U4 small nuclear RNA can function in both the major and minor spliceosomes

Abstract

U4 small nuclear RNA (snRNA) and U6 snRNA form a base-paired di-snRNP complex that is essential for pre-mRNA splicing of the major class of metazoan nuclear introns. The functionally analogous but highly diverged U4atac and U6atac snRNAs form a similar complex that is involved in splicing of the minor class of introns. Previous results with mutants of U6atac in which a substructure was replaced by the analogous structure from U6 snRNA suggested that wild-type U4 snRNA might be able to interact productively with the mutant U6atac snRNA. Here we show that a mutant U4 snRNA designed to base pair with a mutant U6atac snRNA can activate U12-dependent splicing when coexpressed in an in vivo genetic suppression assay. This genetic interaction could also be demonstrated in an in vitro crosslinking assay. These results show that a U4/U6atac di-snRNP can correctly splice a U12-dependent intron and suggest that the specificity for spliceosome type resides in the U6 and U6atac small nuclear ribonucleoproteins. Further experiments suggest that expression of a mutant U4 snRNA that can bind to wild-type U6atac snRNA alters the specificity of some splice sites, providing an evolutionary rationale for maintaining two U4-like snRNAs.

The nuclear pre-mRNA introns of eukaryotes are removed by a large ribonucleoprotein (RNP) complex known as the spliceosome (reviewed in refs. 1–4). In addition to a growing catalog of proteins numbering in the hundreds (reviewed in ref. 5), the spliceosome contains five small nuclear RNAs (snRNAs) that take part in a network of RNA–RNA interactions with the pre-mRNA splice sites and with each other. It is now recognized that there are two types of nuclear pre-mRNA introns widely distributed among higher eukaryotic species. The two types of introns are spliced via the action of two distinct types of spliceosomes. These spliceosomes differ in their snRNA compositions such that the more abundant U2-dependent type contains the snRNAs U1, U2, U4, U5, and U6, whereas the less abundant U12-dependent type contains the snRNAs U11, U12, U4atac, U5, and U6atac.

The four snRNAs that are unique to each spliceosomal type have been shown to be functional analogs. Thus, U11 snRNA appears to be the functional analog of U1 snRNA, U12 snRNA is the analog of U2 snRNA, U4atac snRNA is the analog of U4 snRNA, and U6atac snRNA is the analog of U6 snRNA. U5 snRNA appears to function in both spliceosomes. These analogies are based, in large part, on the similarities in RNA–RNA interactions displayed by the respective snRNAs. Most of these interactions have been validated by genetic evidence or biochemical crosslinking or both.

One set of these analogous interactions occurs between U4 and U6 snRNAs in the U2-dependent spliceosome and U4atac and U6atac snRNAs in the U12-dependent spliceosome. The interactions between these pairs of snRNAs are thought to occur before the addition of a U4/U6.U5 or U4atac/U6atac.U5 tri-snRNP to a nascent spliceosome. The U4 or U4atac snRNAs are subsequently unwound from U6 or U6atac snRNAs during the activation steps leading to an active spliceosome. The requirements for the U4/U6 and U4atac/U6atac base pairing interactions in splicing have been established genetically and biochemically (ref. 6 and references therein).

An essential feature of both U6 and U6atac snRNAs is an intramolecular stem–loop (ISL) structure that is believed to be involved in critical interactions at or near the active site of the spliceosome. The sequences that comprise the ISL regions are also major components of the U4/U6 and U4atac/U6atac base pairing interactions (the stem II regions, see Fig. 1 A and B). In earlier work examining the sequence requirements of the human U6atac snRNA ISL, we showed that several mutants of the human U6atac ISL were functional as long as the ability to form the base-paired stem regions of this structure was preserved (7). However, in many of these mutants, in vivo splicing activity could only be demonstrated when they were coexpressed with U4atac snRNA mutants containing compensatory alterations in the U6atac interaction region.

Fig. 1.

