INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Intron removal from mRNA precursors (pre-mRNA splicing) is an essential step of gene expression in eukaryotes. The precise recognition of the intron boundaries, the 5' and 3' splice sites, is achieved by small nuclear RNPs (snRNPs) and non-snRNP proteins. The 5' splice site is initially recognized by U1 snRNP, and the 3' splice site region is recognized by U2 snRNP. Subsequent addition of the U4/U6/U5 tri-snRNP forms the spliceosome, the macromolecular complex within which splicing catalysis takes place (reviewed in references 6 and 23).

Several sequence elements help to define the 3' splice site region in higher eukaryotes (reviewed in reference 35) : the branchpoint (BP) sequence, usually followed by a pyrimidine-rich sequence (the polypyrimidine tract or Py tract), and a conserved AG dinucleotide at the 3' end of the intron. The BP contains an adenosine residue that forms a 2' to 5' phosphodiester bond with the 5' end of the intron during the first catalytic step of the splicing reaction (39). U2 snRNP binds to the BP through base pairing interactions between this sequence and U2 snRNA (31, 33, 50, 56). U2 snRNP binding requires auxiliary factors, including SF1/mBBP and U2AF (22, 24, 40). SF1/mBBP has been shown to specifically recognize the BP (2, 34) and play a kinetic role in spliceosome assembly (17, 41). U2AF is a heterodimer of 65 and 35-kDa subunits (52). U2AF⁶⁵ binds specifically to the Py tract via its RNA recognition motifs (RRMs) (53) and contacts the BP via its RS domain (11, 44), whereas U2AF³⁵ contacts the AG dinucleotide at the 3' splice site (30, 51, 58).

The 3' splice site AG marks the 3' intron boundary and is involved in exon ligation, the second catalytic step of the splicing reaction. For some AG-dependent substrates, however, this dinucleotide is already required for early steps of spliceosome assembly prior to catalysis (36). AG-dependent substrates typically contain weak Py tracts, and substrates with strong Py tracts generally do not require the presence of the 3' splice site AG before the second catalytic step and are considered AG independent. Interaction between U2AF³⁵ and the 3' splice site AG dinucleotide was shown to stabilize U2AF⁶⁵ binding to a weak Py tract and to be essential for splicing of AG-dependent substrates (51).

An alternative set of interactions has been proposed for U2AF³⁵. The arginine-serine (RS) region of U2AF³⁵ has been shown to establish protein-protein interactions with splicing factors of the SR family (49) (reviewed in references 10, 13, 29, and 45). One type of sequences bound by SR proteins are purine-rich exonic splicing enhancers (ESE), which stimulate splicing of pre-mRNAs containing weak 3' splice sites (reviewed in references 7 and 42). Based upon experiments using purified components, Zuo and Maniatis (60) proposed that SR proteins bound to ESEs facilitate recruitment of U2AF⁶⁵ to the Py tract via bridging interactions mediated by U2AF³⁵ (4, 13, 14, 37). Other results, however, argued that U2AF⁶⁵ recruitment was not the rate-limiting step in ESE-dependent splicing (20, 26).

We have previously shown that splicing of a mouse immunoglobulin M (IgM) M1-M2 pre-mRNA substrate requires both U2AF⁶⁵ and U2AF³⁵ (16). Exon M2 contains the founding member of the purine-rich class of ESEs (48). Because of the different sets of proposed U2AF³⁵-mediated interactions mentioned above, we set out to investigate whether the dependence on U2AF³⁵ for IgM splicing correlated with the presence of the purine-rich ESE or with sequences at the 3' splice site. Our results indicate that IgM M1-M2 is an AG-dependent pre-mRNA and that U2AF³⁵ dependence correlates with AG dependence but not with the presence of the purine-rich ESE.

One common feature of the two models for U2AF³⁵ function is that direct or indirect interactions of U2AF³⁵ with nearby sequences stabilize U2AF⁶⁵ binding to the Py tract. Here we show that although interaction of U2AF³⁵ with the 3' splice site AG can stabilize U2AF⁶⁵ binding, not all the activities of U2AF³⁵ correlate with increased cross-linking of U2AF⁶⁵ to the Py tract. The additional function is strongly dependent on the identity of the first nucleotide of the 3' exon and requires only the pseudo-RRM (RRM) motif present in U2AF³⁵. Taken together, the data indicate that U2AF³⁵ has a dual function in the splicing of AG-dependent pre-mRNAs.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Plasmids. pAdML and pµM (IgM M1-M2) were described previously (48, 57). 5' AdML (for adenovirus major late transcripts)-IgM was prepared by replacing the 3' half of the intron and exon 2 of AdML with the corresponding part of the IgM substrate. In 3' exon AdML-IgM, the enhancer containing exon M2 of IgM was replaced by exon 2 of AdML. In both cases the insert (3' half of IgM and 3' exon AdML) and the corresponding part of the vector (pAdML and pµM) were amplified by PCR and joined by blunt-end ligation. All other mutants were prepared via PCR-based site-directed mutagenesis of the plasmids mentioned above, as described elsewhere (19), using TaqPlus Precision DNA polymerase (Stratagene). All mutant clones were confirmed by sequencing.

