Comparison of polypeptides that are associated with the purified 17S U2 snRNP and SF3a or precipitated from spliceosomes with anti-U2 snRNP antibodies has previously suggested that SF3b consists of at least four subunits of 49, 130, 145, and 155 kD (Behrens et al., 1993b
; Brosi et al., 1993a
; Staknis and Reed, 1994
). Here we have demonstrated that SF3b purified from HeLa cell nuclear extracts comprises four subunits of the expected sizes. Polypeptides of identical apparent molecular mass are highly enriched in glycerol gradient fractions of the 15S U2 snRNP and these fractions substitute for the 12S U2 snRNP and SF3b activity in presplicing complex formation. Moreover, only the presumptive SF3b subunits are precipitated from purified or reconstituted 17S U2 snRNP with an antibody directed against SF3a66 in addition to the SF3a subunits, 12S U2 snRNP-associated proteins, and U2 snRNA.
Together, the SF3a and SF3b subunits account for seven of the 17S U2 snRNP-specific proteins (Behrens et al., 1993b
). Similarly, anti-U2 snRNP antibodies precipitated seven polypeptides in addition to the 12S U2 snRNP proteins from a fraction enriched in U2 snRNP (Staknis and Reed, 1994
). This raises questions about the 35 and 92-kD polypeptides found associated with the purified 17S U2 snRNP by Behrens et al. (1993b)
. It is possible that these proteins are even less tightly bound to the U2 snRNP than SF3a and SF3b (Behrens et al., 1993b
; Brosi et al., 1993a
) and dissociate from the remainder of the particle during purification or immunoprecipitation procedures. The 35- and 92-kD proteins have not been detected in mAb66 immunoprecipitates of the 17S U2 snRNP, suggesting that they are not required for SF3a, SF3b, and the 12S U2 snRNP to remain stably associated with one another. However, this does not exclude the possibility that these proteins function in the assembly of the 17S U2 snRNP or the prespliceosome. Loss of the 35- and 92-kD proteins from the 15S U2 snRNP or SF3b fractions during glycerol-gradient sedimentation could, for example, explain the low activity of these fractions in presplicing complex formation.
Brosi et al. (1993a)
previously established that the 17S U2 snRNP is formed in vitro by stepwise interactions of SF3b and SF3a with the U2 snRNP. The results presented here strongly suggest that the assembly pathway is initiated by contacts of SF3b with the 5′ half of U2 snRNA. First, in the 12S U2 snRNP only the 3′ terminal stem-loops and the Sm-binding site are protected from micrococcal nuclease digestion. In contrast, the region extending from stem-loop I to stem-loop IIa (and to a lesser extent stem-loop IIb) of U2 snRNA, which is fully accessible to micrococcal nuclease digestion in the 12S U2 snRNP, is protected in the reconstituted 15S U2 particle. Second, treatment of the 12S U2 snRNP with micrococcal nuclease before incubation with SF3a and SF3b prevents the assembly of the 17S U2 snRNP, indicating that at least part of the U2 snRNA is required for stable contacts. Moreover, visualization by electron microscopy demonstrated that the 15S U2 particle contains two structurally distinct domains. The smaller domain exhibits the typical features of the single domain of the 12S U2 snRNP, whereas the characteristic structure of SF3b is conserved in the larger domain. The domains are connected by a thin filament, which most likely represents U2 snRNA, in analogy to the 17S U2 snRNP where a corresponding connection is RNase-sensitive (Behrens et al., 1993b
The interpretation of these data is in agreement with and extends results of previous studies. (Although the following results were obtained with the 17S U2 snRNP, we believe that they reflect the situation of the 15S U2 particle, i.e., the binding of SF3b, because the differences in micrococcal nuclease protection in the 5′ half of U2 snRNA in the 15S and 17S U2 snRNP were quantitative rather than qualitative in nature [see Fig. C].) Temsamani et al. (1991)
reconstituted the U2 snRNP in a cytoplasmic S100 extract from synthetic U2 snRNA. Upon incubation in nuclear extract the particles (which exhibited a high mobility in native polyacrylamide gels) were converted to low mobility complexes that almost certainly correspond to the 17S U2 snRNP. Oligonucleotide-directed RNase H cleavage of the 5′ end of U2 snRNA and the branch site interaction sequence before incubation in the nuclear extract abolished 17S U2 snRNP formation. Behrens et al. (1993b)
reported that the 5′ end of the U2 snRNA was cleaved more efficiently by RNase H in the presence of a complementary oligonucleotide in the 17S than in the 12S U2 snRNP. In addition, nucleotides C8 and C10 were more accessible to chemical modification in the 17S than in the 12S U2 snRNP. Together these data suggested a structural change in the 5′ portion of U2 snRNA upon binding of the 17S U2-specific proteins that results in a (partial) melting of stem I and thus an increase in the potential for intermolecular base pairing. Furthermore, strengthening of stem-loop I by mutation inhibited splicing in a mammalian extract (Wu and Manley, 1992
), which could reflect an interference with the binding of SF3b. U2 snRNA sequences on both sides of stem I engage in base pairing interactions with U6 snRNA (for review see Staley and Guthrie, 1998
). Therefore, it is intriguing to speculate that one or more SF3b subunits are directly involved in presenting this region of U2 to U6 snRNA for the formation of helix Ia/Ib which is an essential element of the catalytic center of the spliceosome both in mammals and yeast.
