A Small Subset of Fruitless Subesophageal Neurons Modulate Early Courtship in Drosophila

We show that a small subset of two to six subesophageal neurons, expressing the male products of the male courtship master regulator gene products fruitlessMale (fruM), are required in the early stages of the Drosophila melanogaster male courtship behavioral program. Loss of fruM expression or inhibition of synaptic transmission in these fruM(+) neurons results in delayed courtship initiation and a failure to progress to copulation primarily under visually-deficient conditions. We identify a fruM-dependent sexually dimorphic arborization in the tritocerebrum made by two of these neurons. Furthermore, these SOG neurons extend descending projections to the thorax and abdominal ganglia. These anatomical and functional characteristics place these neurons in the position to integrate gustatory and higher-order signals in order to properly initiate and progress through early courtship.


Introduction
Drosophila melanogaster courtship is a multi-step, progressive series of behaviors established by sex-specific genetic and neurobiological components [1][2][3][4]. Work by our lab and others demonstrated that the expression of male products of the gene fruitless (fru M ) is both necessary and sufficient to specify the potential for male courtship behavior. Approximately 2000 neurons in the central nervous system (roughly 2% of the CNS neuronal population) express Fru M in clusters scattered throughout the central (CNS) and peripheral nervous systems [5][6][7]. In the periphery, expression was detected in subsets of primary sensory neurons of the sensory modalities implicated in courtship. Strikingly, fru M (+) neurons are dedicated to courtship as inactivating them disrupts courtship behaviors, but has no discernible effect on non-sexual behaviors.
Both fruitless Male and another component of the sex determination pathway, doublesex Male are involved in establishment of sexually dimorphic neural circuitry [8]. Activity of Fru M is required for the survival of several male-specific neurons or sexually dimorphic projection patterns [9].
Multiple female sensory cues combine to regulate the activation and performance of male courtship behaviors. A feature of these multi-modal sensory inputs is the partial redundancy of some modalities: loss of visual, olfactory, or gustatory perception does not block courtship from occurring Dros. melanogaster [10][11][12]. Instead loss of any one of these three modalities delays the initiation and decreases the quantity of courtship. These functional redundancies suggest a compensatory integration of these multiple pathways in the courtship circuitry.
Several areas of the CNS have been identified as regions of higher-order processing and integration in the courtship circuitry [13][14][15][16][17][18][19]. fru M (+) projections densely innervate several regions: the lateral protocerebral complex, the mushroom bodies, the mesothoracic triangle in the ventral nerve cord, and the tritocerebral loop. Neurons projecting to the lateral protocerebral complex and mesothoracic triangle induce wing song behavior; subsets of these neurons require Fru M and Dsx M for survival in males [9]. The mushroom bodies are well-characterized regions controlling memory and learning.
One area of interest is the tritocerebral loop-which lies just ventral to the subesophageal ganglion (SOG)-an area of dense innervations targeted by gustatory, protocerebral/neurosecretory, and stomatogastric inputs [20]. Peripheral gustatory axons, from the mouthparts, subsets of the labellum, and stomatogastric nerves, target the tritocerebrum. The termini of descending tracts from the medial superior protocerebrum-notably the pars interecerebralis, a neurosecretory center-innervate the dorsal tritocerebrum. The higher-order interneurons that process and regulate gustatory inputs have not been fully characterized; the tritocerebral loop innervations likely integrate chemosensory and protocerebral inputs.
Here we targeted subpopulations of fru M (+) neurons that regulate chemosensory-dependent courtship initiation. We screened 72 P[GawB] insertions, driving an RNAi construct targeting fru M , UAS-fru M IR [21], for courtship defects that appear only under conditions where melanogaster is visually deficient [10]. The P[GawB]4-57 line exhibited very limited overlap with fru M (+) neurons. P[GawB]4-57 mainly overlapped with two to six fru M (+) neurons in the subesophageal ganglion (SOG), two clusters in the ventral nerve cord (VNC), and inconsistently an area just medial to the antennal lobe (mAL). Knockdown of fru M or inhibition of synaptic fusion limited to the SOG neurons resulted in infrared-specific courtship delays, and a failure to progress to copulatory behaviors. Strikingly, the tritocerebral projections of these neurons were significantly more extensive in males than in females; this male-specific projection pattern required Fru M expression. These fru M (+) SOG neurons likely integrate chemosensory inputs in the tritocerebrum to modulate the initiation and progression of courtship.

