The Drosophila cdc25 homolog twine is required for meiosis

The cdc25 gene was initially identified in Schizosaccharomyces pombe where it acts as an important regulator of entry into mitosis (Russell and Nurse, 1986). The analysis of a S. cerevisiae homolog and of the Drosophila homolog encoded at the string locus emphasized the general importance (Russell et al.,1989; Edgar and O’Farrell, 1989), and studies in a variety of systems have revealed the biochemical basis of cdc25 function (for a recent review see Millar and Russell, 1992).

cdc25 function is required for entry into mitosis. Mutations in the S. pombe cdc25 gene and in string lead to an arrest in the G2-phase of the cell cycle. Without cdc25 function, activation of the mitosis promoting activity of the p34^cdc2-kinase does not occur (Booher et al., 1989; Moreno et al., 1989). This activation requires the association of p34^cdc2 with cyclin B and the removal of inhibitory phosphate modifications from amino acid residues in the ATP binding site of the p34^cdc2-kinase (for a review see Nurse, 1990). Whereas in S. pombe this inhibitory phosphorylation occurs on Tyr-15 only, both Tyr-15 and the adjacent Thr-14 are phosphorylated in higher eukaryotes (Gould and Nurse, 1989; Solomon et al., 1990; Krek and Nigg, 1991; Norbury et al., 1991). The biochemical characterization of cdc25 proteins from humans, Drosophila and S. pombe indicated that they function as phosphatases removing these inhibitory phosphate modifications (Dunphy and Kumagai, 1991, Gautier et al., 1991; Millar et al., 1991; Strausfeld et al., 1991). The dephosphorylation of Tyr-15 by the cdc25 phosphatase has been directly demonstrated and circumstantial evidence has indicated that the cdc25 phosphatase can also dephosphorylate Thr-14.

Sequence comparison revealed low, but significant similarities between phosphotyrosyl phosphatases and cdc25 phosphatases. The greatest similarity, however, was observed with a phosphatase from Vaccinia virus, VH1, which is capable of both tyrosyl and seryl dephosphorylation (Guan et al., 1991). The cdc25 phosphatase, therefore, appears to be a member of a novel class of phosphotyrosyl phosphatases capable of both seryl/threonyl and tyrosyl dephosphorylation.

Despite this apparently relaxed specificity, the cdc25 phosphatase appears to have an exceptionally high preference for the cyclin B-p34^cdc2 complex as a substrate. Recent findings with human cdc25 phosphatases provide a possible explanation for this behaviour. cdc25 phosphatase activity in vitro was found to be stimulated several fold by the addition of purified cyclin B (Galaktionov and Beach, 1991). Furthermore, transient association of the cdc25 phosphatase with the cyclin B-p34^cdc2 complex has been observed during M-phase in humans and Xenopus (Galaktionov and Beach, 1991; Jessus and Beach, 1992). The cdc25 phosphatase might therefore be maximally active only when associated with the cyclin B-p34^cdc2 complex.

In Drosophila, the expression of the string phosphatase is not only required for mitosis, but actually controls the timing of the embryonic cell divisions (Edgar and O’Farrell, 1990). Beginning at the cellular blastoderm stage, zygotic expression of string starts in a spatially defined and highly dynamic pattern which anticipates the pattern of the embryonic mitoses. Premature expression of string from a heat inducible transgene results in a premature entry into mitosis. In contrast, overexpression of the p34^cdc2 kinase or the cyclin proteins does not affect the timing of mitoses (C.F.L., unpublished observations). These results, therefore, have indicated that the cells in the Drosophila embryo remain in the G2-phase until a pulse of string transcription produces cdc25 phosphatase activity which results in the activation of the cyclin B-p34^cdc2 kinase and entry into mitosis.

