Regulation of transcription of meiotic cell cycle and terminal differentiation genes by the testis-specific Zn-finger protein matotopetli

Tissue and cell type specific transcriptional activation is a fundamental feature of cellular differentiation that mediates many changes in cell behaviour and morphology during development. One of the most astonishing developmentally regulated changes in cell morphology occurs in the production of sperm. The cellular differentiation events that constitute spermiogenesis require many gene products used at no other time in development. The unique features of male gamete differentiation are reflected in a developmentally specific transcription programme that initiates in primary spermatocytes in preparation for spermiogenesis. In Drosophila spermatogenesis,transcription is essentially shut down as mature primary spermatocytes enter the meiotic divisions. Therefore, gene products required during differentiation are largely produced from transcripts expressed pre-meiotically (Olivieri and Olivieri,1965; Schafer et al.,1995). A two-part genetic module, the meiotic arrest genes, is responsible for most of this transcriptional activation, and ensures that a large number of genes required for normal cellular morphogenesis are co-expressed in a cell type and developmental stage-specific manner(Lin et al., 1996; White-Cooper et al.,1998).

Males mutant for any of the meiotic arrest genes [including always early (aly), cannonball (can), meiosis I arrest (mia), spermatocyte arrest (sa), cookie monster (comr), achintya/vismay(achi/vis) and no-hitter (nht)] are viable but sterile. Testes from flies mutant for any one of these genes contain morphologically normal spermatogenic cells up to mature primary spermatocytes. However, the mutant cells then arrest and fail to initiate either the meiotic divisions or spermatid differentiation(Ayyar et al., 2003; Jiang and White-Cooper, 2003; Lin et al., 1996). Although transcription of a number of broadly expressed genes (such as cyclin A) is normal in the mutant spermatocytes, transcription of many spermiogenesis genes (e.g. the mitochondrial fusion protein fzo) is very low or undetectable. The aly-class meiotic arrest genes(aly, comr, achi/vis) are likely to act in a pathway distinct from the can-class genes, because they are required for transcription of a wider range of target genes. In addition to a role in transcription of spermiogenesis genes, aly-class genes are also essential for the transcriptional activation in primary spermatocytes of several genes required for progression into the meiotic divisions (namely twine, cycB,boule). By contrast can, mia, sa and nht (can-class genes) are not required for transcription of these cell cycle genes; however,they are required for translation of twine. The can-class meiotic arrest genes analysed to date encode testis specific homologues of more generally transcribed TATA-binding protein associated factors (TAFs)(Aoyagi and Wassarman, 2000; Hiller et al., 2001) (M.H. and M.T.F., unpublished). TAFs are subunits of the basal transcription factor TFIID, which is involved in recruitment of the RNA polymerase II holoenzyme to the proximal promoter region. Thus, the can-class genes are likely to activate full levels of transcription of spermatid differentiation genes through intimate association with target promoters(Hochheimer and Tjian,2003).

Not all genes transcribed in primary spermatocytes are controlled by the meiotic arrest pathway, so it is likely the targeting of particular promoters is mediated through sequence-specific DNA-binding activity. Thus, we would expect that one (or more) of the meiotic arrest genes encode a DNA-binding protein(s), whereas others are likely to encode regulatory proteins. aly is one of two Drosophila homologues of the C. elegans vulval differentiation regulator, lin-9(Beitel et al., 2000; White-Cooper et al., 2000). The SynMuvB pathway, in which lin-9 acts, is thought to negatively regulate adoption of vulval fate and promote hypodermal fate, through activation of a histone deacetylase chromatin remodelling complex(Ferguson and Horvitz, 1989; Lu and Horvitz, 1998; Solari and Ahringer, 2000). comr encodes a novel acidic protein with no significant similarities to other proteins in current sequence databases(Jiang and White-Cooper,2003). achi/vis was the first meiotic arrest locus to be shown to encode products with sequence-specific DNA-binding activity (Ayyar et al., 2003; Wang and Mann, 2003). The TALE class homeodomain proteins encoded by the recent gene duplication pair achi/vis are virtually identical to each other and are homologues of the human TGIF sequence-specific DNA-binding factor(Ayyar et al., 2003; Wang and Mann, 2003). Both achi and vis are expressed in many cells throughout development. However, flies that are null mutant for both genes are viable but male sterile, with a meiotic arrest phenotype. Achi/Vis proteins are localized to chromatin in wild-type primary spermatocytes and in aly mutant testes (Wang and Mann,2003).