Results
We identified subpopulations of fru M (+) neurons involved in chemosensory-specific pathways via a behavioral screen for proper courtship initiation in visually-deficient conditions. To do this we built on the findings of Meissner et al. [21] who screened a collection of approximately 1000 Gal4 P-element, P[GawB], insertions driving expression of two copies of a fru M RNAi construct, UAS-fru M IR (one insertion on the 2 nd and one insertion on the third chromosomes). We screened 65 P[GawB] and 7 candidate Gal4 lines for significant courtship delays to the first unilateral wing extension (courtship latency) in ambient and infrared light ( Figure 1A). Due to lower visual resolution in infrared, we could not reliably identify the first instance of orientation/pursuit, the traditional method of measuring courtship latency. Henceforth courtship latency will refer to the average time to first unilateral wing extension.

P[GawB]4-57-driven fru M -targeted RNAi Leads to Courtship Delays in Infrared
Only under infrared light did we detect a courtship latency defect using the P[GawB]4-57 driver (Figure 2A

P[GawB]4-57-driven Tetanus Toxin also Delays Courtship
We also wanted to determine if repression of synaptic transmission in P[GawB]4-57(+), fru M (+) neurons recapitulated the fru M RNAi results. Due to extensive non-Fru M expression with the P[GawB]4-57 driver, we used an intersectional strategy to inhibit synaptic transmission only in fru M (+), P[GawB]4-57(+) neurons FLP recombinase, under endogenous fru M regulation, limited the expression of tetanus toxin (TNT) to fru M (+), P[GawB]4-57(+) neurons, which we will denote as fru FLP >4-57 [13]. TNT-mediated synaptic inhibition in fru FLP >4-57 males ( Figure  ). The majority of these males performed wing extension/song behaviors but made no detectable attempts at copulation, measured as at least one instance of pronounced curling of the abdomen (.90u curl from horizontal) and/or mounting attempts.
Noting the effect of silencing synaptic transmission on the fru FLP >4-57 neurons, we assayed the effect of depolarization via transgenic manipulation. The UAS.stop.TRPA1 insert, encoding a temperature-sensitive cation channel [10,24] allowed us to depolarize the fru FLP >4-57 neurons by incubation at 28u-32uC ( Figure 2E, J, N). Courtship latency, courtship index, and fraction of males performing courtship did not differ significantly from controls. Unilateral wing extension or abdominal curling was not induced in solitary males. These results suggest that activity from the fru FLP >4-57 neurons, by itself, may not induce wing extension or copulation behaviors.
At this point we will refer to Fru M (+) cells by nomenclature given in [13]. The P[GawB]4-57(+) cluster 7 cells appear to be the DT6 neurons described in that previous study. Also in that study, they characterized 8 subpopulations in the SOG (aSG1-8, pSG1-2). Some of the P[GawB]4-57(+) aSG neurons, characterized here, appear to be either the aSG5 or aSG6 cells. The P[GawB]4-57(+) mSG and pSG cells described here do not appear to confirm to either the pSG1 or pSG2 populations based on anatomy. A diagram of the nervous system showing Fru M expression and cluster locations is shown in Figure 3K, arrows point to locations of the cells in each row of panels.

