INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

Five evolutionarily conserved small nuclear RNAs (U1, U2, U4, U5, and U6 snRNAs) packaged in small nuclear ribonucleoprotein particles (snRNPs) are essential components of the eukaryotic splicing machinery (31). The protein components of the snRNPs are divided into two classes, the specific proteins associated with one particular U snRNP and the common (or core) proteins shared by the U1, U2, U4, and U5 snRNPs. In humans, this last group is composed of eight proteins (B, B', D1, D2, D3, E, F, G), also called Sm proteins, which bind to a highly conserved structural motif called the Sm site (28). The demonstration that yeast snRNPs are immunoprecipitated by anti-Sm antibodies and the observation that yeast snRNAs are able to assemble with the Xenopus Sm proteins upon microinjection into oocytes suggest that yeast snRNPs contain a set of proteins homologous to the metazoan Sm proteins (38, 51, 53). Biochemical studies with yeast indicated that the different yeast snRNPs contain polypeptides with sizes similar to those of the human core proteins (9). Some of these proteins (SmD1, SmD3, and SmE) were identified in the yeast genome sequence by homology to their human counterparts (3, 39, 41, 47). Independently, a conserved Sm motif was identified in the known Sm proteins (5, 18, 47). Scanning the yeast genomic database revealed the presence of several putative Sm proteins, some of which turned out to represent canonical snRNP core proteins (47, 48). Additional proteins from plants, animals, and fungi harboring the Sm motif were called Sm-like proteins (47). Some of them were shown to be components of the yeast U6 snRNP (5, 47). While the present work was in progress, the sequencing of the protein components from the yeast U1 snRNP was published (34). This confirmed the identity of some of the previously characterized Sm proteins, even though in this study, surprisingly, no SmB homolog was found in the U1 snRNP (or U2 snRNP) particles.

The Sm motif consists of two blocks of conserved amino acids separated by a nonconserved region. Of the 70 residues encompassing the Sm domain, only asparagine at position 55 is invariable. At many positions, the physicochemical property of the amino acid, not the primary sequence, is conserved (47). The Sm motif appears to be an important element for Sm protein function. Indeed, truncation of the Sm motif in either the SmB' or the SmD3 protein abolishes SmB'-SmD3 complex formation (18). This result was probably predictable because for most proteins of the family, the Sm domain covers almost all the polypeptide chain.

Assembly of the polymerase II-transcribed U1, U2, U4, and U5 snRNPs is a multistep process following an ordered pathway in higher eukaryotes (12, 20, 30, 36, 37). These snRNAs are transcribed and exported to the cytoplasm. Sm proteins, which are stored in the cytoplasm (29, 44), then assemble on the snRNA Sm site. This binding allows hypermethylation of the snRNA 7-methyl cap by a methylase to form a methyl-2,2,7-guanosine cap structure (29). The assembled snRNP now possesses a bipartite nuclear localization signal composed of the Sm proteins and the cap structure. This signal will permit the transport of the snRNP to the nucleus. Addition of snRNP-specific proteins to the core snRNP, in the cytoplasm and/or the nucleus, completes the assembly of functional snRNP (22, 33, 35). The importance of studying snRNP biogenesis is emphasized by two recent reports showing that defects in this process may be the cause of a genetic disease, spinal muscular atrophy (13, 27).

The stoichiometry of different Sm proteins in Sm core particles has not been precisely determined. Available data suggest that each canonical Sm protein is present as a single copy in every snRNP although the existence of a dimer of an E-F-G subcomplex has been reported (37). Assembly of the Sm core will therefore require many different heterologous protein-protein interactions as well as protein-RNA interactions. The presence of a conserved sequence motif in Sm proteins suggests that they will have similar secondary and tertiary structures (47). It is therefore likely that the Sm core particle forms a pseudoregular structure where distinct but related interactions will take place between different Sm protein pairs.

Most of the studies on snRNP biogenesis have been performed with metazoan systems. Recent studies on the yeast Sm proteins SmD1, SmD3, SmE, and SmG demonstrated that the four corresponding genes are essential for viability (3, 39, 41, 47). Moreover, depletion of the SmD1 and SmD3 proteins results in decreased levels of U1, U2, U4, and U5 snRNAs and in undermethylation of the U snRNA (39, 41). These results suggest that snRNP biogenesis in yeast and mammals may occur by very similar pathways. However, the extent of similarity at the various steps of snRNA biogenesis is unclear. To obtain further insight into the mechanisms of yeast snRNP biogenesis, we have analyzed interactions between all yeast Sm proteins by using the two-hybrid system. Our goal was initially to define new potential partners of yeast Sm core proteins, to identify interacting partners among the Sm core proteins, and to precisely map the residues involved in Sm-Sm protein interactions.

