Although arginine modification has been implicated in a number of cellular processes, the in vivo requirement of protein arginine methyltransferases(PRMTs) in specific biological processes remain to be clarified. In this study we characterize the Drosophila PRMT Capsuléen, homologous to human PRMT5. During Drosophila oogenesis, catalytic activity of Capsuléen is necessary for both the assembly of the nuage surrounding nurse cell nuclei and the formation of the pole plasm at the posterior end of the oocyte. In particular, we show that the nuage and pole plasm localization of Tudor, an essential component for germ cell formation, are abolished in csul mutant germ cells. We identify the spliceosomal Sm proteins as in vivo substrates of Capsuléen and demonstrate that Capsuléen,together with its associated protein Valois, is essential for the synthesis of symmetric di-methylated arginyl residues in Sm proteins. Finally, we show that Tudor can be targeted to the nuage in the absence of Sm methylation by Capsuléen, indicating that Tudor localization and Sm methylation are separate processes. Our results thus reveal the role of a PRMT in protein localization in germ cells.
INTRODUCTION
Establishment and maintenance of the germline is an essential process for all sexually reproducing organisms. Germ cells, which carry the genetic information to the next generation, are simultaneously totipotent and highly specialized (Wylie, 1999). All germ cells throughout the animal kingdom contain in their cytoplasm a distinct cloud-like structure termed the nuage(Eddy, 1975).
One system with a high potential for understanding the assembly and role of the nuage is Drosophila oogenesis. The Drosophila egg chamber consists of a germ line cyst generated from a single cystoblast by four successive mitotic divisions and surrounded by a monolayer of somatic follicle cells. Due to incomplete cytokinesis, the germ cells remain connected to each other through specialized cytoplasmic bridges. The oocyte derives from one of the germ cells, while the remaining 15 cells differentiate into nurse cells (Spradling, 1993)(Fig. 2A). The nuage is concentrated in the perinuclear cytoplasm of nurse cells and can be associated with nuclear pores. It appears as a dense fibrous organelle unbound by membrane and often associated with mitochondrial clusters(Mahowald, 1971). Moreover,the nuage interfaces with sponge bodies, which are abundant RNA-rich particles present in the cytoplasm of nurse cells and, to a lesser degree, in the oocyte(Wilsch-Bräuninger et al.,1997).
The majority of the components identified in the nuage are also present in the pole plasm of the oocyte, and more particularly in the polar granules. The pole plasm constitutes the determinant that is both necessary and sufficient to induce germ cell formation during early embryogenesis(Illmensee and Mahowald,1974). The first step in pole plasm formation is the transport of oskar (osk) transcripts synthesized in nurse cell nuclei to the posterior pole of stage 8 oocytes(Ephrussi et al., 1991; Kim-Ha et al., 1991). At this location Osk is synthesized and serves as an anchor to initiate polar granules assembly (Ephrussi and Lehmann,1992; Smith et al.,1992; Snee and Macdonald,2004). In addition to Osk the polar granule components include Vasa (Vas) (Hay et al., 1988; Lasko and Ashburner, 1990),which interacts directly with Osk(Breitwieser et al., 1996),Tudor (Tud) (Bardsley et al.,1993) and a number of transiently localized factors that are mainly necessary for osk mRNA transport and translation. During late oogenesis and early embryogenesis, the polar granules are maintained at the posterior pole. At the time of blastoderm formation, they are sequestered in the pole cells, the primordial germ cells of the fly, in which they coalesce into a smaller number of large particles(Mahowald, 1968), ultimately disappear and are replaced by the nuage(Mahowald, 1971). This structure appears to evolve from components of the pole plasm and persists only in established germ cells. The only identified nuage-specific component absent from the polar granules is Maelstrom (Mael) which shuttles between the nucleus and the cytoplasm (Findley et al.,2003).