Despite lacking any predicted DNA-binding motifs, both Aly and Comr proteins are concentrated on chromatin in primary spermatocytes. However, the nuclear localizations of Aly and Comr are mutually dependent: Aly remains cytoplasmic in comr mutant testes and vice versa. By contrast, the localization of Aly and Comr to the nucleus is independent of achi/vis (Ayyar et al.,2003; Jiang and White-Cooper,2003; Wang and Mann,2003; White-Cooper et al.,2000). Aly, Comr and Achi/Vis co-immunoprecipitate from testis protein extracts, suggesting that they interact in a common complex in vivo(Wang and Mann, 2003).

We describe the identification, cloning and characterization of the Drosophila aly-class meiotic arrest gene, matotopetli(topi). The similarity in phenotype between topi and aly, comr and achi/vis suggests that these genes act together in a common pathway. topi encodes a testis-specific predicted Zn-finger protein, making Topi the second putative sequence-specific transcription factor to be placed into the aly class. At least two separate Zn-finger containing regions of Topi are sufficient to bind directly to a 100 amino acid region in the middle of the Comr protein. Like achi/vis, topi activity is not required for the nuclear localization of Aly and Comr proteins. We suggest that the DNA-binding activities of Topi and Achi/Vis are required to recruit Aly and Comr to the promoters of target genes in primary spermatocytes. Consistent with this idea,the majority of target genes require both the topi and achi/vis DNA-binding factors for full transcriptional activation. However, rare exceptions to this rule have been revealed via microarray analysis. A small number of target promoters rely to a much greater extent on one or other of these DNA-binding activities, indicating that in at least some contexts the activities of achi/vis and topi are independent.

Flies were raised on standard cornmeal molasses (or sucrose) agar at 25°C. Visible markers and balancer chromosomes are described in FlyBase except where otherwise noted (FlyBase,1999). red e or bw; st/TM6B was used as the wild-type strain. aly⁵ red e/TM6C, w; can³ red e/TM6C, cn achi^Z3922vis^Z3922 bw/CyO, cn comr^Z1340 bw/CyO, mia st/TM6B and sa¹red/TM3 stocks were used to generate homozygotes for comparison to other meiotic arrest loci (Ayyar et al.,2003; Jiang and White-Cooper,2003; Lin et al.,1996). topi^Z3-2139, topi^Z3-0707 and topi^Z3-3767 were provided by C. Zuker from his collection of EMS-generated, viable lines, screened for male sterility by B. Wakimoto and D. Lindsley.

Live testes were dissected, squashed in 4 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. topi mutant testes contained many balls of cells, and no elongating cysts, hence the name matotopetli, which means `balls' in the Aztec language, nahuatl. Aly and Comr proteins were visualized using a BioRad Radiance Plus confocal microscope, using rabbit anti-Aly or rabbit anti-Comr antibodies detected with FITC-conjugated secondary antibodies (Sigma), DNA was co-stained with propidium iodide (Jiang and White-Cooper,2003; White-Cooper et al.,2000). Phase contrast images were captured using a Photometrics cooled CCD camera connected to a Zeiss Axiophot microscope, or a Q-imaging Retiga 1300 CCD camera linked to an Olympus BX50 microscope using IP Lab Spectrum or Openlab software (Improvision) and then imported into Photoshop. Images of RNA in situ hybridization to testes were captured on colour slide film, this was scanned and imported into Photoshop.

Northern blotting of poly-A+ RNA from whole males, whole female, agametic males and embryos using topi and rp49 probes was carried out as described previously (Hiller et al.,2001). RNA in situ hybridization was carried out as previously described (White-Cooper et al.,1998). Dig-labelled RNA probes were generated using dig-RNA labelling mix (Roche). Probes for topi, cyclinA, cyclinB, twine, fzoand Mst87F were generated by transcription off a linearized cDNA-clone plasmid. For TrxT and CG8349 a T3 RNA polymerase promoter was included in the 3′ RT-PCR primer, and purified PCR product was used as the template for transcription of the labelled RNA probe.

Full details of the microarray analysis will be presented elsewhere. Testes and seminal vesicles were dissected from 0- to 1-day-old males homozygous for red e (control), aly⁵, can³,comr^z1340, achi^Z3922vis^Z3922 and topi^Z3-2139 in testis buffer, and placed at–80°C within 30 minutes of dissection. For each genotype, three independent RNA samples were made from 200 testes. Testes were homogenized in Trizol (Roche), then shipped on dry ice to the BBSRC IGF facility in Glasgow for RNA extraction, labelling and Affymetrix array hybridization(http://www.mblab.gla.ac.uk/igf/index.html). Total RNA (6 μg) per sample was used for probe synthesis. Normalized signal intensities were averaged over the three replicates and compared between genotypes. Genes were selected for analysis from a list of genes expressed in wild-type testes where the mean signal for topi was eight times higher or lower than that for achi/vis. P values were calculated with a one tailed t-test on this pair of samples.

