Specific subgroups of Fru^M neurons control sexually dimorphic patterns of aggression in Drosophila melanogaster

Abstract

A great challenge facing neuroscience is to understand how genes, molecules, cells, circuits, and systems interact to generate social behavior. Fruit flies (Drosophila melanogaster) offer a powerful model system to address questions of this magnitude. These animals display genetically specified, sexually dimorphic patterns of fighting behavior via sex-specific splicing of the fruitless gene. Here, we show that sexually dimorphic behavioral patterns displayed during aggression are controlled by specific subgroups of neurons expressing male forms of fruitless proteins (Fru^M). Using the GAL4/UAS system to manipulate transformer expression, we feminized or masculinized different populations of neurons in fly nervous systems. With a panneuronal elav-GAL4 driver, male patterns of fighting behavior were transferred into females and female patterns into males. We screened 60 Gal4 lines that express the yeast transcription factor in different patterns in fly central nervous systems and found five that showed abnormal same-sex courtship behavior. The sexually dimorphic fighting patterns, however, were completely switched only in one and partially switched in a second of these lines. In the other three lines, female patterns of aggression were seen despite a switch in courtship preference. A tight correspondence was seen between Fru^M expression and how flies fight in several subgroups of neurons usually expressing these proteins: Expression is absent when flies fight like females and present when flies fight like males, thereby beginning a separation between courtship and aggression among these neurons.

Throughout the animal kingdom, aggression is commonly seen in competition for resources such as food, territory, or mates. Although described many years ago (1–4), interest has only recently focused on the use of Drosophila melanogaster as a model system for the study of aggression (5–12). The first time two male or female adult flies meet in competition for resources, they are fully capable of generating, recognizing, and appropriately responding to the complex array of behavioral patterns seen during agonistic encounters (8, 9). Aggression, therefore, is a genetically specified behavior in fruit flies, and it is challenging to ask how gene cascades can construct species- and sex-specific patterns of behavior of such complexity.

Like courtship behavior in adult flies (13–16), aggression in D. melanogaster can be subdivided into easily recognizable patterns (modules) that transition from one to the other depending on social interactions. Our previous studies using wild-type Canton-S flies demonstrated that such modules are sexually dimorphic (8, 9). Certain components of fighting behavior like “fencing,” “approach,” and “retreat” are observed during fights in both sexes. Others are sexually dimorphic and can be used to characterize fruit fly fighting behavior as “male-like” or “female-like.” Thus, male flies commonly perform rapid “lunge” movements during fights in which they rise up on their hind legs and snap down on the opponent. Less often, males will “hold” an opponent or rise on their hind legs and use their forelegs to “box.” Females show patterns of behavior that resemble lunging, but they involve less elevation of the body and they do not hold or box during fights. Females, instead, show horizontal shove and headbutt movements that rarely are seen in male fights (7, 9, 12). How such behavioral diversity is encoded at the cellular and circuit levels remains unknown, although recent studies have demonstrated that the fruitless (fru) gene of the sex-determination cascade is both necessary and sufficient to determine whether fruit flies fight like males or females (12).

Sex is cell-autonomous in Drosophila, and available genetic tools allow us to partially or completely change the sex of part or all of the fruit fly nervous system to examine the effects on behavior. In these studies, we used various GAL4 drivers to feminize neurons in male or masculinize neurons in female nervous systems through expression of transformer (tra), a splicing factor of the sex determination cascade (17), using UAS-tra^F (the female form of transformer) in males or reducing or eliminating the expression of tra by using UAS-tra^IR (transformer dsRNA) in females. With a panneuronal GAL4 driver (elav-GAL4), patterns of fighting behavior were switched: Males fought like females and females fought like males. We then screened 60 different lines expressing GAL4 in the nervous system and found same-sex courtship behavior after tra manipulation in five of these. In only one of the five, however, patterns of aggression also were switched. Immunohistochemical examination of the distribution of the male-specific forms of the fru protein (Fru^M) showed that several clusters of Fru^M-expressing neurons showed positive immunostaining only in flies in which masculinized females fought like males. The results suggest that specific subgroups of Fru^M expressing neurons are likely to be important determinants of the sexual dimorphism observed in fighting behavior in D. melanogaster.

