Gemin3: A Novel Dead Box Protein That Interacts with Smn, the Spinal Muscular Atrophy Gene Product, and Is a Component of Gems

The p105 protein was coimmunoprecipitated with anti-SMN mAb 2B1 and the band was excised from a single one-dimensional Coomassie stained polyacrylamide gel and in-gel digested with trypsin (unmodified, sequencing grade; Boehringer Mannheim Corp.) as described in Shevchenko et al. 1996. Tryptic peptides were recovered from gel pieces by extraction with 5% formic acid and acetonitrile. The combined extracts were pooled together, dried in a speed vac, redissolved in 5% formic acid, and analyzed by nanoelectrospray tandem mass spectrometry (nano-ES MS/MS) as described in Wilm et al. 1996. Nano-ES MS/MS was performed on a API III triple quadrupole instrument (PE Sciex) equipped with a nano-ES ion source developed in EMBL (Wilm and Mann 1996).

Comprehensive protein and expressed sequence tag (EST) databases were searched using PeptideSearch v. 3.0 software developed by M. Mann and P. Mortensen (University of Southern Denmark, Odense, Denmark). No limitations on protein molecular weight and species of origin were imposed.

Several peptides of the p105 band analyzed by MS identified a human EST sequence (clone #AA303940) using the peptide sequence tag algorithm. The EST clone was assumed to be incomplete since it lacked a start codon and encoded for a putative protein of only 456 amino acids. The cloning of the full-length Gemin3 cDNA was achieved by hybridization screening of a human leukemia 5′-STRETCH PLUS cDNA library (from CLONETECH) using a subfragment (EcoR1–EcoR1) of the EST clone #AA303940 as a probe. The EST EcoR1 fragment contained a region of the Gemin3 open reading frame (ORF) encoding amino acids 368–548. 12 independent partial cDNA clones with insert sizes ranging from 1–2.5 kb were isolated, all of which contained overlapping regions of the same ORF that encoded for a putative protein of 824 amino acids. Since the clone lacked an in frame stop codon upstream of the start codon, we performed two successive rounds of rapid amplification of 5′-cDNA ends by PCR (5′-RACE PCR) to extend further upstream the cloning of the Gemin3 cDNA. In brief, total mRNA from HeLa-S3 cells was prepared using the TRIZOL reagent (Life Technologies, Inc.). For the generation of cDNA, 2 μg of mRNA was reverse-transcribed using Superscript™II reverse transcriptase (Life Technologies, Inc.). The reaction was primed with Gemin3 cDNA-specific primers, GSPA (5′-TTCCACTTCCAGGCCG) or GSPC (5′-TCTTGGGGCTTTCCTCAGG), in the first or second round of RACE, respectively. The cDNAs were then purified using a GlassMAX spin cartridge, tailed with dCTP and TdT, and amplified by PCR using the Gemin3 specific primer GSP2 (5′-TCTTTCCTCTCCTCCC) and the Universal Amplification primer (UAP) from Life Technologies, Inc. (5′-CUACUACUACUAGGCCACGCTTCGACTAGTAGTAC). PCR products were cloned into the pCR2.1 vector by TA cloning (Invitrogen) and sequenced. The two independent 5′-RACE experiments yielded products terminating at the same 5′ position. The extended 5′ Gemin3 cDNA contained an in frame stop codon upstream of the start codon. By in vitro transcription–translation, we confirmed that the cloned cDNA encoded the full-length Gemin3 protein (see below).

The [³⁵S]methionine-labeled proteins were produced by an in vitro coupled transcription–translation reaction (Promega Biotech) in the presence of [³⁵S]methionine (Nycomed Amersham, Inc.). 6His-Gemin3 and 6His-SMN fusion protein was expressed from a pET bacterial expression system in the Escherichia coli strain BL21(DE3) and purified using nickel chelation chromatography using Novagen His-bind buffer. GST–Gemin3 fusion protein was expressed from a GST expression vector pGEX-5X-3 (Pharmacia Biotech, Inc.) in the E. coli strain BL21 and purified using glutathione-Sepharose (Pharmacia Biotech, Inc.) according the manufacturer's protocol.