Base pairing interactions between U4/U4atac and U6/U6atac snRNAs. (A) Base pairing model of the human U4/U6 di-snRNP showing the sequences and locations of the stem I and stem II helices. (B) Base pairing model of the human U4atac/U6atac di-snRNP showing the sequences and locations of the stem I and stem II helices (29). (C) Base pairing model showing the proposed interactions of human U4 snRNA with a mutant human U6atac snRNA containing the U6 ISL sequence (7). The mutations introduced into the U6atac snRNA are shown in bold. (D) Base pairing model showing the proposed interactions of the U6atac Ath ISL mutant and the compensatory mutant U4 Ath Supp snRNA described in the text. The mutations introduced into U6atac and U4 snRNAs are shown in bold type. Also shown are the location and sequences of the U4 Stem I mutants described in the text.

In another part of this study, we showed that the ISLs of human and yeast U6 snRNA could function in place of the human U6atac snRNA ISL in U12-dependent splicing (7). These heterologous ISL sequences differed from the native human U6atac ISL in nine and eight positions, respectively. Based on our experience with other mutants of the U6atac ISL, we anticipated that these U6/U6atac chimeras would require coexpression of compensatory U4atac snRNA mutants. Surprisingly, however, we found that the chimeric snRNAs showed significant splicing activity even in the absence of a compensatory mutant U4atac snRNA. We suggested that because these chimeric snRNAs now contained the region of U6 snRNA (the ISL) that base pairs to U4 snRNA, the chimeric snRNAs were forming functional complexes with the wild-type U4 snRNA present in the cells. This potential interaction is shown in Fig. 1C.

Here we test this idea by expressing a human U6atac snRNA in which a highly divergent ISL sequence derived from the U6atac snRNA of the plant Arabidopsis thaliana was substituted for the human ISL as shown in Fig. 1D. We previously showed that this mutant U6atac snRNA was functional in vivo only in the presence of a specific U4atac suppressor snRNA designed to base pair to the Arabidopsis ISL (8). To test the hypothesis that U4 snRNA can functionally interact with U6atac snRNA, we altered human U4 snRNA so its interaction region was complementary to the Arabidopsis ISL sequence (Fig. 1D). We show here that these snRNA mutants will base pair with each other as assayed by psoralen crosslinking and that they will functionally interact in U12-dependent splicing when assayed in an in vivo genetic suppression system.

Methods

Construction of U6atac Mutants. The U6atac snRNA mutants were made in the 5′ splice site compensatory mutant GG 14/15 CC expression plasmid as described (9) by using either the pALTER mutagenesis kit (Promega) or PCR sewing techniques and mutagenic oligonucleotides. All mutations were confirmed by DNA sequencing.

Construction of U4 and U4atac Expression Plasmids. The U4 and U4atac expression plasmids were generated by the same method used previously (6, 8). Briefly, the U1 snRNA coding region of a functional U1 gene was replaced by PCR techniques with the coding region of U4 or U4atac snRNA amplified from plasmids obtained from J. Steitz (Yale University, New Haven, CT). For the mutations studied here, oligonucleotides that included the first 35 nucleotides of U4 or U4atac containing the desired mutations were used with wild-type 3′ oligonucleotides in standard PCRs. The PCR products were digested with SalI and BglII restriction enzymes and ligated into a U1 expression vector from which the U1 coding region had been excised (10). The sequences of the mutant and wild-type snRNAs were confirmed by DNA sequencing.

Analysis of in Vivo Splicing. Transient transfection of the P120 minigene and snRNA expression plasmids into cultured Chinese hamster ovary cells was as described (9, 11, 12). For these experiments, 0.5 μg of P120 plasmid and 5 μg of each of the snRNA expression plasmids were added to 1 × 10⁶ cells. Where one or more snRNA plasmids were omitted, a corresponding amount of pUC19 plasmid DNA was substituted. Total RNA was isolated from cells 48 h after transfection, reverse transcribed, and PCR amplified as described (11, 12). The products were analyzed by agarose gel electrophoresis followed by staining with ethidium bromide. Independent transfections and analyses gave substantially similar results. For the experiment shown in Fig. 4, a 5′ ³³P-radiolabeled 3′ primer was used in the PCR. The amplified products were separated on a 10% native polyacrylamide gel and exposed overnight to a PhosphorImager screen.

Fig. 4.

