|Home | About | Journals | Submit | Contact Us | Français|
Two genes coding for transcription factors, fruitless and doublesex, have been suggested to play important roles in the regulation of sexually dimorphic patterns of social behavior in Drosophila melanogaster. The generalization that fruitless specified the development of the nervous system and doublesex specified non-neural tissues culminated with claims that fruitless was both necessary and sufficient to establish sex-specific patterns of behavior. Several recent articles refute this notion, however, demonstrating that at a minimum, both fruitless and doublesex are involved in establishing sexually dimorphic features of neural circuitry and behavior in fruit flies.
Modern biologists are accustomed to thinking in precise terms about how genes control anatomical and physiological traits, by encoding networks of regulatory developmental programs and/or cellular functions. Behavioral traits are inherently complex, and deciphering the connections between genotype and behavioral phenotypes requires understanding how genes control the development and physiological functions of the relevant neural systems. This article focuses on recent findings concerning two genes of Drosophila melanogaster, fruitless (fru) and doublesex (dsx), and their effects on the development of neural systems that generate two sexually dimorphic social behaviors, courtship and aggression. Both fru and dsx are part of the sex determination hierarchy (SDH) of genes in Drosophila that control both morphological and behavioral sexual dimorphisms. The two genes have distinct mutant phenotypes: fru mutations disrupt male sexual behavior, and dsx mutations scramble the normal morphological markers of males and females. These phenotypes suggested that fru and dsx function in distinct anatomical domains, i.e., that fru specifies sex-specific neural development while dsx specifies morphological sexual dimorphisms [1–4]. A few years ago, transgenic manipulations designed to reverse the normal sex-specific patterns of fruitless expression yielded phenotypes in which male flies behaved like females and females like males [5–7]. These results led to suggestions that fru alone was both necessary and sufficient for the generation of sex-specific patterns of behavior [5–8]. That notion has been challenged by more recent findings, and this review will examine evidence indicating that fru and dsx interact to specify many sexually dimorphic features of neural circuitry and behavior. Several comprehensive reviews relating to this topic have been published in recent years [9–11].
Transcripts of both fru and dsx undergo sex-specific splicing that is regulated by upstream genes in the sex-determination hierarchy of Drosophila. The ratio of X-chromosomes to autosomes determines that splicing factors encoded by Sex-lethal (Sxl) and transformer (tra) genes are expressed only in females [12, 13]. Tra and Tra-2, another splicing factor that is not sex-specifically expressed, in turn regulate female-specific splicing of fru and dsx transcripts. Both dsx and fru code for transcription factors, dsx for proteins of the zinc-finger family and fru for proteins resembling the BTB-zinc-finger family of transcription factors. Female-specific splicing of dsx transcript results in distinct male (DsxM) and female (DsxF) proteins [9, 13, 14]. The fruitless gene is more complex, with multiple promoters and alternative exons. Only RNA transcribed from the fruP1 promoter is subject to female-specific splicing by Tra and Tra-2 [15, 16]. This female splice variant is not translated into protein, while default splicing of the fruP1 transcript in males generates FruM protein. In fact, three FruM isoforms with different zinc-finger domains can be produced in males by alternative splicing of 3′ exons [15, 17].
The sex-specifically spliced fruP1 transcript is expressed in neural tissues in both males and females beginning in late third instar larvae, and peaking in 1–2-day old pupae. Generally similar patterns were labeled in the brains of male and female pupae by in situ hybridization, and there were some male-specific cell groups in thoracic and abdominal ganglia [15, 18, 19]. FruM protein was detected only in males, in the nuclei of about 1700 neurons in the pupal and adult CNS. The FruM-expressing neurons occur in 20 clusters or loosely aggregated groups that are widely distributed throughout the male brain, optic lobes, and ventral nerve cord . The locations of many of these cell groups correspond with anatomical regions that had been implicated in the control of sexual behavior by earlier studies of genetic mosaics [20, 21]. Moreover, the effects of various fru alleles on either quantitative or qualitative features of the gene’s cellular expression patterns seem to correlate with the widely differing courtship phenotypes of different fru mutants [18, 22], lending support to the hypothesis that FruM controls the development and/or the function of the neural circuitry for male sexual behavior .
