Tombola, a tesmin/TSO1-family protein, regulates transcriptional activation in the Drosophila male germline and physically interacts with Always early

Differential control of gene expression during development is crucial for specification and maintenance of differentiated cell types. One of the most dramatic gene expression switches occurs in primary spermatocytes during spermatogenesis; many genes are active only in these cells. Some are germline-specific homologues of ubiquitously expressed genes [e.g.β -Tub85D (Kemphues et al.,1979) and testis-specific proteasome components(Ma et al., 2002)], others are spermatogenesis-specific proteins [e.g. the protamines that replace histones to tightly package sperm DNA (Jayaramaiah Raja and Renkawitz-Pohl, 2005)].

In both mammals and insects, germline stem cells divide to produce spermatogonia. After further mitotic amplification divisions (four in Drosophila melanogaster), spermatogonia become primary spermatocytes,committed to differentiation (reviewed by Fuller, 1993). This developmental transition results in transcriptional activation in primary spermatocytes of a large suite of genes required for meiosis and spermiogenesis. In Drosophila, transcription stops before the meiotic divisions, so transcripts for late-acting proteins are made pre-meiotically(Olivieri and Olivieri, 1965). Meiotic-arrest mutant testes accumulate primary spermatocytes, but lack later stages of spermatogenesis because mutant primary spermatocytes fail to initiate transcription of many genes whose products are required after meiosis. Meiotic-arrest mutants also fail to express some meiotic gene products; always early (aly)-class gene products differ from cannonball (can)-class in their regulation of certain cell cycle genes (Lin et al., 1996; White-Cooper et al., 1998). Through their function in controlling production of cell cycle and differentiation gene products, the meiotic-arrest genes coordinate the independent processes of meiosis and spermatid morphogenesis.

The aly class of meiotic-arrest genes have a broader target range than the can class. Four aly-class and five can-class meiotic-arrest loci have been described(Ayyar et al., 2003; Hiller et al., 2004; Hiller et al., 2001; Jiang and White-Cooper, 2003; Perezgazga et al., 2004; Wang and Mann, 2003; White-Cooper et al., 2000; White-Cooper et al., 1998). aly encodes one of two Drosophila homologues of the C. elegans synMuvB gene lin-9, the other homologue being mip130 (Beitel et al.,2000; White-Cooper et al.,2000). cookie monster (comr) encodes a novel protein of unknown function (Jiang and White-Cooper, 2003). achintya/vismay (achi/vis)and matotopetli (topi) encode sequence-specific DNA-binding proteins (Ayyar et al., 2003; Perezgazga et al., 2004; Wang and Mann, 2003). Aly,Comr and Achi/Vis proteins co-immunoprecipitate from testis extracts; Topi was identified in a yeast two-hybrid screen for Comr interactors(Perezgazga et al., 2004; Wang and Mann, 2003). Despite the interactions between aly-class gene products, the alyand comr mutant phenotypes are subtly different from those of topi and achi/vis. Aly and Comr nuclear localisations are mutually dependent, whereas these proteins require topi and achi/vis for their concentration on chromatin. aly or comr (but not topi or achi/vis) mutants display defects in chromatin organisation. Finally, a small subset of genes are much more dependent on topi and/or achi/vis than on alyor comr for their transcription(Jiang and White-Cooper, 2003; Perezgazga et al., 2004).

To find further transcriptional regulators in primary spermatocytes, we screened for Aly-binding proteins. We have identified and characterised a new Drosophila meiotic-arrest gene, tombola (tomb),which is expressed specifically in testis. tomb encodes the second Drosophila member of the tesmin/TSO1 CXC-domain protein family, the other being Mip120, a subunit of the same complex as Mip130. We show that Tomb complexes with Aly and Comr. We identify a tomb mutant and show that tomb mutant testes have an aly-class meiotic-arrest phenotype more like that of aly and comr than of topi and achi/vis mutants. Aly and Comr proteins fail to associate with chromatin in the absence of tomb function. Topi protein also localises to chromatin in wild-type and achi/vis primary spermatocytes, but not in aly, comr or tomb mutant cells. Concentration of ectopically expressed EGFP-Tomb on chromatin in the nucleus is normal in achi/vis or can-class mutants, but is altered in aly and comr mutants.

