Sex-switching of the Drosophila brain by two antagonistic chromatin factors

Abstract

In Drosophila melanogaster, the fruitless (fru) gene encoding BTB-Zn-finger transcription factors organizes male sexual behavior by controlling the development of sexually dimorphic neuronal circuitry. However, the molecular mechanism by which fru controls the sexual fate of neurons has been unknown. Our recent study represents a first step toward clarification of this mechanism. We have shown that: (1) Fru forms a complex with the transcriptional cofactor Bonus (Bon), which recruits either of two chromatin regulators, Histone deacetylase 1 (HDAC1) or Heterochromatin protein 1a (HP1a), to Fru-target sites; (2) the Fru-Bon complex has a masculinizing effect on single sexually-dimorphic neurons when it recruits HDAC1, whereas it has a demasculinizing effect when it recruits HP1a; (3) HDAC1 or HP1a thus recruited to Fru-target sites determines the sexual fate of single neurons in an all-or-none manner, as manipulations of HDAC1 or HP1a expression levels affect the proportion of male-typical neurons and female-typical neurons without producing neurons of intersexual characteristics. Here, we hypothesize that chromatin landscape changes induced by ecdysone surges direct the HDAC1- or HP1a-containing Fru complex to distinct targets, thereby allowing them to switch the neuronal sexual fate in the brain.

fru as the Major Masculinization Factor of the Neural Circuitry

Animals including humans, exhibit sexually distinct behavior patterns which are likely generated by sexually dimorphic neural circuitries.¹ In the central nervous system (CNS) of Drosophila melanogaster, some neurons form male-specifically. The best-characterized example of this is the P1 (also called pMP4)² cluster; it consists of about 20 neurons locating in the dorsal posterior region of the male brain.^2,3 A primary neurite of the P1 neuron extends to the contralateral protocerebrum and bifurcates near the mushroom body peduncle. The P1 neurons play an essential role in the initiation of male-type courtship behavior.³ Some other neurons are present in both sexes yet have dimorphic arborizations. For instance, mAL neurons (also called aDT-b or aDT2 neurons)^2,4 show sexual dimorphisms in the cell number and projection pattern.⁵ The mAL neurons function to integrate gustatory pheromone inputs and transmit the resultant signal from the subesophageal ganglion to the superior lateral protocerebrum.⁶

Loss-of-function alleles of the fruitless (fru) gene disrupt many aspects of courtship behavior and induce male-directed courtship in males.^7,8 Conversely, when male-specific fru transcripts are constitutively expressed in females by gene-targeted modification of the fru gene, they cause masculinization of courtship behavior and female-directed courtship in females.⁹ Based on these observations, the fru gene has been proposed to be the master control gene that organizes the brain centers for male sexual behavior. The Fru protein is detectable in the male but not the female CNS. This suggests that the sexual differences in “fru-expressing” neurons are produced depending on the presence or absence of the Fru protein.^3,5 However, the mechanism by which the Fru protein organizes the sex-specific neural circuitry has been enigmatic.

Bon-Dependent Association of Fru with Two Chromatin Factors

The Fru protein has a BTB domain in the N-terminus and two Zn-finger motifs in the C terminus, implying that it plays a role as a transcription factor.^7,8 Recently, we performed genetic screens for phenotypic modifiers of fru in the compound eye with the aim of identifying proteins that might interact with the Fru protein for neural masculinization. In this attempt, we found that a null mutation in the bon gene suppresses the phenotypic effects of fru overexpression in the eye.^10,11 In our subsequent analysis, a bon null mutation was shown to exacerbate courtship behavior defects observed in male flies of fru hypomorphic mutants (fru²/fru^sat).¹¹ Bon is a fly homolog of mammalian TIF1 family transcription cofactors. TIF1s are known to recruit, when tethered to a promoter, histone deacetylases such as HDAC1 (also known as the protein product of the Rpd3 gene in Drosophila) and HP1a (also known as the protein product of the Su(var)205 gene in Drosophila) to repress transcription through the action of histone modification.^12,13 Our immunoprecipitation assays with Drosophila head extracts or S2 cell lysates showed that Bon mediates associations of Fru with HDAC1 or HP1a, which compete with each other for Bon.¹¹ Moreover, polytene chromosome immunostaining revealed that HDAC1 and HP1a colocalize with Fru at a number of Fru-binding chromosomal sites in a Bon-dependent manner.¹¹

