The septins are a family of proteins involved in cytokinesis and other aspects of cell-cortex organization. In a two-hybrid screen designed to identify septin-interacting proteins in Drosophila, we isolated several genes, including homologues (Dmuba2 and Dmubc9) of yeast UBA2 and UBC9. Yeast Uba2p and Ubc9p are involved in the activation and conjugation, respectively, of the ubiquitin-like protein Smt3p/SUMO, which becomes conjugated to a variety of proteins through this pathway. Uba2p functions together with a second protein, Aos1p. We also cloned and characterized the Drosophila homologues of AOS1(Dmaos1) and SMT3 (Dmsmt3). Our biochemical data suggest that DmUba2/DmAos1 and DmUbc9 indeed act as activating and conjugating enzymes for DmSmt3, implying that this protein-conjugation pathway is well conserved in Drosophila. Immunofluorescence studies showed that DmUba2 shuttles between the embryonic cortex and nuclei during the syncytial blastoderm stage. In older embryos, DmUba2 and DmSmt3 are both concentrated in the nuclei during interphase but dispersed throughout the cells during mitosis, with DmSmt3 also enriched on the chromosomes during mitosis. These data suggest that DmSmt3 could modify target proteins both inside and outside the nuclei. We did not observe any concentration of DmUba2 at sites where the septins are concentrated, and we could not detect DmSmt3 modification of the three Drosophila septins tested. However, we did observe DmSmt3 localization to the midbody during cytokinesis both in tissue-culture cells and in embryonic mitotic domains, suggesting that DmSmt3 modification of septins and/or other midzone proteins occurs during cytokinesis in Drosophila.
The septins are a family of filament-forming, GTP-binding proteins that were first discovered in yeast but are now known to be widely distributed and perhaps ubiquitous in the fungi and animals(Field and Kellogg, 1999;Kartmann and Roth, 2001;Longtine et al., 1996;Longtine and Pringle, 1999;Momany et al., 2001;Trimble, 1999). In some cell types (Hartwell, 1971;Kinoshita et al., 1997;Neufeld and Rubin, 1994),although apparently not in all (Adam et al., 2000; Longtine et al.,1996; Nguyen et al.,2000), septins are essential for cytokinesis. In addition, the yeast septins play a variety of other roles in cell surface organization; in many cases, the septins appear to serve as a scaffold or template for other proteins (Barral et al., 2000;Field and Kellogg, 1999;Longtine and Pringle, 1999;Longtine et al., 2000;Takizawa et al., 2000). Genetic, biochemical, and protein-localization data suggest that the septins probably have important non-cytokinesis roles in animal cells as well(Field and Kellogg, 1999;Kartmann and Roth, 2001;Kinoshita et al., 2000;Larisch et al., 2000;Longtine et al., 1996). Drosophila melanogaster has at least five septins, named Pnut, Sep1,Sep2, Sep4 and Sep5, whose functions are not yet well understood(Adam et al., 2000;Fares et al., 1995;Field et al., 1996;Longtine et al., 1996;Neufeld and Rubin, 1994). Progress will presumably depend, in part, on identifying the proteins with which the septins interact.
Here we used the yeast two-hybrid system to identify proteins that interact with Drosophila septins. Among the positive clones obtained were ones encoding the Drosophila homologues of yeast Uba2p and Ubc9p, which catalyze the activation and conjugation, respectively, of the ubiquitin-like protein Smt3p (Johnson and Blobel,1997; Johnson et al.,1997; Schwarz et al.,1998). Smt3p is one of several ubiquitin-like proteins recently identified and shown to become covalently linked to other proteins as a form of post-translational modification(Hochstrasser, 2000;Melchior, 2000). For example,the mammalian Smt3p homologue SUMO-1 modifies RanGAP1(Mahajan et al., 1997;Matunis et al., 1996;Matunis et al., 1998), several transcriptional regulators including p53 and c-Jun(Gostissa et al., 1999;Muller et al., 2000;Rodriguez et al., 1999), and the RING-finger protein Mdm2 (Buschmann et al., 2000). Yeast Smt3p appears to modify multiple nuclear proteins (Johnson and Blobel,1999; Meluh and Koshland,1995), consistent with the observations that Uba2p and Ubc9p localize predominantly in the nucleus(Dohmen et al., 1995;Seufert et al., 1995). Importantly, it was recently shown that several yeast septins are modified by Smt3p (Johnson and Blobel,1999; Takahashi et al.,1999) (P. Meluh, personal communication).
The functions of modification by Smt3/SUMO and other ubiquitin-like proteins are not entirely clear, but they appear to include regulation of protein localization (RanGAP1), transcriptional activity (p53 and c-Jun), and protein stability (Mdm2). Unlike ubiquitin itself(Hochstrasser, 1996), other ubiquitin-like proteins do not appear to target modified proteins for degradation and in fact may stabilize target proteins by preventing their ubiquitination (Hochstrasser,2000; Melchior,2000).
The ubiquitin-like proteins have differing and generally low levels of sequence similarity to ubiquitin itself; for example, yeast Smt3p and ubiquitin have only 17% identity in their amino acid sequences. Nonetheless,the respective conjugation pathways have many common features(Hochstrasser, 2000;Melchior, 2000). In most cases, ultimate conjugation is by an isopeptide bond between a C-terminal glycine on ubiquitin or the ubiquitin-like protein and the ϵ-amino group of a lysine in the target protein. The conjugation reaction involves an `E2'conjugating enzyme and, at least in some cases, an `E3' protein ligase involved in target-protein recognition. The conjugating enzyme receives the ubiquitin or ubiquitin-like protein from an `E1' activating enzyme, which first adenylates the C-terminus of ubiquitin or the ubiquitin-like protein and then links it by a thiolester bond to a cysteine in the E1 enzyme. The enzymatic mechanisms of the various E1 enzymes are similar, and the enzymes themselves are structurally related; this is also true of the various E2 enzymes. In the case of yeast Smt3p, activation is carried out by a heterodimer of Uba2p and a second protein, Aos1p, and links the C-terminus of Smt3p to the active site cysteine (Cys 177) in Uba2p(Johnson et al., 1997). Smt3p is then transferred and conjugated by another thiolester bond to Cys 93 in Ubc9p (Johnson and Blobel,1997; Schwarz et al.,1998). The available data suggest that the same enzymatic machinery is used for SUMO-1 conjugation in mammalian cells(Desterro et al., 1997;Desterro et al., 1999;Hochstrasser, 2000;Melchior, 2000;Okuma et al., 1999). Although E3 enzymes have not yet been identified for most ubiquitin-like proteins(Hochstrasser, 2000;Melchior, 2000), recent studies have identified two possible E3 enzymes for Smt3/SUMO modification(Johnson and Gupta, 2001;Kahyo et al., 2001;Takahashi et al., 2001).