Among these proteins, Vas plays a cardinal role in the formation of the nuage (Findley et al., 2003). In vas ovaries the nurse cells are devoid of nuage at the ultrastructural level (Liang et al.,1994). The function of Tud in the nuage remains unknown but Tud may play a role in the assembly or modification of specific RNP complexes, as indicated by its requirement for the transfer of mitochondrial ribosomal RNAs from the mitochondria to the polar granules(Amikura et al., 2001). The presence of shared components reinforces the view that the nuage and the polar granules are closely related structures, in which components, such as Vas and Aubergine (Aub), may dissociate from the nuage to reassemble into the polar granules (Snee and Macdonald,2004).
The mechanisms by which nuage components become assembled at the perinuclear region of the nurse cells remains, however, to be identified. Here we report that the catalytic activity of the Capsuléen (Csul)protein-arginine methyltransferase is required for the localization of specific components of the nuage and pole plasm, and in particular of Tud.
MATERIALS AND METHODS
Molecular biology
Plasmid constructs were generated by PCR amplification of the relevant DNA segments (High Fidelity PRC Master; Roche), which were subcloned into appropriate vectors. vas, tud, nos, gcl and SmD3 cDNA plasmids were kindly provided by P. Lasko, R. Boswell, R. Lehmann, T. Jongens,and H. Schenkel, respectively. The plasmids pBS1479 (C-TAP) and pBS1761 (NTAP)were obtained from Cellzome (Heidelberg). SmB and SmD1 cDNAs(BDGP EST project clones LD14049 and RE39488, respectively) were provided by the Drosophila Genomics Resource Center (DGRC), Bloomington, IN, USA. Interstitial deletions in csul were prepared using a QuickChange site-directed mutagenesis kit (Stratagene) with the modifications of Makarova et al. (Makarova et al.,2000). Details of primers and cloning strategies will be provided upon request.
GST pull-down assay
Fragments of tud and SmB cDNAs were cloned into pCITE-4(Novagen). Tud and SmB polypeptides were synthesized using the TNT Coupled Reticulocyte Lysate System (Promega) in the presence of unlabeled amino acids. Full-length SmB proteins were synthesized from the LD14049 cDNA plasmid in the presence of [35S]methionine (Amersham). To cotranslationally inhibit sDMA synthesis, 0.3 mM S-adenosylhomocysteine(Sigma) was included in the reaction. GST pull-down assays were performed as previously described (Anne and Mechler,2005).
Immunocytochemistry
A GST-Csul fusion protein construct containing a 442-amino-acid peptide corresponding to amino acid residues 168-609 of Csul was prepared by cloning the BamHI-XhoI fragment of the csul cDNA in the pGEX-4T2 GST fusion vector (Pharmacia) and expressed in the bacterial strain BL21(DE3). Anti-Csul polyclonal antibodies raised in rabbits against the GST-Csul fusion protein were purified by affinity chromatography on a protein Aagarose column (Roche) and preadsorbed on proteins extracted from Escherichia coli.
For immunostaining, primary antibodies were anti-Osk from rabbit, anti-Vas from rabbit and rat (gifts from P. Lasko and A. Nakamura, respectively),anti-Tud from rabbit (TUD56; a gift from S. Kobayashi), anti-Mael from rabbit(a gift from S. Findley), anti-Nos from rat (a gift from R. Wharton), mouse monoclonal anti-HA (clone 16B12; BAbCO), anti-Sm (clone Y12; NeoMarkers), and anti-dimethyl-arginine (SYM10 and ASYM24) from rabbit (Upstate). DNA was visualized by staining with Oli-Green (Molecular Probes). Biotinylated and Cy3-conjugated secondary antibodies (Jackson ImmunoResearch Laboratories) were used at 1:200.
RESULTS
Maternal-effect phenotype of csul
The csulP mutation was identified in a screen for maternal-effect mutations induced by a P-[white+]transposon. The P-element insertion was mapped to 52F on chromosome 2(data not shown). Following P-element excision, we recovered numerous(92/150) fertile white flies producing fully viable and fertile progeny, indicating that the P-insert was the cause of the maternal-effect phenotype (see below). Moreover, we isolated a series of viable white females producing embryos similar to those laid by csulP females. These alleles resulted from a partial excision of the P-[white+] transposon and are named csulRM.