To determine whether HDAC1 and HP1a mediate functions of endogenous Fru in the male CNS, we examined the effect of null mutations in Rpd3 and Su(var)205 on the courtship activity of hypomorphic fru mutant males. Interestingly, contrary to our expectations, the Rpd3 mutation leads to a phenotype opposite to that of the Su(var)205 mutation: a reduction in Rpd3 enhances the fru mutant phenotype, decreasing male courtship activity, whereas a reduction in Su(var)205 mitigates the fru defects, increasing male courtship activity. The courtship-stimulating effect of the Su(var)205 mutation is lost when the mutant males are also heterozygous for a null mutation in bon.¹¹

Sex Switching of Individual Neurons by Two Antagonistic Chromatin Factors

To investigate whether these chromatin factors are involved in the formation of sexually dimorphic neural circuitry, we examined the effects of changing Fru, HDAC1 and HP1a expression levels on the development of the mAL neuron cluster. Using the MARCM method, neuroblast clones were produced to visualize the entire mAL cluster or single-cell clones to visualize the structure of individual mAL neurons.¹⁴ The mAL cluster exhibits three remarkable sex-specific characteristics:⁵ (1) the cluster is made up of approximately 30 neurons in males, and 5 neurons in females (Fig. 1A and B); (2) the dendritic arbors in the subesophageal ganglion exist exclusively on the contralateral side in females (Fig. 1A), whereas some mAL neurons in males extend to bilateral (ipsilateral in addition to contralateral) arbors (Fig. 1A); (3) the tip of the dendritic tree splits into two branches, giving rise to a “Y-shape” appearance in females, whereas the male counterpart has a simple structure like a horsetail (Fig. 1A). In males with a strong fru (fru^NP21/fru^sat) mutation that lack the Fru protein in their brain, mAL neurons are completely feminized with respect to all three of the characteristics described above (Fig. 1B).⁵ When Rpd3 expression is reduced by knockdown of Rpd3 in males, a fraction of neurons (approximately 20%) in the mAL cluster acquire the female-type structure (Fig. 1B). In addition, the total number of neurons is reduced, phenocopying the female mAL cluster. In moderately hypomorphic fru mutant (fru^NP21/fru²) males, one half of the neurons in the cluster acquire the female-type structure, and the total number of neurons is drastically decreased (Fig. 1B). We next examined the effect of knocking down Su(var)205 in fru^NP21/fru² males. Knockdown of Su(var)205 mitigates the fru mutant effect; i.e., the total number of neurons is increased; more than 70% of neurons now have the male-type structure (Fig. 1B). These results suggest that the wild-type form of HDAC1 functions to masculinize mAL neurons and the wild-type form of HP1a to feminize them. These chromatin regulators appear to function in an all-or-none manner for determination of the sexual fate of mAL neurons, because none of the single neurons exhibit intersexual characteristics. Instead, each mAL neuron has either a male-typical or female-typical structure.¹¹

Figure 1. (A) Structures of mAL neuron clusters and single mAL neurons within an individual cluster. The mAL neuron clusters and single mAL neurons are visualized as neuroblast and single-cell clones using the MARCM method, respectively. (B) Effects of manipulations of fru, Rpd3 and Su(var)205 on the total cell number and proportion of male-typical neurons and female-typical neurons in the mAL cluster. The results reported by Kimura et al. (2005)⁵ and Ito et al. (2012)¹¹ are summarized in this figure.

Axonal projection patterns of foreleg gustatory receptor neurons (GRNs) also exhibit sexual dimorphism; i.e., GRN axons project across the midline of the ventral nerve cord in males but not females. The male-specific midline crossing is impaired by fru mutations.¹⁵ Knockdown of Rpd3 enhances, and that of Su(var)205 alleviates, the fru mutant effect on the male-specific midline crossing. These results support the idea that HDAC1 and HP1a counteract each other in the formation of the sexually dimorphic projections of GRNs.¹¹

How do the Chromatin Factors Control the Neuronal Sex?