Anti-Gemin3 antibodies 11G9 and 12H12 were prepared by immunizing Balb/C mice with 6His-tag COOH-terminal domain of Gemin3 (from amino acids 368–548) purified from nickel chelation chromatography using Novagen His-Bind buffer kit. Hybridoma production, screening, and ascites fluid production were performed as previously described (Choi and Dreyfuss 1984).

Immunoprecipitations of in vitro translated proteins were carried out in the presence of 1% Empigen BB buffer as previously described (Choi and Dreyfuss 1984). Immunoprecipitations of SMN, the Sm proteins, and Gemin3 from cells were carried out using total HeLa lysate in the presence of 1% Empigen BB buffer as previously described (Choi and Dreyfuss 1984). Immunoprecipitations and purifications of the SMN, Gemin2, Sm, and Gemin3 complexes were carried out using total HeLa lysate in the presence of 0.5% Triton X-100 as previously described (Pinol-Roma et al. 1988). For immunoblotting, proteins were resolved on 12.5% SDS-polyacrylamide gels and transferred to a nitrocellulose membrane (Schleicher and Schuell, Inc.) using a BioTrans Model B Transblot apparatus (Gelman Science) according to the manufacturer's instructions. The membranes were then incubated in blocking solution (PBS 5% nonfat milk) for at least 1 h at room temperature, rinsed with cold PBS, and then incubated in blocking solution with primary antibody for at least 1 h at room temperature. The membranes were subsequently washed three times in PBS containing 0.05% NP-40, and bound antibodies were detected using the peroxidase-conjugated goat anti–mouse IgG plus IgM (Jackson ImmunoResearch Laboratories). The antibody-decorated protein bands were visualized by an ECL Western blotting kit (Nycomed Amersham, Inc.) after washing three additional times with PBS containing 0.05% NP-40.

HeLa cells were cultured in DME supplemented with 10% FBS (both from GIBCO BRL). HeLa cells, plated on glass coverslips, were transfected by the standard calcium phosphate method. Following overnight incubation with DNA, cells were washed and fresh medium was added. Transfected cells were fixed and processed by immunofluorescence staining after an additional 24–36 h of incubation.

Immunofluorescence staining was carried out essentially as previously described (Choi and Dreyfuss 1984). Double-label immunofluorescence experiments were performed by separate sequential incubations of each primary antibody diluted in PBS containing 3% BSA, followed by the specific secondary antibody coupled to either FITC or Texas red. All incubations were carried out at room temperature for 1 h. Laser confocal fluorescence microscopy was performed with a Leica TCS 4D confocal microscope. Images from each channel were recorded separately and, where indicated, the files were merged. Antibodies used in these experiments were as follows: mouse IgG1 monoclonal anti-Gemin3 (11G9 and 12H12; this work); mouse IgG1 monoclonal anti-SMN (2B1; Liu and Dreyfuss 1996); rabbit polyserum anti-p80 coilin (R288; Andrade et al. 1991); mouse IgG3 monoclonal anti-Sm (Y12; Lerner et al. 1981); and SP2/O, a nonimmunoglobulin chains secreting mouse hybridoma (ATTC). The rabbit affinity-purified anti-Exon7 antibody was made against the polypetide encoded by the SMN exon7 (Liu et al. 1997).

Purified GST or GST fusion proteins (2 μg) bound to 25 μl of glutathione-Sepharose beads were incubated with 10⁶ cpm of the in vitro translated protein mix in 1 ml of binding buffer (50 mM Tris-HCl, pH 7.5, 200 mM NaCl, 2 mM EDTA, 0.1% NP-40, 2 μg/ml leupeptin and pepstatin A, and 0.5% aprotinin). After incubation for 1 h at 4°C, the resin was washed five times with 1 ml of binding buffer. The bound fraction was eluted by boiling in SDS-PAGE sample buffer, and run on SDS-PAGE. The gels were fixed for 30 min and the signal enhanced by treatment with Amplify solution (Nycomed Amersham, Inc.). For direct in vitro binding (see Fig. 7 C), purified GST or GST–Gemin3 proteins (2 μg) bound to 25 μl of glutathione-Sepharose beads were incubated with 5 μg of purified 6His-SMN or 6His-mB in 1 ml of binding buffer (50 mM Tris-HCl, pH7.5, 100 mM NaCl, 2 mM EDTA, 0.05% NP-40, 2 μg/ml leupeptin and pepstatin A, and 0.5% aprotinin). After incubation for 1 h at 4°C, the resin was washed five times with 1 ml of binding buffer. The bound fraction was eluted by boiling in SDS-PAGE sample buffer, and analyzed by SDS-PAGE and Western blot using a rabbit polyclonal anti-His-tag antibody (Santa Cruz Biotech).

