Evolution of an RNP assembly system: A minimal SMN complex facilitates formation of UsnRNPs in Drosophila melanogaster

Abstract

In vertebrates, assembly of spliceosomal uridine-rich small nuclear ribonucleoproteins (UsnRNPs) is mediated by the SMN complex, a macromolecular entity composed of the proteins SMN and Gemins 2–8. Here we have studied the evolution of this machinery using complete genome assemblies of multiple model organisms. The SMN complex has gained complexity in evolution by a blockwise addition of Gemins onto an ancestral core complex composed of SMN and Gemin2. In contrast to this overall evolutionary trend to more complexity in metazoans, orthologs of most Gemins are missing in dipterans. In accordance with these bioinformatic data a previously undescribed biochemical purification strategy elucidated that the dipteran Drosophila melanogaster contains an SMN complex of remarkable simplicity. Surprisingly, this minimal complex not only mediates the assembly reaction in a manner very similar to its vertebrate counterpart, but also prevents misassembly onto nontarget RNAs. Our data suggest that only a minority of Gemins are required for the assembly reaction per se, whereas others may serve additional functions in the context of UsnRNP biogenesis. The evolution of the SMN complex is an interesting example of how the simplification of a biochemical process contributes to genome compaction.

Splicing of pre-mRNAs is catalyzed by the spliceosome, a macromolecular machine consisting of a large number of protein factors and the uridine-rich small nuclear ribonucleoproteins (snRNPs) U1, U2, U4/6, and U5. The biogenesis of these particles occurs in a stepwise manner. First, nuclear-transcribed, m⁷G-capped snRNAs U1, U2, U4, and U5 are exported into the cytoplasm, where a conserved sequence motif in these RNAs (Sm site) serves as a binding platform for the seven Sm proteins B/B′, D1, D2, D3, E, F, and G. As a consequence, a ring-shaped Sm core domain is formed. This domain is crucial for subsequent steps in the biogenesis of UsnRNPs, such as formation of the hypermethylated m^2,2,7G cap and import of the assembled particle into the nucleus. At a yet to be defined step, additional factors are recruited to form the mature UsnRNP particles that function in splicing (1).

Previous studies have shown that Sm proteins bind spontaneously, albeit in a hierarchical manner, onto UsnRNAs in vitro (2, 3). However, in cellular extracts, this process depends on ATP and the activity of the multisubunit SMN complex (4–6). Recently, a systematic interaction study on the human SMN complex has established its basic architecture (7). A modular composition was deduced where the three factors SMN, Gemin2, and Gemin8 form the backbone of the entire complex. Onto this core, the peripheral building blocks Gemin3/4 and Gemin6/7/UNRIP bind to form the functional unit. In support of this modular architecture, Gemin-containing subcomplexes have been identified composed of SMN/Gemin2, Gemin3–Gemin5, and Gemin6/7/UNRIP (8).

The SMN complex not only functions in the assembly of the Sm core domain, but also influences additional steps in the biogenesis pathway of UsnRNPs. One such step is the nuclear import of the assembled UsnRNP, which is mediated by the SMN complex (or parts thereof) in conjunction with the import factor importin β (9). In addition, specific UsnRNP proteins and the cap hypermethylase Tgs1 have been found in association of the SMN complex (10). This observation indicates that the SMN complex coordinates various events during UsnRNP biogenesis by assuming the role of a binding platform for the respective assisting factors.

The multisubunit composition of the human SMN complex has impeded the mechanistic dissection of the UsnRNP assembly process. Thus, although RNA interference studies indicated essential roles of several Gemins in the assembly reaction (11–13), their precise contributions remain unclear. To facilitate mechanistic studies and to gain insight into the evolution of the SMN complex, we have mined genomic databases for organisms that lack individual Gemins and hence may contain a simpler assembly machinery. Indeed, this was the case for different organisms including dipterans. We chose to further investigate Drosophila because of the wealth of genetic resources. An affinity chromatography strategy has permitted us the purification of an assembly-active complex composed of SMN and Gemin2 only. Remarkably, this complex not only facilitated assembly of the Sm core domain but also discriminated between cognate and noncognate RNAs. Thus, our combined bioinformatic and biochemical approach revealed that the assembly reaction requires only two core proteins in vitro, even though SMN complexes from most metazoans are of considerable complexity. We speculate that Gemins 3–8 have been recruited to the SMN complex in the course of evolution to integrate assembly with additional steps in the biogenesis of UsnRNPs.

