstream neurons.

Figure 2—figure supplement 2.. LM-EM match of SS34963.

Figure 2—figure supplement 3.. LM-EM match of SS33915.

Figure 2—figure supplement 4.. LM-EM match of SS35020.

Figure 2—figure supplement 5.. LM-EM match of SS34979.

Figure 2—figure supplement 6.. LM-EM match of SS34947.

Figure 2—figure supplement 7.. LM-EM match of SS00486.

Figure 2—figure supplement 8.. LM-EM match of SS49308.

Figure 2—figure supplement 9.. LM-EM match of SS49361.

Figure 2—figure supplement 10.. LM-EM match of SS49868.

Figure 2—figure supplement 11.. LM-EM match of SS49352.

Figure 2—figure supplement 12.. LM-EM match of SS32228.

Figure 2—figure supplement 13.. LM-EM match of SS32219.

Figure 2—figure supplement 14.. LM-EM match of SS48890.

Figure 2—figure supplement 15.. LM-EM match of SS48799.

Figure 2—figure supplement 16.. LM-EM match of SS32259.

Figure 2—figure supplement 17.. LM-EM match of SS48341.

Figure 2—figure supplement 18.. LM-EM match of SS35040.

Figure 2—figure supplement 19.. LM-EM match of SS33905.

Figure 2—figure supplement 20.. LM-EM match of SS39538.

Figure 3.. Drivers for MBON downstream and DAN upstream neurons.

(A) Examples of confocal microscopy images of split-GAL4 lines (bottom) and their matching cell types in the hemibrain connectome (top). CsChrimson-mVenus (green); Brp (magenta). (B) The number of cell types that receive synaptic output from MBONs and supply synaptic input to DANs. Only cell types with connection (conn) over the indicated thresholds (i.e. more than 4 synapses for DAN upstream and more than 9 synapses for MBON downstream) were considered. The number of covered cell types are indicated in the brackets. (C) A scatter plot of MB interneuron cell types connected with DANs and MBONs. Cell types covered by Split-GAL4 lines are highlighted in red.

Figure 3—figure supplement 1.. New or improved drivers for MB cell types.

(A) Confocal images and matched EM cell types. CsChrimson-mVenus (green); Brp (magenta). (B) Summary of within-MB split-GAL4 coverage. See Supplementary file 5 for details.

Figure 4.. Driver lines for uni-glomerular projection neurons.

(A) Examples of covered uni-glomerular PN (uPN) cell types. (B) Coverage of the 51 antennal lobe glomeruli. The new collection of split-GAL4 covers uPN in the colored glomeruli. (C) Split-GAL4 coverage summary.

Figure 5.. Driver lines for multi-glomerular projection neurons.

Examples of multi-glomerular projection neurons (mPN) types. M_l2PN3t18 [VC5++ l2PN1], M_lPNm11A [VP4++ lPN], and M_smPNm1 [VP3+ smPN] are predicted to receive majority non-olfactory input (Marin et al., 2020).

Figure 6.. Driver lines for non-olfactory projection neurons and sensory neurons.

Examples of thermo-/hygro-sensory PN types (A) and sensory neurons (B) covered by this collection. Other than thermo-/hygro-sensory receptor neurons (TRNs and HRNs), SS00560 and MB408B also label olfactory receptor neurons (ORNs): ORN_VL2p and ORN_VC5 for SS00560, ORN_VL1 and ORN_VC5 for MB408B.

Figure 7.. Driver lines for gustatory sensory neurons.