HeLa cells were fractionated according to Dignam et al. 1983. S100 fractions (400 μl of ∼20 mg/ml protein) in buffer F (20 mM Tris-HCl, pH 7.4, 0.1 mM EDTA, 1 mM DTT, 10% glycerol, 500 mM KCl) were loaded on a Superose 6 HR 10/30 column (Pharmacia Biotech, Inc.). The column was then washed with buffer A (20 mM Tris-HCl, pH 7.4, 200 mM NaCl, 2.5% glycerol). Fractions (0.5 ml) were collected, and 30 μl of each fraction was resolved on SDS-PAGE, followed by Western blotting.

The Gemin3 GenBank/EMBL/DDBJ accession number is AF171063.

Immunoprecipitations from [³⁵S]methionine-labeled HeLa cell lysates with anti-SMN and anti-SIP1 mAbs revealed the presence of several protein components in the SMN–SIP1 complex (Liu et al. 1997). Among the proteins that can be coimmunopurified with anti-SMN and anti-SIP1 antibodies, only some of the major low molecular mass proteins, identified as the Sm proteins, have been characterized so far (Liu et al. 1997; Fig. 1A and Fig. B). In addition to SMN, SIP1, and the Sm proteins, there is a doublet at ∼97 kD, and additional bands at 175, 95, 60, and 50 kD that coimmunopurified with the anti-SMN antibody. The two proteins of the 97-kD doublet were eluted from the gel, digested with trypsin, and the resulting peptides sequenced by nano-ES MS as described previously (Shevchenko et al. 1996; Wilm et al. 1996; Fig. 2). In this paper, we describe the molecular cloning and characterization of the high molecular weight protein of this doublet (p105). Several peptides from this band identified a human EST sequence (clone #AA303940) using the peptide sequence tag algorithm (Fig. 2). Several additional cDNA clones were obtained by hybridization screening of a human leukemia 5′-STRETCH PLUS cDNA library, using this EST clone as a probe. We isolated 12 independent partial cDNA clones with insert sizes ranging from 1–2.5 kb, all of which contained overlapping regions of the same ORF. 5′ RACE PCR was used to extend this cDNA further upstream. A cDNA clone containing the longest ORF was constructed and conceptual translation of its nucleotide sequence revealed a potential initiator methionine, preceded by an in frame stop codon. This cDNA encodes a putative protein of 824 amino acids with a calculated molecular mass of 92.2 kD and a pI of 6.5. We then determined that this cDNA encodes the p105 protein that coimmunoprecipitates with SMN (see Gemin3 and SMN Colocalize in Gems). Thus, this is a full-length cDNA clone for a novel component of the SMN complex that we termed Gemin3, for component of gems number 3 (see below). Because of the existence of several unrelated proteins called SIP1 (Mylin et al. 1994; Zhang and Wu 1998; Verschueren et al. 1999), we tentatively rename the SMN-interacting SIP1 (Liu et al. 1997) Gemin2, for component of gems number 2 (SMN is the first component of gems identified; Liu and Dreyfuss 1996). Gemin3 has high amino acid sequence similarities with the RNA helicase core region of the human eukaryotic initiation factor 4A-II (eIF4A-II). eIF4A-II is a DEAD box RNA helicase that belongs to the SFII superfamily of helicase (reviewed in De la Cruz et al. 1999). A scheme depicting the modular structure of Gemin3 and the predicted amino acid sequence of Gemin3 aligned with the sequence of eIF4A-II are presented in Fig. 3. This alignment showed the presence of seven motifs in the Gemin3 protein, which are characteristic of the RNA helicase core region. Database searches with the COOH-terminal nonconserved region did not reveal significant homology to any other protein or any recognizable motifs.

