Drosophila dany is essential for transcriptional control and nuclear architecture in spermatocytes Free

Meiotic sex has a deep evolutionary origin in basal eukaryotes. While meiosis has generally remained well conserved, most other aspects of sexual reproduction have diverged rapidly. In animals that differentiate a male and a female gender, male-biased genes and in particular those expressed in the germline are affected by turnover and sequence divergence that is significantly faster than in other gene classes (Parsch and Ellegren, 2013). Analyses in Drosophila melanogaster spermatocytes have provided some of the most striking evidence for rapid evolutionary dynamics even in key elements of transcriptional networks with numerous interactions.

The comparison of different tissues in adult D. melanogaster has clearly revealed that the number of genes with an expression apparently restricted to a single tissue is maximal in the case of testis (Andrews et al., 2000; Chintapalli et al., 2007; Dorus et al., 2006; Vibranovski et al., 2009; Zhao et al., 2010). The large majority of these, as well as of the testis-biased genes, are transcribed in spermatocytes, i.e. in germline cells during a growth phase between the last mitotic division and the onset of the meiotic divisions and spermiogenesis. The transcription of more than 1000 of these genes depends on the function of the testis meiotic arrest complex (tMAC) (Beall et al., 2007; White-Cooper, 2010). tMAC is a testis-specific variant of the MIP/dREAM/SynMuvB complex, a widely conserved somatic transcriptional regulator (Sadasivam and DeCaprio, 2013). tMAC contains subunits encoded by testis-specific paralogs (aly, tomb) that are only present within the genus Drosophila. They were identified based on a characteristic loss-of-function phenotype in which the testes fill up with spermatocytes failing to enter meiotic divisions and postmeiotic differentiation (Jiang et al., 2007; Lin et al., 1996; White-Cooper et al., 2000). Several genes of comparably recent origin and with similar mutant phenotypes (topi, comr, achi, vis) encode proteins interacting with tMAC (Ayyar et al., 2003; Jiang and White-Cooper, 2003; Perezgasga et al., 2004; Wang and Mann, 2003). A second protein complex of paramount importance for spermatocyte-specific transcription is formed by testis-specific TATA-binding protein (TBP)-associated factors (tTAFs) in the genus Drosophila (Chen et al., 2005; Hiller et al., 2004; Li et al., 2009). The tTAF genes [can, mia, nht, rye (Taf12L – FlyBase), sa] were also identified based on their mutant phenotype, which is slightly milder than that associated with tMAC loss.

Here we describe the identification of dany, yet another gene with a recent evolutionary origin and an essential role in spermatocyte-specific gene expression. dany is most similar to the Drosophila genes distal antenna (dan) and distal antenna-related (danr). Dan and Danr have recently been implicated in the control of neuroblast competence in Drosophila embryos, where they inhibit the repositioning of crucial target genes into repressive chromatin associated with the nuclear lamina (Kohwi et al., 2013). Similarly, Dany is required not only for cell type-specific gene expression in spermatocytes but also for normal association of chromatin with the nuclear envelope.

To identify novel genes that might provide meiosis-specific functions, we searched for Drosophila genes expressed predominantly or exclusively in ovaries and testes using various data sources (Brown et al., 2014; Chintapalli et al., 2007; Gan et al., 2010; Graveley et al., 2011; Vibranovski et al., 2009). For further characterization, we targeted candidate genes by transgenic RNA interference (RNAi) in spermatocytes. In the case of the previously uncharacterized gene CG30401, knockdown resulted in male sterility. Therefore, we undertook a more detailed characterization of this gene.

CG30401 encodes a protein with a single N-terminal Psq motif (Fig. 1A,B) (Siegmund and Lehmann, 2002). This motif of ∼50 amino acids contains a helix-turn-helix domain (Fig. 1B) and is present in a family of proteins that includes transposases and the centromere protein CENPB. The Psq motif of CG30401 is most similar to that of the Drosophila gene dan and its tandem duplication danr. Whereas dan/danr orthologs are readily detectable throughout the insect lineage, CG30401 orthologs are restricted to the genus Drosophila. Reflecting its young evolutionary origin, dany will be used as name for CG30401. The similarity between dany and dan/danr is restricted to the Psq motif. dan/danr expression occurs in various somatic tissues during development but not in gonads. By contrast, dany transcripts appear to be restricted to ovaries and testis (Brown et al., 2014; Chintapalli et al., 2007; Graveley et al., 2011).

To evaluate dany function, we generated null mutations by isolating imprecise excision events after mobilizing the transposon P{GSV7}GS22077 (Fig. 1A). Molecular characterization of dany^ex4, dany^ex11 and dany^ex83 revealed deletions removing substantial parts or the entire gene. These alleles were crossed over a deficiency Df(2R)BSC802 that deletes the dany region. dany^ex hemizygous adults eclosed with the expected Mendelian frequency free of obvious morphological abnormalities. Female fertility was at most slightly reduced (Fig. 1C). However, males were sterile (Fig. 1C). A transgene (g-dany) made with a genomic fragment containing only dany (Fig. 1A) completely restored male fertility in dany^ex11/Df(2R)BSC802 (Fig. 1C). We conclude that dany is required for male fertility.

To assess whether dany function is required in the germline, we expressed UAS-V20dany^shmiR17 specifically in early spermatocytes. This also caused male sterility (Fig. 1A,C). The same result was obtained with UAS-W20dany^shmiR41, which targets a different region within dany (Fig. 1A,C). We conclude that dany function is required in spermatocytes.