Modulation of spliceosome choice at a bifunctional 5′ splice site. (A) Splicing pattern of the P120 A1G/C99G mutant. The two splice site mutations are shown in bold. The 5′ GU splice site is joined to either a U12-dependent 3′ splice site AG (–1) or a U2-dependent 3′ splice site AG (–6). (B) RT-PCR analysis of the in vivo splicing of the P120 intron F in transfected Chinese hamster ovary cells. Lane 1 shows the splicing of wild-type P120; lane 2 shows the splicing pattern of the P120 A1G/C99G mutant. The lower band represents U12-dependent splicing to the wild-type 3′ splice site, whereas the upper band represents U2-dependent splicing to the cryptic 3′ splice site at –6. Lanes 3–6 show the effect of adding 2, 5, 10, and 15 μg of a wild-type U4 snRNA expression construct to the transfections. Lanes 7–10 show the effect of adding the same levels of a U4 snRNA expression construct containing the U4atac stem II interaction sequence. Lanes 11–14 show the effect of adding the same levels of a U4 snRNA expression construct containing the Ath stem II interaction sequence (U4 Ath Supp). The use of the U2- and U12-dependent 3′ splice sites is shown below each lane as a percentage of the total spliced signals. Also shown is the ratio of U2- to U12-dependent splicing in each lane.

Preparation of RNAs for Crosslinking. The U4 or U6atac snRNAs were radiolabeled during in vitro transcription. DNA templates were PCR amplified from the expression constructs by using primers that added a T7 RNA polymerase promoter sequence to the 5′ end to generate full-length coding regions of the snRNAs. DNA templates (1 μg) were used in 20-μl in vitro transcription reactions by using a Maxi Script in vitro transcription kit (Ambion) as per manufacturer's instructions. The radiolabeled RNA transcript was purified on a 6% polyacrylamide, 8 M urea gel.

Psoralen/UV Crosslinking. Psoralen/UV crosslinking was performed as described (13). Briefly, radiolabeled U6atac snRNA (10⁵ cpm) and 5 μg of unlabeled U4 snRNA in 5 μl were denatured at 90°C for 2 min and chilled quickly on ice. Then, 5 μl of 2× annealing buffer (50 mM sodium cacodylate buffer pH 7.5/300 mM KCl) was added, and the mixture was incubated at 95°C for 5 min followed by gradual cooling to 37°C. The samples were supplemented with 5 mM MgCl₂ and incubated at 37°C for an additional 15 min. The samples were kept on ice until 2 μl of 4′-aminomethyl-4′, 5′, 8′-trimethylpsoralen (1 mg/ml in RNasefree water; Sigma) was added, followed by further incubation on ice for 15 min in the dark. The samples were exposed to UV irradiation (365 nm) at a distance of 2–3 cm for 5 min. The crosslinked RNA samples were analyzed on a 6% polyacrylamide, 8 M urea gel and subjected to PhosphorImager analysis.

Results

We have previously described the in vivo mutational suppressor assay for the function of several of the snRNAs involved in U12-dependent splicing (8, 9, 11, 12). This assay relies on the genetic suppression of splicing defects due to splice site mutations in a U12-dependent intron by coexpression of compensatory mutant snRNAs. To assay the in vivo function of U6atac snRNA, a mutation in the 5′ splice site of a transfected minigene construct (CC5/6GG), which blocks splicing at the normal sites, is suppressed (i.e., splicing is restored at the normal sites) by cotransfection of expression constructs containing compensatory mutants of U11 and U6atac snRNAs (U11 GG6/7CC and U6atac GG14/15CC, ref. 9).

To test for the ability of U4 to productively interact with U6atac snRNA, we used the U6atac Ath ISL mutant snRNA construct in which the human ISL was exchanged with the ISL from the A. thaliana U6atac snRNA (Fig. 1D). This construct also contained the GG14/15CC 5′ splice site mutant suppressor mutations. To determine whether a U4 snRNA could functionally interact with this U6atac snRNA, we altered the stem II region of human U4 snRNA to make it complementary to U6atac Ath ISL in the stem II region as shown in Fig. 1D.

To demonstrate that these two mutant snRNAs could base pair to each other, we performed in vitro UV crosslinking experiments in the presence of psoralen. As shown in Fig. 2, a prominent crosslinked species involving labeled U6atac Ath ISL snRNA could be detected when incubated in the presence of unlabeled U4 Ath Supp mutant snRNA (Fig. 2, lane 1). The crosslinked band depended on UV irradiation, the presence of psoralen (Fig. 2, lanes 2–4) and the presence of U4 Ath Supp snRNA (Fig. 2, lane 5). Wild-type human U4 snRNA gave much reduced levels of the crosslinked species (Fig. 2, lane 6) as did wild-type human U4atac snRNA (Fig. 2, lane 8). A similar crosslinked species could also be detected with unlabeled U6atac Ath ISL snRNA and labeled U4 Ath Supp snRNA (data not shown). Treatment of the crosslinking reactions with DNA oligonucleotides specific for either U4 or U6atac snRNA and RNase H led to cleavage of the crosslinked band, demonstrating that it was caused by intermolecular crosslinking (see the supporting information, which is published on the PNAS web site).

