Partial proteasomal degradation of Lola triggers the male-to-female switch of a dimorphic courtship circuit

In Drosophila, some neurons develop sex-specific neurites that contribute to dimorphic circuits for sex-specific behavior. As opposed to the idea that the sexual dichotomy in transcriptional profiles produced by a sex-specific factor underlies such sex differences, we discovered that the sex-specific cleavage confers the activity as a sexual-fate inducer on the pleiotropic transcription factor Longitudinals lacking (Lola). Surprisingly, Fruitless, another transcription factor with a master regulator role for courtship circuitry formation, directly binds to Lola to protect its cleavage in males. We also show that Lola cleavage involves E3 ubiquitin ligase Cullin1 and 26S proteasome. Our work adds a new dimension to the study of sex-specific behavior and its circuit basis by unveiling a mechanistic link between proteolysis and the sexually dimorphic patterning of circuits. Our findings may also provide new insights into potential causes of the sex-biased incidence of some neuropsychiatric diseases and inspire novel therapeutic approaches to such disorders. It is unclear how some Drosophila neurons develop sex-specific neurites that contribute to dimorphic circuitries required for gendered behavior. The authors show that sex-specific cleavage by the E3 ubiquitin ligase Cullin1 and 26S proteasome of the pleiotropic BTB-ZF transcription factor Lola confers its sexual fate-inducing ability in these neurons.

F emales and males display distinct behavioral patterns. These differences in behavioral patterns are the outcome of the intersex differences in neural circuits 1 . In Drosophila, it has well been documented that sexual dimorphisms in neural circuits result, in principle, from sex differences in the number, structure and function of individual cells composing the circuit [2][3][4][5][6] . The development of these cellular sex differences relies primarily on either or both of two sex determination genes, fruitless (fru) and doublesex (dsx) 5 . However, it is largely unknown how fru and dsx produce sex differences in individual cells. We discovered that the longitudinals lacking (lola) gene is an essential mediator for the fru action to masculinize the mAL cell structure. The lola locus generates extraordinarily large numbers of mRNA variants ( Supplementary Fig. 1), each of which encodes a unique protein isoform. The majority of Lola isoforms share an N-terminal BTB domain implicated in protein-protein interactions 7 and in anchoring ubiquitin proteasome components, followed by an isoform-specific sequence containing a C-terminal zinc finger motif that is a putative DNA-binding region 8 . Interestingly, the fru gene product FruM represents another, BTB-zinc finger protein group, which includes a set of male-specific proteins (i.e., FruAM, FruBM and FruEM: nomenclature according to Ref. 9; Supplementary Fig. 1) that function to masculinize certain neurons [10][11][12] presumably via chromatin remodeling 9 . For example, FruM represses transcription from roundabout1 (robo1), a negative regulator gene for neuritogenesis, thereby allowing malespecific neurite formation in males 13 . Here we demonstrate that the male-biased Lola29M isoform forms a complex with FruBM. We further demonstrate that Lola29M is a precursor of the female-specific Lola29F isoform, an N-terminal truncation product of Lola29M that counteracts Lola29M action so as to inhibit male-specific neurite formation in females. Surprisingly, the male-specific transcriptional repressor FruBM protects Lola29M from its N-terminal truncation, which is mediated by the E3 ubiquitin ligase Cullin1 (Cul1) and 26S proteasome. Thus, by making the masculinizing protein Lola29M resistant to degradation, the male-specific transcription factor FruBM prevents the production of the feminizing protein Lola29F. As a consequence, the male program for dimorphic circuit formation turns on and the female program turns off in the male brain.
In humans, sexually biased incidence of certain neurological disorders has been well recognized, yet the origin of such sex differences remains largely an enigma. For instance, the male-tofemale incidence ratios have been reported to vary from 1.37 to 3.7 in Parkinson's disease, the etiology of which likely involves mitochondrial dysfunction due to defects in the proteasomal degradation system 14 . Our finding in Drosophila that the neuronal sex-type specification involves proteasomal protein processing will shed light on the hitherto unknown mechanistic link among posttranslational protein modification, neural sex differentiation and complex neurobehavioral traits under normal and disordered conditions.

Results
lola as a phenotypic modifier of fru. To obtain insights into the mechanism of fru actions on neural sex-type specification, we here screened for fru modifier genes. In this screen, we took advantage of a gain-of-function effect of fru + to disrupt the compound eye structure when overexpressed in the developing eye disc. Genome-wide searches for genes that can modify the fru eye phenotype were conducted by the Gene-Search (GS) system 15 . In this system, a P-element vector carrying the GAL4responsive DNA sequence UAS (the GS-element) was randomly inserted into the genome of a fly so that transcription units that happened to flank the GS-element insertion could be transcribed in the presence of an arbitrarily chosen GAL4 transgene (Supplementary Fig. 2a). This study used GMR-GAL4 to drive transcription via the GS-element as well as UAS-fruB + in the developing eye disc ( Supplementary Fig. 2b), yielding several enhancers of the fruB + -induced eye phenotype, which included lola ( Supplementary Fig. 2c-e; for other fru modifiers see Ref. 16). Conversely, a loss-of-function lola mutation dominantly suppressed the fru-induced distortion of the eye ( Supplementary  Fig. 2f,g). In subsequent analyses, we will focus on lola because: (1) it encodes proteins of the BTB-zinc finger superfamily to which Fru also belongs and (2) its functions in neurite guidance have been well established 17 . Indeed, we found that reduced courtship toward a female in fru hypomorphic males was dominantly enhanced by two different lola null alleles (lola 03089 and lola ORE76 ), which are both homozygous lethal 18,19 (Fig. 1a), and lola knockdown per se attenuated male courtship activities (Fig. 1b). These results implicate lola in the fru-dependent formation of courtship circuits.