For further characterization of the defects resulting from loss of dany, testes were dissected. Whereas testes from heterozygous deficient males (+/Df(2R)BSC802) were of wild-type appearance, including prominent sperm tail bundles, those from dany null males (dany^ex4, dany^ex11 or dany^ex83/Df(2R)BSC802) were clearly abnormal (Fig. 2; data not shown). These testes were slightly smaller and did not contain sperm tail bundles. Instead, they were full of spermatocytes, which are normally confined to the distal third of the testis tube. g-dany restored a wild-type appearance to testes in dany null mutant males (Fig. 3A). By contrast, testes after spermatocyte-specific dany knockdown were very similar to dany null mutant testes (Fig. 2; data not shown). We conclude that dany is required for normal spermatogenesis to proceed beyond the spermatocyte stage.

To characterize the pattern of dany expression, we generated transgenic lines that express Dany fused to EGFP under the control of the dany cis-regulatory region (g-EGFP-dany and g-dany-EGFP). To determine whether these transgenes can functionally replace dany, they were crossed into a dany null mutant background (dany^ex11/Df(2R)BSC802). The resulting males were fully fertile (Fig. 1C). We conclude that the EGFP fusion proteins are functional and expressed in a pattern that restores male fertility in dany null mutants.

To analyze the pattern of dany expression during spermatogenesis, testes were dissected from males with either g-EGFP-dany or g-dany-EGFP in a dany null background (Fig. 3A). Nuclear EGFP signals were first observed in young spermatocytes. The signals persisted throughout the spermatocyte stages, but faded rapidly during the postmeiotic stages. To study subcellular localization, testis squash preparations were made using dany null males rescued by g-dany-EGFP (Fig. 3B). In early spermatocytes (stages S1-S3) (Cenci et al., 1994), EGFP signals were mostly diffuse throughout the nucleus. In mid spermatocytes (S4, S5), EGFP signals were enriched on chromosome territories (Cenci et al., 1994). A prominent enrichment in the nucleolus became apparent in late spermatocytes (late S5 and S6), as confirmed by double labeling with anti-Fibrillarin, which marks a nucleolar subcompartment (Fig. 3C). At high magnification it was also evident that the enrichment in chromosome territories occurs primarily at their periphery and less within their central heterochromatic regions as characterized by the brightest DNA staining. EGFP-Dany was localized similarly to Dany-EGFP, except for somewhat lower levels of diffuse nuclear signals (Fig. 3C, Fig. S1).

The subcellular localization of the Dany EGFP fusions in spermatocytes suggested that a fraction of Dany is chromatin associated. To explore this further, UASt transgenes were generated for targeted expression in larval salivary glands, which have large polytene chromosomes. However, initial experiments revealed that ectopic dany expression is highly toxic, not just in salivary glands but also in other somatic tissues. Various GAL4 drivers (Act5C, en, ey, GMR, sev) in combination with UASt-dany, UASt-EGFP-dany or UASt-dany-EGFP caused complete larval or pupal lethality in most cases. In combination with a salivary gland-specific GAL4 driver (F4), larvae developed that did not contain recognizable salivary glands. By contrast, expression of these transgenes in testis with bam-GAL4-VP16 did not have toxic effects. bam-GAL4-VP16-driven UASt-EGFP-dany expression was unable to restore fertility in dany null males, presumably because expression clearly did not perdure to the late spermatocyte stages (data not shown), as is observed with transgenes under control of the dany cis-regulatory region. These results emphasize that precise control of dany expression is crucial. For normal spermatogenesis, dany needs to be expressed in spermatocytes for an appropriate period. Moreover, detrimental ectopic expression in somatic tissues must be prevented.

To minimize confounding toxicity during localization studies in salivary glands, UASt-EGFP-dany was expressed only briefly by means of the Gal4/Gal80^ts system. After 4 h of induction at 29°C, EGFP-Dany was detectable in salivary gland nuclei of third instar larvae (Fig. 3D). EGFP-Dany was not observed when the temperature shift was omitted or when GAL4 and UASt transgenes were not both present (data not shown). Salivary glands were of normal appearance after 4 h of EGFP-Dany expression. Interestingly, EGFP-Dany was enriched on polytene chromosomes, where it was preferentially associated with interbands that have lower DNA staining (Fig. 3D). In conclusion, Dany can associate with chromatin with a preference for euchromatic regions.

In the case of Dan and Danr, there is no direct evidence for DNA binding so far. However, this has been demonstrated for more distant relatives such as human CENPB and various transposases. Structural analyses have revealed several residues within the Psq motif of CENPB that contact the bound DNA (Tanaka et al., 2001). To evaluate the involvement of corresponding residues within the Psq motif of Dany, we generated a mutant transgene, g-dany^AAA-EGFP, that encodes a variant with three alanine residues in place of those in the wild-type protein (Fig. 1B). Complementation tests in a dany null background (dany^ex11/Df(2R)BSC802) revealed that this triple A version is not fully functional. It did not restore male fertility (Fig. 1C). However, the triple A mutation did not completely abolish dany function. When g-dany^AAA-EGFP was present in the dany null mutants, spermatogenesis progressed beyond the spermatocyte stages to a considerable extent. Bundles of spermatids with elongated sperm tails were abundant (Fig. S2A), in contrast to dany null mutants (Fig. 2). Dany^AAA-EGFP expression and localization in spermatocytes appeared to be normal (Fig. S2A,B). However, DNA staining revealed strong nuclear abnormalities in postmeiotic stages (Fig. S2C), and seminal vesicles were always devoid of motile sperm. We conclude that the triple A mutation in the Psq motif of Dany causes a partial loss of function.