An anti-Dsx antibody that detects both DsxM and DsxF labeled about 500 cells in the pupal CNS, with no evident anatomical sexual dimorphism. A marked sexual dimorphism was evident in adults, however, with much the same pattern of DsxM cells in pupal and adult males, and only a few cells labeled in adult females . About half the Dsx-expressing neurons are in the abdominal ganglion, and another 80 are in thoracic ganglia. The brain contains two clusters of Dsx-expressing cells in the posterior protocerebrum as well as a few scattered neurons in the subesophageal ganglion . Considering these results together with evidence for anomalous courtship behavior caused by dsx mutations [24, 25], possible interactions of fru and dsx in specifying certain sex-specific behavioral patterns were suggested . In fact, most dsx-expressing neurons in pro-and mesothoracic ganglia co-express FruM , and about half of the DsxM neurons in the brain and abdominal ganglion of male pupae co-express FruM [17, 26]. Interactions between these two genes in the development of specific sexually dimorphic features of neural circuitry and behavior will be considered below.
Additional features of fruitless expression, beyond those indicated by FruM immunostaining, have been revealed by GFP reporters of several fru promoter-GAL4 drivers [6, 27–29]. These fru-GAL4 drivers are expressed in similar patterns in the central nervous systems of males and females, although each apparently defines a different subset of the cell population revealed by FruM protein staining. Reporters of fru-GAL4 expression also revealed fru expression in many peripheral sensory structures, including olfactory, auditory, gustatory, and mechanosensory neurons in a variety of appendages [6, 27, 28], and much of this was confirmed by anti-FruM staining. Thus, thinking about fruitless function has expanded to include peripheral sensory neurons. Finally, some fru-GAL4 drivers have been used productively to characterize the structure and function of targeted cell groups [29–32], yielding important evidence for sexual dimorphisms in the numbers and neurite projections of some fru-expressing neurons, as described below.
Long before fruitless was suspected of specifying sex-specific neural circuits, three types of sexual dimorphisms in the structure of the adult fly CNS were documented by conventional neuroanatomical methods. These include male-female differences in (1) the size or volume of defined fiber tracts or neuropil regions, (2) the detailed morphology of identified neurons, and (3) patterns of neuroblast divisions, implying different numbers of particular types of neurons. Thus, neuroanatomists counted more mushroom body fibers in females than in males , and determined that certain brain neuropil areas were larger in adult males than in females . Sexually dimorphic projections of sensory neurons into the prothoracic ganglion were visualized by cobalt fills of neurons innervating leg taste bristles . Since the axons of females carrying tra or Sxl mutations were found to have male-specific branches, these gustatory axons provided the first clear evidence for a neuroanatomical dimorphism being controlled by genes of the SDH. Meanwhile, at the back end of the fly, a sexual dimorphism in the developing CNS was discovered by labeling dividing neuroblasts with bromodeoxyuridine . Neuroblasts in the terminal abdominal ganglion that stopped dividing in females as third instar larvae continued to divide in males for another day, thereby producing about 20 extra neurons. Genetic analysis of this neuro-developmental dimorphism indicated that it is controlled by tra, tra-2, and dsx, thus providing evidence that cells intrinsic to the CNS are controlled by genes of the SDH.
Molecular genetic tools and manipulations of fruitless have revealed further examples of all three types of neuroanatomical sex differences, and further evidence that they are specified by the actions of genes of the SDH. Except for a few insights concerning the role of programmed cell death, however, we remain largely ignorant of the developmental mechanisms by which fru and dsx gene products generate these sexual dimorphisms.