Lola isoforms dedicated to sexual differentiation. We suspected that Fru (more specifically, FruBM, see below) might affect expression of lola to exert its neural masculinizing effect in, for example, sexually dimorphic mAL interneurons in the brain ( Supplementary Fig. 2h), although no sex differences in the structure or expression of Lola isoforms has been reported and despite the numerousness of Lola isoforms identified to date 20 . To obtain hints as to which of the Lola isoforms might have a role in the fru-dependent sexual differentiation, we examined possible effects of isoform-specific knockdown for isoforms 11, 17, 22, 26, 28 and 29, for which UAS-RNAi transgenic strains were publicly available, and found that isoforms 22 and 29 interfered with the sex-specific development of fru-expressing neurons (see below). Our extensive analyses of Lola expression with an antibody raised against an exon29-specific sequence (anti-Lola-exon 29; Supplementary Fig. 3) revealed two isoforms, Lola29M of~110kD and Lola29F of~80kD, the relative amounts of which were different between the sexes in wandering stage third instar larvae (Fig. 1c). The isoform containing the sequence encoded by exon 29 was formally designated as type-Q 20 . We found that Lola29M is more abundant in males than females whereas Lola29F is detectable only in females (Fig. 1d). Immunostaining of the wandering-stage larval CNS with the anti-Lola-exon 29 antibody revealed that Lola29M/F expression is confined to differentiating neurons, including those expressing FruM, but not detected in neuroblasts ( Supplementary Fig. 4). The Lola29M and Lola29F isoforms shared exon29, which was connected to the same 5' exons, and transcribed from the same promoters. Because no sex difference was detected in the mRNA species with exon 29 by 5' and 3' RACE ( Supplementary Fig. 5), the size difference between the Lola29F and Lola29M proteins was considered to result from posttranslational modification of a protein product. Intriguingly, fru mutant (fruM -) males that are null for male-specific FruM proteins expressed the female-specific Lola29F isoform with a concomitant reduction in Lola29M (Fig. 1c, d). In contrast, females sexually transformed into males by the transformer 1 (tra 1 ) mutation lost Lola29F with an elevated level of Lola29M (Fig. 1c, d). These results imply that the posttranslational modification of the Lola protein is under the control of the sex-determination cascade downstream of tra and fru.
Female-specific Lola29F is a truncation product of Lola29M. To examine whether female-specific Lola29F is a truncation product of male-enriched Lola29M, we attempted to determine the structures of Lola29F and Lola29M and the mechanism whereby these two isoforms are produced in a sex-dependent manner. We presumed that Lola29F might be a proteolytic product of Lola29M. To test this idea, we overexpressed a lola29m gene decorated with an N-terminal HA tag and a C-terminal V5 tag in flies under the control of fru-GAL4, and analyzed the lysates from CNS cells by western blotting with antibodies against HA and V5 (Fig. 1e). Remarkably, the anti-V5 antibody detected a shorter band similar to Lola29F (referred to as Lola29F-like hereafter) in addition to the full-length Lola29M, whereas the anti-HA antibody detected only the latter (Fig. 1e). Therefore, the difference between the two isoforms must reside in their Ntermini. Overexpression of a series of N-terminal deletants of lola29m in fru-GAL4-positive cells resulted in the expression of mutant Lola29 proteins; deletion of the N-terminal 300 (Δ1-300) residues yielded a single band of a size similar to that of Lola29F on western blots (Fig. 1f). We infer that Lola29F is produced by the truncation of Lola29M somewhere between residues 250 and 300 of Lola29M, and later analyses by Edman degradation revealed that this is indeed the case (see below).
Because lola plays an important role in neurite patterning, the effects of lola knockdown on courtship behavior (Fig. 1b) likely result from a disturbance in the formation of courtship circuits. It has been established that FruM plays a key role in the courtship circuit formation in the male CNS, and some of the fru mutant defects in mating behavior have been ascribed to impairments of sex-specific development of fru-expressing neurons [10][11][12] . Therefore, to examine the possible role of Lola29F/M on the courtship circuit formation, we next examined fru-expressing mAL interneurons, which exhibit remarkable sex-differences   Fig. 1 lola exon 29 contributes to a set of Lola isoforms differentially expressed between females and males. a, b Male courtship activities. Courtship defects in fru hypomorphic males were enhanced in the lola heterozygous background (a) and knockdown by lola-COM RNAi (3rd bar from the top) or by lola-exon 29 RNAi (5th bar from the top) suppressed male-to-female courtship (b). ***P < 0.001 by the Kruskal-Wallis analysis of variance followed by Steel-Dwass post hoc test (a, b). c Western blotting with anti-Lola-exon 29 of CNS extracts obtained from wild-type (CS) males, wild-type females, fru sat /fru NP21 males and tra 1 homozygous females. d Quantification of Lola29M (left-side panel) and Lola29F (right-side panel), as normalized by the values for the wild-type male. The number of replicates is indicated in parentheses. **P < 0.01; *P < 0.05 by the Kruskal-Wallis analysis of variance followed by Steel's nonparametric multiple comparisons. The box plot shows the median and 10th, 25th, 75th, and 