Preparation of HeLa nuclear extract. HeLa nuclear extract was prepared as described by Dignam et al. (8).

Depletion of U2AF by oligo(dT)-cellulose chromatography. HeLa nuclear extract was depleted of U2AF exactly as described previously (46) by passing the extract over an oligo(dT)-cellulose column at 1 M KCl. The column flow-through of this procedure yields the depleted nuclear extract (odTNE).

Expression and purification of recombinant proteins. U2AF⁶⁵ was expressed as a glutathione S-transferase (GST) fusion protein in Escherichia coli as described previously (27). The plasmid used for expression was described previously (53). The purified protein was dialyzed against buffer D (20 mM HEPES [pH 8.0], 0.5 mM EDTA, 20% glycerol, 1 mM dithiothreitol [DTT], 0.05% NP-40) with 100 mM KCl.

In vitro transcription of splicing substrates. Transcription templates were generated by PCR using plasmids harboring the sequences of AdML, IgM M1-M2 (pµM), or mutant derivatives of both substrates (Fig. 1) preceded by an SP6 promoter. SP6 primer and a reverse primer annealing to the IgM exon enhancer sequence or AdML exon 2 were used to generate templates for full-length transcription templates. The same reverse primers were used for the 3'-half substrates in combination with a forward primer containing the T7 promoter followed by a sequence annealing approximately 20 nucleotides upstream of the BP.

View larger version (19K): [in this window] [in a new window]   FIG. 1.   Pre-mRNA splicing substrates. (A) Schematic representation of AdML, IgM, and mutant splicing substrates. Black boxes and thick lines represent the AdML exons and intron, respectively; white boxes and thin lines depict the IgM M1-M2 exons and intron. yBP indicates the yeast consensus BP sequence, TACTAAC. (B) Sequence of splicing substrates at the 3' splice site, including BP, Py tract, and part of the 3' exon (in capitals). The BP in the wild-type substrates is underlined; sequences in bold indicate mutated nucleotides.

In vitro splicing assays and spliceosome assembly reactions. Splicing reactions, splicing complementation assays, and spliceosome assembly reactions were performed as described previously (16). Spliced products were resolved on 13% denaturing polyacrylamide gels in Tris-Borate-EDTA buffer and spliceosomal complexes were resolved on native 4% acrylamide:bisacrylamide (80:1)-0.5% agarose gels in 50 mM Tris base-50 mM glycine buffer. Gels were exposed to PhosphorImager screens (Fuji BAS-MP).

UV cross-linking and immunoprecipitation. The UV cross-linking and immunoprecipitation experiments were performed exactly as described elsewhere (16). Gels were exposed to PhosphorImager screens, and the intensity of the bands was quantified.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Mapping sequences that make IgM M1-M2 pre-mRNA U2AF³⁵ dependent. We have used oligo(dT) chromatography at 1 M KCl to deplete U2AF from HeLa nuclear extracts (46). The flow-through of the column represents an extract unable to support in vitro splicing assays unless complemented with U2AF activity. For some splicing substrates that contain strong 3' splice site signals, e.g., AdML or -globin, complementation can be achieved with U2AF⁶⁵ alone (16, 44, 46, 53) (Fig. 2A). For other pre-mRNAs, e.g., IgM M1-M2 pre-mRNA (referred to as IgM), which contains relatively weak Py tract and BP sequences, significant levels of complementation require the presence of both U2AF subunits (16) (Fig. 2B). As a first step to determine sequence elements responsible for the U2AF³⁵ dependence exhibited by IgM pre-mRNA, chimeric RNAs comprising parts of AdML and IgM as well as additional IgM RNAs containing mutations within the 3' splice site signals were generated. Figure 1 shows the mutants used in this study.