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

Yeast strains, media, and genetic methods. Two-hybrid screening and assays were performed with the Y190 strain (7) obtained from S. Elledge. The strain, containing pAS2-SmE, was cotransformed with a yeast cDNA library (constructed by M. White, Cold Spring Harbor Laboratory) inserted in the pGAD vector. Putative interacting clones were identified on selective plates containing 40 mM 3-aminotriazole and screened for -galactosidase activity. Positive clones were recovered in bacteria and retransformed in yeast strain Y190. Clones activating the reporter on their own were eliminated. Clones interacting specifically in combination with SmE were partially sequenced, and genes in the yeast genome were identified by using the Blast program (1).

ura3-52 lys2-801 ade2-101 trp1-63 his3-200 leu2-1 sme1::HIS3

ura3-52 lys2-801 ade2-101 trp1-63 his3-200 leu2-1 sme1::HIS3

Plasmids. Plasmids were constructed by standard methods (43). The SmE protein (3) (accession no. X92449 and U55020) used in the two-hybrid system was cloned into the pACTII vector (7, 17) by PCR amplification of the Sm protein coding region by using 5' and 3' oligonucleotides containing BamHI (BAM5, 5'-GAAAAACAAGAGGATCCGAATGTCGAAC-3') and EcoRI (ECO3, 5'-ACTTGCGAATTCTCTATTTCTCTCG-3') restriction sites, respectively. The sequence of the coding region was verified after cloning. A BamHI-EcoRI fragment (blunt ended by Klenow treatment at the EcoRI site) was isolated from pACTII-SmE and cloned into the pAS2 vector (7, 17) digested with SalI, blunt ended by Klenow treatment, and cut with BamHI. The resulting plasmid was named pAS2-SmE.

Screening for SmE mutants. To create a library of SmE point mutants, the SmE coding region was amplified by PCR using pACTII-SmE as the DNA source and the BAM5 and ECO3 oligonucleotides (see above). Extension was done under normal conditions to minimize the occurrence of multiple mutations in the same molecule. BamHI-EcoRI fragments were then gel purified on a 6% acrylamide gel and cloned into the pACTII vector. After amplification in Escherichia coli, the library was used to transform strain Y190 containing pAS2-SmF or pAS2-SmG. Transformants were selected on SC-Trp-Leu plates (SC is synthetic complete medium), replicaplated on nitrocellulose filters, and regrown on selective medium. The filters were then used for the assay of -galactosidase activity. White, noninteracting SmE mutants were recovered, plasmid was extracted, and DNA was sequenced (see Results for further analyses).

Site-directed mutagenesis. Oligonucleotide-directed mutagenesis of the SmE gene was performed by using the megaprimer strategy (15, 49) with appropriate primers. The following mutagenic oligonucleotides were synthesized for this study (5'-3'): D47A, CATGAACTCAGCAAATCCAA; M50A, ACAACATTAGCGAACTCATC; N51A, GATGACAACAGCCATGAACT; E59A, CACAGGAATTGCCACGGCTT; K83V, TTATCGCCTACCAACAGGAT; G84A, GATATTATCGGCTTTCAACA; N86A, CAATGTGATAGCATCGCCTT.

Production of recombinant SmE protein, preparation of anti-SmE antibodies, and Western blot analysis. pGEX-2T- and pGST-containing fusion proteins were transformed in E. coli and induced by the addition of 0.5 mM IPTG (isopropyl--D-thiogalactopyranoside). Recombinant GST and GST fusion proteins were purified by using glutathione-Sepharose beads (Sigma) essentially as described previously (52). Anti-SmE antiserum was prepared in rabbits by using E. coli-produced SmE protein purified by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) after thrombin cleavage of the GST-SmE fusion protein.

In vitro binding assays. In vitro transcription with T7 RNA polymerase and translation of Sm protein genes by using rabbit reticulocyte lysate was done as described in the manufacturer's instructions (Promega-Biotech) in the presence of [³⁵S]methionine. For in vitro protein-protein interaction studies, 1 µg of purified GST fusion proteins or GST alone as a negative control was incubated with 10 µl of glutathione agarose and 20 µl of in vitro-translated [³⁵S]methionine-labeled Sm protein in 200 µl of binding buffer (50 mM Tris HCl [pH 7.4], 150 mM NaCl, 5 mM MgCl₂, 0.1% Nonidet P-40, 100 µg of bovine serum albumin/ml). After 20 min of incubation at 25°C followed by 20 min of incubation at 4°C with constant rotation, the beads were pelleted and washed five times with 1 ml of binding buffer. Bound proteins were fractionated on a Tricine-13% polyacrylamide gel (45). The gel was then dried and analyzed by fluorography.