In addition to cyclin B and p34^cdc2, a variety of different cyclin proteins and cdc2-like kinases have been identified in Drosophila (Lehner and O’Farrell, 1989; Lehner and O’Farrell, 1990a; Leopold and O’Farrell, 1991) and in other higher eukaryotes (Swenson et al., 1986; Elledge and Spottswood, 1991; Koff et al., 1991; Lew et al., 1991; Matsushime et al., 1991; Paris et al., 1991; Tsai et al., 1991; Xiong et al., 1991). Therefore, various additional cyclin-kinase complexes are likely to exist. In fact, such a complex has been clearly documented in the case of cyclin A and the cdk2 kinase in humans (Pines and Hunter, 1990; Tsai et al., 1991). In order to address whether multiple cdc25 homologs are involved in the regulation of such complexes, we have started a search for additional Drosophila cdc25 homologs. Here we report the existence of a second Drosophila cdc25 homolog. The same gene has been independently isolated with a different approach involving complementation of S. pombe cdc25 and has been named twine (Jimenez et al., 1990; Alphey et al., 1992). Our analyses of twine expression and of the phenotype resulting from a mutation in twine indicates that this second cdc25 homolog is a meiotic regulator.

The stock mat(2)synHB5 cn bw/CyO,DTS-513 and other stocks with female sterile mutations on the second chromosome have been isolated by Schüpbach and Wieschaus (1989) and were kindly provided by T. Schüpbach, Princeton University, Princeton. The stock Df(2L)RN2/In(2LR)O carrying a defiency with the breakpoints 35E1.2;36A4.5 was kindly provided by M. Ashburner, University of Cambridge, Cambridge.

For the identification of additional cdc25 homologs in Drosophila three degenerate primers corresponding to the most highly conserved regions in cdc25 phosphatases were synthesized:

primer 1: 5′-GGA GGATCC AT(ACT) GA(CT) TG(CT) (AC)GN T(AT)(CT) CCN TA(CT) GA-3′

primer 2: 5′-GGA TCTAGA (AG)TA NAT (CT)TC NGG (AG)TA NTG NA(AG) NGC NGG (AG)TA-3′

primer 3: 5′-GGA TCTAGA NGG NCC NC(GT) NTC NG(AT) N(CG)(AT) (AG)(AT)A (CT)TC (AG)CA-3′

In the putative amino acid sequence of the twine product (Fig. 1) the corresponding regions are found at the positions 241–248 (primer 1), 287–295 (primer 3) and 313-322 (primer 2). Primer 1 included a BamHI restriction site at the 5′ end, and primer 2 and 3 a XbaI site.

The primer combinations 1+2 and 1+3 were used in enzymatic amplification reactions (Saiki et al., 1985) using DNA isolated from a lambda zap ovary cDNA library. The isolation of the template DNA and the reaction conditions for enzymatic amplification have been described previously (Lehner and O’Farrell, 1990b). The temperature cycle, however, was changed to 1 minute at 94° C, 2 minutes at 48° C, 3 minutes at 72° C, and the cycle was repeated 35 times. With both primer combinations, two main products in the expected size range were amplified. The two main reaction products obtained with the primer combination 1+2 were gel purified and used as a template in a reaction with the primer combination 1+3. This reamplification also yielded two products with the same size as those obtained with the same primer combination in primary enzymatic amplifications from total library DNA. The larger products were also obtained when a string cDNA was used as a template with both primer combinations. The smaller reaction product obtained with the primer combination 1+2 was gel purified and inserted into a Bluescript vector.

For the analysis of mutations in the twine gene, the two primers corresponding to the regions flanking the coding sequence were synthesized:

primer 4: 5′-CGA GGATCC ATG GAA TAC CAG GCC AAA AGG-3′

primer 5: 5′-CGA TCTAGA GCC TGA AAT GGT CAC TC-3′

Primer 4 included a BamHI restriction site, primer 5 a XbaI restriction site. Size comparison of the reaction products enzymatically amplified from genomic DNA or cDNA templates indicated that the coding sequence of the twine gene is not interrupted by introns. For the isolation of the twine coding sequence from mat(2)synHB5 flies, 1 μg of genomic DNA isolated from homozygous males was used in an enzymatic amplification reaction using primers 4 and 5. The temperature cycle was adjusted to 1 minute at 94° C, 2 minutes at 53° C, 3 minutes at 72° C, and was repeated 35 times. The reaction products obtained in four independent reactions were gel purified and inserted into a Bluescript vector.