In this study, we sought septin-interacting proteins in Drosophilaand attempted to elucidate the relationship between the septins and the Smt3 conjugation system in this organism. In addition, we used biochemical and cell biological approaches to characterize the Drosophila Smt3 conjugation pathway and investigate its possible functions.
Materials and Methods
Strains and molecular biology methods
The wild-type Canton S strain of D. melanogaster and the Clone-8 Drosophila cultured cell line(Peel and Milner, 1992) were used. S. cerevisiae strains EGY48R (α his3 trp1 ura3-52 leu2::pLEU2-lexAop6 [pSH18-34])(DeMarini et al., 1997) and JD90-1A (α his3 leu2 lys2 trp1 ura3 uba2Δ::HIS3)containing the uba2ts10 plasmid pIS2-ts10(Johnson et al., 1997) were used for two-hybrid analyses and for expressing Dmuba2, respectively,and were grown under standard conditions(Guthrie and Fink, 1991). Escherichia coli strains DH5α and DH12S (Life Technologies,Grand Island, NY) were used for routine plasmid propagation, and strains JMB9r-m+ΔtrpF(Sterner et al., 1995) and JM109 (Promega, Madison, WI) were used as described below.
Standard recombinant DNA techniques(Ausubel et al., 1994;Gietz et al., 1992) were used except where noted. PCR used either the Expand High Fidelity kit (Boehringer Mannheim, Indianapolis, IN) to generate fragments for cloning into fusion-protein and two-hybrid vectors or Taq DNA polymerase (Promega)for other applications. Oligonucleotide primers were obtained from Integrated DNA Technologies (Coralville, IA).
Two-hybrid assays and screening
Two-hybrid assays were performed as described previously(DeMarini et al., 1997;Fields and Sternglanz, 1994;Gyuris et al., 1993) using the LexA DNA-binding domain (DBD) plasmid pEG202(Ausubel et al., 1994) and activation domain (AD) plasmid pJG4-5(Ausubel et al., 1994) or pJG4-5PL (DeMarini et al.,1997). The baits used for screening contained full-length sequences of pnut, sep1, and sep2 (J. C. Adam et al.,unpublished). The libraries used (provided by R. Finley, Harvard Medical School, Boston, MA) were RFLY1, a Drosophila embryonic cDNA library,and RFLY5, an imaginal-disc cDNA library, both in pJG4-5. Each bait plasmid was co-transformed into yeast strain EGY48R with each of the two libraries. Transformants were plated onto Synthetic Minimal plates(Guthrie and Fink, 1991)containing 2% galactose + 1% raffinose and grown for 3-6 days to select Leu+ clones, which were further evaluated using both filter and liquid-culture β-galactosidase assays(Ausubel et al., 1994). Plasmids from positive clones were rescued into E. coli strain JMB9 and retested by cotransformation into EGY48R together with one of the pEG202-based plasmids and measurement of β-galactosidase activity. Inserts from positive plasmids were then sequenced. We screened>106 transformants for each library with each bait.
Additional two-hybrid tests used the following plasmids. pJG4-5PL,expressing full-length anillin, was provided by Julie Brill (Hospital for Sick Children, Toronto, Canada). Construction of plasmids expressing N- or C-terminal fragments of Pnut (Pnut-N, amino acids 1-426; Pnut-C, amino acids 407-539), Sep1 (Sep1-N, amino acids 1-325; Sep1-C, amino acids 306-361), or Sep2 (Sep2-N, amino acids 1-337; Sep2-C, amino acids 318-419) in pEG202 will be described in detail elsewhere (J. C. Adam et al., unpublished). Full-length Dmuba2 and Dmaos1 and the 5′ (amino acids 1-350) and 3′ (amino acids 351-700) halves of Dmuba2 were amplified by PCR using the cloned genes (see below) as templates and the primers shown inTable 1. The amplified Dmuba2 and its fragments were cloned into pEG202 and pJG4-5PL at their EcoRI sites, and Dmaos1 was cloned into pEG202 and pJG4-5PL at their XhoI sites. The full-length genes and the Dmuba2 C-terminal constructs contain the original stop codons,whereas the Dmuba2 N-terminal constructs use a stop codon in the vector downstream of the polylinker.
Isolation of full-length Dmuba2, Dmaos1, and Dmsmt3clones
Two independent positive clones from the two-hybrid screen encoded C-terminal fragments (including the putative stop codon) of the same gene. To isolate the 5′ end of this gene, we identified an EST clone (LD03967)from the Berkeley Drosophila Genome Project (BDGP) that extended further in the 5′ direction and then performed PCR as described previously (McCurdy and Kim,1998), using a Schneider cell cDNA library in vector pDB20(Becker et al., 1991) as template and primers (Table 1)that corresponded to vector sequences and to the 5′ end of the LD03967 sequences, respectively. In aggregate, the cDNA sequences revealed an apparently complete ORF of 2,100 bp that had similarity to yeast UBA2(see Results). Full-length Dmuba2 was then cloned by PCR using the primers shown in Table 1, the Schneider cell cDNA library as the template, and plasmid pGEM-T (Promega). We mapped Dmuba2 to position 66B6-66B10 on chromosome arm 3L using a high-density filter of P1 clones (Genome Systems, St. Louis, MO), as subsequently confirmed by data from the genome project (see FlyBase).