90th percentiles. e Overexpression of a Lola29M-encoding sequence decorated with a HA-tag at the N-terminus and a V5-tag at the C-terminus (HA-lola29m-V5) yielded the Lola29M protein in males and less abundantly in females, as well as a Lola29F-like protein in females. Shown is a western blot of larval CNS extracts probed with anti-V5 that recognizes the C-terminus of Lola29M. f Lola29M[Δ1-300] lacking a.a. 1-300 resulted in a protein, Lola29F-like, with a molecular weight close to that of Lola29F in both sexes. α-Tubulin (α-Tub) was used as a loading control. g-l mAL neurons in the male brain triply stained with antibodies against GFP (g, h, k), FruM (i, l) and Lola29M/F (j-l). GFP was preferentially expressed in mAL neurons by the intersection of fru FLP and 9-189-GAL4. Scale bars: 50 μm (g) or 10 μm (h). m-q Analysis of single-cell mAL clones that express lola-exon29 RNAi. Examples of single-cell clones (m-p) and the proportion of flies that carried clones with or without the ipsilateral neurite (q). Scale bar: 50 µm. ***P < 0.001 by the Fisher's exact probability test. Source data are provided as a Source Data file ( Supplementary Fig. 2h), which have been shown to play a role in the processing of sex pheromone information [21][22][23] . One of the prominent sex differences found in mAL neurons is the absence (in females) or presence (in males) of the ipsilateral neurite. In a subsequent analysis, therefore, we focused on the male-specific ipsilateral neurite formation, where male-specific FruM directly represses transcription of robo1 in males, thereby allowing mAL neurons to extend the male-specific neurite 13 . Gain of the malespecific ipsilateral neurite (i.e., masculinization) in females or fruM-null mutant males can be assessed in neuroblast clones that contain up to 5 mAL neurons, because none of the mAL neurons in control females and control fruM-null males possesses this neurite. Loss of the male-specific neurite (i.e., feminization) in males, in contrast, is only detectable in single-cell clones, because any remaining ipsilateral neurite of unaffected cells hinder the neurite loss from affected cells 13 . We therefore examine neuroblast clones for quantifying the masculinizing effect of genetic manipulations and single-cell clones for the feminizing effect.
We first confirmed the immunoreactivity of mAL neurons to the anti-Lola-exon 29 antibody (Fig. 1g-l). We then tested the effect of lola29 knockdown in these cells. Notably, the expression of RNAi targeting lola-exon29 markedly increased the formation of mAL neurons without the male-specific ipsilateral neurite in male mAL neurons ( Fig. 1m-q). It is noteworthy that the lola-exon29 RNAi used here effectively inhibited Lola29M production in flies, which retained a high level of expression of other lola isoforms ( Supplementary Fig. 6).
To further clarify roles of Lola29M and Lola29F in the sex-type specification of neurites, we first wanted to define their molecular structures and the mechanism whereby two Lola29 forms are produced. To determine the exact site of truncation in producing Lola29F from Lola29M, we employed Edman degradation analysis (Fig. 2a). To obtain a large amount of Lola29F sufficient for amino acid sequencing, we overexpressed lola29m-6xV5 under the control of fru-GAL4 in female larvae, from which the CNSs were dissected to extract proteins for immunoprecipitating lola29m-6xV5 products. The band representing Lola29F on the PVDF membrane visualized with Coomassie brilliant blue (CBB; Fig. 2a) was cut out and analyzed, allowing us to determine that the N-terminal five residues were SSTAA (Fig. 2b). This fiveresidue sequence was uniquely identified as a.a. 264-268 in Lola29M. We therefore conclude that a.a. 1-263 of Lola29M are degraded to yield Lola29F.
In a separate experiment, we detected Lola29F-like, in addition to the expected Lola29M, in S2 cells that had been transfected with lola29m alone without a lola29f minigene when lysates were prepared with a sufficient delay (~2 days) after transfection ( Supplementary Fig. 7). Importantly, the production of Lola29Flike in S2 cells was inhibited by the administration of a ubiquitinproteasome inhibitor, MG-132 ( Fig. 2c) or Lactacystin (Fig. 2d). This result suggests that Lola29F-like is produced by ubiquitinmediated proteolysis of Lola29M in S2 cells.
Polyubiquitination of ubiquitin at lysine 48 (K48) is known to direct substrate proteins to proteasome-mediated degradation 24 . We found that a K48-linkage-specific polyubiquitin antibody coimmunoprecipitates Lola29M/F in lysates from S2 cells transfected with lola29m, and fruBM cotransfection markedly diminishes the yield of immunoprecipitates ( Supplementary  Fig. 8). This result is taken as evidence that Lola29M is indeed ubiquitinated in the absence of FruM.
Ubiquitination takes place at lysine (K) residues. Lola29M has 40 lysine residues, and 15 of them are within the N-terminal 263 amino acid region, which is likely removed in Lola29F (Fig. 2e). We replaced these 15 lysine residues with arginine (R) one by one to examine whether any of these mutations confer to Lola29M the resistance against proteolysis. Notably, four mutants, K41R, K44R, K47R and K67R, showed a marked reduction in the Lola29F-like production (Fig. 2f). The K41R mutation similarly protected Lola29M from proteolysis in vivo (Fig. 2g). Taking all these results into account, we conclude that Lola29F is a proteolytic product of Lola29M (Fig. 2h).

Lola29M masculinizes whereas Lola29F feminizes the neurons.
