Drosophila retained/dead ringer is necessary for neuronal pathfinding, female receptivity and repression of fruitlessindependent male courtship behaviors

Courtship in Drosophila provides a genetic, molecular, and neurological model for behavioral development. During courtship, males and females perform gender-specific behaviors (reviewed by Greenspan and Ferveur, 2000). The male begins by following the female, tapping her abdomen, and extending and vibrating one wing to produce a species-specific `love song'. A virgin female initially runs from the male, but if receptive, she slows and positions herself to facilitate copulation.

This binary behavioral system is controlled by the sex differentiation cascade (Hall, 1994; Yamamoto et al., 1998; O'Kane and Asztalos, 1999; Goodwin, 1999; Christiansen et al., 2002). Sex-lethal (Sxl), transformer (tra) and transformer 2 (tra2) catalyze splicing of the next step of the pathway, leading to the activation of sex-specific forms of doublesex (dsx) and fruitless (fru). dsx controls external differentiation, yolk protein synthesis,aspects of male song production (Villella and Hall, 1996) and potentially some aspects of female neural differentiation (Waterbury et al.,1999). fru determines many aspects of male courtship and copulatory behaviors, but has no apparent role in female sexual development(Ryner et al., 1996; Ito et al., 1996; Gailey et al., 1991; Villella et al., 1997). dissatisfaction (dsf) females resist males during courtship,whereas dsf males are bisexual(Finley et al., 1997; Finley et al., 1998). Many male courtship mutants have been identified, while few mutations linked to female receptivity have been characterized(Yamamoto et al., 1997).

We identified retained/dead ringer (retn) from a genetic screen for female behavioral mutations. retn females are resistant to courtship, and show fru-independent male-like courtship behaviors, while retn males are behaviorally normal. These sex-specific effects on behavior do not correlate with sexually distinct expression or splicing patterns in the CNS. Examination of retn cells in retn mutant backgrounds reveals aberrant projections by mushroom body, photoreceptor and subesophageal neurons. retn affects development of sex-specific neurons, and may repress male behavior patterns in the female CNS.

Fly stocks for the EMS screen are from Charles Zuker. UAS-retn and balanced retn-Gal4⁸⁹, retn-Gal4²⁸, retn-Gal4¹⁰⁸, retn-Gal4¹¹⁷, retn-Gal4¹³⁹, dri¹, dri², dri³, dri⁵, dri⁶, dri⁸, dri^B142 stocks were donated by Tetyana Shandala. retn^RU50, retn^RO44 lines are from Trudi Schupbach. UAS-tra is from Ralph Greenspan. Additional lines were provided by the Bloomington Stock Center, Illinois. Control is Canton-S. Flies were raised on standard media.

Female resistance and male courtship indices were tested as previously described (Finley et al.,1997). Male-male courtship and female bisexuality were tested in groups of 10 animals and quantitated as number of courtship events per 5-minute interval. A courtship event was counted as one fly following, tapping or singing to a target fly for a minimum of 2 seconds. Multiple trials were carried out for each genotype and age. All P-values are derived from two-tailed paired t-tests. Multiple retn-Gal4 lines were used to drive UAS-retn, UAS-Tra^F and UAS-GFP. All generated the same pattern and had similar effects. retn-Gal4⁸⁹, a lethal insertion, showed the most complete rescue of female resistance behavior and egg laying, and was primarily used for studies of retn function and expression. Rescue of male-like behaviors in females was complicated by inconsistency of these behaviors in retn-Gal4/retn mutants.

Genomic DNA from retn^z2-428, retn^RO44, retn^RU50, retn^dri1 and retn^driB142 flies was amplified by PCR. Purified PCR product was sequenced at the Salk Sequencing Facility (La Jolla, CA). Sequences were assembled using DNA Sequencher (Gene Codes Corp, Ann Arbor, MI). Sequence comparison and database searches utilized BLAST (Altschul et al., 1990)and/or FASTA (Pearson and Lipman,1988).

Full-length retn cDNA was generated by PCR, using genomic DNA from UAS-retn flies. The cDNA product was cloned into pBluescript-SK+(pBS, Stratagene) and sequenced on both strands.

To produce ARID-box subclones, the pBS-retn plasmid was used as template for further PCR. The subsequent retn_ARID product encodes amino acids 230-500 of RETN, and includes the ARID domain plus flanking sequence. This was subcloned into pBS and sequenced on both strands. pGEX-retn_ARID was produced by inserting a BamHI-XhoI fragment of pBS-retn_ARID into the BamHI and SalI sites of pGEX-KG.