The expression program and mutant phenotype observed for dany corresponded to those of ‘meiotic arrest’ genes (White-Cooper, 2010). The majority of these genes encode tTAFs or tMAC components. To evaluate whether a loss of dany causes similar transcriptome changes as tTAF or tMAC mutations, we probed microarrays with RNA isolated from testes of control males and of males lacking either dany, aly, sa or topi function. Hierarchal clustering of the data revealed highest similarities between biological replicates, as expected (Fig. 4A). Moreover, also in accord with expectations, the samples from the two analyzed tMAC mutants (aly and topi) were more similar to each other than to the tTAF mutant sa. The sa samples were slightly more similar to the control than were the dany samples.

The comparison of control and dany mutants (Fig. 4B) revealed a considerable fraction of genes with differential transcript levels (14% after filtering for probes with a change in signal intensity higher than 4-fold and an FDR-adjusted P-value less than 0.05). Compared with the control, ∼1000 genes had lower (‘danyDown’) and 1600 genes had higher (‘danyUp’) transcript levels in the mutant. In controls, the danyDown genes were 12-fold more strongly expressed overall than the danyUp genes (Fig. 4C). Moreover, the danyDown set predominantly contained genes with testis-specific or testis-biased expression, whereas such testis genes were strongly depleted in the danyUp gene set (Fig. S3A). These results suggest that Dany is important not just for the transcriptional activation of spermatocyte-specific genes but also for the repression of genes that function specifically in other tissues.

Our finding that more genes had increased rather than decreased transcript levels in dany mutants indicated that dany functions in a manner distinct from tMAC and tTAFs. These were reported to be required primarily for transcriptional activation of spermatocyte-specific genes (Doggett et al., 2011; Lu and Fuller, 2015; White-Cooper and Caporilli, 2013). Our direct comparison of transcriptome changes in dany mutants with those in sa, aly and topi mutants confirmed this notion (Fig. 4D-F). In general, those transcripts that were increased in dany mutants relative to the control were not also increased in the other mutants (Fig. 4D). However, transcripts that were decreased in dany mutants were generally also decreased in the other mutants (Fig. 4E). The greatest number of genes showing decreased transcript levels and the strongest decreases were observed in aly and topi mutants (Fig. 4F; data not shown), consistent with previous analyses (Doggett et al., 2011; Lu and Fuller, 2015). Most of these aly/topi-dependent genes also had decreased transcript levels in sa mutants, although in general to a more modest extent (Fig. 4E, Fig. S3B,C). In addition, most of these genes also had decreased transcript levels in dany mutants, where the decreases were usually slightly less severe than in the sa mutants (Fig. 4E, Fig. S3B,C).

The transcriptome changes observed in dany mutants were distinct from those caused by loss of tTAF and tMAC function, suggesting that dany might work independently of these transcriptional activators of spermatocyte-specific genes. Supporting this notion, loss of dany did not affect the transcript levels of tTAF and tMAC genes and, vice versa, dany transcript levels were not changed in aly, topi and sa mutants (Fig. 5A). Reassuringly, however, dany transcript levels were abolished in the dany null mutants, as expected (Fig. 5A).

For further analysis of potential functional relations between tTAFs, tMAC and dany, we performed microscopic analyses. In contrast to tMAC proteins, which accumulate on spermatocyte chromatin but not in the nucleolus (Doggett et al., 2011; Jiang et al., 2007), tTAF proteins including Sa accumulate primarily in the nucleolus (Chen et al., 2005). As nucleolar enrichment of Dany EGFP fusions had been observed as well (Fig. 3C), we analyzed the colocalization of Sa and Dany in further detail. Double labeling of g-EGFP-dany spermatocytes with anti-Sa revealed that the two proteins are not strictly colocalized at stage S5 (Fig. 5B). As expected (Chen et al., 2005), Sa was detected primarily in the nucleolus. In addition, weak nuclear signals that presumably represent Y loops were also obtained with anti-Sa. As these non-nucleolar signals were not observed in sa-EGFP testis (see below), they most likely reflect non-specific anti-Sa staining. In comparison to Sa, the presence of EGFP-Dany in chromosome territories outside of the nucleolus was far more prominent. Moreover, subcellular localization was even more distinct at the onset of sa and dany expression in early spermatocytes (Fig. 5C). Whereas Sa-EGFP accumulation in the nucleolus occurred already at the onset of expression, nucleolar enrichment of Dany EGFP fusions was observed only in late spermatocytes.