Cha-Gal80 Repressed P[GawB]4-57-driven GFP Primarily in DT6 and SOG Neurons
We noted that Cha-Gal80 repressed a significant fraction of the P[GawB]4-57-driven UAS-mCD8GFP expression pattern. In Figure 3F-J, hollow arrowheads represent Fru M (+) cells where we failed to detect P[GawB]4-57-driven GFP in combination with Cha-Gal80. Cha-Gal80 repressed GFP fluorescence mainly in the two P[GawB]4-57(+) mSG and two-three of the aSG neurons ( Figure 3G-H, hollow arrowheads). Conversely Cha-Gal80 reduced but did not eliminate GFP fluorescence in P[GawB]4-57(+) DT6 cells ( Figure 3F, see subpanels 1-2 for clarity). In pSG neurons ( Figure 2H) and ventral nerve cord neurons ( Figure 2I-J), GFP fluorescence appeared unaffected. Combined with Cha-Gal80 in behavioral studies, we inferred that some of the aSG and both mSG cells were primarily responsible for the courtship defects seen with P[GawB]4-57-driven constructs. We must note, however, that using Cha-Gal80-mediated repression of GFP may not fully reflect the relationship between P[GawB]4-57 expression and behavioral phenotype. Temporal differences between Cha-Gal80 and P[GawB]4-57 expressions or incomplete repression of GFP are caveats to this inference. In 3/14 brains ( Figure 5A), the aDT6>4-57 neurons presented as one soma located in the periphery of the median bundle, with projections ramifying in the region proximal to the esophagus (eo), just dorsal to the tritocerebrum (Supplemental Figure 1S). Projections continued along the median bundle to the medial superior protocerebrum.
Most aSG>4-57 GFP cells bodies in the SOG had no detectable projections ( Figure 4C, arrowheads), but a single aSG>4-57 neuron was seen, in two brains, with extensive arbors throughout the SOG with a collateral ramifying the inferior lateral protocerebrum (Supplemental Figure 1S, arrowhead, asterisks). pSG>4-57 neurons sent projections to the ventral nerve cord with no detectable arbors in the brain.
Two mSG>4-57 somas were found in the ventral medial SOG in 14/14 brains examined ( Figure 4A-F, arrows). These exhibited similar bilateral projection patterns extending towards the tritocerebrum (TC) along with two descending tracts into the ventral nerve cord (Figure 4K-L).

mSG>4-57 Neurons make Sexually Dimorphic Projections in the Tritocerebrum
We noted a sexual dimorphism in the mSG projections. In female fru FLP /+ brains, fluorescence from the tritocerebral neurites was significantly reduced in extent and intensity compared to male projection patterns (compare Figure 4D-F to 4A-C). This difference was more striking in sagittal reconstructions of the tritocerebral projections ( Figure 4G-H, arrows).
The behavioral defects caused by knockdown of fru M and the dimorphic projection pattern suggested a direct role for Fru M in regulating neurite morphology in these neurons. In order to determine whether the sexual dimorphism required fru M expression, we examined the mSG>4-57 projections in fru FLP /fru LexA mutant males (two different genetic backgrounds were used for the fru LexA chromosome). In these fru mutant males, the tritocerebral projections were similar to those seen in fru FLP /+ females ( Figure 4I, arrow) indicating that fru M expression was required for proper male-specific arbors in the tritocerebrum.

mSG>4-57 Neurons make Descending Projections that Target the VNC
Descending tracts from the mSG>4-57 and the pSG>4-57 cells terminated as presynaptic arbors in prothoracic/mesothoracic ( Figure 4K) and faintly in the mesothoracic/abdominal boundaries (not shown). We utilized the pre-synaptic marker nsyb-GFP and the somatodendritic marker Dscam17.1-GFP [13] to determine   Figure 4L). These data are consistent with a role for the mSG>4-57 neurons in relaying signals targeting the tritocerebrum to the VNC and modulating targets within the ventral tritocerebrum. These were confirmed to be descending termini using the tshirt-Gal80 transgene [13] to repress Gal4 activity in VNC fru M (+) neurons ( Figure 4K  behavioral defects due to Cha-Gal80 repression mainly in the mSG neurons, ( Figure 2B). Two, strong, consistent intersectional expression in mSG>4-57 cells ( Figure 5A). Three, sexual dimorphism of the tritocerebral arbors from the mSG>4-57 cells ( Figure 5B). Four, courtship and copulation defects seen upon fru FLP >4-57 tetanus toxin expression ( Figure 5C). Five, the behavioral defects were primarily seen under visually-deficient conditions (infrared) where male depend more on chemosensory cues, consistent with the putative function of tritocerebral dendritic projections of the mSG>4-57 cells.