Fig. 2.

Psoralen crosslinking of snRNAs in vitro. Labeled U6atac Ath ISL RNA was incubated with unlabeled U4, U4 Ath Supp, or U4atac snRNA as indicated and treated with UV light in the presence or absence of psoralen. Samples were analyzed by denaturing polyacrylamide gel electrophoresis. The positions of labeled U6atac Ath ISL RNA and a slower migrating crosslinked species are indicated.

Although the above data suggest that appropriately modified U6atac and U4 snRNAs can form a base-paired complex, it does not show that such a complex can be functional in nuclear pre-mRNA splicing. To examine this possibility, we asked whether the U4 Ath Supp mutant snRNA could rescue the function of the U6atac Ath ISL snRNA in vivo by using the genetic suppression assay for U12-dependent pre-mRNA splicing.

In Fig. 3 we show the effect of substituting the U4 Ath Supp mutant snRNA construct for the previously described U4atac Ath Supp snRNA suppressor. In this experiment, splicing of the U12-dependent intron F in the P120 minigene is monitored after transient transfection of the minigene along with various snRNA constructs. Fig. 3 shows the RT-PCR products by using primers located in the exons flanking the U12-dependent intron F. Lane 3 in Fig. 3 shows that the transfected wild-type P120 minigene produces almost exclusively correctly spliced RNA whereas lane 4 shows that no correctly spliced RNA is produced by the CC5/6GG 5′ splice site mutant alone. We have previously shown that the U11 GG6/7CC snRNA suppressor alone is unable to restore splicing to this mutant (12). This result is also shown in lane 5 of Fig. 3. Lanes 6 and 7 of Fig. 3 show that the U6atac GG14/15CC suppressor with the Ath ISL is unable to restore splicing either in the presence or absence of U11 GG6/7CC snRNA. Fig. 3, lane 8 confirms that the addition of the U4atac Ath Supp snRNA construct that restores base pairing to the U6atac Ath ISL stem II region can restore U12-dependent splicing in vivo (8). Lanes 9 and 10 of Fig. 3 repeat our previous results (7) showing that the U6atac with the human U6 ISL is active in the absence of a compensatory U4atac mutant although correct splicing is stimulated by the addition of this U4atac mutant (U4atac Hu U6 supp, Fig. 3, lane 10).

Fig. 3.

In vivo suppression assay of mutant snRNAs. RNA was extracted from cells transiently transfected with the P120 minigene and snRNA constructs as shown. The splicing pattern of the U12-dependent P120 intron F was analyzed by RT-PCR amplification using primers in the adjacent exons. The positions of bands corresponding to unspliced and correctly spliced products are shown as well as the cryptic spliced product. The cryptic spliced band is caused by splicing between a pair of U2-dependent splice sites located within the P120 F intron. The constructs used for the various lanes are: lane 1, mock transfected cells; lane 2, empty pCB6 expression vector; and lane 3, wild-type P120 minigene. Lanes 4–13 used the P120 5′ splice site mutant CC5/6GG and the following snRNA expression constructs: lane 4, the P120 mutant alone; lane 5, U11 GG6/7CC; lane 6, U6atac GG14/15CC Ath ISL; lane 7, U6atac GG14/15CC Ath ISL plus U11 GG6/7CC; lane 8, U6atac GG14/15CC Ath ISL plus U4atac Ath Supp plus U11 GG6/7CC; lane 9, U6atac GG14/15CC Hu U6 ISL plus U11 GG6/7CC; lane 10, U6atac GG14/15CC Hu U6 ISL plus U4atac Hu U6 Supp plus U11 GG6/7CC; lane 11, U6atac GG14/15CC Ath ISL plus U4 Ath Supp plus U11 GG6/7CC; lane 12, U6atac GG14/15CC Ath ISL plus U4 Ath Supp SI-1 plus U11 GG6/7CC; and lane 13, U6atac GG14/15CC Ath ISL plus U4 Ath Supp SI-2 plus U11 GG6/7CC. Lane M contains molecular size markers.