Recent advances in understanding of the Drosophila olfactory system  include new evidence for neuroanatomical sexual dimorphisms. Sensory neurons in olfactory sensilla of the antennae send axons through the antennal nerve to terminate in glomeruli of the antennal lobe. While it has long been known that enlarged sex-specialized olfactory glomeruli exist in other insect species , clear evidence for sexually dimorphic antennal glomeruli in Drosophila is relatively recent . The subset of olfactory receptor neurons in the antennae that express fru-GAL4 appear to project to a small number of sexually dimorphic glomeruli [6, 8, 28]. The volumes of these glomeruli were larger in males than in females of control genotypes, but not in genotypes engineered to prevent sex-specific splicing of fru transcripts, showing that sex-specific splicing of fruitless transcripts specifies the larger size of these glomeruli in male flies . The possibility that synaptic contacts might be made between FruM-expressing sensory neurons and second order projection neurons that also express FruM was raised by the presence of FruM-expressing cell bodies near the antennal lobes [8, 28]. Such connections were elegantly demonstrated recently using a photoactivatable form of GFP to specifically label fru-GAL4-expressing neurons that innervate a single glomerulus . With this approach, the processes of six projection neurons that innervate the sexually dimorphic DA1 glomerulus were traced to their terminal branches in the lateral horn of the protocerebrum. A small sex difference was detected in the fine structure of these arborizations, with a small male-specific ventral branch . This sexually dimorphic pattern of axonal branching appears to be specified by FruM, since small ventral branches were observed in the DA1 projection neurons of females expressing FruM, but were not detected in fru mutant males .
Since the olfactory receptor neurons and the projection neurons that innervate the DA1 glomerulus respond to cis-vaccenyl acetate [40–43], a male anti-aphrodisiac compound transferred from male to female flies during copulation [44, 47], it was suggested that this higher order morphological difference might account for behavioral differences in male and female responses to this pheromone . That remains to be established, however, as many complexities exist in relation to the behavioral significance of cVA (it acts as an anti-aphrodisiac in males, but can serve as an attractant in both males and females ), in the receptors involved in detecting it [46, 47], and in the physiological role of the circuitry described by these studies. Moreover, FruM is expressed in olfactory receptor neurons that project to other than sexually dimorphic glomeruli, and neurons expressing different olfactory receptors project to some of these glomeruli [48, 49]. Nonetheless these results illustrate that FruM–expression can define at least the initial stages of a sensory system concerned with courtship behavior, and they suggest the possibility that FruM–expression might enable further tracing of neural circuitry that is concerned with sexual behavior.
In the original mapping of cells in the CNS that express FruM, approximately 1700 neurons were classified into 20 groups . This classification scheme has been refined recently by analyzing both cell body positions and trajectories of the primary neurites labeled by GFP reporters of fruNP21-GAL4, which is expressed in 82% of the FruM staining neurons in male brains. As a result, the fruNP21- expressing neurons in the brain have been subdivided into 48 identifiable clusters . For most of these clusters, similar numbers of cells express fruNP21 in male and female adult brains, but two clusters are male-specific, and seven have significantly more cells in males than in females . On average, in each brain hemisphere, fruNP21-GFP labeled 480 neurons in males, and 336 neurons in females. Recognizing that these sexually dimorphic neuron numbers could arise from differences in either neuroblast proliferation or cell death, the MARCM method (Mosaic Analysis with a Repressible Cell Marker ) was used to ask whether female cells could be rescued by eliminating three cell-death genes, hid, grim and rpr [29, 30]. Small clonally-derived groups of cells that were homozygous for a deletion that eliminated the cell-death genes were generated in females. In two of the sexually dimorphic cell groups, the mAL cluster just medial to the antennal lobes and the P1 cluster in the dorsal posterior brain, increased cell numbers were found in hid− grim− rpr−clones of adult females [29, 30]. The results suggested that a cell-death program normally removes these fru-expressing neurons from female brains. Interestingly, genetic evidence suggests that the cell-death programs are regulated differently in these two cell clusters: that cell-death is inhibited in mAL cells of males by FruM , and that it is activated in P1 cells of females by DsxF .