(A) A summary of new transgenic lines generated by this study. In addition to Gr64f promoters, LexADBD lines were generated with Gr64f, Gr43a, Gr66a, Gr28b promoters and other GMR or VT promoters. The six Gr64f-LexADBD lines are with different insertion sites, and with the presence or absence of the p10 translational enhancer (see Supplementary file 2 for details). (B) Schematic of the screening strategy used here to subdivide the Gr64f-DBD pattern by intersecting it with GMR and VT AD lines. (C) A schematic of the sensory neuron projection types that compose the Gr64f-DBD pattern. (D) A summary of expression patterns of 6 of the Gr64f split-GAL4 lines derived from the screening strategy in (B), using the anatomical notation described in (C). (E) Expression pattern of SS87269 in the brain and VNC. The arrow indicates an ascending projection of atGRN. (F) Expression of SS87269 in the labellum. (G) Expression of SS87269 in tarsi of fore (f), middle (m) and hind (h) legs. (H) Expression of SS87272 in the labral sense organ (LSO). (I) Expression of SS88801 in tarsi. (J) Expression of SS87278 in the abdominal body wall. (K–Q) Expression patterns of designated driver lines in the Gnathal Ganglia (GNG) and VNC. The arrow in K indicates the absence of ascending projections from stGRN. Magenta in F-J indicates muscle counterstaining with phalloidin (actin filaments); magenta in other panels indicates neuropil counterstaining of Brp. All scale bars are 50 µm.

Figure 7—figure supplement 1.. Expression pattern of Gr64f-Gal4.

Expression pattern of Gr64f-GAL4 driving 5xUAS-myr-smFLAG in VK0005 in the brain and VNC (green) with neuropil counterstaining of Brp (magenta). The line used contains two copies of Gr64f-Gal4, with one copy on the second and one copy on the third chromosome. The insert on the left shows a magnified view of each cell type in different colors. The insert on the right shows a sagittal view of the VNC.

Figure 7—figure supplement 2.. Expression pattern of Split-LexA lines from Gr43a and Gr66a.

(A) Split-LexA line derived from sugar-sensory Gr43a-LexADBD. SL99997 labels internal fructose sensory neurons (Miyamoto et al., 2012). The labeling of T1 neurons in SL99998 presumably comes from ectopic expression of Gr43a-LexADBD. (B) Split-LexA line derived from Gr66a-LexADBD labels different subsets of bitter-sensory neurons.

Figure 7—figure supplement 3.. Examples of covered SEZ neurons.

(Top two rows) New or improved split-GAL4 drivers for cell types already covered by the SEZ Split-GAL4 Collection (Sterne et al., 2021). (Bottom row) Split-GAL4 drivers for cell types not included in the SEZ Split-GAL4 Collection. Cell types are from Hemibrain 1.2.1, Sterne et al., 2021 or mn2V (McKellar et al., 2020).

Figure 8.. Behaviors with Gr64f-split-GAL4 lines.

(A) In the optogenetic olfactory arena, odors are delivered from the periphery and sucked out from the central vacuum port to deliver odors only to defined quadrants. Red LEDs (627 nm peak) enable programmed optogenetic activation of neurons expressing CsChrimson. (B) Training and testing protocol for experiments shown in (E). The training protocol consisted of 3x20s optogenetic activation training followed by the first preference test, 1x1 min training followed by the 2nd test, and additional 2x1 min training followed by the last test. Odors delivered to the two zones and odor durations in each period are indicated. LED intensities were 4.3 µW/mm² in early 20 s training and 34.9 µW/mm² in later training. Activation LED was applied with 1 s ON and 1 s OFF during pairing with odor A. Odors were trained reciprocally. Pentyl acetate and Ethyl lactate were used as odor A and B, respectively, in one half of the experiments and the two odors were swapped in the other half of experiments. (C) Protocol to characterize Gr64f split-GAL4 activation phenotypes in the absence of an odor. During each trial, flies were illuminated with a red LED light continuously for 10 s. (D) Summary diagram of the expression patterns of the original Gr64f-GAL4 (far left) and 6 Gr64f-split-GAL4s. The expression of the original Gr64f-GAL4 in olfactory sensory neurons is not depicted here. (E) Associative memory scores after the training protocol in (B). Mean, standard error of the mean (SEM), and the number of groups are shown. (F) The kinematic parameters of trajectories measured with Caltech FlyTracker during split-GAL4 activation in the absence of odor as shown in (C). Return behavior was assessed within a 15 s time window. The probability of return (P return) is the number of flies that made an excursion (>10 mm) and then returned to within 3 mm of their initial position divided by the total number of flies. Curvature is the ratio of angular velocity to walking speed. Each group of flies received 6 activation trials. Summarization was based on the trial average of each group. The number of groups is indicated. The thick lines and shadows are mean and SEM. Gray lines are Empty-split-GAL4 control. Dashed lines are time bins for data summary in Figure 8—figure supplement 2. (G) Average walking speed in each of 6 trials. (H) An image of a tethered fly on a floating ball. Flies were tracked for proboscis extension (PE) activity with the Animal Part Tracker (Kabra et al., 2022). The annotated points, in the order of numbers, consisted of the tip of the abdomen (1), the highest point on the thorax (2), the midpoint between the root of the antennae (3), the base of the proboscis (4) and the tip of the proboscis (5). PE activity was quantified as the change of proboscis length, i.e., the distance from the tip to the base of the proboscis, or the distance between points 4 and 5. (I) SS87269 and SS88801 activation and proboscis extension. Each fly was recorded over 6 activation trials in which the 624 nm LED was turned on for 1 s. LED intensity for SS87269 and SS88801, 11 µW/mm²; for empty Gal4 (pBDPGal4), 50 µW/mm². Less saturated traces indicate behavior during LED off trials, while more saturated traces indicate behavior during LED on trials.