Results

An Elaborate SMN Complex Is Characteristic of Metazoans.

To understand how the UsnRNP assembly machinery has evolved, we performed homology searches for all Gemin proteins constituting the human SMN complex in genomic databases of a variety of organisms (see Experimental Procedures for details). Because of its diverse functions (14, 15) and its transient cytoplasmic interaction with the SMN complex, the UNRIP protein has been excluded from this analysis.

SMN and Gemin2 orthologs (termed Yab8p and Yip1p) but no other Gemins can be found in the fungus Schizosaccharomyces pombe [see Fig. 1; for protein identifiers see supporting information (SI) Table S1]. Importantly, both orthologs interact physically and may hence form a functional unit (16). In Saccharomyces cerevisiae, however, only the distantly related Gemin2 ortholog Brr1p, but no SMN ortholog, could be identified. Brr1p has been proposed to be an ortholog of human Gemin2 in an early publication (5). However, because of its limited homology to Yip1p this finding has been questioned (16). Taking advantage of the dramatically increased genome databases and novel search algorithms (PSI BLAST), Brr1p can be defined as the single significant homolog of human Gemin2 (NP_003607.1, PSI BLAST third iteration, E = 3 × 10⁻⁵) and S. pombe Yip1p (NP_594775.1, fourth iteration, E = 3 × 10⁻⁵). Because reciprocal searches further support this homology, Brr1p is the ortholog of human Gemin2. Thus, S. cerevisiae retained only one Gemin and hence is unlikely to form a functional SMN complex. Interestingly, like S. pombe, the plants Arabidopsis thaliana and Oryza sativa contained orthologs of SMN and Gemin2 only. We therefore conclude that SMN and Gemin2 represent the most primitive and ancestral version of the SMN complex. Surprisingly, the genome of Dictyostelium discoideum, a facultative multicellular organism, encoded orthologs of Gemin3 and Gemin5. Given that D. discoideum is basal to fungi and metazoans, the bioinformatic data suggest that both Gemins have been lost during evolution in the fungi branch but were retained in metazoans. Interestingly, the presence of a Gemin5 ortholog in Ostreococcus tauri but its absence in land plants indicates an independent gene loss in this phylum. Moreover, we found SMN and Gemins 2, 3, 5, 6, 7, and 8 to be present in the cnidarian Nematostella vectensis, a basic metazoan. Gemin4 first appeared in the sea urchin Strongylocentrotus purpuratus, whereas it is absent in all ecdysozoans under study. This suggests that Gemin4 has joined the SMN complex only recently in evolution, most likely with the appearance of deuterostomians. Consequently, it was found to be part of the SMN complex in vertebrates such as Danio rerio but also cephalochordates like Branchiostoma floridae and Ciona intestinalis (urochordates). Thus, plants and some fungi possess a core complex composed of SMN and Gemin2 only, whereas an elaborate SMN complex has developed only in animal branches by addition of Gemin proteins.

Fig. 1.

Evolution of the SMN complex. Complete genome assemblies of indicated organisms have been screened for orthologs of human SMN complex components. The presence of the proteins was mapped on a published phylogenetic tree (38). A homolog of Gemin6 was found in the algae O. tauri, but the homology was restricted to the C-terminus and therefore is not shown. Gemin5 orthologs in dipterans are evolving significantly faster than in other organisms. Because this may indicate a change of function, they are shown as squares.

Absence of Most Gemin Orthologs in Genomes of Dipterans.

The bioinformatic data indicated an evolutionary trend in the animal kingdom toward a multisubunit SMN complex. Interestingly, however, we failed to identify orthologs of most Gemins in the dipterans Drosophila melanogaster and Anopheles gambiae although they were present in closely related Apis mellifera and Nasonia vitripennis. We have restricted our further analysis to D. melanogaster in this study. Besides the known SMN ortholog (17, 18), we found a Gemin2 ortholog encoded by CG10419 and putative orthologs of Gemin3 (Dhh1) and Gemin5 (Rigor mortis). Dhh1 protein shows high conservation in the N-terminal DEAD box helicase domain but possesses a diverged C terminus. Rigor mortis displays moderate homology to Gemin5 over the entire protein length. A phylogenetic analysis revealed that both evolve significantly faster (P < 0.001, see also Experimental Procedures) than their orthologs in other organisms. This released evolutionary pressure might indicate the emergence of a novel function or the loss of a common one for these factors. These data suggest that D. melanogaster possesses a much simpler SMN complex as compared with vertebrates.