Not only do males have a greater number of mAL cells than females (30 vs. 5 per hemisphere), there are clear sex-specific differences in the arborizations of these neurons . In order to visualize the morphological details of individual mAL neurons, the MARCM method was used to limit the expression of the GFP reporter to small clonal groups of cells. The axonal processes of mAL neurons were similar in males and females, with varicose arborizations in the contralateral superior lateral protocerebrum. The dendritic branches, however, are sexually dimorphic. All five female mAL neurons have a single contralateral dendritic branch in the subesophageal ganglion, which is forked, whereas the larger male mAL cluster includes cells with exclusively contralateral dendrites and others with dendritic branches on both sides of the subesophageal ganglion, none of which were forked . Analysis of the morphology of mAL neurons in fru mutant males revealed only female-like dendritic branches in most cells, suggesting that FruM specifies the development of the male-specific dendritic morphology .
A distinctive group of large fru-expressing serotonergic neurons at the posterior tip of the abdominal ganglion (SAbg neurons) innervates male reproductive organs and is implicated in the transfer of sperm and seminal fluids during mating [22, 51]. These neurons were the subject of a recent study that provided important evidence for transcriptional codes involving combinations of different Fru and Dsx isoforms in specifying neuronal identity. This study involved a new fru allele, fruΔC, that contained a small deletion in the coding sequence of exon C, one of four alternative C-terminal exons that encode zinc-finger domains . By combining the fruΔC allele and another fru allele that lacks all FruM isoforms with fru-GAL4 targeted expression of individual FruM isoforms, it was possible to evaluate the role of different isoforms in specifying neuronal phenotypes and aspects of sexual behavior. Thus, for the male specific Muscle of Lawrence (MOL) in the dorsal abdomen, whose appearance requires FruM expression in the motor neuron that innervates it , the FruMC isoform was found to be both necessary and sufficient for formation of the MOL and the associated motoneuron . A more complex picture emerged for the male-specific serotonergic neurons in the abdominal ganglion, however. These cells were completely eliminated in FruM-null males, and greatly reduced in number in FruMC-null males, which still produce FruMA and FruMB isoforms . Results of earlier studies using anti-5HT staining to try to label these cells in females suggested that FruM controls 5-HT expression rather than the presence or absence of the SAbg cells . When FruM-immunostaining was used to count neurons in abdominal ganglia of FruMC-null males, however, the results suggested that a subset of the SAbg neurons failed to develop in FruMC-null mutants . Moreover, efforts to rescue the SAbg cells by expressing different FruM isoforms in FruM-null males suggested that each FruM isoform rescued only a subset of the normal male complement of SAbg neurons. These results indicate that multiple FruM isoforms are required for the development of these cells. Furthermore, cell counts in different sex determination gene mutants suggested that both FruM and DsxM contribute to the formation of normal numbers of these male-specific serotonergic neurons . These findings resonate with older evidence that dsx produces extra neurons in male abdominal ganglia by prolonging neuroblast proliferation . These results clearly indicate that transcriptional codes involving different combinations of fru and dsx gene products specify different components of the neural circuitry concerned with sexually dimorphic patterns of behavior.
The hypothesis that FruM expression in pupal and adult nervous system is sufficient to generate male courtship behavior was tested by forcing expression of FruM in female flies. Two research groups used gene targeting by homologous recombination to generate engineered fru alleles that forced either male- or female-specific splicing of the fruP1 transcript. One team directly engineered sites required for sex-specific splicing of fru P1 transcripts such that fruM and fruΔtra alleles encode transcripts that are always spliced into the male form encoding FruM, and a fruF allele prevents male-specific splicing . The other group produced a fruP1-GAL4 driver by recombining the GAL4 coding sequence into fru, then used it to drive expression of UAS-RNAi transgenes that disrupted sex-specific regulation of the endogenous fruP1 transcripts . Female flies that expressed FruM by either of these strategies performed some distinctly male patterns of courtship towards wild-type virgin females, providing evidence of the sufficiency of FruM to specify the neural circuitry responsible for some male-specific behavioral patterns. Many features of courtship behavior in these masculinized females, however, were different from that of normal males. Females with their fruP1-cells masculinized by expressing tra2IR performed early steps of courtship, but not later stages (wing extension, licking, and copulation attempts) . The fruM and fruΔtra females reportedly performed all steps of courtship except for attempting to copulate, except that data in this paper showed that licking was significantly subnormal in these masculinized females . Furthermore, in a later study, audio recordings of fruM and fruΔtra females revealed that their courtship songs were abnormal and much less frequent than those of control males . Since dsx mutations also produce abnormal song phenotypes , and most FruM-expressing neurons in the mesothoracic segment of the CNS co-express DsxM , it is likely that both genes contribute to specifying the neural systems for song. Thus, in spite of claims to the contrary, FruM expression in females is not sufficient to specify normal male courtship behavior.