For RT-PCR, total RNA was extracted from dissected testes with Trizol reagent. The RNA was resuspended in RNAse free water at a concentration of three testes worth/μl. First-strand cDNA from 4 μl of this sample was generated using oligo-dT primers with the SuperScript II reverse transcriptase system (Invitrogen) following the manufacturer's instructions. As a negative control, wild-type testis RNA was used as the template and reverse transcriptase was omitted from the reaction. cDNA derived from 0.18 testes 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.

PCR primers (synthesized by MWG) were designed to amplify 400-1000 bp fragments from the transcript of the genes of interest. The T3 RNA polymerase promoter site, preceded by 6 bp of random sequence, was incorporated into the 3′ primer to facilitate production of RNA probes directly from the RT-PCR product. Primer sequences were: CG8349-5′ GCTCCTTCAGCGCTACATGC;CG8349-3′T3 GCAACGAATTAACCCTCACTAAAGGGCGCATAGGCACATCG; TrxT-5′TCGGCGAGGGCAGAGCTC; TrxT-3′T3 GCAACGAATTAACCCTCACTAAAGGGCATTCTCGTCGTGGGC.

topi was mapped by recombination between th and cu, and further localized by deficiency complementation to polytene interval 85E9-13. The topi region was defined by the right breakpoints of Df(3R)GB104, which complemented topi, and Df(3R)by10, which failed to complement topi. This identified a 60 kb region of genomic DNA containing 23 predicted genes. We used representation of ESTs from different tissues as a crude guide to the expression pattern of these genes. Seventeen were not represented in the EST set derived from testis. Four had ESTs from both testis and other tissues, and two (CG8484 and CG8526)had ESTs from only testis libraries. The premature stop codons in topi^Z3-2139 and topi^Z3-0707 and the mis-sense mutation in topi^Z3-3767 were identified by PCR amplifying the genomic region containing CG8484 from homozygous males, and sequencing the bulk PCR product. Predicted Zn-finger motifs within Topi were identified by eye and using the InterPro analysis tool(http://www.ebi.ac.uk/InterProScan/). topi homologues from Anopheles gambiae (on AAAB01008944.1)and Drosophila pseudoobscura (on Contig815_Contig5737) were identified using Blast searches of the genome sequence databases.

A Comr-Gal4-DNA-binding-domain fusion construct was made by subcloning the ORF from a full-length comr cDNA clone into the vector pGBKT7 using NdeI (at the start codon) and NotI. As no testis cDNA libraries suitable for two-hybrid screening were available, we generated and screened a testis cDNA-Gal4-Activation Domain (AD) fusion protein library by in vivo recombination using the Matchmaker library construction and screening kit (Clontech). Total RNA (1 μg), isolated from wild-type testes with Trizol, was used to synthesize first-strand cDNA using an oligo d(T) primer. Double-strand cDNA was synthesized with SMART III and CDS III anchors. The AD fusion library construction and two-hybrid screen were carried out in one step by co-transforming the yeast strain AH109 with ds cDNA, pGADT7-Rec and pGBKT7/comr. Colonies were picked from SD/-Ade/-His/-Leu/-Trp/X-α-Gal selection plates after 7 days. A screen of 10⁶ independent co-transformants yielded 39 colonies that grew under selective conditions and were blue in the presence of X-α-Gal. AD/library plasmids were isolated from each positive yeast colony, transformed into E. coli and sequenced; one of the positive clones contained the full ORF of CG8484 – topi.

Determination of the protein interaction domains necessitated construction of comr and topi deletion series. PCR products for the deletion derivatives were subcloned into pGBKT7 using NdeI and NotI sites engineered into the primers. Co-transformation of HA109 cells was with pGADT7-RecTopi (full length Topi) or pGADT7-RecComr (full length Comr) as appropriate. A 622 bp fragment covering five zinc fingers in Kruppel was generated by PCR from wild-type genomic DNA.

The Topi ORF was subcloned into HA-tagged pCDEF3 and the Comr ORF into FLAG-tagged pCDEF3 respectively at NdeI and NotI sites. 293T cells were cultured in DMEM with 10% FCS to 90% confluence then co-transfected with 10 μg HA-tagged Topi and Flag-tagged Comr using Lipofectamine 2000 reagent (Invitrogen). As a negative control, 293T cells were co-transfected with HA-tagged Topi and Flag-tagged human Smad1. Forty-eight hours after transfection, cells were collected and processed for immunoprecipitation(Bennett and Alphey, 2002),except that protein G sepharose incubation was carried out at 4°C overnight.