Biochemical Investigation of Dhh1 and Rigor Mortis.

To investigate whether Dhh1 and Rigor mortis have retained their function in the context of the D. melanogaster SMN (dSMN) complex, we made use of a novel epitope tag. This tag consists of the first 30 aa of human SMN protein, which are specifically recognized by the monoclonal antibody 7B10 (19). Importantly, competition with synthetic peptide comprising this epitope allows native elution of tagged proteins from this antibody. Because D. melanogaster SMN protein lacks these 30 aa, we constructed a plasmid allowing the expression and subsequent purification of a protein fused to this epitope (termed TagIt epitope) after stable transfection of Schneider2 cells. In a TagIt-Dhh1 affinity purification, only small amounts of dSMN protein could be detected under physiological conditions but not at salt concentrations exceeding 250 mM (Fig. 2A). Thus, Dhh1 is only weakly associated with dSMN. Similarly, we have investigated the role of Rigor mortis (Fig. 2B; see also Fig. S1). No binding of Rigor mortis to dSMN has been observed, arguing against a stable association of this protein with the dSMN complex. These data suggest that Rigor mortis either functions in UsnRNP core formation in a manner different from vertebrate Gemin5 or has completely lost its function in the pathway of UsnRNP biogenesis.

Fig. 2.

TagIt–dSMN binds dGemin2 and Sm proteins. (A) Coimmunoprecipitation of TagIt–Dhh1 with dSMN from Schneider2 cells at increasing salt conditions (lanes 4–6). The coprecipitated dSMN protein was detected by Western blot. Lanes 1–3 show control immunoprecipitations. (B) Extracts from cells expressing TagIt–Rigor mortis (lanes 4–6), TagIt–dSMN (lane 7), or no tagged protein (lanes 1–3) were immunoprecipitated with antibody 7B10. Immunoprecipitates were analyzed by Western blotting with antibodies against dSMN and Rigor mortis, respectively. (C) Extracts from Drosophila Schneider2 cells stably expressing TagIt–dSMN were separated on glycerol gradients and analyzed by Western blotting with 7B10 (Upper) an anti-dGemin2 antibody (Lower). Estimated sedimentation value is indicated. (D) Isolation of TagIt–dSMN from Schneider2 extracts. Proteins were separated by SDS/PAGE under reducing (lanes 1 and 2) and nonreducing (lane 3) conditions and visualized by silver staining. Lane 1 shows a control elution of nontransfected cells. The indicated proteins were identified by mass spectrometry. (E) Immunoblot analyses of TagIt–dSMN (Upper) and TagIt–dGemin2 (Lower) purifications with indicated antibodies. Lanes 1 and 3 show mock controls.

Biochemical Isolation of the Drosophila SMN Complex Using the TagIt Epitope.

To gain detailed insight into the composition of the D. melanogaster SMN complex, a TagIt–dSMN-expressing Schneider2 cell line was generated. Importantly, TagIt–dSMN was incorporated into a high-molecular-weight complex [≈20 Svedberg (S) units] that also contained Gemin2 (Fig. 2C). This implied that the tagged dSMN protein engages in interactions similar to those of its endogenous counterpart. We next affinity-purified the SMN complex from extracts by means of 7B10 affinity chromatography (see Experimental Procedures for details). Affinity-purified proteins were separated by SDS-PAGE under reducing and nonreducing conditions and identified by protein mass spectrometry and Western blotting (Fig. 2 D and E). Whereas the tagged SMN protein and its interactor dGemin2 could be readily identified, neither Dhh1 nor Rigor mortis was found under the purification conditions applied here.

It is known that the human SMN complex consists of the core machinery (i.e., SMN and Gemins) as well as the transiently interacting substrates that are transferred onto the UsnRNA during assembly. These are the Sm proteins and some UsnRNP-specific proteins. Strikingly, the entire set of Sm proteins, namely SmB, SmD1 (gene snRNP69D), SmD2 (CG1249), SmD3, SmE (CG18591), SmF (DebB), and SmG (CG9742), was prominently present in the elution. Furthermore, we also found the UsnRNP-specific factors U1 70K (20), U2A′ (21), the U2B″/U1A ortholog SNF (22), and the ortholog of the U5 specific protein (CG4849) U5 116kD (23) reproducibly in the purified complex. However, the abundance of these specific proteins varied among preparations and was often substoichiometric.