Aggression is a second sexually dimorphic behavior in Drosophila that is influenced by fruitless . Both male and female flies display patterns of aggression in same sex pairings, and many of the behavioral patterns seen are sex-specific . By using the fruF and fruM alleles described above , it was demonstrated in same sex pairings that fruF males displayed female patterns of aggression and fruM females displayed male patterns . Similar results were obtained when male nervous systems were feminized or female nervous systems were masculinized by pan-neural expression of either UAS-traF or UAS-traIR, respectively , except that the traIR-expressing females did not perform male-type wing threats. The conclusion of these studies too, was that FruM expression was necessary and sufficient to transform patterns of aggressive behavior from female-like to male-like. Note, however, that forcing both FruM and DsxM expression in female nervous systems (with traIR) was not sufficient to produce male-like wing-threat displays.
In an attempt to identify neurons responsible for sex-specific patterns of aggression, a collection of 60 enhancer trap-GAL4 drivers was screened for the ability to induce sex-reversed behavioral patterns when used to manipulate transformer expression . Five of these GAL4 drivers produced partially masculinized (GAL4-traIR) females that performed male courtship behavior, and two of these genotypes also displayed male patterns of aggression. The brains of these five types of GAL4-traIR females were stained with anti- FruM to determine whether FruM expression in any particular cluster(s) might be correlated with the female’s ability to fight like males. While there was no single cluster where FruM-staining correlated absolutely with male-like fighting patterns, four clusters were identified where the numbers of FruM–staining cells in GAL4-traIR females that fought like males were comparable to those in normal males, and where FruM–staining was consistently weaker in GAL4-traIR females that fought like females . Three of these four clusters are close to the midline in the brain (fru-mAL, fru-aSP2, fru-mcAL), and one is in the ventral nerve cord (fru-PrMs). They include the mAL cluster that has sexually dimorphic cell numbers and arborizations  and also was selectively missing from the FruM-staining pattern in fru1 mutant males that show high levels of male-male courtship .
Understanding of the structure and function of specific subsets of neurons requires manipulations that target identifiable cells or clusters. In a recent attempt to localize the behavioral functions of fru to a particular cell cluster, the MARCM method was used to generate small clonal groups of fruNP21-GAL4 –expressing cells that were both labeled by GFP and homozygous for tra1, thereby masculinizing the cells . By assaying courtship behavior and evaluating the spatial patterns of GFP/tra1 neurons in 205 MARCM females, this study was designed to determine whether masculinization of any fruNP21-expressing cells or clusters correlated with the ability of the MARCM females to court like males. Only 16 of the 205 mosaic females displayed early steps of courtship behavior, and none performed advanced steps or sustained bouts of courtship . That may not be surprising since most of their neural circuitry was still female. As above, in no single cell cluster did the tra1 genotype correlate perfectly with male-like behavior. The best correlation between genotype and behavior was found for the P1 cluster, a group that is normally found to express fruNP21 only in males. This cluster was masculinized in 13 of the 16 MARCM females that performed any courtship, and in every one of the 10 most active courters . The P1 cells were not masculinized in three weak courters, however, thus one cannot conclude that FruM/DsxM expression of these cells is absolutely necessary for weak displays of courtship. Nor are members of this cluster sufficient for male courtship, since P1 cells also were masculinized in 12 of the 189 MARCM females that displayed no courtship in these behavioral assays . Given the complexity of the behavioral phenotype and the numbers and diversity of FruM-expressing neurons, it may not be reasonable to expect that masculinization of any one cell cluster will prove to be both necessary and sufficient to initiate courtship in an otherwise female brain. It is fair to say that the genetic sex of cells in the P1 cluster has an important influence on a fly’s ability to perform early steps of male courtship behavior. A note of caution, however, is that early steps in aggression in both males and females involve approach and tapping behaviors that may be difficult to distinguish from early steps of courtship behavior.