Homozygous csulP females produce embryos(csulP embryos) displaying defects in abdominal patterning and lacking pole cells (Fig. 1B). Flies hatching from csulP larvae are fully agametic. The severity of the abdominal defects is variable, ranging from the presence of only one to two abdominal segments(Fig. 1D) to a fully normal abdomen. On the basis of the absence of pole cells and the occurrence of abdominal defects, we classified csulP as belonging to the`posterior-grandchildless' class of mutations(Schüpbach and Wieschaus,1986).
To determine whether csul acts in the posterior-grandchildless pathway, we investigated two gene transcripts, nanos (nos),the posterior determinant (Wang and Lehmann, 1991), and germ cell-less (gcl),specifically involved in pole cell formation(Jongens et al., 1992). The posterior localization of their transcripts in preblastoderm embryos depends on this pathway. In one half of csulP embryos, nos mRNA was absent or barely detectable (data not shown), whereas in the other half it was correctly localized, albeit in much lower amounts than in wild-type embryos (Fig. 1E,F). Analysis of the Nos protein revealed a Nos gradient emanating from the posterior pole in only 15 percent of csulP embryos, whereas Nos remains undetected in the other embryos (data not shown). In situ hybridization showed that gcl mRNA is evenly distributed in csulP embryos instead of being localized to the posterior pole (Fig. 1G,H). From these data we conclude that csul is required for the posterior localization of at least two transcripts, nos and gcl.
The posterior-grandchildless group comprises a relatively small number of genes hierarchically organized with osk orchestrating the activity of the other genes. As ectopic expression of osk induces pole cell and abdomen formation at the anterior of the embryo(Ephrussi and Lehmann, 1992),we tested whether expression of an osk-bcd3′UTRtransgene would induce pole cells formation in csulP(Fig. 1I-L). No pole cell could be detected in csulP; osk-bcd3′UTRblastoderm embryos. Cuticle examination revealed the absence of duplicated abdominal structures and a marked atrophy of the head in the majority of the larvae, indicating a downregulation of osk activity. We therefore suggest that csul is essential for pole plasm assembly and acts downstream of osk.
Distribution of pole plasm components in early csulembryos
We then determined whether csul activity is needed for the localization of Osk and Vas (Fig. 1M-T). We found that Osk and Vas are present at the posterior pole of early syncytial csulP embryos, albeit at a reduced level for Vas. However, starting from nuclear cycle 6, Osk and Vas signals significantly decrease and are nearly completely absent from the posterior pole of late syncytial csulP embryos. In a few pre-blastoderm embryos we detected Osk- and Vas-stained particles closely associated with nuclei at the posterior pole (inserts in Fig. 1P,T). These structures may represent remnants of polar granules, similar to those detected in hypomorphic tud mutant embryos(Hay et al., 1990). In contrast to Osk and Vas, no posterior accumulation of Tud could be detected in csulP, even in very early embryos(Fig. 1V). From these data we conclude that csul activity mediates Tud and (to some degree) Vas accumulation and the maintenance of Osk and Vas at the posterior pole of the embryo.
csul-dependent pole plasm assembly during oogenesis
As pole plasm assembly occurs during oogenesis we investigated the distribution of pole plasm components in csulP egg chambers. osk mRNA and Osk protein distributions appear normal(Fig. 2A,B; data not shown) and western blot analysis of ovarian protein extracts revealed similar levels of Osk proteins and phosphorylation of the short Osk isoform in csulP and wild-type egg chambers (data not shown). We then used a GFP-Vas transgene (Sano et al., 2002) to monitor the distribution of Vas. GFP-Vas is absent from the pole plasm in the majority of stage 10 csulRM egg chambers (data not shown). Only a minority of csulRMoocytes display a small amount of GFP-Vas at the posterior pole(Fig. 2D). Immunostaining of csulP egg chambers using anti-Vas antibodies showed a similar distribution (data not shown). By comparison with early embryos, the smaller number of stage 10 oocytes displaying a posterior localization of Vas indicates that Vas may accumulate in the pole plasm during late csulRM oogenesis to become detectable during early embryogenesis.