Northern blots. Total yeast RNA was prepared by using the guanidinium thiocyanate method (54). Electrophoresis, transfer, and hybridization conditions were performed as described previously (2).

-Galactosidase assays. Assays of -galactosidase activity on plates were done as described previously (4). For quantitative analyses in liquid, strains carrying the two-hybrid plasmids were grown to mid-log phase at 26°C in SC-Trp-Leu medium to an optical density at 600 nm of 1.0 to 1.5. Cells were then prepared and -galactosidase activity was measured by using ONPG (o-nitrophenyl--D-galactopyranoside) as described previously (21, 23). Results obtained independently in our collaborating laboratories were consistent even though some slight variations in values of -galactosidase activity were observed (e.g., compare Tables 1 and 2).

Immunoprecipitations and quantitation of cap hypermethylation. Immunoprecipitations were performed as previously described (2). Quantitation was performed by scanning Northern autoradiography and by determining the intensities for each band by using MacBass software. The value, representing cap hypermethylation efficiency, was determined by the formula (P_RNA/S_RNA)_t = 12/(P_RNA/S_RNA)_t = 0 where P_RNA and S_RNA represent intensities of bands in the pellet and supernatant, respectively, 12 h after (t = 12) or before (t = 0) shift to glucose-containing medium.

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

Isolation of proteins interacting with SmE. To characterize new factors implicated at various steps in snRNP biogenesis and function, we decided to use the two-hybrid assay to screen a library of yeast cDNA for sequences interacting with the yeast SmE protein (3). We constructed a plasmid harboring the SmE coding sequence fused to the Gal4p DNA binding domain (GBD) of the pAS2 backbone. We determined that a gene fusion encoding the Gal4p DNA binding domain and the yeast SmE protein complements the lethal disruption of the chromosomal copy of the SmE gene, suggesting that the fusion protein is stable and active and could therefore be used in a screen (data not shown). A strain containing the GBD-SmE fusion protein was transformed with a yeast cDNA library inserted in the pGAD vector. From 2 × 10⁵ transformants, we isolated candidates which coded for fusion proteins that interacted specifically with SmE. We recovered cDNA clones encoding two other Sm core proteins, SmF and SmG. The GAD-SmF and GAD-SmG fusions isolated from the cDNA library start, respectively, 50 and 7 nucleotides upstream of the ATG codon of the corresponding open reading frames. In addition, we recovered a clone encoding the Sm-like protein SmX4, which is a U6-associated protein (47). The GAD-SmX4 fusion isolated starts 32 nucleotides upstream of the ATG codon of the SmX4 coding sequence. All GAD-Sm fusions contained complete coding sequences, with additional amino acids encoded by the 5' nontranslated region at the N terminus2 amino acids for SmG, 16 for SmF, and 10 for SmX4. As a first test to verify that the selected proteins interact, we inverted the orientation of the two-hybrid assay. The SmE protein was fused to the GAD (in the pACTII vector), while the coding sequences for the SmF, SmG, and SmX4 proteins were cloned into the pAS2 vector (see Materials and Methods). Interactions between SmE, SmF, SmG, and SmX4 were quantified by measuring -galactosidase activities for the different combinations. As shown in Table 1, the interactions between SmE and SmF and between SmE and SmG are relatively strong (10 to 20 U) in comparison to that of other known interacting proteins, SNF1 and SNF4 (0.33 U). In contrast, the interaction between SmE and SmX4 appears weak since we obtained only 0.4 to 0.5 U of -galactosidase activity. This value is, however, higher than the value obtained for the SNF1-SNF4 control usually used to test interactions in the two-hybrid system (7). No interactions between SmX4 and SmF or SmG could be detected (Table 1). To verify the specificity of the interaction between SmX4 and SmE, we tested whether SmE also interacts with SmX5, another known Sm-like U6 protein. As shown in Table 1, SmX5 also interacts weakly (0.55 U) with SmE. This result suggests that the putative interaction between SmX4 and SmE may not be specific (see Discussion).