BS-retn^RO44_ARID and BS-retn^z2-428_ARID vectors were generated using PCR-based site-directed mutagenesis of the pBS-retn_ARIDtemplate. Positive clones were confirmed by sequencing and transferred into pGEX-KG.

DNA-binding analysis was performed as described by Pitman et al.(Pitman et al., 2002). RETN wild-type and mutant fragments were expressed as GST-RETN fusion proteins in BL21 pLysS bacteria. Proteins were purified and eluted(Kaelin et al., 1992). EMSA analysis used 2 μl of eluted protein. Proteins were tested for relative expression on a western blot, using rabbit anti-GST antibodies.

For examination of retn RNA, CNS tissue (sans imaginal discs) was isolated from both sexes of late third instar larvae or mid-stage pupae. Total RNA was extracted using RNeasy Mini Kit (Qiagen). An antisense primer targeted to either exon 11 or 12 of retn was used to prime DNA synthesis by M-MLV Reverse Transcriptase (Sigma). A first round of PCR was carried out using primers against exons 1 and 4, 4 and 8, 8 and 11, 8 and 12, 1 and 8, and 4 and 11. A second round of PCR was then carried out using primers internal to those used in the first round. For examination of fru P1-derived RNAs in retn mutants, RNA was isolated from mid-pupal CNS tissue and from adult heads. For analysis of fru P1 RNAs in retn fru double mutants, RNA was isolated from adult heads. The RT-PCR procedure was as above with fru primers. For all fru RNA tests, reverse transcription was primed from within exon 3, which is common to all fru RNAs. The 3′ primer for both first and second round PCR was placed just inside (more 5′ on the RNA) to the RT primer. For analysis of the fruM RNA, first round PCR was primed at the 5′ side from within promoter P1-derived exon 2. Second round PCR used a primer just 3′ of this. For analysis of the fruF RNA, first and second round 5′ primers were just upstream of the TRA/TRA2 activated splice site of fru.

Confocal images were obtained on Zeiss LSM 480 and LSM510 Meta microscopes,using Renaissance 410 (Microcosm, Columbia, MD) software. Antibodies to Fas2 were obtained from the Developmental Studies Hybridoma Bank (University of Iowa). The brains of mutant and wild-type males and females were labeled with anti-Fasciclin 2 (Fas2) (1:20) and then with an anti-mouse secondary Alexa 488(1:200; Molecular Probes) using standard methods(Finley et al., 1997).

We conducted a genetic screen similar to the screen that identified dsf (Finley et al.,1997), testing a collection of viable EMS-treated chromosomes developed in C. Zuker's laboratory. One of these lines, z2-428,showed substantial alterations in female behavior and fertility. Recombination and deficiency mapping place z2-428 in salivary chromosome region 59F, between the right-hand breakpoint of Df(2R)bw5 and the left-hand breakpoint of Df(2R)HB132. These deletions complement z2-428. Additional testing revealed that z2-428 is allelic to retn,an uncloned female sterile locus(Schupbach and Wieschaus,1991).

Alleles of dead ringer (dri), an extended ARID (AT-rich interaction domain) Box-Family embryonic DNA-binding factor(Gregory et al., 1996; Iwahara et al., 2002) also fail to complement retn. We sequenced the exons and exon/intron boundaries of dri in z2-428, retn^RO44 and retn^RU50 (Fig. 1). Each allele has a single nucleotide change in the dri-coding region, corresponding to an ARID box amino acid substitution. Two dri lethal alleles, dri¹ and dri^B142, encode premature stop codons, truncating the protein (Fig. 1). Thus,missense alleles retn^z2-428, retn^RU50and retn^RO44 encode a protein with sufficient function that mutant progeny survive to adulthood, while nonsense alleles are lethal.