Based on knowledge of their molecular properties, we attempted to determine the roles of Lola29M and Lola29F on sex-specific neurite identity. We found that overexpression of Lola29F-like inhibits the ipsilateral neurite formation; the proportion of singlecell clones without the ipsilateral neurite was significantly increased upon Lola29F-like overexpression in males (Fig. 3a-c). However, mAL neurons with the ipsilateral neurite were still produced, though at a reduced rate, in male flies with Lola29Flike overexpression, implying that an additional, Lola29Fresistant mechanism operates for the ipsilateral neurite formation.
We next examined the effect of overexpression of Lola29M  Fig. 9b) and some fru mutant males (Fig. 3e), which otherwise lack it without exception ( Fig. 3d and Supplementary Fig. 9a). It is conceivable that endogenous Lola29F present in females and fru mutant males (see Fig. 1c) hampered the masculinizing action of overexpressed Lola29M[K41R], such that formation of the ipsilateral neurite in mAL neurons was observed in only 20-40% of flies examined ( Fig. 3h and Supplementary Fig. 9d). This notion was supported by the observation that the ability of Lola29M[K41R] to restore the male-specific neurite in fru mutant males was canceled out by the additional overexpression of Lola29F-like (Fig. 3g). In contrast, overexpression of Lola29F-like alone via an Nterminally deleted lola29m[Δ1-300] transgene in females (Supplementary Fig. 9c) or fru mutant males (Fig. 3f) had no effect on the mAL neurite structure, which continued to be the female-type in all flies examined.
To examine whether Lola29M and Lola29F are involved in the neural sexual differentiation of neurons other than mAL, we overexpressed truncation-resistant Lola29M or Lola29F-like in fru-expressing pheromone receptor neurons ( Fig. 4a-g), the central projection of which is sexually dimorphic; male afferent fibers cross the midline (Fig. 4a, b) whereas female counterparts terminate in the ipsilateral neuromere in the prothoracic ganglion 25 . fru hypomorphic mutant (fru 2 /fru sat ) males have a reduced midline crossing (Fig. 4c), offering a sensitized genetic condition in which subtle phenotypic changes can be unambiguously detected 9 . We found that truncation-resistant Lola29M significantly increased (Fig. 4d) and Lola29F-like significantly decreased (Fig. 4e) the midline crossing of sensory axons (Fig. 4g), suggesting that Lola29M and Lola29F contribute to certain dimorphic features of different types of fru-expressing neurons. We conclude that Lola29M promotes the male-typical neurite formation and Lola29F counteracts the action of Lola29M.
Lola29F and Lola29M are required for mating behavior. To evaluate the contributions of Lola29M and Lola29F to the execution of mating behavior, we measured the courtship index for male courtship vigor and the time to copulation for female receptivity in flies which expressed lola exon 29-RNAi together with lola29m[K41R] encoding truncation-resistant Lola29M or lola29m[Δ1-300] encoding Lola29F-like in fru-GAL4-positive cells. We found that lola exon 29-RNAi effectively reduced the courtship index to~40% of the value in control males (Fig. 4h), a level of inhibition comparable to that obtained with lola-COM RNAi, which knocked down nearly all types of lola mRNAs ( Fig. 1b). As mentioned above, our western blot analysis confirmed that both lola-COM RNAi and lola-exon29 RNAi effectively reduced Lola29M expression ( Supplementary Fig. 6  Lactacystin:  whereas Lola29F-like failed to do so (Fig. 4h). Conversely, in females, Lola29F-like overexpression via fru-GAL4 mitigated the inhibitory effect of lola-exon29 RNAi expression on female mating, whereas Lola29M[K41R] did not (Fig. 4i, j). Because male flies paired with test females of different genotypes were equally active in courtship (Fig. 4k), the observed reduction in mating success in this experiment was ascribable to a change in female receptivity. These results suggest that Lola29F and Lola29M play sex-specific functions in females and males, respectively, to ensure successful courtship and copulation. We note, however, that neither Lola29M nor Lola29F-like was able to resume the mating activity to the normal level in male or female flies expressing lola-exon29 RNAi. The partial rescue of mating behavior by overexpressed Lola29M or Lola29F-like might suggest that mRNAs derived from the transgenes were also targeted by the lola-exon29 RNAi to a certain extent.
Lola29F prevents Lola29M from binding to its target. To explore the mechanism by which Lola29F counteracts Lola29M, we first examined the possible effect of Lola29F/M on the transcription of robo1, a direct downstream target gene of FruBM 13 . Of note, Crowner et al. 18 have reported that a robo1 mutant copy dominantly enhances the axon misrouting phenotype in a weak hypomorph of lola, lola ORE120 , which barely manifests this phenotype on its own. Our quantitative RT-PCR experiment with male fly extracts revealed that overexpression of Lola29M decreased robo1 transcripts, whereas overexpression of Lola29Flike had no effect on its own (Fig. 4l). Interestingly, the reduction in robo1 transcript levels by Lola29M was not detected when Lola29F-like was co-overexpressed (Fig. 4l). In keeping with the RT-PCR result, overexpression of Lola29M[K41R] in the larval CNS reduced its immunoreactivity to the anti-Robo1 antibody ( Supplementary Fig. 10). These results support the hypothesis that Lola29F inhibits the transcriptional repressor activity of Lola29M.
To further explore the mechanism by which Lola29F counteracts Lola29M, we conducted reporter assays. A series of reporters with a luciferase-coding sequence fused to robo1 promoter fragments of varying length (Fig. 5a) were transfected into S2 cells with or without Lola29M and/or FruBM. The reporter activity was repressed by Lola29M, irrespective of whether FruBM was present (Fig. 5b), provided that the reporter contained a robo1 promoter fragment of 0.9 kb or longer (Fig. 5c). Truncation-resistant Lola29M[K41R] was equally effective as wild-type Lola29M in repressing the robo1 reporter activity (Fig. 5c). Notably, Lola29F-like encoded by lola29m[Δ1-300] had no effect on the robo1 reporter activity (Fig. 5d). Intriguingly, the ability of Lola29M to repress reporter activities was completely blocked by the addition of Lola29F-like (Fig. 5e).