By contrast, we found that Tud fails to concentrate at the posterior pole in csulP oocyte of stage 10 egg chambers(Fig. 2F) indicating that csul activity is essentially required during oogenesis for the localization of Tud, and to a lesser degree for that of Vas, in the pole plasm.
csul-dependent nuage assembly
As a number of pole plasm components localize in the nuage, we analyzed their pattern of distribution in csul egg chambers. GFP-Vas shows a normal, albeit weak, nuage localization in csulRM nurse cells during early oogenesis (Fig. 2H). Subsequently, GFP-Vas progressively fades away and is barely detected in the nuage of stage 10 egg chambers(Fig. 2D). By contrast, no Tud is detected in the nuage in csulP egg chambers, although Tud accumulates in mutant oocytes (Fig. 2J) and transiently localizes at their anterior margin in stage 6/7 egg chambers (data not shown). In wild-type nurse cells, Mael displays a punctate perinuclear distribution (Fig. 2K), which is not detected in csulP egg chambers (Fig. 2L). Hence, csul is needed for the perinuclear localization of Mael, Tud and Vas,indicating that csul contributes to the assembly of the nuage.
Molecular characterization of the csul gene
To gain insight into csul function, we isolated the gene by inverse PCR amplification of a genomic DNA fragment adjacent to the P-[white+] insertion site. This fragment was used to isolate DNA segments overlapping the P-element which in turn were used to recover cDNA clones. Northern blot analysis using genomic DNA probes from both sides of the P-insert revealed transcripts of ∼2kb in size. Alignment of the cDNA sequences with genomic DNA revealed two types of transcripts oriented in opposite directions. The P-element was inserted in one of the transcription units, 51 bp downstream of its 5′end and 7 nucleotides downstream of the putative translation initiation site. This transcript was assigned to the csul gene, extending from the P-insert down to the adjacent Kinesin heavy chain(Khc) gene. The 3′ end of Khc is located 5 nucleotides downstream of the 3′ end of csul. To validate the identification of the csul gene, we generated a transgene and tested its ability to complement csul mutations. We found that the fertility of csulRM females was restored by a single copy of the XhoI-XhoI P-[csul] transgene(Fig. 3A). Constructs containing either the downstream Khc(Saxton et al., 1991) or the upstream fidipidine (data not shown) genes failed to restore fertility to csulRM females.
Alignment of cDNA and genomic DNA sequences revealed that csul is composed of five exons (Fig. 3B). The csul transcript has a size of about 2 kb(Fig. 3C) and potentially encodes a 610 aminoacid protein (accession number: AJ002740). In vitro translation of a full-length cDNA produced a single polypeptide with an apparent molecular mass of ∼67 kDa corresponding to the predicted molecular mass of Csul (data not shown). Polyclonal antibodies were raised against the carboxyl-terminal half of bacterially synthesized Csul. On western blots of ovarian and embryonic protein extracts, these antibodies reacted with a protein band displaying a molecular mass of ∼65 kDa, consistent with the expected mass of Csul (Fig. 3D). No such protein was detected in ovarian extracts of csulP females. However, long exposure of the immunoblot revealed a faint band in the 65 kDa range, suggesting that csulP females may produce a low level of Csul, possibly by use of an in-frame initiation codon present in the P-white+ inverted repeat and located 16 nucleotides upstream of the insertion site (data not shown). This finding suggests that csulP is a hypomorphic allele.
Csul acts as a protein methyltransferase
Csul protein sequence was found to exhibit significant similarity to protein-arginine methyltransferases (PRMTs), which catalyse the formation of symmetric di-methyl arginyl (sDMA) residues(Fig. 3E). Csul homologues can be identified in organisms ranging from the budding yeast Saccharomyces cerevisiae to humans (Ma,2000).
To ascertain that Csul acts as a methyltransferase, we investigated whether amino acid substitutions in the catalytic domain of the enzyme would inactivate its function. Csul family proteins contain a conserved core region characterized by motif I [GXGRG], which, together with the Post-I motif, forms the S-adenosyl-L-methionine binding module(Ma, 2000). As the GXGRG motif is highly conserved in the Csul homologs and is required for protein methylation (Pollack et al.,1999) we selected this region for mutational analysis,substituting the conserved residues Gly343 and Arg344with Ala and Leu residues, respectively.