Mutations in matotopetli (topi) block both progression of the meiotic cell cycle in males and onset of spermatid differentiation. Wild-type adult Drosophila testes contain a spatially organized temporal array of differentiating male germ cells(Fig. 1A) (reviewed by Fuller, 1993). A small population of germline stem cells resides at the testis apical tip. Division of a germline stem cell and associated somatic cyst progenitor cells gives one spermatogonial daughter, which is committed to differentiation and encapsulated by two somatically derived cyst cells. Four mitotic amplification divisions of the spermatogonium results in a cyst of 16 interconnected primary spermatocytes, which then enter an extended period of growth and gene expression followed by meiotic division to produce a cyst of 64 round spermatids. By a programme including extensive elongation and other morphological changes, the spermatids differentiate into mature sperm. Testes lacking topi function contained germ cell cysts up to and including mature primary spermatocytes, but lacked cells in meiotic division or spermatid differentiation stages, placing topi into the meiotic arrest class of genes (Fig. 1B). topi null flies were viable and female fertile. Three alleles of topi were identified by screening a collection of EMS induced, viable, male sterile mutations (generated by E. Koundakjian and C. Zuker, screened for male fertility by B. Wakimoto and D. Lindsley). topi^Z3-2139 was identical in phenotype to topi^Z3-0707. topi^Z3-3767 appeared to be a hypomorphic allele, as occasional progression of mature primary spermatocytes into the meiotic divisions was observed in homozygotes and hemizygotes (data not shown).

The mature spermatocytes present in topi^Z3-3767/Df(3R)by10 or topi^Z3-3767 homozygous mutant testes resembled wild-type mature primary spermatocytes by phase contrast microscopy(Fig. 1C,F). However the chromatin morphology of mutant spermatocytes was subtly different from wild type when visualized by inclusion of Hoechst 33342 in the squash buffer. Mature wild-type primary spermatocytes have decondensed chromosomes, which are typically found as three chromatin domains close to the nuclear membrane(Fig. 1D,E). As primary spermatocytes pass through the G2-Meiosis I transition, the chromosomes condense and move away from the nuclear envelope. Chromosomes in topiarrested cells were somewhat more compact in appearance than in mature wild-type cells in late G2 (Fig. 1G,H), suggesting a block in the G2-M transition of meiosis I in topi mutant males. topi mutant chromatin most closely resembled that found in achi/vis mutant testes, and did not show the fuzzy, decondensed morphology typical of aly and comr(Ayyar et al., 2003; Jiang and White-Cooper,2003).

RNA in situ hybridization analysis of the effects of topi mutants on transcription of meiotic cell cycle and spermatid differentiation genes placed topi into the aly class of meiotic arrest mutants. The two strong alleles of topi showed greatly reduced levels of expression in primary spermatocytes of the meiotic cell cycle transcripts twine, cyclin B (Fig. 2C-F) and boule (not shown). The transcript levels of several spermatid differentiation genes, including fzo and Mst87F, were also dramatically reduced compared with wild-type levels(Fig. 2G-J). Some expression of fzo was detected in topi^Z3-3767 mutant testes,consistent with the weaker phenotype seen in this allele by phase-contrast microscopy (data not shown).

Our phenotypic analysis indicated that topi is a new member of the aly-class of meiotic arrest genes, which includes aly, comrand achi/vis. All three of these proteins localize to nuclei of wild-type primary spermatocytes. The behaviour of Aly and Comr proteins in achi/vis mutant testes and topi mutant testes was extremely similar. Immunofluorescence analysis with anti-Aly and anti-Comr antibodies revealed that both Aly and Comr proteins localized to nuclei in spermatocytes from males carrying strong loss of function mutations in topi(Fig. 3). In wild-type primary spermatocytes, Aly and Comr proteins were concentrated on chromatin as well as being distributed throughout the nucleoplasm. By contrast, in topimutant testes, Aly and Comr proteins did not appear to concentrate on chromatin, but were uniformly distributed within the nucleus.