To investigate the interaction of Gemin3 with SMN and to characterize Gemin3 further, we generated mAbs to it by immunizing mice with a purified bacterially produced recombinant 6His-Gemin3 fragment (amino acids 368–548). Two hybridomas, 11G9 and 12H12, were selected for additional studies. Several lines of evidence demonstrate that these hybridomas indeed produce mAbs that recognize Gemin3 specifically. First, both 11G9 and 12H12 specifically immunoprecipitate Gemin3 produced by in vitro transcription and translation from the Gemin3 cDNA, but do not immunoprecipitate similarly produced hnRNP A1 or SMN proteins (Fig. 4 A). Second, the mAb 11G9 efficiently recognized purified 6His-Gemin3 on Western blots, but did not recognize similarly produced and purified 6His-Gemin2 (Fig. 4 B). Finally, on an immunoblot of total HeLa lysate, both 11G9 and 12H12 recognized a single protein of ∼105 kD (Fig. 4 C). mAbs 11G9 or 12H12 did not recognized a specific protein on a Western blot with total mouse 3T3 cell lysate or Xenopus laevis XL-177 cell lysate. However, 11G9 specifically immunoprecipitated a single protein of ∼105 kD from these cell lysates (data not shown), suggesting that Gemin3, like SMN, is conserved in vertebrates.

Indirect laser confocal immunofluorescence microscopy using antibodies 11G9 and 12H12 was performed on HeLa cells to determine the subcellular localization of Gemin3. Fig. 5 A shows that Gemin3 is found throughout the cytoplasm and also displays intense staining of prominent discrete nuclear bodies that are readily discernible by differential interference contrast (DIC; Fig. 5 B). This pattern is similar to that seen for SMN and Gemin2 (Liu and Dreyfuss 1996; Liu et al. 1997), except that the nucleoplasmic staining of Gemin3 is stronger. To determine if the nuclear structures stained by 11G9 are gems or coiled bodies, we performed double-label immunofluorescence experiments using antibodies against Gemin3 and either p80 coilin as a marker of coiled bodies (Andrade et al. 1991) or SMN as a marker of gems (Liu and Dreyfuss 1996; Fig. 5). In many cell lines, gems and coiled bodies entirely overlap by antibody staining, however, in the HeLa PV strain used here, these two bodies are frequently found separate from each other (Liu and Dreyfuss 1996; Matera and Frey 1998). Therefore, we used HeLa PV cells to examine whether Gemin3 is located in gems or in coiled bodies. As can be seen in Fig. 5C and Fig. D, the nuclear structures that contain Gemin3 are clearly distinct from coiled bodies, but Gemin3 completely colocalized with SMN in gems (Fig. 5 E). The colocalization of Gemin3 with SMN strongly supports the conclusion that these two proteins exist as a complex in the cell. Gemin3 is thus the third constituent of gems described so far. Compared with visual observation, the confocal photographs in Fig. 5, slightly overestimates the nuclear staining and underestimates the cytoplasmic staining for Fig. 5C and Fig. D.

To characterize further the Gemin3 complex, immunoprecipitations using anti-Gemin3 mAbs and [³⁵S]methionine-labeled HeLa cells were carried out in the presence of either Triton X-100 or the more stringent detergent, Empigen BB (Matunis et al. 1994). The immunoprecipitated proteins were then analyzed by SDS-PAGE. As references for these immunoprecipitations, we also included an immunoprecipitation with the anti-Sm mAb Y12 (Lerner and Steitz 1979; Lerner et al. 1981) and an immunoprecipitation with the anti-SMN mAb 2B1. As shown in Fig. 6 A, several proteins can be coimmunoprecipitated with Gemin3 and the pattern of immunoprecipitated proteins is very similar to that obtained with the anti-SMN antibody. In addition to Gemin3, SMN, and Gemin2, there are several prominent bands at 175, 95, and 50 kD. The two groups of proteins at 28 and 15 kD have been identified previously as the Sm B/B′, D1-3, E, F, and G proteins of snRNPs (Liu et al. 1997). In addition, there are bands that coimmunoprecipitate only with anti-SMN (at 60 kD) or anti-Gemin3 (at 115 kD) mAbs. As further evidence for the specificity of the antibodies used, the immunoprecipitations were performed in the presence of Empigen BB. Under these conditions, anti-Gemin3 and anti-SMN antibodies immunoprecipitate Gemin3 and SMN proteins respectively (Fig. 6, + Empigen BB, lane 11G9 and lane 2B1, respectively). Interestingly, a protein of 95 kD is still present under these conditions in both of these immunoprecipitations, but not in the control SP2/O immunoprecipitation, suggesting that this unidentified protein interacts tightly with both Gemin3 and SMN.