After sequence analysis of the fragment obtained by enzymatic amplification from the lambda zap ovary cDNA library, 10 cDNA clones were isolated from this library and plaque purified using the sequenced fragment as a probe after random primer labeling. After plasmid rescue from the lambda zap phages, the sizes of the cDNA inserts were determined, as well as the distance between the 5′ end of the insert and the region annealing to primer 3. For this purpose, the size of the products obtained by enzymatic amplification with primer 3 and either M13 primer or reverse primer (which flank the insert in the rescued plasmids) were determined after gel electrophoresis.

The largest insert was completely sequenced using double-stranded plasmid DNA as a template and a Sequenase kit (USB). In addition, the 5′ ends of two additional cDNA clones which, according to the PCR assay described above, extended far towards the 5′ end were sequenced. One of the clones started at position 35 of the nucleotide sequence shown in Fig. 1, the other at position 44.

For the analysis of the mat(2)synHB5 mutation, one clone containing the twine coding sequence isolated by enzymatic amplification from mat(2)synHB5 DNA was completely sequenced. To exclude the possibility that the identified mutation was introduced during enzymatic amplification, the result was confirmed by sequencing the corresponding region in three additional clones derived from independent PCR reactions.

The wild-type cDNA and the twine coding sequence isolated from mat(2)synHB5 flies was inserted into the vector pSM1 (Russell, 1989). Transformation of S. pombe was done as described (Okazaki et al., 1990). The strain cdc25–22 leu1-32 h-^S was transformed to leucine prototrophy at 25° C and colonies were then tested for the ability to divide at the restrictive temperature 36° C. Media used in genetic and physiological procedures have been described (Gutz et al., 1974; Nurse, 1975). These were Yeast extract (YE) containing 50 mg/ml each of adenine, uracil, leucine, lysine and histidine, Malt extract (ME), and EMM2 minimal, supplemented as required.

Hybridizations to polytene chromosomes from salivary glands of wandering third instar larvae were carried out with a digoxi-geninylated twine cDNA probe using reagents from the Genius kit (Boehringer).

In situ hybridization experiments with embryos were done according to Tautz and Pfeifle (1989) with minor modifications as described in Lehner and O’Farrell (1990b). In situ hybridization experiments with ovaries were done as described by Ephrussi et al. (1991). This method was also used for experiments with testis.

Testes and ovaries were dissected in Schneider’s medium. Testes were fixed for 10 minutes in 4% formaldehyde/PBS (8 mM Na₂ HPO₄, 1.5 mM KH₂ PO₄, 138 mM NaCl, 4 mM KCl), stained with Hoechst 33258 (1 μg/μl in PBS) and mounted in 70% glycerol/PBS. Ovaries were fixed following the method of Puro (1991) with slight modifications. After dissection, ovaries were either directly fixed or incubated for 3 minutes in 75 mM KCl before fixation. Fixation was done during 1 hour in Carnoy’s fixative (6:3:1 ethanol:chloroform:acetic acid). Subsequently, ovaries were incubated for 1 hour in ethanol and rehydrated by incubation in a series of reducing concentrations of ethanol in water. Ovaries were incubated in 1 N HCl for 10 minutes at room temperature and for 8 minutes at 60 ° C. After Feulgen staining, ovarioles were dissected and mounted in 70% glycerol in 15 mM NaHSO₃, 15 mM HCl.

Embryos from homozygous mat(2)synHB5 females or wild-type females were collected for 10 minutes after several rounds of pre-collections. They were fixed and stained with Hoechst 33258 as described previously (Lehner and O’Farrell, 1989). Immunofluorescent labeling with a monoclonal antibody recognising the sperm tail (Karr, 1991) was done as described by Karr (1991).

Testes, ovarioles and embryos were analysed in a Zeiss Axiophot microscope using phase contrast, differential interference contrast and epifluorescence. Kodak technical pan film was used for photography.