Drosophila convention favors earlier over later names of the same locus. Thus, FlyBase now refers to retn and dri as retn. We distinguish retn-class alleles, which are adult viable with behavioral and reproductive defects, from dri-class alleles, which are embryonic lethal. We denote dri¹, dri-Gal4⁸⁹ and other lethal alleles as retn^dri1, retn-Gal4⁸⁹, etc.

retn mutant proteins have residual DNA-binding ability (data not shown) consistent with survival of some mutant individuals to adult stages. We asked whether these mutations alter the vital function of retn and to what extent phenotypes may be limited to later functions. In examining viability of retn heteroallelic combinations, we found variability in eclosion rates (Fig. 2A) with most lethality in the larval stages. Allelic strength in terms of pre-adult mortality is retn^dri2 > retn^dri1> retn-Gal4⁸⁹ > retn^dri8 > retn^z2-428 > retn^RU50 >wild-type. retn^z2-428/retn^dri2 flies eclose with only 8% of expected rates, while retn^z2-428/retn^dri1 flies eclose with 25% of expected rates. retn^RU50/retn^dri1 and retn^RU50/retn^dri2 eclose with 65% and 68%, respectively, of expected numbers. P-element insertion alleles show full or nearly full viability with retn-class alleles. retnlethal alleles show no dominant lethality. Thus, all retn-class alleles at least partially complement the vital functions of retn. In addition, the retn cDNA rescues the partial lethality of retn^z2-428/retn-Gal4⁸⁹(Fig. 2B). retn^z2-428/retn-Gal4⁸⁹ flies eclose with 33% of expected numbers, while retn^z2-428/retn-Gal4⁸⁹;UAS-retn flies eclose with 100% of expected numbers.

retn females are strikingly resistant to male courtship(Fig. 3A). Wild-type females,as well as retn/+, copulate after an average of three minutes or less of courtship. retn^RU50 and retn^z2-428females showed significant increase in time of courtship prior to copulation: retn^RU50/retn^dri2 females average 34±6 minutes (P=0.00004), and retn^z2-428/retn^dri2 females typically resisted male advances for the entire hour in which we monitored courtship,averaging 58±2 minutes (P=5 ×10^-17). retn^RU50/retn^z2-428 females showed a less severe phenotype, with an average of 8.8±2 minutes(P=0.013). Females showed virgin resistance behaviors of running,kicking, wing flicking and bending the abdomen away from males. Following copulation, females showed normal mated responses of ovipositor extrusion.

A retn-Gal4 enhancer trap(Brand and Perrimon, 1993) that is known to match the RETN protein pattern(Shandala et al., 1999) (J. Sibbons, personal communication) driving a UAS-controlled long form retn cDNA (Shandala et al.,1999) (see below) rescues female resistance behavior(Fig. 3B). retn-Gal4⁸⁹/retn^RU50; +/+ females resist courtship for an average of 25±4.6 minutes. retn-Gal4⁸⁹/retn^RU50;UAS-retn/+ females copulate in 4.8±1 minutes, comparable with wild type, and are fertile (L. M. Ditch, PhD thesis, University of California,2002). This indicates that retn-Gal4 activates expression of UAS-retn in cells necessary for female behavior in a positionally and temporally correct pattern, and that overexpression of a non-sex-specific embryo-derived cDNA is sufficient to carry out some female neuronal functions.

retn females show one behavior not shown by dsf, dsx or fru females: male-like courtship of females and males, especially as they age (Fig. 3C-F). retn females follow, tap and appear to sing. Although not as robust as male courtship - following is not as sustained, full wing extension and vibration are not seen, and copulatory bending is weak or absent - these behaviors highly resemble courtship. Fig. 3 shows still frames of this behavior, directed towards females(Fig. 3C,D) or a courting male(Fig. 3E). These behaviors vary between and within allelic combinations, but when the behaviors are seen they are striking and continue for hours. retn^z2-428/retn^dri8 females, which show the most consistent behaviors, with maximum penetrance at 3-4 weeks post-eclosion, averaged 42 courtship events per 5-minute observation period(Fig. 3F), while control females display fewer than three courtship-like events in the same period. Although male behaviors are evident, the fruM-dependent Muscles of Lawrence are not seen in retn females (not shown and L. M. Ditch, PhD thesis, University of California, 2002).

Aspects of the retn female behaviors are similar to wild-type female defenses of food and egg-laying resources. One study on Drosophila aggressive behaviors(Ueda and Kidokoro, 2002)indicated that aggression in wild-type females increases if females are raised individually before pairing for observation. We found no increase in male-like behaviors in females kept separately from eclosion until testing (not shown;L. M. Ditch, PhD thesis, University of California, 2002). This suggests that these behaviors are not an exaggerated defense response. Other indications that these behaviors are not based on access to food come from observations of wild-type females starved overnight on moistened filter paper and transferred back onto food. These females showed short head-to-head and head-to-side interactions, but did not show behavior resembling male courtship. Courting retn females, by contrast, primarily show posterior orientation(Fig. 3C,D), and will follow other females on and off a food source for minutes at a time.

