1. Genetics and Genomics
  2. Neuroscience
Download icon

Sex-determining genes distinctly regulate courtship capability and target preference via sexually dimorphic neurons

  1. Kenichi Ishii
  2. Margot Wohl
  3. Andre DeSouza
  4. Kenta Asahina  Is a corresponding author
  1. Molecular Neurobiology Laboratory, Salk Institute for Biological Studies, United States
  2. Neuroscience Graduate Program, University of California, San Diego, United States
Research Article
  • Cited 2
  • Views 1,482
  • Annotations
Cite this article as: eLife 2020;9:e52701 doi: 10.7554/eLife.52701

Abstract

For successful mating, a male animal must execute effective courtship behaviors toward a receptive target sex, which is female. Whether the courtship execution capability and upregulation of courtship toward females are specified through separable sex-determining genetic pathways remains uncharacterized. Here, we found that one of the two Drosophila sex-determining genes, doublesex (dsx), specifies a male-specific neuronal component that serves as an execution mechanism for courtship behavior, whereas fruitless (fru) is required for enhancement of courtship behavior toward females. The dsx-dependent courtship execution mechanism includes a specific subclass within a neuronal cluster that co-express dsx and fru. This cluster contains at least another subclass that is specified cooperatively by both dsx and fru. Although these neuronal populations can also promote aggressive behavior toward male flies, this capacity requires fru-dependent mechanisms. Our results uncover how sex-determining genes specify execution capability and female-specific enhancement of courtship behavior through separable yet cooperative neurogenetic mechanisms.

Introduction

Social behaviors, such as courtship and aggressive behaviors, are consequential for the fitness of many animal species. Successful male courtship behavior is vital for the propagation of offspring, whereas intra-specific aggressive behavior is often the key to ensure access to potential mates. These two behaviors therefore serve to directly and indirectly increase the reproductive success of males. Likely owing to the contribution to a common goal, e.g. reproduction, courtship and aggressive behaviors are often simultaneously up- or down-regulated. In vertebrates, sex hormones, such as estrogens and testosterones, orchestrate sexually dimorphic reproductive behaviors by organizing underlying neural circuits during development, and activating these behaviors during adult stages (McCarthy, 2008; McEwen, 1981; Tinbergen, 1951). However, expression of either courtship or aggression to the wrong target sex can be costly. Thus, an animal must be equipped with a neural mechanism that co-regulates these two behaviors while also executing each action selectively toward the right target. How the nervous system manages these two seemingly conflicting demands remains a central question in the neurobiology of social behaviors (Anderson, 2016; Chen and Hong, 2018; Li and Dulac, 2018).

One possible mechanism that can account for the target sex-selective execution of an appropriate behavior is that sex-specific sensory cues are channeled into a dedicated circuit, which in turn triggers a specific behavior toward each sex (Kohatsu et al., 2011; Ruta et al., 2010; von Philipsborn et al., 2011). This hypothetical chain of neurons are sometimes referred to as a ‘labeled line’ (Ishii et al., 2017). In this relatively rigid circuit, behaviorally relevant sensory information represented in the brain is transformed into the motor execution of sexually dimorphic behavior in a stereotypical manner (Chen and Hong, 2018; Manoli et al., 2013). Simply put, the labeled line hypothesis predicts that the activation of any neuron that is downstream of sensory perception (‘releaser’ in a classical term Tinbergen, 1951) should in principle trigger the behavior of interest, regardless of the presence of external stimuli (von Philipsborn et al., 2011). Sex-specific sensory cues are differentially represented in the brains of both males and females (Bayless et al., 2019; Bergan et al., 2014; Datta et al., 2008; Haga et al., 2010; Kohatsu et al., 2011; Kohl et al., 2013; Li et al., 2017; Remedios et al., 2017), which is consistent with a simple two-step mechanism that generates sexually dimorphic behaviors: recognition of sex, followed by execution of sex-specific behaviors.

Interestingly, certain populations of neurons are known to be capable of promoting both courtship and aggressive behaviors. In mice, stimulation of neurons in ventromedial hypothalamus (Lee et al., 2014) and medial amygdala (Hong et al., 2014) induces both behaviors. In fruit flies, neurons located at the posterior medial part of the male brain (commonly referred to as ‘P1’ or ‘pC1’ neurons) show a similar dual functionality (Hoopfer et al., 2015; Koganezawa et al., 2016). Although several mechanisms have been proposed to account for the generation of two behaviors by a seemingly single population of neurons (Anderson, 2016; Koganezawa et al., 2016), an underlying implicit assumption remains that the activation pattern of the given neurons determines the behavioral outcome. However, for mouse medial amygdala and ventromedial hypothalamus (Li et al., 2017; Remedios et al., 2017), target sex-specific activation patterns are not hard-wired, but instead emerge as a result of intra-species interactions. Moreover, certain aspects of aggressive behavior induced by the artificial stimulation of mouse ventromedial hypothalamus can be modulated by social contexts, including the sex of target animals (Lin et al., 2011; Yang et al., 2017). These observations raise the possibility that the hierarchy between the sex-recognition mechanism and the execution mechanism may not be as unidirectional as a ‘labeled line’ model suggests. Increasing cellular precision for neural manipulation, especially paired with the temporal precision of optogenetics, provides a unique opportunity to address how the sex-recognition mechanism and the execution mechanism interact with one another, and where in the neural circuit behavior specificity toward a given sex is generated.

Here, we addressed the genetic and neural origins of the two mechanisms – sex recognition and execution of behaviors – underlying sexually dimorphic social behaviors of the fruit fly Drosophila melanogaster. Instead of sex hormones, Drosophila uses two sex-determining genes, doublesex (dsx) and fruitless (fru), to specify sexual dimorphisms at the cellular level. We found that optogenetic stimulation of a specific subset of P1/pC1 neurons induces more courtship behavior toward females than toward males. Neuroanatomical sexual dimorphism of this P1/pC1 subset is specified predominantly by dsx, and its capacity to promote courtship is retained regardless of fru function. However, fru is necessary to enhance courtship behavior specifically toward female targets after optogenetic stimulation. We also found evidence that P1/pC1 neurons consist of genetically and functionally heterogeneous populations, one of which requires the contribution of both dsx and fru for specification of neuroanatomical sexual dimorphism. Lastly, in contrast to courtship behavior, aggressive behavior requires a fru-dependent execution mechanism. Our studies illuminate the previously under-appreciated importance of the sex of a target animal as a behaviorally relevant biological variable, and suggest that the relationship between the sex-recognition mechanism and the behavior execution mechanism is interactive rather than hierarchical. For courtship behavior, the function of dsx resembles the organizational role of the vertebrate steroid hormone, whereas the function of fru can be framed as the activation role. This separation of functions does not extend to aggressive behavior, suggesting that execution mechanisms for different types of sexually dimorphic social behaviors may be specified through separable genetic mechanisms. The neurogenetic approach we employed presents a path to dissect the genetic and circuitry origins of sexually dimorphic social behaviors.

Results

The target fly’s sex affects the function of social behavior-promoting neurons

Both dsx and fru genes control the sexually dimorphic specification of Drosophila neurons that are critical for sexual behaviors both in males and females (Dickson, 2008; Ellendersen and von Philipsborn, 2017; Yamamoto and Koganezawa, 2013). Namely, a cluster of up to 60 sexually dimorphic neurons located at the posterior medial part of the male Drosophila brain, collectively referred to as ‘P1’ (Cachero et al., 2010; Kimura et al., 2008; Kohatsu et al., 2011; Lee et al., 2000; Pan et al., 2012; Ren et al., 2016; von Philipsborn et al., 2011; Yu et al., 2010; Zhou et al., 2015) or ‘pC1’ (Deutsch et al., 2020; Kohatsu and Yamamoto, 2015; Lee et al., 2002; Palavicino-Maggio et al., 2019; Ren et al., 2016; Rideout et al., 2010; Robinett et al., 2010; Sanders and Arbeitman, 2008; Wang et al., 2020; Zhou et al., 2014) neurons, are considered central for various aspects of male and female reproductive behaviors (Auer and Benton, 2016; Ellendersen and von Philipsborn, 2017). Artificial activation of male P1/pC1 neurons can induce courtship behavior in the absence of a target fly (Bath et al., 2014; Inagaki et al., 2014; Kohatsu et al., 2011; von Philipsborn et al., 2011), suggesting that these neurons can serve as an execution mechanism for courtship. However, activation of certain P1/pC1 subsets are reported to promote aggressive as well as courtship behavior when a male target fly is present (Hoopfer et al., 2015; Koganezawa et al., 2016), raising a possibility that the function of P1/pC1 neurons is not entirely independent of the target sex.

To address this, we generated ‘tester’ flies in which the red-shifted channelrhodopsin CsChrimson (Klapoetke et al., 2014) was expressed in dsx, fru-co-expressing neurons by combining dsxGAL4 (Rideout et al., 2010) and fruFLP (Yu et al., 2010), which are knock-in alleles of dsx and fru, respectively. We then used a transgenic element in which the coding sequence for CsChrimson is placed downstream of both the upstream activation sequence (UAS), to which the yeast transcription activator GAL4 binds and drives transcription, and a transcriptional termination signal that can be excised only in the presence of the DNA recombinase Flippase (FLP). Under this configuration, CsChrimson proteins are expressed only in cells in which both GAL4 and FLP are present, which in this case should be the neurons that express dsx and fru (dsxGAL4 ∩ fruFLP neurons). We visualized neuronal morphology and soma by detecting immunoreactivity to a red fluorescent protein tdTomato that tags CsChrimson. This approach eliminates possible discrepancies of labeling patterns between marker genes and untagged effector proteins, which cannot be directly visualized.

We observed CsChrimson expression in specific neuronal clusters that correspond to previously characterized dsx, fru-co-expressing neurons in the male brain and ventral nerve cord, including P1/pC1 neurons (Figure 1A,B, Figure 1—figure supplement 1ARideout et al., 2007; Rideout et al., 2010; Sanders and Arbeitman, 2008; Zhou et al., 2014). In contrast, we found very few labeled cells in the female brain (Figure 1B, Figure 1—figure supplement 1B). We note that the number of cells labeled by our approach was significantly fewer than neurons reported to co-express dsx and fru by immunohistochemistry (Rideout et al., 2007) likely because of a mismatch between knock-in alleles and endogenous gene expression patterns (Stockinger et al., 2005; Yu et al., 2010), a difference in the expression levels of UAS transgenic elements (Pfeiffer et al., 2010; Pfeiffer et al., 2012), or an incomplete excision by FLP of the transcriptional termination signals (Nern et al., 2011).

Figure 1 with 3 supplements see all
Sex of the target fly influences behaviors triggered by the optogenetic activation of social behavior-promoting neurons.

(A) Expression of CsChrimson:tdTomato under the control of dsxGAL4 and fruFLP (red in A1, black in A2,3) in a male brain is visualized together with a neuropil marker BRP (blue in A1) by immunohistochemistry. Labeled cell body clusters are enlarged in A3. Scale bar: 100 μm (A1), 10 μm (A3). (B) Mean number of cell bodies per hemibrain visualized by anti-DsRed antibody in male (left) and female (right) brains. (C) Schematics of the design of behavioral assays. (D) Schematics of the optogenetic stimulation paradigm. Time windows 1–4 represent periods in which behavioral parameters are pooled and calculated in subsequent panels. (E, G) Rasters of behaviors (indicated in left) performed by male tester flies that express CsChrimson:tdTomato under the control of dsxGAL4 and fruFLP. Pink bar: LED-on periods, horizontal bar: 10 min (also see D), vertical bar: 10 flies. LED stimulation condition is indicated at the bottom. (F, H): Boxplots of time orienting (F1, H1), lunges (F2, H2), and wing extension (F3, H3) by the tester flies during the time windows 1–4 (see D). Testers’ genotypes and pair numbers are indicated below the plots. Gray lines represent single testers. In gray: **p<0.01, n.s. p>0.05 (Kruskal-Wallis one-way ANOVA and post-hoc Wilcoxon signed rank test). In black: **p<0.01, n.s., p>0.05 (Kruskal-Wallis one-way ANOVA and post-hoc Mann-Whitney U-test). Target flies are either wild-type group-housed males (E, F) or mated females (G, H). (I) Comparison of time orienting (I1), lunges (I2), and wing extension (I3) performed by tester flies that express CsChrimson:tdTomato under the control of dsxGAL4 and fruFLP in males (data replotted from F, H) toward male or female target flies (indicated above). Number of pairs tested and time windows compared are indicated below the panels. **p<0.01 (Mann-Whitney U-test). For detailed methods to quantify behaviors, see Materials and methods.

We then observed behaviors of these CsChrimson-expressing flies (‘tester’ flies) toward either a wild-type male or female fly (‘target’ fly) in a behavioral arena equipped with programmable light-emitting diodes (LEDs) for photostimulation (Inagaki et al., 2014Figure 1C). After 1 min of a pre-stimulation period, we applied three blocks of 1 min LED illumination, separated by 2 min inter-stimulus intervals (ISIs) during which the LEDs were turned off (Figure 1D). We used three parameters to quantify social interactions: 1) the duration of ‘time orienting’ toward a target fly (see Materials and methods for the definition), which is a behavior-neutral parameter associated with levels of interaction; 2) number of lunges, which is a highly expressed male-type aggressive action (Chen et al., 2002; Jacobs, 1960; and 3) duration of unilateral wing extensions (henceforth referred to as wing extensions), which is a frequent action during male-type courtship behavior (Hall, 1994; Murthy, 2010). For lunges and wing extensions, we developed automated behavioral classifiers using the machine learning-based platform JAABA (Kabra et al., 2013Figure 1—figure supplement 2A, B1-2; see also Materials and methods for details).

When the target fly was male, optogenetic activation of dsxGAL4 ∩ fruFLP neurons induced both wing extensions and lunges across a range of stimulation LED frequencies (Figure 1E,F, Figure 1—figure supplement 3A; see also Video 1 – part 1). Toward a female target fly, the same optogenetic activation induced robust wing extensions (Figure 1G3–H3, Figure 1—figure supplement 3B3) but no lunges (Figure 1G2, H2, I2, Figure 1—figure supplement 3B2) (see also Video 1 – part 2). The presence of female target flies induced more wing extensions from tester flies than did the presence of male target flies (Figure 1I3). Optogenetic activation of dsxGAL4 ∩ fruFLP neurons also increased time orienting toward females more than it did toward males (Figure 1E1, I1), suggesting that the lack of lunges toward female targets is not due to the lack of opportunities to interact (due to immobility, for instance). We therefore conclude that activation of dsxGAL4 ∩ fruFLP neurons can induce more wing extensions toward female than toward male targets, and lunges only if a male target fly is present. In other words, the activation of this neuronal population by itself does not determine the behavior of the tester fly. Even under optogenetic manipulations, behaviors of tester flies are influenced by target flies, likely through sex-specific sensory cues or subtle behaviors of target flies that are currently not quantified.

Video 1
Representative behavior of a male tester fly that expresses.

CsChrimson:tdTomato under the control of dsxGAL4 and fruFLP toward a wild-type male (Part 1) or a wild-type female (Part 2) target fly, at the onset and offset of LED stimulation.

NP2631 ∩ dsxFLP neurons contain subpopulations of the P1/pC1 cluster that promote both courtship and aggression dsxGAL4 ∩ fruFLP neurons in the brain contain several distinct clusters (Figure 1A3Rideout et al., 2007; Zhou et al., 2014). As discussed above, the P1/pC1 cluster is the most likely neural substrate among them that trigger the above-described social behaviors (Auer and Benton, 2016; Ellendersen and von Philipsborn, 2017; Koganezawa et al., 2016; Pan and Baker, 2014; Pan et al., 2011) (see also Deutsch et al., 2019; Kohatsu and Yamamoto, 2015 for contributions of the pC2 cluster). Our finding that activation of dsxGAL4 ∩ fruFLP neurons as a whole triggers social behaviors in a target sex-dependent manner prompted us to ask whether the previously reported ‘dual functionality’ of P1/pC1 subsets (e.g., neurons that induce both courtship and aggression upon stimulation; Hoopfer et al., 2015; Koganezawa et al., 2016) is also influenced by the sex of target flies.

We first focused on the neurons that are labeled by the combination of an enhancer trap GAL4 line NP2631 and dsxFLP, a knock-in allele of dsx (Rezával et al., 2014) (we hereafter refer to this population as ‘NP2631 ∩ dsxFLP neurons’). This population includes a specific subset of P1/pC1 neurons that is reported to promote both courtship and aggression toward a male target (Koganezawa et al., 2016). Consistent with this, we found that the optogenetic activation of NP2631 ∩ dsxFLP neurons robustly increased both wing extensions and lunges toward male target flies (Figure 2A, Figure 2—figure supplement 1A; see also Video 2 – part 1). Interestingly, however, the same manipulation did not trigger lunges when the target flies were females (Figure 2B2, Figure 2—figure supplement 1B). Instead, activation of NP2631 ∩ dsxFLP neurons in the presence of female target flies induced more wing extensions during LED stimulations than during ISIs (Figure 2B3; see also Video 2 – Part 2). Time orienting toward both male and female targets increased after optogenetic stimulation of NP2631 ∩ dsxFLP neurons (Figure 2A1, B1), excluding the possibility that tester males did not have an opportunity to lunge toward female targets. These behavioral changes were largely consistent across different stimulation LED frequencies (Figure 2—figure supplement 1C). However, in contrast to the previous observation (Koganezawa et al., 2016), we did not observe a lunge-like behavior or change in speed when we optogenetically stimulate NP2631 ∩ dsxFLP neurons in the absence of target flies (Figure 2—figure supplement 1D). Thus, the function of NP2631 ∩ dsxFLP neurons depends on the sex of target flies.

Figure 2 with 3 supplements see all
NP2631 ∩ dsxFLP neurons that express both dsx and fru can promote courtship and aggressive behaviors.

(A–C): Boxplots of time orienting (A1–C1), lunges (A2–C2), and wing extension (A3–C3) by the tester flies during the time windows 1–4 (see Figure 1D). Testers’ genotypes and pair numbers are indicated below the plots. Gray lines represent single testers. Target flies are either group-housed wild-type males (A, C) or mated females (B). Gray lines represent single testers. In gray: **p<0.01, n.s. p>0.05 (Kruskal-Wallis one-way ANOVA and post-hoc Wilcoxon signed rank test). In black: **p<0.01, n.s. p>0.05 (Kruskal-Wallis one-way ANOVA and post-hoc Mann-Whitney U-test). (D). A schematic summary. A subset of NP2631 ∩ dsxFLP neurons that express fru (labeled by the fruP1.LexA allele) (orange) can promote both lunges and wing extensions.

Video 2
Representative behavior of a male tester fly that expresses.

CsChrimson:tdTomato under the control of NP2631 and dsxFLP toward a wild-type male (Part 1) or a wild-type female (Part 2) target fly, at the onset and offset of LED stimulation.

With the genetic reagents we used (UAS > stop > CsChrimson:tdTomato as a reporter as well as an effector element: see Supplementary file 1 for details), we consistently detected expression of CsChrimson:tdTomato in a single pair of neuronal clusters located at the medial posterior region of male brains (23 out of 23 brains examined in Figure 2—figure supplement 2D1; orange circles in Figure 2—figure supplement 2A1; see also Video 3). As was reported in Koganezawa et al. (2016), we found that nearly all of these cell bodies express Dsx proteins, and a subset of them also co-express FruM proteins (Figure 2—figure supplement 2B,C), indicating that they belong to the ‘P1/pC1’ neuronal cluster. In addition, we occasionally observed a neuron with a large cell body and a tract that is not shared with the P1/pC1 cluster. This neuron has a prominent arborization in the contralateral ventromedial neuropil and a long descending fiber through the entire ventral nerve cord (Figure 2—figure supplement 2A2). We could not resolve its arborizations that overlap with those of the above-mentioned P1/pC1 cluster. The overall morphology resembles the previously documented dsx-expressing pMN1 neuron (Deutsch et al., 2019; Bogovic et al., 2018; Robinett et al., 2010), but unlike the descending neurons we found, the arborization in the ventromedial neuropil seems absent in pMN1 (Kimura et al., 2015). This descending neuron was present in up to 35% of male brains, all of which had only one such neuron in one of the two hemispheres (Figure 2 – figure supplement D1). Importantly, we observed an increase in both lunges and wing extensions upon optogenetic stimulation of NP2631 ∩ dsxFLP neurons far more consistently than the presence of the descending neuron (Figure 2—figure supplements 2D2 and 3). Based on this observation, we conclude that the behavioral effects by the activation of NP2631 ∩ dsxFLP neurons are due to its P1/pC1 neuronal cluster. For the sake of simplicity, we will hereafter use ‘NP2631 ∩ dsxFLP neurons’ to refer to its P1/pC1 cluster. We do not exclude the possibility that the occasionally labeled descending neuron can contribute to social behaviors as well.