Results

Sex-specific patterns of fighting behavior are neuronally based. In the first studies, we used the panneuronal elav-GAL4 driver to express either the active female form of transformer (tra^F) in male nervous systems or a transformer dsRNA (tra^IR) to eliminate tra^F in female nervous systems. No major morphological differences were observed in these flies except in feminized males, where small changes in the abdominal banding patterns and in the genitalia were observed as reported (18). Pairs of feminized male or masculinized female flies were placed in our standard fighting chambers. The patterns of behavior observed were scored and compared with the patterns typically seen in fights between pairs of Canton-S males or females. In addition, because high levels of courtship behavior were seen in the experimental lines, this too was scored by measuring wing extension and attempted copulation. Other patterns like orientation do not clearly distinguish between courtship and aggression, and these were not scored.

In Canton-S flies, large differences are seen between males and females in the easily recognizable midintensity behavioral patterns: lunge, shove, and headbutt (Fig. 1A). Much higher numbers of lunges are seen in males than females. Comparable differences, but in the reverse direction, are seen in shove and headbutt patterns in females. Female flies do display small numbers of lunge-like hopping behaviors in which they slightly elevate their bodies and move forward (9), but these are easily differentiated from the lifting of forelegs high off the surface and rapid snapping down on an opponent shown by males. Other behavioral components like boxing are rare in male fights and are never seen in female fights. Finally, extended-wing threats are seen in male fights, and these too are not observed in female fights. Movie clips of individual behavioral components are provided as supporting information (SI) Movies 1 a–d. Courtship behavior is rare between Canton-S males and is never seen in female pairings (Fig. 1B). In additional to differences in the frequency of the individual behavioral components, the dynamics of male and female fights also differ. Male flies defend the entire territory (food surface) and engage in repeated lunging to drive away opponents, whereas female flies usually share the territory (9, 12). Movie clips of sample male and female fights that demonstrate these differences are included as SI Movies 1 e and f.

Fig. 1.

Behavioral patterns seen during courtship and aggression in wild-type and mutant flies. (A) Percentages of sexually dimorphic behavioral patterns seen during aggressive interactions in Canton-S, masculinized female and feminized male flies and in a Fru^F mutant background. Masculinized females (elav-GAL4/UAS-tra^IR) show male fighting patterns similar to Canton-S males, whereas feminized males (elav-GAL4/UAS-tra^F) show female fighting patterns similar to Canton-S females. No changes in fighting patterns were observed in masculinized females in a Fru^F background (elav-GAL4/UAS-tra^IR; fru^F/fru^F), confirming that Fru^M expression is responsible for the behavioral phenotypes we observed. “Lunge,” “Shove,” and “Headbutt” are the main components used to distinguish male-like and female-like patterns of fighting behavior. (B) Percentages of aggressive, courtship, and mixed interactions between pairs of flies. Both masculinized females and feminized males show increased same-sex courtship behavior (i.e., wing extension, singing, licking, or attempted copulation) when compared with wild-type flies. (C) Patterns of aggression shown by females masculinized by using five selected GAL4 drivers. With the l (3)31-GAL4 driver male fighting patterns are observed, others (c1003-, 60IIA-, 227-GAL4) show female fighting patterns, and the 1407-GAL4 driver shows intermediate patterns of fighting behavior. (D) Percentages of interactions masculinized females of the five selected GAL4 driver lines spend in aggression, courtship, and mixed interactions. (n) represents the total number of interactions.

In the experimental lines, the fighting patterns were completely switched (Fig. 1A and SI Movies 1 g and h). Feminized males (elav-GAL4;UAS-tra^F) show mainly shove and headbutt in fights instead of lunge, whereas masculinized females (elav-GAL4;UAS-tra^IR) lunge at frequencies and with patterns comparable with those seen in control males. The two transgenic lines also display male–male and female–female courtship behavior (Fig. 1B). If three or more transgenic male or female flies are put together, chaining behavior is seen.