Matrix of protein-protein interactions. The analysis described above demonstrated that Sm-Sm protein interactions can be identified with the two-hybrid assay. We therefore decided to extend this approach to include all Sm proteins. The remaining Sm protein coding sequences, identified by database searches (47, 48), were cloned by PCR from genomic DNA, and the sequences were verified to ensure that no mutation was present before these fragments were transferred to the pAS2 and pACTII vectors. All pairwise combinations of the resulting plasmids, including the control vectors without insert, were introduced into yeast cells. Three independent transformants were picked from each transformation and used to determine what level of -galactosidase is produced under the control of a given combination of Sm proteins. The strain carrying the pAS2-SmF fusion grew very poorly and gave consistently variable results with some combinations of pACTII-derived plasmids, probably owing to the toxicity of this particular hybrid protein. It was therefore excluded from this analysis. Our results are summarized in Table 2. We observed an interaction between the yeast SmD3 and SmB proteins as well as interactions between the SmE and SmG proteins and the SmF and SmE proteins as already described above. SmE-SmG and SmB-SmD3 were tested in two orientations in this particular experiment, and they were positive in both assays. The other combinations of Sm proteins did not induce the production of significant -galactosidase activity (see Discussion), demonstrating the specificity of the assay.

View larger version (50K): [in this window] [in a new window]   FIG. 1.   In vitro binding of yeast Sm proteins. (A) [35S]methionine-labeled SmD3, SmF, and SmG prepared by in vitro transcription and translation were mixed with GST-SmE or GST-SmB produced in E. coli. Proteins bound to glutathione-Sepharose beads were washed, denatured, and separated on gels in Tricine buffer. Input represents aliquots of radioactive proteins corresponding to 25% of that used in each of the binding reactions. The predicted molecular sizes of the proteins are as follows: SmD3, 11.2 kDa; SmG, 8.4 kDa; SmF, 9.6 kDa. Additional bands observed in input lanes correspond to readthrough and premature translation termination products. Exposure times were identical for both panels. (B) [35S]methionine-labeled SmF and SmG were mixed with GST, GST-SmE, and SmE mutant Q22L fused to GST (GST-Q22L) produced in E. coli. Proteins bound to glutathione-Sepharose beads were treated as described for panel A.

Dissection of the SmE protein. The results presented above demonstrate that the two-hybrid system is an assay of choice to study Sm-Sm protein interactions. We decided to analyze these interactions in detail. We chose to study the SmE protein because it interacts with both the SmF and the SmG proteins. We first tested whether the SmE protein could be divided into independent domains. The Sm motif is composed of two conserved regions separated by a variable number of residues with no sequence conservation and which are probably part of a loop (5, 18, 47). We decided to test whether the two conserved regions would be able to function independently in interaction with Sm proteins. The SmE protein was divided in the middle of the loop into N (residues 1 to 66)- and C (residues 67 to 94)-terminal fragments. These fragments (designated SmE-N and SmE-C, respectively) were cloned independently in the pAS2 and pACTII vectors and introduced in yeast in combination with constructs containing either SmE, SmF, SmG, or no insert. The -galactosidase produced by these transformants was then assayed. Interestingly, we did not detect any interaction between the SmE-N and SmE-C fragments in this assay (Table 2), suggesting that the two halves of the Sm domain are not able to function independently.

Isolation of SmE interaction mutants. The results presented above suggest that the Sm domain cannot be divided into smaller functional units. Nevertheless, only a limited number of residues can be involved in Sm-Sm protein contacts. We decided to find residues of the SmE protein which, when mutated, affect protein interaction. SmE was a good candidate for this study because it allowed us to look for mutants defective in interaction with SmF but still able to interact with SmG and vice-versa. We mutagenized the SmE coding region by PCR amplification under mild conditions (see Materials and Methods) to minimize the production of multiple mutants. The amplified DNA was subcloned as a BamHI-EcoRI fragment in pACTII vector to generate a library of GAD-SmE mutant fusion proteins. Y190 yeast strains carrying either pAS2-SmG or pAS2-SmF were transformed with this library. Transformants were replica plated onto nitrocellulose filters that were incubated on SC-Trp-Leu plates for one night before a -galactosidase test was performed. On average, of the colonies obtained, approximately 15% were white on each tested filter. Plasmids were recovered from colonies displaying a white phenotype and sequenced. Plasmids carrying a mutation(s) leading to a single amino acid change were transformed back into Y190 strains carrying either pAS2-SmF or pAS2-SmG and reanalyzed for the corresponding -galactosidase activities on plates. All single mutants and all multiple missense mutants obtained in this experiment were analyzed further. For each mutant, two transformants were used to quantify -galactosidase activities. Sequencing a limited number (i.e., 60) of randomly selected clones from the library also allowed us to identify four additional SmE mutants carrying a single amino acid change. Finally, we used site-directed mutagenesis to substitute seven evolutionarily conserved residues which, with the exception of residue D47, were not found mutated by the two previous strategies (see above). These mutants were also analyzed in the two-hybrid system for interaction with the SmF and SmG proteins.