Flies were raised on standard corn meal (or wheat meal) sucrose agar at 25°C. Visible markers and balancer chromosomes are described in Flybase(FlyBase Consortium, 1999). P[GS]12862/CyO was from the Kyoto Drosophila Stock Centre. Mutant alleles used were aly⁵, comr^Z1340, achi^Z3922vis^Z3922, topi^Z3-2139,nht^Z2-5946. w¹¹¹⁸ or red e were used as wild-type controls.

Df(2L)cl-h3/SM6b and Df(2L)cl-h1/CyO,amos^Roi-1 (Bloomington Drosophila Stock Centre) were crossed to P[GS]12862/CyO. Excision of the P[GS]12862insertion was carried out by crossing w;P[GS]12862/CyO,Δ2-3 males to w; Tft/CyO females,recovering individual white-eyed progeny and back-crossing to w;Tft/CyO to establish stocks. Excision lines were analysed by testis squashing and PCR and sequencing of the ORF. Females were crossed to their balanced brothers to test for fertility; male fertility was tested by crossing to virgin w¹¹¹⁸ females. The testis phenotype was scored by phase contrast microscopy after dissection and squashing.

An Aly(C-terminus)-Gal4-DNA-Binding Domain [Aly(C)-DB] fusion construct was made by subcloning the ORF (equivalent to amino acids 275 to 534) from a full-length aly cDNA clone into pGBKT7. We generated and screened a testis cDNA-Gal4-Activation Domain (AD) fusion protein library using the Matchmaker Library Construction and Screening Kit (Clontech) as previously described (Perezgazga et al.,2004). Colonies were picked from SD -Ade -His -Leu -Trp selection plates after 7 days. One million independent co-transformants yielded 90 colonies that grew under selective conditions and were blue in the presence of X-α-Gal. To test for interaction between Tomb and Comr, AH109 yeast cells were co-transformed with pGADT7-Rec-Tomb and pGBKT7-Comr or pGBKT7-CG15031 (CG15031 was another clone isolated in the yeast two-hybrid screen) as a negative control. Transformed cells were plated on SD -Ade -His-Leu -Trp selection plates containing X-α-Gal.

PCR products for Tomb deletion derivatives (amino acid residues: 1-73;1-136; 73-243; 136-243) and full-length Tomb were subcloned into pACT2. Co-transformation of AH109 yeast cells was with pGBKT7-Aly(C) or pGBKT7-Kr(Zn-finger) as the negative control(Perezgazga et al., 2004). Transformed cells were plated on SD -Ade -His -Leu -Trp/X-α-Gal plates.

For semi-quantitative RT-PCR, total RNA was extracted from dissected testes with Trizol (Invitrogen) and resuspended in RNAse-free water (three testes-worth per μl). First-strand cDNA was generated from 4 μl of this sample using oligo-dT primers with the SuperScript II Reverse Transcriptase System (Invitrogen). cDNA derived from 0.18 testes (0.3 μl of RT reaction)was used for each RT-PCR reaction and amplified with Taq DNA polymerase(Qiagen) with 24 amplification cycles. Genomic DNA from wild-type flies was used as a positive PCR control, and a no-reverse-transcriptase (no-RT)reaction on wild-type RNA served as a negative control. For RT-PCR from various developmental stages, total RNA was extracted with Trizol, cDNA prepared as above, and PCR amplification carried out for 30 cycles. For re-amplification, 0.5 μl of the first PCR product was used as the template for a further 30-cycle PCR reaction.

A 3′ RACE Kit was used following the manufacturer's (Invitrogen)instructions. The RACE products were either directly sequenced, or were subcloned into pGEM-T-Easy for sequencing. For the 5′ end, RT-PCR was performed using a series of primers upstream of the ATG, paired with a 3′ primer within the coding sequence.

The full-length tomb ORF was subcloned into the mammalian tissue culture expression vector HA-tagged pCDEF3. The full-length aly ORF and Kruppel (Kr) zinc-finger region(Perezgazga et al., 2004) were similarly subcloned into FLAG-tagged pCDEF3. 293T human kidney cells were co-transfected with plasmids for expression of HA-Tomb and FLAG-Aly, or HA-Tomb and FLAG-Kr(Zn-finger), respectively, with lipofectamine 2000 reagent(Invitrogen). Co-immunoprecipitation was as previously described(Perezgazga et al., 2004).