To address whether sa expression and localization depended on dany function, we analyzed Sa-EGFP in dany null mutants. We did not detect an abnormal Sa-EGFP behavior in dany mutants (Fig. 5C). Loss of dany function also did not affect expression and subcellular localization of the can product, another tTAF (Chen et al., 2005), which we studied with the help of a functional can-myc transgene (Fig. S4A). tTAFs have been proposed to function in part by sequestering Polycomb (Pc), a core component of the PRC1 repressive chromatin complex (Steffen and Ringrose, 2014), in the nucleolus (Chen et al., 2005,, 2011; White-Cooper and Caporilli, 2013), although this notion has been challenged recently (El-Sharnouby et al., 2013). We were unable to detect abnormalities in the expression and localization of Pc-EGFP in dany mutants (Fig. S4B). Nevertheless, the list of genes that were strongly derepressed in dany mutant testis was significantly enriched for genes that had been identified as Pc targets previously in several independent ChIP experiments (Table S1).

Beyond the above finding that dany function is dispensable for the expression and nucleolar localization of tTAFs such as Sa and Can, we analyzed whether, conversely, sa might be required for Dany accumulation or localization. In addition, with analyses in aly and topi mutants, we addressed whether tMAC function is required for Dany accumulation or localization. g-EGFP-dany and g-dany-EGFP were therefore crossed into males with transheterozygous combinations of strong alleles that are thought to abolish the function of sa, aly or topi completely (Hiller et al., 2004; Perezgasga et al., 2004; White-Cooper et al., 2000). Expression of EGFP-tagged Dany was still observed in mutant testis, starting in early spermatocytes as in wild type, and reaching apparently normal levels. Intranuclear localization, however, was affected in mutant testis. The enrichment of the Dany EGFP fusions that eventually occurs during wild-type spermatogenesis within chromosome territories and nucleolus in late spermatocytes was largely abolished in the mutants (Fig. 5D; data not shown). This abnormal localization of Dany in late spermatocytes might reflect a failure of normal spermatocyte differentiation in the absence of tMAC and tTAFs, and hence possibly secondary consequences of the mutations in sa, aly and topi. Importantly, however, dany expression clearly does not depend on tMAC and tTAFs. Vice versa, Dany does not function by allowing expression of tMAC and tTAFs.

In the context of our microscopy analyses of EGFP-tagged Dany, we noticed a surprisingly abrupt transition from EGFP-positive to EGFP-negative spermatocytes within the proximal part of sa, aly and topi mutant testes (data not shown). Those aspects of the spermatocyte differentiation program that bring about the postmeiotic disappearance of Dany during normal spermatogenesis might therefore still occur in these mutants. Alternatively, the eventual disappearance of Dany EGFP fusions in mutant testes might reflect a degeneration process rather than an aspect of normal spermatogenesis. However, in dany mutant spermatocytes, at least some differentiation processes appear to progress beyond the spermatocyte stage. In the proximal region of dany mutant testes, nucleolar Sa-EGFP disappeared abruptly concomitant with enhanced chromatin condensation, nuclear lamina depolymerization and formation of abnormal spindles (data not shown), suggesting the occurrence of meiotic divisions. Anti-phospho-histone H3, a widely used M-phase marker, labeled some spermatocytes close to the terminal epithelium in most dany mutant testes (in 13 of 16 analyzed testes) (Fig. 2D). To analyze these meiotic divisions further, we crossed the centriole marker EYFP-Asl into dany mutant spermatocytes and performed double labeling with anti-tubulin (Fig. 5E). This staining confirmed that dany mutant spermatocytes eventually enter meiotic divisions. Although spindles were readily observed, they lacked normal centrosomes with regular numbers of centrioles at the spindle poles. Moreover, they were often kinked or multipolar. Anaphase figures were never observed, indicating that the chromosomes did not attach normally to spindle fibers. We conclude that loss of dany function does not result in a complete arrest before entry into meiotic divisions, as characteristically reported in the case of mutations in tTAF and tMAC genes (White-Cooper, 2010).

To identify early consequences of dany loss, we analyzed spermatocyte morphology at the stages when dany expression normally starts. As dany is required for the realization of a normal transcriptional program, we examined the distribution of DNA within spermatocyte nuclei as delimited by Lamin (Lam) staining. Control and dany mutant spermatocytes were undistinguishable at stage S2 (not shown), but differences were clearly apparent at stage S3 (Fig. 6A). In controls (w, as well as dany^ex11/Df(2R)BSC802 rescued by one copy of g-dany), the forming chromosome territories seemed as if pulling the nuclear lamina around them, resulting in conspicuous nuclear lamina deformations. By contrast, dany mutants (dany^ex11/Df(2R)BSC802) had a nuclear lamina that was balloon-like, lacking deformations correlated with chromatin distribution. At the same stage, the nuclear lamina in sa, aly and topi mutants was of wild-type appearance (Fig. 6A). During all subsequent stages (S4-S6), the intriguing difference in nuclear lamina appearance between dany mutants and controls was maintained (Fig. 6B; data not shown). At these later stages, nuclear lamina appearance in sa, aly and topi mutants was more similar to that in dany mutants, i.e. distinctly more balloon-like than in the controls (Fig. 6B). In conclusion, the first phenotypic abnormalities that we have been able to detect in dany mutants concern the association between chromatin and the nuclear envelope.

To assess whether loss of dany has any major effects on active and repressive chromatin marks, we compared the level and intranuclear localization of immunofluorescent signals in control and dany mutant spermatocytes after labeling with antibodies against modified histones: acetylated histone H4 (H4ac) and histone H3 with trimethylated K4 (H3K4me3), K9 (H3K9me3) or K27(H3K27me3) (Fig. 6C). We could not detect any striking differences in global signal intensities between the dany mutant and control. However, these analyses clearly confirmed that association of chromatin with the nuclear envelope region is strongly reduced in dany mutant spermatocytes.