(i) SmE mutants unable to interact with SmF and SmG (class I). Eight SmE mutants were unable to interact with SmF and SmG both by the plate assay (Table 3) and by quantitative measurement of -galactosidase (Fig. 2). Of these, three mutants are nonsense mutants (W30Stop, G45Stop, and L82Stop) while the others harbor single amino acid substitutions. The mutation in one mutant, Q22L, affects a residue that is not conserved in the Sm protein family, while four other mutants have substitutions in evolutionarily conserved positions: the mutations in mutants V27E, I29K, and I54N change hydrophobic residues into negatively charged, positively charged, and polar amino acids, respectively, while the mutation in mutant G41D changes the highly conserved uncharged glycine at position 41 into a negatively charged aspartic acid. As mentioned above, the GST-Q22L mutant fusion protein is unable to bind SmF and SmG in vitro (Fig. 1B).

View larger version (35K): [in this window] [in a new window]   FIG. 2.   Protein-protein interactions of mutant SmE fusion proteins. (A) -Galactosidase activities shown correspond to averages of two independent transformants assayed in duplicate and are presented relative to those of the wild-type pACTII-SmE interactions with pAS2-SmF and pAS2-SmG. (B) The amino acid sequence of SmE protein is given. Cons., consensus inferred from Sm and Sm-like protein sequences (47). The positions of the point mutants are indicated above the sequence by small circles. Stop mutants are represented by flags.

(ii) SmE mutants interacting with SmF but not with SmG (class II). We recovered a single point mutant (I90R) of the SmE protein that still interacted with SmF but that had lost the ability to interact with SmG (Table 3). This mutant harbors a substitution at a conserved hydrophobic position that replaces a nonpolar isoleucine with a positively charged arginine near the C terminus of the SmE protein. This mutant reduces -galactosidase production to background levels when in combination with pAS2-SmG, while in the presence of pAS2-SmF it reduces -galactosidase levels only to 40% of the value obtained with the wild-type protein (Fig. 2). A multiple mutant producing a similar phenotype resulted from the presence of a stop codon at residue 89, producing a deletion of the last 5 amino acids (Table 3). Quantitative -galactosidase assays indicate that this mutant retains some ability to interact with SmF (30% of the wild-type -galactosidase level) but is totally unable to interact with SmG (Fig. 2).

(iii) SmE mutants interacting specifically with SmG but not with SmF (class III). Two SmE mutants with mutations in conserved residues at positions D47 and L81 were recovered from our genetic screen as specifically interacting with SmG (Table 3). Mutant D47H carries a change of aspartic acid at position 47 to a positively charged histidine, whereas mutant L81P possesses a nonpolar proline in place of leucine at position 81. We also characterized the D47A mutant created by site-directed mutagenesis. Both D47H and D47A SmE mutants have identical phenotypes when tested for interaction properties and viability (Table 3). Quantitation of -galactosidase activities indicates that mutant L81P retains almost a wild-type interaction capacity for SmG whereas mutants D47H and D47A reduce -galactosidase levels in the presence of pAS2-SmG to 40% of the value obtained with the wild-type protein (Fig. 2). In all three cases, no interaction above background could be observed with SmF (Fig. 2).

(iv) SmE mutants unaffected in SmF and SmG binding (class IV). Class IV is the largest group, with 13 mutants: 6 mutants were obtained by site-directed mutagenesis of residues at conserved positions in SmE, 4 mutants were isolated by random sequencing, and 3 mutants, which were selected by the in vivo screen, were actually not affected in their interaction properties after retransformation into pAS2-SmF- and pAS2-SmG-containing strains. As shown in Table 3, mutants with mutations in residues that are poorly conserved in the Sm protein family, such as M3T, KE6, M10V, G45V, and P76S, are still able to interact with both SmF and SmG. Interestingly, mutants with mutations in residues that are evolutionarily conserved, such as L31S, M50A, N51A, I80F, G84A, and N86A as well as the double mutant E59A-N63D, are able to bind to their partners. This is also the case for mutant K83V, which has a mutation at the SmE specific position 83. We noticed nevertheless that the levels of -galactosidase produced in the presence of pAS2-SmF were somewhat lower for most of these mutants than for the wild type (Fig. 2).