Wild-type csul and csulG343A;R344L gene sequences were fused to the tandem affinity purification (TAP)(Rigaut et al., 1999) tag,cloned into a transformation vector and transgenic strains were generated. After introducing the transgenes in a csulRM genetic background we determined, by western blot, that they both express at similar level (data not shown) and examined their ability to abolish the csulRM maternal-effect embryonic phenotype. csul-TAP and TAP-csul transgenes rescue the development of csulRM embryos, whereas the csulG343A;R344L-TAP transgene shows no rescue activity(data not shown and Fig. 7D). These results strongly suggest that csul encodes a methyltransferase whose activity requires an intact catalytic domain.
Sm proteins are substrates of the Csul-Vls complex
We used anti-SYM10 antibodies (α-SYM10), which specifically recognize proteins containing multiple sDMA-glycine repeats(Boisvert et al., 2002), to investigate the pattern of methylated proteins extracted from wild-type, csul, valois (vls), vas, aub and oskovaries. As shown in Fig. 4A,α-SYM10 reacted with five protein bands of a relatively low molecular mass, ranging from 14 to 30 kDa, in wild-type, osk, vas and aub protein extracts. In csul and vls the intensity of four of these bands was strongly reduced. Probing similar protein blots with the anti-ASYM24 antibodies, which identify proteins containing asymmetric di-methyl arginyl residues (aDMA) (Boisvert et al., 2002), we found no change in the pattern of the reactive protein bands in all extracts. These results corroborate our finding that Csul and Vls are part of a methylosome complex contributing to sDMA synthesis in specific proteins (Anne and Mechler,2005), and indicate that Vls acts as a co-factor of Csul. Moreover, these data show that Vas, Osk and Aub exert no function in the Csul-dependent synthesis of sDMA residues.
As the size of the Drosophila proteins recognized by α-SYM10 corresponds to that of components required for pre-mRNA splicing in human cells (Boisvert et al., 2003),we analyzed whether Sm proteins might be in vivo substrates of Csul. As shown in Fig. 4B,C, making use of available mutations in Drosophila genes encoding Sm proteins, we identified SmB and SmD3 as targets of the Csul-Vls complex. In particular we found that the amount of an ∼26 kDa α-SYM10 reactive protein was reduced by half in western blots of heterozygous SmBBG02775 in comparison to wild-type ovarian extracts. A similar reduction in intensity was obtained with the Y12 monoclonal antibody(α-Y12) that specifically recognizes methylated SmB(Paterson et al., 1991). Similarly, comparison of protein extracts of wild-type and homozygous SmD3l(2)k131-07 larvae(Schenkel et al., 2002)revealed that the α-SYM10 signal of a 16 kDa protein is nearly abolished in the mutant, identifying this protein as SmD3.
To further confirm the Csul-dependent symmetrical methylation of SmB, we stained ovaries using α-Y12. As compared with wild-type ovaries, in which methylated SmB accumulates in the nuclei of the nurse cells and somatic follicular cells, α-Y12 signal was dramatically reduced in csulRM egg chambers(Fig. 4D). Taken together, our data demonstrate that Csul is required for the symmetrical methylation of SmB in vivo.
Csul and Vls interact with Sm proteins in vitro
As Sm proteins can be methylated by the Csul-Vls methylosome, we investigated by binding assays whether these proteins can physically interact together. We found that Csul and Vls can bind to SmB(Fig. 5), SmD1, and SmD3 (data not shown). We then determined the reciprocal binding domains between Csul and SmB. First we synthesized 35S-labeled S-tag SmB1-199(full-length), SmB1-106 (N-terminal moiety with Sm domains) and SmB83-199 (C-terminal moiety with RG repeats) polypeptides and incubated them with GST-Csul, GST-Vls, or GST alone. We found that the deletion of the RG domain had no effect on SmB binding to Csul or Vls,indicating that the N-terminal moiety containing the Sm domains mediates SmB interaction with Csul or Vls. Then, by using a similar approach we assigned the SmB binding domain of Csul to the N-terminal part (amino acid residues 1-200) (Fig. 6). Further analysis showed that both N- or C-terminal truncations in this domain(fragments 20-610 and 1-111, respectively) prevent SmB binding to Csul(Fig. 6 and data not shown),indicating that the N-terminal part of Csul contains at least two domains necessary for interaction with SmB.