Intriguingly, the list of strongly derepressed genes in dany mutant testes contained a significantly higher number of genes strongly derepressed by Lam knockdown in the fat body (Chen et al., 2013) than expected stochastically (Table S1). For an initial evaluation of the effects of dany loss on intranuclear localization at the gene level by fluorescence in situ hybridization (FISH) to chromosomal DNA, inaC was selected as the gene most strongly derepressed in dany mutant testis according to our microarray data and as it is also associated with the nuclear lamina according to Lam-ID in Kc cells (Pickersgill et al., 2006), as well as strongly derepressed after Lam knockdown in fat body (Chen et al., 2013). In addition, we analyzed a testis-specific gene cluster at 22A1, which has been shown to be associated with the nuclear lamina in Kc cells and male germ cells during the early stages of spermatogenesis followed by relocalization into the nuclear interior in late spermatocytes (Pickersgill et al., 2006; Shevelyov et al., 2009). The genes within this cluster were at most mildly downregulated in dany mutant testis according to our microarray data. Immuno-FISH with anti-Lam and the inaC or the 22A1 FISH probe (Fig. 6D) indicated that both loci were preferentially close to the nuclear periphery in early spermatocytes (S1) in both control and dany mutants, as expected. In late control spermatocytes (S6), 22A1 tended to be farther away from the nuclear lamina, similar to previous reports (Shevelyov et al., 2009), whereas inaC was still relatively close to the nuclear periphery. Strikingly, compared with control, both loci were far more internal in dany mutant S6 spermatocytes, confirming that dany is required for normal chromatin association with the nuclear periphery.

Our identification and characterization of dany has uncovered a factor crucial for the realization of the spermatocyte-specific gene expression program in Drosophila. Loss of dany compromises the transcriptional activation of a large number of spermatocyte-specific genes and derepresses genes that are normally inactive in spermatocytes. Moreover, dany is required for normal association of spermatocyte chromatin with the nuclear periphery.

With regard to transcriptional activation, Dany is similar to tTAFs and tMAC. Mutations in tMAC genes have a severe effect on spermatocyte-specific gene expression (Doggett et al., 2011; Lu et al., 2013) (Fig. 4). In the case of tTAF mutants, fewer genes are affected and the reduction in transcript levels is often less severe. Loss of dany has even slightly milder effects, but there are still ∼1000 genes with significantly reduced transcript levels. Many of these genes are also dependent on tMAC and tTAFs.

The fact that Dany is just as important for silencing of non-spermatocyte genes as for activation of spermatocyte-specific genes indicates that this protein does not function like the activators tTAFs and tMAC, which have not been implicated in gene silencing. Several additional observations support our conclusion. Dany intracellular localization in early spermatocytes is distinct from that of the tTAF Sa (Fig. 5C). dany transcript and protein levels depend neither on Sa nor on the tMAC components Aly and Topi (Fig. 5A,D). Vice versa, mutations in dany affect neither transcript levels of tTAF and tMAC genes (Fig. 5A) nor accumulation and intracellular localization of representative protein products (Sa and Can; Fig. 5C, Fig. S4).

dany orthologs cannot be detected outside the genus Drosophila, whereas the primordial dan/danr genes are present throughout the insect lineage. The testis-specific tMAC and tTAF genes are comparably young in evolutionary terms. The fact that several functionally independent factors of paramount importance for the highly complex spermatocyte-specific gene expression program have a recent origin provides further testimony of the surprising evolutionary dynamics of genes that function in the male germline. After a gene duplication event, dany has evolved to control the expression of thousands of genes in the male germline. Dany is a potent regulator; ectopic expression of dany in somatic tissues is highly toxic.

How does Dany exert its function? We demonstrate that Dany is a nuclear protein. After ectopic expression in larval salivary glands, it binds preferentially to all euchromatic interband regions of polytene chromosomes. In maturing spermatocytes, where it is normally expressed, Dany also appears to associate preferentially with euchromatic regions within chromosome territories and not with the regions of maximal DNA staining intensity corresponding to pericentromeric heterochromatin. As the Psq motif of some other proteins, including CENPB and transposases, is involved in sequence-specific DNA binding, it is conceivable that Dany and its closest relatives Dan and Danr bind to DNA as well. The Psq domain structure of Dan as revealed by NMR [Protein Data Bank (PDB) code 2ELH] is highly similar to that observed by X-ray crystallography in CENPB (PDB code 1HLV) (Tanaka et al., 2001). However, the DNA-binding region within CENPB is not restricted to the Psq motif but includes an additional helix-turn-helix domain that is not present in Dan, Danr and Dany. Similarly, the DNA-binding region of transposases extends considerably beyond the Psq motif. By mutating three amino acid codons within the Dany Psq motif, which correspond to positions contacting bound DNA in the case of CENPB, we were able to demonstrate the functional importance of this motif. However, Dany^AAA still retains some function; it promotes spermatogenesis beyond the stage when dany null mutant spermatocytes arrest. Moreover, the intracellular localization of Dany^AAA in spermatocytes appears to be normal. Additional analyses will be required to resolve whether the Psq motif of Dany indeed binds DNA. We point out that the Psq motif in CENPB contacts only four base pairs (Tanaka et al., 2001). Such a limited DNA sequence specificity could not explain the observed preferential association of Dany with euchromatic regions. Additional factors will have to be identified.