The topi gene was localized by a combination of recombination mapping and deficiency complementation (see Materials and methods). Sequence analysis of genomic DNA fragments amplified from topi^Z3-3767/Df (3R)by10,topi^Z3-2139 and topi^Z3-0707 mutant animals by PCR revealed that each of the three topi alleles caused lesions in CG8484, identifying this predicted gene as matotopetli. topi^Z3-0707 and topi^Z3-2139 introduced stop codons, in the first third and in the last third of the predicted protein, respectively. The weaker allele (topi^Z3-3767)carried a mis-sense mutation.

topi encodes a 814 amino acid multiple Zn-finger predicted protein with a predicted molecular weight of 92 kDa(Fig. 4). A near full length cDNA isolated from testes (AT25463) was identified in the BDGP EST project and was fully sequenced. The conceptually translated Topi protein contains 10 C2H2 and one C2HC class Zn fingers predicted by the PFAM and prosite protein domain analysis tools. The predicted Zn fingers are clustered in the central region of the protein (amino acids 230-650), with the first two fingers being separated from the remaining nine by a 64 amino acid spacer. The mis-sense mutation in topi^Z3-3767 alters a conserved residue,G(516)D, just after the second Zn-binding cystine in the seventh predicted Zn finger. The non-sense mutation in topi^Z3-0707 would truncate the protein after the first predicted Zn finger (at amino acid 261). The non-sense mutation in topi^Z3-2139 would truncate the predicted Topi protein within the seventh predicted Zn finger (amino acid 521), potentially encoding a protein with six Zn fingers. Blast searches were used to identify putative topi orthologues, with sequence conservation extending beyond the Zn finger domains, in the mosquito Anopheles gambiae and in Drosophila pseudoobscura. Unlike both fly genes, the mosquito protein had only 10 predicted Zn fingers owing to lack of three crucial Zn-binding residues from the second Zn finger.

The matotpetli gene encodes a 2.7 kb transcript that was detected in northern blots of polyA+ RNA from whole adult males but not from adult males lacking a germ line, females or embryos(Fig. 5A). RNA in situ hybridization to whole testes revealed that topi transcript was expressed specifically in primary spermatocytes. topi transcript was not detected in spermatogonia or elongating spermatids(Fig. 5B). topitranscript was low or not detectable in spermatocytes from topi^Z3-0707/Df males(Fig. 5C).

A full-length clone of the topi predicted Zn-finger protein was identified independently in a yeast-two hybrid screen for testis cDNAs encoding proteins that interact with Comr (see Materials and methods). Full-length Comr did not interact in the two-hybrid system with a domain containing five C2H2 Zn fingers from the transcriptional repressor protein Kruppel, indicating that the Comr-Topi interaction is specific. The interaction between Topi and Comr proteins was confirmed using co-expression and immunoprecipitation from mammalian tissue culture cells(Fig. 6C). FLAG-tagged Comr and HA-tagged Topi were co-expressed in 293T tissue culture cells. HA-tagged Topi protein had an apparent M_r of 92 kDa, correlating well with the predicted protein size. Immunoprecipitation with anti-FLAG antibodies, followed by western blotting with anti-HA antibodies showed that Topi co-immunoprecipitated with Comr. The reciprocal experiment,immunoprecipitation with anti-HA antibodies and blotting with anti-FLAG, also showed co-immunoprecipitation of Comr and Topi. To control for non-specific interactions in the cultured cells we co-expressed HA-Topi with FLAG-Smad1. We were not able to co-immunoprecipitate this pair of proteins.

The domains of Comr and Topi responsible for the interaction were mapped by constructing N-terminal and C-terminal deletion series for each of the proteins and assaying interactions in yeast(Fig. 6A,B). Fragments of Comr containing residues 200-300 were capable of interacting with full-length Topi,whereas deletions lacking this region did not bind Topi. The Topi deletion series indicated that the Comr binding activity lies within the Zn-finger region. Specifically, fragments containing either the first two Zn fingers,fingers 7-10, or both, were capable of interacting with full-length Comr.

To explore whether topi and achi/vis might differ in their target gene specificities, we carried out a set of DNA microarray experiments to examine gene expression in wild-type and meiotic arrest mutant testes (see Materials and methods) (H.W.-C., unpublished). About 1000 genes were expressed at reduced levels (at least fourfold) in aly-class meiotic arrest mutants than in wild type, most of these transcriptional targets of the meiotic arrest genes depended equally on the functions of aly, comr, topi and achi/vis. However, transcription of a subset (about 42) of genes was much more dependent on achi/vis than on any of the other meiotic arrest loci. Transcription of a different subset (about 44) of genes seemed to be much more dependent on topi. The raw and log ratio array data for these genes are presented in Table S1,along with data for genes whose expression in wild-type and meiotic arrest mutant testes has already been examined by RNA in situ hybridization or northern blotting (White-Cooper et al.,1998). Further analysis of representative genes by RT-PCR from wild-type and mutant testes confirmed the achi/vis versus topi differential dependence.