Testes dissected from EGFP-Tomb-expressing flies were homogenised in lysis buffer (50 mM Tris-HCl pH 7.5-8.0, 0.5% Triton X-100, 150 mM NaCl, protease inhibitors) (146 testes, 500 μl buffer used), incubated with ethidium bromide (400 μg/ml) for 30 minutes at 4°C, then cleared by centrifugation. 20 μl was retained as the `input' sample, the remainder was pre-cleared with protein G-sepharose, then incubated with mouse anti-GFP(Roche) and precipitated with protein G-sepharose. Beads were washed, then bound proteins were eluted by boiling in SDS sample buffer. Wild-type testes were processed in parallel as a negative control.

The tomb ORF was subcloned in frame into pUAST-EGFP(Parker et al., 2001). Numerous independent P-element-mediated insertions were recovered using standard transformation protocols after injection of w¹¹¹⁸embryos. EGFP-Tomb fusion protein expression was driven using Bam-Gal4-VP16,which expresses just before the onset of meiotic-arrest gene expression and functions in all the mutant backgrounds(Chen and McKearin, 2003). Bam-GAL4-VP16 (on chromosome 3) was recombined with a third-chromosome UAS-EGFP-Tomb insertion, and the chromosome was used homozygous to express tagged protein in testes homozygous for second chromosome male steriles(tomb, achi/vis, comr, nht). Bam-GAL4-VP16 was recombined with aly⁵ to allow expression from a homozygous second chromosome-linked UAS-EGFP-tomb insertion in this mutant background.

Anti-peptide antibodies were raised by Moravian-Biotechnology. The synthesised oligopeptide KNNPTKPIFSDTYL from the Topi C-terminus was coupled to BSA and used to immunise two rats. The staining patterns for these sera were indistinguishable.

Live testes were dissected, squashed in 2 μg/ml Hoechst 33342 in testis buffer (183 mM KCl, 47 mM NaCl, 10 mM Tris pH 6.8) and examined by phase contrast and fluorescence microscopy. Images were captured with a Q-imaging Retiga 1300 monochrome CCD camera linked to an Olympus BX50 microscope using Openlab software (Improvision) or on a JVC KY-F75U three-colour CCD camera with KY-Link software, and imported into Photoshop (Adobe). Aly, Comr and Topi proteins were visualised by indirect immunofluorescent staining using rabbit anti-Aly (1:2000), rabbit anti-Comr (1:1000) or rat anti-Topi (1:1000)antibodies detected with FITC-conjugated secondary antibodies (Jackson), as described (Jiang and White-Cooper,2003; White-Cooper et al.,2000). DNA was co-stained with propidium iodide. Cells were imaged using a Bio-Rad Radiance Plus confocal microscope mounted on a Nikon E800.

Dig-labelled antisense probes for Cyclin B, Mst87F and polo were generated as previously described(White-Cooper et al., 1998). To synthesise RNA probes for CG3330, CG3927 and CG12907, we generated 400-600 bp RT-PCR products using total testis RNA as template. For tomb, the PCR amplified the entire ORF. The 3′ PCR primers included a T3 RNA polymerase promoter site for in vitro transcription of dig-labelled antisense RNA probes. In situ hybridisation was carried out as described (White-Cooper et al.,1998). Primer sequences are available on request.

We conducted a yeast two-hybrid screen to identify proteins that act with Aly to control transcriptional activation in Drosophila primary spermatocytes. Using the C-terminal half of Aly as bait we recovered seven independent clones of CG14016; we named this gene tombola(tomb) based on the testis phenotype - mutant testes resemble a tube full of balls, as in the lottery game. Co-immunoprecipitation of transiently expressed tagged proteins from tissue culture cells confirmed the interaction between Aly and Tomb proteins. 293T cells were co-transfected to express HA-tagged Tomb (HA-Tomb) and FLAG-tagged Aly (FLAG-Aly). Immunoprecipitation with anti-FLAG antibodies, followed by western blotting with anti-HA antibodies showed that Tomb co-immunoprecipitated with Aly(Fig. 1A). This was confirmed with the reciprocal experiment - immunoprecipitation with anti-HA followed by blotting with anti-FLAG. We detected no co-immunoprecipitation in cells co-expressing HA-Tomb and FLAG-Kr [FLAG fused to the five-zinc-finger motif region of Kruppel (Perezgazga et al.,2004)]. To test the in vivo interaction between Aly and Tomb, and to test whether DNA was implicated in the interaction, we made extracts from EGFP-Tomb-expressing testes (see below), incubated the extracts with ethidium bromide, and immunoprecipitated using anti-GFP. We found that Aly co-immunoprecipitated with EGFP-Tomb, showing that the interaction occurs in testes, and is DNA independent (Fig. 1B). To define the Aly-interaction region of Tomb, we generated deletion constructs and tested their ability to interact in the two-hybrid system. The Aly-interaction domain is found in the C-terminal half of Tomb(amino acids 136-243).