Discussion
Initiation of unilateral wing extension is heavily dependent on visual, olfactory, and gustatory cues. By forcing males to depend on non-visual pathways for courtship and co-expressing tissuespecific fru M RNAi, we screened for fru M (+) neurons that likely regulate chemosensory-dependent processes in courtship, which manifested as infrared-specific courtship latency defects. The P[GawB]4-57 line driving UAS-fru M IR possessed normal courtship latency in ambient light and significant infrared-specific delays. Notably fru M overlap was strongest in the SOG, while lacking any detectable peripheral expression. Behavioral and anatomical studies using Cha-Gal80, to subdivide the P[GawB]4-57 expression pattern, highlighted a small subpopulation of fru M (+) neurons in the SOG, two-four anterior SG>4-57 neurons and two medial SG>4-57 neurons as responsible for the courtship defects.

mSG>4-57 Neuronal Anatomy Suggests a Direct Role in Regulating Wing Extension and Copulatory Behavior Initiation
Several lines of evidence suggest a direct role for the mSG>4-57 neurons in regulating the initiation of wing extension and copulatory behaviors. First, we detected expression of fluorescent markers in the mSG>4-57 neurons driven by P[GawB]4-57 in all brains, whereas fluorescence was only detected in a subset of animals for the other fru M >4-57 subpopulations. The mSG>4-57 neurons made sexually dimorphic arbors in the tritocerebrum ( Figure 6A), where male arbors were significantly larger than in wild type female and fru mutant male brains. The mSG>4-57 neuronal tracts extended into the VNC where presynaptic innervation of the mesothoracic triangle was seen ( Figure 6B). The mesothoracic triangle is a target of descending command neurons that control wing song [15][16][17]. Faint projections were detected in the posterior metathoracic/anterior abdominal ganglia, which suggest possible regulation of motor circuitry needed for abdominal curling during copulatory behaviors.
The sexually dimorphic projections of the mSG>4-57 suggest sex-specific roles in receiving tritocerebral signals in males. In males, fru M knockdown and silencing of fru FLP >4-57 neurons resulted in a failure to progress to copulation, a behavior that follows proboscis contact with a female (''licking''). The internal mouthparts house gustatory sensilla that likely detect contact female pheromones accessed via licking behavior.
We cannot rule out functions for the non-mSG>4-57 neurons, particularly the DT6>4-57 (aSG) neurons ( Figure 6C) in regulating courtship initiation, however. Our approach infers, but does not conclusively demonstrate that the mSG>4-57 neurons are responsible for the courtship initiation and copulation defects. Further studies are required to conclusively identify the neurons responsible for each behavioral phenotype and their exact roles.

A Comparison to fru M (+) SOG Neuron in other Studies
Several studies have examined the projections of fru M (+) neurons in the SOG. Antibody staining using anti-Fru M identified 1262 total Fru M (+) nuclei in the SOG in the 2-day pupal brain [5]. An intersectional study, using 131 Gal4 lines with sparse overlap with fru FLP , identified 8 fru FLP (+) SOG neuronal classes divided into six anterior, aSG1-6, and two posterior neuronal types, pSG1-2 [13]. At least one aSG>4-57 neuron's projection pattern, identified here, is consistent with the aSG5 class identified in that larger-scale study. Cachero et al. [14] used mosaic analyses of fru Gal4 to identify larval neuroblast clonal populations of fru Gal4 (+) neurons. Cachero et al. identified six clones in SOG, however, none appear to correspond to neurons identified here. It appears that these broad mapping studies, while extensive, have not exhaustively identified fru-expressing neurons in the SOG.
Using tdc2-Gal4, three studies [24][25][26] characterized three octopaminergic Fru M (+) neurons in the SOG: designated VPM1 and VPM2 (ventral paired median) and one VUM1 (ventral unpaired median) neuron. Expression of tdc2-Gal4-driven UASfru M IR leads to courtship latency delays but no copulation defect. The VUM1 neuron tritocerebral projections appear similar to the mSG>4-57 projections, however, no descending tracts to the VNC were reported. The VPM1 and VPM2 appear to correspond to the DT8 neurons Repression of fru M using tdc2-Gal4 appeared to primarily disrupt male-female discrimination, resulting in significant male-male courtship, whereas we detected no significant male-male courtship using P[GawB]4-57.