Expression of Fru^M is necessary and sufficient to elicit male forms of courtship behavior in female flies and to transform patterns of aggression from female-like to male-like (12, 19–21). We anticipated, therefore, that the behavioral phenotypes resulting from the above experimental manipulation of tra were due to changes in Fru^M expression. Because tra is involved in splicing doublesex (dsx) as well as fru into male and female forms, we felt it necessary to confirm that the observed effects were mainly due to altering the expression of Fru^M. For this purpose, we masculinized female flies that cannot express Fru^M (elav-GAL4;UAS-traIR;fru^F/fru^F) and found no changes in courtship behavior or in the patterns of aggression (Fig. 1 A and B).

Sexually Dimorphic Patterns of Aggression and Courtship Can Be Separated.

To ask whether distinct subgroups of neurons might be concerned with courtship and aggression, we screened a library of 60 GAL4 lines, all of which were reported to show different expression patterns in the nervous system (SI Table 2). When these lines were crossed with UAS-tra^IR, we found abnormal behavioral phenotypes in five of the lines. Expression of tra^IR in female flies from these crosses led to female–female courtship behavior. Male-specific patterns of courtship behavior were seen in these masculinized females (Fig. 1D), and when three or more of the progeny of these crosses were placed together in a chamber, chaining behavior was observed (SI Movie 2).

Although all five lines of masculinized females displayed abnormal female–female courtship behavior, different results were obtained in examining their patterns of aggression. Expression of tra^IR in female flies in crosses with the c1003, 60IIA, or 227-GAL4 drivers did not change the normal female patterns of fighting behavior (Fig. 1C). Instead, these lines showed the shove and headbutt patterns commonly seen in fights between pairs of wild-type females and not the lunge pattern seen in male fights. To illustrate this, movie clips of fights between pairs of 227-GAL4 masculinized females are included as SI Movie 3a. In contrast, females that had been masculinized by using the l (3)31-GAL4 driver showed clear male fighting patterns using lunge rather than shove and headbutt in fighting behaviors (SI Movie 3b). Female flies that expressed tra^IR using the 1407-GAL4 driver showed an intermediate fighting pattern, with similar levels of lunge, shove, and headbutt seen. No changes in patterns of aggression were observed in any control lines generated by using the various GAL4 driver lines to express red fluorescent protein (driver lines crossed with UAS-RedStinger, data not shown). All male progeny of crosses of these five lines with UAS-tra^F flies showed male–male courtship behavior, but because they also all showed male patterns of aggression (data not shown), these crosses were not examined further.

Identification of Fru^M-Expressing Neurons Important in Aggression.

With the experimental manipulations carried out above, we expected to see Fru^M expression in the nervous systems of masculinized female flies but expected no Fru^M expression in feminized males. In the brains of the masculinized females (elav-GAL4;UAS-tra^IR) clear Fru^M expression was seen (Fig. 2A), and most of the anticipated Fru^M expressing neuronal clusters in the central brain region were identifiable (for nomenclature see refs. 22 and 23). The general expression levels decreased after early pupal stages, but Fru^M-staining neurons still were detectable in newly eclosed adults. What we did not anticipate, however, was that we would find similar Fru^M expression patterns in the brains of feminized males (elav-GAL4;UAS-tra^F) (Fig. 2C). Thus, despite the clear switch in behavioral phenotypes (Fig. 1 A and B), ectopic expression of Tra^F using the elav-GAL4 driver was insufficient to completely shut down Fru^M expression in feminized males. A more complete knockdown of Fru^M expression was achieved when we used the Act5c-GAL4 driver, suggesting that a stronger driver than elav-GAL4 was required to eliminate Fru^M expression (data not shown). Because feminized males and masculinized females display switched-sex fighting patterns and because fru splicing is responsible for this dimorphism, it is interesting and particularly valuable for us that some Fru^M expression still is seen in the feminized males. It allows us to compare this expression with Fru^M seen in masculinized females and wild-type males, to begin to identify candidate clusters of neurons important in aggression. In examining the numbers of Fru^M-positive neurons in pupal brains, seven clusters of neurons (fru-aSP2, fru-mAL, fru-AL, fru-mcAL, fru-P, fru-PrMs, and fru-MsMt) showed significant reductions of Fru^M-positive cells in feminized males compared with masculinized females and wild-type males (SI Table 3). We used pupal brains in these studies because that is the time when sexual behavioral patterns are specified (24), and that also is when the highest levels of mRNA for Fru^M proteins are found in development (22). As flies mature, the numbers of Fru^M-positive cells decrease in both feminized males and masculinized females. However, significant reductions in cell number still can be found in some of the seven clusters.