Since aly and comr have identical mutant phenotypes, we suspect that these gene products probably act together in a complex, but we have not detected direct interaction between these proteins. We tested the ability of Tomb to bind Comr by two-hybrid analysis. Yeast co-transformed with Tomb-AD and Comr-DB grew under selective conditions, demonstrating that Comr can interact with Tomb. The Tomb-Aly and Tomb-Comr interactions were specific,as yeast co-transformed with Tomb-AD and CG15031-DB were unable to grow under selective conditions. Co-expression and co-immunoprecipitation experiments in tissue culture cells confirmed that FLAG-Tomb can interact with HA-Comr (data not shown).

The tomb genomic region is complex(Fig. 2A). The tombORF is embedded within, but in the opposite orientation to, the 3′ UTR of CG31989, which is predicted to encode a conserved protein (Cap-D3)of unknown function. As no tomb cDNA clones have been sequenced, we mapped the 5′ and 3′ ends by RACE and RT-PCR. The 5′ end of tomb overlaps the 3′ end of the adjacent gene CG14015. Translation of the tomb ORF gave a 243 amino acid, 26 kD conceptual protein with a theoretical pI of 9.4. The predicted Tomb protein contains a nuclear localisation signal and a CXC motif of the tesmin/TSO1 family(Fig. 2B). Tesmin has been described in vertebrates (it also known as Mtl5/MTL5 in mouse and human),whereas TSO1 is from Arabidopsis, indicating that this domain is conserved between animals and plants. The only other Drosophilatesmin/TSO1 CXC-domain protein, Mip120, has been found in a complex with the second Drosophila lin-9 (aly) homologue, Mip130(Beall et al., 2002; Korenjak et al., 2004; Lewis et al., 2004). A second tesmin/TSO1 CXC-domain protein, which we refer to as tesmin-like (tesl), was found in humans and mouse. C. elegans has a single member of this family, LIN-54 (JC8.6), sea urchin and Ciona each have one homologue. Including TSO1, the A. thaliana genome has 11 tesmin/TSO1 family members.

Comparison of the tesmin/TSO1-domain proteins revealed that tombwas unusual in only having a single CXC domain(Fig. 2C). All other family members have either two domains (vertebrate tesl, worm LIN-54 and plant TSO1),or one and a half CXC domains (vertebrate tesmin and plant SOL2). These domains were separated by a conserved, 42 amino acid spacer in animals (50 amino acids in plants). The first and second CXC domains contain several residues in common; however, they are distinguished by characteristic amino acids conserved within repeat 1 or 2, but not between repeats(Fig. 2C). The one and a half CXC-domain proteins lack the N-terminus of the first domain, whereas tomb has only the second CXC domain. E(z) CXC-like domains fall into a separate family.

We also identified a 52 amino acid additional region of homology between the animal proteins near the C-terminus (31% identity, 50% similarity between mip120 and human tesmin (hs-tes); 33% identity and 42%similarity between tomb and hs-tes). Although the primary sequence conservation is low these sequences are strongly predicted to form a helix-coil-helix secondary structure (PSIpred)(McGuffin et al., 2000)(Fig. 2D). The Aly-interaction domain of Tomb includes this conserved motif but not the CXC domain.

We investigated the tomb developmental expression profile by RT-PCR. tomb is entirely included within the CG319893′ UTR; however, they are encoded on opposite strands. tombcontains a 62 bp intron, whereas the CG31989 3′ UTR lacks introns, allowing us to distinguish the transcripts. tomb transcript was detected only in testis (Fig. 3A). Unspliced products, derived from CG31989transcripts, were not produced from the testis sample, but were found after re-amplification in gonadectomised adults (both males and females) and embryos(0-16 hours) (data not shown).