During UsnRNP assembly, the SMN complex physically contacts the UsnRNAs (4). In vertebrates, this interaction has been proposed to be mediated, at least in part, by Gemin5 and to occur in the cytoplasm (12). Interestingly, despite the absence of Rigor mortis in the TagIt–dSMN complex, snRNAs U1, U2, U4, and U5 were specifically coprecipitated with dSMN and dGemin2 antibodies from total Schneider2 cell extract (Fig. 3A). Identical results were obtained when the SMN complex was purified from the cytosol, where SMN is predominantly localized (Fig. 3B and Fig. S1B). Hence, in D. melanogaster, the SMN complex is sufficient to recruit a set of substrate proteins similar to those in vertebrates. In addition, the complex interacts specifically with U snRNAs in the cytoplasm, which reflects a situation previously observed in Xenopus laevis oocytes (4).

Fig. 3.

dSMN complex contains UsnRNAs and is active in UsnRNP assembly. (A) Northern blot analysis revealed snRNAs U1, U2, U4, and U5 but not Met-tRNA_i in anti-dSMN (lane 3) and anti-dGemin2 (lane 3 and 4) immunoprecipitates from Schneider2 total cell extract. Lane 1 shows the extract before immunoprecipitation, and lane 2 shows a control immunoprecipitation with a normal rabbit serum (NRS). (B) The same set of snRNAs was purified from cytosolic extract of TagIt–dSMN-expressing cells. U4 snRNA was detected after longer exposure of the film. (C) In vitro transcribed U1snRNA, U1ΔSm, and U85scaRNA were incubated with purified TagIt–dSMN complex at the indicated temperatures (lanes 2–4, 6–8, and 10–12) and separated by native gel electrophoresis. Lanes 1, 5, and 9 show the indicated RNAs in the absence of SMN complex; in lanes 4, 8, and 12 monoclonal antibody Y12 was added after the assembly reaction had been completed. The assembled dU1 Sm core domain and the supershift are indicated by arrows.

Isolated SMN Complex Is Active in UsnRNP Assembly.

A series of studies has suggested that Gemins 2, 3, 4, and 5 are essential players in the assembly reaction in vertebrates (7, 12, 13). Hence, we tested whether the minimal SMN complex in D. melanogaster is sufficient to mediate formation of the UsnRNP core domain. To this end, ³²P-labeled dU1snRNA was incubated with affinity-purified dSMN complex (Fig. 3C). dU1snRNA lacking the Sm site (dU1ΔSm) and an unrelated small Cajal body-associated RNA [U85scaRNA (24)] served as negative controls. The reaction products were analyzed by native gel electrophoresis. Incubation of dU1snRNA but not the control RNAs with the dSMN complex at 25°C led to the formation of a specific complex whose formation was inhibited at 4°C (Fig. 3C, lanes 2 and 3). This complex could be shifted to lower mobility upon incubation with the anti-Sm antibody Y12 (Fig. 3C, lane 4). Thus, isolated dSMN complex mediates UsnRNP core assembly.

Reconstituted dSMN Complex Binds Sm Proteins and Catalyzes UsnRNP Assembly.

Because we could not exclude that substoichiometric factors contributed to UsnRNP core formation, we reconstituted the SMN complex from recombinant proteins, assuming that it consists of SMN and Gemin2 only. His-dSMN and His-GST-dGemin2 were coexpressed in Escherichia coli and purified on Ni-NTA-Sepharose. After immobilization on glutathione-Sepharose the dimer was loaded with recombinant Sm heterooligomers B/D3, D1/D2, and E/F/G. Remarkably, individual oligomers bound only weakly to the dSMN/dGemin2 dimer (Fig. 4A). However, binding of D1/D2 was greatly enhanced in the presence of E/F/G, indicating cooperative binding of these units. Likewise, B/D3 bound only efficiently to the complex when the other Sm proteins were present. Simultaneous incubation of all heterooligomers with the dSMN/dGemin2 dimer resulted in its efficient loading with the complete set of Sm proteins (Fig. 4A, lane 15).