Fig. 2.

An illustration of the Fru^M expression patterns in elav-GAL4/UAS-tra^IR masculinized female and elav-GAL4/UAS-tra^F feminized male pupal brains. Immunostaining of the posterior brain regions and adult brains are shown in SI Fig. 4. (A) Anterior brain regions of 48-h-old pupae of elav-GAL4/UAS-tra^IR masculinized females were dissected, fixed, and immunostained (see Methods) for Fru^M (green) and NC82 (magenta). (B). Schematic drawing of elav-GAL4/UAS-tra^IR masculinized female brains. (C) Anterior brain regions of 48-h-old pupae of elav-GAL4/UAS-tra^F feminized males. (D) Schematic drawing of elav-GAL4/UAS-tra^F feminized male brains. Reductions in Fru^M cell number are seen in fru-mAL and fru-aSP2 clusters in feminized males. (Scale bars, 50 μm.)

Next, we examined Fru^M expression patterns in masculinized female flies derived from crosses between the five GAL4 lines showing courtship abnormalities and UAS-tra^IR flies (SI Fig. 5). This is quantified in Table 1, where the numbers of Fru^M expressing neurons in pupae are measured. Groups 1–3 of Table 1 are the seven Fru^M clusters that showed reductions in cell numbers in males feminized by using the elav-GAL4 driver. Groups 4 and 5 are clusters that showed no differences in cell numbers between masculinized females and feminized males in the elav-GAL4 studies.

Table 1.

Cell counts of individual Fru^M-expressing neuron clusters in 48-h-old pupae by using selective GAL4 driver lines

Group	Cell cluster	Canton-S (M)	l(3)31-GAL4/UAS-traIR (F)	1407-GAL4/UAS-traIR (F)	c1003-GAL4/UAS-traIR (F)	60IIA-GAL4/UAS-traIR (F)	227-GAL4/UAS-traIR (F)
1	fru-mAL	23 ± 2	25 ± 3	23 ± 5	17 ± 4	9 ± 4	0
	fru-PrMs	73 ± 7	68 ± 6	64 ± 6	54 ± 6	33 ± 3	43 ± 6
2	fru-aSP2	42 ± 5	37 ± 6	38 ± 5	26 ± 6	12 ± 2	1 ± 1
	fru-mcAL	31 ± 1	23 ± 4	25 ± 3	16 ± 4	12 ± 2	4 ± 2
3	fru-AL	52 ± 6	40 ± 3	45 ± 8	37 ± 5	39 ± 5	31 ± 6
	fru-P	56 ± 7	31 ± 5	32 ± 7	31 ± 5	38 ± 8	20 ± 4
	fru-MsMt	38 ± 8	30 ± 4	33 ± 4	28 ± 6	18 ± 2	27 ± 6
4	fru-aSP1	12 ± 2	12 ± 2	11 ± 3	11 ± 3	10 ± 2	0
	fru-aSP3	29 ± 5	31 ± 4	31 ± 3	24 ± 2	28 ± 5	18 ± 4
	fru-SG	11 ± 2	3 ± 1	6 ± 2	1 ± 1	2 ± 1	1 ± 1
	fru-pSP1	9 ± 2	3 ± 2	5 ± 2	3 ± 2	4 ± 1	1 ± 1
	fru-pSP2	18 ± 3	16 ± 1	15 ± 1	12 ± 2	13 ± 2	2 ± 3
	fru-pL	12 ± 2	10 ± 1	8 ± 4	5 ± 3	6 ± 3	1 ± 1
	fru-Pr	17 ± 3	16 ± 2	13 ± 2	12 ± 2	8 ± 2	10 ± 2
	fru-MtAb + AB	139 ± 14	40 ± 5	46 ± 4	47 ± 8	38 ± 4	37 ± 3
5	fru-Lv	17 ± 3	16 ± 2	15 ± 2	18 ± 3	17 ± 3	16 ± 1
	Fighting pattern	Male	Male	Intermediate	Female	Female	Female
	Courtship	+	+	+	+	+	+