Genetic data indicate that males lacking fruM (P1 derived)transcripts show a `complete absence of sexual behavior'(Anand et al., 2001). However,we observe male-like courtship by retn mutant females, which should lack fruM (Ryner et al.,1996). This suggests three possibilities: (1) retnmutants could lead to an up regulation of fruM in females; (2) there could be a very low level of fruM in wild-type and retnfemales, which, in the absence of retn, is sufficient to induce some male behavior; or (3) there could be an intrinsic, but weak, fru-independent pathway for male behavior that is repressed by retn or retn-expressing neurons (see Discussion for a model incorporating this idea). We have tested these possibilities.

As fruM RNA expression is male specific and is eliminated in females by TRA- and TRA2-mediated splicing of P1 transcripts into the fruF RNA form, we expect no increase in fruM in retn females. We addressed whether retn loss-of-function leads to upregulation of fruM in females. RT-PCR with one round of amplification using primers against fruM gave no detectable fruM product in Canton S or retn^- midpupal or aged-adult female CNS tissue (data not shown). A second round of amplification showed an extremely low signal for fruM in equal amounts in both wild-type and retn^- CNS tissue (data not shown). These results indicate that fruM is not upregulated in retn^- CNS tissue, although the small amount of fruM detected in the second round of amplification might be responsible for the male-like behaviors in retn females.

We tested the dependence of the male-like behaviors in retnfemales upon the observed amount of fruM. Df(3R)fru^4-40 removes the P1 (responsible for transcripts under tra/tra2 control) and P2 promoters, leaving the P3 and P4 promoters intact. Df(3R)fru^AJ96u3 removes P4 and the entire fru protein coding region(Song et al., 2002). fru^4-40/fru^AJ96u3 flies lack P1 derived transcripts, but are healthy because of P3 and P4 activity(Song et al., 2002). RT-PCR analysis with two rounds of amplification upon CNS tissue from these females indicated a complete absence of fruF and fruM (data not shown), as expected. We tested for male-like behaviors by retn^-; fru^- females(retn^z2-428/retn^dri8; fru^4-40/fru^AJ96u3). Such females aged for ∼2.5 weeks, produced retn-like male behaviors(Fig. 3G,H), indicating an independence of such behaviors from fruM. In addition, similarly aged retn^- females carrying a different fruM null allelic combination[Df(3R)fru^sat15/Df(3R)fru^4-40(Anand et al., 2001)] also display substantial male-like courtship behavior (not shown). Taken together,these data indicate that the male-like behaviors observed in retnfemales are specified by a means independent of fruM.

We tested if retn alters male behaviors or functions. retn males court females, are not delayed in copulation(Fig. 4A), do not show significant courtship of other males (Fig. 4B) and have normal Muscles of Lawrence. retn males produce motile sperm and copulate normally, but show defects in sperm transfer and are partially sterile (L. M. Ditch, PhD thesis, University of California,2002).

To test if any retn cells have important sexual identities in males, we used retn-Gal4⁸⁹ to drive UAS-Tra^F in males. XY; retn-Gal4⁸⁹/UAS-Tra^F animals have male pigmentation patterns and sex combs, but genitalia are underdeveloped(data not shown; L. M. Ditch, PhD thesis, University of California, 2002). They court females with normal courtship indices, and court other males. Wild-type males do not court the retn-Gal4⁸⁹/UAS-Tra^F males (data not shown; L. M. Ditch, PhD thesis, University of California, 2002). These results indicate that, although retn mutations do not alter male behavior, some retn-Gal4⁸⁹-expressing cells have sex-specific identities essential for male sexual orientation.