Fig. 4.

Reconstitution of functional dSMN complex. (A) Recombinant H₆-GST-dGemin2/H₆-dSMN complex was immobilized on glutathione-Sepharose beads and incubated with purified Sm protein heterooligomers as indicated (lanes 9–15). As a specificity control, H₆-GST was immobilized on beads and mock-loaded with Sm proteins (lane 6). After removal of unbound proteins, reconstituted complexes were analyzed by SDS/PAGE. Lanes 1–5 show proteins used for the reconstitution assay. Bands indicated by an asterisk are degradation products or aggregates. In vitro assembly assay of reconstituted dSMN complex with dU1snRNA (B), dU1ΔSm (C), and dU85scaRNA (D) is shown. Radiolabeled RNAs were incubated with His-GST control (lanes 2, 7, and 12), dSMN/dGemin2 dimer (lanes 3, 8, and 13), or dSMN/dGemin2 bound to Sm proteins (lanes 4, 9, and 14), or with free Sm proteins (lanes 5, 10, and 15). Lanes 1, 6, and 11 show the RNA in the absence of protein. Reaction mixtures were separated by native gel electrophoresis, and complexes were visualized by autoradiography. The light upper band seen most prominently in lanes 12 and 14 denotes a conformer of dU85scaRNA.

To test its function in UsnRNP assembly, the dSMN/dGemin2 dimer was immobilized on glutathione-Sepharose beads and incubated with all Sm proteins. After washing, the protein complexes were eluted with glutathione and incubated with dU1snRNA, dU1ΔSm snRNA, or dU85scaRNA (Fig. 4 B–D). As a control, the same procedure was performed with His-GST protein instead of the SMN/Gemin2 complex. Sm core formation was observed only upon incubation of the SMN complex with dU1snRNA but not with the His-GST control (Fig. 4B). Importantly, Sm proteins bound to the SMN complex could not be transferred onto RNAs lacking a functional Sm site (Fig. 4 C and D). Thus, in vitro reconstituted dSMN complex is sufficient for Sm core assembly. To test whether a vertebrate SMN/Gemin2 dimer is likewise able to mediate this reaction alone, similar experiments were conducted using human SMN and Gemin2 (Fig. S2). Indeed we found not only binding of all Sm proteins to hSMN and hGemin2 (Fig. S2A) but also UsnRNP core assembly in an in vitro reaction (Fig. S2B).

UsnRNP assembly can occur spontaneously in vitro but depends on the SMN system in vivo. We therefore considered the possibility that the SMN complex confers specificity to the assembly reaction and hence prevents misassembly. To test this, we compared the specificity of spontaneous and SMN complex-mediated assembly. Remarkably, spontaneous binding of isolated Sm proteins occurred even under stringent conditions not only to the dU1snRNA target (Fig. 4B, lane 5), but also to the control RNAs lacking an Sm site (Fig. 4B, lanes 10 and 15). In contrast, Sm proteins bound to the SMN complex were exclusively transferred onto the cognate snRNA (Fig. 4B, lane 4), whereas noncognate RNAs were ignored (Fig. 4B, lanes 9 and 14). Hence, the minimal SMN complex of D. melanogaster acts not only as an assembly device but also as a chaperone that discriminates target and nontarget RNAs of Sm proteins.

Discussion

The human SMN complex belongs to a growing list of factors that assist assembly of RNA–protein complexes (25–27). However, because of its considerable complexity, only little is known about its mode of action and the contribution of individual factors to the assembly reaction. We therefore searched databases for organisms with simpler SMN complexes. This has not only allowed insight into the evolution of this assembly machinery, but has also led to the identification of the D. melanogaster SMN complex as a minimal unit amenable to biochemical investigation of the assembly process.

Previous studies have indicated that Gemins interact within the SMN complex in a modular manner (7, 8). Interestingly, our homology searches for components of the SMN complex in a variety of organisms recapitulated this finding on an evolutionary scale. The most simple SMN-containing complex is composed of SMN and Gemin2 only and can be found in unicellular organisms such as the fission yeast S. pombe (16) and in plants. The next level of complexity is characterized by the appearance of Gemin3 in D. discoideum, thus predating the emergence of the Fungi/Metazoa clade. The absence of Gemin5 from genomes of fungi and land plants and its presence in the green algae O. tauri and in D. discoideum indicate independent secondary gene loss in fungi and plants. This may be due to a role of Gemin5 outside of the SMN complex, which is not retained in these organisms (see also below).