Full-length clones of Dmaos1 and Dmsmt3 were obtained by identifying and sequencing BDGP EST clones (GM10027 and GM01812) that contained apparently complete ORFs with similarity to yeast Aos1p and Smt3p,respectively (see Results).
Antibodies, immunoblotting, and immunoprecipitation
Rabbit polyclonal anti-DmUbc9 antibodies(Joanisse et al., 1998) were provided by Robert Tanguay (University of Laval, Quebec, Canada), mouse monoclonal anti-Pnut antibody 4C9 (Neufeld and Rubin, 1994) and anti-myc antibody 9E10(Evan et al., 1985) were obtained from the Developmental Studies Hybridoma Bank (University of Iowa,Iowa City, IA), and anti-β-tubulin antibody was obtained from Amersham-Pharmacia Biotech (Piscataway, NJ).
To prepare DmUba2-specific antibodies, DNA encoding DmUba2 amino acids 553-700 was amplified by PCR using one of the two-hybrid clones as the template and the primers shown in Table 1. The products were cut with EcoRI and cloned into pGEX-1 (Amersham-Pharmacia), producing pGEX-DmUba2, and into pMAL-c2 (New England Biolabs, Beverly, MA), producing pMAL-DmUba2. The resulting glutathione-S-transferase (GST)- and maltose-binding protein(MBP)-fusion proteins were expressed, purified, and used to raise antibodies in rabbits by standard methods (Ausubel et al., 1994) (Cocalico Biologicals, Reamstown, PA). After five boosts, antibodies raised against GST-DmUba2 were affinity purified using MBP-DmUba2 that had been subjected to SDS-PAGE and blotted to nitrocellulose(Pringle et al., 1989).
To prepare DmSmt3-specific antibodies, a His6-tagged DmSmt3 protein was generated. Dmsmt3 codons 1-88 were amplified by PCR using Canton S genomic DNA (Sullivan et al.,2000) as the template and the primers shown inTable 1. The resulting product and plasmid pQE30 (Qiagen) were digested with BamHI and HindIII and ligated together to produce pQE30-DmSmt3, which encodes DmSmt3 tagged at its N-terminus with His6 and truncated after the two glycines that presumably represent the C-terminus of the mature endogenous protein (Johnson et al., 1997;Kamitani et al., 1997). His6DmSmt3 was expressed in strain JM109 and purified on Ni-NTA columns as recommended by Qiagen, and rabbit antibodies were raised using standard protocols (Cocalico Biologicals). After four boosts, DmSmt3-specific antibodies were affinity purified using His6DmSmt3 that had been subjected to SDS-PAGE and blotted to nitrocellulose (see above).
For immunoblotting, samples were diluted or resuspended in 5×- or 2×-concentrated Laemmli buffer to achieve a 1× final concentration, treated at 100°C for 5 minutes (except where noted),analyzed by SDS-PAGE, and blotted to nitrocellulose. Blots were blocked for 1 hour at 23°C with 5% non-fat dry milk in TBS(Ausubel et al., 1994) buffer containing 0.1% Tween-20 (TBS-T), incubated with primary antibodies in the same buffer for 1 hour at 23°C, washed three times in TBS-T at 23°C(10 minutes per wash), incubated with secondary antibodies in TBS-T containing 5% non-fat dry milk for 1 hour at 23°C, and washed three times in TBS-T at 23°C (5 minutes per wash). Proteins were then detected using the enhanced chemiluminescence (ECL) kit (Amersham-Pharmacia). The primary antibodies used were purified anti-DmUba2 (at 1:1500), anti-DmUbc9 (at 1:2000), and purified anti-DmSmt3 (at 1:750). The secondary antibody was HRP-conjugated goat anti-rabbit-IgG (at 1:5,000; ICN Pharmaceuticals, Costa Mesa, CA). Fly extracts were prepared from 0-16-hour-old embryos as described previously(Pai et al., 1996) using NET buffer (50 mM Tris-HCl, pH 7.5, 400 mM NaCl, 5 mM EDTA, 1% NP-40) containing phosphatase and protease inhibitors. In one case, the buffer also contained 20 mM N-ethylmaleimide (NEM; Sigma).
To express Dmuba2 in yeast under GAL-promoter control, we amplified full-length Dmuba2 by PCR using the cloned cDNA as template and the primers shown in Table 1, cut the product with SalI, and cloned it into YCpIF2(Foreman and Davis, 1994),producing YCpIF2-DmUba2. Strain JD90-1A [pIS2-ts10] (see above) was transformed with YCpIF2-DmUba2 and grown under conditions selective for both plasmids and inducing or noninducing for the GAL promoter. Yeast extracts were then prepared as described previously(Ausubel et al., 1994)
To prepare fly extracts for immunoprecipitation (IP), 0-16-hour-old embryos were collected, dechorionated, and rinsed as described previously(Pai et al., 1996). 100 μl of embryos were homogenized at 0°C in extraction buffer (500 μl of CER I plus 27.5 μl of CER II; both reagents from the NE-PER kit, Pierce,Rockford, IL). The extract was added to 1 ml of NET buffer containing phosphatase and protease inhibitors (Pai et al., 1996) and centrifuged for 10 minutes at 15,000 g in a microfuge at 4°C, and the supernatant was collected. IPs were carried out as described previously(Peifer, 1993) using purified anti-DmUba2 (at 1:75), anti-DmUbc9 (at 1:200), or monoclonal anti-myc 9E10 (at 1:20). IPs were analyzed by SDS-PAGE and immunoblotting.