To confirm the coimmunopurification results, we tested for interaction of Gemin3 with SMN, Gemin2, and the Sm proteins in HeLa cells in vivo by immunoprecipitations and Western blot experiments. The anti-Gemin3 mAb 11G9 was used for immunoprecipitation from total HeLa cell extracts, and these were then resolved by SDS-PAGE and an immunoblot was probed with the anti-SMN antibody (Liu and Dreyfuss 1996). As shown in Fig. 6 C (lane 11G9 IP), 2B1 readily detects SMN in the 11G9 immunoprecipitates, indicating that SMN is coimmunoprecipitated with Gemin3. Because SMN is associated with Gemin2 to form a stable complex in vivo and in vitro (Liu et al. 1997), we also investigated whether Gemin3 could be coimmunoprecipitated with Gemin2. As shown in Fig. 6 C, the anti-Gemin2 mAb 2S7 clearly detects Gemin2 in the anti-Gemin3 11G9 immunoprecipitates (Fig. 6 C, lane 11G9 IP). In a reciprocal experiment, the Gemin3 protein could also be coimmunoprecipitated by the anti-SMN mAb 2B1 (Fig. 6 D, lane 2B1 IP) and the anti-Gemin2 mAb 2S7 (Fig. 6 D, lane 2S7 IP). Because SMN and Gemin2 are found in a complex with the Sm proteins, we asked whether Gemin3 can be coimmunoprecipitated with the spliceosomal snRNP Sm core proteins as well. Fig. 6 D shows that Gemin3 is present in the anti-Sm mAb Y12 immunoprecipitates (Fig. 6 D, lane IP Y12) like SMN and Gemin2 (Liu et al. 1997). No Gemin3, SMN, Gemin2, or Sm proteins were detected in a SP2/O immunoprecipitate (data not shown). These results demonstrate that Gemin3, SMN, Gemin2, and the Sm proteins are associated in vivo in a complex that can be immunoprecipitated by either anti-SMN, anti-Gemin2, anti-Sm, or anti-Gemin3 antibodies.

Further support for the existence in vivo of a complex that contains SMN, Gemin2, and Gemin3 was obtained from gel filtration experiments. HeLa cytoplasmic S100 extract was fractionated on a Superose 6 HR 10/30 high performance gel filtration column and each fraction was subjected to SDS-PAGE, followed by Western blot with anti-Gemin3, anti-SMN, and anti-Gemin2 mAbs. Gemin3, SMN, and Gemin2 comigrated and showed a peak at ∼800 kD, demonstrating that they are components of a large macromolecular complex (Fig. 6 E). Interestingly, a second pool of SMN-Gemin2, lacking Gemin3, is observed in a lower molecular weight complex that peaks at ∼150 kD, suggesting that at least two different SMN–Gemin2 subcomplexes exist in vivo. However, we cannot exclude the possibility that the 150-kD subcomplex corresponds to a fraction of SMN-Gemin2 that dissociates from Gemin3 during cell fractionation and/or chromatography. We reported previously that a SMN–Gemin2 complex migrates at >300 kD after filtration of a cytoplasmic S100 extract on a TSK-GEL G3000-SW column (Liu et al. 1997). The Superose 6 HR 10/30 gel filtration column used here allowed us to obtain a better resolution of the cytoplasmic SMN complex and to better estimate its size as ∼800 kD.