Only later in evolution at the level when first metazoans developed, the building block composed of Gemins 6, 7, and 8 was added to the set of the Gemin family. From this time on, organisms had the potential to express an SMN complex similar in architecture to the human one. The only component that was not present at that point was Gemin4, which can be found only in the genomes of deuterostomians. Thus, our data suggest that the SMN complex evolved by a blockwise addition of Gemins to an ancient core complex of SMN and Gemin2 in a manner corresponding to their mutual biochemical association.

In striking contrast to this overall evolutionary trend, we found a remarkable simplification of this complex in the dipterans A. gambiae and D. melanogaster. In these animals, we failed to identify orthologs of Gemin4, as expected, but also of Gemins 6–8. However, these latter Gemins were clearly present in hymenopterans. Although orthologs of Gemin3 and Gemin5 were found in dipterans, they show a significantly higher evolutionary rate in dipterans than in other clades. These computational findings have been experimentally challenged by a biochemical approach that has allowed us to isolate an assembly-active SMN complex from D. melanogaster. Indeed, the composition of the complex was remarkably simple and consisted of SMN and Gemin2 as the only stoichiometric components. Dhh1 (Gemin3) bound to this core complex only at low salt concentrations, and Rigor mortis (Gemin5) was not present at all. The D. melanogaster SMN complex therefore equals its counterpart in S. pombe and plants although the function of SMN and Gemin2 orthologs in these organisms has not been demonstrated. It is conceivable that Dhh1 and Rigor mortis have adopted novel functions in a different context because they rapidly diverge from their ancestors (28). Consistent with this notion, a function of Rigor mortis in ecdysone signaling has been described (29).

Despite the obvious simplicity of the SMN complex in D. melanogaster, we provide evidence that this unit is functionally related to the SMN complex of mammals. First, a set of UsnRNP-related substrates, namely the common Sm proteins, UsnRNP specific factors (U1 70K, U2A′, U2B″/U1A, and U5 115K), and UsnRNAs were found to be part of the complex. Most of these factors have previously been shown to bind to SMN complexes of vertebrates (30). Second, affinity-purified dSMN complex mediated the assembly of the Sm core domain in vitro. Similar to the situation in humans, we found a strong dependence of UsnRNP core assembly on temperature but not on ATP (30). However, at present we cannot rule out the possibility that assembly of UsnRNPs in D. melanogaster cytosolic extracts requires ATP hydrolysis as observed for the same reaction in vertebrates (6, 31).

The obvious simplicity of the assembly system of D. melanogaster allowed the reconstitution of the dSMN complex from recombinant proteins and the investigation of its mode of action. Interestingly, we observed strong cooperativity in Sm protein binding onto the complex. Heterooligomers D1/D2 and B/D3 had only little affinity for the complex, but binding was greatly enhanced in the presence of E/F/G. Further studies are required to determine the precise binding sites of all Sm proteins on the SMN complex and to test the influence of arginine methylation on Sm protein binding (32, 33). It was an open question why UsnRNP assembly is strictly dependent on the SMN complex in vivo even though this reaction is spontaneous in vitro. Our assembly studies with the D. melanogaster SMN complex show that precise assembly of the Sm core domain on UsnRNA was possible only when Sm proteins were prebound to the SMN complex, whereas misassembly of isolated Sm proteins occurred under the same conditions. In addition, we demonstrate that human SMN and Gemin2 are likewise sufficient to specifically transfer Sm proteins onto UsnRNA. Hence, these data and similar studies performed in vertebrates argue for a dual role of the SMN complex as an RNP assembler and chaperone (31).

From an evolutionary point of view, our findings raise the question why dipterans can afford a minimized assembly system, whereas apparently other branches in the animal kingdom require a multicomponent SMN complex. The most plausible explanation for this paradox is that Gemins 3–8 are not primarily involved in the assembly reaction per se but rather in other steps during the UsnRNP biogenesis. Thus, it is known that the human SMN complex integrates several steps in biogenesis, such as cap hypermethylation (10) and nuclear import (9). We speculate that these steps will occur in dipterans independent of the SMN complex and may hence allow for the omission of individual Gemins. Further studies will be needed to test whether this is indeed the case.