Degenerate primers corresponding to the most conserved regions in cdc25 phosphatases were used for enzymatic amplification from an ovary cDNA library template (for details see Materials and methods). One of the two resulting products was clearly different from string, but revealed strong sequence similarities to cdc25 phosphatases. Using this fragment as a probe, several cDNA clones were isolated from the ovary cDNA library. The DNA sequence derived from the clone with the largest cDNA insert is shown in Fig. 1. The open reading frame starting from the first methionine is 395 amino acids in length. Sequence comparison revealed that the corresponding gene was identical to the gene twine (L. Alphey and D. Glover, personal communication) which has been isolated independently in a screen for Drosophila cDNAs complementing the temperature sensitive cdc25 allele in S. pombe (Jimenez et al., 1990).

The alignment of the putative amino acid sequence of the twine product with other cdc25 homologs revealed extensive similarities in the C-terminal domain as shown in Fig. 2A. The most extensive conservation was found in the region with the HC motif, a signature sequence which contains residues essential for catalysis in all tyrosyl phosphatases (arrows in Fig. 2A). The N-terminal domain did not show similarities with other cdc25 homologs except for two motifs that were also present in the string protein (not shown). Comparison of sequence identities obtained after pairwise alignments clearly demonstrated that the two Drosophila sequences, string and twine, are most closely related to each other (43% identity, see Fig. 2B) and significantly less homologous to cdc25 of S. pombe or to the three human homologs (22–28% identity).

The pattern of twine expression was analyzed by whole-mount in situ hybridization with a digoxygeninylated twine cDNA probe. twine expression was not observed in imaginal and larval tissues of third instar larvae. Labeling was also not detected in late embryonic stages (not shown). In contrast to string, which is expressed zygotically in dynamic patterns (Edgar and O’Farrell, 1989), twine, therefore, does not appear to be expressed zygotically in the embryo. In very young embryos, however, intense labeling was observed, uniformly distributed throughout the embryo (not shown), as has also been described in the case of string (Edgar and O’Farrell, 1989). Since zygotic transcription has not yet started in these early embryos, this labeling must reflect maternally derived twine mRNA (see below). twine mRNA was detected during all of the early, syncytial stages. After the last syncytial division (mitosis 13), concomittant with cellularization, signals were fading rapidly and were detected predominantly in the yolky interior. After gastrulation, only background levels were observed.

The expression of the maternal twine mRNA was analyzed by in situ hybridization experiments with ovaries (Fig. 3; for a description of oogenesis see Mahowald and Kambysellis, 1980). Strong labeling was observed in the nurse cells of stage 10 oocytes (Fig. 3B). In stage 11 egg chambers (Fig. 3C), labeling was seen not only in the nurse cells, but also in the oocyte, indicating that twine mRNA is translocated from the nurse cells into the oocyte during nurse cell contraction. During the early stages of oogenesis, signals above background were not seen prior to stage 7 and only weak signals were observed during stage 8 and 9 (Fig. 3A). Thus, twine expression does not occur in the somatic follicle cells while they proliferate mitotically during stages 2-5. In contrast to twine, (and in contrast to the results of Alphey et al., 1992), string was found to be expressed also during earlier stages of oogenesis (Fig. 3D). In these early stages, however, string expression occurred predominantly in the somatic follicle cells (Fig. 3D, arrow). After the end of follicle cell proliferation (stage 6), string mRNA was no longer found in the follicle cells, and in later stages (Fig. 3E,F), string expression was observed in germ line cells where it closely paralleled the expression of twine. The expression pattern of the twine cdc25 homolog during oogenesis is consistent with a role either in meiosis or in early embryonic mitoses which are known to rely on maternally provided components. In order to distinguish between these two possibilities, it was of interest to analyze the expression of twine in male gonads, where only meiosis and no concurrent production of stores for the early embryo occurs (for a description of spermatogenesis see Lindsley and Tokuyasu, 1980; Gönczy et al., 1992). The results of these experiments are shown in Fig. 4. twine expression was not detected at the apical tip of the testis where the stem cells divide mitotically to form the cysts containing 16 syncytial spermatocytes (Fig. 4B). However, during the premeiotic growth phase of the cysts, labeling was clearly present. As illustrated in Fig. 4C, labeling was not only present in the premeiotic, 16-cell cysts (arrow-head), but also after meiosis I in 32-cell cysts (arrow). No labeling was seen after meiosis II, during the process of spermatid differentiation (star).