We determined the testes tomb expression pattern by RNA in situ hybridisation, and found that tomb is highly expressed in early primary spermatocytes (arrow in Fig. 3C), with transcript abundance declining as primary spermatocytes mature (arrowhead in Fig. 3C). The tomb expression pattern in testes was essentially identical to the other aly-class meiotic-arrest loci. We found, using RT-PCR, that tomb was expressed in other meiotic-arrest mutants (aly, comr,achi/vis, topi, mia, sa, nht), indicating that tomb expression does not depend on the activity of any known meiotic-arrest gene (alyand mia shown, Fig. 3B). Some elevation of the tomb transcript level was seen in mutant testes by RT-PCR. This apparent increase in transcript abundance was not due to uniformly increased expression; rather, the transcript appeared specifically more abundant in mutant than wild-type mature primary spermatocytes (Fig. 3D).

To further characterise tomb function, we searched P-element mutagenesis databases (BDGP, Baylor, Cambridge and Kyoto) and found a potential tomb mutant allele in the Kyoto P-collection(Toba et al., 1999). Inverse PCR and sequencing of the flanking DNA of P[GS]12862 confirmed that the element was inserted in codon 174 of tomb. The P[GS]12862 line was homozygous viable, but male sterile. Homozygous females were initially semi-sterile; however, this phenotype was later lost from the stock.

The mutant phenotype of P[GS]12862 could be due to disruption of the function of tomb or CG31989 or both, or could be unrelated to the P-insertion. We tested the contribution of CG31989to the phenotype using P[EY]00456, a P-element insertion in the CG31989 ORF. P[EY]00456 mutant flies were homozygous viable and male and female fertile, as were P[GS]12862/P[EY]00456trans-heterozygotes, indicating that the phenotype of P[GS]12862 was not due to CG31989 loss-of-function. The male fertility defect of P[GS]12862 was uncovered by both Df(2L)cl-h3 and Df(2L)cl-h1, which delete 25D2-3;26B2-5 and 25D4;25F1-2, respectively (tomb is at 25E5). P[GS]12862/Df females were fully fertile, confirming that the male and female fertility defects of P[GS]12862 were separable. Transposase-mediated excision of P[GS]12862 resulted in full reversion of the mutant phenotype, indicating that the male sterility is caused by the insertion into tomb.

Phase contrast examination of squash preparations of tomb^GS12862 homozygous or tomb^GS12862/Df testes revealed that tomb is a meiotic-arrest gene. tomb testes contained morphologically normal stages of spermatogenesis, up to and including mature primary spermatocytes,but no meiotic division or post-meiotic stages(Fig. 4A,B).

We expressed an EGFP-Tomb fusion protein in primary spermatocytes using a Bam-GAL4-VP16 driver, and found that tagged Tomb protein was able to rescue the meiotic-arrest phenotype of tomb^GS12862 homozygous males. This confirmed that the expressed protein is functional, and provided final confirmation that the meiotic-arrest phenotype is due to loss of tomb function (Fig. 4C,D). When expressed in a wild-type background, EGFP-tagged Tomb protein was initially both nuclear and cytoplasmic (at lower levels) in early primary spermatocytes. In more mature primary spermatocytes, EGFP-Tomb was restricted to the nucleus and concentrated on chromatin(Fig. 4E-H): three brightly labelled major chromosome bivalents apposed to the nuclear membrane were visible in every nucleus.

aly-class meiotic-arrest mutant primary spermatocytes fail to express Cyclin B mRNA, whereas can-class mutants express normal levels of Cyclin B mRNA(White-Cooper et al., 1998). tomb mutant testes did not accumulate significant levels of Cyclin B mRNA (Fig. 5I-L), so tomb is aly-class. RNA in situ hybridisation confirmed that tomb, like all known meiotic-arrest genes, is also required for expression of spermatid differentiation genes,including Mst87F (Fig. 5E-H). tomb mutant testes again resembled other meiotic-arrest loci in that transcription was not completely blocked; for example, they accumulated polo transcripts normally(Fig. 5A-D).