Interaction between Csul and Tud in vivo is essential for Tud localization
The defective accumulation of Tud in the nuage of csulPnurse cells led us to evaluate whether Tud might directly interact with Csul. As shown in Fig. S1 in the supplementary material, we performed pull down assays, using tagged fragments of Tud(Golumbeski et al., 1991; Anne and Mechler, 2005). In vitro translated fragments of Tud were incubated with immobilized GST-Csul proteins. After washing, the bound S*Tag-Tud proteins were revealed by immunodetection. This procedure showed that the JOZ fragment and the 9A1 polypeptide and derivatives display a strong binding, whereas the 3ZS+L fragments exhibit more moderate binding to GST-Csul whereas GST alone is unable to bind to Tud fragments. We then used Tud9A1-N to map the binding site of Tud in Csul.
As shown in Fig. 7A, we found that the 9A1-N fragment of Tud can bind to the N-terminal region of Csul(amino acid residues 1-111). These results suggest that a domain in the N-terminal region of Csul mediates a direct and specific interaction with Tud. To delimit more precisely the N-terminal domain of Csul critical for Tud binding, we generated serial N-terminal truncations of Csul. In this way we identified a critical region for Tud binding between amino acids 60 and 80(Fig. 7B,C).
We next tested the in vivo relevance of the N-terminal region of Csul with respect to the localization of Tud in the nuage and the formation of pole cells. For this purpose we constructed four csul transgenes each bearing an internal deletion (between amino acids 21-40, 41-60, 61-80 and 81-100, respectively), fused in-frame to the TAP tag, and introduced these transgenes into the csulRM genetic background. Immunostaining revealed that the Csul proteins carrying the deletionsΔ21-40 and Δ41-60 could restore Tud localization in the nuage. By contrast, deletions Δ61-80 and Δ81-100 failed to do so(Fig. 7D, left panels). Finally we observed that deletions Δ21-40 and Δ41-60 can give rise to the formation of pole cells and rescue the csulRMmaternal-effect phenotype, whereas deletions Δ61-80 and Δ81-100 do not (Fig. 7D, right panels). Together with our in vitro interaction results, these findings indicate that the localization of Tud in the nuage requires interaction with Csul.
sDMA-Sm proteins can bind to Tud in vitro
As sDMA Sm proteins can bind to the Tudor domain of SMN(Brahms et al., 2001; Selenko et al., 2001) we investigated the requirement of sDMA synthesis for Sm binding to Tud. For this purpose we took advantage of the occurrence of an endogenous type II PRMT activity in the reticulocyte lysate system(Brahms et al., 2001). By using S-adenosyl-homocysteine (SAH), which blocks the activity of protein methyltransferases (Brahms et al.,2001), we tested whether SAH could inhibit sDMA synthesis in SmB. Addition of SAH to the reticulocyte lysate significantly reduced the amount ofα-SYM10-immunoreactive SmB polypetides, but did not inhibit SmB synthesis (see Fig. S2A in the supplementary material). We next incubated bacterially produced Tud fragments (see Fig. S2B in the supplementary material) with SmB polypeptides synthesized in the presence or absence of SAH. As shown in Fig. S2C in the supplementary material, methylation of SmB promotes its binding to all three tested Tud fragments.