In situ hybridization experiments with the string probe yielded complementary results. String expression was detected at the apical tip where the mitotic divisions of the germ line cells lead to the formation of the cysts. During the meiotic stages, signals with the string probe were not above background (C.F.L., unpublished observation, see also Alphey et al., 1992).

The results of the in situ hybridization experiments suggested that the twine product serves a role during meiosis in both sexes Accordingly, we anticipated that mutations in twine would result in female and male sterility. Interestingly, the female sterile mutation, mat(2)synHB5, isolated by Schüpbach and Wieschaus (1989) had been mapped to the same genomic region as the twine gene (35F; according to in situ hybridizations to polytene chromosomes, not shown). The analysis of homozygous mat(2)synHB5 males revealed that this mutation also resulted in complete male sterility (see below). Male and female sterility were also observed in flies with the mat(2)synHB5 chromosome over the deficiency Df(2L)RN2, which deletes the twine gene. Thus, the mat(2)synHB5 mutation had the characteristics expected for a twine allele, and was characterized further. By in situ hybridization experiments, twine mRNA was detected at wild-type levels in early embryos derived from mothers homozygous for mat(2)synHB5 (not shown). This observation suggested that mat(2)synHB5 was either a mutation in the coding sequence or not a twine allele at all. This was analyzed by sequencing the coding region of the twine gene isolated from mat(2)synHB5 flies by enzymatic amplification (see Material and methods). As shown in Fig. 5, the sequence revealed the presence of one missense mutation. The codon for a proline, which is conserved in all the known cdc25 homologs and is located C-terminally in the HC motif containing the active site, was found to be changed into a leucine codon.

Fig. 4. Expression of twine during spermatogenesis A whole-mount preparation of a testis after in situ hybridization with a twine probe is shown in A. The apical tip is shown at higher magnification in B and the region were meiosis occurs in C. twine expression is not found in the proliferation center at the apical tip, but starts in cysts during the growth phase. It is detected in premeiotic cysts (arrowhead) and in cysts after meiosis I (arrow), but not during the postmeiotic stages (star).

In order to demonstrate that the observed mutation would impair the function of the twine product, we constructed plasmids that allowed complementation experiments in S. pombe (not shown). The construct with the wild-type twine cDNA transformed into an S. pombe strain with a temperature sensitive mutation in cdc25 was able to restore growth at the restrictive temperature. In contrast, the construct with the mutant twine allele did not allow growth at the restrictive temperature. These results indicate that the wild-type twine product is a functional cdc25 homolog and that the mutation present in mat(2)synHB5 does impair its function.

We cannot exclude, however, that the mat(2)synHB5 product has some residual activity.

The phenotypic consequences of the mat(2)synHB5 mutation were studied cytologically. Fig. 6 illustrates the results obtained with testes. Testes developed in homozygous mat(2)synHB5 males (Fig. 6A′,B′) and displayed a similar morphology as testes from wild-type males (Fig. 6A,B). Inspection of the apical tip at high magnification (Fig. 6C and C′) clearly demonstrated that premeiotic cysts with normal appearance were formed in the mutant testis. The arrowheads in Fig. 6C′ point to several clearly visible nuclei in a cyst of a mutant. However, whereas in wild-type testis meiotic figures with condensed chromosomes during either meiosis I (Fig. 6D) or meiosis II (not shown) were readily observed after staining with a DNA stain, meiotic figures were never observed in mutant testes. Accordingly, the cysts with densely packed clusters of either 32 or 64 nuclei produced by meiosis I or II, respectively, were only detected in the wild type (Fig. 6E, arrowheads) and not in the mutants (Fig. 6E′). Despite the absence of meiosis, the processes of sperm differentiation including elongation of sperm tails and compaction of the nuclear DNA into sperm heads occurred in the mutant. Nevertheless, spermatids were clearly abnormal, and the mutant sperm heads were always considerably larger and never thightly bundled as in the wild type (compare Fig. 6F and F′). Moreover, in contrast to wild type, no motile sperm was present in the seminal vesicle of mutant males (Fig. 6G and G′). In conclusion, these observations demonstrate that meiotic divisions do not occur in the mutant males.