Microarray assays revealed that a large number of genes are up- or downregulated by Fru in whole bodies (236 genes) or CNS tissues (94 genes) of male pupae at 48 h after puparium formation (APF).¹⁶ Similarly, numerous genes are up- or downregulated by individual Fru isoforms, Fru^MA (156 genes), Fru^MB (116 genes) or Fru^MC (109 genes), in male pupae at 48 h APF.¹⁶ Apart from the genes predicted by this microarray analysis, proapoptotic genes, head involution defective (hid), grim and reaper (rpr), are potential targets of Fru, because female-specific cell death is responsible for the smaller number of cells in the female mAL cluster. In females with MARCM clones that are homozygous for Df(3L)H99 in which the hid, grim and rpr genes are deleted, the number of mAL neurons increases to a level similar to that in wild-type males.⁵ In addition, males with a strong fru mutation (fru^NP21/fru^sat) that lack the Fru protein in the brain carry 5 mAL neurons, just as wild-type females do.⁵ These results suggest that Fru suppresses cell death in the male mAL cluster presumably by downregulating expression of the hid, grim and rpr genes. Other genes reported to function downstream from fru include defective proboscis extension response (dpr),¹⁷ yellow (y),¹⁸ takeout (to)¹⁹ and neuropeptide F,²⁰ although it is unknown whether Fru controls them directly or indirectly. Our recent immunostaining of polytene chromosomes with an anti-FruCOM antibody that detects all isoforms of Fru revealed that Fru binds to approximately 130 distinct loci including five X-chromosomal bands: 1C, 2B, 3C, 3D and 3F.¹¹ The immunostaining with anti-HDAC and anti-HP1a antibodies indicated that HDAC1 binds to all five loci, while HP1a binds to only three: 3C, 3D and 3F.¹¹ Chip-seq analyses in the modENCODE project have shown that all five loci contain HDAC1-target regions, and that the 3C, 3D and 3F loci contain HP1a-target regions (www.modencode.org/).

Our recent data for mAL neurons and GRNs described above indicated that Rpd3 masculinizes (its loss-of-function mutant demasculinizes) both types of neuron, whereas Su(var)205 demasculinizes (its loss-of-function mutant masculinizes) them (Fig. 2A At first, it seems puzzling that bon mutant males have the neural demasculinizing phenotype rather than displaying a composite phenotype of masculinization and demasculinization, despite the fact that Bon recruits demasculinizing protein HP1a [= Su(var)205] as well as masculinizing protein HDAC1 [= Rpd3] (Fig. 2A). A possible scenario is that HP1a can exert its effect only after HDAC1 has modified chromatin, and thus the phenotype of bon mutation resembles the effect of eliminating early-acting HDAC1 (Fig. 2B). It remains an open question how the counteracting effects of two antagonistic complexes, Fru-Bon-HDAC1 and Fru-Bon-HP1a, are balanced so as to tune sexually dimorphic characteristics of fru-expressing neurons.

Figure 2. (A) Effects on the sexually dimorphic characteristics of mAL neurons and GRNs of the manipulation of fru, bon, Rpd3 or Su(var)205 expression. (B) Fru-Bon-dependent recruitment of HDAC1 or HP1a, which exerts masculinizing or demasculinizing effects on neurons. Lower panel shows the developmental time window when the male-specific structures of mAL neurons form and the ecdysone titer²¹ elevates during pupal-adult metamorphosis of Drosophila.

Dynamic changes in chromatin structure are known to occur during larval-pupal and pupal-adult metamorphosis, which is orchestrated by sequential transcriptional cascades initiated by an action of the molting hormone ecdysone. Fru proteins are detected in the CNS of males from the late third-instar larval through adult stages.²² In mAL neurons, female-specific cell death commences around 15 h APF. mAL neurons start to form neurite branches at exactly this time (~15 h APF), and adopt their adult structure by 48 h APF in both sexes (Ito H, unpublished data) (Fig. 2B). Note that a high-titer ecdysone pulse at the end of the third-instar larval stage triggers puparium formation. Subsequent ecdysone pulses are observed at ~12 and ~36 h APF, and the ecdysone pulse at ~12 h APF triggers head eversion. Terminal differentiation of pupal development occurs at ~96 h APF, followed by the eclosion of an adult fly (Fig. 2B).²² It is envisaged that distinct ecdysone peaks could signal which of the two antagonistic complexes should act on the Fru-target genes, as the timing of the neuritogenesis and cell death of mAL neurons coincides with the two ecdysone pulses at the early pupal stage. Dalton et al. (2009)¹⁶ reported that about one third of the genes regulated downstream from fru contain ecdysone receptor (EcR) binding sites. A reduction in EcR isoform-A (EcR-A) in males results in increased male-male courtship activity and a reduction in the sizes of two sexually dimorphic antennal lobe glomeruli, two phenotypes reminiscent of those induced by the loss of fru.^16,23 Moreover, genes associated with ecdysone signaling are overrepresented in fru modifiers isolated by genetic screens (reference ²⁴ and Ito H, unpublished data). It is tempting to postulate that EcR-A or other EcR isoforms and their cofactors specify which of the mutually antagonistic Fru complexes is to be recruited to particular target genes for the formation of sex-specific neural circuitry in response to changes in the ecdysone titer. Notably, EcR-B expression peaks earlier than EcR-A does at the early pupal stage.²⁵