Video 3
3D-rendered average image of NP2631 ∩ dsxFLP neurons in male (green) and in female (magenta).

Gray represents a standard unisex Drosophila brain (Bogovic et al., 2018).

The previous report (Koganezawa et al., 2016) concluded that the fru-expressing subset of NP2631 ∩ dsxFLP neurons specifically promotes courtship, whereas the fru-negative subset specifically promotes aggression. This division of functions is in contrast to our observation with dsxGAL4 ∩ fruFLP neurons. We therefore took the same genetic approach as the previous report (Koganezawa et al., 2016) to limit the expression of CsChrimson in the neurons that do not express fruP1.LexA (a knock-in allele that expresses a bacterial transcription factor LexA in place of male-specific fru isoforms) (Mellert et al., 2010). We confirmed that this genetic approach reduced the number of labeled NP2631 ∩ dsxFLP neurons (3.1 ± 1.3, n = 16) to a degree that is predicted from the result of FruM immunohistochemistry within this cluster (Figure 2—figure supplement 2E–H). When the expression of CsChrimson was suppressed in fruP1.LexA-expressing subpopulations, the amount of optogenetically induced lunges and wing extensions were both significantly decreased compared to controls that have CsChrimson in all NP2631 ∩ dsxFLP neurons (Figure 2C). The amount of lunges during LED stimulations in this genotype was still significantly higher than those before LED stimulations or during ISIs (Figure 2C2), which can be due to some fruP1.LexA-negative neurons that specifically promote aggression (Koganezawa et al., 2016), or incomplete suppression of NP2631 in fruP1.LexA-expressing neurons in some animals. Regardless, the significant reduction of lunges and wing extensions in the absence of the fruP1.LexA-expressing subgroup favors the idea that the fru-expressing subpopulation (NP2631 ∩ dsxFLPfruP1.LexA) is capable of promoting both behaviors. These results led us to conclude that NP2631 ∩ dsxFLP neurons contain dsx and fru-expressing subset of P1/pC1 neurons that can promote courtship and aggressive behaviors depending on the sex of target flies (Figure 2D).

Courtship or aggressive behaviors of target flies can feed back to tester flies and change their behaviors. Therefore, we also quantified behaviors of male and female target flies while the NP2631 ∩ dsxFLP neurons of tester flies were optogenetically activated. For this purpose, we created a classifier for headbutt, a female-type aggressive action (Nilsen et al., 2004; Ueda and Kidokoro, 2002Figure 1—figure supplement 1A, B3). We observed that the behavioral levels of group-housed target flies, for the most part, remained low regardless of the genotypes of tester flies (Figure 2—figure supplement 3A,B). Although male target flies under this condition showed a statistically significant increase in lunges (Figure 2—figure supplement 3A2), the magnitude was very small. We therefore conclude that target flies do not alter tester flies’ behaviors by actively performing courtship or aggressive behaviors. For clarity, we henceforth only show behaviors of the tester flies.

Dsx specifies NP2631 ∩ dsxFLP neurons

Considering the importance of dsx and fru-co-expressing neurons on both courtship and aggressive behaviors, we next asked the role of the two sex-determining genes on the specification of NP2631 ∩ dsxFLP neurons. As a result of sex-specific splicing (Figure 3A), males (whose sex chromosome composition is XY) transcribe male-specific mRNA isoforms of dsx (dsxM) and fru (fruM) (green in Figure 3B), whereas females (whose sex chromosome composition is XX) have female-specific mRNA isoforms of dsx (dsxF) and fru (fruF) (magenta in Figure 3B). To dissociate the role of dsx and fru, we took advantage of two genetic alleles of fru, fruM and fruF (Demir and Dickson, 2005). These fru alleles force male and female-specific splicing, respectively, regardless of the sex chromosome composition. As a result, a chromosomal male (XY) that carries the fruF allele expresses dsxM and fruF (brown in Figure 3B), whereas a chromosomal female (XX) that carries the fruM allele expresses dsxF and fruM (blue in Figure 3B), dissociating the sex-specific isoforms of dsx and fru. Because the dsx isoform determines almost all somatic gender characteristics (Baker and Ridge, 1980; Nagoshi and Baker, 1990), chromosomally male fruF-bearing gynandromorphs (‘fruF males’) appear anatomically male, whereas chromosomally female fruM-bearing gynandromorphs (‘fruM females’) appear anatomically female (Demir and Dickson, 2005). These four genotypes exhaust all possible combinations of sex-specific dsx and fru splicing species, affording us an opportunity to address which ‘genetic sex’ (that is, sex-specific splicing of dsx or fru) is responsible for observed sexual dimorphism.

Figure 3 with 1 supplement see all
dsx specifies the sexual dimorphism of NP2631 ∩ dsxFLP neurons.

(A) Schematics of the sex-determination pathway in Drosophila. (B) Schematic of the four sex genotypes defined by dsx and fru splicing, and how NP2631 ∩ dsxFLP neurons are specified in each genotype (see following panels for details). (C–F) Expression of CsChrimson:tdTomato under the control of NP2631 and dsxFLP (red in C1–F1, black in C2,3-F2,3) in brains of a male (C), fruM female (D), fruF male (E), and female (F) is visualized together with a neuropil marker BRP (blue) by immunohistochemistry. Circle: soma (right cluster is enlarged in C3–F3). Scale bar: 100 μm (C1–F1), 10 μm (C3–F3). (G) Mean number of cell bodies per hemibrain visualized by anti-DsRed antibody in each genotype represented in C–F) and Figure 3—figure supplement 1A–D. (H–K) Z-projection of the registered and averaged images of CsChrimson:tdTomato expression under the control of NP2631 and dsxFLP (black in H1–K1) in male (H), fruM female (I), fruF male (J), and female (K). A part of the standard Drosophila brain is shown in gray in H2–K2). Number of used hemibrains are indicated in H2–K2). Lateral junction (orange), lateral vertical projection (green), and superior medial projection (magenta) are segmented and overlaid in H2-K2. L-M: Boxplot of volumes of lateral junction (L), lateral vertical projection (M), and superior medial projections (N). Genotypes and number of hemibrains are indicated below the plot. **p<0.01, n.s. p>0.05 (Kruskal-Wallis one-way ANOVA and post-hoc Wilcoxon signed rank test).

Sexual dimorphism of NP2631 ∩ dsxFLP neurons has not been characterized in detail (Koganezawa et al., 2016). We observed that more NP2631 ∩ dsxFLP neurons are labeled in males than in females (Figure 3G). NP2631 ∩ dsxFLP neurons in both sexes extend a single tract that innervates the so-called lateral protocerebral complex (Yu et al., 2010) including commissural projections at the superior dorsal protocerebrum (Figure 3C2, F2). To better visualize the population-level sex differences in neuroanatomy, we registered the z-stack images of immunohistochemically labeled brains into a standard unisex Drosophila brain template (Bogovic et al., 2018). Visual inspection of 3D-reconstructed standardized neurons and quantification of specific neuronal segments revealed several notable sexual dimorphisms. Firstly, male neurons have thicker projections at the lateral junctions (orange in Figure 3H,K; Figure 3LYu et al., 2010). Secondly, male neurons consistently have a lateral segment branching within the ‘ring’ projection (Yu et al., 2010) (‘lateral vertical projection’) that female neurons almost completely lack (green in Figure 3H,K; Figure 3M). Thirdly, female neurons have more extensive projections at the superior medial part of the brain (‘superior medial projection’) (magenta in Figure 3H,K; Figure 3N) (see also Video 3). Thus, NP2631 ∩ dsxFLP neurons are sexually dimorphic.

We next asked whether dsx or fru is responsible for this sexual dimorphism. In fruF males, NP2631 ∩ dsxFLP neurons (Figure 3E) are similar to male neurons in cell body number and all the three sexually dimorphic neuroanatomical characteristics listed above (Figure 3G,J,L–N (brown); see also Video 4), while NP2631 ∩ dsxFLP neurons in fruM females (Figure 3D) are similar to those of females in cell body number and two of the three characteristics (Figure 3G,I,L–M (blue); see also Video 5). The only exception was the size of superior medial projection, which in this genotype was comparable to that in males (Figure 3N). These observations suggest that morphological sexual dimorphisms of NP2631 ∩ dsxFLP neurons are established predominantly by the sexual dimorphism of dsx. Complete transformation of this class of neurons to the female type may also require the cooperative contribution of fru. In addition, we cannot exclude the possibility that subset-specific neuroanatomical changes in either fruF males or fruM females escaped our detection.

Video 4
3D-rendered average image of NP2631 ∩ dsxFLP neurons in male (green; duplication from Video 3) and in fruF male (yellow).

Gray represents a standard unisex Drosophila brain (Bogovic et al., 2018).

Video 5
3D-rendered average image of NP2631 ∩ dsxFLP neurons in female (magenta, duplication from Video 3) and in fruM female (blue).

Gray represents a standard unisex Drosophila brain (Bogovic et al., 2018).

We used the fruM and fruF alleles in trans with a male isoform-specific deletion allele, fru4-40, following previous studies (Rideout et al., 2007; Vrontou et al., 2006). This configuration was necessary to eliminate potential confounds by incomplete dominance of the fruM allele, which cannot be made homozygous because females that carry the fruM allele do not mate (see also Wohl et al., 2020 for discussion regarding incomplete dominance of the fruM allele). Presence of the fru4-40 allele by itself did not affect overall sexually dimorphic characteristics of NP2631 ∩ dsxFLP neurons (Figure 3G, Figure 3—figure supplement 1A–H). Activation of this population induced qualitatively similar behavioral changes in males with the fru locus of +/+, +/fru4-40, or fruM/fru4-40 (Figure 3—figure supplement 1I).

Activation of NP2631 ∩ dsxFLP neurons in the absence of fruM induces courtship

The presence of male-like NP2631 ∩ dsxFLP neurons in fruF males raises a question about the behavioral role of these neurons in this genotype. fruF males are defective in enhancing courtship specifically toward conspecific females, but they are still capable of executing courtship actions (Demir and Dickson, 2005; Fan et al., 2013; Pan and Baker, 2014; Shirangi et al., 2006; Villella et al., 1997). These observations suggest that the courtship deficits in fruF males are not necessarily due to the absence of the courtship execution mechanism, but specifically in the mechanism to recognize proper targets for courtship. Importantly, activation of the entire dsx-expressing neurons in fruF males increases courtship behavior (Pan et al., 2011), suggesting that neural substrates for the courtship execution mechanism are present in fruF males. We asked whether NP2631 ∩ dsxFLP neurons in fruF males might be a part of this mechanism.

Optogenetic activation of NP2631 ∩ dsxFLP neurons in fruF males increased both time orienting (Figure 4A1, B1) and wing extensions (Figure 4A3, B3), toward both males and females (Figure 4—figure supplement 1A,B; see also Video 6). This result suggests that this neuronal population is indeed capable of generating at least an aspect of male-type courtship behavior in a fruM-independent manner. Interestingly, in fruF males, the amount of wing extensions toward male and female target flies was comparable after optogenetic activation of NP2631 ∩ dsxFLP neurons (Figure 4C, right). This is in contrast to the situation in regular males, which performed more wing extensions toward females (Figure 4C, left). Nonetheless, optogenetic activation of NP2631 ∩ dsxFLP neurons in regular males and in fruF males induced a similar amount of wing extensions toward male targets (Figure 4C), suggesting that the lack of fruM specifically abrogates tester flies’ capacity to enhance the intensity of courtship behavior selectively toward females. fruF males do not show lunges (Lee and Hall, 2000; Vrontou et al., 2006), which suggests that the execution mechanism for male-type aggressive behavior depends on fruM. In fact, the same optogenetic manipulation induced almost no lunges toward either target sex (Figure 4A2, B2; see also Video 6). We therefore conclude that sexual dimorphism in dsx and fru genes have different impacts on the function of NP2631 ∩ dsxFLP neurons. For courtship, dsx specifies NP2631 ∩ dsxFLP neurons as an execution component, while amplification of this function specifically toward female targets is fruM-dependent (Figure 4D). On the other hand, a fruM-dependent mechanism is necessary for NP2631 ∩ dsxFLP neurons to promote male-type aggressive behaviors, even though gross neuroanatomy of this population is specified primarily by dsx.

Figure 4 with 1 supplement see all
NP2631 ∩ dsxFLP neurons in fruF males promote courtship, but not aggressive, behavior.

(A, B) Boxplots of time orienting (A1, B1), lunges (A2, B2), and wing extension (A3, B3) by the tester flies during the time windows 1–4 (see Figure 1D). Testers’ genotypes and pair numbers are indicated below the plots. Gray lines represent single testers. Target flies are either group-housed wild-type males (A) or mated females (B). In gray: **p<0.01, n.s. p>0.05 (Kruskal-Wallis one-way ANOVA and post-hoc Wilcoxon signed rank test). In black: **p<0.01, n.s. p>0.05 (Kruskal-Wallis one-way ANOVA and post-hoc Mann-Whitney U-test). (C) Comparison of wing extension performed by flies that express CsChrimson:tdTomato under the control of NP2631 and dsxFLP in males (data from Figure 3—figure supplement 1I) or in fruF males (data from A, (B), during the time window 4. Sex of tester and target flies is indicated above the panels. Number of pairs tested is indicated below the panels. **p<0.01, n.s. p>0.05 (Mann-Whitney U-test). (D) Models of the function of NP2631 ∩ dsxFLP neurons in the context of dsx and fru.

Video 6
Representative behavior of a fruF male tester fly that expresses.

CsChrimson:tdTomato under the control of NP2631 and dsxFLP toward a wild-type male (Part 1) or a wild-type female (Part 2) target fly, at the onset and offset of LED stimulation.

P1a neurons are morphologically and functionally distinct from NP2631 ∩ dsxFLP neurons

A P1/pC1 subset that promotes both courtship and aggressive behaviors is also labeled by the P1a split GAL4 line (Hoopfer et al., 2015; Inagaki et al., 2014), which is the genetic intersection of two promoter fragments R15A01 and R71G01. Interestingly, we found that LexA lines made from these two promoters in the male brains labeled largely separate populations from NP2631 ∩ dsxFLP neurons (Figure 5A–D). Projections of standardized P1a (Figure 5E) and NP2631 ∩ dsxFLP neurons (Figure 5F) only partially overlap (Figure 5G; see also Video 7). Moreover, the temporal dynamics of behaviors induced by optogenetic activation of NP2631 ∩ dsxFLP and P1a neurons are different. Activation of NP2631 ∩ dsxFLP neurons increased both lunges and wing extensions mostly during LED stimulations, coinciding with an increase in time orienting toward target flies (Figure 2A,B, Figure 2—figure supplement 1C). In contrast, stimulation of P1a neurons increased lunges toward males and wing extensions toward females mostly during ISIs (Figure 5—figure supplement 1A2, B3) (Hoopfer et al., 2015; Jung et al., 2020). The increase of wing extensions in the presence of male target flies, which is largely confined to LED stimulations, was not accompanied by the increase of time orienting (Figure 5—figure supplement 1A1, A3). In addition, while we found that all P1a neurons are immunoreactive to Dsx (Figure 5—figure supplement 1C,D), intersectional combination of dsxFLP and P1a-GAL4 labeled almost no neurons (Figure 5—figure supplement 2A). This can be because of the low efficacy of dsxFLP in some dsx-expressing cells. Regardless, these differences further support the idea that NP2631 ∩ dsxFLP and P1a neurons contain non-overlapping and functionally distinct populations.

Figure 5 with 3 supplements see all
P1a neurons are distinct from NP2631 ∩ dsxFLP neurons.

(A, C) Expression of CsChrimson:tdTomato under the control of NP2631 and dsxFLP (yellow), GCaMP6f under the control of R15A01-LexA (green in A) or R71G01-LexA (green in C), and a neuropil marker BRP (blue) in a representative male brain are visualized by immunohistochemistry. Scale bar: 100 μm. An inset represents a magnified view of the posterior cell body cluster in the white rectangle (scale bar: 10 μm). An open arrowhead in C) indicates a neural tract from LexA-expressing neurons that appear distinct from the tract from neurons labeled by NP2631 and dsxFLP (white arrowhead). (B, D) Mean number of cell bodies per hemibrain with immunohistochemical signal by anti-DsRed antibody (orange), anti-GFP antibody (green), and both antibodies (purple) in brains of the genotype represented in A (B) or in C (D). In D, all LexA-expressing neurons located near NP2631 and dsxFLP neurons are included, although some may belong to different neuronal clusters. Error bars: S.D. Gray lines represent single hemibrains. (E–G) Z-projection of the registered and averaged images of CsChrimson:tdTomato expression under the control of P1a-GAL4 (cyan in E1, G, black in E2), and NP2631 and dsxFLP (orange in F1, G, black in F2; duplication from Figure 3H) neurons. A part of the standard Drosophila brain is shown in gray. Number of hemibrains used are indicated in E2 and F2.

Video 7
3D-rendered average image of P1a neurons (cyan) and NP2631 ∩ dsxFLP neurons (orange) in male.

Gray represents a standard unisex Drosophila brain (Bogovic et al., 2018).

Optogenetic stimulation of the fruFLP-expressing subpopulation within P1a neurons (R15A01R71G01fruFLP ‘triple intersection’; Figure 5—figure supplement 2A) promoted both lunges and wing extensions toward male target flies (Figure 5—figure supplement 2B), similar to the entire P1a population (Figure 5—figure supplement 1A). Previously, the artificial activation of R15A01fruFLP and R71G01fruFLP neurons was reported to promote only lunges (Hoopfer et al., 2015; Watanabe et al., 2017). However, P1a neurons are also known to alter the relative ratio of courtship and aggressive behaviors depending on the activation level (Hoopfer et al., 2015), leaving a possibility that a difference in activation strength due to different stimulation conditions or genetic reagents used may account for this perceived discrepancy. Consistent with the result of the entire P1a population, the P1a ∩ fruFLP neurons are morphologically distinct from NP2631 ∩ dsxFLP neurons (Figure 5—figure supplement 2C–E).

In addition, we found that R15A01 and R71G01 promoters label largely non-overlapping neurons with another P1/pC1 subset NP2631 ∩ fruFLP neurons (von Philipsborn et al., 2011; Yu et al., 2010Figure 5—figure supplement 3A–D), underscoring the potential diversity among P1/pC1 neurons (Costa et al., 2016; Zhang et al., 2018; see also Discussions).

Together, our results suggest that P1a neurons contain a subset of dsx and fru-co-expressing P1/pC1 neurons that are morphologically and functionally distinct from NP2631 ∩ dsxFLP neurons. While we do not exclude the possibility that behavior-specific subsets exist within each population, we conclude that activation of distinct dsx and fru-co-expressing P1/pC1 subsets can promote courtship and aggression with different dynamics. Activation of these neurons does not necessarily serve as a rigid command for specific behaviors, but instead is modulated by another mechanism that enhances courtship or aggression according to the sex of target flies.

Cooperative roles of dsx and fru on the specification of P1a neurons

P1a neurons appeared only in male flies (Hoopfer et al., 2015Figure 6A,D). The lack of labeling in females likely reflects the absence of P1a neurons, since dsx and fru-co-expressing neurons in the posterior part of the brain are known to undergo DsxF-dependent apoptosis (Kimura et al., 2008; Sanders and Arbeitman, 2008). If this is the case, we expect that P1a neurons are specified in fruF males (which express dsxM), but not in fruM females (which express dsxF), similar to male-type NP2631 ∩ dsxFLP neurons.

Figure 6 with 2 supplements see all
Both dsx and fru specify the sexual dimorphism of P1a neurons.