As in the males, the premeiotic mitoses in the germ line were also not affected by the mat(2)synHB5 mutation in the female (Fig. 7). Completely normal egg chambers containing nurse cells and oocyte were seen in the mutant up to the onset of the meiotic nuclear envelope break down in stage 13 of oogenesis (see below). A single nucleus with the DNA characteristically compacted in the karyosome was present in all the oocytes (arrowheads in Fig. 7).

Surprisingly, in contrast to the mutant males, entry into the first meiotic division occurred at the correct stage of oogenesis in mutant females (Fig. 8). Whereas early stage-13 oocytes still contained a single nucleus with compacted DNA (Fig. 8A) in a clearly recognisable nuclear envelope (Fig. 8A′), entry into the first meiotic division as evidenced by chromosome appearance (Fig. 8B) and nuclear envelope breakdown (Fig. 8B′) was observed in stage-13 oocytes. Inspection at high magnification also revealed fibrillar structures in the region of the condensing chromosomes presumably indicating the formation of a meiotic spindle (see arrowheads in Fig. 8B′).

Analysis of mature stage-14 oocytes, however, clearly revealed abnormalities in the mutants. In the wild type, mature stage-14 oocytes are naturally arrested in metaphase of meiosis I. Normally, meiosis is completed only after egg activation, which occurs during egg laying. Egg activation and consequentially resumption of meiosis can be induced experimentally by incubation in hypotonic medium (Mahowald et al., 1983; Theurkauf and Hawley, 1992). An early anaphase figure of the first meiotic division which was typically observed, if mature stage-14 oocytes were fixed after a 3 minute incubation in hypotonic medium, is shown in the inset in Fig. 8E. The white arrows presumably mark the small fourth chromosome univalents which are known to be segregated even before the onset of anaphase (Puro, 1991, Theurkauf and Hawley, 1992). The other, not yet completely segregated chromosomes are found in more central positions. Such normal anaphase figures were essentially never observed in mutant stage-14 oocytes. Instead, abnormalities of varying severity were detected. Fig. 8C illustrates a mild phenotype and Fig. 8D a more severe phenotype. In the mutant oocyte shown in Fig. 8C, three discrete clusters of condensed chromosomes were detected (see arrowheads 1-3 and corresponding insets in Fig. 8C). Many mutant oocytes had even more dispersed clusters of chromatin of varying size which sometimes appeared condensed (not shown) and sometimes decondensed as in Fig. 8D.

Despite defective meiosis, mutant eggs could be fertilized with sperm as revealed by labeling with an antibody recognizing the sperm tail (not shown). As in the case of the mature stage-14 oocytes, freshly laid eggs of mutant mothers displayed various degrees of abnormalities (Fig. 9A–C). Often two (Fig. 9A,B) or more (Fig. 9C) clusters of intensely labeled chromatin of either decondensed (Fig. 9A) or condensed appearance (Fig. 9B, inset) were recognized. Normal eggs with three polar bodies (arrows and upper inset in Fig. 9D) and two pronuclei (arrowhead and lower inset in Fig. 9D) as seen in freshly laid eggs from wild-type mothers were never detected among the progeny of mutant mothers.

All these observations clearly indicate that meiosis is defective in the mutant females. The defects, however, start only after entry into meiosis. Our observations suggest that the arrest at metaphase of meiosis I which normally occurs in mature stage-14 oocytes does not occur in the mutant oocytes. Instead, chromosomes become dispersed, eventually thoughout the oocyte. According to the signal intensities observed after DNA labeling, this chromatin dispersal must involve DNA replication. The observed phenotype is most easily explained by assuming that the mutant oocyte continues to progress through aberrant cell cycles during which chromosomes are replicated and segregated irregularily.