Brains of masculinized female flies of each of the different lines were stained with Fru^M antibodies, and the numbers of cells in each Fru^M expressing cluster were counted. Shown is the average number of Fru^M-positive cells in each neuron cluster ± standard deviation (N = 10). Groups highlighted in bold are Fru^M clusters that show significant differences in cell number (one-way ANOVA, P < 0.01) between masculinized females and feminized males (see Fig. 1 and SI Table 3). mAL, above the medial antennal lobe; PrMs, between pro- and mesothoracic ganglia; aSP2, anterior region of superior protocerebrum; mcAL, anterior to mechanosensory neuropile of the antennal lobe; AL, cells above antennal lobe; P, posterior region; MsMt, between meso- and metathoracic ganglia; see ref. 22 for other clusters. M, male; F, female.

Among the Fru^M-expressing clusters in the first three groups, the cell numbers in group 1 (fru-mAL and fru-PrMs) showed the strongest correlation with the male-aggression pattern phenotype. Both l (3)31-GAL4 and 1407-GAL4 masculinized females fought like males (only partially in the 1407 group), and these showed similar numbers of Fru^M-immunostaining neurons as Canton-S males. In contrast, the c1003-, 60IIA-, and 227-GAL4 masculinized females fought like females and showed reduced numbers of Fru^M-positive cells in Group 1 clusters. The close correlation between the presence or absence of Fru^M staining in the fru-mAL and fru-PrMs clusters of neurons and whether flies fight like males or females suggests that these two groups may play essential roles in how flies fight. Within Group 2 (fru-aSP2 and fru-mcAL), similar patterns were observed, but smaller reductions and greater heterogeneity in cell numbers were seen when compared with Canton-S males, suggesting that further experiments will be needed to clarify any role of these clusters in aggression. In contrast, cell numbers in group 3 (fru-AL, fru-P, and fru-MsMt) were similar in all five of the masculinized female lines, suggesting that these clusters serve no role in establishing patterns of aggression. Among all Fru^M clusters, only the fru-Lv group (group 5) showed no reduction in numbers of immunostained cells in any of the five lines. Because we observe female–female courtship in all lines, this raises the possibility that the fru-Lv cluster may be particularly concerned with courtship.

Females masculinized with the 1407-GAL4 driver showed bisexual patterns of aggression and same-sex courtship behavior. Fru^M expression levels in these females were similar to those seen with the l (3)31-GAL4 driver, the line that showed the clear switch to male patterns of aggression in the female progeny of crosses with UAS-fru^IR flies but only at the pupal stage. In adults, the numbers of cells showing Fru^M immunostaining with the 1407-GAL4 driver were similar to those seen with the three lines that showed no switch in aggression patterns (the c1003, 60IIA, and 227-GAL4 lines). The levels of Fru^M immunostaining decreased in all clusters of neurons as flies matured, and only females masculinized with the l (3)31-GAL4 driver had clear Fru^M expression upon eclosion as adults. Fru^M expression levels in the other four lines were weak to nondetectable in newly emerged adults, and no Fru^M immunostaining was seen in 5-day-old adults derived from any of the five lines (data not shown). Possibly a more rapid fall off of Fru^M expression during development in progeny of crosses with the 1407-GAL4 driver when compared with the l (3)31-GAL4 driver could account for the bisexual fighting patterns.

Analysis of GAL4 Expression Patterns.

Because the expression of tra^IR in females using the five GAL4 drivers led to changes in behavioral phenotypes in courtship and/or aggression, we asked next whether common features in the expression patterns of the driver lines might be correlated with the behavioral changes. To visualize the GAL4 expression patterns of the five lines, we crossed the various drivers with a reporter line (UAS-RedStinger), a nuclear localized red fluorescent protein, to identify the GAL4-expressing neurons in adult male brains. In the same tissues, we searched for possible colocalization of GAL4 with Fru^M. The GAL4 expression patterns were widespread and varied greatly from driver to driver (SI Fig. 6 A–E), but no obvious shared expression patterns were observed. Even more surprising, however, was the lack of any significant overlap between the GAL4 expression patterns and the Fru^M immunostaining patterns.