CG8349 encodes a predicted cytidine deaminase. 11 testis ESTs for this gene have been sequenced, no CG8349 ESTs have been sequenced from other tissue libraries, indicating that CG8349 transcription is highly testis enriched. Microarray analysis indicated that expression of this gene was approximately five times lower in aly, comr and topi than in wild type, while expression of CG8349 was reduced by two-hundred times in achi/vis compared with wild type. By RT-PCR, we found that CG8349 transcript was detectable, but levels were dramatically reduced in aly, comr, topi, mia and samutant testes. However, no transcript was detected in achi/vis mutant testes (Fig. 7A). To further investigate these changes in gene expression, we used in situ hybridization against wild-type and mutant testes. The relative level of transcript can be compared between wild-type and mutant testes using this method as the tissues were mixed before hybridization. In situ hybridization revealed that CG8349 was expressed in primary spermatocytes, and the transcript persisted into mid-late stages of spermatid elongation(Fig. 7B). As predicted by the RT-PCR and array analysis, the in situ hybridization signal intensity for CG8349 was lower than wild type in aly, comr and topi testes (Fig. 7C,E), and the transcript was undetectable by in situ hybridization in achi/vis mutant testes(Fig. 7D).

The microarray analysis indicated that transcripts of TrxT, a testis specific isoform of thioredoxin(Svensson et al., 2003), were 2.5 times less abundant in aly and comr than in wild-type or achi/vis mutant testes. However, TrxT transcript was at least fifty times less abundant in topi mutant testes than in wild type. Consistent with these data, we were unable to RT-PCR TrxTfrom topi mutant testes, while the RT-PCR of TrxT revealed only a mild reduction in transcript levels compared with wild type in aly,comr and achi/vis mutant testes(Fig. 7F). No reduction in transcript level was evident by RT-PCR in mia or sa mutant testes. TrxT expression was first detected by in situ hybridization in primary spermatocytes in wild-type testes, the transcript persisted after the meiotic divisions and was detected until mid-elongation stages of spermiogenesis (Fig. 7G). aly and comr mutant testes showed expression in the primary spermatocytes; however, the signal intensity in these cells was lower than in the wild-type control (Fig. 7H). Thus, although significant expression of TrxT was detected in aly and comr testes, aly and comr are both required for full expression of TrxT. Signal intensity was similar to control levels in achi/vis mutant testes(Fig. 7I). TrxTtranscript was undetectable in topi mutant testes by in situ hybridization (Fig. 7J),confirming that topi is absolutely required for expression of TrxT.

The aly class meiotic arrest genes of Drosophila form a functional module required for transcription in primary spermatocytes of a number of meiotic cell cycle regulatory components and spermatid differentiation genes. aly encodes a homologue of the C. elegans SynMuvB gene lin-9, which is thought to be involved in recruiting the NURD nucleosome remodelling and histone deacetylase complex to repress target genes involved in vulval development(Beitel et al., 2000; White-Cooper et al., 2000). The Drosophila Aly protein is found in a complex with Comr, a novel protein that also associates with chromatin in Drosophila primary spermatocytes (Jiang and White-Cooper,2003; Wang and Mann,2003) (J.J. and H.W.-C., unpublished). A pathway by which Aly and Comr are assembled onto chromatin in primary spermatocytes is now beginning to emerge (Fig. 8). Interaction between Aly and Comr is required for each of the two proteins to localize to the nucleus, because absence of Aly leads to cytoplasmic accumulation of Comr and vice versa (Jiang and White-Cooper,2003). Once the Aly/Comr complex has entered the nucleus,wild-type function of the predicted DNA-binding proteins Topi and Achi/Vis appears to be required for concentration of the Aly/Comr complex on chromatin. The Aly/Comr complex interacts structurally with Achi/Vis in primary spermatocytes (Wang and Mann,2003) and Topi directly binds Comr. We propose that the Zn-finger protein Topi, via its physical interaction with Comr, and the TALE-class homeodomain proteins Achi/Vis, via their physical interaction with Aly/Comr,recruit an Aly/Comr complex to chromatin. We propose that a typical target gene, e.g. dj, has both topi and achi/vis target sequences in its promoter. Topi and Achi/Vis would bind to these regions,either independently or co-operatively, and recruit a pre-formed Aly/Comr complex from the nucleoplasm onto the promoter. Both Topi and Achi/Vis would need to be present to efficiently recruit Aly/Comr. Alternatively, Topi and Achi/Vis could bind to an Aly/Comr complex in the nucleus, with the target sequence for either DNA-binding protein sufficient to localize the entire assemblage to the target promoter. The aly-class meiotic arrest gene topi plays a crucial role in function of this regulatory module required for activation of a novel transcriptional programme in primary spermatocytes. Loss of function of topi results in the arrest of mutant cells as mature primary spermatocytes, and wild-type function of topi is required to generate a store of transcripts to support meiosis and the post-meiotic morphological changes that occur during spermatid differentiation.