As retn has female-specific phenotypes, we asked if it is a direct target of regulation by Tra/Tra2-mediated alternative splicing focusing on central nervous system RNAs, as retn has non-sex-specific functions in other tissues (Gregory et al.,1996; Shandala et al.,1999; Shandala et al.,2002; Bradley et al.,2001; Iwaki et al.,2001). We analyzed RNA from the larval CNS, prior to the most sensitive period for sexual nervous system differentiation, and the early/mid pupal CNS, the primary period of sex-specific nervous system determination(Belote and Baker, 1987; Arthur et al., 1998).

retn has 12 exons, most of which are separated by small (fewer than 100 nucleotides) introns (Fig. 1). Exons 1 and 2, 4 and 6, and 6 and 7 are separated by large(multiple kb) introns, while exons 11 and 12 are separated by a 182 base intron. We used RT-PCR to analyze alternative processing between exons 1 and 4, 1 and 8 (pupal only), 4 and 8, 4 and 11 (pupal only), 8 and 11, and 8 and 12 (not shown). The data (Fig. 5) show the expected products, and two novel variants. None of these is sex-specific, which is completely consistent with the rescue of retn female behavioral (Fig. 3A,B) and egg-laying phenotypes using a common form cDNA.

The first novel form is rare (not visible in Fig. 5A) relative to the previously described major RNA form, and joins exons 1 and 4, skipping exons 2 and 3. This creates an in frame deletion in the RNA, removing 318 bases and 106 amino acids, much of the N-terminal non-conserved region of the protein,but leaving the extended ARID box and C terminus intact. The second novel form is approximately equally abundant with the major form and joins exons 1 and 6,creating an in frame deletion removing 756 bases and 252 amino acids. This deletes from very near the protein start into the N-terminal region of the extended ARID box shared with the mammalian Bright/Dril family of factors,leaving the C terminus intact. It is possible that this variant encodes the`95 kDa' form seen by Valentine et al.(Valentine et al., 1998).

To map retn expression in the CNS, we examined retn-driven GFP expression using retn-Gal4 insertions that rescue retn phenotypes with the retn cDNA. These Gal4 enhancer traps, in addition to rescuing retn viability and behaviors,exactly reproduce Retn antibody patterns in embryos and larval eye tissue(Shandala et al., 1999) (J. Sibbons, personal communication); therefore, they should represent the later CNS expression to a high degree of accuracy. Expression and projections were monitored using membrane-associated UASCD8::GFP (UAS-mGFP). retnexpression in the CNS begins in the embryo(Gregory et al., 1996; Shandala et al., 2002), and continues through adulthood, in specific subsets of neurons. As we were primarily interested in neurons involved in adult behaviors, we focused on expression of retn in the periods before and during metamorphosis,when adult neurons are born and larval neurons are remodeled into adult-specific forms. Notably, we see expression in the mushroom bodies,subesophageal ganglion, ventral ganglion and developing photoreceptors. These patterns are essentially the same in both sexes.

In the third instar, MB expression is seen in the Kenyon cell (KC) bodies lying in the dorsoposterior of the central brain, with staining in the calyx,containing KC dendrites, and the pedunculus and lobes, containing KC axons(Fig. 6D). Between 12 and 18 hours after puparium formation (APF), the calyx retracts, the α andβ lobes narrow and what appears to be axonal debris can be seen at the lobe tips (arrow, Fig. 6E). At this stage there are slightly more retn cells in females than in males, perhaps reflecting the greater axon number in female MBs(Technau, 1984). By 36 hours APF, the adult α, α′, β, β′, and γlobe projections are visible, although retn expression is stronger inα/β projections (Fig. 6F). Between 24 and 48 hours APF, expression in all lobes exceptα/β gradually fades, and by 48 hours only the α/β lobes can be seen. This pattern remains through the rest of metamorphosis.

In the larval SOG, two central groups of six or seven neurons and two anterior groups of five neurons send projections towards the protocerebrum and ventral nerve cord (Fig. 6G). Laterally to these neurons are four additional neurons per side. The projections of these neurons form a dense pattern, and individual projections cannot be discerned. Retraction of larval-specific processes can be seen beginning six hours APF (Fig. 6H, 18 hours APF); by 36 hours APF, new processes are evident. The number of SOG neurons expressing retn remains constant, but projections become increasingly dense (Fig. 6I, 48 hours APF) through the pupal period (see Fig. 6C).

In the larval ventral nerve cord (VNC), 18 paired dorsal lateral neurons,nine per side, send projections towards the midline(Fig. 6J). These may mediate signaling to or from the nine larval abdominal segments. By 24 hours APF, the abdominal neurons are now six pairs, residing at the abdominal tip(Fig. 6K). Beyond 36 hours APF and continuing into adulthood, three sets of paired abdominal neurons are visible (Fig. 6L). These final neurons may project outwards from the CNS. A small subset of adult peripheral sensory neurons that innervate the female reproductive structures also send their

retn-Gal4⁸⁹ is expressed posterior to the morphogenetic furrow, in photoreceptor cells R1-R6, which project to the lamina and R8,which projects to the medulla (not shown), as is also seen with Retn antibody staining (J. Sibbons, personal communication). Beyond 48 hours APF, R8 expression and projections fade, although lamina projections remain (48 hour pupal eye, Fig. 7J). Expression in the eye, MB, SOG and ventral nerve cord is still visible post-eclosion(Fig. 6C, early adult).