Are the Fru-Containing Complexes Activators or Repressors?

Various chromatin factors for post-translational modifications of histones are involved in switching between the euchromatic and heterochromatic states in Drosophila chromosomes. In transcriptionally active genes in euchromatin, lysine 9 of histone H3 (H3K9) is normally acetylated by a histone acetylase. Several sequential histone modifications are required for the transition from an active to a silent form of chromatin: (1) acetylation of H3K9 is removed by histone deacetylases such as HDAC1; (2) H3K9 is methylated by Su(var)3–9 (one of the histone methyltransferases); (3) HP1a binds to the methylated H3K9 and oligomerizes to induce chromatin compaction; (4) HP1a also recruits Su(var)4–20 (another histone methyltransferase); (5) lysine 20 of histone H4 is methylated by Su(var)4–20, which further facilitates chromatin condensation.^26,27 These observations have led to the traditional view that HDAC1 and HP1a coordinately act to silence target genes by the deacetylation of histones H3/H4 and chromatin compaction.^26,28 However, our genetic and molecular data lead us to pose the intriguing hypothesis that, when associated with Fru, HDAC1 and HP1a acquire mutually antagonistic functions, i.e., masculinizing vs. demasculinizing neurons, which might result from their counteracting effects on transcription, i.e., activation and repression (Fig. 2B). The Rpd3 mutations enhance the demasculinizing effects of loss-of-function fru mutations, whereas Su(var)205 mutations oppose these effects, implying that HDAC1 would confer transcription-activating ability on the Fru complex, whereas HP1a would impede it. Growing bodies of evidence in recent years point to the notion that HDAC1 and HP1a both activate transcription in some instances. In mammals, HDAC1 is involved in glucocorticoid-induced activation of the mouse mammary tumor virus promoter.²⁹ In Drosophila, HP1a upregulates the expression of some euchromatic genes through the association with heterogeneous nuclear ribonucleoproteins and RNA transcripts.³⁰ Similarly, it was reported that the association of HP1a with chromosomes leads to the expression of several HP1a target genes in euchromatin.^31,32 These findings are in concert with our idea that HDAC1, and perhaps HP1a also, in the Fru-protein complex not only silence but also activate target genes, depending on the developmental context, contributing to either the masculinization or demasculinization of fru-expressing neurons. We do not know how the activator role is assigned to one of these chromatin factors or how the silencer role is assigned to the other factors in every specific transcription event in a Fru-target gene. A tantalizing possibility is that HDAC1 and HP1a learn their roles from target chromatin landscape differences, which could have been shaped by the preceding binding of distinct transcription factors, e.g., an EcR-containing complex, to a site near the Fru target gene.

To test this possibility, the temporal dynamics of the Fru-Bon-HDAC1 and Fru-Bon-HP1a complexes on a defined target site within an identified neuronal cluster must be analyzed. Now this approach is feasible; FACS allows one to purify GFP-marked cells, which can be restricted to a defined fru-expressing cluster with the intersectional in vivo labeling technique developed by Yu et al. (2010).² The fru-expressing neurons thus purified will be subjected to ChIP-seq and RNA-seq analyses to determine the spatial and temporal binding of Fru protein complexes on a given target gene. Such analysis will help to elucidate the mechanistic link among chromatin dynamics, single neuron ontogeny and behavioral traits, and thus will enable us to refine our view on gene-environment interactions in shaping the psychosocial characteristics of an animal.

Ito H, Sato K, Koganezawa M, Ote M, Matsumoto K, Hama C, et al. Fruitless recruits two antagonistic chromatin factors to establish single-neuron sexual dimorphism. Cell. 2012;149:1327–38. doi: 10.1016/j.cell.2012.04.025.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.