Interestingly, the first phenotypic abnormalities that we have detected in dany mutants occur in young spermatocytes soon after the onset of dany expression. Whereas chromosome territories tend to become enwrapped in the nuclear envelope when they are formed in wild-type spermatocytes during the S3 stage (Cenci et al., 1994; Fuller, 1993), dany mutant spermatocytes are devoid of such characteristic nuclear envelope deformations at the corresponding stage (Fig. 6). Thus, Dany appears to be required for normal chromatin association with the nuclear envelope. Since Dany is not enriched at the nuclear periphery, it obviously is very unlikely to function as a factor that directly establishes physical contact between chromatin and the nuclear envelope. But the extensive chromatin reorganization, which presumably occurs in early spermatocytes when thousands of previously repressed genes become active while precursor-specific genes are inactivated, might be abnormal in the absence of dany. Thereby, chromatin properties that normally cause localization to the nuclear periphery might also be affected.

Dany's closest relatives Dan and Danr, which act partially redundantly in somatic tissues (Curtiss et al., 2007; Emerald et al., 2003; Kohwi et al., 2011,, 2013), are crucial for the control of intranuclear position and silencing of the hunchback (hb) genomic locus in neuroblasts during embryogenesis (Kohwi et al., 2013). hb relocalization to the nuclear lamina, which is correlated with hb silencing and termination of Hb response competence, depends on timely Dan downregulation.

Changes in the association of genes with the nuclear periphery have previously been implicated in the control of the spermatocyte-specific gene expression program in Drosophila. In cell types other than spermatocytes, most spermatocyte-specific genes appear to be packaged into a repressive type of heterochromatin designated ‘BLACK’ heterochromatin, which is associated with Lam at the nuclear envelope (El-Sharnouby et al., 2013; Filion et al., 2010; Shevelyov et al., 2009; van Bemmel et al., 2013). In case of two representative testis-specific gene clusters (at 60D1 and 22A1), loss of Lam was shown to result in their derepression in somatic larval and cultured cells, and the normal transcriptional activation was accompanied by cluster repositioning into the nuclear interior of late spermatocytes (Shevelyov et al., 2009). Our immuno-FISH analyses support the notion that a repressive chromatin type that is normally associated with the nuclear periphery is impaired in dany mutants. The derepression of the eye-specific inaC gene in dany mutant spermatocytes was accompanied by delocalization away from the nuclear periphery into the interior. Extensive future work will be required to clarify the molecular basis and functional significance of our initial observations.

Even if an extensive correlation between intranuclear position and transcriptional activity were established by systematic studies, it remains to be considered whether loss of dany might affect the organization of chromatin within the nucleus indirectly. dany mutations could alter the expression program of particular chromatin regulators or proteins of the nuclear periphery. Indeed, some important chromatin regulators and nuclear envelope proteins have been shown to undergo drastic changes in expression levels during normal spermatogenesis. Su(Hw), a multi-zinc finger DNA-binding protein that can function as a transcriptional insulator and modulator of chromatin association with the nuclear lamina (Soshnev et al., 2013; van Bemmel et al., 2010), and the PRC2 complex components E(z) and Su(z) (Chen et al., 2011) are strongly downregulated in early spermatocytes, whereas Lamin C (LamC) expression is induced (Chen et al., 2013). However, in these particular cases, we did not detect an abnormal expression program in dany mutant testis by immunolabeling (data not shown).

We point out that loss of dany also results in the derepression of genes that appear to be Pc targets. Such genes are primarily within ‘BLUE’ heterochromatin (Filion et al., 2010). Pc-mediated repression of spermatocyte-specific genes has been proposed to be counteracted in spermatocytes by the combined action of tMAC and tTAFs. Although not identical, the dany mutant phenotype shares similarities with the meiotic arrest phenotype caused by loss of tMAC and tTAF function. Future work will be required to clarify the functional interactions between Dany and these major activators of spermatocyte-specific genes.

Extensive spatial intranuclear chromatin reorganization might guide many differentiation processes in complex eukaryotes. For example, hundreds of genes were reported to move towards or from the nuclear envelope during differentiation of mouse embryonic stem cells, concomitant with reduced or increased expression, respectively (Peric-Hupkes et al., 2010). Similar observations were made during adipogenic cell differentiation (Lund et al., 2013). Our identification and initial characterization of dany should help support future progress towards a molecular understanding of the mechanisms that coordinate change in spatial genome organization and cell type-specific expression programs.

The transposon insertion P{GSV7}GS22077 (DGRC #203614) (Toba et al., 1999) was used for the isolation of the imprecise excision events dany^ex4, dany^ex11 and dany^ex83 as described in the supplementary Materials and Methods. Df(2R)BSC802 (Cook et al., 2012) [Bloomington Drosophila Stock Center (BDSC) #27374] deletes the dany gene.

Plasmid constructs used for the generation of transgenic lines are described in detail in the supplementary Materials and Methods. Lines for GAL4-regulated transgenic RNAi, UAS-V20dany^shmiR17 and UAS-W20dany^shmiR41 were generated after integration of the constructs pVALIUM20-dany^shmiR17 and pWALIUM20-dany^shmiR41 into the attP2 landing site (Groth et al., 2004).