We observed MB-specific abnormalities in three different retnmutant genotypes: retn-Gal4⁸⁹/retn^Z2-428 larvae and pupae; retn^dri8/retn^Z2-428, and retn^Ro44/retn^RO44 adults(Fig. 7B,C,E) MB neurons diverge within the nerve tracks and β-lobe neurons cross the midline and join with the opposite β-lobe neurons, causing β-lobe fusion,compared with retn-Gal4⁸⁹/+. This is more common in females than males (4/10 larval females, 0/11 larval males, 7/12 pupal females, 2/19 pupal males for retn-Gal4⁸⁹/retn^Z2-428), but phenotypes of retn; fru males (not shown) indicate that retnfunctions in male neurons. Using antibodies to Fas2, which is expressed in MB axons projecting to the α- and β-lobes in retn^dri8/retn^Z2-428 and retn^R044/retn^R044 adults, we found that in a subset of mutant (4/6 retn^dri8/retn^Z2-428 and 1/3 retn^RO44/retn^RO44) females, axons in the posterior part of the β-lobe crossed the midline, leading toβ-lobe fusion (Crittenden et al.,1998). In addition, in those animals with β-lobe fusion,there were fewer Fas2-positive axons in the α-lobe. These MB fusion phenotypes are similar to the β-lobe fusion phenotypes reported in other mutants, such as linotte/derailed, Drosophila fragile X mental retardation 1, fused lobes, ciboulot and α-lobe absent(Moreau-Fauvargue et al., 1998; Boquet et al., 2000; Michel et al.,2004). Resistance is shown by the vast majority of females of these genotypes, thus MB fusion is unlikely to be causal for resistance.

To determine retn neuronal birth dates and the neural phenotypes of dri-class alleles, we used the MARCM system(Lee and Luo, 1999), which can simultaneously create homozygous mutant cells and allow them to express Gal4-regulated marker genes. retn-expressing MB neurons are born throughout the larval and pupal stages and eye clones appear at all embryonic and larval stages. The VNC neurons are born only within 48 hours of egg laying, and SOG retn neurons are born in 8-hour-old or younger embryos.

Homozygous retn-Gal4⁸⁹ clones show striking mis-projection phenotypes in SOG neurons. The normal elaboration and symmetry of arbors in mid-pupae is diminished; ventral dendritic branches do not show normal density (compare arrowhead in Fig. 7F with arrowhead in Fig. 7G), and anterior projections wander and fail to extend (compare arrows in Fig. 7F and Fig. 7G). Neurons also fail to fasciculate normally. A central SOG midline-crossing tract, visible throughout metamorphosis, contains tightly bundled projections (arrow, Fig. 7H). In mutant clones,projections stray from this tract, apparently losing some adherent ability(arrow, Fig. 7I). Photoreceptor neurons also mis-project. In retn^dri clones, induced in the embryo, R1-R6 cells overshoot the lamina, and a number now target the medulla (ME, arrows; Fig. 7J,wild type; Fig. 7K, mutant). Although retn mutations alter neuronal projection patterns, and projection differences are consistent with changes in behavior, we have not yet mapped retn behavioral functions to a particular set of neurons,nor have we demonstrated that the projection differences, as opposed, for example, to retn-induced reductions in neural activity, are responsible for behavioral changes.

retn functions in multiple, separable processes during development. It acts in differentiation and control of gene expression along the anterior posterior and dorsal ventral axes in embryos(Shandala et al., 1999; Valentine et al., 1998). It also acts in the production of various tube structures such as salivary ducts and gut (Bradley et al., 2001; Iwaki et al., 2001). Failures in these or other embryonic processes with dri-class (null or near null) alleles lead to embryonic death. retn-class (hypomorphic missense) alleles can perform the embryonic functions but show defects in neural development and projections. Correlating with this are changes in female behavior, including resistance to male courtship and, strikingly,generation of male-like courtship behaviors. Additional functions in development of internal genital ducts and fertility have been observed (L. M. D., B. J. T. and M. M., unpublished) and will be discussed elsewhere.