A possible explanation for this might be that the effects of tra^IR are latent. In fact, RNAi effects in certain cases, have been reported days after the initial synthesis or injection of the polynucleotide (25, 26). If the tra dsRNA is made at an early developmental stage in neurons or in neuronal precursor cells before Fru^M expression, it might influence patterns of gene expression at later stages in development when GAL4 no longer is being expressed by these cells. To test this hypothesis, we used the “flp-out” system, with the different GAL4 drivers used to drive the expression of FLP protein in flies carrying the Act5C(FRT.polyA)lacZ.nls reporter gene (27) (see Methods). Expression of FLP in GAL4-expressing neurons at any developmental stage will splice out the polyA sequence in these cells, resulting in continuous expression of a lacZ.nls reporter. This can be detected at later developmental stages whether or not GAL4 still is being made in those neurons. Using this approach, we found that all five GAL4 lines have broader and far more extensive expression patterns than were observed above (SI Fig. 6 F–J). Clear colocalization of lacZ with Fru^M immunostaining was found in almost all clusters of Fru^M-expressing neurons in all five lines. These results show that GAL4 was expressed in most of the Fru^M-positive neurons or in their precursor cells at developmental stages earlier than those examined above. No lacZ immunostaining was observed in controls without GAL4 drivers (data not shown).

When individual Fru^M-positive clusters were examined at higher magnification, most of the Fru^M-immunostained cells overlapped with the GAL4-expressing neurons, but a few exceptions were found (Fig. 3). No overlap of Fru^M and LacZ.nls staining was found in the fru-mAL and fru-aSP1 clusters in the 227-GAL4 line, suggesting that GAL4 never was synthesized in those neurons. This is consistent with the Fru^M staining pattern found in the 227-GAL4 masculinized females, where Fru^M expression is absent in the same two clusters of neurons (Table 1) and where females show normal female patterns of aggression (Fig. 1C). Both of these clusters show GAL4 expression with the l (3)31-GAL4 driver (Fig. 3 A–C), and females masculinized by using this driver show clear Fru^M expression (Table 1) and switched patterns of fighting behavior (Fig. 1C). In several other neuronal clusters (e.g., fru-aSP2), GAL4 expression is seen by using the 227-GAL4 driver and the flp-out technique, but no or very limited Fru^M expression is observed in masculinized females. This could result from inadequate levels of Tra^IR expression in those neurons or their precursors.

Fig. 3.

Colocalization of Fru^M (green) and β-galactosidase (magenta) staining reveals GAL4 expression at any time during development (see above text) in the l (3)31-GAL4 and 227-GAL4 lines. Images were obtained by crossing GAL4 drivers with w;UAS-FLP; Act5c(FRT.polyA)lacZ.nls reporter lines. White arrows indicate cell clusters that are Fru^M-positive but do not express GAL4. (Scale bars, 20 μm.)

Discussion

Aggression in Fruit Flies: A Second Model System for Exploring the Establishment of Sexually Dimorphic Patterns of Social Behavior in Nervous Systems.

Studies of courtship behavior in D. melanogaster and the subsequent identification of fru as a master gene involved in establishing the circuitries involved, has provided an elegant and important model with which to ask how genes, molecules, cells, circuits, and systems interact to generate social behavior (17, 19–21, 28–30). Recently, aggression was added to the list of genetically specified sexually dimorphic behaviors in fruit flies, and again, fru was shown to be both necessary and sufficient to determine how flies fight (9, 12). In this study focused on aggression, we began studies aimed at dissecting the neuronal circuitry underlying the behavior. Male and female flies compete for resources in same-sex pairings: some of the behavioral patterns observed during the resultant fights are similar; others are different. The present study demonstrates that the differences seen are neuronal in origin, because altering the sex of neurons alone through manipulation of the transformer gene, changes how flies fight. Male-selective behavioral components like lunge now are seen in female fights, and female selective components like shove and headbutt are seen in male pairings. However, not all of the components normally seen in fights are switched, nor are they changed to the same extent. For example, we saw no extended “wing threats” in flies with masculinized female brains. Although anatomical differences between males and females could partially explain such discrepancies, more than likely, genes in addition to fru and tra will be shown to be essential in establishing the entire repertoire of complex behavioral patterns observed during aggression.