(A–D) Expression of CsChrimson:tdTomato under the control of P1a-GAL4 (red in A1–D1, black in A2,3-D2,3) in brains of a male (A), fruM female (B), fruF male (C), and female (D) is visualized together with a neuropil marker BRP (blue) by immunohistochemistry. Circle: soma (right cluster is enlarged in A3 and C3), red asterisk: sex-invariant background labeling (see Materials and methods for details). Scale bar: 100 μm (A1–D1), 10 μm (A3, C3). (E) Mean number of cell bodies per hemibrain visualized by anti-DsRed antibody in each genotype represented in A–D) and Figure 6—figure supplement 1A–D. (F, G) Z-projection of the registered and averaged images of CsChrimson:tdTomato expression under the control of P1a-GAL4 in male (F) and fruF male (G). A part of the standard Drosophila brain is shown in gray in F2, G2. Number of hemibrains used are indicated in F2) and G2). Lateral junction (cyan) and dorsal posterior projection (yellow) are segmented and overlaid in F2) and G2. H, I. Boxplot of volumes of lateral junction (H) and dorsal posterior projection (I). Genotypes and number of hemibrains are indicated below the plot. **p<0.01, n.s. p>0.05 (Kruskal-Wallis one-way ANOVA and post-hoc Mann-Whitney U-test).

Indeed, we found P1a neurons in fruF males (Figure 6C), but not in fruM females (Figure 6B). The number of P1a cell bodies was comparable between males and fruF males (Figure 6E). Interestingly, registered P1a neurons in male and fruF male brains showed quantitative differences in their neuroanatomy. Most notably, innervations of P1a neurons in fruF males were generally thinner. We specifically found that the volume of projections at the lateral junction was smaller in fruF males than in males (Figure 6F,G,H). In contrast, neural processes emanating from the posterior side of the lateral junction (‘dorsal posterior projection’) was more prominent in fruF males than in males (Figure 6F,G,I; see also Video 8). These observations indicate that sexually dimorphic characteristics of the P1a neurons are specified by both dsx and fru. Such cooperative roles of dsx and fru are reminiscent of the previous report on the roles of the two genes in a fru-expressing P1/pC1 neuronal cluster labeled by different genetic reagents (Kimura et al., 2008).

Video 8
3D-rendered average image of P1a neurons in male (green) and in fruF male (yellow).

Gray represents a standard unisex Drosophila brain (Bogovic et al., 2018).

We also asked whether P1a neurons in fruF males can influence male-type courtship or aggressive behaviors. Similar to the case in NP2631 ∩ dsxFLP neurons, presence of the fru4-40 allele did not affect the overall morphology of P1a neurons (Figure 6E, Figure 6—figure supplement 1A–F) or the behavioral effect of optogenetic stimulations (Figure 6—figure supplement 1G), although lunges toward males were not as robustly induced in the presence of the fru4-40 allele. Optogenetic activation of P1a neurons in fruF males robustly increased time orienting toward both male and female target flies (Figure 6—figure supplement 2A1, B1), but its capacity to promote wing extensions was severely impaired. We observed sporadic wing extensions during LED stimulations and ISIs, which were significantly more frequent than wing extensions performed by genetic controls (Figure 6—figure supplement 2A3, B3). However, the amount of optogenetically induced wing extensions was significantly less than what P1a neurons in normal males induced toward male targets (Figure 6—figure supplement 2C), which is in contrast to the situation in male NP2631 ∩ dsxFLP neurons.

Similar to NP2631 ∩ dsxFLP neurons, optogenetic activation of P1a neurons in fruF males induced virtually no lunges (Figure 6—figure supplement 2A2, B2). Although it is possible that a cell-specific rescue of fruM within P1a neurons in otherwise fruF males may restore their capacity to promote both courtship and aggression, we have evidence suggesting that the male type aggression-promoting function of both NP2631 ∩ dsxFLP and P1a neurons requires other fruM-expressing neurons (see discussion and Wohl et al., 2020 for details).

Overall, these results demonstrate that NP2631 ∩ dsxFLP and P1a neurons are distinct not only in neuroanatomy and behavioral roles, but also in the genetic specification mechanisms. Genetic and functional diversity within the dsx and fru-co-expressing P1/pC1 cluster suggests that Drosophila social behavior can be tuned at a multitude of circuit nodes. Our result also underscores the importance of the precise identification of cell types at the focus of a study.

Discussion

In this study, we uncovered distinct yet cooperative roles of dsx and fru on male-type social behaviors through a specific subset of P1/pC1 neurons. For courtship behaviors, we found that NP2631 ∩ dsxFLP neurons are specified in a fru-independent manner, and in males, their capacity to generate courtship behaviors does not require fruM. However, activation of NP2631 ∩ dsxFLP neurons in fruF males failed to increase courtship selectively toward female targets. These results suggest that dsx plays a major role in establishing a neuronal circuit that enables the male flies to execute courtship behavior, whereas fru is critical for enhancing courtship behavior toward females, likely through proper recognition of target sex. The fact that the sex of the target flies influences the function of P1/pC1 subsets implies that information about target sex can modulate the neural circuit units downstream of these neurons, and encourages us to revise the linear circuit model for sexually dimorphic social behaviors. In contrast, the complete specification and courtship-promoting functions of P1a neurons require both dsx and fru, revealing genetic and functional heterogeneity within P1/pC1 neurons. Lastly, NP2631 ∩ dsxFLP and P1a neurons require a fruM-dependent mechanism to promote male-type aggressive behavior. This suggests that neither of these neurons are part of the execution mechanism for male-type aggressive behavior, and that the genetic mechanisms specifying execution components for courtship and aggressive behaviors are different.

Sex of the target animals is an important biological variable

Electrical stimulation of various parts of the brain has been known to elicit complex behaviors, including social behaviors, for almost a century (Koolhaas, 1978; Miller, 1957). Recent technological advances have allowed researchers to identify specific, genetically labeled populations of neurons that can induce mating and aggressive behaviors upon acute optogenetic stimulation in both mice and in flies (Anderson, 2016; Bayless and Shah, 2016; Chen and Hong, 2018; Li and Dulac, 2018), even toward suboptimal targets (such as inanimate objects) (Asahina et al., 2014; Duistermars et al., 2018; Hong et al., 2014; Lin et al., 2011; Pan et al., 2012). These findings seem consistent with the idea that neuronal activation can override most contexts and generate specific behaviors depending on the identity of stimulated cells. However, interactions with a target animal can transmit important information which a tester animal may use to choose appropriate behaviors (Chen and Hong, 2018). In fact, attacks triggered by optogenetic stimulation of ventrolateral hypothalamus (VMH) in male mice tend to last longer toward castrated males than toward female targets (Lin et al., 2011), and chemogenetic activation of progesterone receptor-expressing VMH neurons appears to induce more attacks toward male than toward female targets (Yang et al., 2017). While effects on target sex are not consistently documented, our results and above-mentioned previous observations in mice show that the target sex has a significant impact on behavioral choice even for optogenetically induced social behaviors. These results suggest that sensory or behavioral feedback from target animals can impact the operation of what may appear to be an ‘execution mechanism’ for a given behavior.

Identification of neural sites where the information about the target sex is integrated with the activity of both NP2631 ∩ dsxFLP and P1a neurons will be an important step in understanding how such context cues modulate ongoing neural activity and, ultimately, behavioral outcome. While a ‘command’-like center that irreversibly executes courtship or aggressive behaviors, like recently characterized egg-laying controlling neurons (Wang et al., 2020), may exist, it is also possible that information about target sex (and its behavioral response) can be injected at multiple levels of a neural circuit, thereby ensuring the target sex-specific execution of sexually dimorphic social behaviors. This is conceptually analogous to the neural control of fine motions, which can be constantly adjusted by sensory feedback and efference copies all the way down to the motoneuron level (Azim and Seki, 2019).

Recently, the importance of addressing sex as a biological variable has been widely recognized (Klein et al., 2015; McCarthy et al., 2017). In the context of social behaviors, this variable in the tester animals can be critical for uncovering the underlying neural mechanisms.

Organizational function of dsx on Drosophila courtship behavior

The functional segregation of dsx and fru that we observed can be considered analogous to the organizational and activation functions of sex hormones in mammals (McCarthy, 2008). Differential exposure to gonadal steroid hormones, mostly through estrogen receptors (Wu et al., 2009), specifies neural circuits that are necessary for sex-specific reproductive behaviors, whereas hormonal surges in the adult stage (such as testosterone (Juntti et al., 2010) or progesterone Dey et al., 2015; Inoue et al., 2019) orchestrate activation of sex-specific behaviors. We postulate that dsx has an organizational function for the courtship execution circuit, whereas fru is important for the appropriate activation of the circuit.

Our results do not mean that fru is not necessary for the establishment of all neuronal components involved in courtship (see below). Nonetheless, our result suggests that the wing extension execution circuit that connects NP2631 ∩ dsxFLP neurons and relevant motoneurons is specified even in the absence of fruM (Pan et al., 2011), which is consistent with previous observations that fruF males are capable of expressing at least a part of courtship behavior (Demir and Dickson, 2005; Hall, 1978; Villella et al., 1997). While a specification role for dsx on P1/pC1 neurons has been previously reported (Kimura et al., 2008; Pan et al., 2011; Rideout et al., 2010; Sanders and Arbeitman, 2008; von Philipsborn et al., 2014), our study showed for the first time the behavioral role of a specific P1/pC1 subset (NP2631 ∩ dsxFLP neurons) in fruF males. dsx is important for the specification of a few other behaviorally relevant sexual dimorphisms in the Drosophila nervous system. For instance, the sexually dimorphic axon development of leg gustatory receptor neurons, which includes aphrodisiac pheromone sensors (Lu et al., 2012; Starostina et al., 2012; Thistle et al., 2012; Toda et al., 2012), requires dsx function (Mellert et al., 2010; Mellert et al., 2012). The neural connectivity and function of TN1 neurons (Rideout et al., 2010; Robinett et al., 2010; Shirangi et al., 2016), which are pre-motor neurons important for the production of pulse song, are also specified by dsx (Shirangi et al., 2016). Several classes of abdominal ganglia neurons involved in male copulation also express dsx (Crickmore and Vosshall, 2013; Pavlou et al., 2016). Although relatively few in number, these examples display the importance of dsx in key neuronal populations for organizing circuit components that are essential for the execution of courtship behaviors. It is noteworthy that dsx is involved in sex-determination across a variety of animal phyla (Kopp, 2012; Matson and Zarkower, 2012), whereas fru’s role in sex-determination seems confined to insects (Gailey et al., 2006). This suggests that dsx may be evolutionarily more ancient in the context of sex-determination than fru, which can account for its dominance over fru when specifying sexually dimorphic neurons that co-express dsx and fru.

Function of fru in activating courtship toward female targets

Our proposal that fruM may be important for enhancing courtship behavior specifically towards females is consistent with the fact that many characterized fru-expressing neurons are involved in processing sex- and species-specific sensory cues (Clowney et al., 2015; Fan et al., 2013; Kallman et al., 2015; Kohl et al., 2013; Kurtovic et al., 2007; Lin et al., 2016; Lu et al., 2012; Ribeiro et al., 2018; Ruta et al., 2010; Starostina et al., 2012; Stockinger et al., 2005; Thistle et al., 2012; Toda et al., 2012; Yu et al., 2010; Zhou et al., 2015). Namely, P1a neurons, as well as more broadly defined P1/pC1 neurons accessed by different genetic reagents, are known to respond to sex-specific chemical cues (Clowney et al., 2015; Kallman et al., 2015; Kohatsu et al., 2011), underscoring their critical role in sensory integration for courtship (Auer and Benton, 2016; Ellendersen and von Philipsborn, 2017). fruM can play the ‘activation’ role for courtship by establishing sensory circuits that transmit sex-specific sensory information to P1/pC1 neurons (Ito et al., 2012; Ito et al., 2016; Kimura et al., 2005; von Philipsborn et al., 2014), or by enabling P1/pC1 neurons to properly integrate and transform such neural inputs. Neuroanatomical defects of P1a neurons in fruF males could disrupt either process.

Gain control of sex-specific sensory cues can be one neuronal mechanism for the ‘activation’ function, but courtship behavior can be enhanced in other ways as well. For instance, behavioral persistence or context-dependent intensity adjustment (Clemens et al., 2018; Coen et al., 2016; Grosjean et al., 2011; Keleman et al., 2012; Zhang et al., 2016) can result in an increase of the overall courtship vigor. Recently, a new class of fru-expressing neurons downstream of P1a neurons has been found to mediate the persistence of courtship behavior triggered by P1a neuronal activation (Jung et al., 2020). Even if fruM is not absolutely necessary for the formation of the minimal wing extension execution circuit, it can have a significant impact on the generation of effective wing extension toward female target flies (Fan et al., 2013; Pan and Baker, 2014; Pan et al., 2011).

While we conclude that the role of fru is not necessarily to specify the execution mechanism for courtship behavior, fruM females can still perform wing extensions (Demir and Dickson, 2005; Rideout et al., 2007). Moreover, artificial stimulation of either fru- (Clyne and Miesenböck, 2008; Pan et al., 2011) or dsx- (Rezával et al., 2016) expressing neurons in females can elicit wing extensions, suggesting that the residual execution mechanism for at least a part of courtship behavior may be specified in a sex-invariant manner. The presence of a latent mating execution circuit in female brains is also suggested in mice (Kimchi et al., 2007; Li and Dulac, 2018). Because the courtship songs produced by females or fruM females are defective (Pan et al., 2011; Rezával et al., 2016; Rideout et al., 2007), male-type splicing of dsx nonetheless seems to be instrumental in organizing the proper execution mechanism for Drosophila courtship behavior.

In striking contrast to wing extensions, we found that activation of neither NP2631 ∩ dsxFLP nor P1a neurons in fruF males induced lunges. This result points to the existence of a fruM-dependent execution mechanism for male-type aggressive behaviors (Vrontou et al., 2006), likely downstream of these neurons. In Wohl et al. (2020), we found that at least one group of fruM-dependent neurons can promote male-type aggressive behaviors independent of dsx. Therefore, a separation of the courtship execution mechanism and the aggression execution mechanism by two sex-determining genes is likely accomplished by a partial separation of underlying neural circuits. The aggression-promoting function of NP2631 ∩ dsxFLP and P1a neurons likely reflects their roles to coordinate aggression and courtship depending on internal and external conditions, instead of a simple decision switch that triggers fixed types of behavior.

An organismal sex and a cellular sex

Both dsx and fru encode transcription factors. The sexually dimorphic morphology and wiring specificity of many fru-expressing neurons are determined in a cell-autonomous manner (Kimura et al., 2008; Kohl et al., 2013), suggesting that dsx and fru define Drosophila sex at a cellular level through regulation of a specific set of target genes (Dalton et al., 2013; Ito et al., 2012; Neville et al., 2014; Vernes, 2015). Mammalian sex hormones ultimately exert their effects through nuclear steroid receptors, which serve as transcription factors. Thus, both in flies and in mammals, organismal sex can be regarded as a collective phenotype of genetic ‘sexes’ that can be reduced down to the cellular level (Robinett et al., 2010; Williams and Carroll, 2009). To understand how sex at the neuronal level influences sexually dimorphic behaviors, cell-type specific manipulation of sex-determining genes is required. Our current study focused on neural functions in a whole animal mutant, which prevents us from addressing the role of either dsx or fru specifically within NP2631 ∩ dsxFLP or P1a neurons. For example, we do not know whether NP2631 ∩ dsxFLP neurons in fruF males failed to enhance courtship behavior toward female targets because of the absence of fruM within this population, or because of the lack of fruM in other neuronal populations, or both. In addition, our four genotypes approach does not address if it is the presence of dsxM or the absence of dsxF that is important for the specification of male-type NP2631 ∩ dsxFLP neurons or P1a neurons.

It is important to note that sex specification is a developmental process of transformation. Both at genetic and organismal levels, one sex is not a loss-of-function mutant of the other. Loss-of-function manipulations at the cellular level, by cell type-specific RNA interference (Dietzl et al., 2007; Ni et al., 2009) or CRISPR interference (Qi et al., 2013)-based approaches, may show that either dsx or fru is necessary for the proper development or function of the given neurons, but may be insufficient to illuminate the genetic origin of the sex-specific transformation at the cellular level. In addition, temporally and spatially precise manipulation of genes during development remains difficult. This can create a difficulty interpreting the effects of either knock-down or over-expression of sex determining genes, which are dynamically regulated from early developmental stages (Baker and Ridge, 1980; Lee et al., 2000; Lee et al., 2002; Mellert et al., 2012; Sanders and Arbeitman, 2008). Creation of neuronal mutant clones (Lee and Luo, 1999) may circumvent this problem, but the tra mutation (Kimura et al., 2008; Kohl et al., 2013), which has been previously used to convert a ‘neuronal sex’, cannot dissociate the roles of dsx and fru (see Figure 3A).

Faced with these often overlooked limitations of cell-type specific gene manipulations, it would be informative to characterize what types of transformations are observed in mutants of sex-specific splicing at an organismal level, as in this and other studies (Datta et al., 2008; Kimura et al., 2008; Kohl et al., 2013; von Philipsborn et al., 2014; Zhou et al., 2015). Although a constitutive mutants have above-mentioned limitations, they nonetheless establish fundamental functional differences among sex-determining genes, as well as benchmarks for the efficacy for cell-specific manipulations techniques. Although clearly out of the scope of the current study, electron microscopy-based connectome reconstructions of fruF male and fruM female brains (Zheng et al., 2018) could provide useful information for understanding the transformative nature of sex specification in the brain.

Uncovering functional heterogeneity of social behavior-controlling neurons

Lastly, our serendipitous finding that NP2631 ∩ dsxFLP and P1a neurons contain genetically and functionally distinct populations underscores the importance of characterizing neuronal cell types in greater detail. How to determine cell types remains a challenge in neuroscience, but genetic access to a finely defined population of neurons even within what is considered as a single class of neurons can be the key to understand how a neural circuit generates complex behaviors such as social behaviors (Luo et al., 2018).

In the posterior part of male brains, ‘P1’ neurons, as defined by fru-expressing cluster (Kimura et al., 2008; Lee et al., 2000), and pC1 neurons, as defined by dsx-expressing cluster (Lee et al., 2002; Rideout et al., 2010), extensively overlap (Rideout et al., 2007; Rideout et al., 2010; Sanders and Arbeitman, 2008; Zhou et al., 2014) This raises a question about the distinction between ‘P1’ and ‘pC1’ neurons (Asahina, 2018). Furthermore, recent single cell level analyses of the neurons that belong to the male ‘P1’ cluster or ‘pC1’ cluster revealed surprising neuroanatomical and functional diversity (Costa et al., 2016; Zhang et al., 2018), raising a possibility that P1/pC1 neurons may be functionally heterogeneous as well (Asahina, 2018).

Surprisingly, we found that behaviorally relevant NP2631 ∩ dsxFLP and P1a neurons, as well as NP2631 ∩ fruFLP and P1a neurons, seldom overlap. Optogenetic stimulation of NP2631 ∩ dsxFLP and P1a neurons triggers social behaviors in temporally distinct manners. Moreover, fru has a different impact on the specification and function of these two neuron groups, suggesting that little overlap of NP2631 ∩ dsxFLP and P1a neurons does not necessarily reflect arbitrary labeling bias within a single homogeneous neuronal population by different genetic reagents. Instead, these observations support the idea that of P1/pC1 neurons consist of functionally diverse subtypes.

We acknowledge that the genetic reagents used in our study are likely insufficient to resolve the possible heterogeneity within either NP2631 ∩ dsxFLP or P1a neurons. Differential expression patterns of FruM proteins within both clusters (Figure 2—figure supplement 2C, Figure 5—figure supplement 1D) alone suggest that such heterogeneity almost certainly exists. Recent advances in whole-brain neural reconstruction using electron microscopy images (Zheng et al., 2018) will provide a foundation for precise characterization of Drosophila neurons, as has been recently used for the female-type ‘pC1’ cluster (Deutsch et al., 2020; Wang et al., 2020). A large number of ‘split-GAL4’ collections (Jenett et al., 2012; Kvon et al., 2014; Luan et al., 2006) will allow universal access to the specific subpopulations. These types of tools will facilitate cross-study comparisons of neuroanatomical and behavioral data, and will serve as a catalyst to understand the logic of neural control of behavior in general. With the advance of single cell-level genetic and epigenetic profiling techniques, the importance of precisely characterizing the targeted neuronal types will only grow not only in Drosophila, but in every model organism. Reproducible access to each neuronal type can uncover functional units for a given behavior at even finer detail (Robie et al., 2017), which will be fundamental for deconstructing the dynamics of neural circuits that are responsible for generating social behaviors in a context-dependent manner. Such knowledge will be also critical for establishing theoretical models that account for brain operations (Kingsbury et al., 2019; Zhang and Yartsev, 2019) and population-level dynamics of animals (Ramdya et al., 2017) engaging in social interactions.