The protection of the region encompassing stem-loop IIa in the 15S (and 17S) U2 snRNP is in good agreement with the result that chemical modification of nucleotides C40, G42, and C45 is less efficient in the 17S than the 12S U2 snRNP (Behrens et al., 1993b
). Moreover, deletion of nucleotides 46–49 prevents the assembly of the 17S U2 snRNP (Temsamani et al., 1991
). In addition, a mutation in Cus1p, the S
homologue of SF3b145 (Gozani et al., 1996
; Wells et al., 1996
), suppresses mutations in yeast U2 snRNA that correspond to nucleotides 39–44, 52, and 61 of human U2 snRNA (Wells et al., 1996
; Yan and Ares, 1996
). Based on these results and the evolutionary conservation of SF3b145 and Cus1p (Gozani et al., 1996
; Wells et al., 1996
) we assume that the protection of stem-loop IIa reflects the binding of SF3b145. Preliminary evidence indicates that SF3b145, SF3b49, and one of the other large SF3b subunits cross-link to U2 snRNA (Gröning, K., and A. Krämer, unpublished observation). Moreover, Cus1p is apparently associated with an RNA-binding activity (see Igel et al., 1998
). Thus, it is highly likely that SF3b145 binds directly to stem-loop IIa of U2 snRNA.
Whether or not the branch site interaction sequence of U2 snRNA is involved in interactions with the SF3b subunits is unclear. We have not detected any micrococcal nuclease-sensitive sites between stem-loops I and IIa in the 15S or 17S U2 snRNPs. However, this region is equally well accessible to oligonucleotide-targeted RNase H digestion and chemical modification in the 12S and 17S U2 snRNPs (Behrens et al., 1993b
). This discrepancy could be explained by a different availability of the branch site interaction region to oligonucleotide-targeted RNase H or micrococcal nuclease digestion or to relatively mild conditions used for nuclease treatment in this report. The observation that U2 snRNAs carrying mutations within the branch site interaction region fail to assemble into the 17S U2 snRNP (Temsamani et al., 1991
) argues in favor of a contribution of these sequences to the assembly of the active U2 snRNP. Clearly, this issue needs further clarification.
During the formation of the 15S U2 snRNP we also observed changes in the protection pattern in the 12S domain, in that the protection of stem-loops III and IV is more efficient and a weak protection of the Sm-binding site is apparent. Equivalent changes in the structure of the subdomains of the 15S U2 snRNP compared with isolated SF3b or the 12S U2 snRNP are not visible in the electron microscope. In fact, the domain containing the A′ and B′′ proteins appears to be located at a distance from the SF3b domain; therefore, we believe that direct contacts between these proteins are limited. Although we cannot rule out interactions between SF3b and the Sm proteins that may not be resolved in the electron microscope, we favor the notion that binding of SF3b to the 5′ half of U2 snRNA results in a subtle conformational change in the 12S domain.