An in silico search for the putative Lola-binding motif within the Lola-responsive 0.9 kb region of the robo1 promoter revealed an 18-bp direct repeat (DR), G C A C T A A A G A G C A G G A A A, which we named DR1 (Fig. 5f, g). The 0.9 kb reporter construct lost its sensitivity to Lola29M when DR1 was deleted from the promoter fragment (i.e., the 0.9 kb ΔDR1 reporter; Fig. 5h). Interestingly, DR1 was located immediately 3' to Pal 1, a palindrome sequence shown to be essential for a 42-bp Fig. 3 Lola29M promotes whereas Lola29F impedes male-typical neuritogenesis. a-c Analysis of single-cell mAL clones overexpressing Lola29F-like. Examples of single-cell clones (a, b) and the proportion of flies that carry single-cell mAL clones with or without the ipsilateral neurite (c). Scale bar: 50 µm. Statistical differences were evaluated by the Fisher's exact probability test (*P < 0.05). d-h mAL neurons in fru mutant (FruM protein-null) males without (d) or with overexpression of truncationresistant Lola29M[K41R] (e) or Lola29F-like (f), or both (g). Representative MARCM clones are shown. h Quantitative comparisons of the percentage of mAL neuroblast clones that carried ipsilateral neurites **P < 0.01 by the Fisher's exact probability test. Scale bars: 50 μm UAS-lola29m  a UAS-lola29m   FruBM-binding region, i.e., the FruBM response obligatory sequence (FROS) 13 , to bind to FruBM (Fig. 5g). We conclude that DR1 is critical for the response of the robo1 promoter to Lola29M.
To determine whether the observed reduction in robo1 transcript levels by Lola29M overexpression is a result of Lola29M binding to the robo1 promoter, we carried out electromobility shift assays (EMSA) for robo1 promoter DNA fragments with V5-tagged Lola29M[K41R] in the absence and presence of unlabeled competitor DNAs (Fig. 5i). We anticipated that Lola29M binds to the DNA fragment carrying FROS (probe DNA B in Fig. 5f). Our EMSA results supported this scenario. A    /fru sat , poxn-GAL4 *** ***  Notably, probe DNA B without DR1 (Probe DNA B ΔDR1) was unable to produce a retarded band (Fig. 5k). We conclude that , which, on its own, did not produce the retarded band either (Fig. 5m). We conclude that Lola29F inhibits Lola29M binding to the robo1 promoter, and consequently derepresses robo1 transcription. DR1 lies just outside FROS and, therefore, is dispensable for FruBM binding to the robo1 promoter 13 . Remarkably, however, male flies homozygous for robo1 Δ4 , a robo1 mutant carrying a 10 bp deletion that removes DR1, exhibited precocious wing switching during courtship (Supplementary Fig. 11 and Supplementary Movie 1), which represents a behavioral phenotype uniquely observed in flies with defects in the male-specific neurite formation of mAL neurons 13 . For example, robo1 Δ1 and robo1 Δ2 mutations that delete a part of the core palindrome Pal1 in FROS dominantly induce the precocious wing switching in male flies 13 . The robo1 Δ4 mutation was recessive in inducing the precocious wing switching and had no dominant effect ( Supplementary  Fig. 11), suggesting that Lola29M bound to DR1 plays a distinct role that is needed for FruBM to fully repress robo1 transcription. Next, we examined whether a manipulation of the robo1 gene dosage could modify the effect of Lola29M in vivo. More specifically, the effect of Lola29M[K41R] overexpression on the male-specific neurite formation was compared between females with the wild-type robo1 and females heterozygous for a lethal loss-of-function allele, robo1 GA285 , which carries a premature termination codon in place of Q411 (http://flybase.org/reports/ FBgn0005631.html). We found that the male-specific neurite was produced by overexpressed lola29m[K41R] at a higher rate in the presence of robo1 GA285 than in its absence (Fig. 6a, b), and the difference was statistically significant (Fig. 6c). Thus, the effect of Lola29M[K41R] in promoting the male-specific neurite was enhanced by the robo1 mutant heterozygosity, which, by itself, has no phenotypic effect. This finding supports the notion that lola and robo1 interact in vivo.

Lola29M cleavage involves the ubiquitin proteasome pathway.
To identify molecules involved in the proteolytic degradation of the Lola29 N-terminus, we analyzed the protein complex that was pulled down with Lola29M in immunoprecipitation assays. We overexpressed Lola29M that lacked a.a. 1-150 (Lola29M[Δ1-150]), as we thought that the N-terminal loss might stimulate further proteolysis of Lola29M. Mass spectrometric analysis identified 121 proteins that were precipitated with Lola29M, including Cul1, an E3 ubiquitin ligase 26 , raising the possibility that this enzyme might be involved in the observed Lola29M truncation (Supplementary Table 1). Expression of Cul1 in mAL neurons was confirmed by the immunostaining with an anti-Cul1 antibody ( Supplementary Fig. 12). We conducted a MARCM analysis in which Cul1 was knocked down in mAL clones.