In conclusion, our studies have shown that the integration of bioinformatics and biochemistry can be used to analyze cellular pathways functionally and evolutionarily. Similar strategies may prove to be powerful tools in the analysis of even more complex systems such as the spliceosome.

Experimental Procedures

Bioinformatic Strategies.

Ortholog identification.

To identify orthologs of SMN complex members, a PSI BLAST profile was generated for each protein by searching for five iterations with the human protein against the National Center for Biotechnology Information nonredundant peptide database (nr). The resulting profile was searched against the proteomes of the selected organisms. For validation, each candidate ortholog was manually searched backwards against nr and accepted only if known members of the SMN complex were retrieved significantly as best hits. In cases of unclear orthologous relationships caused for example by widespread domains in the query protein (Gemin3 and Gemin5) a phylogenetic tree was calculated. This allowed the separation of the orthologous subgroup from paralogs. Additionally, each protein was searched manually against nr to identify orthologs missing in the genome-specific gene prediction as well as against the National Center for Biotechnology Information's whole genome shotgun reads to identify orthologs missing in the genome assembly.

Evolutionary rate analysis for dipteran Gemin3 and 5.

Gemin3 and Gemin5 protein sequences from all organisms under investigation were aligned, and a phylogenetic tree was calculated by using proml from the PHYLIP package (34). In both cases of Gemin3 and Gemin5, the tree followed the species tree. To test whether the evolutionary rate of the dipteran proteins significantly differs from the other Gemin3 and Gemin5 orthologs, respectively, we used the codeml program of the PAML package (35). Using the topology of the calculated tree, two models were assumed and the fit of the data to the models was tested. First, a constant evolutionary rate was assumed all over the tree. Second, the dipteran branch (i.e., the one leading from the branching of A. mellifera to the speciation of Drosophila and Anopheles) was allowed a different rate. For both Gemin3 and Gemin5, the second model generated a faster evolutionary rate for the dipteran branch. To test whether this result was significant, a log-likelihood ratio test was performed. Here, the duplicated difference between the log-likelihoods of both models has to exceed a given cutoff drawn from the χ² distribution. Because the only difference between the models was the rate on one branch, one degree of freedom was assumed. In both cases, the log-likelihood ratio exceeded the χ² critical value for P = 0.001. A list of cDNAs and oligonucleotides used in this study is provided in Tables S2 and S3.

Generation of Stable Cell Lines and Affinity Purification of Complexes.

Drosophila Schneider2 cells (37) were cultured in Schneider's Drosophila medium (Biowest) containing 10% vol/vol FCS (PAA Laboratories). Cells were transfected with the respective pMTagIt expression construct using calcium phosphate. For the generation of stable cell lines, the pCoBlast vector (Invitrogen) was cotransfected and resistant cells were selected with blasticidin at 25 μg/ml (Invivogen). Induction of protein expression was performed with copper sulfate in a concentration determined for each construct. For extract preparation, cell pellets were resuspended in 3 volumes of PBS supplemented with 0.01% Igepal CA630 (PBS-I; Sigma) and disrupted by sonication. The lysate was cleared by centrifugation and subsequent filtration through a 0.25-μm syringe filter. For cytosolic extract preparation, Schneider2 cells were resuspended in 20 mM Hepes/KOH (pH 7.4), 80 mM potassium acetate, 4 mM magnesium acetate, and 50 μg/ml digitonin with protease inhibitors. After lysis for 10 min at room temperature, extract was cleared by centrifugation at 25,000 × g.

For TagIt purifications, extracts were incubated monoclonal antibody 7B10 (6) coupled to protein G Sepharose (GE Healthcare). After removal of unbound proteins by extensive washing, complexes were eluted with a 5-fold molar excess of synthetic peptide comprising the epitope recognized by 7B10.

For gradient centrifugation, 200 μl of extract prepared from S2 cells was layered on top of a 5–30% vol/vol glycerol gradient and centrifuged at 4°C at 38,500 rpm for 5 h in a Sw60Ti rotor. Fractions were separated by SDS/PAGE, and proteins were detected by Western blot analysis as described (30).

Additional Details.

For additional experimental procedures see SI Text.