SF3a does not stably interact with either SF3b or the 12S U2 snRNP alone (Brosi et al., 1993a
; data presented here), indicating that both components contribute to the binding site for SF3a. Consistent with this, in the 17S U2 snRNP we observed an extended protection from micrococcal nuclease digestion that encompasses the Sm binding site and stem-loops III and IV of U2 snRNA as well as a more efficient protection of the 5′ half. In addition, we observed differences in the morphology of the 17S U2 snRNP in the electron microscope. Like the 15S U2 snRNP, the 17S particle is composed of two domains, both of which approximately correspond in size to the SF3b domain of the 15S U2 snRNP. However, structural details of neither the SF3b nor 12S domain are apparent in the two globular structures of the 17S U2 particle. Together, the results from micrococcal nuclease protection and electron microscopic analysis suggest that the main binding site for SF3a is the 12S domain of the U2 snRNP, thus contributing to the increase in size. The chemical modification pattern in the 3′ half of U2 snRNA was identical in the 12S and 17S U2 snRNP (Behrens et al., 1993b
), which was interpreted to mean that most 17S U2 snRNP-specific proteins associate with the 5′ portion of the U2 snRNP. On the other hand, it is possible that binding of SF3a does not occur by tight protein–RNA interactions or induce substantial structural changes in the RNA, but may rely mainly on protein–protein interactions. Good candidates for interacting proteins are the A′ and/or B′′ proteins, because deletion of their binding site (nucleotides 154–167) prevents the assembly of the 17S U2 snRNP (Temsamani et al., 1991
). Thus, the integrity of the 3′ end of U2 snRNA and/or the presence of the A′ and B′′ proteins may be necessary for binding of SF3a. An interaction of SF3a with these proteins (or this region of U2 snRNA) may also explain why U2 snRNAs carrying mutations in loop IV were inactive in mammalian splicing (Wu and Manley, 1992
). Contacts with the Sm proteins are also possible, given that no micrococcal nuclease sensitive sites were detected between stem-loop IV and the Sm-binding site. In fact, the major micrococcal nuclease-sensitive site between sequences complementary to oligonucleotides Sm and J in the 15S U2 snRNP is completely protected in the 17S U2 particle, which locates SF3a close to the Sm proteins.
The increased efficiency of micrococcal nuclease protection in the 5′ half of U2 snRNA observed upon binding of SF3a can be interpreted by stabilization of the interaction of SF3b with the U2 snRNA. This could either be the result of a structural change induced by binding of SF3a to the 3′ portion of U2 snRNP or by direct contacts between the SF3a and SF3b subunits. The idea of direct contacts between these proteins is supported by the result that the SF3a and SF3b subunits as well as the B′′, B, and B′ proteins are coprecipitated after micrococcal nuclease digestion of the reconstituted 17S U2 snRNP. Evidence for interactions between SF3a and SF3b or U2 snRNA comes from studies in S
. First, mutations in homologues of the SF3a subunits (Prp9p, Prp11p, and Prp21p) are synthetic lethal with a number of mutations in stem-loops IIa and IIb of U2 snRNA and the region between the branch site interaction sequence and stem-loop IIa (Ruby et al., 1993
; Wells and Ares, 1994
; Yan and Ares, 1996
). The same pattern of phenotype was observed with all three mutant proteins, which suggested that they interacted with the U2 snRNP as a functional unit. A candidate for interaction is Cus1p, the homologue of SF3b145. Mutations in Cus1p suppressed mutations in U2 snRNA downstream of the branch site interaction sequence and in stem-loop IIa (Wells et al., 1996
; Yan and Ares, 1996
). The synthetic lethality between mutations in the SF3a subunits and stem IIb (Wells and Ares, 1994
) may also explain the weak but apparent protection within this region. Second, extra copies of wild-type Cus1p partially suppressed the growth defect of prp11-1, but not prp9-1 or prp21-1, suggesting interactions between SF3b145 and SF3a66 (Wells et al., 1996
). Third, in a yeast two-hybrid screen the gene product of Yml049c, which represents the yeast orthologue of SF3b130 (Caspary et al., 1999
; Krämer, A., unpublished data), was found as a partner of Prp9p (SF3a60; Fromont-Racine et al., 1997
Given these possible interactions, it may be surprising that the 17S U2 snRNP consists of two distinct globular domains in the electron microscope (Fig. D; Behrens et al., 1993b
). On the one hand, this discrepancy could be explained if interactions between SF3a and SF3b were indirect and merely resulted in a structural change of the SF3b domain, which may be reflected in a loss of the typical features of this domain upon formation of the 17S U2 snRNP. On the other hand, interactions between SF3a and SF3b could be relatively weak and thus not be visible in the electron microscope. Another possibility is that protein–protein contacts involve structures that are beyond the detection limit of the electron microscope.
In summary, the combined data from biochemical and electron microscopic analyses can be incorporated into a model in which SF3b initiates the assembly of the active U2 snRNP by interactions with the 5′ half of U2 snRNA, which positions this splicing factor close to U2 snRNA sequences that are essential for the catalysis of splicing. SF3a then associates with the 3′ portion of the U2 snRNP and, most likely, directly interacts with one or more of the SF3b subunits.