retn neural and behavioral phenotypes are substantially different from those of dsf or fru. dsf females, like retn-females, are sterile and resist male courtship(Finley et al., 1997). For dsf, sterility results from loss of motor synapses on the circular muscles of the uterus (Finley et al.,1997). By contrast, these synapses are intact in retnfemales. dsf females show no male behaviors(Finley et al., 1997), while retn females do. dsf males are bisexual and slow to copulate, owing to inefficient abdominal bending, correlated with abnormal synapses on the muscles of ventral abdominal segment 5(Finley et al., 1997). retn males court and mate with normal kinetics and have normal A5 synapses. This suggests that retn and dsf have largely separate functions.

retn and fru also have different phenotypes. In a wild-type background retn behavioral phenotypes are restricted to females. fru behavioral phenotypes are restricted to males and include failure to attempt copulation, bisexual and homosexual courtship, and,in the strongest allelic combinations, complete lack of male courtship. In addition, fru males lack the male-specific muscles of Lawrence in dorsal abdominal segment 5. retn males have normal muscles of Lawrence, and retn females do not have muscles of Lawrence. In addition, the larval and pupal expression patterns of retn (this paper) and the sex-specific products of the fru P1 promoter(Lee et al., 2000), notably the active male-specific fru proteins, show little or no overlap. This all suggests that fru and retn are unlikely to interact intracellularly and would be expected to be involved in different aspects of behavioral control.

The latter conclusion seems to be contradicted by the male-like courtship generated by retn females, as previous work demonstrates that otherwise wild-type males require FRU-M to generate male behavior(Anand et al., 2001). We have operationally and molecularly shown that the male behavior generated by retn females occurs even in the absence of fru P1 transcripts (Fig. 3G,H).

We have developed a plausible working model that reconciles the data on the necessity of fruM in males and male-like courtship by retnfemales. The largely non-overlapping expression patterns of fru and retn suggests that the formal interactions of this model will result from interactions between networks of fru- and retn-influenced neurons rather than by intracellular regulatory interactions involving FRU-M and RETN, although the model can accommodate either situation.

Our model posits that in the absence of fruM and retn the nervous system has an inherent tendency to set down some rudiments of neural pathways for male courtship behavior (Fig. 8A).

When retn is wild type and fruM is not expressed, as in wild-type females, retn, or cells expressing retn [perhaps in conjunction or parallel with other factors such as dsxF (below)],act to suppress the basal male courtship pathway(Fig. 8B). This blocks male courtship behaviors. This is the case in wild-type females, as shown.

Finally, in wild-type males, fruM or cells expressing fruM, perhaps along with other factors such as dsxM, act to strengthen the male courtship pathway such that the repressive action of retn-expressing cells is overpowered(Fig. 8C). This makes fru the switch that results in male behavior and captures both the requirement for fru⁺ in males, and the male-like courtship by retn females.

This model does not rule out involvement of other components. For example,work by Waterbury et al. (Waterbury et al., 1999) suggests that dsxF can suppress male behaviors in a retn⁺ background. This can be fitted into the model as an additional female-specific block to male behavior in both Fig. 8A and 8B. A simple prediction of such a role for dsx is that reduction of dsx expression in a retn mutant background will enhance the retn phenotype. Recent work involving expression of fru RNAi in a subset of fru neurons suggests a role for temporally repression in the sequencing of male behaviors in courtship(Manoli and Baker, 2004).

An extensive series of experiments is in progress to test predictions of this model. Experiments are also in progress to determine if dsxparticipation fits within the context of the model, and to identify the molecules and mechanisms downstream of retn in the control of behavior.

We thank Erin Gross and Michael Benedetti for substantial effort in the genetic screen yielding retn, Rebecca Wagaman for work mapping the exon 1-6 splice form of retn RNA, and Michael Ludwig for technical assistance. This work was supported by grants from the NIH (MH57460) and NSF(IBN-0315660) to M.McK. and from the NIH (GM-56920, NS033352) to B.J.T. J.L.P. was supported by an American Cancer Society Postdoctoral Fellowship(PF-00-324-01-DDC), while K.D.F. was supported by an National Institute of Neurological Diseases and Stroke Postdoctoral Fellowship. T.S. is currently an NIH Predoctoral Trainee (GM07601).