View larger version (50K): [in this window] [in a new window]   FIG. 2.   In vitro splicing reconstitution assay. Radioactively labeled pre-mRNAs were incubated under splicing conditions, and RNAs were isolated and fractionated by electrophoresis on denaturing 13% polyacrylamide gels. Pre-mRNAs were incubated in HeLa nuclear extract (NE) or odtNE in the absence or presence of recombinant GST-U2AF65 and His-U2AF35 (for protein concentrations, see below). Splicing substrates and products are indicated schematically on the left of each panel, represented as in Fig. 1A. (A) AdML pre-mRNA, with 90 nM GST-U2AF65 in lanes 3 and 4 and 210 nM His-U2AF35 in lane 4; (B) mouse IgM M1-M2 minigene, with 90nM GST-U2AF65 in lanes 3 and 4 and 210 nM His-U2AF35 in lane 4 and 90, 180, and 270 nM GST-U2AF65 in lanes 7, 8, and 9, respectively; (C) 5' AdML-IgM, a chimeric RNA comprising the 5' half of the AdML and the 3' half of the IgM substrate, with 90nM GST-U2AF65 in lanes 3 and 4 and 210 nM His-U2AF35 in lane 4.

View larger version (60K): [in this window] [in a new window]   FIG. 3.   A U-rich Py tract, but not a consensus BP, renders IgM U2AF35 independent. Radioactively labeled yBP-IgM (A) or Py AdML-IgM RNA (B) was incubated in HeLa nuclear extract (NE) or odtNE in the absence or presence of 90 nM GST-U2AF65 and 210 nM U2AF35 and were analyzed as described for Fig. 2. Splicing products and intermediates are indicated on the left of each panel and are represented as in Fig. 1A.

IgM is an AG-dependent substrate. Substrates with weak Py tracts are usually AG dependent, i.e., the AG dinucleotide at the 3' splice site is already required for spliceosome assembly and the first catalytic step. In at least one -globin pre-mRNA derivative, such AG dependence correlated with a requirement for U2AF³⁵ (51). To investigate whether the requirement for U2AF³⁵ in IgM splicing also correlated with the AG dependence of this substrate, the sequence at the 3' splice site of IgM was changed from AG/G into GA/C (designated 3'ss GA/C-IgM; Fig. 1B). This pre-mRNA remained completely unspliced (Fig. 4A, lane 4) under conditions that allowed accumulation of substantial amounts of spliced products from the wild-type RNA (Fig. 4A, lane 2). Concurrently, formation of splicing complexes was significantly reduced in the 3' splice site GA/C mutant (Fig. 4B, compare lanes 2 and 4). This is in contrast with results obtained with AdML, where the same mutation (designated 3'ss GA/C-AdML; Fig. 1B) still allowed the first catalytic step of splicing to occur, leading to an accumulation of splicing intermediates (Fig. 4A, lane 15). Consistent with our predictions and previous observations (36), combination of the GA/C substitution with replacement of the IgM Py tract by that of AdML (designated Py AdML 3'ss GA/C-IgM; Fig. 1B) resulted in a pre-mRNA which could undergo the first step of splicing, leading to accumulation of splicing intermediates (Fig. 4A, lanes 7 and 8). For this substrate the first step of splicing could be partially restored by addition of U2AF⁶⁵ alone (Fig. 4A, lane 10), as was the case for AdML. As the AdML Py tract rendered IgM pre-mRNA splicing both AG independent (Fig. 4A) and U2AF³⁵ independent (Fig. 3B), these results are consistent with the idea that interaction between U2AF³⁵ and the 3' splice site AG is important for splicing of the U2AF³⁵-dependent IgM pre-mRNA.

View larger version (33K): [in this window] [in a new window]   FIG. 4.   IgM is an AG-dependent substrate. (A) Splicing assays. The splicing substrates indicated at the top of each panel were incubated in HeLa nuclear extract (NE) in the presence or absence of ATP or in odtNE in the presence of recombinant U2AF subunits (90 nM U2AF65, 210 nM U2AF35) as indicated. Products were analyzed as described for Fig. 2. IgM splicing products are indicated on the left, and AdML splicing products are indicated on the right. (B) Spliceosome assembly assay. IgM and 3'ss GA/C-IgM RNAs were incubated in HeLa nuclear extract (NE) in the absence or presence of ATP. IgM was also incubated in odtNE (right panel) in the presence of recombinant U2AF subunits or the U2AF-containing column eluate (GUA). After incubation for 20 min the mixtures were loaded onto native polyacrylamide composite gels to separate ATP-independent hnRNP complexes (complex H) from ATP-dependent prespliceosomes (complex A) and two conformations of the spliceosome (complex B/C). wt, wild type.