Remarkably, mAL neurons with Cul1 knockdown led to the malespecific neurite formation in females (Fig. 6d). Moreover, the production of Lola29F was markedly suppressed in females expressing Cul1 RNAi via fru-GAL4 (Fig. 6g). These observations are in keeping with the idea that Cul1 is involved in the Nterminal truncation of Lola29M. Ubiquitinated proteins are known to be degraded by the 26S proteasome 27 . We thus examined the effect of knocking down Proteasome subunit β5 (Pros-β5), a gene encoding a 26S proteasome subunit, on the mAL structure in the female brain. As shown in Fig. 6e, the expression of Pros-β5 RNAi induced formation of the male-specific neurite in female mAL neurons (Fig. 6f). This observation is compatible with the hypothesis that the ubiquitin-proteasome pathway is involved in the N-terminal truncation and resulting conversion of Lola29M into Lola29F in female mAL neurons.

FruM inhibits the Lola29M truncation and Lola29F formation.
A question arises as to why proteolysis of Lola29M occurs in females but not males. An interesting possibility is that the malespecific protein FruM protects Lola29M against proteolysis. First, we tested the possibility that Lola29M forms a complex with FruM, because both proteins have a BTB domain, through which they may interact directly with each other 28 . As shown in Fig. 6h, an anti-Lola-exon29 antibody precipitated FruM together with Lola29M in flies. Moreover, a Lola29M-recognizing antibody failed to precipitate FruM, when the BTB domain of either FruM or Lola29M was deleted in an experiment with cotransfected S2 cells ( Fig. 6i and Supplementary Fig. 13). These results support the hypothesis that Lola29M and FruBM form a complex through each other's BTB domains.
Second, we examined the effect of increased FruBM expression on the production of Lola29F in S2 cells transfected with lola29m. FruBM has been shown to be the major FruM isoform that contributes to neural sex differences 13,29 . We found that as the amount of FruBM increased the amount of Lola29F decreased (Fig. 6j, k). Notably, FruBM lost its ability to inhibit the Lola29F production when its BTB domain but not its zinc finger domain was deleted (Fig. 6l). These results demonstrate that FruBM binding to Lola29M, presumably as mediated by the BTB domains of two proteins, protects Lola29M against proteolysis, thereby preventing the production of Lola29F in males. Because female neurons lack FruM, Lola29M can be processed into Lola29F.

Discussion
In this study, we showed that Lola29M, and its truncation product, Lola29F, play important roles in the masculinization and feminization of one of the three types of sexual dimorphism exhibited by fru-expressing mAL neurons: Lola29M promotes the male-specific ipsilateral neurite formation in males, whereas Lola29F counteracts Lola29M and prevents the male-specific neurite from forming in females (Fig. 7). Our results support the hypothesis that males have only the full-length protein Lola29M, because male-specific FruBM binds to Lola29M and protects it from truncation, whereas females have, in addition to full-length Lola29M, N-terminally truncated Lola29F due to the absence of FruBM. As a consequence, the Lola29M action is inhibited by Lola29F in females. This would mean that Lola29M exerts its masculinizing action only when FruBM is present, representing an efficient and secure means to induce masculine characteristics in fru-expressing neurons.
Our EMSA results demonstrated that Lola29M and FruBM share the same transcriptional target gene robo1, the transcription of which is repressed by each protein or by both proteins in synergy, according to our reporter assay data. The robo1 gene encodes a transmembrane receptor of the immunoglobulin superfamily, Robo1 30 , which is a key effector to prevent the male-specific ipsilateral neurite from forming in  13 . Lola29M induces the male-specific ipsilateral neurite by transcriptional repression of robo1 (acting as a negative regulator of neuritogenesis) in males, whereas Lola29F appears to compete with Lola29M, allowing robo1 to be transcribed, and as a result, Lola29F blocks the male-specific neurite formation in females. Our EMSA data identified two distinct DNA stretches, one harboring a Lola29M-binding site and the other a FruBM-binding site 13 in the robo1 promoter region, which were located side-by-side. On the other hand, our immunoprecipitation assays indicated that Lola29M and FruBM form a complex in vivo. It might be that Lola29M and FruBM each bound to their own binding sites interact in trans, leading the robo1 promoter to a conformation unfavorable for transcriptional activation.
We showed that the replacement of four lysine residues, K41, K44, K47 and K67, which are potential ubiquitination-targets, with arginine residues conferred resistance to N-terminal truncation on Lola29M, suggesting that the ubiquitin proteasome pathway is involved in the Lola29M processing. We also showed that the E3 ubiquitin ligase Cul1 is required for the Nterminal truncation of Lola29M, and that the male-specific neurite formation is induced by Cul1 knockdown in female mAL neurons. In natural killer T-cell thymocytes, the E3 ubiquitination enzyme Cul3 is recruited to a chromatin-modifying complex, where it then induces changes in the ubiquitination patterns of other components of the same complex 26 . This recruitment of Cul3 to the chromatin-modifying complex is mediated by BTB zinc finger proteins such as PLZF and BCL6 26 . Therefore, it may be possible that Lola29M truncation is similarly mediated by a component in the protein complex to which Lola29M contributes. In this context, it would be worth mentioning that Cul1 and Cul3 play contrasting roles in the regulation of hedgehog (hh) signaling; in the absence of Hh, Cul1 is recruited to the Hh downstream zinc finger protein Cubitus interruptus (Ci155), which undergoes, as a consequence, partial degradation to yield the transcriptional repressor Ci75, whereas upon the Cul3 recruitment in the presence of Hh, Ci155 is completely degraded 31 . It remains to be determined how Cul1 is recruited to Lola29M for its partial degradation in fru-positive neurons.