Materials and methods

See Supplementary file 1 for details of reagents used in this study.

Fly strains

Request a detailed protocol

See Table 1 for the complete genotypes of Drosophila strains used in each figure panel. NP2631 (Yu et al., 2010) is a gift from Daisuke Yamamoto (Tohoku University). P1a-GAL4 (R15A01-p65AD:Zp (in attP40) (RRID:BDSC_68837); R71G01-Zp:GAL4DBD (in attP2) (RRID:BDSC_69507)) (Hoopfer et al., 2015; Inagaki et al., 2014) and R15A01-LexA (in attP2) (Hoopfer et al., 2015) were gifts from David Anderson (California Institute of Technology). 20XUAS-IVS-Syn21-CsChrimson:tdTomato (in VK00022), 20XUAS > myr:TopHAT2 >CsChrimson:tdTomato (in VK00022 and VK00005) (Duistermars et al., 2018; Watanabe et al., 2017) and 13XLexAop2-IVS-Syn21-GCaMP6f (codon-optimized)-p10 (in su(Hw)attP5) were created by Barret Pfeiffer in the lab of Gerald Rubin (HHMI Janelia Research Campus) and kindly shared by David Anderson. fruM (RRID:BDSC_66874), fruF (RRID:BDSC_66873) (Demir and Dickson, 2005), and fruFLP (RRID:BDSC_66870) (Yu et al., 2010) flies are gifts from Barry Dickson (HHMI Janelia Research Campus); dsxGAL4 (Rideout et al., 2010) and dsxFLP (Rezával et al., 2014) are gifts from Stephen Goodwin (University of Oxford); fruP1.LexA (RRID:BDSC_66698) (Mellert et al., 2010) is a gift from Bruce Baker (HHMI Janelia Research Campus). fru4-40 (RRID:BDSC_66692), and 8XLexAop2-GAL80 (in attP40) (RRID:BDSC_32214) flies were obtained from Bloomington Drosophila Resource Center in the University of Indiana.

Table 1
Complete genotypes of Drosophila strains used in this study.
FigurePanelAbbreviated
genotype
COMPLETE GENOTYPE (‘Y’
represents the Y chromosome)
Figure 1A, BdsxGAL4 ∩ fruFLPw/Y; 20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00022/+; dsxGAL4/fruFLP
Figure 1—figure supplement 1A
Video 1
Figure 1BdsxGAL4 ∩ fruFLPw/w; 20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00022/+; dsxGAL4/fruFLP
Figure 1—figure supplement 1B
Figure 1E-IdsxGAL4, UAS > stop > CsChrimson, fruFLPw/Y; 20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00022/+; dsxGAL4/fruFLP
Figure 1—figure supplement 1C, D
Figure 1E-HdsxGAL4, UAS > stop > CsChrimsonw/Y; 20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00022/+; dsxGAL4/+
Figure 1—figure supplement 1C, D
Figure 1E-HUAS > stop > CsChrimson, fruFLPw/Y; 20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00022/+; fruFLP/+
Figure 1—figure supplement 1C, D
Figure 1—figure supplement 2BCanton-S ♂+/Y; +/+; +/+ (Canton-S)
Figure 1—figure supplement 2BCanton-S ♀+/+; +/+; +/+ (Canton-S)
Figure 2A, BNP2631, UAS > stop > CsChrimson, dsxFLPw/Y; NP2631/20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00022; dsxFLP, fru4-40/+
Figure 2—figure supplement 1A-C
Figure 2—figure supplement 3A, B
Video 2
Figure 2A, BNP2631, UAS > stop > CsChrimsonw/Y; NP2631/20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00022; fru4-40/+
Figure 2—figure supplement 3A, B
Figure 2A, BUAS > stop > CsChrimson, dsxFLPw/Y; +/20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00022; dsxFLP, fru4-40/+
Figure 2—figure supplement 3A, B
Figure 2CNP2631, UAS > stop > CsChrimson, dsxFLP, fruP1. LexA, LexAop2-GAL80w/Y; NP2631/8XLexAop2-GAL80 in attP40; 20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00005, dsxFLP/fruP1. LexA
Figure 2CNP2631, UAS > stop > CsChrimson, dsxFLP, LexAop2-GAL80w/Y; NP2631/8XLexAop2-GAL80 in attP40; 20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00005, dsxFLP/+
Figure 2CNP2631, UAS > stop > CsChrimson, dsxFLP, fruP1. LexAw/Y; NP2631/+; 20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00005, dsxFLP/fruP1. LexA
Figure 2—figure supplement 1A-CNP2631, UAS > stop > CsChrimson, dsxFLPw/Y; NP2631/20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00022; dsxFLP, fru4-40/+
Figure 2—figure supplement 2A-DNP2631 ∩ dsxFLPw/Y; NP2631/20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00022; dsxFLP/+
Figure 2—figure supplement 2E, HNP2631, UAS > stop > CsChrimson (VK00005), dsxFLP, fruP1. LexA, LexAop2-GAL80w/Y; NP2631/8XLexAop2-GAL80 in attP40; 20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00005, dsxFLP/fruP1. LexA
Figure 2—figure supplement 2F, HNP2631, UAS > stop > CsChrimson (VK00005), dsxFLP, LexAop2-GAL80w/Y; NP2631/8XLexAop2-GAL80 in attP40; 20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00005, dsxFLP/+
Figure 2—figure supplement 2G, HNP2631, UAS > stop > CsChrimson (VK00005), dsxFLP, fruP1. LexAw/Y; NP2631/+; 20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00005, dsxFLP/fruP1. LexA
Figure 3C, G, H, L-NNP2631 ∩ dsxFLP, XY, fru locus: fruM/fru4-40w/Y; NP2631/20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00022; dsxFLP, fru4-40/fruM
Figure 3—figure supplement 1F
Video 3, 4
Figure 3D, G, I, L-NNP2631 ∩ dsxFLP, XX, fru locus: fruM/fru4-40w/w; NP2631/20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00022; dsxFLP, fru4-40/fruM
Video 5
Figure 3E, G, J, L-NNP2631 ∩ dsxFLP, XY, fru locus: fruF/fru4-40w/Y; NP2631/20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00022; dsxFLP, fru4-40/fruF
Video 4
Figure 3F, G, K-NNP2631 ∩ dsxFLP, XX, fru locus: fruF/fru4-40w/w; NP2631/20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00022; dsxFLP, fru4-40/fruF
Figure 3—figure supplement 1H
Video 3, 5
Figure 3GNP2631 ∩ dsxFLP, XY, fru locus: +/+w/Y; NP2631/20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00022; dsxFLP/+
Figure 3—figure supplement 1A, E
Figure 3GNP2631 ∩ dsxFLP, XX, fru locus: +/+w/w; NP2631/20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00022; dsxFLP/+
Figure 3—figure supplement 1B, G
Figure 3GNP2631 ∩ dsxFLP, XY, fru locus: +/fru4-40w/Y; NP2631/20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00022; dsxFLP, fru4-40/+
Figure 3—figure supplement 1C
Figure 3GNP2631 ∩ dsxFLP, XX, fru locus: +/fru4-40w/w; NP2631/20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00022; dsxFLP, fru4-40/+
Figure 3—figure supplement 1D
Figure 3—figure supplement 1INP2631, UAS > stop > CsChrimson, dsxFLP, fru locus: +/+w/Y; NP2631/20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00022; dsxFLP/+
Figure 3—figure supplement 1INP2631, UAS > stop > CsChrimson, dsxFLP, fru locus: +/fru4-40w/Y; NP2631/20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00022; dsxFLP, fru4-40/+
Figure 3—figure supplement 1INP2631, UAS > stop > CsChrimson, dsxFLP, fru locus: fruM/fru4-40w/Y; NP2631/20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00022; dsxFLP, fru4-40/fruM
Figure 4A-CfruF ♂, NP2631, UAS > stop > CsChrimson, dsxFLPw/Y; NP2631/20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00022; dsxFLP, fru4-40/fruF
Figure 4—figure supplement 1A, B
Video 6
Figure 4A, BfruF ♂, NP2631, UAS > stop > CsChrimsonw/Y; NP2631/20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00022; fru4-40/fruF
Figure 4A, BfruF ♂, UAS > stop > CsChrimson, dsxFLPw/Y; +/20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00022; dsxFLP, fru4-40/fruF
Figure 4C♂, NP2631, UAS > stop > CsChrimson,dsxFLPw/Y; NP2631/20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00022; dsxFLP, fru4-40/fruM
Figure 5A, BNP2631 ∩ dsxFLP, R15A01-LexAw/Y; NP2631/13XLexAop2-IVS-GCaMP6f-p10 in su(Hw)attP5; R15A01-LexA in attP2/20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00005, dsxFLP
Figure 5C, DNP2631 ∩ dsxFLP, R71G01-LexAw/Y; NP2631/13XLexAop2-IVS-GCaMP6f-p10 in su(Hw)attP5; 20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00005, dsxFLP/R71G01-LexA in attP2
Figure 5E, GP1a-GAL4w/Y; R15A01-p65AD:Zp in attP40/20XUAS-IVS-CsChrimson:tdTomato in VK00022; R71G01-Zp:GAL4DBD in attP2/+
Video 7
Figure 5F, GNP2631 ∩ dsxFLPw/Y; NP2631/20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00022; dsxFLP, fru4-40/fruM
Video 7
Figure 5—figure supplement 1AP1a-GAL4, UAS-CsChrimsonw/Y; R15A01-p65AD:Zp in attP40/20XUAS-IVS-CsChrimson:tdTomato in VK00022; R71G01-Zp:GAL4DBD in attP2/+
Figure 5—figure supplement 1AP1a-GAL4w/Y; R15A01-p65AD:Zp in attP40/20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00022; R71G01-Zp:GAL4DBD in attP2/+
Figure 5—figure supplement 1AUAS-CsChrimsonw/Y; +/20XUAS-IVS-CsChrimson:tdTomato in VK00022; +/+
Figure 5—figure supplement 1BP1a-GAL4, UAS-CsChrimsonw/Y; R15A01-p65AD:Zp in attP40/20XUAS-IVS-CsChrimson:tdTomato in VK00022; R71G01-Zp:GAL4DBD in attP2, fru4-40/fruM
Figure 5—figure supplement 1BP1a-GAL4w/Y; R15A01-p65AD:Zp in attP40/10XUAS-IVS-GFP-p10 in VK00022; R71G01-Zp:GAL4DBD in attP2, fru4-40/fruM
Figure 5—figure supplement 1BUAS-CsChrimsonw/Y; +/20XUAS-IVS-CsChrimson:tdTomato in VK00022; fru4-40/fruM
Figure 5—figure supplement 1C, DP1a-GAL4w/Y; R15A01-p65AD:Zp in attP40/20XUAS-IVS-CsChrimson:tdTomato in VK00022; R71G01-Zp:GAL4DBD in attP2/+
Figure 5—figure supplement 2AP1a-GAL4, UAS-CsChrimson (VK00022)w/Y; R15A01-p65AD:Zp in attP40/20XUAS-IVS-CsChrimson:tdTomato in VK00022; R71G01-Zp;GAL4DBD in attP2/+
Figure 5—figure supplement 2AP1a-GAL4, UAS > stop > CsChrimson (VK00022), fruFLPw/Y; R15A01-p65AD:Zp in attP40/20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00022; R71G01-Zp;GAL4DBD in attP2/fruFLP
Figure 5—figure supplement 2AP1a-GAL4, UAS > stop > CsChrimson (VK00005), fruFLPw/Y; R15A01-p65AD:Zp in attP40/+; 71G01-Zp;GAL4DBD in attP2/20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00005, fruFLP
Figure 5—figure supplement 2AP1a-GAL4, UAS > stop > CsChrimson (VK00005), dsxFLPw/Y; R15A01-p65AD:Zp in attP40/+; 71G01-Zp;GAL4DBD in attP2/20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00005, dsxFLP
Figure 5—figure supplement 2C, EP1a-GAL4 ∩ fruFLPw/Y; R15A01-p65AD:Zp in attP40/+; 71G01-Zp;GAL4DBD in attP2/20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00005, fruFLP
Figure 5—figure supplement 2D, ENP2631 ∩ dsxFLPw/Y; NP2631/20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00022; dsxFLP/+
Figure 5—figure supplement 2BP1a-GAL4,UAS > stop > CsChrimson, fruFLPw/Y; R15A01-p65AD:Zp in attP40/+;R71G01-Zp:GAL4DBD in attP2/20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00005, fruFLP
Figure 5—figure supplement 2BP1a-GAL4, UAS > stop > CsChrimsonw/Y; R15A01-p65AD:Zp in attP40/+; R71G01-Zp;GaL4DBD in attP2/20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00005
Figure 5—figure supplement 2BUAS > stop > CsChrimson,fruFLPw/Y; R15A01-p65AD:Zp in attP40/+; +/20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00005, fruFLP
Figure 5—figure supplement 3A, BNP2631 ∩ fruFLP, R15A01-LexAw/Y; NP2631/13XLexAop2-IVS-GCaMP6f-p10 in su(Hw)attP5; R15A01-LexA in attP2/20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00005, fruFLP
Figure 5—figure supplement 3C, DNP2631 ∩ fruFLP, R71G01-LexAw/Y; NP2631/13XLexAop2-IVS-GCaMP6f-p10 in su(Hw)attP5; 20XUAS > myr:TopHAT2 > CsChrimson:tdTomato in VK00005, fruFLP/R71G01-LexA in attP2
Figure 6A, E, F, H, IP1a-GAL4, XY, fru locus: fruM/fru4-40w/Y; R15A01-p65AD:Zp in attP40/20XUAS-IVS-CsChrimson:tdTomato in VK00022; R71G01-Zp:GAL4DBD in attP2, fru4-40/fruM
Figure 6—figure supplement 1F
Figure 6B, EP1a-GAL4, XX, fru locus:fruM/fru4-40w/w; R15A01-p65AD:Zp in attP40/20XUAS-IVS-CsChrimson:tdTomato in VK00022; R71G01-Zp:GAL4DBD in attP2, fru4-40/fruM
Figure 6C, E, G-IP1a-GAL4, XY, fru locus: fruF/fru4-40w/Y; R15A01-p65AD:Zp in attP40/20XUAS-IVS-CsChrimson:tdTomato in VK00022; R71G01-Zp:GAL4DBD in attP2, fru4-40/fruF
Figure 6D, EP1a-GAL4, XX, fru locus:fruF/fru4-40w/w; R15A01-p65AD:Zp in attP40/20XUAS-IVS-CsChrimson:tdTomato in VK00022; R71G01-Zp:GAL4DBD in attP2, fru4-40/fruF
Figure 6E, H, IP1a-GAL4, XY, fru locus: +/+w/Y; R15A01-p65AD:Zp in attP40/20XUAS-IVS-CsChrimson:tdTomato in VK00022; R71G01-Zp:GAL4DBD in attP2/+
Figure 6—figure supplement 1A, E
Figure 6EP1a-GAL4, XX, fru locus: +/+w/w; R15A01-p65AD:Zp in attP40/20XUAS-IVS-CsChrimson:tdTomato in VK00022; R71G01-Zp:GAL4DBD in attP2/+
Figure 6—figure supplement 1B
Figure 6EP1a-GAL4, XY, fru locus: +/fru4-40w/Y; R15A01-p65AD:Zp in attP40/20XUAS-IVS-CsChrimson:tdTomato in VK00022; R71G01-Zp:GAL4DBD in attP2, fru4-40/+
Figure 6—figure supplement 1C
Figure 6EP1a-GAL4, XX, fru locus: +/fru4-40w/w; R15A01-p65AD:Zp in attP40/20XUAS-IVS-CsChrimson:tdTomato in VK00022; R71G01-Zp:GAL4DBD in attP2, fru4-40/+
Figure 6—figure supplement 1D
Figure 6—figure supplement 1GP1a-GAL4, UAS-CsChrimson, fru locus: +/+w/Y; R15A01-p65AD:Zp in attP40/20XUAS-IVS-CsChrimson:tdTomato in VK00022; R71G01-Zp:GAL4DBD in attP2, +/+
Figure 6—figure supplement 1GP1a-GAL4, UAS-CsChrimson, fru locus: +/fru4-40w/Y; R15A01-p65AD:Zp in attP40/20XUAS-IVS-CsChrimson:tdTomato in VK00022; R71G01-Zp:GAL4DBD in attP2, fru4-40/+
Figure 6—figure supplement 1GP1a-GAL4, UAS-CsChrimson, fru locus: fruM/fru4-40w/Y; R15A01-p65AD:Zp in attP40/20XUAS-IVS-CsChrimson:tdTomato in VK00022; R71G01-Zp:GAL4DBD in attP2, fru4-40/fruF
Figure 6—figure supplement 2A-CfruF ♂, P1a-GAL4, UAS-CsChrimsonw/Y; R15A01-p65AD:Zp in attP40/20XUAS-IVS-CsChrimson:tdTomato in VK00022; R71G01-Zp:GAL4DBD in attP2, fru4-40/fruF
Figure 6—figure supplement 2AfruF ♂, P1a-GAL4w/Y; R15A01-p65AD:Zp in attP40/10XUAS-IVS-GFP-p10 in VK00022; R71G01-Zp:GAL4DBD in attP2, fru4-40/fruF
Figure 6—figure supplement 2AfruF ♂, UAS-CsChrimsonw/Y; R15A01-p65AD:Zp in attP40/20XUAS-IVS-CsChrimson:tdTomato in VK00022; fru4-40/fruF
Figure 6—figure supplement 2C♂, P1a-GAL4, UAS-CsChrimson

R71G01-LexA (in attP2) was created by targeting the GMR71G01-LexA plasmid construct (in pCR8/GW/TOPO backbone: gift from Gerald Rubin, HHMI Janelia Research Campus) into the P{CaryP}attP2 landing site. The plasmid was injected into embryos of the nos-phiC31, y1, sc1, v1, sev21; P{CaryP}attP2 (RRID:BDSC_25710) strain (BestGene Inc, Chino Hills, CA), and transformants were recovered using on the mini-white eye color marker.

Immunohistochemistry

Request a detailed protocol

The following antibodies were used for immunohistochemistry with dilution ratios as indicated: rabbit anti-DsRed (1:1,000, Clontech # 632496, RRID:AB_10013483), mouse anti-BRP (1:100; Developmental Studies Hybridoma Bank nc82 (concentrated), RRID:AB_2314866), chicken anti-GFP (1:1,000, Abcam # ab13970, RRID:AB_300798), rabbit anti-FruM (1:10,000, a gift from Barry Dickson; Stockinger et al., 2005), guinea pig anti-FruM (1:100), rat anti-DsxM (1:100) (both gifts from Michael Perry (University of California, San Diego)), goat anti-chicken Alexa 488 (1:100, ThermoFisher Scientific Cat# A11039, RRID:AB_2534096), goat anti-rat Alexa 488 (1:100, ThermoFisher Scientific Cat# A11006, RRID:AB_2534074), goat anti-rabbit Alexa 568 (1:100; ThermoFisher Scientific Cat# A11036, RRID:AB_10563566), goat anti-mouse Alexa 633 (1:100; ThermoFisher Scientific Cat# A21052, RRID:AB_2535719), goat anti-guinea pig Alexa 633 (1:100, ThermoFisher Scientific Cat# A21105, RRID:AB_2535757). Immunohistochemistry of the fly brains followed the protocol described in Van Vactor et al. (1991). Briefly, the fly brains are dissected in 1XPBS and fixed in 1XPBS with 2% formaldehyde and 75 mM L-lysine for 75–90 min at room temperature. The brains were then washed in PBST (1XPBS, 0.3% TritonX-100) and were incubated in the blocking solution (10% heat-inactivated normal goad serum, 1XPBS, 0.3% TrintonX-100) for 30 min. Primary antibodies were diluted in the blocking solution and were applied to samples, which were then incubated at 4°C for 2 days. The brains were then washed in PBST and then incubated in the blocking solution for 30 min. Secondary antibodies were diluted in the blocking solution and were applied to the samples, which were then incubated at 4°C overnight. The brains were then washed in PBST, and then either incubated in 1XPBS, 50% glycerol for 2 hr at room temperature before mounted in Vectashield (Vector Laboratories, Cat# H-1000) onto a slide glass, or incubated in FocusClear (CelExplorer Labs, Taiwan, Cat# FC-101) medium for 2 hr at room temperature before being mounted in MountClear (CelExplorer Labs, Taiwan, Cat# MC-301) medium. A small well was made by cutting vinyl tape fixed on a slide glass, and one brain was transferred to each well before a cover slip (#1.5) was placed on the well and was sealed with nail polish. All reactions were carried out in a well of 6 × 10 microwell mini tray (ThermoFisher Scientific Cat# 439225).