Lane 11 of Fig. 3 provides functional evidence that U4 snRNA can interact with U6atac snRNA to promote U12-dependent splicing in vivo. In this experiment, the U6atac Ath ISL mutant, which is inactive by itself (Fig. 3, lane 7), can promote in vivo splicing in the presence of a U4 snRNA mutant that can base pair to the Ath ISL sequence (human U4 Ath Supp). Neither transfection of the human U4 Ath Supp construct alone nor transfection of a wild-type U4 construct with the U6atac Ath ISL construct restored correct splicing to the P120 mutant (data not shown).

All of the changes in the human U4 Ath Supp mutant were limited to the stem II interaction region of U4. As shown in Fig. 1, the stem I interaction regions of U4 and U4atac snRNAs are very similar such that U6atac might be able to form virtually identical stem I structures with either U4atac or U4 snRNAs. To determine whether human U4 Ath Supp was forming an essential stem I interaction with U6atac, two additional mutations were tested in the in vivo suppression assay. The five boxed nucleotides in U4 snRNA shown in Fig. 1D were either deleted from the human U4 Ath Supp construct to make human U4 Ath Supp SI-1 or simultaneously mutated to their Watson–Crick complements to make human U4 Ath Supp SI-2. These U4 mutants were tested for in vivo suppression activity, and the results are shown in lanes 12 and 13 of Fig. 3. Neither mutant showed suppressor activity, suggesting that the stem I interaction between U4 and U6atac snRNAs is essential for activity as we have shown for the U4atac/U6atac Stem I interaction (6). Taken together, these results provide in vivo evidence that U4 snRNA can cooperate with U6atac snRNA in U12-dependent splicing in vivo.

The results in Fig. 3 show that a modified U4 snRNA can functionally replace U4atac snRNA in the splicing of a U12-dependent intron. This result raises the interesting question of why the cell maintains two distinct U4-like snRNAs when it appears that a single bifunctional snRNA might suffice. One possibility is that having two distinct U4-like snRNAs that function in each spliceosome separately helps to differentiate U2- from U12-dependent introns. We have found that relatively minor changes to a U12-dependent 5′ splice site can transform it into a functional U2-dependent 5′ splice site (14). We have argued that this kind of shift lies behind the low abundance of U12-dependent introns in modern genomes (15). If the identity of the U4-like snRNA in the tri-snRNP complexes contributes specificity information as to which type of spliceosome is assembled at a 5′ splice site, then this could be detected at bifunctional 5′ splice sites. Such bifunctional sites, which can be spliced to either a U12-dependent 3′ splice site or a U2-dependent 3′ splice site, have been described (14). The specific case we use here is a double mutant of the U12-dependent P120 intron F where both the 5′ and 3′ terminal intron nucleotides have been mutated to G residues (Fig. 4A). This construct splices either to the normal U12-dependent 3′ splice site (designated –1) or to the cryptic U2-dependent 3′ splice site located six nucleotides upstream of the normal site (designated –6). The snRNA dependence of these sites has been established by in vitro experiments using specific antisense oligonucleotides (14).

To investigate the effect on spliceosome specificity by substituting U4 for U4atac snRNA, we cotransfected this P120 mutant with several snRNA expression constructs and assessed the effect on use of the two alternative 3′ splice sites (Fig. 4B). Fig. 4B, lane 2 shows that the P120 A1G/C99G mutant by itself splices to both the U2-dependent and U12-dependent 3′ splice sites with a ratio of 1.9. Addition of a construct coding for the wild-type U4 snRNA had only a minor effect on the relative use of the two 3′ splice sites increasing the ratio of U2- to U12-dependent splicing to 2.6 (Fig. 4B, lanes 3–6). In contrast, a mutant U4 snRNA construct in which the stem II interaction region was replaced by the U4atac sequence led to a significant increase in U2-dependent splicing over U12-dependent splicing to give a ratio of 5.2 (Fig. 4B, lanes 7–10). As a control, we used the U4 Ath Supp construct in which the same interaction region of U4 was mutated, but in this case to a sequence that cannot interact with the wild-type human U6atac snRNA. Cotransfection of this construct had the same effect as wild-type U4, slightly increasing the ratio to 2.6 at the highest concentration (Fig. 4B, lanes 11–14), suggesting that the activity of the U4+U4atac stem II mutant was due to its ability to base pair with human U6atac snRNA.