Although the meiotic arrest genes are crucial for transcription of a large suite of primary spermatocyte specific genes, they are not global regulators,as many transcripts accumulate to normal levels in the primary spermatocytes of mutant testes. Cis-acting sequences at target promoters would make them dependent on the function of aly and comr. Targeting of the Aly/Comr complex activity to particular sites on the genome may be crucial for understanding the molecular mechanism of this genetic module. The identification of two potential DNA-binding activities within the aly-class of meiotic arrest genes raises questions about the requirements for each of these proteins at different target gene promoters. In our microarray analysis, all genes we found to depend on aly or comr were also dependent on topi and/or achi/vis. The overwhelming majority of topi-dependent genes are also dependent on aly, comr and achi/vis for their full activation; genes in this class include Mst87F, dj and fzo. This indicates that both Topi and Achi/Vis are usually needed at the promoters of meiotic arrest target genes for full activity. Many transcription factors function as homo- or heterodimers, and many developmentally regulated promoters have binding sites for more than one transcription factor. These characteristics allow a combinatorial mechanism to ensure spatiotemporal accuracy of the transcription of any particular gene. The simplest explanation for the requirement for both Topi and Achi/Vis is that the promoters of the majority of terminal differentiation genes have binding sites for both factors. achi and vis are expressed at many stages of development,yet their testis target genes remain silent in these other tissues. This silence is likely to be due, at least in part, to the lack of topiexpression in other tissues. Thus, the cooperation between these two transcription factors is essential for the developmental accuracy of expression of genes required for spermatogenesis.

It is intriguing that transcription of some genes (including TrxTand CG8349) showed a strong requirement for topi but not achi/vis (or vice versa), and that expression of these genes was less strongly dependent on the activity of the other meiotic arrest genes. This indicates that, at least in these contexts, the binding of the Achi/Vis and Topi transcription factors to DNA is independent. Another feature these genes have in common is that they are members of clusters of related genes,with different expression patterns. TrxT is found adjacent in the genome to the female specific thioredoxin, deadhead(Svensson et al., 2003). CG8349 is the 3′-most of a cluster of three predicted cytidine deaminase genes (the others being CG8360 and CG8353). The structure of these promoters may differ from the canonical structure for meiotic arrest target genes, perhaps by having several binding sites for only one of the aly-class transcription factors. In this situation partial activation of gene expression would be achieved on binding of one transcription factor. Further recruitment of Aly and Comr (and usually the other DNA-binding factor) would be needed for the robust expression levels seen in wild type.

The meiotic arrest gene topi encodes a protein with multiple Zn fingers. The final five Zn-fingers of Topi form an evolutionarily constrained unit, which we suggest may function via sequence specific DNA binding activity. topi has a large number of transcriptional targets(H.W.-C., unpublished). If, as is likely, Topi binds directly to cisacting regulatory elements associated with target genes in a sequence-specific manner, we would expect the protein sequence in the DNA-binding domain to be evolutionarily constrained. Changes in the affinity of Topi for any particular DNA sequence would have to be compensated for by changes in the sequence of a great many promoters, and are therefore likely to be selected against. Conservation of topi protein sequences within Diptera is extremely high across the final five Zn-finger motifs, and much lower across the first six Zn-fingers and at the termini. This conservation is particularly apparent in the loop regions of the final five Zn fingers, which are likely to be exposed, and important for interaction with DNA. Additionally these final five Zn finger motifs are more similar to those found in many transcription factors from other taxa. topi^Z3-2139 has a non-sense mutation that would truncate the protein before this predicted DNA binding region, and would presumably therefore be unable to bind DNA. The mis-sense mutation in the hypomorphic allele topi^Z3-3767 may reduce the affinity of the protein for its target site, but not totally abolish DNA-binding activity,thus allowing some residual function.