aly-class mutants fall into two subgroups based on primary spermatocyte DNA morphology (Ayyar et al.,2003; Jiang and White-Cooper,2003; Lin et al.,1996; Perezgazga et al.,2004). Hoechst 33342 labelling revealed that the tomb DNA chromosomes were somewhat condensed and fuzzy, like aly or comr mutants, rather than more condensed and away from the nuclear envelope as seen in achi/vis or topi mutants(Fig. 6A-D′). achi/vis and topi also differ slightly from aly and comr in their target gene specificities(Perezgazga et al., 2004). Although all genes that depend on aly or comr for expression also depend on achi/vis and/or topi, there are a few genes,including CG3927 and CG12907, whose transcription depends on achi/vis and topi but not on aly or comr. Several other genes, including CG3330, depend on all the aly-class meiotic-arrest genes to some extent for their expression,but differ between aly or comr and achi/vis or topi in that their expression is undetectable in testes from the latter two mutants, but is detected at very low levels in aly or comr testes. The tomb phenotype was indistinguishable from that of aly or comr with respect to expression of CG3927, CG12907 and CG3330(Fig. 6E-P). Phenotypic comparison data are summarised in Table 1.

The tomb phenotype is also more like aly and comr than like achi/vis with respect to Topi localisation. Immunofluorescence revealed that Topi, like the other aly-class meiotic-arrest proteins, localises to primary spermatocyte nuclei, and concentrates on chromatin (Fig. 7A-C). topi mutant testes showed no staining, confirming the antibody specificity (data not shown). Topi protein was nuclear, but less concentrated on chromatin in aly and comr mutant spermatocytes (Fig. 7D-F, comr data not shown), indicating that aly and comrfunctions are not required for Topi's nuclear localisation or DNA binding per se, but are required for efficient accumulation of Topi on chromatin. In achi/vis mutants, Topi localisation was similar to wild type, being nuclear and more concentrated on chromatin. Thus, Topi nuclear localisation and chromatin accumulation are independent of achi/vis function(Fig. 7G-I). Topi protein localisation in tomb and aly mutant spermatocytes were indistinguishable (Fig. 7J-L). Therefore tomb, like aly and comr, is required for accumulation of Topi protein on chromatin.

Aly and Comr proteins both localise to chromatin in wild-type primary spermatocytes; however, if either gene is mutant, the other protein remains cytoplasmic (Jiang and White-Cooper,2003). By contrast, mutation of achi/vis or topidoes not prevent nuclear translocation of Aly and Comr, although these proteins fail to concentrate on chromatin and show a uniform nuclear localisation in achi/vis or topi mutant spermatocytes(Ayyar et al., 2003; Perezgazga et al., 2004). Immunofluorescence revealed that in tomb mutant spermatocytes, Aly and Comr proteins were localised to the nucleus, but were excluded from chromatin (Fig. 7M-R). Therefore, tomb function is not required for nuclear import of Aly and Comr, but is required to load these proteins onto chromatin.

When expressed in achi/vis(Fig. 8A,C), or nht (a can-class meiotic-arrest gene, data not shown) mutant testes,EGFP-tagged Tomb protein also localised to primary spermatocyte nuclei. The protein was concentrated on chromatin, but was also found throughout the nucleoplasm. Therefore, the functions of achi/vis and the can-class genes are not required to establish or maintain the correct subcellular localisation of Tomb, although they might be required to enhance the association of Tomb with chromatin.

By contrast, EGFP-Tomb protein localisation was altered when expressed in comr (Fig. 8B,D) or aly (data not shown) mutant testes. The fusion protein was able to localise to nuclei and chromatin of early primary spermatocytes. However, as spermatocytes matured, the nuclear staining was lost, so that in late primary spermatocytes only very weak, cytoplasmic EGFP fluorescence could be detected. We conclude that aly and comr functions are not required for the localisation of Tomb to the nucleus or chromatin per se, but are required to maintain the nuclear concentration of Tomb by preventing either nuclear export or Tomb degradation.

The meiotic-arrest genes of Drosophila regulate a developmental transition and associated gene expression switch, during which many hundreds of genes whose products are required during sperm formation are upregulated(Andrews et al., 2000; Parisi et al., 2004). Most meiotic-arrest genes described to date have been identified through classical genetics. To find additional gene products that act with those already isolated, we undertook a reverse genetics approach; we identified tomb while screening for proteins that could bind Aly in a yeast two-hybrid system.