Lines with transgenes under control of the dany cis-regulatory region (g-dany, g-EGFP-dany, g-dany-EGFP, g-dany^AAA-EGFP) were obtained after integration (BestGene) of pattB constructs (Bischof et al., 2013) into the attP40 landing site (Markstein et al., 2008). For analysis of transgene expression and function in dany mutants they were recombined with dany^ex11 and the recombinant chromosome was crossed over Df(2R)BSC802. We obtained indistinguishable results after recombination of the transgenes with the deficiency, followed by crossing the recombinant chromosome over the dany allele.

Transgenic lines allowing GAL4-regulated expression of dany with or without EGFP extension (UASt-dany, UASt-EGFP-dany, UASt-dany-EGFP) were obtained with pUASt constructs after microinjection into w¹¹¹⁸ (BestGene). Multiple independent insertions were established and analyzed. The driver lines bam-GAL4-VP16 (Chen and McKearin, 2003) and esg-GAL4 (P{PZ}esg[05730] P{GawB}05730B), P{w{+mC}=tubPGAL80TS}20 (Keisman et al., 2001; McGuire et al., 2003) have been described. The experiments involving temporally regulated expression in the salivary gland were performed with P{w⁺, UASt-EGFP-dany}III.1.

The following alleles of meiotic arrest mutants were used in heteroallelic combinations: sa¹ and sa² (Hiller et al., 2004) (kindly provided by M. Fuller, Stanford University, CA, USA), aly² and aly⁵ (White-Cooper et al., 2000) (kindly provided by H. White-Cooper, Cardiff University, UK), topi^Z3-0707 (BDSC #38441) and topi^Z3-2139 (BDSC #38442) (Perezgasga et al., 2004). For the analysis of g-EGFP-dany expression in these meiotic arrest mutants, we used testis from heteroallelic mutants carrying one copy of the transgene.

Strains with the transgenes can-6myc (X) (Chen et al., 2005) and sa-EGFP (X) (Chen et al., 2005) were kindly provided by M. Fuller, Pc-EGFP (III) (Dietzel et al., 1999) by R. Paro (ETHZ, Basel, Switzerland) and pUbi-EYFP-asl (X) (Rebollo et al., 2007) by C. Gonzalez (IRB, Barcelona, Spain). The expression of these transgenes was analyzed in testis isolated from dany^ex11/Df(2R)BSC802 males carrying one copy of a given transgene.

Primers used in plasmid construction and analysis of genomic DNA by PCR are listed in Table S2.

Five to ten males of the genotype to be tested were crossed individually, each with three tester virgins. The tester strain was w. For analysis of female fertility, three virgin females of the genotype to be tested were pooled and crossed with three tester males. Five replicate crosses were set up per genotype for female fertility determination. Two days after setting up the crosses, flies were transferred to a fresh vial and discarded 48 h later. Adult progeny flies that eclosed in the vial were counted. For the control genotype Df(2R)BSC802/+ we obtained on average 217 progeny flies from a single male and 232 from three females.

For whole-mount testis preparations, dissection was performed in testis buffer (183 mM KCl, 47 mM NaCl, 10 mM Tris-HCl, pH 6.8). Testes were fixed in depression slides for 10 min in PBS containing 4% formaldehyde and 0.1% Triton X-100. For DNA staining, testes were incubated for 10 min in PBS, 0.1% Triton X-100 (PBTx) containing 1 μg/ml Hoechst 33258. After three washes with PBS, testes were transferred to a drop of mounting medium (70% glycerol, 1% n-propyl gallate, 0.05% p-phenylenediamine) on a slide before adding a coverslip.

Testis squash preparations were made and stained essentially as described previously (White-Cooper, 2004), according to protocol 3.3.2, except that we used the mounting medium described above. For accurate comparison of signal intensities obtained with the antibodies against modified histones (H4ac, H3K4me3, H3K9me3, H3K27me3) in dany mutant and control spermatocytes, we used pUbi-EYFP-asl males for isolation of control testes and mixed these with dany mutant testes before squashing, fixation and staining on the same slide. After imaging with constant settings, control and dany mutant spermatocytes were identified based on the presence and absence of EYFP-Asl, respectively.

For inactivation of GAL80^ts, we transferred third instar wandering stage larvae grown at 18°C onto apple juice agar plates, which were then floated on a 29°C water bath for 4 h before dissection and staining of salivary glands as described (Radermacher et al., 2014).

Detailed information on antibodies is provided in the supplementary Materials and Methods. Images were acquired using a Zeiss Cell Observer HS wide-field microscope, except for those in Fig. 6A-C for which we used an Olympus FluoView 1000 laser-scanning confocal microscope. Single focal planes were captured with a 10× objective when imaging entire testes. z-stacks were acquired with 40×/0.75, 60×/1.42 or 63×/1.4 objectives for analysis of testis subregions at high resolution. Maximum intensity projections were generated using AxioVision (Zeiss) for wide-field images and Imaris (Bitplane) for confocal images. For the images in Fig. 3D, z-stacks of five focal planes with 250 nm spacing were acquired with a 100×/1.4 objective and deconvolved (Huygens Professional software) before maximum intensity projection (ImageJ).