Finally, it may be significant that these first attempts to relate the genetic sex of subgroups of FruM-expressing neurons with behavioral patterns of courtship and aggression have all implicated neurons that lie within FruM-clusters close to the brain midline [22, 30, 54].
Behavioral traits are more complex than neuroanatomical ones, and can be difficult to measure reliably. This is especially true for social behaviors that are influenced by multiple behavioral and sensory interactions with another fly. Male courtship behavior, for example, has proven to be highly sensitive to assay conditions . In addition, general statements of necessity and sufficiency for genes that regulate behavioral phenotypes often need to be revised and qualified in light of new evidence. While this may be disconcerting for Drosophila geneticists, who are accustomed to dealing with Mendelian traits, it is less surprising from the perspective of human behavioral genetics, where it has long been recognized that Mendelian behavioral phenotypes are the exception rather than the rule.
While expression of FruM proteins is necessary to specify the neural systems that generate male patterns of behavior, FruM alone is not sufficient to generate normal male patterns of behavior. At a minimum, dsx expression also is required and other genes may be involved as well . Yamamoto [10, 30] has suggested that three categories of neurons might be involved in establishing the circuitry underlying sexually dimorphic patterns of behavior: sex specific neurons requiring dsx and fru [17, 26]; sex specific neurons requiring either dsx or fru [17, 57]; and non-sex specific neurons requiring neither gene [31, 47]. Much of evidence described in this review supports this view.
Much remains to be done, however, to unravel the roles of genes in establishing sexually dimorphic patterns of behavior in fruit flies. At the genetic level, only a start has been made on deciphering the transcriptional codes that specify neuronal fate within sexually dimorphic circuitry [17, 26, 30]. It will be important to determine which of the three FruM isoforms are required to specify the phenotypes of identified subsets of FruM -expressing neurons, and how DsxM and DsxF influence neuronal numbers and phenotypes. It also will be important to identify genes that are directly regulated by Fru or Dsx proteins in the developing nervous system . Current evidence suggests that Dsx proteins will regulate genes that control neuroblast proliferation and programmed cell-death, and that FruM isoforms will potentially regulate a broader array of targets, encompassing processes that specify both neuronal numbers and cellular phenotypes. At the neural systems level, important beginnings have been made on mapping the arbors of and connectivity between neurons within sexually dimorphic circuits [29, 32], but far greater detail is required in the examination of these neurons, many more neurons must be examined, and the patterns of connectivity must be established at both light and electron microscopic levels. At the cell biological level, neurotransmitter and neurohormone identification and receptor, second messenger and ion channel distribution among neurons within FruM-associated circuitry has been lacking [22, 59]. Information about these essential signaling functions will be needed to understand how sex-related neural systems functions. Although many questions remain concerning how sex-specific patterns of behavior get established and maintained in even the relatively simple fruit fly nervous system, with an ever-expanding array of powerful genetic and physiological methods available to address such questions, progress is likely to continue at a rapid rate.
We thank the members of the Kravitz laboratory for helpful discussions. This work was supported by grants from the National Institute of General Medical Sciences (GM074675 and GM067645 to E.A.K) and the National Science Foundation (IDS-075165 to E.A.K.). K.K.S. was supported by the Radcliffe Institute for Advanced Studies.
Ethics in publishing statement: The authors declare no conflicts of interest and agree with the journal’s general statement on ethics and conflicts of interest.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.