Growth phenotypes of SmE mutants. We next tested the ability of the mutated GAD-SmE fusion proteins to complement yeast strain YRB15 disrupted at the SME1 locus and carrying a wild-type copy of the SME1 gene on a URA3-marked plasmid (3). After transformation with SmE mutant fusion protein constructs, cells were grown on plates containing 5-fluoroorotic acid (5-FOA) to select against the URA3-marked plasmid. As shown in Table 3, all mutant fusion proteins in classes I, II, and III, with the exception of I54N, are unable to support growth. Of the class IV mutants, which are able to interact with SmF and SmG, all complement the SME1 disruption, with the single exception of the G45V substitution mutant. Interestingly, the mutant G45V is lethal as a GAD fusion protein, although it is able to interact with both SmF and SmG (Fig. 2).

View larger version (31K): [in this window] [in a new window]   FIG. 3.   Western analysis of mutant GAD-SmE fusion proteins. Similar amounts of cell extracts from Y190 transformants carrying the indicated mutated SmE fusion proteins were fractionated by SDS-PAGE and immunoblotted with anti-SmE antibodies. Control extracts were made from cells carrying the wild-type SmE fusion protein pACTII-SmE (lanes 1 and 16) or pACTII vector alone (lanes 2 and 17).

Characterization of SmE mutant defects. The lethality of the SmE mutants producing normal protein levels may be a consequence of defects in snRNP stability and/or formation. Alternatively, the mutated protein might allow snRNP biogenesis but prevent snRNP function in the spliceosome or intranuclear snRNP localization. To distinguish these possibilities, we analyzed the snRNA levels in strains expressing the mutated alleles.

View larger version (31K): [in this window] [in a new window]   FIG. 4.   Defects of SmE mutant G45V. (A) Amounts of spliceosomal snRNAs are decreased after shift of SmE mutant G45V to the restrictive temperature. Northern analysis was performed on RNA isolated from strains carrying wild-type (pSmE) or mutant G45V (pSmE-G45V) SmE genes before and 4 h after shift to 37°C. Probes were oligonucleotides complementary to U4, U5, and U6 snRNAs. The US snRNA is represented by two bands corresponding to a long (U5l) and a short (U5S) form. (B) Stability of SmE mutant G45V protein is affected at the restrictive temperature. Strains carrying wild-type (pSmE) or mutant G45V (pSmE-G45V) genes under the natural promoter as the sole source of SmE protein were grown and shifted for the indicated times at 37°C. Similar amounts of cell extracts were separated by SDS-PAGE in Tricine buffer and immunoblotted with anti-SmE antibodies.

View larger version (60K): [in this window] [in a new window]   FIG. 5.   snRNA stability in SmE mutants. RNA was extracted from endogenous sme1-disrupted strains bearing the GAL1::SME1 allele and either the indicated mutant SmE genes under the natural promoter (plasmid pUN-PrE) (lanes 1 to 8), the wild-type SmE gene (lanes 9 and 10), or vector pUN alone (lanes 11 and 12) before and 12 h after shift to glucose (GAL1-repressing conditions). RNA was separated on a 6% polyacrylamide-8 M urea gel, subjected to Northern analysis, and hybridized with probes specific for the yeast U4, U5, and U6 snRNAs. The poor transfer of the U1 and U2 snRNAs precluded their analyses in this experiment.

Cap hypermethylation in SmE mutants. In higher eukaryotes, it has been shown that the core complex is necessary for cytoplasmic cap trimethylation and nuclear localization of snRNPs (11, 29, 36). Since our previous results suggest that the I54N, L81P, and I90R mutants are able (at least in part) to assemble into snRNPs, the defect of these mutants may be explained by a deficiency in cap hypermethylation. To test this hypothesis, total RNA extracted from cells carrying a given SmE mutant gene before and after depletion of wild-type SmE protein was immunoprecipitated with antitrimethylguanosine (TMG) antibody and subjected to Northern analysis. As shown in Fig. 6, in a wild-type strain, essentially all the snRNAs are recovered in the pellet independently of the growth medium (containing galactose [lane 10] or glucose for 12 h [lane 12]) whereas very few snRNAs remain in the supernatant (lanes 9 and 11). Quantitation of the results shown in Fig. 6 is given in Table 4. For the wild type, the hypermethylation efficiency, which represents the relative proportion of snRNAs present in the pellet versus that in the supernatant after (t = 12) and before (t = 0) shift to glucose, is 0.97 to 1.17, indicating no effect on hypermethylation. In contrast, Fig. 6 shows that after depletion of the SmEp (cells carrying the pUN10 vector alone), some snRNAs remain in the supernatant 12 h after shift to glucose (lane 15). As shown in Table 4 (row 2), the efficiency of hypermethylation is around 0.51 to 0.61, indicating that depletion of SmE protein affects the snRNA hypermethylation process. Similar results have been obtained following depletion of the yeast SmD1 and SmD3 core proteins (39, 41). Concerning the SmE mutants and as shown in Fig. 6 and Table 4, analysis of the relative proportions of U5 and U4 snRNAs present after and before depletion of wild-type SmE protein in L81P and I54N SmE mutants reveals that hypermethylation is affected in the L81P mutant (efficiency of 0.49 to 0.53) but not in the I54N mutant strain (efficiency of 0.89 to 1.05). These results suggest that hypomethylation of snRNAs may be the defective step in SnRNP assembly in SmE mutant L81P but not in mutant I54N.