Figure 8—figure supplement 1.. Olfactory arena learning experiment, fly tracking and data analysis.

(A) To test associative learning in the olfactory circular arena, two different odors (A) and (B) were delivered to interleaved quadrants, defined as Zone 1 and Zone 2. The trajectory of an example fly during the odor choice period after automatic tracking of the experiment movies is shown. (B) Experiment protocol for learning as in Figure 8B. Odors (green-A, purple-B) and LED delivered to the two zones are indicated. The transition between sessions that were not video-recorded are masked by gray. In the example data, the odor ethyl lactate (EL) served as CS+ and was paired with CsChrimson-activation of SS87269, while the other odor pentyl acetate (PA) served as CS- and was unpaired. The full data set included an additional reciprocal group with PA as CS + and EL as CS-. (C) Upwind displacement towards CS + and CS- during the experiment was quantified by a change in the mean distance-to-wall for all flies (airflow came from the periphery of the arena and was drawn out in the center). In addition to the three testing periods (0–20 s after odor onset), the mean distance-to-wall during training in the odor periods before LED onset (0–10 s after odor onset) were also quantified. ‘CS+’ and ‘CS-’ groups were compared with a multi-comparison t-test with Bonferroni-Dunn’s correction. *, p<0.05; ***, p<0.001.

Figure 8—figure supplement 2.. Summary data of Gr64f-split-Gal4 activation phenotypes.

The data points during the time bin shown in Figure 8F. (A) Probability of return during the post-activation period. (B) Change of walking speed during the onset of activation period. (C) Change of curvature during the onset of activation period. (D) Change of curvature during the post-activation period. One-way ANOVA followed by Dunnett’s multiple comparisons test. ***, p<0.001. Individual data points are shown with minimum, maximum, median, and interquartile ranges.

Figure 8—figure supplement 3.. Consistency of Gr64f-split-GAL4 phenotypes over repeated activations.

Flies expressing the CsChrimson-mVenus were repeatedly exposed to 10-s red LED as in Figure 8F but with lower intensity (4.3 µW/mm²) and more repetitions. N=8. (A) The averaged time course of probability of return binned over 10 trials. LED was turned on at 10 s and lasted for 10 s. (B) The mean probability of return in the 10-s period after the LED is turned off. (C) The walking speed in the 5-s period before LED onset. Averages of 10-trial bins are displayed, except for the trial zero. Gradual increase of walking speed was observed in the Empty-split-GAL4 control, likely because of dehydration caused by constant exposure to the airflow. (D) Change of walking speed during the activation onset period.

Figure 9.. Examples of covered ascending neurons.