For simultaneous detection of DsxM and FruM (Figure 2—figure supplement 2B,C, Figure 5—figure supplement 1C,D), brains were fixed in 1XPBS with 4% formaldehyde for 15 min at room temperature. Brains were then transferred to a 1.5 mL microtube, in which the remaining steps were carried out.

All z-stack images were acquired by FV-1000 confocal microscopy (Olympus America; kindly shared by Dr. Samuel Pfaff at Salk Institute) except samples for Figure 5C,D and Figure 5—figure supplement 3C,D, which were acquired by a Zeiss 710 confocal microscopy (Carl Zeiss Microscopy) at the Salk Institute Biophotonics Core, and were processed in Fiji software (Schindelin et al., 2012) (RRID:SCR_002285; https://fiji.sc/). The despeckle function was applied before a z-projection image was generated using maximum intensity projection. Minimum or maximum intensity thresholds were adjusted for enhanced clarity. Source image files used in all figures can be found in Source data 1.

Registration and analysis of immunohistochemical samples

Request a detailed protocol

All data points for anatomical quantifications used in all figures, as well as all statistical results with exact p values, can be found in Source data 1.

Parametric tests were applied as indicated in figure legends to compare cell body numbers among different genotypes. All data points have biological replicates of at least 8, which has sufficient power to detect mean changes in cell body number of larger than 1.75 when assuming a mean cell body number of 5 and a standard deviation of 1.5 (which is reasonable based on our data). As indicated in figures, genotype-dependent changes in cell body number had a larger effect size than two in all cases.

We first split each channel of a z-stack file and resaved each as individual. nrrd files in Fiji. We then used the Fiji plugin for Computational Morphometry Tookit (CMTK) (RRID:SCR_002234; https://www.nitrc.org/projects/cmtk) (Rohlfing and Maurer, 2003) to register the brain image stacks to the template brain as described in Jefferis et al. (2007) (https://github.com/jefferis/fiji-cmtk-gui). Briefly, for each brain, the image of neuropil visualized by anti-BRP antibody was used to transform the z-stack to the template brain, and the same transformation was subsequently applied to additional channels. Registration was performed by using the same parameters implemented in ‘Cachero et al. (2010)’ (exploration = 26, coarsest = 8, grid spacing = 80, refine = 4, accuracy = 0.4) in CMTK plugin on Fiji. We used JRC2018 INTERSEX (Bogovic et al., 2018) as our template, since we experienced more robust registration result than with the FCWB template brain (Costa et al., 2016) (data not shown).

After the registration of z-stacks, we calculated the average of the images in a hemibrain for each genotype. We did this by horizontally flipping each transformed z-stack image, and calculated the average signal intensity in each voxel. The resulting averaged images are bilaterally symmetrical. Therefore, only the left hemisphere is shown for z-projection images.

To calculate the volumes of specific neuronal structures, we first visualized each z-stack in a 3D space using a rendering software FluoRender (Wan et al., 2009) (RRID:SCR_014303; https://github.com/SCIInstitute/fluorender). We then segmented the target structures using the Paint Brush function and calculated the volume of each structure using the Volume Size function (threshold = 800) (Wan et al., 2017). Statistical analyses were carried out using MATLAB (The Mathworks, Inc, RRID:SCR_001622). The Kruskal-Wallis test (‘kruskalwallis’) was used to evaluate whether a volume of the given structure was significantly different among different genotypes. When the p-value was below 0.05, the post-hoc the Mann-Whitney U-test (‘ranksum’) was used to detect significant differences between testing and control genotypes. In both cases, the Bonferroni correction was applied to p values. Non-parametric tests were applied for volume data since we could not necessarily assume a normal distribution for this data type.

Social behavior analysis

Subject preparation

Request a detailed protocol

Flies were collected on the day of eclosion into vials containing standard cornmeal-based food, and were kept either as a group of up to 16 flies per vial, or singly at 25 °C with 60% relative humidity, and a 9AM:9PM light:dark cycle. For optogenetic experiments, the tester flies were reared on food containing 0.2 mM all-trans retinal (MilliporeSigma, Cat#R2500, 20 mM stock solution prepared in 95% ethanol), and vials were covered with aluminum foil to shield light. Every 3 days, flies were transferred to vials containing fresh food. Tester flies for Figures 14 (including figure supplements) and Figure 5—figure supplements 2 and 3 were aged for 14–16 days to ensure consistent labeling of targeted neurons (data not shown). Tester flies for Figures 5 and 6, and Figure 5—figure supplement 1 were aged for 5–7 days.

Male target flies were group-reared Canton-S (originally from the lab of Martin Heisenberg, University of Würzberg) virgin males. To prepare mated wild-type target females, 5 Canton-S males were introduced into vials with 10 virgin females at 4 days old, and were reared for two more days to let them mate. At 3 days old, both male and female target flies were briefly anesthetized with CO2, and the tip of either one of their wings were clipped by a razor to create a ‘mark’. This clipping treatment did not reduce the amount of each behavior (lunge, wing extension, and headbutt) detected under our experimental settings (Figure 1—figure supplement 2B).

Behavioral assays

Request a detailed protocol

All behavior assays were conducted in the evening (from 4 to 9PM) at 22–25°C. Social behavior assays were performed in a ‘12-well’ acrylic chamber (Asahina et al., 2014) with food substrate (apple juice (Minute Maid) supplemented with 2.25% w/v agarose and 2.5% w/v sucrose; Hoyer et al., 2008) covering the entire floor of arena. The wall was coated with Insec-a-Slip (Bioquip Products, Inc, Cat# 2871C) and the ceiling was coated with Surfasil Siliconizing Fluid (ThermoFisher Scientific, Cat# TS-42800), both to prevent flies from climbing, as described previously (Asahina et al., 2014; Hoyer et al., 2008). The arenas were lit by LED backlights, which were controlled by a custom-built switch box. For optogenetic experiments, 850 nm infrared LED backlights (Sobel Imaging Systems, CA, Cat# SOBL-150 × 100–850) were used. Flies were introduced into the chamber by gentle aspiration, and were allowed to acclimate for 5 min before recording started.

Recording was done by USB3 digital cameras (Point Grey Flea3 USB3.0, FLIR Inc, Cat# FL3-U3-13Y3M-C) controlled by the BIAS acquisition software (IORodeo, CA; https://bitbucket.org/iorodeo/bias). The camera was equipped with a machine vision lens (Fujinon, Cat# HF35HA1B), and an infrared longpass filter (Midwest Optical Systems, Cat# LP780-25.5) when the infrared light sources were used. Movies were taken at 60 frames per second in the AVI format, either for 10 min in the optogenetic experiments or for 30 min for non-optogenetic experiments. Flies were discarded after each experiment. The food substrates were changed to a new one after 2 recordings for 30 min movies, or after 3 recordings for 10 min movies.

The setup for optogenetic experiments was assembled as described previously (Inagaki et al., 2014). Briefly, the red light (655 nm) LEDs were controlled via an Arduino Uno board (Arduino, Italy) using a custom program. As illustrated in Figure 1D, the stimulation paradigm (10 min in total) consists of 1 min pre-stimulation (time window ‘1’ in Figure 1D), three blocks of 1 min stimulation at an indicated frequency (time window ‘2’, 3 min in total) each followed by 2 min inter-stimulus intervals (ISIs, time window ‘3’, 6 min in total). The recording and LED control were manually started simultaneously.

Creation of behavioral classifiers

Request a detailed protocol

Lunges, wing extensions, and headbutts were quantified using behavioral classifiers developed in JAABA (Kabra et al., 2013) (https://sourceforge.net/projects/jaaba/files/), which runs on MATLAB. The details of the classifier development and quantification of performance will be described elsewhere (Leng X, Wohl M, Ishii K, Nayak P, and Asahina K, manuscript in preparation). Briefly, we trained the lunge classifiers using 20 male pair movies, nine male-female pair movies, five female pair movies, five fruM female-fruF male pair movies, and five male-fruF male pair movies; wing extension classifiers using 28 male pair movies, 12 male-female pair movies, eight female pair movies, and three fruM female-fruF male pair movies; headbutt classifier using 15 male pairs movies, seven male-female pair movies, 13 female pair movies, four fruM female-fruF male pair movies, and two male-fruF male pair movies. We then evaluated the performance of these classifiers using 17 male pair movies (223.5 min in total), four male-female pair movies (120 min in total), five female pair movies (150 min in total), and four fruM female-fruF male pair movies (34 min). These pairs were not used as training samples, and their behaviors were manually annotated prior to evaluation independently by at least two people (referred to as ‘true’ behaviors).

Recall rate was calculated as: (number of ‘true’ behavior bouts detected by a classifier)/(total number of ‘true’ behavior bouts).

While precision rate was calculated as: (number of ‘true’ behavior bouts detected by a classifier)/ [(total number of behavior bouts detected by a classifier) + (total number of false positive bouts detected by a classifier)].

Quantification of social behavior data

Request a detailed protocol

All behavioral data points used in all figures, as well as all statistical results with exact p values, can be found in Source data 1.

Acquired movies were first processed by the FlyTracker program (Eyjolfsdottir et al., 2014) (http://www.vision.caltech.edu/Tools/FlyTracker/), which runs on MATLAB (The Mathworks, Inc). The regions of interest were defined as circles corresponding to the chamber of each arena. The identities of tester and target flies were manually confirmed. The fly pair was removed from further analysis when (1) one of the two flies were killed during introduction to the chamber, (2) the wings of either fly was stuck at the extended position, or (3) the discrimination of the two flies was impossible due to wing damage of a tester fly. The amount of behavior is the number of bouts for lunges and headbutts, and the total duration (seconds) for wing extensions. These amounts were binned per minute for quantification. Extremely short bouts detected by a classifier were almost always false positives, and were eliminated from quantification. For lunges and headbutts, events with duration of less than 50 milliseconds were discarded. For wing extensions, events with duration of less than 100 milliseconds were discarded. The post-processing of data was done using custom MATLAB codes which are available on Github (https://github.com/wohlmp/Ishii_Wohl_DeSouza_Asahina_2019Wohl, 2019; copy archived at https://github.com/elifesciences-publications/Ishii_Wohl_DeSouza_Asahina_2019).

The duration of time a tester fly orients toward a target fly (‘time orienting’) was defined as the duration in which the following three conditions are met: (1) a target fly is within ±60° of the heading direction of a tester fly, (2) the distance between the ellipsoid centers of a tester and a target fly is within 5 mm (approximately two fly body length), and (3) the moving speed of the tester fly was above 0.1 mm/sec. The second and third conditions were used to eliminate frames in which a tester fly was standing still (and often grooming) while the target fly walked past in front of the tester fly. Frames that simultaneously fulfilled these three conditions were directly calculated from the output file of the tracker (‘trx.mat’ file) using MATLAB. To further eliminate incidents caused by erroneous flipping of fly orientations and other tracking errors, bouts with a duration of less than 200 milliseconds were discarded. This time period includes ‘chasing’, which has been observed in the context of both courtship and aggressive behaviors (Chen et al., 2002; Dankert et al., 2009; Hall, 1994).

The frame in which the infrared indicator LED turned on during the first LED stimulation period was used to align frames of movies. Statistical analyses were carried out using MATLAB. After behavior within each time window were calculated (see Figure 1D), the Kruskal-Wallis test (‘kruskalwallis’) was used to evaluate whether a given behavior was significantly different among different illumination windows (windows 1, 2, and 3; Figure 1D) or among different genotypes. When the p-value was below 0.05, the post-hoc Mann-Whitney signed rank test (‘signrank’) was used to detect significant differences between illumination period, and the Mann-Whitney U-test (‘ranksum’) was used to detect significant differences between testing and control genotypes. In both cases, Bonferroni correction was applied to p values. When the uncorrected p value was less than 0.05 but the corrected value did not pass the significant level, the uncorrected value was shown on panels with parenthesis.

In Figure 2—figure supplement 2D2, each tester fly was classified as ‘lunge increased’ in window two relative to window 1 or three if (1) lunges per minute increased more than 50%, or (2) more than one lunge per minute was performed when no lunge was detected in either window 1 or 3. In Figure 2—figure supplement 2D3, each tester fly was classified as ‘wing extension increased’ in window two relative to window 1 or three if (1) wing extension duration per minute was increased more than 50%, or (2) more than 500 milliseconds per minute of wing extension was performed when no wing extension was detected in either window 1 or 3. Fisher’s exact test (‘fishertest’) was used to determine whether the frequency of flies that increased lunging or wing extension was significantly different from the frequency of brains with a descending neuron (as shown in D1).

Generally, noticeably less lunges were detected in fruF males compared to in males. Often the corresponding classifiers did not detect any behaviors. In such zero-inflated datasets, a few false positive incidents can greatly impact the result of statistical tests. To avoid this pitfall, classifier results of lunges from fruF males were manually validated, and obvious false positives (caused by tracking errors, when a fly was near or on the wall, or when a fly suddenly jumped) were eliminated before statistical tests were applied.

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
    Sex and the single cell. I. on the action of major loci affecting sex determination in Drosophila melanogaster
    1. BS Baker
    2. KA Ridge
    (1980)
    Genetics 94:383–423.
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
  35. 35
  36. 36
  37. 37
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
  44. 44
  45. 45
  46. 46
  47. 47
  48. 48
  49. 49
  50. 50
  51. 51
  52. 52
  53. 53
  54. 54
  55. 55
  56. 56
  57. 57
  58. 58
  59. 59
  60. 60
  61. 61
  62. 62
  63. 63
  64. 64
  65. 65
  66. 66
  67. 67
  68. 68
  69. 69
  70. 70
  71. 71
  72. 72
  73. 73
  74. 74
  75. 75
  76. 76
  77. 77
  78. 78
  79. 79
  80. 80
  81. 81
  82. 82
  83. 83
  84. 84
  85. 85
  86. 86
  87. 87
  88. 88
  89. 89
  90. 90
  91. 91
  92. 92
  93. 93
  94. 94
  95. 95
  96. 96
  97. 97
  98. 98
  99. 99
  100. 100
  101. 101
  102. 102
  103. 103
  104. 104
  105. 105
  106. 106
  107. 107
  108. 108
  109. 109
  110. 110
  111. 111
  112. 112
  113. 113
  114. 114
  115. 115
  116. 116
  117. 117
  118. 118
  119. 119
  120. 120
  121. 121
  122. 122
  123. 123
    The Study of Instinct
    1. N Tinbergen
    (1951)
    Clarendon Press.
  124. 124
  125. 125
  126. 126
  127. 127
  128. 128
    Extended reproductive roles of the fruitless gene in Drosophila melanogaster revealed by behavioral analysis of new fru mutants
    1. A Villella
    2. DA Gailey
    3. B Berwald
    4. S Ohshima
    5. PT Barnes
    6. JC Hall
    (1997)
    Genetics 147:1107–1130.
  129. 129
  130. 130
  131. 131
  132. 132
  133. 133
  134. 134
  135. 135
  136. 136
  137. 137
  138. 138
  139. 139
  140. 140
  141. 141
  142. 142
  143. 143
  144. 144
  145. 145
  146. 146
  147. 147
  148. 148

Decision letter

  1. Mani Ramaswami
    Reviewing Editor; Trinity College Dublin, Ireland
  2. K VijayRaghavan
    Senior Editor; National Centre for Biological Sciences, Tata Institute of Fundamental Research, India

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

The evidence in this paper broadly supports the interesting conclusion that two distinct classes of sexually dimorphic neurons control male-specific courtship behaviour and its targeted display towards females. This is concluded based on the different requirement of Doublesex and Fruitless for each of these aspects of male courtship and the evidence that they function in different subsets of neurons.

Decision letter after peer review:

[Editors’ note: the authors submitted for reconsideration following the decision after peer review. What follows is the decision letter after the first round of review.]

Thank you for submitting your work entitled "A genetic code that specifies sexually dimorphic circuits and their control of social behaviors" for consideration by eLife. Your article has been reviewed by a Senior Editor, a Reviewing Editor, and three reviewers. The reviewers have opted to remain anonymous.

Our decision has been reached after consultation between the reviewers. In the discussions, all the referees expressed interest in the main message (dissociating fru and dsx functions and relating these to vertebrate mechanisms) as well as enthusiasm for several very interesting and clear findings from an impressively extensive set of experiments. However, there were also substantial concerns leading to the consensus opinion that additions and revisions required were too extensive to be accomplished in the two months required by eLife.

Based on these discussions and the individual reviews appended below, we regret to inform you that, in present form, your work will not be considered further for publication in eLife. However, if you choose to revise the paper on the lines suggested by the referees and resubmit, then eLife would be happy to consider a fresh submission (or two) that we would endeavour to have reviewed by the same referees.

In addition to the specific comments below, the referees offered some specific overall suggestions for revision that you may find useful. These include: (a) to refocus the paper on solely on the fru/dsx story; (b) to include new rigorous anatomical analysis of the "P1" cluster in P1a and NP2631 flies as requested in the reviews; and (c) providing a detailed discussion of limitation of using systemic/constitutive alleles of fru/dsx in relation to behavior as well as anatomy. The referees also felt that the story be clarified and better communicated by splitting into two separate but simpler manuscripts, with a potentially a separate manuscript on Tk, but you should of course make your own decision on this rather too interventional opinion.

Reviewer #1:

In both flies and mammals, multiple sexual differentiation mechanisms guide display of adult social behaviors. In flies it is dsx and fru, whereas in mammals it is largely testosterone and estrogen. Ishii et al. present a wide-ranging set of observations that they use to suggest that various aspects of sexually dimorphic courtship and aggression displays in flies are controlled by either dsx or fru. This seems an over-interpretation of the findings. I am supportive of the study but have a few suggestions.

1) Data from Figure 3 is used to conclude that dsx specifies both P1 and NP2631 neurons. This is inconsistent with the anatomy shown in figure panels, which shows that both dsx and fru contribute. It is true that cell number is dependent on dsx, but the morphology of the arbors indicates that fru is essential as well: E3/J3/K3 are different than C3/H3/I3. Similarly, data from Figure 6 is used to conclude that fruM but not dsx specifies Tk neurons; comparison of E3 and F3 indicates that dsxM is also important. The authors should quantify arborization rather than rely on gross examination for Figure 3 and Figure 6.

Presentation of the histology could be better elsewhere too. For example, from panels Figure 2A,C, I conclude that there is no overlap of R15A01 and R71G01, raising the question of the intersectional identification of P1a neurons.

2) It is clear that FruM is essential for male type lunges toward other males and enhancing wing extension to females. However, FruM is also essential for the pattern of wing extension upon activation of P1a neurons; in controls (Figure 1), wing extensions in phase 2>3 and 2<3 toward males and females, whereas in FruF males (Figure 4A-D), wing extensions in phase 2<3 toward males and females. In addition, it is clear that FruM is essential for regulating intensity of wing extensions. It is a gross simplification to conclude that dsxM enables execution of male courtship whereas FruM enforces choice of target sex. Clearly, FruM is also critical for execution of this behavior.

3) Why do males not lunge toward other males in the absence of activation of P1a or NP2631 or Tk neurons (Figure 1, Figure 6)?

4) There is some over-simplification of the literature. It might be true in flies that most studies have only used one sex as a target individual to analyze function of particular neurons (subsection “Effect of a target fly’s sex on optogenetically induced social interactions”), but this is certainly not true in mice (studies from Anderson, Dulac, Lin, Shah).

Rather than draw parallels to the four core genotype in the mouse (using which not a single gene has been identified that regulates social behavior independent of sex hormones, subsection “dsx, and not fru, specifies both P1a and NP2631 ∩ dsxFLP neurons”), it would be more appropriate to discuss the role of estrogen vs. testosterone in regulating male behaviors. It is clear that estrogen masculinizes the brain during a critical window whereas testosterone signaling amplifies male behaviors in adult life. This would be a more instructive analogy when trying to dissect the specific functions of dsx and fru.

Reviewer #2:

Unfortunately, I have some major technical concerns with the manuscript, and the lack of quantitative anatomical analyses throughout. The rather cursory analysis offered in this manuscript has several critical caveats that are not discussed and undermine the conclusions that can be drawn from the experimental data. Therefore, this manuscript is not suitable for publication without significant experimental clarification of these points.