Discussion

In the work described here, we show that the human U4 snRNA can functionally interact with the human U6atac snRNA to promote splicing of a U12-dependent intron. Both the in vivo and in vitro assays suggest that U4 and U6atac can form a functional di-snRNP complex. Our current understanding of the pathway of spliceosome formation holds that U6 or U6atac snRNAs are delivered to the forming U2- or U12-dependent spliceosomes, respectively, in a tri-snRNP complex also involving U5 and U4 or U4atac snRNPs. Because the U5 snRNA is common to both spliceosomes, it was expected that the U12-dependent splicing pathway would involve a U4atac/U6atac.U5 complex that was functionally equivalent to the U4/U6.U5 tri-snRNP of the U2-dependent pathway. Such a complex has been recently identified biochemically (16). The work described here shows that a different composition of snRNPs, namely U4/U6atac.U5, can also function to promote correct U12-dependent splicing.

Although there is ample evidence that U4 and U4atac snRNAs are required for splicing by the U2- and U12-dependent spliceosomes, respectively (6, 17–19), precisely why they are required has been poorly understood. Both snRNAs appear to be destabilized or dissociated from the spliceosome during the activation stage of spliceosome formation, and thus it was believed that they played the role of RNA chaperones by base pairing with U6 or U6atac snRNAs (19–21). Mutagenesis experiments appeared to support this idea by showing that the functional regions of U4 or U4atac snRNAs were largely confined to the two regions that base pair with U6 or U6atac snRNAs (6, 22). The exceptions were the Sm protein binding site, required for proper snRNA processing and targeting, and the 5′ stem loop region. More recent biochemical and genetic investigations have focused on this 5′ stem loop element. Nottrott et al. (23) have shown that this region is conserved between U4 and U4atac and that it is the binding site for a 15.5-kDa protein. The binding of this 15.5-kDa protein to the 5′ stem loop structure serves as a recognition element for the assembly of other proteins onto the tri-snRNP particle (24). The U4/U4atac snRNAs are now believed to help structure the tri-snRNP and to aid in the delivery of both proteins and snRNAs to the forming spliceosomes. Recent results have also shown that all of the known proteins of the U4/U6.U5 tri-snRNP are also present in the U4atac/U6atac.U5 tri-snRNP (16).

From the studies described here and the previous work described above, it seems clear that U4 and U4atac snRNAs are structurally and functionally homologous and can substitute for each other (or at least that U4 can substitute for U4atac). The only functional distinction between them appears to be the ability to base pair to the U6 or U6atac stem II region. As we have shown previously, however, the U6atac stem II region can be replaced by the U6 stem II region to produce a functional snRNA that probably interacts with endogenous U4 snRNA (7). The question arises then, why are two types of U4-like snRNAs maintained in the genome and why are the stem II regions of U6 and U6atac snRNAs consistently different?

The existence of two parallel spliceosomes predates the separation of plants and animals (25). In comparing the snRNAs between spliceosomes and between kingdoms, it is clear that the U2-dependent snRNAs are much more highly conserved than are the U12-dependent snRNAs (8). The U6 ISL region is identical in plants and mammals. The stem II pairing region of U4 is similarly highly conserved. In contrast, the U6atac ISL region of Arabidopsis differs in 9 of 22 positions compared to mammals. To our knowledge, the plant U4atac homolog has not yet been identified. Nevertheless, it seems virtually certain that a plant U4atac homolog does exist because the plant U4 cannot base pair with the plant U6atac stem II region. Given the billions of years during which the two spliceosomes have coexisted in the genome, it is striking that the ISLs of U6 and U6atac have not converged so that they can interact with a single, bifunctional U4-like snRNA.

The most likely explanation for this continued separation of function is that a U4/U6atac.U5 tri-snRNP complex, such as we produced in this study, may have a relaxed specificity with respect to splice site choice or may hinder the proper recognition of the correct splice site pairs. We have provided evidence for this by using an unusual 5′ splice site that can splice to both types of 3′ splice sites. We showed that expression of U4 snRNA mutants that could base pair with U6atac snRNA altered the specificity of this 5′ splice site in favor of the U2-dependent splicing pathway.