The Drosophila cdc25 phosphatase encoded at the string locus is required for entry into mitosis (Edgar and O’Farrell, 1989), and the regulated expression of string directs the intricate patterns of the embryonic cell divisions (Edgar and O’Farrell, 1990). We have identified a second Drosophila cdc25 homolog. Expression of this second homolog, twine, was only detected in the germ line starting at relatively late stages of gametogenesis, well after the end of the mitotic germ cell divisions. In both male and female germ cells, twine transcripts are present during the stages of meiosis, and our genetic analyses indicate that twine is required for meiosis.

The notion that the control of meiotic and mitotic divisions is mechanistically similar is old and based on a large body of evidence. In Xenopus oocytes, for example, entry into the meiotic divisions can be induced by injection of the active cyclin B-p34^cdc2 kinase complex. The same result is also observed after injection of string mRNA (Gautier et al., 1991). Moreover, in S. pombe, cdc25 is not only involved in mitosis but also in meiosis (Grallert and Sipiczki, 1989). The existence of a distinct, meiotic cdc25 homolog in Drosophila presumably reflects a need for additonal levels of controls required for meiosis in the context of gametogenesis in higher organisms.

The twine mRNA that accumulates during oogenesis persists in the oocyte beyond meiosis and is detectable in the early embryo until cellularization. Therefore, twine might not have an exclusively meiotic function, but could also act during the early embryonic mitoses. However, during the early embryonic stages, not only twine mRNA but also maternally derived string mRNA is present (Edgar and O’Farrell, 1989). The function of either twine or string during these stages has not yet been analyzed. It remains possible that the twine protein is inactivated after meiosis, and that the string protein is the only cdc25 phosphatase active during the early embryonic mitoses. At least in later mitotic divisions, when twine is no longer expressed, string function is sufficient for entry into mitosis. Whereas string is expressed in apparently all mitotically dividing cells in the embryo (Edgar and O’Farrell, 1989), in the larvae (J. Knoblich and C.F.L., unpublished observations) and in the adult (e.g. during the follicle cell proliferation, see Fig. 3), twine expression, with the sole exception of the early embryonic cycles, is not observed during mitotic cell cycles. Therefore, we propose that twine is specialized for meiosis and that it is not active in the early embryo.

The observation that twine is not only expressed during oogenesis but also during spermatogenesis where, in contrast to oogenesis, no stockpiling of components for early embryonic development occurs, strongly suggested that twine might function during meiosis. A mutation in twine was therefore expected to result in both female and male sterility. Our analysis of mutant fly lines, which had been isolated in a screen for female sterile mutations on the second chromosome by Schüpbach and Wieschaus (1989) and which had been mapped to the same genomic region as twine, revealed that one line, mat(2)synHB5, was not only female sterile but also male sterile. Sequence analysis of the twine gene from mat(2)synHB5 flies revealed one missense mutation in a position that is conserved in all the known cdc25 homologs. Moreover, this missense mutation in twine is in the immediate neighbourhood of amino acid residues that are known to be part of the active site in phosphotyrosyl phosphatases (Gautier et al., 1991; Guan et al., 1991; Millar et al., 1991). The results of our complementation experiments demonstrate that the function of twine which allows complementation of a mutant cdc25 allele in S. pombe is impaired by this missense mutation.