Tud localization and Sm protein methylation may be unrelated processes, both driven by Csul
To further explore the relationship between Sm methylation and pole plasm function, we investigated the methylation status of Sm proteins in extracts of transgenic csulRM ovaries producing Csul proteins with interstitial deletions in the N-terminal region. Western blot analysis usingα-SYM10 revealed that all four mutant Csul proteins failed to symmetrically methylate SmB (Fig. 7E). The finding that the Csul proteins containing the two most distal deletions (Δ21-40 and Δ41-60) are still able to restore Tud targeting to the nuage of csulRM egg chambers indicates that only the methylation of Sm proteins is impaired. In light of the in vitro binding results (Fig. 6) we envisage that the absence of methylation of Sm proteins is primarily due to a defective binding of SmB to Csul. Whether the two other deletions(Δ61-80 and Δ81-100) only affect the binding of Tud to Csul or structurally compromise other functions of Csul remains an open question but primarily our data show that we can separate Tud localization in the nuage from Sm methylation.
DISCUSSION
Csul is a Type II PRMT
csul encodes a Type II PRMT, which transfers methyl groups from S-adenosyl-L-methionine to the guanidinium group of arginyl residues. PRMTs can be divided into two major categories, catalyzing the synthesis of aDMA (Type I) or sDMA (Type II) residues, respectively. The mammalian PRMT5(Pollack et al., 1999; Lee et al., 2000; Branscombe et al., 2001),homologous to Csul, and the recently identified PMRT7 and PRMT9(Lee et al., 2005; Cook et al., 2006) are responsible for Type II methylation.
By using α-SYM10 antibodies that recognize proteins harbouring two spaced sDMA-glycine motifs (Boisvert et al., 2002) we identified four major reactive proteins bands as specific targets of Csul. These proteins are distinct from aDMA-containing proteins, whose methylation is independent of Csul. Among the sDMA proteins,we genetically confirmed that the spliceosomal components SmB and SmD3 are Csul targets. The corresponding mammalian targets have been identified for PMRT5 (Brahms et al., 2001; Meister et al., 2001; Friesen et al., 2001a). Asα-SYM10 may only recognize a subset of sDMA proteins methylated by Csul,further proteomic analysis of ovarian Csul complexes may identify additional targets of Csul.
The Csul-Vls and human methylosomes interact with sDMA-proteins and Tudor-domain proteins
As indicated by the physical interaction of Csul with Vls(Anne and Mechler, 2005) and the size of the native Csul complexes, with a molecular mass of ∼500 kDa(data not shown), Csul is part of a large protein complex. In the present work we show that Vls, the Drosophila homolog of human MEP50, itself a partner of PRMT5 (Friesen et al.,2002), is also required in sDMA synthesis on identical target proteins. However, in the case of pIcln, a component of the human methylosome of yet unknown function (Friesen et al.,2001a), we detected no interaction between DrosophilapIcln and Csul in pull-down assays (data not shown). Furthermore, we found that both Csul and Vls interact with the N-terminal moiety of SmB. This is in contrast to PRMT5, which appears to bind to the RG-rich C-terminal domain of Sm proteins (Friesen et al.,2001a). Differences in protein interaction and quaternary structure between the human and Drosophila methylosome may reflect divergences in the activities of the methylosome between the two species.
Both human (Friesen et al.,2001a; Meister et al.,2001) and Drosophila methylosomes lead to sDMA synthesis on Sm proteins (this study). Similarly to the requirement of sDMA synthesis for the association of human Sm proteins with the SMN Tudor domain(Côté and Richard,2005), we found that Drosophila Sm proteins need to be symmetrically methylated to bind Drosophila Tud. The binding of sDMA-Sm to non-overlapping Tud polypeptides indicates that these proteins may bind to several, if not all Tudor domains in Tud.
The association of human SMN protein with the PMRT5 complex suggests direct interactions between PMRT5, MEP50 and SMN(Meister and Fisher, 2002). Similarly, Drosophila Tud can directly bind to Csul and Vls(Anne and Mechler, 2005). However, in contrast to Sm, which binds to multiple sites on Tud, Csul and Vls more strongly interact with the N-terminal than the C-terminal moiety of Tud,suggesting a distinct mechanism of association with Tud. Although the specific binding sites of Csul and Vls on Tud remain to be determined, preliminary results indicate that each protein binds to a distinct site.