We propose that the first Zn finger of Topi is most important for the binding to Comr. Zn-finger motifs were first identified as nucleic acid binding domains. However Zn fingers are also known to be important for some protein-protein interactions (Evans and Hollenberg, 1988; Tsai and Reed, 1998). There is precedent among known Zn-finger transcription factors, as we propose for Topi, for certain of the fingers to bind proteins, while others contact the DNA, allowing the transcription factor to target a multi-subunit complex to specific promoters(Kalenik et al., 1997). The Topi protein interacts structurally with Comr, and the region of Topi containing the first two Zn fingers was sufficient for interaction with Comr in yeast two hybrid assays. We propose that the first Zn finger is more important for Comr binding because the second of the two Zn fingers in mosquito Topi lacks crucial Zn-binding residues, so is unlikely to form a conventional finger structure in the folded protein. The final five Zn fingers, which we propose function primarily as DNA-binding motifs, also showed some Comr-binding activity in the yeast two-hybrid system. This interaction may stabilize, or increase the binding affinity of, the Topi-Comr complex. Strikingly, the N-terminal Comr binding region of topi is much less well conserved than the last five Zn finger containing domain, even among Diptera (D. melanogaster, D. pseudoobscura, A. gambiae). Also apparent is the relatively low conservation within the loop regions of the Zn fingers. If comr is rapidly evolving, the Comr interaction domains within Comr-binding proteins would also be expected to show low levels of conservation. It is possible that the N-terminal Zn-finger Comr-binding domain of Topi has evolved rapidly in concert with the rapid evolution of Comr. Searches of sequence databases had previously indicated failed to identify a homologue of D. melanogaster comr in the genome sequence of A. gambiae, which diverged from Drosophila 250 Mya(Zdobnov et al., 2002). Partial sequence of a comr homologue (H.W.-C., unpublished) was found in the set of unassembled genome sequence contigs from another Drosophilid, D. pseudoobscura, which diverged from D. melanogaster 46 Mya(Bergman et al., 2002). The conservation with D. melanogaster comr (31% identical, 52% similar)was remarkably low considering the relatedness of the species. This relative lack of conservation was apparent even in the Topi-binding region.

The exact mechanism by which the aly homologue lin-9functions is not well understood. However, some insight has been provided by the molecular analysis of other genes in the C. elegans Syn-MuvB pathway, including lin-35 (Rb, Retinoblastoma), lin-53(RbAp48) and hda-1 (histone deacetylase)(Lu and Horvitz, 1998). Other components of the NURD histone deacetylase and chromatin remodelling complex also have SynMuvB activity (Solari and Ahringer, 2000). The biochemical function of NURD is to de-acetylate histones and re-position nucleosomes(Xue et al., 1998; Zhang et al., 1998). Typically, histone de-acetylation is associated with transcriptional repression, although exceptions to this rule have been found(De Rubertis et al., 1996; Struhl, 1998). Another gene in the SynMuvB pathway, lin-13, has been shown to encode a putative Rb-interacting protein with multiple C2H2 Zn-fingers, which may be important for targeting Rb and the NURD complex to specific sites(Melendez and Greenwald,2000). The loop regions of the Zn fingers in LIN-13 showed no significant similarity with those in Topi, however the two proteins could be playing similar roles in linking lin-9 and NURD to target promoters. Analogous to the C. elegans system, we have proposed that aly, and by extension other genes in the same pathway as aly, functions through a NURD-complex chromatin-remodelling activity. Many genes require the aly-class meiotic arrest genes for activation of expression in primary spermatocytes, while a few may require these genes for transcriptional repression (H.W.-C., unpublished). Several SynMuvB genes have also been shown to be important for expression of transgenes in repetitive extrachromosomal arrays in C. elegans, indicating that, at least in certain specific contexts, the SynMuvB pathway activates gene expression (Hsieh et al.,1999). We favour a model in which the putative sequence-specific DNA-binding proteins, Topi and Achi/Vis target Aly and Comr to specific promoters, where Aly/Comr may then recruit a NURD complex to alter local chromatin structure. In the second part of the model, transcription of specific terminal differentiation genes in primary spermatocytes in some way requires the altered chromatin structure set up by the function of the Aly/Comr complex. The can-class meiotic arrest genes appear to form a different functional module that is also required for normal levels of transcription of many of the same target genes that require aly/comr. So far, all of the can class meiotic arrest genes molecularly identified encode homologues of subunits of TFIID. A TFIID complex containing the testis specific TAFs encoded by the can-class meiotic arrest genes may have a higher affinity for the altered chromatin conformation than conventional TFIID, and thus be required for full transcriptional activity.

We thank James Wakefield and Daimark Bennett for critical reading of the manuscript; and Luke Alphey, Myles Axton and members of the Fuller and White-Cooper laboratories for helpful discussions throughout this work. We thank Cricket Wood for help with the recombination mapping of topi. We are indebted to Barbara Wakimoto, Dan Lindsley and Charles Zuker for the topi alleles. L.P. thanks Alberto Darszon and Mario Zurita for allowing some of the experiments to be carried out in their respective laboratories. The microarray analysis was conducted by the BBSRC funded Drosophila IGF facility in Glasgow – thanks to Steve Russell,Julian Dow and Jing Wang. This work was supported by a Stanford Medical School Dean's postdoctoral fellowship and a CONACyT Fellowship (990544) to L.P., an NIH grant (1R01 1HD32936) to M.T.F., and Royal Society, MRC and Wellcome Trust grants to H.W.-C.