In vitro protein-binding assays
Codons 1-76 of a Drosophila ubiquitin gene(Lee et al., 1988) were amplified by PCR using the Schneider cell cDNA library as the template and the primers shown in Table 1. The resulting fragment was then cloned into pQE30 using KpnI and HindIII. The resulting plasmid encodes ubiquitin (DmUb) tagged at its N-terminus with His6 and truncated after the Gly-Gly that presumably represents the C-terminus of the mature endogenous protein. His6-DmSmt3 (see above) and His6-DmUb were expressed separately in strain JM109 as recommended by Qiagen and purified using the B-PER kit (Pierce). Fly embryo extracts were prepared as described above except that homogenization was in 1 ml of RIPA buffer (50 mM Tris-HCl, pH 8.5,300 mM NaCl, 0.5% sodium deoxycholate, 0.1% SDS, 1% NP-40) containing 20 mM imidazole, 2 mM ATP, 5 mM MgCl2, 0.1 mM DTT, 50 mM NaF, and protease inhibitors (see above). Extract was centrifuged for 10 minutes at 8000 g in a microfuge at 4°C, and 5 μg of His6-DmSmt3 or His6-DmUb was added to 450 μl of supernatant. After incubation at 23°C for 15 minutes, 15 μl of Ni-NTA beads were added, and the mixture was incubated at 4°C for 45 minutes. The beads were washed four times in RIPA containing 20 mM imidazole, 50 mM NaF,and protease inhibitors (as above). For analysis under reducing conditions,beads were suspended in standard 1× Laemmli buffer and analyzed by immunoblotting as described above. For analysis under nonreducing conditions,beads were suspended in 1× Laemmli buffer without β-mercaptoethanol and heated at 68°C for 10 minutes prior to SDS-PAGE and immunoblotting.
Embryos collected after development at 25°C were washed and dechorionated as described above, then fixed, devitellinized, and stained as described (Cox et al., 1996). The ages of syncytial-blastoderm embryos were determined by counting the nuclei in longitudinal sections. Standard morphological criteria(Roberts, 1986) were used to identify other developmental stages. Clone-8 cells were washed three times with PBS, fixed with 2% paraformaldehyde in PBS for 10-15 minutes at 23°C,washed twice with PBS, treated with primary antibodies in Solution H (PBS containing 0.1% Triton X-100 and 1% normal goat serum) for 1 hour at 23°C,washed once with PBS, treated with secondary antibodies in Solution H for 1 hour at 23°C, and washed twice again with PBS. The primary antibodies used were purified anti-DmUba2 (at 1:50) and anti-DmSmt3 (at 1:75), monoclonal anti-Pnut antibody 4C9 (at 1:3), and anti-β-tubulin (at 1:100). FITC- and rhodamine-conjugated goat anti-rabbit-IgG and goat anti-mouse-IgG (Jackson ImmunoResearch, West Grove, PA) were used at 1:500; Alexa Fluor 488-conjugated goat anti-rabbit-IgG (Molecular Probes, Eugene, OR) and Cy3-conjugated goat anti-mouse-IgG (Jackson ImmunoResearch) were used at 1:1000. Embryos and cells were mounted in AquaPolymount (Polysciences, Warrington, PA), then observed and photographed using a Zeiss LSM 410 confocal microscope.
Identification of genes by two-hybrid interactions with Drosophila septins
To identify proteins interacting with the Drosophila septins, we conducted two-hybrid screens using Pnut, Sep1, and Sep2 as baits (see Materials and Methods). These screens identified 27 positive clones that proved to represent eight genes (Table 2). Among these were the other septins, as expected from other data indicating that septins interact with one another (see Discussion). In addition, the screens identified Drosophila homologues(Dmuba2 and Dmubc9) of yeast UBA2 and UBC9, whose products are involved in the activation and conjugation of the ubiquitin-like molecule Smt3p/SUMO (see Introduction). These screening results and the recent discovery of Smt3p modification of septins in S. cerevisiae (Johnson and Blobel,1999; Takahashi et al.,1999) (P. Meluh, personal communication) stimulated us to study the Smt3p/SUMO conjugation machinery in Drosophila.
In further two-hybrid analyses, the C-terminal portion of DmUba2 interacted strongly with full-length Sep1 and Sep2 and with the N-terminal portion of Sep1. Interactions were also detected with the N-terminal portions of Pnut and Sep2 and with the C-terminal portions of Sep1 and Sep2(Table 2). In contrast, a full-length AD-DmUba2 fusion showed none of these interactions(Table 2), although other studies (see Table 3) indicated that this fusion was functional for other interactions. Interestingly,full-length AD-DmUbc9 showed a pattern of interactions very similar to those seen with the C-terminal portion of DmUba2.
The other genes identified in the screens had not been described previously; we designated them sip1-sip4 (for septin-interacting protein). sip1 (Accession No. AF221101; Drosophila genome annotation No. CG7238) encodes a protein with predicted P-loop and coiled-coil domains; it appeared to interact specifically with the C-terminal portion of Pnut (Table 2). sip2-sip4 encode proteins without obviously informative motifs. sip2 (CG9188) encodes a protein that interacted with full-length Sep1 and Sep2 but not with Pnut. sip3 (CG1937) encodes a protein that appeared to interact specifically with the C-terminal portion of Sep2. sip4 was identified independently as dip2 (Dorsal interacting protein 2) (Bhaskar et al.,2000); it encodes a protein that interacted with all of the Sep1 and Sep2 fusions and (weakly) with the N-terminal portion of Pnut(Table 2).