By manipulating tra expression, which is required for splicing fru into male and female forms, we generate flies that fight like their counterparts of the opposite sex. A similar approach could have been used to attempt to manipulate fru expression directly, but it probably would not have worked in these studies because we saw no overlap in expression between the GAL4 lines used here and fru. In addition, misexpression of Fru^M in neurons that normally do not express these proteins can lead to changes in courtship behavior (22) and possibly also generate other behavioral phenotypes. In such cases, it would be difficult to untangle the role of different Fru^M clusters in establishing the sexually dimorphic patterns behavior seen during courtship and aggression.

Specific Fru^M-Expressing Neuronal Clusters Are Involved in the Control of Sexually Dimorphic Patterns of Aggression.

By using the panneuronal elav-GAL4 driver to manipulate tra expression, we switched the key aggressive behavioral patterns displayed by male and female flies. Recent studies using fru^M and fru^F mutant lines demonstrated that splicing of fru into male or female forms determines how flies fight (12, 19). The changes in behavioral patterns observed in fru mutants were similar to the changes we observed here with tra manipulation. Thus the manipulation of either gene leads to behavioral changes that are likely to be governed by fru-related circuitry. The control experiment in which we manipulated tra expression in flies that were incapable of making Fru^M (in a Fru^F genetic background) adds strong support to the notion that the effects we observe mainly are due to changes in Fru^M expression.

In examining Fru^M expression patterns, we found that only seven clusters of neurons showed significant differences in the numbers of stained cells between feminized males and masculinized females. Among the seven clusters of neurons showing reductions in numbers of Fru^M-stained cells with the elav driver, four (fru-aSP2, fru-mAL, fru-mcAL, and fru-PrMs) appear to be the most relevant to aggression (Table 1) based on studies with the five GAL4 driver lines that influence courtship and/or aggression. With the latter lines, in masculinized female flies that show male patterns of aggression, Fru^M must be expressed in these four clusters to change the aggression behavioral phenotype. Only weak or no Fru^M immunostaining is observed in the same groups of neurons in masculinized females that show normal female patterns of aggression. Despite a strong association between the numbers of Fru^M-expressing cells in these clusters and the switch in behavioral phenotypes, we cannot eliminate the possibility that patterns of aggression might be altered by only small numbers of the Fru^M-containing neurons within the clusters. At present, we cannot recognize individual neurons within clusters, and cell counts cannot tell us whether particular neurons are missing. Even so, these findings do identify which of the 20 clusters of Fru^M-expressing neurons are essential for aggression.

mAL Comprises a Sexually Dimorphic Cluster of Neurons.

One of these candidate clusters is the fru-mAL group. Kimura et al. (30) have shown that neurons of this cluster are sexually dimorphic both in number and in morphology. Thus the numbers of neurons in the fru-mAL cluster are lower in females than males (5 vs. 30) as a possible result of programmed cell death during development in female brains. In both males and females, the fru-mAL neurons extend processes to the subesophageal ganglion (SOG), a primary sensory receiving area for gustatory information (31). Recent studies from our laboratory demonstrate that a subgroup of three octopamine (OCT)-containing neurons in the SOG coexpress Fru^M and effect the behavioral choice between courtship and aggression (32). These three neurons show extensive arbors of processes within the SOG. It would be interesting to ask whether any functional overlap exists in the SOG between the arbors of processes of neurons of the fru-mAL cluster important in aggression and the OCT/Fru^M neurons important in behavioral choice. In addition, because 227-GAl4/UAS-tra^IR masculinized female flies show no Fru^M expression in the fru-mAL cluster but still display male-like courtship behavior, we anticipate that these neurons are not essential for the initiation of courtship behavior.

Elucidating the Circuitry for Courtship and Aggression.

Extensive studies of other investigators on courtship behavior in fruit flies have raised the suggestion that male forms of fru identify the circuitry concerned with courtship. An important recent review summarizes these findings and identifies regions of the brain associated with different aspects of the courtship ritual (17). Our findings with Tra^IR masculinized females are in general agreement with these studies. For example, we find consistent high levels of expression of Fru^M in the fru-Lv cluster in all five experimental lines in which courtship behavior is altered, suggesting that this cluster may be of particular importance in male-specific courtship behavior. The present study also shows that, although both courtship behavior and aggression are controlled by Fru^M-expressing circuitry, they are separable and regulated by different subsets of Fru^M-positive neurons. Unfortunately, morphological information on the projection patterns of the aggression-related neurons exists only for the mAL cluster (30), and the transmitter identity of these neurons is unknown.