To further analyze the Gemin3 complex, we performed in vitro protein-binding assay between Gemin3 and several components of the SMN complex. For in vitro binding assays, Gemin3 was produced as a fusion protein with glutathione S-transferase (GST), and SMN and Gemin2 were produced and labeled with [³⁵S]methionine by in vitro transcription and translation in rabbit reticulocyte lysate. Purified GST or GST–Gemin3 fusion immobilized on gluthatione-Sepharose were incubated with labeled SMN or Gemin2 proteins. After extensive washing, bound proteins were eluted by boiling in SDS-containing sample buffer and the eluted material was analyzed by SDS-PAGE and detected by fluorography. Full-length SMN, but not Gemin2, bound specifically to immobilized GST–Gemin3 (Fig. 7 A), but not to GST alone (data not shown). To investigate whether Gemin3 interacts with Sm proteins, purified GST or GST–Gemin3 recombinant proteins were used for binding assays with in vitro [³⁵S]methionine-labeled Sm proteins B, D1, D2, D3, E, F, and G (Lehmeier et al. 1994; Herrmann et al. 1995; Raker et al. 1996). The results, shown in Fig. 7 B, demonstrate that the Sm proteins B and D3 bind to GST–Gemin3, whereas there is no detectable binding to GST alone (data not shown). D2 binds Gemin3 only weakly and we note that the profiles of Sm protein binding to SMN and Gemin3 are not identical (see Liu et al. 1997). For example, SMN binds to D1 whereas Gemin3 does not (Fig. 7; Liu et al. 1997; Pellizzoni et al. 1999).

To address the possibility that some component of the rabbit reticulocyte lysate mediates these interactions, wild-type full-length SMN and SmB were produced as recombinant 6His-tagged proteins and incubated with GST or GST–Gemin3. After several rounds of washing, bound proteins were solubilized by boiling in SDS sample buffer, resolved by SDS-PAGE, immunoblotted, and probed with a rabbit polyclonal antibody specific to the 6His-tag. As shown in Fig. 7 C, SMN and SmB bind specifically to Gemin3, but not to GST alone. We conclude that both SMN and SmB interact directly with Gemin3.

To further characterize the interaction between Gemin3 and SMN, we first tested whether SMN carrying two well-characterized mutations found in SMA patients, the Y272C point mutant (SMNY272C) and the exon7 deletion mutant (SMNΔEx7), the major product of the SMN2 gene (reviewed in Burghes 1997; Talbot et al. 1997), are able to interact with Gemin3. SMN wild-type and mutants were produced and labeled with [³⁵S]methionine by in vitro transcription and translation in rabbit reticulocyte lysate. Full-length wild-type SMN bound specifically to immobilized GST–Gemin3 (Fig. 7 D). However, SMNY272C and SMNΔEx7 are severely defective in their ability to bind GST–Gemin3. No detectable binding was observed to GST alone. Similar results were observed using purified recombinant 6His-SMN wild-type and mutant proteins instead of in vitro translated products (data not shown).

SMN oligomerization and Sm binding are not mutually exclusive, and in fact, Sm binding is strongly enhanced by SMN oligomerization (Pellizzoni et al. 1999; Fig. 7 E). To determine whether SMN self-association enhances Gemin3 interaction, GST–SMN, or GST as a control, was preincubated with a molar excess of recombinant 6His-tag SMN to form SMN oligomers. After removing the unbound 6His-SMN by washing, in vitro translated [³⁵S]methionine-labeled Gemin3 and SmB were added and assayed for binding (Fig. 7 E). As expected, SmB binding was strongly enhanced by SMN oligomerization (Pellizzoni et al. 1999), however, Gemin3 binding was not affected.