As this work was being completed, another group reported the identification of the csul gene, termed dart5(Gonsalvez et al., 2006), and showed that disruption of this gene (mutant e00797 from the Exelixis collection) leads to the absence of sDMA synthesis of spliceosomal Sm proteins without impairing spliceosomal function. This work and ours confirm the previous finding of Khusial et al.(Khusial et al., 2005),indicating that sDMA synthesis on Sm proteins is not required for sRNP assembly and transport, a critical process for Drosophiladevelopment. In addition, Gonsalvez et al.(Gonsalvez et al., 2006) also characterized the maternal requirement of csul for pole cell formation.
Tud localization in the nuage
In addition to their role in sDMA synthesis, Csul and Vls are required for Tud localization in the nuage. Our data indicate that csul activity is also necessary for the proper nuage accumulation of Vas. However, despite the occurrence of Vas in the nuage of early csul egg chambers, Tud is absent from this structure, suggesting that the activity of Csul in Tud localization is independent from that exerted on Vas.
How the Csul/Vls methylosome directs Tud localization in the nuage remains an open question. The restoration of fertility by mutated csultransgenes defective in SmB binding, and hence in sDMA synthesis on SmB,points out the occurrence of a yet unknown protein which should act as a substrate of Csul and specifically function in germline formation. We favour a cytoplasmic association of the Csul/Vls methylosome with this substrate and Tud. Upon methylation the substrate is then transferred to Tud, as indicated by the preferential binding of sDMA-SmB to Tud polypeptides. In our view, the interaction between Csul/Vls, the substrate, and Tud may be critical to position Tud in the vicinity of the site where sDMA synthesis takes place,thus facilitating the association of Tud with the sDMA-protein. A similar model has been proposed for the targeting of high-affinity Sm protein substrates to the SMN complex (Friesen et al., 2001b). Following the docking of the sDMA protein on Tud, the Csul/Vls methylosome is released, and the Tud/sDMA protein complex becomes positioned in the nuage. The docking of the sDMA protein might induce an allosteric change in Tud, increasing its affinity for a component of the nuage.
Finally, although Vas is not properly localized at the perinuclear region of nurse cells in csul and vls mutant egg chambers we notice that its distribution pattern differs in each mutant. In particular, the level of Vas in the nuage is comparatively smaller in csul than in vls mutants (Anne and Mechler,2005), suggesting that Csul acts independently of Vls in the localization of Vas to the nuage. Moreover, the finding that Vls specifically accumulates in the nuage and pole plasm whereas Csul displays a ubiquitous distribution (data not shown) suggests that both proteins may exert additional independent functions.
Although the functional relationship between the nuage and pole plasm remains unresolved, events occurring in the nuage may affect pole plasm formation. In csul mutant egg chambers, Tud is absent from both the nuage and the pole plasm and, similarly, a reduced amount of Vas in the nuage correlates with a decreased level of this protein in the pole plasm. However,it has been reported recently that a Tud protein containing the Tudor domains 1 and 6-10 could localize to the pole plasm, albeit at a moderate level compared to full-length Tud, but fail to properly localize to the nuage(Arkov et al., 2006). Additional work on the requirement of Csul for Tud localization in the nuage will be critical for understanding the assembly of this structure, its dynamical relationship with the pole plasm, and the role of arginine methylation in protein targeting.
Acknowledgements
J.A. dedicates this work to the memory of his grandparents, Andréand Henriette Costentin. We are particularly grateful to Jean-RenéMartin, Susan Sather, and Anna Raibaud in the initial phase of this work and to Seth Findley, Thomas Jongens, Satoru Kobayashi, Paul Lasko, Ruth Lehmann,Peter Tolias and Robin Wharton for providing reagents. We would like to thank Christof Niehrs and the members of the DKFZ group, Gunter Merdes, Istvan Török, Joachim Marhold, Dennis Strand, Rolf Schmitt, and Ioannis“PAOK Olé” Illiopoulos, for their constant help. This study was supported by grants to B.M.M. from the European Commission (BMH1-CT94-1572 and QLRI-CT2000-00915).