Subsection “Effect of a target fly’s sex on optogenetically induced social interactions” and throughout the manuscript:

Neither the P1a split-Gal4 nor NP2631 ∩ dsxFLP driver are 'P1/pC1'-specific, some additional neurons in the brain express Gal4 in both cases (see both Hoopfer et al., 2015 and Koganezawa et al., 2016). Therefore, by using these drivers, the authors cannot conclude that the observed behavioral phenotypes were caused purely by the purported 'P1/pC1' neurons. Further intersectional approaches to rigorously target these neurons are essential.

Subsection “dsx, and not fru, specifies both P1a and NP2631 ∩ dsxFLP neurons”:

The expression 'P1a neurons were specified only in the flies that had dsxM' is not accurate and is likely even wrong. Whether the presence of the P1 neurons depends on fru or dsx was originally examined in the very first P1 paper, Kimura et al., 2008. Kimura et al. previously demonstrated by using dsx mutations that the absence of dsxF, but not the presence of dsxM, is the most striking factor for determining the presence of the P1 neurons.

In Figure 3, the dsx locus was always wild type and therefore the authors could not dissociate the presence of dsxM and the absence of dsxF. It is essential for the authors to do the similar experiments in combination with the appropriate dsx mutant alleles.

The authors mentioned that 'processes of P1a neurons in fruF males appeared thinner than those in males'. What does thinner mean? What form of scientific measurement is "thinner"? This needs to be quantified properly. Image co-registration or any other quantitative analyses for the size differences of the neuronal processes between these genotypes must be carried out.

The authors also mention ‘sex differences both in terms of cell body numbers and branching patterns’. If the authors declare the branching patterns were different between genotypes, they must carry out quantitative analyses.

They also declare the neuronal morphology was 'similar' between particular genotypes, but what does 'similar' mean? The authors seem to conclude that the neuronal morphology is different when they expect it is different and that it is similar when they expect it is similar, without any quantitative criteria for measurement.

Subsection “P1a and NP2631 ∩ dsxFLP neurons in fruF males can induce courtship, but not male-type aggressive behaviour”:

The logical link between the anatomy in Figure 3 and behavior in Figure 4 is unclear. In Figure 3, the authors looked at the contributions of the fru and dsx genes in the purported 'P1/pC1'-specific neurons, but in Figure 4, they carried out behavioral assays with the fru mutants, in which fru gene function is disrupted not only in P1/pC1 neurons but also all other FruM-expressing neurons throughout the PNS and CNS, they then base all their conclusions on only the behavioral function of P1/pC1 neurons. These animals will have a plethora of fru mutant phenotypes that are associated with lack of FruM in other neurons in the CNS and PNS.

To make any relevant conclusions the authors need to re-do the behavioral assays in Figure 4 using the flies in which only the P1/pC1 neurons are mutant for fru (e.g., mutant MARCM clones, UAS-Cas9 system, or fru-RNAi).

Subsection “fruM specifies aggression-promoting Tk-GAL4FruM neurons”:

The logic of transition from P1/pC1 to Tk-Gal4 neurons is very difficult to follow. The relationship between the former and the latter need to be explained. Are they completely different stories?

Figure 6:

As a control, the authors should demonstrate that Tk-GAL4 neurons do not express dsx.

Subsection “Differential roles of fru isoforms on male-male interactions and specification of Tk-GAL4FruM neurons”:

The authors showed that the Tk-Gal4 neurons exist in fru∆B mutant males but behavior is affected in these flies. The simplest explanation would be that the neurites are malformed in these flies.

Although the authors mentioned that 'fruMA and fruM mutants did not affect […] their arborization pattern', they didn't seem to carry out any quantitative morphological analyses. Such analyses should be included.

Subsection “The Tk-GAL4FruM neurons in fruMB mutants can induce male aggression”:

The authors mention that 'these mutants have Tk-GAL4FruM neurons that appear to retain their morphology'. But how did they examine this?

Reviewer #3:

This paper is a paradox. As far as I can tell it is rigorous on all levels. The scholarship is excellent, deep and well rounded. In addition, the issues are explored thoroughly and there is an enormous amount of data. In addition, an enormous number of fly lines (the table that describes the lines is phenomenal). But on the other hand, it is too much. I would like to discuss the argument and conclusions that surface relating dsx and fru expressing neurons with sexually dimorphic choice vs. sexually dimorphic motor patterns and the interesting differences between reproductive behavior and aggression, but the paper is simply too dense at almost every point. Reading it is a painful experience and this does the work a tremendous disservice. I say this recognizing and admiring the care that has gone into preparing this text. It seems flawless, just impossibly dense.

What to do? Here is a list of suggestions.

1) The text and figures are impeccably constructed but for the average reader they are overly complicated and far too dense. The flow of the manuscript would benefit greatly from severe editing of the text, figures and figure legends. It is with a heavy heart that I ask the authors to do this knowing the amount of effort it took to construct the manuscript.

2) Remove raster plots from figures throughout and move to supplemental material if needed. See Figure 1, Figure 4, Figure 5, Figure 6, Figure 7, Figure 9.

3) Figure 2 is not required. Move to supplemental material and concisely state result in text.

4) Much of the immunohistochemistry could be removed from figures and moved to supplemental material. See Figure 3, Figure 6, Figure 8.

5) The summary panels at the end of some of the figures are helpful. But a single unifying summary/model would be best. Though organized in a systematic and rational way the results still feel piecemeal and it is tiresome to continually have refer back to previous figures or supplemental material to regain your bearing. A single model would help immensely.

Alternatively, it might make sense to turn this into two papers, one about sex, the other about aggression?

[Editors’ note: further revisions were suggested prior to acceptance, as described below.]

Thank you for submitting your article "Sex-determining genes distinctly regulate courtship capability and target preference via sexually dimorphic neurons" for consideration by eLife. Your article has been reviewed by K VijayRaghavan as the Senior Editor, a Reviewing Editor,, and two reviewers. The reviewers have opted to remain anonymous.

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

Summary:

The evidence in this paper broadly supports the interesting conclusion that two distinct classes of sexually dimorphic neurons control male-specific courtship behaviour and its targeted display towards females. This is concluded based on the different requirement of Doublesex and Fruitless for each of these aspects of male courtship and the evidence that they function in different subsets of neurons.

Essential revisions:

Despite the notable revisions, e.g. valuable inclusion of quantitative anatomical analyses of the P1/pC1 neurons, there are remain issues that remain to be addressed.

1) Although the cover letter refers to Figure 1E of Hoopfer et al. (2015) to justify the view that the driver is P1/pC1-specific, Figure 1—figure supplement 1C of Hoopfer et al. (2015) appears to show that there are several additional neurons close to the lateral horn labelled by this split-Gal4 driver in both males and females. The main text of Hoopfer et al. (2015) states that "in addition to P1 neurons, the P1a spGAL4 driver labelled a minor and variable population of neurons (<4 per hemi- brain) with cell bodies in the lateral horn (LH) that were neither FruM-positive nor male-specific".

2) The same issue applies to the NP2631 ∩ dsxFLP driver. The text and cover letter reference Figure 4D of Koganezawa et al. (2016) to justify that their driver is P1/pC1-specific. However, the referenced figure appears to show a single descending axon per hemisphere is going out of the ventral end of the GNG. This axon crosses the midline at the neuropil region just dorsal to the oesophageal foramen, which is a typical neurite structure of another dsx-expressing neuronal cluster called pMN1 (see Figure S1 of Kimura et al., 2015). Similarly, in Figure S3A of Koganezawa et al. (2016), a descending axon of pMN1 is more obvious in this female specimen (pMN1 is present in both sexes with dimorphic neurite morphology, according to Figure S1 of Kimura et al., 2015). I would encourage the authors to show the expression patterns in the VNC. Are there any cells in this tissue labelled by the NP2631 ∩ dsxFLP driver?

The above two Points 1 and 2 reflect concerns on the specificity of driver lines used. To address ambiguities that arise from this, unless there are additional strong arguments in support of the authors' interpretations, ideally all behavioural assays should be re-done under the conditions of P1/pC1-specific activation using either further intersection or a stochastic/mosaic approach.

3) It still remains unclear how fru function is dissociated from that of dsx in the sexual differentiation of P1/pC1 neurons. To clarify potentially distinct contributions of fru and dsx to this developmental process, it appears that dsx mutant analyses are required. Have these been attempted or done?

4) The experiments use fruM, fruF and single-isoform fru mutant alleles for behavior and attribute the behavioural phenotypes to P1/pC1 (this paper) and Tk-GAL4FruM neurons (the other paper, titled “Layered roles of fruitless isoforms in specification and function of male aggression-promoting neurons in Drosophila”). Given that Fru isoforms are expressed in other neurons in the CNS and the PNS, surely the effects of Fru isoform on behavior in P1/pC1 and Tk-GAL4FruM neurons can only really be assessed by looking at animals were the specific neurons are mutant for fru (e.g. fru isoform-specific mutant MARCM clones, UAS-Cas9 system, or fru-RNAi)? Otherwise, the data would need to be interpreted with obvious caveats.

5) The paper should include better images of Dsx immunolabeling (Figure 2—figure supplement 2E) as the current one is of very poor quality.

https://doi.org/10.7554/eLife.52701.sa1

Author response

[Editors’ note: the authors resubmitted a revised version of the paper for consideration. What follows is the authors’ response to the first round of review.]

Our decision has been reached after consultation between the reviewers. In the discussions, all the referees expressed interest in the main message (dissociating fru and dsx functions and relating these to vertebrate mechanisms) as well as enthusiasm for several very interesting and clear findings from a impressively extensive set of experiments. However, there were also substantial concerns leading to the consensus opinion that additions and revisions required were too extensive to be accomplished in the two months required by eLife.

Based on these discussions and the individual reviews appended below, we regret to inform you that, in present form, your work will not be considered further for publication in eLife. However, if you choose to revise the paper on the lines suggested by the referees and resubmit, then eLife would be happy to consider a fresh submission (or two) that we would endeavour to have reviewed by the same referees.

In addition to the specific comments below, the referees offered some specific overall suggestions for revision that you may find useful. These include: (a) 1 to refocus the paper on solely on the fru/dsx story; (b) to include new rigorous anatomical analysis of the "P1" cluster in P1a and Np2631 flies as requested in the reviews; and (c) providing a detailed discussion of limitation of using systemic/constitutive alleles of fru/dsx in relation to behavior as well as anatomy. The referees also felt that the story be clarified and better communicated by splitting into two separate but simpler manuscripts, with a potentially a separate manuscript on Tk, but you should of course make your own decision on this rather too interventional opinion.

We decided to split our original manuscript into two, following two reviewers’ suggestions that this would increase clarity. We agree that our attempts to integrate our findings on 2 genes (dsx and fru), 2 behaviors (courtship and aggression), and 3 neuronal populations (NP2631 ∩ dsxFLP, P1a, and Tk-GAL4FruM neurons) caused difficulties with the flow of logic. We believe that two separate manuscripts allow us to focus on two relatively simpler messages individually, and help us present relevant data without causing confusion. The current paper, “Sex-determining genes distinctly regulate courtship capability and target preference via sexually dimorphic neurons”, also serves to incorporate your first major recommendation: “to refocus the paper on solely on the fru/dsx story”. In line with this recommendation, we now include a new set of data showing that dsx and fru-co-expressing neurons in males can promote both courtship and aggressive behaviors in a target sex-dependent manner (see Figure 1 of this manuscript).

Regarding your second recommendation to provide rigorous neuroanatomical analysis, we thoroughly re-evaluated our anatomical data by registering brains to a common template and analyzing specific volumes with 3-dimentional segmentation. Please see Figure 3L-M, Figure 6H, I, in “Sex-determining genes distinctly regulate courtship capability and target preference via sexually dimorphic neurons”, and Figure 1N-P, and Figure 5I-K in “Layered roles of fruitless isoforms in specification and function of male aggression-promoting neurons in Drosophila” for new quantification of neuronal morphologies. Standardized immunohistochemical data enabled us to directly compare the morphology of specific neurons across multiple genotypes. Moreover, we could detect statistical differences in neuroanatomy by quantitatively comparing segmented volumes of specific neuronal structures. This rigorous quantification revealed differences in P1a neurons between males and fruF males, which compelled us to revise our original conclusions for this class of neurons.

In response to your third recommendation, we added a discussion of the limitations of studying sex-determining genes through constitutive mutants and through cell-specific genetic manipulation (see “An organismal sex and a cellular sex” section in Discussion of “Sex-determining genes distinctly regulate courtship capability and target preference via sexually dimorphic neurons”). We believe that cell-specific manipulation of sex-determining genes is a powerful method that is not without weaknesses, and will serve complimentary roles to constitutive mutants when studying their effects on neuroanatomy and behaviors.

Reviewer #1:

In both flies and mammals, multiple sexual differentiation mechanisms guide display of adult social behaviors. In flies it is dsx and fru, whereas in mammals it is largely testosterone and estrogen. Ishii et al. present a wide-ranging set of observations that they use to suggest that various aspects of sexually dimorphic courtship and aggression displays in flies are controlled by either dsx or fru. This seems an over-interpretation of the findings. I am supportive of the study but have a few suggestions.

1) Data from Figure 3 is used to conclude that dsx specifies both P1 and NP2631 neurons. This is inconsistent with the anatomy shown in figure panels, which shows that both dsx and fru contribute. It is true that cell number is dependent on dsx, but the morphology of the arbors indicates that fru is essential as well: E3/J3/K3 are different than C3/H3/I3. Similarly, data from Figure 6 is used to conclude that fruM but not dsx specifies Tk neurons; comparison of E3 and F3 indicates that dsxM is also important. The authors should quantify arborization rather than rely on gross examination for Figure 3 and Figure 6.

We agree that we should have quantified our findings to draw conclusions about the neuroanatomy of these neurons. Reviewer #2 also raised concerns about our anecdotal statements regarding neuroanatomy. We therefore thoroughly re-analyzed our anatomical data by registering brains to a “template” Drosophila brain, and by using 3-dimentional segmentation and volume measurements. Please see Materials and methods section for technical details. In essence, we used a non-rigid transformation technique described by Jefferis et al., 2007, to register individual sample brains to the standard unisex Drosophila brain described in Bogovic et al., 2018. This allowed us to compare labeling patterns across multiple brains and sexes in the single reference space.

Immunohistochemical labeling can have technical as well as biological variability. Such inter-sample variability might have been perceived as inter-genotype differences. To better represent labeling patterns that are consistent across samples, we averaged the signal intensity of imaged neurons after registration in the standard brain. The resulting images are shown in 2-dimensional projections as well as in 3-dimensional movies. A similar approach was taken to show the neuroanatomy of various fru-expressing neurons (Yu et al., 2010; von Philipsborn et al., 2014), and we feel this is the best approach to concisely summarize our observations from multiple samples.

As a quantification, we compared the volumes of specific neuronal innervations across different genotypes. We used the 3-dimensional rendering program FluoRender to segment structures of interest, and calculated the volumes of each structure.

With these new and rigorous approaches, we are now able to quantitatively support our initial conclusions that (1) NP2631 ∩ dsxFLP neurons are largely specified by dsx, and (2) Tk-GAL4FruM neurons are specified by fru.

In “Sex-determining genes distinctly regulate courtship capability and preference toward females through sexually dimorphic neurons” Figure 3H-K and Video 3, Video 4 and Video 5, we demonstrate that the overall morphology of NP2631 ∩ dsxFLP neurons in males and fruF males overlap well, and that these neurons in females and fruM females are also similar. We quantified three neuropils that show noticeable sexual dimorphism, and in all cases, volumes of these neuropils in fruF males are statistically indistinguishable from those in males. Volumes in fruM females are comparable to values in females except in one neuropil. We acknowledge that our quantification is based on the entire neural population, and that we may not have sufficient sensitivity to detect either subtype-specific contributions of fru, or differences at finer scales such as number or length of branches. We mention this possibility in the main text. Overall, however, our analysis provided us with evidence arguing that sexual dimorphism of NP2631 ∩ dsxFLP neurons is mainly, if not entirely, specified by dsx.

Likewise, data in “Layered roles of fruitless isoforms in specification and function of male-type aggression-promoting neurons in Drosophila” Figure 1L, M and Video 3 clearly show that average Tk-GAL4FruM neurons from males and from fruM females almost perfectly overlap (light cyan in Video 3 indicates the overlap between male (green) and fruM female (dark blue) Tk-GAL4FruM neurons). As is shown in “Layered roles of fruitless isoforms in specification and function of male-type aggression-promoting neurons in Drosophila” Figure 1N-P, three prominent innervations of these neurons show similar volumes across males (both fru +/+ and fruM/fru4-40) and fruM females. We therefore retain our original conclusion that Tk-GAL4FruM neurons are predominantly specified by fru.

In contrast, we found that both dsx and fru are important for specifying P1a neurons. In the previous version, we hastily concluded that dsx specified P1a neurons mainly because P1a neurons appear in fruF male brains, but not in fruM female brains. However, quantification of volumes of two neuropils clearly indicates that P1a neurons in fruF males have distinct morphology from those in males (see “Sex-determining genes distinctly regulate courtship capability and preference toward females through sexually dimorphic neurons”, Figure 6F-I). Results from this and re-analysis of behavioral data (discussed below) compelled us to revise our conclusion about P1a neurons. We now fully acknowledge the role of fru in both specification and function of P1a neurons. We feel this is a striking contrast to NP2631 ∩ dsxFLP neurons, and further supports the idea that these two neurons are distinct subpopulations within what has been collectively described as P1 or pC1 neurons.

We thank the reviewer for encouraging us to take an extra step for verification (which turned out to require substantial manual and computational works), as we feel our revised conclusions are now better supported by the data.

Presentation of the histology could be better elsewhere too. For example, from panels Figure 2A,C, I conclude that there is no overlap of R15A01 and R71G01, raising the question of the intersectional identification of P1a neurons.

We admit that images in our original version were not sufficiently clear, especially on printed media. It is a challenge to show all visible neurons clearly without over-saturating strongly labeled cells, or without dominating the figure panel. We enlarged the images by 20%, and would like to draw attention to original images provided as source data.

Regarding the overlap of R15A01 and R71G01 promoters, our initial submission used the R71G01-LexA transgene in attP40. It is known in the community that transgene expression at the attP40 landing site is not as robust as at the attP2 landing site, which is used for creating the Rubin GAL4 collection. We had created a transgenic animal with R71G01-LexA in attP2, and tested its overlap with NP2631 ∩ dsxFLP and with NP2631 ∩ fruFLP (“Sex-determining genes distinctly regulate courtship capability and preference toward females through sexually dimorphic neurons”, Figure 5C, D and Figure 5—figure supplement 3C, D, respectively). As anticipated, we observed more R71G01-LexA-labeled neurons from the attP2 transgene than from the attP40 transgene. The overlap with NP2631 ∩ dsxFLP or NP2631 ∩ fruFLP neurons is still very small, confirming our initial conclusion that P1a neurons have minimal overlap with these neurons. We strengthen this conclusion by comparing registered and averaged 3D images of NP2631 ∩ dsxFLP and P1a neurons (“Sex-determining genes distinctly regulate courtship capability and preference toward females through sexually dimorphic neurons”, Figure 5E-G). Neuroanatomy of these two populations shows significant differences.

In addition, Hoopfer et al., 2015 showed that optogenetic stimulation of R15A01-LexA neurons recapitulates the behavioural effects of the P1a neuronal activation (see Figure 3H in the citation). This data suggest that R1501-LexA indeed includes P1a neurons. We therefore remain confident with our conclusion which this piece of data is most relevant: NP2631 ∩ dsxFLP and P1a neurons are largely separate subpopulations.

We would like to add that split GAL4 lines are often created from lines labeling seemingly different neurons at a population level, especially in z-projection images. In fact, it may be difficult to discern overlapping neurons from the images of R15A01-GAL4 and R71G01-GAL4 in Figure 1C of Hoopfer et al., 2015 (which described P1a neurons).

2) It is clear that FruM is essential for male type lunges toward other males and enhancing wing extension to females. However, FruM is also essential for the pattern of wing extension upon activation of P1a neurons; in controls (Figure 1), wing extensions in phase 2>3 and 2<3 toward males and females, whereas in FruF males (Figure 4A-D), wing extensions in phase 2<3 toward males and females. In addition, it is clear that FruM is essential for regulating intensity of wing extensions. It is a gross simplification to conclude that dsxM enables execution of male courtship whereas FruM enforces choice of target sex. Clearly, FruM is also critical for execution of this behavior.