A recent study documented neural activity-dependent activation of lola transcription, leading to the proposition that lola is an immediate early gene in insects 32 . The transient expression common to immediate early gene products is caused by the rapid degradation that follows their rapid induction. An intriguing possibility is that Lola proteins are vulnerable to rapid degradation through ubiquitination, unless their ubiquitination target sites are masked by a binding partner, such as FruBM, that interacts through the BTB domain. The known pleiotropic functions of Lola invite the supposition that other Lola isoforms expressed in a variety of cell types might form a complex with the different BTB proteins expressed there, accomplishing diversified roles unique to the individual cell type.
The present work thus provides new insights into the molecular mechanism whereby a common transcription factor such as Lola can play distinct roles in different cells in a contextdependent manner. A growing body of evidence indicates that fate changes in a variety of cell lineages in both vertebrates and invertebrates are induced by rapid epigenetic remodeling by chromatin regulators, such as BTB-zinc finger transcription factors, some of which are known to recruit E3 ubiquitin ligases to the chromatin-modifying complex and thus alter its ubiquitination pattern. It is therefore plausible that the proteolytic removal of the N-terminal BTB domain from these transcription factors within the chromatin-modifying complex is a prevalent means to convert the epigenetic marks.
Moreover, there is accumulating evidence that the association of BTB proteins with ubiquitin ligases plays multilayered regulatory roles in developmental decisions. For example, Germ cellless (Gcl), a BTB protein conserved from C. elegans to humans with a transcriptional repressor activity 33,34 , plays a key role in the soma-germ fate switch via a non-transcriptional mechanism. When complexed with Cul3, Gcl exits the nucleus to degrade the somatic fate determinant Torso and promotes the primordial germ cell fate in Drosophila 35 . In vertebrates, the Cul3-KBTB18 complex switches the translational program so as to specify the neural crest fate 36 . The present study further expanded the roles of the ubiquitin proteasome system-namely, the system was shown to function in the specification of sex-types of neurons via regulated proteolysis of a transcription factor that functions in the sex-determination molecular machinery.

Methods
Fly strains. Flies were reared on cornmeal-yeast medium at 25°C. Canton-S served as a wild-type control. The following fru alleles were used: fru sat , fru 2 , and fru NP21 . The GAL4 driver fru NP21 used in this study is a recessive allele of fru induced by a P-element insertion into the fru second intron ( Supplementary Fig. 1a), which exhibited no discernible defect in neurite structures. A subset of mAL neurons lacked the ipsilateral neurite in fru NP21 /+ heterozygous males, as were the case in males of some flylight lines 37 , which carried driver GAL4 insertions at genomic sites unrelated to the fru locus 3 . The GS2169 strain (#200307) carrying a GS vector inserted into the 5' region of the lola gene was obtained from the Drosophila Genetic Resource Center (Kyoto, Japan). elav-Gal4 C155 (elav-Gal4, #458), Modifier screens. The female flies with both GMR-GAL4 and UAS-fru-typeB + (denoted as UAS-fruB + in Supplementary Fig. 2) transgenes were crossed with male flies from GS P-element insertion stocks 15 or those from mutant stocks reported to have developmental defects in the nervous system (FlyBase: https:// flybase.org/). In this screen, we overexpressed a FruCOM protein rather than a FruM protein when inducing the rough eye phenotype, as the former yielded more viable offspring. The nomenclature for Fru isoforms is adapted from our previous study 38 and different from that of other groups 39 . 1364 stocks were screened, resulting in 40 dominant suppressors of the rough-eye phenotype induced by fru-typeB + overexpression. Among the 5 Fru C-terminal variants, TypeB was most effective in rescuing the fru sat mutant phenotype 38 and thus was most likely to yield modifiers that were relevant to the in vivo functions of fru. Images of the compound eye surface were obtained with a scanning electron microscope (SU8000; Hitachi High-Technologies, Tokyo, Japan).
Antibody production. A rabbit polyclonal anti-Lola-exon 29 antibody (Fig. 1c) was raised against a mixture of two 19-mer peptides, HARQEYIKIDTSRLEDKML and YRSDLRKHMNQKHADSGEA, which were both encoded by exon 29 of the NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-018-08146-1 ARTICLE NATURE COMMUNICATIONS | (2019) 10:166 | https://doi.org/10.1038/s41467-018-08146-1 | www.nature.com/naturecommunications mAL neurons, larvae at 5-6 days AEL were heat shocked at 37°C for 1 h. Flies to be tested were reared at 29°C after the heat shock in order to enhance the expression of transgenes. The number of clones examined was different from experiment to experiment as a consequence of the stochastic nature of mosaic generation. This might complicate interpretations of data, particularly in the experiment shown in Figs. 3c, 6c and 6f, where the proportion of neurons with the ipsilateral neurite was quantitatively compared. We therefore performed a post hoc power analysis of these data with the aid of Webpower 43 , which yielded the following results: Power=0.7194, α=0.05 for Fig. 3c; power=0.3095, α=0.05 for Fig. 6c; power=0.9774, α=0.05 (fru NP21 /+ vs. fru NP21 /UAS-Cul1 RNAi) and power = 0.8209, α = 0.05 (fru NP21 /+ vs. fru NP21 /UAS-Pros-β5 RNAi) for Fig. 6. Here, the sample number is considered to be large enough when it gives a power value larger than 0.7. The data shown in Figs. 3c and 6f fulfilled this criterion but those in Fig. 6c did not. Our calculations indicate that power values over 0.7 would be obtained if 23 or more samples were subjected to analysis in the experiment shown in Fig. 6c (cf. 8-15, the current sample numbers). However, the flies with the genotype robo1 GA285 /+; fru NP21 /UAS-lola29M K41R were extremely difficult to obtain as most of the individuals died before eclosion. Because the statistical difference at P<0.01 found in this set of data stands regardless of whether the power exceeds 0.7 or not, we consider that the result in Fig. 6c supports our conclusion that robo1 and lola (Lola29M) interact in vivo.