The unwinding activity of DEAD box RNA helicases may not be sequence specific. The target specificity of these proteins is, at least in some cases, provided by their interaction with specific proteins of the RNP substrate. These interactions appear to be mediated via the unique auxiliary domain that each RNA helicase contains (Hamm and Lamond 1998; Staley and Guthrie 1998). Therefore, we investigated the role of the unique COOH-terminal domain of Gemin3 (amino acids 430–825) in the interaction and cellular localization of this novel DEAD box RNA helicase. To do so, we constructed three deletion mutants of Gemin3 and first tested their binding to GST–SMN. Wild-type and mutant myc-Gemin3 constructs were transcribed and translated in rabbit reticulocyte lysate in the presence of [³⁵S]methionine, and the resultant translated products were assayed for binding to GST–SMN as described above. As Fig. 8 B indicates, the wild-type myc-Gemin3 protein and myc-ΔN368C277Gemin3 mutant proteins interact specifically with GST–SMN, but not with GST alone. The myc-ΔC328Gemin3 and myc-ΔN548 Gemin3 mutant proteins clearly do not interact with GST–SMN. Thus, the COOH-terminal domain of Gemin3 (amino acids 456–547) mediates the interaction of SMN with Gemin3.

We then monitored the expression and cellular localization of the myc-tagged mutants in transfected HeLa cells. Because gems represent a marker of a concentrated pool of the SMN complex in the nucleus, it is likely that the incorporation of a myc-tagged Gemin3 protein into gems results from its interaction with SMN in vivo. Double-label immunofluorescence microscopy, using anti-myc tag antibodies to detect either the transfected wild-type myc-Gemin3 or the myc-ΔGemin3 mutants, and the anti-SMN mAb 2β1 antibody, showed accumulation of the wild-type (data not shown) and myc-ΔN368C277Gemin3 mutants into gems (Fig. 8 C). However, the myc-ΔC328Gemin3 and myc-ΔN548 (data not shown) Gemin3 mutant proteins accumulate essentially in the cytoplasm without any gems or nucleoplasm localization. This strongly suggests that the COOH-terminal region of Gemin3 (amino acids 456–547) is responsible for the localization of Gemin3 into gems.

The molecular characterization of the spinal muscular atrophy gene product, SMN, demonstrated that it is concentrated in novel nuclear structures called gems (Liu and Dreyfuss 1996; Liu et al. 1997). Coiled bodies and gems represent nuclear structures that appear to be involved in RNA metabolism, and in many of the cell lines studied, these two bodies are often found in association (Lamond and Carmo-Fonseca 1993; Gall et al. 1995; Liu and Dreyfuss 1996; Liu et al. 1997; Matera and Frey 1998; Pellizzoni et al. 1998). SMN is also found in the cytoplasm, where, together with its tightly associated partner, Gemin2, it functions in the assembly of snRNP particles (Fisher et al., 1997; Pellizzoni et al. 1998). In the nucleus, SMN is required for pre-mRNA splicing, and likely serves to assemble and maintain the splicing machinery in an active form (Pellizzoni et al. 1998). To perform these functions, SMN must either have an intrinsic activity or recruit other proteins that can actively affect structural transitions to the complex in certain RNP targets. Several factors that have the capacity to serve in such functions, including assembly and disassembly of components of the splicing machinery, have been described. Many of these factors are DEAD/DEAH box RNA helicases that are essential for splicing (reviewed in Staley and Guthrie 1998). Prp43, for instance, is required for the disassembly of the snRNP–intron lariat complex (Arenas and Abelson 1997), Prp22 is needed to release the mature mRNA from the spliceosome (Company et al. 1991), and Prp24 acts as a recycling factor for U4 and U6 snRNP (Raghunathan and Guthrie 1998).

Using a biochemical approach to characterize new components of the SMN complex, we have identified a novel DEAD box RNA helicase termed Gemin3. Gemin3 forms a stable complex with SMN in vivo and in vitro, and it colocalizes with SMN in nuclear gems. Several lines of evidence suggest that Gemin3 and SMN function as a complex in vivo. SMN and Gemin3 can be coimmunoprecipitated and both are present in a large (∼800 kD) complex that also contains Gemin2. Anti-SMN, anti-Gemin2, or anti-Gemin3 mAbs immunoprecipitate the spliceosomal snRNP core Sm proteins, as well as several other unidentified proteins. Gemin3 interacts directly with SMN and with several snRNP Sm core proteins, including B/B′, D2, and D3. In addition, Gemin3 is uniformly distributed in the cytoplasm, where snRNP assembly takes place, and it can be specifically coimmunoprecipitated with the cytoplasmic pool of Sm proteins (data not shown). Together, these findings suggest that Gemin3 may play an important role in spliceosomal snRNP biogenesis.