The reviewer was right in pointing out the importance of fru on the function, as well as on neuroanatomy (discussed above), of P1a neurons. We confirmed that optogenetic stimulation of P1a neurons in fruF males induced significantly less wing extensions than in regular males regardless of target sex (“Sex-determining genes distinctly regulate courtship capability and preference toward females through sexually dimorphic neurons”, Figure 6—figure supplement 2). This means that the courtship-executing function, as well as specification, of P1a neurons require the activity of both dsx and fru. Currently, we do not know whether dsx and/or fru is necessary within P1a neurons, or whether the proper function of P1a neurons requires another population of dsx- or fru-dependent neurons (see also discussions regarding cell type-specific manipulation of sex-determining genes). We therefore reasoned that it would be prudent to avoid speculations about the roles of dsx and fru on P1a neurons, and decided to re-frame P1a-related data in the context of discussing the potential diversity of the P1/pC1 neuronal cluster.

On the other hand, optogenetic stimulation of NP2631 ∩ dsxFLP neurons in fruF males induced a similar amount of wing extensions as the same manipulation in regular males did toward male target flies (“Sex-determining genes distinctly regulate courtship capability and preference toward females through sexually dimorphic neurons”, Figure 4C). Together with our observation that the neuroanatomy of NP2631 ∩ dsxFLP neurons in regular males and in fruF males showed little difference, we conclude that (1) fruF males are capable of executing wing extensions, (2) NP2631 ∩ dsxFLP neurons are one neuronal component that belongs to a dsx-dependent wing extension execution circuit, and (3) fruM is important for enhancing courtship behavior specifically toward females.

By referring to the “execution” of courtship behavior, we are specifically focusing on the animal’s capability to express courtship behavior. We aim to uncover a neurogenetic process that enables animals to perform courtship behavior at all, in contrast to the ability to adjust the intensity of courtship behavior toward females. Please note that we do not argue that fru is not required for proper courtship behavior. What we argue is that fru is not necessary for all aspects of male courtship behavior. The importance of fru on male sexual behavior has been extensively demonstrated, leading some to state that fruM specifies all aspects of male sexual behaviors. Here, we wish to shed light on the fundamental characteristics of fru mutants, which is the unusually high level of courtship behavior toward males, not the absence of courtship behavior altogether. This phenotype has been consistently and repeatedly reported (literature include Gailey and Hall (1989), Villella et al. (1997), Ito et al. (1996), Demir and Dickson (2005), Shirangi et al. (2006), Pan and Baker (2014), among others), even though these observations may not have received sufficient attention. Moreover, many of previously characterized fru-expressing neurons are sensory neurons or early-stage interneurons that transmit sensory information. We feel that our conclusions are entirely consistent with past literature. Our novel functional characterization of NP2631 ∩ dsxFLP neurons provides a neuronal substrate that accounts for the courtship behavior-execution capability of fruF males.

3) Why do males not lunge toward other males in the absence of activation of P1a or NP2631 or Tk neurons (Figure 1, Figure 6)?

The tester flies were all housed in groups for 6 days (P1a neurons) or 14 days (for NP2631 ∩ dsxFLP neurons), which potently reduces spontaneous aggressive behavior. Socially isolated animals show high levels of spontaneous aggression, which can cause a ceiling effect when trying to observe further increases of aggression. Both in flies and mice, group-housing is a common treatment to characterize behavioral changes in response to neuronal manipulations.

4) There is some over-simplification of the literature. It might be true in flies that most studies have only used one sex as a target individual to analyze function of particular neurons (subsection “Effect of a target fly’s sex on optogenetically induced social interactions”), but this is certainly not true in mice (studies from Anderson, Dulac, Lin, Shah).

Rather than draw parallels to the four core genotype in the mouse (using which not a single gene has been identified that regulates social behavior independent of sex hormones, subsection “dsx, and not fru, specifies both P1a and NP2631 ∩ dsxFLP neurons”), it would be more appropriate to discuss the role of estrogen vs. testosterone in regulating male behaviors. It is clear that estrogen masculinizes the brain during a critical window whereas testosterone signaling amplifies male behaviors in adult life. This would be a more instructive analogy when trying to dissect the specific functions of dsx and fru.

We appreciate the reviewer’s constructive suggestion to re-frame the roles of dsx and fru. Indeed, studies in mice often use both males (often castrated) and females as intruders. While that these neuronal manipulations may appear to trigger social behaviors largely independent of target sex, we found that some target sex-dependence has been previously reported. For instance, Lin et al. (2010) reported that optogenetic stimulation of the mediolateral part of the ventromedial hypothalamus (VMHml) triggered aggression more consistently toward males than toward females (Figure 4L,M in the citation). Likewise, close examination of data from Yang et al. (2017), which investigated the behavioral changes caused by chemogenetic activation of progesterone receptor expressing neurons in VMHml, suggests that aggression toward males was more consistent that toward females, although it was not explicitly tested (Figure 1F,G in citation). These past observations reinforce our idea that target sex in neuronal manipulation experiments can be an important variable in other experimental models.

We also agree that the “four core genotypes” model has not been particularly successful in providing mechanical insight into the origin of sexual dimorphism. Although the temporal dynamics between mammalian sex hormones and Drosophila sex-determining genes can be quite different, the reviewer’s opinion that the roles of dsx and fru can be analogous to those of estrogen and testosterone is indeed a quite useful framework for our study. We incorporated this suggestion into our Discussion section.

Reviewer #2:

Unfortunately, I have some major technical concerns with the manuscript, and the lack of quantitative anatomical analyses throughout. The rather cursory analysis offered in this manuscript has several critical caveats that are not discussed and undermine the conclusions that can be drawn from the experimental data. Therefore, this manuscript is not suitable for publication without significant experimental clarification of these points.

As discussed in response to reviewer #1, we incorporated quantification of our neuroanatomical data. Additional details are provided below, and we hope these new analyses address the reviewer’s concerns.

Subsection “Effect of a target fly’s sex on optogenetically induced social interactions” and throughout the manuscript:

Neither the P1a split-Gal4 nor NP2631 ∩ dsxFLP driver are 'P1/pC1'-specific, some additional neurons in the brain express Gal4 in both cases (see both Hoopfer et al., 2015 and Koganezawa et al., 2016). Therefore, by using these drivers, the authors cannot conclude that the observed behavioral phenotypes were caused purely by the purported 'P1/pC1' neurons. Further intersectional approaches to rigorously target these neurons are essential.

We are a little confused by the reviewer’s assertion that “Neither P1a-GAL4 or NP2631 ∩ dsxFLP are ‘P1/pC1’-specific”, after citing the previous publications. In both cases, we see in the citation figures that these drivers in fact label a single cluster of neurons cleanly, both in original publications as well as in our own studies.

For P1a-GAL4 neurons, Figure 1E of Hoopfer et al., 2015 shows clearly that only a single cluster of neurons at the posterior end of a male brain were labeled. We confirmed this specificity repeatedly, as shown in “Sex-determining genes distinctly regulate courtship capability and preference toward females through sexually dimorphic neurons”, Figure 6A, Figure 6—figure supplement 1A, C. NP2631 ∩ dsxFLP neurons described in Figure 4D of Koganezawa et al., 2016 also seems to label a well-defined single cluster of neurons per hemisphere. Again, we observed a bilateral single cluster of neurons at the posterior edge of male brains (“Sex-determining genes distinctly regulate courtship capability and preference toward females through sexually dimorphic neurons”, Figure 3C, Figure 3—figure supplement 1A, C). Please also see the movies showing the 3-dimentional structure of P1a and NP2631 ∩ dsxFLP neurons.

We wondered whether the reviewer was raising the possibility that this seemingly single cluster of neurons may indeed consist of heterogeneous populations. Although the number of cells that are labeled by both drivers is rather small (“Sex-determining genes distinctly regulate courtship capability and preference toward females through sexually dimorphic neurons”, Figure 3G and Figure 6E), further intersection could reveal a specific subset that is responsible for the observed behaviors. We currently do not have a means to reliably separate such subpopulations, and we also feel that such experiments would better belong to a future study. Here, we focus on the existence of dsx-dependent neurons that can execute wing extensions, and a separate, fru-dependent mechanism that enhances wing extensions toward females.

Subsection “dsx, and not fru, specifies both P1a and NP2631 ∩ dsxFLP neurons”:

The expression 'P1a neurons were specified only in the flies that had dsxM' is not accurate and is likely even wrong. Whether the presence of the P1 neurons depends on fru or dsx was originally examined in the very first P1 paper, Kimura et al., 2008. Kimura et al. previously demonstrated by using dsx mutations that the absence of dsxF, but not the presence of dsxM, is the most striking factor for determining the presence of the P1 neurons.

In Figure 3, the dsx locus was always wild type and therefore the authors could not dissociate the presence of dsxM and the absence of dsxF. It is essential for the authors to do the similar experiments in combination with the appropriate dsx mutant alleles.

We admit that we were careless in stating that either dsxM or dsxF is responsible for neuroanatomical or behavioral differences from regular males. With the reagents we used, we could only state whether sexually dimorphic splicing of dsx has an impact on NP2631 ∩ dsxFLP or P1a neurons. We do not know whether altered neuroanatomy of P1a neurons in fruF males is due to the lack of dsxM, or due to the presence of dsxF. As the reviewer points out, we must use dsx mutants (especially dsxDom, which forces male-type splicing of dsx regardless of sex chromosome composition) to gain insight into which of the two isoforms is responsible.

However, we would like to emphasize that it is not the specific role of dsxM or dsxF, but rather the genetic origin of sexual dimorphism, that we wished to investigate in this study. That is, we are interested in addressing which sexually dimorphic characteristics (neuroanatomical and behavioral) are controlled by the dimorphism (e.g., sex-specific splicing) of dsx and fru. As we state in the Discussion, one sex is not a mutant of the other sex, and sex-determining genes often have different functions in each of the two sexes – although fru was often considered silent in females, we presented circumstantial evidence that the fruF allele may not be truly null (“Layered roles of fruitless isoforms in specification and function of male-type aggression-promoting neurons in Drosophila”, Figure 1—figure supplement 3A-C). In this context, splicing mutants such as fruM and fruF provide a useful platform to understand the genetic nature of sex-specific transformation. These two alleles have the added benefit that experiments could be conducted in a symmetrical manner – we can exhaust all four possible combinations of dsx and fru sex-specific isoforms (“Sex-determining genes distinctly regulate courtship capability and preference toward females through sexually dimorphic neurons”, Figure 3B). Currently, no mutation that forces female-type splicing of dsx is known.

We believe our data present strong evidence that dsx and fru have distinct impacts on the specification and function of sexually dimorphic neuronal circuits. We revised our statements and referred to the importance of dsx as a gene, not of sex-specific splicing isoforms.

The authors mentioned that 'processes of P1a neurons in fruF males appeared thinner than those in males'. What does thinner mean? What form of scientific measurement is "thinner"? This needs to be quantified properly. Image co-registration or any other quantitative analyses for the size differences of the neuronal processes between these genotypes must be carried out.

The authors also mention ‘sex differences both in terms of cell body numbers and branching patterns’. If the authors declare the branching patterns were different between genotypes, they must carry out quantitative analyses.

They also declare the neuronal morphology was 'similar' between particular genotypes, but what does 'similar' mean? The authors seem to conclude that the neuronal morphology is different when they expect it is different and that it is similar when they expect it is similar, without any quantitative criteria for measurement.

As stated above in reply to a similar request from reviewer #1, we did an extensive re-analysis of our neuroanatomical data. We followed reviewer #2’s request, and hope that our new set of data will directly address the reviewers’ concerns.

Subsection “P1a and NP2631 ∩ dsxFLP neurons in fruF males can induce courtship, but not male-type aggressive behaviour”:

The logical link between the anatomy in Figure 3 and behavior in Figure 4 is unclear. In Figure 3, the authors looked at the contributions of the fru and dsx genes in the purported 'P1/pC1'-specific neurons, but in Figure 4, they carried out behavioral assays with the fru mutants, in which fru gene function is disrupted not only in P1/pC1 neurons but also all other fruM-expressing neurons throughout the PNS and CNS, they then base all their conclusions on only the behavioral function of P1/pC1 neurons. These animals will have a plethora of fru mutant phenotypes that are associated with lack of fruM in other neurons in the CNS and PNS.

To make any relevant conclusions the authors need to re-do the behavioral assays in Figure 4 using the flies in which only the P1/pC1 neurons are mutant for fru (e.g., mutant MARCM clones, UAS-Cas9 system, or fru-RNAi).

Our experiments address what the NP2631 ∩ dsxFLP or P1a neurons can do in fruF males. We would like to clarify that our original submission did not state that fru is required cell autonomously in either NP2631 ∩ dsxFLP or P1a neurons, precisely because we were aware that we could not attribute the behavioral differences from regular males to the deficit specifically within either neuronal population. We apologize if our wording in the previous submission caused any confusion.

We agree with the reviewer’s critique that the use of constitutive mutants has a limitation. We cannot answer whether the phenotype (anatomical or behavioral) is due to the lack of dsx or fru functions in the neurons of interest, or due to their cell non-autonomous functions (likely in other neurons). However, we would like to reiterate that the cell-autonomous function of dsx or fru is not the focus of this study. We wished to address how sexual dimorphism of dsx and fru (in a form of sex-specific splicing) may contribute to the specification of sexually dimorphic circuits and sexually dimorphic social behaviors. As stated above, simple loss-of-function of dsx or fru does not necessarily distinguish which male-like or female-like characteristics are specified by a given gene, because the loss of a gene is not synonymous to the transformation of sex from one to the other. We respectfully disagree with the reviewer’s assertion that we cannot make “any relevant conclusions” with these mutants. Rather, we believe that systematic investigation into what changes in fruF males and fruM females at circuit and behavioral levels (e.g., comparison of all 4 genotypes shown in “Sex-determining genes distinctly regulate courtship capability and preference toward females through sexually dimorphic neurons”, Figure 3B) is crucial for understanding the transformative nature of sex determination.

We would also like to be mindful of the limitations of cell-type specific manipulations of dsx and fru. For instance, successful gene knockdown using ‘dead Cas9’ (CRISPRi) crucially depends on the sequence of 20-base guide RNA. RNA interference (RNAi) relies on the specificity of ~21-base double-strand RNA sequence. In both cases, off-target effects as well as incomplete (or even ineffective) knockdown of the target gene are prevalent and well documented. Incomplete knockdown can be also caused by varied strengths of GAL4 drivers. To make matters more complicated, dsx and fru are likely crucial during development. This means that a GAL4 driver needs to turn on very early and remain active consistently during development to provide interpretable results. In previous studies, characterization of both the efficacy of fru RNAi knockdown or the temporal activity or GAL4 drivers is disturbingly sparse, raising a question about the interpretability of such data. Presumably because of this limitation, results of RNAi against sex-determining genes are often presented only when such manipulation results in reproduction of phenotypes in systemic mutants (that is, when the result is interpretable). Considering that cell-autonomous role of dsx or fru is not necessarily the focus of our study, we do not think either RNAi or CRISPRi will further support our conclusions.

MARCM can circumvent the problem of timing and strength because the mutations become homozygous at the cell division. However, MARCM also has an often-overlooked weakness that any cell clones in which a GAL4 driver is not active remain undetected. Moreover, creation of animals with mutant cells only in the desired neuronal population is exceedingly difficult, often requiring hundreds of individuals to identify a handful of suitable samples. Because the distribution of cell clones can be analyzed only after behavioral experiments, and since there is substantial inherent variability of behavior, an unreasonable amount of labor would be required to obtain interpretable data using MARCM.

We are aware of the limitation of constitutive mutants, but systematic analysis of such mutants is fundamental to interpret cell-specific manipulations of sex-determining genes. We hope reviewers will understand the value in what we can observe, and allow us to address functions of dsx or fru within specific neuronal populations in future studies.

Subsection “fruM specifies aggression-promoting Tk-GAL4FruM neurons”:

The logic of transition from P1/pC1 to Tk-Gal4 neurons is very difficult to follow. The relationship between the former and the latter need to be explained. Are they completely different stories?

As mentioned at the beginning of this letter, we have split the original submission into 2 parts. This point was also raised by reviewer #3, and we hope the current format better conveys our central messages.

Figure 6:

As a control, the authors should demonstrate that Tk-GAL4 neurons do not express dsx.

Please see Figure 1 —figure supplement 2C.

Subsection “Differential roles of fru isoforms on male-male interactions and specification of Tk-GAL4FruM neurons”:

The authors showed that the Tk-Gal4 neurons exist in fru∆B mutant males but behavior is affected in these flies. The simplest explanation would be that the neurites are malformed in these flies.

Although the authors mentioned that 'fruMA and fruM mutants did not affect […] their arborization pattern', they didn't seem to carry out any quantitative morphological analyses. Such analyses should be included.

Subsection “The Tk-GAL4FruM neurons in fruMB mutants can induce male aggression”:

The authors mention that 'these mutants have Tk-GAL4FruM neurons that appear to retain their morphology'. But how did they examine this?

We now show the averaged images of Tk-GAL4FruM neurons in fruMA and fruMB mutants, which is in fact very similar to those in wild type males (“Layered roles of fruitless isoforms in specification and function of male-type aggression-promoting neurons in Drosophila”, Figure 5F-H, Video 7). The volume quantification of three major neural processes in these three genotypes indeed shows that they are indistinguishable from each other.

Again, this level of analysis may not reveal subcellular differences in neuromorphology among regular males, fruMA mutants, and fruMB mutants. However, we would like to point out that our conclusion that fruMB does not affect the specification of Tk-GAL4FruM neurons is based on physiological and behavioral data, as well as neuroanatomical data (“Layered roles of fruitless isoforms in specification and function of male-type aggression-promoting neurons in Drosophila”, Figure 6 and Figure 5, respectively). The fact that optogenetic stimulation of Tk-GAL4FruM neurons in fruMB mutants triggers aggression as robustly as in regular males suggests that the overall integrity of Tk-GAL4FruM neurons remains largely unaffected in fruMB mutants.

Reviewer #3:

This paper is a paradox. As far as I can tell it is rigorous on all levels. The scholarship is excellent, deep and well rounded. In addition, the issues are explored thoroughly and there is an enormous amount of data. In addition, an enormous number of fly lines (the table that describes the lines is phenomenal). But on the other hand, it is too much. I would like to discuss the argument and conclusions that surface relating dsx and fru expressing neurons with sexually dimorphic choice vs. sexually dimorphic motor patterns and the interesting differences between reproductive behavior and aggression, but the paper is simply too dense at almost every point. Reading it is a painful experience and this does the work a tremendous disservice. I say this recognizing and admiring the care that has gone into preparing this text. It seems flawless, just impossibly dense.

We understand a reservation that echoes with other reviewers’ concerns regarding the clarity of our original submission. We would like to address the reviewer’s specific suggestions in the following sections.

What to do? Here is a list of suggestions.

1) The text and figures are impeccably constructed but for the average reader they are overly complicated and far too dense. The flow of the manuscript would benefit greatly from severe editing of the text, figures and figure legends. It is with a heavy heart that I ask the authors to do this knowing the amount of effort it took to construct the manuscript.

We significantly reorganized our original submission. Namely, we first understood the value of the last suggestion and split the original version into two new manuscripts as we already described. We hope this will substantially improve the clarity of our messages.

2) Remove raster plots from figures throughout and move to supplemental material if needed. See Figure 1, Figure 4, Figure 5, Figure 6, Figure 7, Figure 9.

We moved the raster plots of behaviors largely to figure supplement panels.

3) Figure 2 is not required. Move to supplemental material and concisely state result in text.

We feel the evidence suggesting that P1a and NP2631 ∩ dsxFLP neurons are separate populations has significance, especially now that we detected different levels of contribution of fru on these two neuronal populations. We therefore decided to keep the panels in “Sex-determining genes distinctly regulate courtship capability and preference toward females through sexually dimorphic neurons”, Figure 5.

4) Much of the immunohistochemistry could be removed from figures and moved to supplemental material. See Figure 3, Figure 6, Figure 8.

This is in contrast to requirements from reviewers #1 and #2 to perform more rigorous neuroanatomical analyses. We followed a custom for most Drosophila papers and kept representative images, but we would be open to further suggestions for improving clarity.

5) The summary panels at the end of some of the figures are helpful. But a single unifying summary/model would be best. Though organized in a systematic and rational way the results still feel piecemeal and it is tiresome to continually have refer back to previous figures or supplemental material to regain your bearing. A single model would help immensely.