Cloning and sequence analysis of Dmuba2 and Dmubc9
We obtained a full-length clone of Dmuba2 as described in the Materials and Methods. Sequencing (Accession No. AF193553) showed that the predicted DmUba2 contains 700 amino acids and has 29% sequence identity to yeast Uba2p and 48% identity to human hUba2, as observed also by others(Bhaskar et al., 2000;Long and Griffith, 2000;Donaghue et al., 2000). Like its homologues and other E1-type enzymes, DmUba2 contains an ATP-binding motif (amino acids 26-31) and the consensus Cys (C175)corresponding to those essential for thiolester bond formation in other E1-type enzymes (Desterro et al.,1999; Dohmen et al.,1995). Our original two-hybrid clone of Dmubc9 appeared to be full length by comparison to yeast UBC9 and our sequence for Dmubc9 agreed with that reported by Joanisse et al.(Joanisse et al., 1998).
We then raised antibodies to DmUba2 (see Materials and Methods). The affinity-purified antibodies recognized mainly one polypeptide of apparent molecular weight ∼97 kDa (Fig. 1A), which is presumably DmUba2 (predicted molecular weight, 77.5 kDa). Support for this conclusion was obtained by expressing Dmuba2under GAL-promoter control in yeast. When cells were grown under inducing conditions, the antibodies recognized primarily a polypeptide of apparent molecular weight ∼97 kDa (Fig. 1B, lane 1) that was absent when cells were grown under repressing conditions (Fig. 1B, lane 2). Similarly anomalous low mobility on SDS-PAGE has been noted for both Uba2p and hUba2 (Desterro et al., 1999;Dohmen et al., 1995).
Identification of DmAos1 and its interaction with DmUba2
In S. cerevisiae, the E1 enzyme for ubiquitin activation is the 1024 amino-acid Uba1p (McGrath et al.,1991). In contrast, Smt3p is activated by a heterodimer of the 636 amino-acid Uba2p, which is related to the C-terminal part of Uba1p, and the 347 amino-acid Aos1p, which is related to the N-terminal part of Uba1p(Johnson et al., 1997). Similarly, the 700 amino-acid DmUba2 is related in sequence (∼40% identity over the ∼225 amino acids of the three similarity boxes defined for other Uba1-type and Uba2-type enzymes (Johnson et al., 1997; Okuma et al.,1999)) to the C-terminal part of the putative Drosophilaubiquitin-activating enzyme DmUba1 (EMBL# Y15895). Therefore, we sought and identified a Drosophila homologue of yeast AOS1 among the ESTs from the Berkeley Drosophila Genome Project (BDGP) (see Materials and Methods). Sequencing (Accession No. AF193554) showed that Dmaos1 encodes a polypeptide of 337 amino acids that has 28% sequence identity to yeast Aos1p and 40% identity to the human Aos1p homologue Sua1(Okuma et al., 1999), as also observed by others (Bhaskar et al.,2000; Long and Griffith,2000). As expected, DmAos1 is related in sequence to the N-terminal part of DmUba1 (∼37% identity over the ∼202 amino acids of the similarity boxes as defined previously(Johnson et al., 1997;Okuma et al., 1999)),suggesting that a heterodimer of DmUba2 and DmAos1 is the DrosophilaSmt3/SUMO-activating enzyme. In support of this hypothesis, we detected an interaction between full-length DmUba2 and full-length DmAos1 in the two-hybrid system (Table 3). An attempt to use the two-hybrid system to delimit the region of DmUba2 responsible for its interaction with DmAos1 was unsuccessful(Table 3).
Identification of DmSmt3 and of DmSmt3-conjugated proteins
We next sought and identified a homolog of SMT3/SUMO-1 among the BDGP EST clones (see Materials and Methods). The predicted DmSmt3 contains 90 amino acids with 48% identity to yeast Smt3p and 54% identity to human SUMO-1,as also observed by others (Bhaskar et al.,2000; Huang et al.,1998; Lehembre et al.,2000). We generated polyclonal antibodies (see Materials and Methods) and performed immunoblots on fly extracts, expecting to detect both free DmSmt3 and DmSmt3-modified proteins. Because yeast and mammalian cells contain Smt3p/SUMO-1-specific isopeptidases(Gong et al., 2000;Kim et al., 2000;Li and Hochstrasser, 1999;Li and Hochstrasser, 2000),which remove Smt3/SUMO from modified proteins, we prepared extracts both with and without N-ethylmaleimide (NEM), an isopeptidase inhibitor. In both extracts, the purified antibodies recognized both a polypeptide of ∼16 kDa(presumably free DmSmt3) and many polypeptides of higher molecular weight(presumably DmSmt3-conjugated proteins)(Fig. 1C). As expected, the higher molecular-weight species were both less abundant and of lower average molecular weight when extracts were prepared without NEM.
Interactions of DmUba2 and DmUbc9 with each other and with DmSmt3
To test the hypothesis that DmUba2/DmAos1 and DmUbc9 are activating and conjugating enzymes for DmSmt3, but not for ubiquitin (DmUb), we used in vitro protein-binding assays to investigate the interactions among these proteins. Because ubiquitin and ubiquitin-like proteins undergo proteolytic cleavage of their C-termini to leave the sequence Gly-Gly, which is essential for both activation and conjugation (Kamitani et al., 1997; Melchior,2000), we cloned DmSmt3 and DmUb such that they terminated with Gly88 (DmSmt3) or Gly76 (DmUb) and were tagged with His6 at their N-terminal ends (see Materials and Methods). We then incubated purified His6DmSmt3 and purified His6DmUb with fly extracts,isolated the His6-tagged proteins using Ni-NTA beads, and analyzed the associated proteins. As expected, we found that both DmUba2 and DmUbc9 associated only with His6DmSmt3 and not with His6DmUb(Fig. 2A,C). The anti-DmUba2 antibodies detected not only the free form of DmUba2 (∼97 kDa) but also species whose lower mobilities (Fig. 2A) suggested that they might represent DmUba2 conjugated to one,two, or three molecules of DmSmt3 (and/or some other ubiquitin-like molecule). To ask if the interactions of DmUba2 and DmUbc9 with DmSmt3 involved thiolester bonds, we repeated the experiments but omittedβ-mercaptoethanol (which reduces thiolester bonds) during sample preparation. As expected, the most abundant species now observed with the anti-DmUba2 antibodies had an apparent molecular weight of ∼116 kDa(Fig. 2B), consistent with its being DmUba2 with a single His6DmSmt3 linked by a thiolester bond. Similarly, the anti-DmUbc9 antibodies now revealed an additional species with an apparent molecular weight of ∼30 kDa(Fig. 2D), presumably representing DmUbc9 with a single His6DmSmt3 linked by a thiolester bond.