Two possible explanations for this result are that (i) the modified U4 is binding the functional U6atac to form a less active complex, thus favoring the U2-dependent pathway, or (ii) U4 and perhaps U4atac play a role in recruiting factors specific for each splicing system to the 5′ splice site. This second possibility raises interesting questions about the process of splice site identification. The results of Maroney et al. (26) suggest that U1 snRNP is stabilized at the 5′ splice site in part by the binding of a U4/U6.U5 tri-snRNP to the exon immediately upstream of the 5′ splice site. The molecular interactions involved in this binding are unknown. The ability of the chimeric U4 snRNA, which is presumed to be in a U4/U6atac.U5 tri-snRNP, to promote U2-dependent splicing at a bifunctional 5′ splice site suggests that this tri-snRNP may recruit U1 snRNP to this 5′ splice site. This would, in turn, suggest that either U4 snRNA itself or a protein associated with U4 snRNA might interact with U1 snRNP in this early complex. We do not know whether there is an analogous interaction of the U4atac/U6atac.U5 tri-snRNP with U11 that promotes the activation of U12-dependent 5′ splice sites. However, this potential early function for the spliceosomal tri-snRNPs in splice site activation may provide an explanation for the conservation of separate U4-like and U6-like snRNAs in the two splicing systems.

Although the bifunctional 5′ splice site used in this experiment was created by the mutation of a normal U12-dependent 5′ splice site, we have found several examples of genes in which a single 5′ splice site is alternatively spliced to 3′ splice sites with either U12-type or U2-type features. In these cases, the 5′ splice site sequences deviate from the U12-type consensus (data not shown). These results suggest, then, that to the extent that separation in function between the two spliceosomal systems is important for proper expression of the genome, there is selective pressure to maintain separate U4-like snRNAs for the two splicing systems.

Our finding that U6atac snRNA can be functional in a hybrid U4/U6atac.U5 tri-snRNP complex also raises the issue of what determines the specificity of the spliceosome with regard to intron type. Although the experiment discussed above suggests that some specificity resides in the U4-like snRNA, the fact that U4 snRNA can be made to function with U6atac snRNA suggests that the major determinant of spliceosomal specificity lies elsewhere. Because all of the known tri-snRNP proteins are shared between the U4/U6.U5 tri-snRNP of the major class spliceosome and the U4atac/U6atac.U5 tri-snRNP of the minor class spliceosome (16), only the U6atac snRNA appears to be uniquely required in the U12-dependent tri-snRNP. Where in U6atac would the spliceosome specificity be located? An obvious candidate is the 5′ splice site interaction region immediately 5′ of the conserved ACAGA sequence. Both U6 and U6atac snRNAs have nucleotides in this region that form base pairs with their respective 5′ splice sites (ref. 9 and references therein). If these sequences were the primary specificity elements, it should be possible to reprogram a U4/U6.U5 tri-snRNP to functionally interact with a U12-dependent intron by altering the 5′ splice site pairing region of U6.

We have tested this idea by introducing a mutated U6 snRNA construct containing the 5′ splice site pairing sequence from the U6atac GG14/15CC suppressor mutant into cells along with our mutated P120 test plasmid as described above. No suppression of the U12-dependent splicing defect was detected, suggesting that this U6 snRNA could not productively participate in this process (data not shown). Although this is a negative result, it suggests the possibility that there are other features of U6atac snRNP that target it to U12-dependent introns. These could be other unidentified RNA–RNA interactions with U11 and/or U12 snRNAs or protein–protein interactions between as yet unidentified U6atac-specific proteins and U11/U12-specific proteins. Although most of the known RNA–RNA interactions in the U2-dependent splicing pathway have identified counterparts in the U12-dependent pathway, a notable exception is the U2–U6 helix II interaction. No similar U12–U6atac interaction can be made because the 5′ end of U12 snRNA immediately follows the helix I interaction region with U6atac snRNA. The U2–U6 helix II interaction is required for in vivo splicing of U2-dependent introns in mammalian cells (27, 28), suggesting that a counterpart should be required for U12-dependent splicing. Further investigation of U12–U6atac interactions may reveal such a counterpart that might supply the specificity to select which snRNPs enter the two spliceosomes.