We reduced the number of models to one for “Sex-determining genes distinctly regulate courtship capability and preference toward females through sexually dimorphic neurons” (Figure 4D), and two for “Layered roles of fruitless isoforms in specification and function of male-type aggression-promoting neurons in Drosophila” (Figure 2D and Figure 6H). It was a challenge to summarize our findings in a single model, given that dsx and fru have different roles on 3 populations of neurons and 2 types of social behaviors. We hope the new models make our core finding clear.

[Editors’ note: what follows is the authors’ response to the second round of review.]

Essential revisions:

Despite the notable revisions, e.g. valuable inclusion of quantitative anatomical analyses of the P1/pC1 neurons, there are remain issues that remain to be addressed.

We appreciate the reviewers’ comments on key issues. We would like to respond to each critique below.

1) Although the cover letter refers to Figure 1E of Hoopfer et al. (2015) to justify the view that the driver is P1/pC1-specific, Figure 1—figure supplement 1C of Hoopfer et al. (2015) appears to show that there are several additional neurons close to the lateral horn labelled by this split-Gal4 driver in both males and females. The main text of Hoopfer et al. (2015) states that "in addition to P1 neurons, the P1a spGAL4 driver labelled a minor and variable population of neurons (<4 per hemi- brain) with cell bodies in the lateral horn (LH) that were neither FruM-positive nor male-specific".

2) The same issue applies to the NP2631 ∩ dsxFLP driver. The text and cover letter reference Figure 4D of Koganezawa et al. (2016) to justify that their driver is P1/pC1-specific. However, the referenced figure appears to show a single descending axon per hemisphere is going out of the ventral end of the GNG. This axon crosses the midline at the neuropil region just dorsal to the oesophageal foramen, which is a typical neurite structure of another dsx-expressing neuronal cluster called pMN1 (see Figure S1 of Kimura et al., 2015). Similarly, in Figure S3A of Koganezawa et al. (2016), a descending axon of pMN1 is more obvious in this female specimen (pMN1 is present in both sexes with dimorphic neurite morphology, according to Figure S1 of Kimura et al., 2015). I would encourage the authors to show the expression patterns in the VNC. Are there any cells in this tissue labelled by the NP2631 ∩ dsxFLP driver?

The above two Points 1 and 2 reflect concerns on the specificity of driver lines used. To address ambiguities that arise from this, unless there are additional strong arguments in support of the authors' interpretations, ideally all behavioural assays should be re-done under the conditions of P1/pC1-specific activation using either further intersection or a stochastic/mosaic approach.

We certainly share the concern over the specificity of GAL4 drivers. When we attempted to make a causal link between the manipulated neurons and observed behavioral phenotypes, it is critical to know in which cells the proteins that manipulate neural activities (effector proteins) are present. This is why we chose to directly visualize the protein (CsChrimson:tdTomato) that serves as the effector of neuronal manipulation. As we wrote in our main text in “Sex-determining genes distinctly regulate courtship capability and preference toward females through sexually dimorphic neurons”, subsection “The target fly’s sex affects the function of social behavior-promoting neurons”, this should eliminate the necessity to deduce the expression pattern of the effector protein that cannot be visualized based on expression pattern of a reporter protein from other UAS elements. The same GAL4 line shows a dramatically different expression pattern when combined with different UAS elements (see Pfeiffer et al., 2010 and 2012 for examples), requiring caution when extrapolating the expression of reporter proteins (GFP for both Hoopfer et al., 2015: Figure 1—figure supplement 1C, and Koganezawa et al., 2016: Figure 4D) to deduce expression of untagged, and therefore invisible, effector proteins (dTRPA1). Such interpretation is widely accepted in numerous publications.

Keeping this in mind, we first wish to reiterate that we did not detect any cell bodies other than P1a neurons when examining the expression pattern of CsChrimson:tdTomato driven by the P1a split GAL4 from any brain (Figure 6A-G and Figure 5—figure supplement 1C, D).

In contrast, we occasionally detected a CsChrimson:tdTomato-expressing neuron outside of P1/pC1 cluster of NP2631 ∩ dsxFLP neurons (which in contrast were found in all male brains examined: shown in Figure 3C-K, Figure 2—figure supplement 2A1). We apologize for not mentioning the presence of this neuron in our initial submission. As shown in the new Figure 2—figure supplement 2A2, this neuron projects contralaterally and to the VNC. We appreciate the reviewer’s eye to identify this neuron in Figure 4D of Koganezawa et al., 2016, the image quality of which was too poor for us to discern a descending projection from noise, and to characterize this as “pMN1” neuron described in Supplementary Figure S1 of Kimura et al., 2015, without an image from VNC. In our samples, this descending neuron seems to have dense arborization in the ventromedial neuropil (nomenclature based on Ito et al., 2016), which is missing in examples shown in the above-mentioned Kimura et al., 2015.

However, the presence of this descending neuron is not consistent enough to account for behaviors induced by the optogenetic stimulation of NP2631 ∩ dsxFLP neurons. The descending neuron is labeled only in 35% of the male brains analyzed in Figure 2—figure supplement 2D1. These ratios are significantly lower than the ratio of the flies that showed LED stimulation-dependent lunges (Figure 2—figure supplement D2) or wing extensions (Figure 2—figure supplement 2D3). While the contribution of the descending neuron on social behaviors cannot be excluded, these observations strongly argue that they cannot account for the behavioral effect that we observed as a result of the stimulation of NP2631 ∩ dsxFLP neurons.

We therefore conclude that we found little evidence that re-doing behavioral experiments using “further intersection” will change our conclusions. While we agree that additional neurons may be labeled by using different UAS reporter elements, the combination of the GAL4 driver and UAS effector we used in this manuscript provides sufficient specificity to maintain that “P1/pC1” clusters labeled by NP2631 and dsxFLP are causally responsible for the behaviors we observed. Of course, the specificity of observed expression depends on other technical limitations such as the antibody sensitivity, detection threshold of the microscope, and so on, but these limitations also apply to other published works.

We added a paragraph that summarizes our statements above in “Sex-determining genes distinctly regulate courtship capability and preference toward females through sexually dimorphic neurons”, subsection “NP2631 ∩ dsxFLP neurons contain subpopulations of the P1/pC1 cluster that promote both courtship and aggression”. Also, we speculate that both (P1/pC1 cluster of) NP2631 ∩ dsxFLP neurons and P1a neurons can be further subdivided into distinct neuronal classes, as we mentioned in our discussion (“Sex-determining genes distinctly regulate courtship capability and preference toward females through sexually dimorphic neurons”, subsection “Uncovering functional heterogeneity of social behavior-controlling neurons”).

3) It still remains unclear how fru function is dissociated from that of dsx in the sexual differentiation of P1/pC1 neurons. To clarify potentially distinct contributions of fru and dsx to this developmental process, it appears that dsx mutant analyses are required. Have these been attempted or done?

We are a little perplexed by this comment. As we explained in the schematics of Figure 3B, our 4 genotypes allow us to dissociate the contribution of dsx isoforms and fru isoforms on circuit formation and behavior. For instance, if the male-like phenotype appears in the dsxM/fruF genotype, but not in the dsxF/fruM genotype (as is the case for NP2631 ∩ dsxFLP neuronal morphology), we can certainly conclude that it is the dsx isoform, and not the fru isoform, that controls the phenotype. We are not sure exactly what alternative possibilities the reviewers think there would be, and how additional experiments with dsx mutants would help differentiate such possibilities. While dsx mutants (especially the one that forces specific isoform splicing, as described in Nagoshi and Baker, 1990) will likely provide additional evidence, we do not think that such experiments are necessary for our conclusions.

Note that we did not attempt to investigate whether it is the lack of dsxM, or the presence of dsxF, that is responsible for the phenotype. We acknowledge that we cannot differentiate these two possibilities using our 4 genotypes (see “Sex-determining genes distinctly regulate courtship capability and preference toward females through sexually dimorphic neurons”, subsection “An organismal sex and a cellular sex” and “Layered roles of fruitless isoforms in specification and function of male-type aggression-promoting neurons in Drosophila”, subsection “fru specifies a neural circuit for sexually dimorphic aggressive actions”), but it is not the question we tried to answer.

4) The experiments use fruM, fruF and single-isoform fru mutant alleles for behavior and attribute the behavioural phenotypes to P1/pC1 (this paper) and Tk-GAL4FruM neurons (the other paper, titled “Layered roles of fruitless isoforms in specification and function of male aggression-promoting neurons in Drosophila”). Given that ru isoforms are expressed in other neurons in the CNS and the PNS, surely the effects of Fru isoform on behavior in P1/pC1 and Tk-GAL4FruM neurons can only really be assessed by looking at animals were the specific neurons are mutant for fru (e.g. fru isoform-specific mutant MARCM clones, UAS-Cas9 system, or fru-RNAi)? Otherwise, the data would need to be interpreted with obvious caveats.

As we had acknowledged in the “An organismal sex and a cellular sex” section of the Discussion of “Sex-determining genes distinctly regulate courtship capability and preference toward females through sexually dimorphic neurons”, we had been aware that our experiment does not clarify whether dsx and fru acts cell-autonomously or through other parts of the nervous system. While the molecular mechanism by which dsx and fru specify neuronal sexual dimorphism is an important question, we would also like to clarify that “the effects of Fru isoform on behavior in P1/pC1 and Tk-GAL4FruM neurons (in the other manuscript, titled “Layered roles of fruitless isoforms in specification and function of male aggression-promoting neurons in Drosophila”)” is not the question we aimed to answer.

Based on the fact that dsx and fru are both transcription factors, a straightforward prediction is that these genes work cell-autonomously, but it is certainly possible that dsx and fru in other neurons direct the morphogenesis or behavioral impact of P1a and NP2631 ∩ dsxFLP neurons. Such a cell non-autonomous role for dsx has been previously reported in the gonad and the genital imaginal disc, but we are not aware of any published results suggesting that dsx or fru functions cell non-autonomously in the nervous system. In fact, one of the reasons why we did not address this question is because none of the approaches the reviewer suggested (RNAi, dCas9, MARCM) can immediately differentiate these two possibilities. We stated this limitation in subsection “An organismal sex and a cellular sex” (“Sex-determining genes distinctly regulate courtship capability and preference toward females through sexually dimorphic neurons”), and will clarify our arguments below.

An overlooked limitation of RNAi, dCas9, or MARCM approaches is that results are immediately interpretable only if it recapitulates the phenotype of the systemic mutant. To further elaborate:

1) Both RNAi and dCas9 are limited by the temporal as well as spatial expression of the GAL4 driver. This means that a GAL4 driver that turns on late may not be effective in suppressing gene activity at the critical developmental phase. For instance, dsx-dependent sexual dimorphism of leg gustatory receptor neurons begins to be established as early as 8 hours after pupal formation (Mellert, Robinett, and Baker, 2012). This type of situation leaves open the possibility that knock-down of fru using P1a-GAL4 or NP2631 may occur too late to impact the fate of the neurons. We do not say it is too late: we merely say that we cannot interpret the “no phenotype” result (which would be the case when the RNAi turns on too late to impact the cell-autonomous role of fru, and when fru is indeed required in non-cell-autonomous manner) unless we characterize the expression patterns of a given GAL4 used throughout development. A GAL4 line that fulfills temporal as well as spatial specificity for the targeted cells will be exceedingly difficult to identify.

2) Another seldom verified limitation of RNAi and dCas9 is its efficacy. Both approaches rely on the specificity of 17-23 base homology of RNA substrates, and how efficiently the double-stranded RNA can recruit effector proteins (RISC complex for RNAi, Cas9 for dCas9). Most likely, they provide only partial knockdown of the targeted gene. As we showed in “Layered roles of fruitless isoforms in specification and function of male-type aggression-promoting neurons in Drosophila”, a hypomorph of fru can result in an intermediate phenotype (see Figure 1—figure supplement 3E of that submission). Such incomplete phenotypes can complicate the interpretation when the phenotypic differences of systemic mutants are quantitative rather than qualitative, as is the case for NP2631 ∩ dsxFLP neurons (Figure 3G-N) and P1a neurons (Figure 6E-I)).

3) A limitation of MARCM that evades scrutiny is that the mutant cell clones may be generated in GAL4 negative cells as well as GAL4 positive cells. Only GAL4 positive cells are visible, but this does not mean that they are the only mutant cells. This leaves a possibility that a heat-shock condition that generates NP2631 ∩ dsxFLP or P1a neuronal clones may systematically generates other clones together, due to similar developmental time course (note that MARCM works during cell division). This means that MARCM data is inherently correlational, not causal. In fact, a well-cited paper using MARCM for this type of analysis (Kimura et al., 2008) does not exclude the possibility that behaviorally relevant neurons are actually fruNP21-negative neurons (such as dsx+/fru- subclass within P1/pC1 neurons) that become tra mutant clones together with the fruNP21-expressing “P1” neurons. Considering the shared lineage of this cluster (the DM4 lineage described in Ren et al., 2016), this is an alternative scenario that remains to be addressed.

4) The other practical limitation of the MARCM approach is the difficulty of characterizing labeled cell types, especially in the P1/pC1 cluster. Our data (Figure 5, Figure 5—figure supplement 1) as well as previous publications (Costa et al., 2016; Zhang et al., 2018) point to the heterogeneity of this neuronal population (discussed in “Uncovering functional heterogeneity of social behavior-controlling neurons”). Morphology of these neurons are often very similar to each other, and it will be extremely challenging to accurately classify labeled neurons especially if multiple types are simultaneously labeled.

These issues can be properly addressed with adequately validated tools (such as truly effective RNAi or gRNA, GAL4 lines with verified temporal and spatial expression patterns, etc.). However, we think conclusive evidence will require a level of effort that amounts to a separate project. In fact, we think it is highly valuable to characterize the neuronal morphology and function in the systemic mutants as we did in our manuscript, precisely because this will be an important foundation on which we can further investigate the molecular mechanism underlying the roles of dsx and fru on sexual dimorphisms (“Sex-determining genes distinctly regulate courtship capability and preference toward females through sexually dimorphic neurons”, subsection “An organismal sex and a cellular sex”). We hope reviewers will understand both the merit of our current study, as well as the challenges to obtain definitive conclusions regarding the cell-autonomy of dsx and fru.

5) The paper should include better images of Dsx immunolabeling (Figure 2—figure supplement 2E) as the current one is of very poor quality.

We performed additional immunohistochemistry against DsxM and FruM using a new batch of antibodies, which gave us superior signal to noise. We thank Dr. Michael Perry of UCSD for providing us with these antibodies. The newly captured images clearly show that almost all of NP2631 ∩ dsxFLP neurons (Figure 2—figure supplement 2B, C), and all P1a neurons (Figure 5—figure supplement 1C, D), express DsxM. We updated the figure panels and texts accordingly.

https://doi.org/10.7554/eLife.52701.sa2

Article and author information

Author details

  1. Kenichi Ishii

    Molecular Neurobiology Laboratory, Salk Institute for Biological Studies, La Jolla, United States
    Contribution
    Investigation
    Contributed equally with
    Margot Wohl
    Competing interests
    No competing interests declared
  2. Margot Wohl

    1. Molecular Neurobiology Laboratory, Salk Institute for Biological Studies, La Jolla, United States
    2. Neuroscience Graduate Program, University of California, San Diego, San Diego, United States
    Contribution
    Software, Investigation
    Contributed equally with
    Kenichi Ishii
    Competing interests
    No competing interests declared
  3. Andre DeSouza

    1. Molecular Neurobiology Laboratory, Salk Institute for Biological Studies, La Jolla, United States
    2. Neuroscience Graduate Program, University of California, San Diego, San Diego, United States
    Contribution
    Investigation
    Competing interests
    No competing interests declared
  4. Kenta Asahina

    1. Molecular Neurobiology Laboratory, Salk Institute for Biological Studies, La Jolla, United States
    2. Neuroscience Graduate Program, University of California, San Diego, San Diego, United States
    Contribution
    Conceptualization, Resources, Data curation, Formal analysis, Supervision, Funding acquisition, Validation, Investigation, Visualization, Methodology, Project administration
    For correspondence
    kasahina@salk.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-6359-4369

Funding

National Institute of General Medical Sciences (GM119844)

  • Kenichi Ishii
  • Margot Wohl
  • Kenta Asahina

Naito Foundation

  • Kenichi Ishii

Japan Society for the Promotion of Science

  • Kenichi Ishii

Mary K. Chapman Foundation

  • Margot Wohl

Rose Hills Foundation

  • Margot Wohl

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank Drs. David Anderson, Gerald Rubin, and Barret Pfeiffer for sharing unpublished transgenic Drosophila strains with us; Drs. Stephen Goodwin and Daisuke Yamamoto for other Drosophila strains; Eyrun Eyjolfsdottir and Dr. Pietro Perona for developing and improving the FlyTracker program; Dr. Michael Perry for sharing with us anti-DsxM and anti-FruM antibodies; Dr. Yong Wan for development and support for FluoRender, Pavan Nayak and Vivian Shaw for their help on the development of behavior classifiers, Dr. Samuel Pfaff for sharing the Olympus FV-1000 confocal microscopy with us, Drs. Eiman Azim, Weizhe Hong, Samuel Pfaff, Carla Shatz, John Thomas, and members of the Asahina lab for critical comments on the manuscript, and David O’Keefe for scientific editing on the manuscript. The antisera nc82 (anti-BRP), developed by E Buchner, were obtained from the Developmental Studies Hybridoma Bank, created by the NICHD of the NIH and maintained at The University of Iowa, Department of Biology, Iowa City, IA 52242. Stocks obtained from the Bloomington Drosophila Stock Center (NIH P40OD018537) were used in this study. This work was also made possible in part by software funded by the NIH: FluoRender: Visualization-Based and Interactive Analysis for Multichannel Microscopy Data, 1R01EB023947-01 and the National Institute of General Medical Sciences of the National Institutes of Health under grant number P41 GM103545-18.

Senior Editor

  1. K VijayRaghavan, National Centre for Biological Sciences, Tata Institute of Fundamental Research, India

Reviewing Editor

  1. Mani Ramaswami, Trinity College Dublin, Ireland

Publication history

  1. Received: October 12, 2019
  2. Accepted: April 3, 2020
  3. Version of Record published: April 21, 2020 (version 1)

Copyright

© 2020, Ishii et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 1,482
    Page views
  • 225
    Downloads
  • 2
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Genetics and Genomics
    Daniel Wells et al.
    Research Article Updated

    During meiosis, homologous chromosomes pair and recombine, enabling balanced segregation and generating genetic diversity. In many vertebrates, double-strand breaks (DSBs) initiate recombination within hotspots where PRDM9 binds, and deposits H3K4me3 and H3K36me3. However, no protein(s) recognising this unique combination of histone marks have been identified. We identified Zcwpw1, containing H3K4me3 and H3K36me3 recognition domains, as having highly correlated expression with Prdm9. Here, we show that ZCWPW1 has co-evolved with PRDM9 and, in human cells, is strongly and specifically recruited to PRDM9 binding sites, with higher affinity than sites possessing H3K4me3 alone. Surprisingly, ZCWPW1 also recognises CpG dinucleotides. Male Zcwpw1 knockout mice show completely normal DSB positioning, but persistent DMC1 foci, severe DSB repair and synapsis defects, and downstream sterility. Our findings suggest ZCWPW1 recognition of PRDM9-bound sites at DSB hotspots is critical for synapsis, and hence fertility.

    1. Epidemiology and Global Health
    2. Genetics and Genomics
    Jacob A Tennessen et al.
    Short Report Updated

    Schistosomiasis is a debilitating parasitic disease infecting hundreds of millions of people. Schistosomes use aquatic snails as intermediate hosts. A promising avenue for disease control involves leveraging innate host mechanisms to reduce snail vectorial capacity. In a genome-wide association study of Biomphalaria glabrata snails, we identify genomic region PTC2 which exhibits the largest known correlation with susceptibility to parasite infection (>15 fold effect). Using new genome assemblies with substantially higher contiguity than the Biomphalaria reference genome, we show that PTC2 haplotypes are exceptionally divergent in structure and sequence. This variation includes multi-kilobase indels containing entire genes, and orthologs for which most amino acid residues are polymorphic. RNA-Seq annotation reveals that most of these genes encode single-pass transmembrane proteins, as seen in another resistance region in the same species. Such groups of hyperdiverse snail proteins may mediate host-parasite interaction at the cell surface, offering promising targets for blocking the transmission of schistosomiasis.