View larger version (32K): [in this window] [in a new window]   FIG. 6.   Decreased 5' cap hypermethylation of spliceosomal U snRNAs in SmE mutant L81P. RNA blot analysis after immunoprecipitation with anti-TMG antibody. Endogenous sme1-disrupted strains containing the GAL1::SME1 allele and plasmids carrying the wild-type SmE gene, the mutant L81P and I54N genes, or the vector pUN alone were shifted to glucose-containing medium for 0 and 12 h. RNA was extracted and immunoprecipitated with anti-TMG antibodies. Supernatants (S) and pellets (P) were subjected to Northern analysis using U4 and U5 as probes. For mutants I54N and L81P, the amounts of snRNA were increased twofold to compensate for the observed slight decrease in U snRNAs 12 h after shift to glucose (see Fig. 5). Quantitation of the results is presented in Table 4.

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

We have utilized the two-hybrid technique to examine protein-protein interactions between yeast Sm proteins. Our results can be summarized as follows. First, we identified SmE interacting partners in a two-hybrid screen. Second, we analyzed the interaction between Sm core proteins, giving us information about the protein-protein contacts occurring inside the Sm complex. Finally, we selected mutants of the SmE protein unable to interact with SmF and/or SmG and completed this analysis by mutating conserved positions in the SmE protein. This allowed us to identify residues of the SmE protein involved in various functions and protein-protein interactions.

SmE interacting partners. We have isolated potential binding partners for the spliceosomal core SmE protein by screening a yeast cDNA library by use of the yeast two-hybrid system. The three clones (SmX4, SmF, and SmG) that we isolated are members of the Sm and Sm-like protein families (5, 18, 47). The interaction detected between SmX4 and SmE was the most unexpected since these proteins reside on different snRNPs (47). One explanation for this result is that because all Sm and Sm-like proteins are likely to have similar shapes (47) and because Sm proteins interact with other Sm proteins, members of the family will display a low affinity for each other. This interpretation suggests that the SmX4-SmE interaction detected is nonspecific. This may be the case since we detected an equivalent weak interaction between SmE and SmX5, another Sm-like U6-associated protein (Table 1). The detection of a low level of interaction of the SmE protein with itself (Table 2) further supported the possibility that the two-hybrid system allows the detection of weak nonspecific interactions between Sm proteins. On the other hand, we did not detect interaction of yeast SmX4 with either SmF or SmG. Furthermore, the levels of -galactosidase detected in the two-hybrid system with the SmE-SmX4, SmE-SmX5, and SmE-SmE combinations exceeded by far the background levels observed with any combination of different Sm core proteins (Table 2). However, although potential interactions between these proteins could take place in the U4/U6 and U4/U6.U5 particles, the biological relevance of these interactions, if any, are presently unclear.