The tomb predicted protein contains a tesmin/TSO1-family CXC domain that probably mediates DNA binding. Other tesmin/TSO1-family members have either two full CXC domains, or one truncated domain and one full domain,separated by a conserved spacer. Tomb is exceptional in having a single CXC domain and no spacer sequence. In addition to the CXC domain, we identified a second region of homology shared between tomb and the other animal tesmin/TSO1 CXC-domain-containing proteins. This C-terminal domain has conserved secondary structure, and might be responsible for the Tomb-Aly interaction.

Direct interactions have been demonstrated between Comr and Topi, whereas Aly, Comr and Achi/Vis have been found in a complex in vivo(Perezgazga et al., 2004; Wang and Mann, 2003). Here, we additionally show that Aly and Comr can interact with Tomb. In support of our interaction data, Beall et al. (Beall et al., 2007) have purified a complex of proteins containing Aly,Topi, Comr, Tomb and other factors from Drosophila testes extracts and these components were not detected in ovary-specific extracts. The known aly-class meiotic-arrest gene products localise primarily on chromatin in wild-type primary spermatocytes, although Aly and Tomb are also detected at significant levels in early primary spermatocyte cytoplasm. Only when all five aly-class gene products are present is full chromatin-binding activity achieved. There are subtle differences in aly and comr phenotypes as compared with achi/visand topi. Most notably, achi/vis and topi have broader ranges of target genes than aly and comr(Perezgazga et al., 2004). We have previously shown that the nuclear localisations of Aly and Comr are mutually dependent, i.e. Aly remains cytoplasmic in comr mutants and vice versa (Jiang and White-Cooper,2003). We have also shown that topi and achi/visact later in the localisation pathway, both gene products being required for the efficient loading of Aly and Comr onto chromatin(Ayyar et al., 2003; Perezgazga et al., 2004). We can now place tomb into the pathway of complex assembly and activity(Fig. 9). We propose that Tomb,Achi/Vis and Topi enter the nucleus independently, whereas Aly and Comr can only become (or remain) nuclear as a complex. Topi and Achi/Vis probably have inherent sequence-specific DNA-binding activity, which allows them to localise independently, albeit inefficiently, to their targets. Like Mip120, Tomb might also have DNA-binding activity. When in the nucleus, Aly and Comr interact with Tomb; this complex then promotes Topi and Achi/Vis interactions with target promoters. Tomb protein is destabilised in the absence of Aly and Comr;hence, the phenotypes of tomb, aly and comr mutants are identical with respect to target gene expression levels. DRM, a complex containing the proteins encoded by the C. elegans aly and tomb homologues (lin-9 and lin-54), has recently been described (Harrison et al.,2006). Formation of the DRM complex was sensitive to loss of lin-9 or lin-54, just as aly and tomb are crucial for formation of the aly-class gene product complex in testis. Mammalian tesmin is cytoplasmic in early pachytene cells,and normally translocates to the nucleus during late pachytene and diplotene stages of male meiosis, in a similar manner to fly aly and tomb (Matsuura et al.,2002; Sutou et al.,2003).

modulo (mod), which encodes Drosophilanucleolin, has recently been implicated in transcriptional activation of spermiogenesis genes (Mikhaylova et al.,2006). mod-null mutants are lethal, but a viable weak allele is male sterile. Mod was shown to bind sequence elements in certain testis-specific promoters. Many, but not all, mod target genes are also meiotic-arrest gene targets. An alternative form of Mod, expressed only in testis, has an acidic N-terminal domain that probably allows Mod to act as a transcriptional activator. The can-class meiotic-arrest genes,which encode testis-specific homologues of the basal transcription factor complex TF_IID (testis TAFs), might activate transcription by sequestering the polycomb repressor complex away from active chromatin, i.e. they might activate genes by repressing a repressor(Chen et al., 2005; Hiller et al., 2004; Hiller et al., 2001). In normal primary spermatocytes, Pc and testis TAFs are primarily nucleolar,although the proteins are also detected uniformly on chromatin. ChIP analysis revealed that Sa protein binds promoters of target genes in primary spermatocytes, suggesting a direct transcriptional activator role for testis TAFs (Chen et al., 2005).