We also used immunoprecipitation to ask whether DmUba2 and DmUbc9 interact in vivo. When immunoprecipitates were prepared from embryo extracts using antibodies to either protein, the other protein was detected by immunoblotting(Fig. 3). Taken together, the results of in vitro binding assays and coimmunoprecipitation suggest that DmUba2/DmAos1 and DmUbc9 are indeed the activating and conjugating enzymes,respectively, for DmSmt3 and that they may form a complex containing both the E1-type and E2-type enzymes.
Localization of DmUba2 during embryogenesis
To begin investigating the roles of the DmSmt3-conjugation pathway in Drosophila, we used immunofluorescence and confocal microscopy to characterize the intracellular localization of DmUba2. Yeast Uba2p is concentrated in the nucleus (Dohmen et al., 1995), but the localization of the homologous enzyme has not been examined in multicellular organisms. Interestingly, we observed that DmUba2 is not exclusively nuclear during early embryogenesis. Before migration of the nuclei to the embryo cortex after nuclear division 9, DmUba2 was found largely in the cortex, and its distribution there appeared homogenous (data not shown). During the interphase preceding nuclear division 10, DmUba2 gradually became organized into a cap corresponding approximately to the cortical actin cap that forms over each nucleus(Fig. 4, A1-A3, C1-C3). DmUba2 was also found in the deeper cytoplasm(Fig. 4, B1-B3, C1-C3) and gradually moved into the nuclei (Fig. 4,B2-B3, C2-C3). During mitosis, DmUba2 was dispersed in the cortex and in the cytoplasm near the cortex (Fig. 4, A4, B4, C4). During the three subsequent nuclear cycles, the cap-like localization (Fig. 4, D1-D3,F1-F3, G1-G2, I1-I2), progressive nuclear accumulation(Fig. 4, E1-E3, F1-F3, H1-H2,I1-I2), and dispersion during mitosis(Fig. 4, D4, E4, F4, G3, H3,I3) of DmUba2 remained evident. However, during the successive cycles, the cap-like cortical localization became more organized and the degree of nuclear enrichment became more pronounced. By cycle 13, although some DmUba2 still localized to the cortex during interphase, it was predominantly nuclear (Fig. 4, G1-G2,H1-H2, and I1-I2).
Because DmUba2 was identified by its two-hybrid interaction with Sep1 and Sep2, we examined carefully whether DmUba2 colocalized with the septins either at the cellularization front or in cleavage furrows in mitotic domains after gastrulation. During cellularization, most DmUba2 localized to nuclei,accumulating preferentially at their apical ends(Fig. 5A,C). We did not detect DmUba2 at the cellularization front. However, some DmUba2 remained at the cortex, where diffuse septin staining was also observed(Fig. 5A,C). During mitosis in older embryos, DmUba2 spread throughout the cell but did not become detectably concentrated in cleavage furrows (Fig. 5E-G, cells 1 and 2 and inset); it then moved back into the nuclei after mitosis, with no detectable concentration at the midbody(Fig. 5E-G, cells 3 and 4). Thus, we detected no substantial colocalization of DmUba2 and septins in early embryogenesis. However, it remains possible that DmUba2 could interact with septins at the cortex in syncytial-blastoderm embryos, during cellularization,or in mitotic cells. DmUba2 was concentrated in nuclei of nondividing cells and dispersed throughout the cell during mitosis throughout embryonic stages 6 to 15 (Fig. 4J,K) (data not shown). This was particularly striking in the CNS, where the septins, in contrast, are enriched in axons (Fares et al., 1995; Neufeld and Rubin,1994) (J. C. Adam et al., unpublished)(Fig. 6D).
We also examined DmUba2 localization during oogenesis. DmUba2 localized to the nuclei of both germ cells and somatic follicle cells in the germarium(Fig. 6A, green arrowhead;Fig. 6C). After encapsulation of the germ-line cells by the follicle cells, DmUba2 remained localized to follicle cell nuclei (Fig. 6A,white arrow; Fig. 6B). DmUba2 localization to nurse cells decreased as egg chamber development progressed(Fig. 6A,B, green arrows), but it remained enriched in the oocyte nucleus(Fig. 6A, white arrowhead). In contrast, Pnut localizes primarily to the basal surface of the follicle cells and is excluded from nuclei (Fig. 6A).
Localization of DmSmt3 in embryos and cultured cells
The two-hybrid interactions between the septins and DmUba2 and DmUbc9 might reflect a physiologically significant but transient interaction, such as might occur if Drosophila septins, like yeast septins(Johnson and Blobel, 1999;Takahashi et al., 1999), are Smt3 modified. To explore this possibility, we used immunofluorescence to examine DmSmt3 localization in cultured cells and in cellularizing and older embryos. DmSmt3 did not colocalize detectably with the septins at the cellularization front (Fig. 5B,D). Instead, DmSmt3 localized to nuclei, with a particular enrichment at their apical ends, as did DmUba2(Fig. 5A-D). In cultured cells and in cells of post-gastrulation embryos, DmSmt3 was concentrated in nuclei throughout interphase (Fig. 5H,cells 1 and 2; Fig. 5K-M, cells 1-3). During mitosis, DmSmt3 initially appeared to spread throughout the cell(Fig. 5K-M, cell 5). However,during metaphase, DmSmt3 appeared to concentrate in the region of the chromosomes (Fig. 5I;Fig. 5K-M, cell 4); this was confirmed by localizing DmSmt3 relative to the mitotic spindle(Fig. 5J). Strikingly, DmSmt3 was also found concentrated in a spot at cleavage furrows and midbodies both in cultured cells (Fig. 5H,cells 3 and 4) and in dividing embryonic cells(Fig. 5K-M, cells 6-8, 10 and 13); this spot overlapped, but did not appear to coincide fully, with the concentration of the septins in these furrows. DmSmt3 was not enriched at the cleavage furrows early in furrow formation(Fig. 5K-M, cells 9, 11 and 12), but only during later stages, and it remained concentrated in the midbody after most DmSmt3, and essentially all of the DmUba2, had reaccumulated in nuclei (Fig. 5H, cell 4; K-M,cells 6-8, 10, and 13; and compare with E-G, cells 3 and 4).