Protein-protein interactions inside the Sm core complex. We originally identified interactions of the yeast SmE with both SmF and SmG in a two-hybrid screen. To extend this observation, we identified the remaining yeast core Sm proteins in the database and analyzed interactions between all possible pairwise combinations of these proteins. Our results are consistent with previous data indicating the existence of multimeric complexes of mammalian Sm proteins such as D1.D2, E.F.G, D1.D2.E.F.G, and B/B'.D3 (18, 25, 37). Indeed, we show here the existence in vivo and in vitro of specific interactions between yeast Sm proteins SmE-SmF, SmE-SmG, and SmB-SmD3. Furthermore, no interaction between yeast SmF and SmG could be detected either in vivo (Table 2) or in vitro by using in vitro-translated SmF protein and purified GST-SmG fusion protein (data not shown). The SmB-SmD3 interaction has also been identified recently in an exhaustive two-hybrid screen by using yeast SmD3 as bait (14). Interestingly, the yeast SmB protein was not found in purified yeast U1 and U2 snRNPs (34). Our interaction results suggest that the yeast SmB protein would be part of the Sm core complex. This has been confirmed by showing that the yeast SmB protein coimmunoprecipitates the U1, U2, U4, and U5 snRNAs (40). We failed to detect an interaction between SmD1 and SmD2 (Table 2). This does not mean that SmD1-SmD2 interaction does not occur in yeast, and our failure to detect such a complex in vivo might be explained by poor folding of the SmD1 and/or SmD2 fusion proteins. Indeed, the yeast SmD1 is highly divergent (41, 42), with the insertion of 27 residues in the variable region separating two main conserved blocks of the Sm motif (47) (see below). Because of this insertion, the yeast SmD1 protein might be able to fold or stably interact with other proteins only in the presence of its cognate RNA binding site. Further studies will address this possibility. However, taken together, our results indicate a conservation of the protein-protein interaction sites in the yeast and mammalian Sm proteins. This suggests that the snRNP structure will be conserved and that snRNP assembly will follow similar pathways in the two species.

Mapping interaction surfaces of the yeast SmE protein. All the Sm proteins contain a conserved structural domain which extends over 100 conserved residues interrupted by a nonconserved region of up to 30 amino acids (18, 47). A previous analysis of the human SmB'-SmD3 complex suggested that the Sm domain may be involved in protein-protein interactions (18). However, these studies were limited since the mutants used therein were truncation mutants which might have prevented the folding of an independent interaction domain. Our results support this idea. Indeed, we first analyzed the interaction properties of the two conserved regions of the Sm domain of the SmE protein expressed independently. Neither of these two domains were able to interact with other Sm proteins, and they did not interact with each other (Table 2). These data support the results from structure prediction that suggest that the nonconserved residues are located in a loop at the surface of a single globular domain (47). Further support for this model came from sensitive database searches which indicated that the two conserved regions are always associated in cis (47). These data suggest that the Sm domain should not be divided into two substructures (14, 18) but rather should be regarded as a single entity (47).

Assembly of snRNPs in the SmE mutants. We have shown that the mutations in only four SmE mutants (V27E, W30Stop, G41D, and G45Stop) give rise to unstable proteins and that mutants I54N, L81P, and I90R are assembled into snRNPs. Our results also suggest that hypomethylation may be the defective step in snRNP assembly for mutant L81P but not for mutant I54N. In addition, with two exceptions (G45V and I54N), we have shown a perfect correlation between the viability phenotypes observed for GAD-SmE mutant fusion proteins and those for mutant proteins produced from their natural promoter, suggesting that properties carried by the GAD domain do not interfere with SmE function in the majority of the mutants. This is not the case for the G45V and I54N mutants. Indeed, the presence of the GAD domain in the G45V mutant impairs function of the fusion protein, whereas the G45V mutant produced from its natural promoter is functional at a permissive temperature (Table 3). For mutant I54N, the GAD fusion protein is functional, in contrast to the protein produced from its natural promoter. This is surprising since in the two-hybrid system, this mutant shows background-level interaction with SmF and interacts only very poorly (10% of wild-type level) with SmG (Table 3 and Fig. 2). The fact that the GAD-I54N fusion protein is functional suggests that it is assembled into snRNPs able to follow a complete assembly pathway. Therefore, the GAD-I54N protein must present residual low interactions with both SmF and SmG which must be sufficient to allow core assembly. How can we explain the lethality of the SmE-I54N protein produced from its natural promoter? An explanation is that the GAD fusion positively influences the folding of the I54N mutant. Another possibility is that the I54N mutant protein is unstable in contrast to the GAD-I54N fusion protein. This does not seem to be the case, since our studies suggest that, at least in part, the I54N mutant assembles into snRNPs (Fig. 5). If snRNP biogenesis proceeds in yeast and mammals in similar manners, the low residual interaction with SmF and SmG of this mutant may not be sufficient to allow the formation of the bipartite signal for nuclear localization. Hence, nuclear targeting of the snRNPs would be impossible, explaining the lethality of the cells carrying the I54N mutant. Conceivably, the nuclear localization signal present in the GAD-I54N fusion protein could help to target the fusion protein to the nucleus. Further analysis is needed to resolve this issue.