The aly-class meiotic-arrest mutant phenotype is most easily explained in terms of transcriptional activation rather than through the repression of a repressor. The aly-class gene products accumulate on chromatin in primary spermatocytes in transcriptionally active regions, and not in the nucleolus. Their function depends on the chromatin localisation. In addition, lack of testis TAF gene activity results in low (but readily detectable) levels of target gene expression, whereas expression of many target genes in aly-class mutant testes is undetectable.

tomb and mip120 (CG6061) are the only Drosophila tesmin/TSO1 CXC-motif proteins. Likewise, aly and mip130 (twit, CG3480, EG86E4.4) are the only Drosophila homologues of lin-9(White-Cooper et al., 1998). Mip120 and Mip130 have been described as components of the dREAM/Myb-MuvB complex found in embryos and tissue culture cells(Beall et al., 2002; Korenjak et al., 2004; Lewis et al., 2004). The dREAM complex contains, in addition to Mip120 and Mip130, Myb, Caf1p55, Dp, Mip40,E2F2 and Rbf or Rbf2 (Korenjak et al.,2004). The MybMuvB complex was purified independently and contains all the subunits of the dREAM complex as well as several additional proteins including Rpd3, Lin-52 and l(3)MBT (Lewis et al., 2004). dREAM/Myb-MuvB regulates DNA replication at chorion gene amplification origins in ovarian follicle cells(Beall et al., 2004; Beall et al., 2002; Cayirlioglu et al., 2001; Frolov et al., 2001). In addition to this role in controlling developmentally regulated DNA replication, the dREAM/Myb-MuvB complex acts as a transcriptional repressor,primarily of genes involved in differentiation(Korenjak et al., 2004; Lewis et al., 2004). This transcriptional repressor role is also developmentally regulated as there are different transcriptional targets for Rbf2 and E2F2 in ovaries, early embryos and S2 tissue culture cells (Stevaux et al., 2005).

DRM, a complex containing the C. elegans homologues of the dREAM subunits has recently been described(Harrison et al., 2006). The genes encoding DRM components act together in the SynMuvB genetic pathway that regulates vulval development redundantly with the SynMuvA and SynMuvC pathways[see the following studies (Ceol and Horvitz, 2004; Ceol et al.,2006; Poulin et al.,2005) and references therein; see Lipsick(Lipsick, 2004) for commentary]. All the dREAM/Myb-MuvB genes are also conserved in mammals, and recently LIN9, the human homologue of aly/Mip130, has been shown to have tumour suppressor activity and to work in concert with Rb to promote differentiation (Gagrica et al.,2004). LIN9, LIN54 (human Mip120) and hMip40 are all also capable of binding directly to Rb (Korenjak et al., 2004).

Drosophila E2f2- and Rbf2-null mutants are viable and male fertile, but E2F2 females have reduced fertility(Cayirlioglu et al., 2001; Frolov et al., 2001; Stevaux et al., 2005), whereas Myb-, Dp- and Rbf-null mutants are lethal(Duronio et al., 1995; Manak et al., 2002; Royzman et al., 1997) and males mutant for weak Dp alleles are sterile(Duronio et al., 1998) but do not show a meiotic-arrest phenotype. Thus, the mutant phenotypes of the DNA-binding subunits dE2F2, Rbf, Rbf2, Dp and Myb are not consistent with them functioning in testes with aly and tombto activate gene expression. Indeed, Rbf2 function in ovaries is implicated in repression of some testis-specific genes(Stevaux et al., 2005).

There is remarkable evolutionary conservation of the interaction between dREAM/Myb-MuvB gene products in somatic tissues in mammals, flies and worms. We suggest that gene duplications in Drosophila of lin-54(tomb/mip120), lin-9 (aly/mip130) and lin-52 (CG12442/lin52) (J.J., K.D. and H.W.-C.,unpublished), has led to the evolution of a complex paralogous to the dREAM/MybMuvB complex, but using different DNA-binding subunits, dedicated to testis-specific transcriptional regulation.

We thank Michael Botchen for critical reading of the manuscript and for communicating results prior to publication; Carine Barreau and Jasmin Kirchner for reading the manuscript and members of the Zoology fly groups, especially Luke Alphey, for helpful discussions throughout this work. The P[GS]12862 line was kindly provided by the Kyoto DrosophilaStock Centre. This work was funded by Wellcome Trust and BBSRC grants to H.W.-C. E.B. was supported by an MRC doctoral studentship, H.W.-C. is a Royal Society University Research Fellow.