We also examined DmSmt3 localization during other stages of embryogenesis. During the syncytial cell cycles, DmSmt3 was concentrated in the nuclei during interphase (Fig. 6G) and appeared to localize to the chromosomes during mitosis(Fig. 6F). Like DmUba2, DmSmt3 localized primarily to the nuclei of non-mitotic cells throughout the rest of embryogenesis (Fig. 6H),including in the CNS (where the septins, in contrast, localized to axons)(Fig. 6E).
The concentration of DmSmt3 in late cleavage furrows and midbodies suggested that one or more of the Drosophila septins might be modified by DmSmt3. We thus tried various experimental approaches to look for DmSmt3-modified septins. First, we used immunoblotting to look for higher-molecular-weight forms of Pnut, Sep1 or Sep2, which might represent Smt3-modified proteins, in extracts from embryos, adults, and cultured Clone-8 and S2 cells. Second, we immunoprecipitated Pnut, Sep1, and Sep2 from embryonic extracts and analyzed the precipitates by immunoblotting using anti-DmSmt3 antibodies. Third, we prepared immunoprecipitates from embryonic extracts using anti-DmSmt3 antibodies and analyzed the precipitates by immunoblotting using anti-Pnut, anti-Sep1, and anti-Sep2 antibodies. In each case, we did the experiments both in the presence and the absence of the isopeptidase inhibitor NEM. None of these approaches detected DmSmt3 modification of Pnut, Sep1, or Sep2 (data not shown). These negative results may mean that the septins are not Smt3 modified. However, our immunofluorescence studies detected colocalization of Smt3 with the septins only for a brief period at the end of mitosis, and even at this stage the overlap was not complete. Thus even if septins are Smt3 modified during this period, they would probably comprise only a very small fraction of the total septins in the embryo and thus might have escaped our detection.
Possible interaction of the septins and the Smt3 conjugation system
The Drosophila septins appear to be essential for cytokinesis in at least some cell types, and it is likely that they have a variety of non-cytokinesis roles as well (see Introduction). Because studies in yeast suggest that a primary function of the septins is to serve as a matrix or template for the organization of other proteins at the cell surface, the identification of septin-interacting proteins should be critical to the elucidation of septin function in Drosophila. This study began with an attempt to identify such proteins using the yeast two-hybrid system. Of 27 positive clones identified with three septin baits, 17 contained fragments of the septin genes themselves. Because other evidence suggests strongly that the septins interact with each other in vivo(Beites et al., 1999;De Virgilio et al., 1996;Fares et al., 1995;Field et al., 1996;Frazier et al., 1998;Hsu et al., 1998;Longtine et al., 1996), this result suggests that the baits used were good and that the screen was otherwise of high fidelity. Thus, it seems likely that at least some of the other positive clones represent genes whose products really interact with septins in vivo.
Of the six non-septin genes identified in the screens, only two have been investigated in detail as yet. These genes, Dmuba2 and Dmubc9, encode the Drosophila homologues of yeast Uba2p and Ubc9p, which catalyze the activation and conjugation, respectively, of the ubiquitin-like molecule Smt3p (see Introduction). Several lines of evidence suggest that the two-hybrid interactions observed between DmUba2, DmUbc9, and the septins reflect physiologically significant interactions. First, among the 10 positive clones that did not encode septins, four contained either Dmuba2 or Dmubc9, and Dmuba2 fragments were isolated with two different septin baits. Second, because DmUba2 and DmUbc9 also interact with each other (see Results), the identification of both genes independently in screens using septin baits suggests that the interactions are relevant. Third, while our studies were in progress, it became clear that yeast septins are extensively modified by Smt3p, although the functional significance of that remains uncertain(Johnson and Blobel, 1999;Takahashi et al., 1999;Takahashi et al., 2001;Johnson and Gupta, 2001) (P. Meluh, personal communication). Fourth, several other groups also isolated Uba2 and/or Ubc9 in two-hybrid screens using other protein baits. Several of these interactors, including Drosophila calcium/calmodulin-dependent kinase II (CAM-kinase II; Long and Griffith, 2000) and Dorsal(Bhaskar et al., 2000), were subsequently shown to be Smt3 modified.
Other genetic data may also reflect an interaction between the septins and the DmSmt3 system. The pnut septin mutation was originally identified as an enhancer of a sina mutation that affects R7 photoreceptor development (Carthew et al.,1994; Neufeld and Rubin,1994). Although the significance of this genetic interaction remains unclear, it may reflect the recently discovered crosstalk between Smt3/SUMO modification and ubiquitination. Mammalian SUMO-1 is conjugated to the protein Mdm2, a RING-finger E3 ubiquitin ligase involved in p53 degradation (Buschmann et al.,2000). Modification by SUMO-1 appears to regulate Mdm2 activity and hence the level of p53, probably by regulating the ubiquitination and degradation of Mdm2 itself. Other RING-finger proteins, including Drosophila Sina, interact with Ubc9 family proteins and/or are modified by Smt3/SUMO (Duprez et al.,1999; Hu et al.,1997). Thus, it seems possible that DmSmt3 modification regulates the activities of Sina, such as its role in downregulating the transcriptional repressor Tramtrack (one of whose isoforms is itself DmSmt3 modified)(Lehembre et al., 2000;Li et al., 1997;Neufeld et al., 1998;Tang et al., 1997), and that Pnut plays a role in mediating the requisite interactions.