DEAD box proteins have been found to be involved in many aspects of RNA metabolism, including pre-mRNA splicing, translation, snRNP–snRNP interactions, mRNA degradation, and mRNA transport in eukaryotes and prokaryotes (Company et al. 1991; Ohno and Shimura 1996; Arenas and Abelson 1997; Hamm and Lamond 1998; Staley and Guthrie 1998; De la Cruz et al. 1999). One of the major questions about the function of each DEAD/DEAH box RNA helicase is the identification of the specific RNA target for it. Some of the enzymes of this family can unwind generic RNA substrates in vitro (Laggerbauer et al. 1998; Schwer and Gross 1998; Wang et al. 1998). For these enzymes, the specificity towards particular RNAs, therefore, appears to be determined by factors that interact with their unique auxiliary domains. For example, the DEAH box RNA helicase Prp16 is recruited to the spliceosome via its unique NH₂-terminal domain (Wang and Guthrie 1998). The specific substrate for Gemin3 has not yet been identified and this remains a central question of interest. In a series of preliminary experiments, we have not been able to detect RNA helicase or RNA-dependent ATPase activity for recombinant Gemin3, so far. It is possible that such activity will only manifest itself when Gemin3 is associated with other proteins as part of a complex or that it will be detectable once a specific RNA or RNP target is found. The interaction of Gemin3 with SMN is direct, and we found that amino acids 456–547 of Gemin3 mediate this interaction and, likely as a consequence of this, the localization of Gemin3 to the gems. Thus, we propose that Gemin3 provides the enzymatic activity of the SMN complex to affect structural transitions in its RNA targets.

The SMN protein is capable of forming an oligomer of >400 kD in vitro (Pellizzoni et al. 1999) and we show here that SMN comigrates with an ∼800-kD complex that also contains Gemin2 and Gemin3. It is likely that SMN oligomerization is critical for the nucleation of this large complex. In addition to Gemin3 and Gemin2, several Sm proteins interact with SMN, and we therefore propose that SMN forms a docking platform to bring together, in the appropriate spatial arrangement, the multiple proteins that are involved in the de novo assembly and regeneration of its RNP (e.g., snRNP) substrates. Interestingly, the interaction of SMN with Gemin3 is severely reduced by mutations found in SMA patients, such as the point mutant SMNY272C or the exon 7 deletion. Thus, the formation of the SMN platform seems critical for SMN function because SMA affects both the capacity of SMN to oligomerize, as well as to interact with several Sm proteins (Pellizzoni et al. 1999) and Gemin3. As a likely consequence of these defective interactions, the function of SMN in the regeneration of the splicing machinery is abolished (Pellizzoni et al. 1998, Pellizzoni et al. 1999).

Coiled bodies contain the highest local concentration of p80 coilin and are enriched in components of three major RNA processing pathways: pre-mRNA splicing; histone mRNA 3′ maturation; and pre-mRNA processing (reviewed in Lamond and Earnshaw 1998). Gems contain the highest local concentration of SMN, Gemin2, and Gemin3 and are often found associated with coiled bodies (Liu and Dreyfuss 1996; Liu et al. 1997; and this work). Although the definitive function of these two nuclear bodies has not been completely elucidated, the characterization of their protein and RNA contents represents an important step toward the understanding of their functions. Further studies of Gemin3, a novel DEAD box containing protein and component of gems, should shed light on the functions of the SMN complex and gems.

We thank Dr. Eng M. Tan for the anti-p80 coilin antibody and Dr. Joan Steitz for the anti-Sm Y12 mAb. We are grateful to members of our laboratory for stimulating discussions and, in particular, Drs. Wes Friesen, Zissimos Mourelatos, and Sara Nakielny for helpful discussions and critical comments on this manuscript.

This work was supported by a grant from the National Institutes of Health. G. Dreyfuss is an investigator of the Howard Hughes Medical Institute.