In vertebrate systems (rodents), pathways for copulatory and defensive behavior have been mapped in circuits that begin with olfactory cues and project through the medial region of the amygdala to the hypothalamus (33, 34). Different parts of the medial amygdala concerned with the two behaviors are identified by expression of different LIM homeodomain transcription factors. Neurons from these regions project to hypothalamic nuclei in glutamate and GABA-associated subcircuits that suggest a route toward mutual inhibition of the two behaviors.

In the studies reported here, we have identified clusters of Fru^M-expressing neurons important for aggression. Hopefully this can serve to begin a subdivision of the Fru^M-expressing neuron clusters into those of special importance to courtship and those essential for aggression. Examining the transmitter identity and detailed morphology of neurons within these clusters and asking whether and how they interact with each other seem logical next steps in attempting to unravel the Fru^M-identified circuitries concerned with each of these behaviors.

Methods

Fly Stocks and Rearing Conditions.

Flies were raised on standard cornmeal medium at 25 ± 1°C in a 12-h/12-h light/dark cycle. Wild-type Canton-S, elav^c155-GAL4, w;UAS-tra^F, and all GAL4 lines (SI Table 2) were obtained from the Bloomington Stock Center. w;UAS-tra^IR was a kind gift from B. Dickson (Research Institute of Molecular Pathology, Vienna).

Behavioral Assays.

Fights were conducted in chambers constructed as described (9), with flies isolated as pupae and paired for fights 5 days after eclosion. Yeast paste was used as a lure, and fights were analyzed as shown in earlier reports (8, 9). All fights were carried out between 1–3 h after lights-on at 25 ± 1°C. All interactions between pairs of Canton-S controls (20 pairs) and experimental lines (10 pairs) were recorded and analyzed.

For screening 60 GAL4 lines, assays were performed in 12-well plates. Four 4- to 5-day-old flies, lightly anesthetized with CO₂, were introduced into each well of a Multiwell 12-chamber plate (Falcon), with 3 ml of standard fly food. The next day, starting 1 h after lights-on, plates were videotaped from above, and the behavior of each line was examined and analyzed for 30 min. Three replicates of each line were tested. Those showing abnormal levels of chaining and singing (courtship behavior), and/or lunging and boxing (aggression) in all replicates were examined in standard fight chambers.

Immunohistochemistry.

For pupal staining, late third-instar larvae were collected and separated by sex. Vials were checked every hour, and dissection was performed 48 h after the formation of white prepupae. Brains were dissected in 4°C. PBS (GIBCO) and fixed with 4% paraformaldehyde in 1× PBS for 20 min at room temperature. For newly eclosed adults, flies were collected within 2 h after eclosion, and brains were dissected and fixed as with pupae. Rabbit antibodies against Fru^M (1:3,000) (21), mouse antibodies against NC82 (1:100) (35), and mouse antibodies against β-galactosidase (1:200; Promega) were used. Confocal images were taken by using a Zeiss LSM META 510 microscope and processed with LSM 510 image examiner and Adobe Photoshop 7.0.

Analysis of GAL4 Expression Patterns.

Individual GAL4 lines were crossed to w;UAS-RedStinger (Bloomington Stock Center) flies and 3- to 5-day-old male progeny were dissected and prepared for Fru^M immunohistochemical examination. To label all neurons expressing GAL4 throughout development, we generated w;UAS-FLP;Act5c(FRT.polyA)lacZ.nls flies from Act5c(FRT.polyA)lacZ.nls and yw;UAS-FLP1.D (Bloomington Stock Center) using standard crosses. Individual GAL4 driver lines were crossed with the newly generated w;UAS-FLP;Act5c(FRT.polyA)lacZ.nls flies. Male progeny were dissected, and immunohistochemistry was performed for Fru^M and β-galactosidase as above.