Finally, in several types of dividing Drosophila cells, we found that DmSmt3 colocalizes with septins in the cleavage furrows and/or midbodies during cytokinesis. Interestingly, DmSmt3 is not enriched in the furrow during the early stages of furrow formation but only later, at a time when most DmUba2 has moved back into the nuclei. Although we were unable to detect DmSmt3 modification of Sep1, Sep2, or Pnut (despite considerable effort), it remains possible that one or more of these proteins is modified at low levels or that DmSmt3 is conjugated to Sep4 or Sep5. However, it is also possible that the DmSmt3 in the midzone is conjugated to other proteins, such as the`chromosomal passenger proteins' (Adams et al., 2001), which are associated with the chromosomes and then relocate to the spindle midzone in mitosis. The involvement of such proteins in chromosome segregation as well as in cytokinesis might help to explain the observations suggesting that the Smt3/SUMO system is involved in chromosome segregation (Apionishev et al.,2001; Meluh and Koshland,1995; Tanaka et al.,1999).
It also remains unclear whether the Drosophila septins ever serve as a matrix/template for the localization of the DmSmt3 conjugation system. Although our immunofluorescence studies show that DmUba2 and the septins are sometimes in the same compartment, so that interaction would be possible, we observed no persuasive colocalization of the proteins. Thus, elucidation of the possible interactions between the septins and the Smt3 system in Drosophila, and of their functional significance, will need to await further studies using other approaches.
Conservation and possible functions of the Smt3/SUMO conjugation pathway in Drosophila
Despite these residual uncertainties, in the course of our studies we generated other valuable information about the Smt3/SUMO conjugation system in Drosophila that complement and extend recent work by others on this system. For example, our biochemical and two-hybrid studies indicate that there are multiple DmSmt3-modified proteins, that DmSmt3-specific isopeptidases probably exist, that DmUba2 and DmAos1 interact with each other,and that both DmUba2 and DmUbc9 become conjugated to DmSmt3, but not to DmUbiquitin, by thiolester bonds. While our studies were in progress, related findings were also made by other investigators who were led by other routes to the Smt3/SUMO system in Drosophila. In particular, several other proteins were shown to interact with DmUba2 and DmUbc9 using the two-hybrid system, and multiple DmSmt3-modified proteins were observed(Bhaskar et al., 2000;Donaghue et al., 2001;Joanisse et al., 1998;Lehembre et al., 2000;Long and Griffith, 2000;Ohsako and Takamatsu, 1999),supporting the hypothesis that DmSmt3 is indeed conjugated to a variety of proteins in vivo. In addition, the DmUba2/DmAos1 interaction and the conjugation of DmSmt3 to DmUba2 and DmUbc9 by thiolester bonds were also observed using other methods (Lehembre et al., 2000; Long and Griffith,2000). Finally, it was shown that Dmubc9 can functionally complement a yeast ubc9 mutation(Joanisse et al., 1998;Ohsako and Takamatsu, 1999). Taken together, these results make clear that the Smt3/SUMO conjugation system is closely conserved between Drosophila, yeast, and mammals. Our data also show that DmUba2 and DmUbc9 form a complex in vivo, suggesting that the conjugation machinery may act in a concerted fashion.
The studies in other laboratories also provided clues to possible functions of modification by DmSmt3. In particular, the semushi(Epps and Tanda, 1998) and lesswright (Apionishev et al.,2001) mutations are both in Dmubc9, suggesting (from their mutant phenotypes) that DmUbc9 has roles in the nuclear import of the transcription factor Bicoid and in meiotic chromosome segregation. In addition, two other transcriptional regulators, Dorsal and Tramtrack, as well as CAM-kinase II, have also been shown to be modified by DmSmt3(Bhaskar et al., 2000;Lehembre et al., 2000;Long and Griffith, 2000). In the case of Dorsal, as with Bicoid, DmSmt3 conjugation appears to promote nuclear localization, whereas Tramtrack modification may help to regulate its activity and/or its degradation by the proteasome, as discussed above. The modification of CAM-kinase II may regulate its activity.
Although most suggested functions of the Smt3/SUMO system in Drosophila, as well as in yeast and mammalian cells, center on nuclear proteins, our immunofluorescence and two-hybrid data support the hypothesis that there are also cytoplasmic targets. First, as discussed above,it remains likely that in Drosophila, as in yeast, there is an interaction between the Smt3/SUMO system and the septins, which appear to be exclusively cytoplasmic proteins. Second, some DmSmt3-modified proteins appear to remain both at the embryo cortex during cellularization and in the midbodies that remain after the nuclear envelopes have reformed at the end of cytokinesis. Third, although DmUba2 (this study) and DmUbc9(Joanisse et al., 1998;Lehembre et al., 2000) are found primarily in nuclei, considerable DmUba2 is also found in the cytoplasm and at the cortex during the syncytial-blastoderm stage, suggesting that DmSmt3 modification of cortical and/or cytoplasmic proteins could occur. Finally, the cell-cycle-regulated translocation of DmUba2 between cytoplasm and nucleus both in syncytial-blastoderm and in post-cellularization embryos may suggest that the partitioning of the DmSmt3-conjugation system between the cytoplasm and the nucleus is important and thus well regulated during embryonic development.
We thank Erica Johnson and Pam Meluh for stimulating discussions; Robert Tanguay for antibodies; Julie Brill for constructs; Tony Perdue and Susan Whitfield for help with confocal microscopy and photography; and members of the Pringle and Peifer laboratories for encouragement and comments on the manuscript. This work was supported by National Institute of Health grant GM 52606 to J.R.P. and M.P. and by NIH Postdoctoral Fellowship GM19981 to K.G.H.