(A) Activation preference screen of 581 split-GAL4 lines (342 lines from this study). SS01159 (blue arrow) is one of the lines that flies showed strong avoidance at optogenetic activation. (B) Time course of flies’ preference to quadrants with red LED light by SS01159>CsChrimson (blue) or empty-GAL4>CsChrimson (gray). A preference score to red LED quadrants was quantified from the distribution during the last 5s of two 30s activation periods. n = 8 groups for SS01159, n = 15 for empty Gal4. Mean (thick lines) and SEM (shadow) are plotted. (C) The mean normalized movement speed at the LED onset for flies in the LED quadrants. The 3-s period before LED onset was used as the baseline for normalization. (D) The mean cumulative turning angles in 5 movie frames (total elapsed time of 167 ms) when flies encountered the LED boundary. The boundary was defined as a 4-mm zone in between the LED and dark quadrants. Trajectories too close to the center (< 0.2*radius) or the wall (> 0.6*radius) of the arena were not analyzed. (E) Examples of split-GAL4 lines for ascending neurons. SS35227 and SS35256 shared a split half (R41C05-AD) with SS01159. SS32217 matched with TPN1 (Kim et al., 2017). No cell types (or only EM BodyIds) were assigned to the other lines shown due to missing information in the hemibrain volume.

Figure 10.. Stereotypy and erroneous projections.

(A) V_l2PN from both hemispheres send axonal projections to the right hemisphere in an atypical case (arrowhead). (B) In this atypical case, there are two MB-DPM neurons in one hemisphere (arrowhead) and the arborizations extend beyond the pedunculus into the calyx. (C) MBON-α1 occasionally has additional arborizations in the ellipsoid body (arrowhead). (D) The localization of MBON-α3 soma widely varied along the dorso-ventral axis. It occasionally had an additional axonal branch (arrowhead). (E) A table to summarize normal and erroneous projections of MBONs and MB-DPM. In all the cases except for the DPM, ‘different number of cells’ was likely due to stochastic expression of the drivers (i.e. lack of labeling) rather than biological difference. We defined ‘mislocalization’ when axons or dendrites projected to outside of the normally targeted brain regions. For instance, dendrites of typical MBON07 are usually confined inside the α1, but were extended to outside the MB lobes in 22.9% of samples. Variable branching patterns inside the normally targeted brain regions were not counted as mislocalization here. In some MB310C-split-GAL4 samples, we observed a third soma attached to MBON-α1 but they lacked any neurites. We did not observe obvious mislocalization of dendrites or axons for MBON03, 5, 6, 12, 18, and 19. See Figure 11 for variability of MBON08/09 in MB083C.

Figure 11.. Individuality and asymmetry of MBON08 and MBON09.

(A) A typical image of two MBONs in MB083C-split-GAL4 driver. (B) Abnormal axonal projection of MBON08/09 observed in one of 169 samples. (C, D) MCFO images of MB083C driver from different flies show that the two cells can either be both MBON09-γ3β’1 (C) or one MBON09-γ3β’1 and one MBON08-γ3 (D). (E–G) An example of MCFO image of MB083C, which visualized one MBON08 and two MBON09 in the same brain. The projection (E) and image sections at the γ3 (F) or β’1 (G) are shown. (H) Diagrams of the three MBONs shown in E-G. (I) A summary table for observation of MBON08 and MBON09 in male and female brains. (J) All possible variations of 4 MBONs in MB083C driver, and estimated probability for each case based on the observations summarized in I. (A) and (C) were adapted from Figure 8 of Aso et al., 2014a.

Figure 11—figure supplement 1.. MB083C invariantly label two MBONs.

Maximum intensity projection images of individual samples are shown. Note that two somas in each hemisphere are labeled in all samples. Reporters used are indicated in each panel.

Figure 11—figure supplement 2.. Subdivisions of medial and lateral γ3 compartments.

(A) Overlay between the dendrites of MBON08/09 and axons of PAM-γ4 DANs. (B) Overlay between the dendrites of MBON08/09 and axons of PAM-γ3 DANs. (C) A subset of PAM-γ3 arborizes in the medial part of the γ3 compartment. Asterisk indicates the position of the lateral γ3 that is not labeled in this line.

Figure 12.. Examples of split-LexA conversion.

Split-LexA shares the same enhancers with split-GAL4 but with the Gal4DBD replaced by LexADBD. Among 34 conversions tested, 22 were successful, with the split-LexA showing identical or similar expression patterns as the split-GAL4. The remaining 12 had weak/no expression or showed unintended intersectional patterns. See Supplementary file 2 for the hemidriver lines with p10 translational enhancers to enhance expression level.