Meiosis in males homozygous for mat(2)synHB5 is completely blocked. Meiotic figures with condensed chromosomes were never observed in mutant testes indicating that twine is required for entry into meiosis. Other aspects of spermatogenesis are not affected. The development and growth of premeiotic cysts appears completely normal. Moreover, postmeiotic differentiation processes also continue despite the absence of meiosis. Sperm tails elongate and the formation of sperm heads is attempted in mutant testes, although the compaction of the premeiotic, presumably 4N nuclei into the typical rod shape does not occur to the same extent as in the case of the postmeiotic 1N nuclei in wild-type testes.

twine expression starts during the growth phase of cysts many hours before the onset of the meiotic divisions. string expression on the other hand is not observed during the meiotic stages. If this does not simply reflect an inability to detect transient, low level expression, it indicates that entry into meiosis is controlled by a mechanism different from that controlling entry into mitosis in the cellularized embryo, where the transcriptional control of string expression is thought to determine the time of entry into mitosis (for a review see Lehner, 1991). During the embryonic cell division cycles, string expression is sufficient to force G2-cells into mitosis, and starts immediately (about 25 minutes) before mitosis in an intricate pattern which accurately anticipates the pattern of the subsequent division (Edgar and O’Farrell, 1990).

As in testes from homozygous mat(2)synHB5 males, mitotic divisions of germ line cells were completely normal in mutant females and defects were only observed during meiosis. However, whereas entry into meiosis appears to be completely blocked in mutant males, it is still accomplished in mutant females. Nuclear envelope breakdown and chromosome condensation occur at the correct stage in mutant oocytes. But, in contrast to wild-type oocytes, mutant oocytes do not arrest at metaphase of the first meiotic division. Instead, chromosomes are replicated and dispersed into irregular nuclei of variable size. Such irregular nuclei were never observed in mutant testes. The extent of this chromatin dispersal in mature mutant oocytes varies considerably. This variability most likely reflects the fact that mature stage-14 oocytes can be retained for several hours before fertilization and egg deposition. We assume that the severity of the phenotype increases with increasing retention time, which is affected by feeding of the adult females. Our observation that the proportion of eggs with few nuclei is higher in collections from well fed females than in collections from starved females is consistent with this assumption (C.F.L., unpublished observation).

Our analysis of string expression suggests an attractive explanation for the phenotypic differences observed in males and females. Whereas we have been unable to detect string transcripts during the meiotic stages of spermatogenesis, string transcripts are clearly present in oocytes undergoing meiosis. Normal levels of string transcripts were also detected in mutant oocytes (data not shown). It is therefore possible that the string activity allows entry into meiosis in mutant oocytes. The results of experiments in S. pombe clearly indicate that string and twine have similar activities since both can complement cdc25 alleles (Edgar and O’Farrell, 1989; Jimenez et al., 1990; this paper). Moreover, the presence of string activity in mutant oocytes might not only allow entry into meiosis but also further cell cycle progression. Our observations, however, showed that the oocyte nucleus does not proceed through regular mitotic cycles but is fragmented rapidly in the mutant oocytes. While an ordered cell cycle progression might be hampered for a variety of reasons, we would like to point out that the centrosome which might well be required for an ordered mitotic cycle is missing in late oocytes and is contributed only during fertilization by the sperm (Mahowald and Kambysellis, 1980).

According to our interpretation, twine would act formally analogous to CSF (cytostatic factor), an activity that causes the arrest at metaphase II during Xenopus egg maturation (Masui and Markert, 1971; for recent references on CSF see Hunt, 1992). Mechanistically, CSF acts by stabilizing the activity of MPF (maturation promoting factor), a factor which according to recent results represents the active complex of cyclin B and the p34^cdc2 kinase. In Drosophila oocytes, MPF activity might be stabilized by twine activity and cause the arrest at metaphase of meiosis I until twine is inactivated after egg activation. Testing this idea will require a biochemical analysis of string and twine activity during female meiosis. Fortunately, recent progress in the development of methods allowing mass isolation and in vitro activation of mature oocytes should render such investigations feasible. In addition, such studies might also reveal which of the various cyclin-cdc2 (or cdc2-like) kinases is a target of twine activity.

We would like to thank T. Schüpbach and M. Ashburner for providing fly strains, L. Alphey and D. Glover for communicating results prior to publication, and T. Karr for the gift of his monoclonal antibody recognizing the sperm tail. This work would not have been possible without expert technical help from E.-M. Illgen and M. Langegger. We thank W. Seufert and the members of the lab for their comments on the manuscript.