The Tritocerebrum is a Major Site of Signal Convergence for fru M Circuitry
Given the extensive projections of fru M (+) innervations, the tritocerebrum appears to be a site of gustatory integration with higher-order information in male courtship. The extensive, sexually dimorphic arbors from the mSG>4-57 receive signals in the tritocerebrum that serve to regulate the progression to copulation in males and the performance of courtship. The tritocerebrum is targeted directly by gustatory afferents from the mouthparts via the pharyngeal nerves, indirectly via the SOG interneurons, which could relay signals from proboscis gustatory afferents entering via the labial nerve, and by descending tracts from the par interecerebralis of the superior medial protocerebrum (SMPR in Figure 6), which contains many neurosecretory cells [27][28]. These mSG>4-57 cells could then relay signals to circuitry controlling wing extension/song in the metathoracic triangle and copulation/abdominal curling in the anterior abdominal ganglia.
The decision to perform courtship by males likely weighs the receptivity of the female versus the cost of female rejection via escape, with greater costs associated with later steps in the ritual, i.e. copulation. In open environs, escape behaviors exhibited by rejecting females likely results in the cessation of the courtship unless the male correctly gauges receptivity. We propose that the fru M (+) SOG neurons identified here play a vital link between detection of female receptivity cues and integration of higher-order signals in order to appropriate initiate wing extension and copulatory behaviors.

Behavioral Assays
Courtship assays. Courtship assays were conducted according to established protocols [29]. Males were entrained in isolation for 3-5 days post-eclosion and then single males were presented with a 1-2-day-old Canton S virgin female. Single male and female pairs were placed into custom-made plexiglass chambers with hollowed circular chambers, 10 mm in diameter and 6 mm in height, separated by plastic transparencies. Contact between courtship pairs was initiated by removal of the transparencies. Courtship behaviors assayed in ambient fluorescent light and in infrared light, recorded for 10-15 minutes, and logged using the LifesongX annotation program. Initial screening of P[GawB] lines involved 9-12 males, while testing of candidate lines involved 20-40 males.
Activity monitoring. Recording of basal activity was done on according to established protocols [30]. Individual males, entrained and aged in the same method for courtship assays, were placed into glass tubes, sealed at one end and plugged with Drosophila media at the other end. The average number of line crossings, measured by an infrared laser, were recorded over a 24 hour period for 16-32 males using the Drosophila Activity Monitoring System I (TriKinetics).

Confocal Microscopy
Tissues were analyzed using a Zeiss LSM510 with a 40X oil objective. Images were taken at 102461024 pixels with slices at 1.0 to 1.5 mm intervals.

Imaging Analyses
Image analyses were conducted using ImageJ and Fiji. The fluorescence density of the neurites was analyzed by sampling the integrated density of the target area and correcting for background by subtracting the fluorescence from three neighboring regions with no detectable labeled neurites. Normalization of tritocerebral arbors was achieved by establishing the average male, female, and fru mutant male brain size. Measurements were adjusted by a normalization factor per axis (sample brain axis/average brain axis).
Partial least squares. PLS does an iterative extraction on both the predictor (P[GawB]:fru P1-LexA overlap) and response (courtship latency) data sets to derive latent variables with the constraint that these latent variables explain the covariance between the data sets [33]. Significance testing was done using random resampling without replacement of the original overlap . Gustatory inputs from the labellum, mouthparts innervate the anterior-medial SOG and the tritocerebrum. Inputs from the superior medial protocerebrum also innervate the tritocerebrum. Descending tracts from the mSG>4-57 innervate prothoracic/metathoracic and mesothoracic/abdominal ganglia boundaries B). The mSG>4-57, and possibly aSG>4-57 neurons, function to regulate initiation of wing extension and copulatory behaviors. SMPR = superior medial protocerebrum, and CC = cervical connective. TC marks the tritocerebrum. The SOG is marked by a dashed blue line. doi:10.1371/journal.pone.0095472.g006 data set. 500 resampled matrices were constructed and analyzed with PLS to estimate confidence intervals.