ImageJ was used for estimating the fraction of intranuclear immunofluorescent signal that was close to the nuclear envelope in the testis squash preparations. Image stacks of well-squashed cysts with sixteen spermatocytes in a monolayer were imaged with the laser-scanning confocal microscope. First, the focal plane within an image stack which contained a maximum number of equatorial sections through spermatocyte nuclei was selected manually. The anti-Lam signal was then used for image segmentation. Signal intensity obtained with the antibody against a particular modified histone (H4ac, H3K4me3, H3K9me3 or H3K37me3) within the regions confined by the rims of anti-Lam signal was integrated (total nuclear anti-histone signal). After shrinkage of the measured region by 10 pixels (length of one pixel=103 nm), signal intensity in the anti-histone channel was again integrated (internal nuclear anti-histone signal). Subtraction of the internal from the total nuclear signal yielded the nuclear envelope-associated anti-histone signal. The nuclear envelope-associated signal was converted into percent of total nuclear signal. For each genotype and anti-histone antibody, this percentage was determined for each of three independent image stacks followed by averaging. A t-test was performed for analysis of statistical differences between the average observed in control and dany mutants, respectively. The finding that the average percentage of nuclear envelope-associated anti-histone signal is consistently larger in control than dany mutant spermatocytes was also observed when the nuclear region was shrunk by a pixel value other than 10 (5, 15 or 20 pixels) and when the analysis was extended to include all focal planes of the image stack.

Immuno-FISH was performed essentially as described (Blattner et al., 2016). Anti-Lam and anti-LamC were used as a mixture. FISH probes were generated as described (Dernburg, 2011). The fosmid FF031161 (Ejsmont et al., 2009) was used for preparation of the 22A1 gene cluster probe, and pooled PCR products amplified with primers AB168-AB177 (Table S2) from w genomic DNA for the inaC probe. ChromaTide Alexa Fluor 568-5-dUTP (Thermo Fisher Scientific) was used for probe labeling. The denaturation step was performed at 95°C for 3 min and hybridization for 18 h at 30°C. Slides were washed twice in 50% formamide, 2×SSCT at 30°C for 1 h each. Thereafter, additional washes were performed at room temperature in 25% formamide, 2×SSCT for 10 min and three times in 2×SSCT for 10 min each. DNA staining and mounting were performed as described above. Image stacks with 250 nm spacing between focal planes were acquired with a 63×/1.4 oil-immersion objective on a Zeiss Cell Observer HS microscope. Imaris ‘section view’ was used for analysis of the shortest distance between the FISH signal and the nuclear lamina.

Total RNA was isolated from testes dissected from young adult males (0-2 days after eclosion). Five genotypes were analyzed: (1) w¹¹¹⁸/Y; Df(2R)BSC802/+, (2) w¹¹¹⁸/Y; Df(2R)BSC802/dany^ex11, (3) w*/Y; aly² red e/aly⁵ red e, (4) +/Y; bw; topi^Z3-0707/topi^Z3-2139 and (5) +/Y;th sa¹ red/ru h sa² ca. After dissection of 20 testes, these were transferred into 5 µl ice-cold testis buffer, snap frozen in liquid nitrogen and stored at −80°C. Eighty testes per genotype were pooled for extraction of total RNA with Trizol (Ambion). RNA was resuspended in 10 µl DEPC-treated water before digestion of residual DNA (DNA-Free Kit, Ambion). Four independent RNA samples were prepared and analyzed from genotype 1, three from genotype 2, and two for genotypes 2-5. RNA samples were used for the generation of Cy3-labeled cRNA (Low RNA Input Linear Amplification Kit, Agilent Technologies) and hybridization to Drosophila gene expression microarrays (4×44 K microarrays, G2519F-021791, Agilent Technologies). Before hybridization, cRNA was purified (Absolutely RNA Nanoprep Kit, Agilent Technologies). A NanoDrop 2000 spectrophotometer (Thermo Scientific) was used for quantification, a 2100 Bioanalyzer (Agilent Technologies) for quality control, and Agilent Feature Extraction software for quantification of signal intensities.

Data were processed and analyzed using R (v3.0.2) and the Limma package (Ritchie et al., 2015; Smyth, 2005). The normexp convolution method was used for background correction. Corrected signal intensities were quantile normalized. Signal intensities observed with technical replicate probes present more than once on the microarray were averaged. Probes with low signals were filtered out from the total of 32,162 probes. 27,005 probes with a signal ≥5.8 (log₂ transformed intensity value) on at least two microarrays were retained for further analyses. Hierarchical clustering was performed for the comparison of all the different hybridizations. A two-group analysis was performed for the comparison of +/Df(2R)BSC802 and dany^ex11/Df(2R)BSC802, and an ANOVA for the comparison of dany and meiotic arrest mutants (sa, aly, topi). Probes were only considered to have differential signal intensities if the associated false discovery rate (FDR)-corrected P-value was less than 0.05. FlyMine (http://www.flymine.org/) (Lyne et al., 2007) was used for the analysis of tissue specificity of genes that were upregulated or downregulated in dany mutants. Microarray data are available at GEO (accession number GSE76298).

We thank Konrad Basler, Minx Fuller, Pamela Geyer, Cayetano Gonzalez, Rick Jones, Jürg Müller, Renato Paro and Helen White-Cooper for providing fly strains and/or antibodies; Norbert Perrimon for vector plasmids; Rita Lecca and the Functional Genomics Center Zurich for supporting the microarray hybridizations; Charity Law, Helen Lindsay and Mark Robinson for assisting during the microarray data analysis; and Emmanuel Caussinus for assisting during immunofluorescent signal quantification.