Midline crossing score analysis. The MCS (Fig. 4g) was calculated as described previously 9 . Briefly, stacked images of each sample were summed up using ImageJ, and a circle of 8 μm (marked as "a" in Fig. 4f) was drawn on the resultant image so that the circle was centered at the prothoracic midline, where trans-midline axons are expected to run in males. The fluorescent intensity within the circle marked "a" was measured to quantify the level of midline crossing by fibers. Similarly, the fiber tracts locating lateral to the midline were quantified within circles "b1" and "b2" for the fluorescent intensity. To normalize for the background fluorescent level, the areas with no fibers delineated by circles "c1" and "c2" were also measured.  Supplementary Fig. 11), virgin males of the indicated genotypes were collected at eclosion and aged for 5-7 days. Canton-S virgin females were similarly prepared as mating partners of test males. In the behavioral assays, a male of each genotype and a Canton-S virgin female were paired in a small chamber of 8 mm in diameter and 3 mm in height. The flies were video recorded for 5 min. The courtship index (CI) was determined as the percentage of time that the male courted the females during a 5 min observation period. In calculating the CI, the time spent for all courtship elements, i.e., orientation, tapping, following, wing extension/vibration and attempted copulation, was included. The wing switching index was estimated by the method as described in ref. 13 . For the analysis of female mating behavior (Fig. 4i-k), virgin females of each genotype were collected at eclosion and aged for 5-7 days. Each female fly was transferred to a small chamber (8 mm in diameter and 3 mm in height) with a Canton-S virgin male. The behavior of the fly pair was recorded using a video recorder. To estimate the level of female sexual receptivity, the cumulative number of copulating pairs in a 1 h observation period was counted and compared among the fly groups of different genotypes.
RACE PCRs. 5ʹ and 3ʹ RACE PCRs ( Supplementary Fig. 5) were performed using the SMARTer RACE 5ʹ/3ʹ kit (Clontech, Z4858N). Poly A(+) RNA was extracted from the CNS of a wandering third instar larva of a Canton-S female using the Micro-FastTrack 2.0 kit (Life Technologies, K1520-02). RACE-ready cDNA was synthesized from the extracted RNA. For 5' RACE, the forward primer was a mixture of oligo named "5' RACE primer (F)", which was used in conjunction with the reverse primer named "Primer 1 (R)", a gene-specific primer. For 3' RACE, the forward primer was the gene-specific "Primer 2 (F), which was used in conjunction with the "3 RACE primer (R)", a mixture of oligo. For 3ʹ and 5ʹ RACE PCR, the following gene-specific primers were used: Primer 1 (accatccagcaatcgcagacgcatgc) and Primer 2 (cgcaggagcatcttgtcctccagcct), respectively. The cycling parameters for 3' and 5' RACE-PCR were as follows: 1 cycle of 94°C for 2 min, 30 cycles of 94°C for 30 seconds and 63°C for 30 seconds and 68°C for 2 min, followed by 68°C for 10 min. DNA fragments obtained by 5ʹ and 3ʹ RACE PCR were subcloned into the pGEM-T vector and sequenced with ABI 3500 Genetic Analyzer (Applied Biosystems).
Mass spectrometric analysis. Aiming at identifying molecules that contribute to the N-terminal truncation of Lola29M, we carried out mass spectrometric analysis of proteins precipitated with a partially truncated Lola29M in the presence of the anti-V5 antibody. More specifically, the pMT-HA-lola29m[Δ1-150]-V5 vector was transfected into S2 cells, and the expression of proteins was induced by the addition of copper sulfate. The HA-Lola29M[Δ1-150]-V5 products were immuno-purified using Anti-V5-tag mAb-Magnetic Beads (MBL, M167-11), trypsinized, and then directly subjected to LC-MS/MS analysis as previously described 44 . Briefly, extracted peptides were analyzed by ESI-MS/MS using an LTQ velos Orbitrap ETD instrument (Thermo Fisher Scientific, Pittsburgh, PA, USA). MS spectra were recorded over a range of 321−1600 m/z, followed by data-dependent collisioninduced dissociation (CID) MS/MS spectra generated from the 15 highest intensity precursor ions. For protein identification, spectra were processed using Proteome Discoverer 1.3.0.339 (Thermo Fisher Scientific) against MASCOT algorithm. For protein database searches, SwissProt (Taxonomy: Drosophila (fruit flies)) was used.
The following parameters were used for the searches: tryptic cleavage, up to two missed cleavage sites, and tolerances of ± 10 ppm for precursor ions, and ± 0.8 Da for MS/MS fragment ions. MASCOT searches were performed allowing optional methionine oxidation and N-terminal acetylation, and fixed cysteine carbamidomethylation. Peptide data were filtered using a Mascot significance threshold of 0.05 as valid identification.
Statistical analysis. Statistical analyses were done by GraphPad Prism 7.0b software.
Reporting Summary. Further information on experimental design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The data sets generated for this manuscript are available from the corresponding author upon reasonable request. List of proteins that were immuno-purified with Lola29M protein is found in Supplementary