Dual function of U2AF³⁵. Wu et al. (51) demonstrated that in vitro splicing of an AG-dependent derivative of the -globin pre-mRNA was strongly stimulated by U2AF³⁵ and that U2AF³⁵ increased binding of U2AF⁶⁵ to the Py tract of this substrate. To test whether this was also the case for IgM under our experimental conditions, we carried out UV cross-linking and immunoprecipitation of U2AF⁶⁵ from reactions set up under splicing conditions. The 3'-half RNAs corresponding to wild-type IgM, 3'ss AG/C-IgM, and 3'ss GA/C-IgM, comprising approximately 20 nucleotides upstream from the BP, Py tract, and the downstream exon including the purine-rich ESE, were utilized. U2AF⁶⁵ cross-linking to these RNAs was specific to the Py tract and could no longer be detected when the Py tract was mutated or deleted (16 and data not shown).

View larger version (36K): [in this window] [in a new window]   FIG. 5.   Cross-linking of U2AF65 to IgM and 3'ss GA/C-IgM. The radioactively labeled 3'-half RNAs of IgM and 3' splice site mutants (3' ss AG/C-IgM and 3'ss GA/C-IgM) were incubated under splicing condition in HeLa nuclear extracts (NE) or odtNE, with or without GST-U2AF65 (concentrations [conc.] are indicated above each lane) or 210 nM His-U2AF35. The mixtures were irradiated with UV light and U2AF65 immunoprecipitated with specific anti-U2AF65 antibodies. The precipitates were fractionated on sodium dodecyl sulfate-10% polyacrylamide gels, and the dried gels were exposed to a phosphorimager screen. The positions of GST-U2AF65, endogenous U2AF65, and U2AF35 are indicated on the left. Quantification of the phosphorimager signals corresponding to GST-U2AF65 cross-linking is shown in the lower panel. The signal obtained in odtNE without added protein was used as background, and the value was deducted from values obtained for GST-U2AF65 cross-linking. The value for 90 nM GST-U2AF65 was set to 300 arbitrary scan units, and the remaining values were scaled accordingly to be able to directly compare them. wt, wild type.

View larger version (33K): [in this window] [in a new window]   FIG. 6.   A consensus 3' splice site, but not the ESE, is required for optimal U2AF35 function on IgM. (A) Splicing complementation assay using 3' exon AdML 3'ss AG/C-IgM and 3' exon AdML 3'ss AG/G-IgM pre-mRNA substrates in HeLa nuclear extract (NE) or odtNE supplemented with 90 nM GST-U2AF65 and 210 nM His-U2AF35 as indicated above each lane. After incubation the RNA was isolated and splicing products were separated on denaturing 13% polyacrylamide gels. (B) Cross-linking of U2AF65 in the reactions shown in panel A. The 3'-half RNAs corresponding to 3' exon AdML 3'ss AG/C-IgM and 3' exon AdML 3'ss AG/G-IgM were incubated under the same conditions as those described for panel A. The mixtures were then irradiated with UV light, and U2AF65 was immunoprecipitated with specific anti-U2AF65 antibodies. Precipitated proteins were separated on sodium dodecyl sulfate-10% polyacrylamide gels and exposed to a phosphorimager screen. The positions of GST-U2AF65, U2AF65, and U2AF35 are indicated on the left. The lower panel shows a quantification of the signals corresponding to U2AF65. (C) Splicing complementation assay using the 3'ss AG/C-IgM pre-mRNA performed as described for panel A.

The additional function of U2AF³⁵ is dependent upon the presence of a consensus 3' splice site. To investigate whether the purine-rich ESE present in IgM exon M2 plays a role in U2AF³⁵ dependence, this sequence was deleted. Accumulation of spliced products from this substrate was significantly reduced under our experimental conditions (data not shown), thus precluding further analysis in depleted extracts. As a second attempt to study the role of the M2 ESE, a mutant was generated in which exon M2 was replaced by exon 2 of AdML pre-mRNA, which does not contain an identifiable purine-rich ESE. This chimeric RNA (designated 3' exon AdML 3'ss AG/C-IgM; Fig. 1A and B) could be spliced in nuclear extracts, arguing that the purine-rich ESE present in exon M2 is not essential for splicing of the IgM intron. Surprisingly, however, addition of U2AF⁶⁵ and U2AF³⁵ failed to restore significant levels of splicing in U2AF-depleted extracts (Fig. 6A, lanes 1 to 4). One possible explanation for this observation was that although it is dispensable for splicing in complete extracts, the M2 ESE was required for complementation by the U2AF heterodimer in oligo(dT)-depleted extracts. An alternative possibility was that other sequences within AdML exon 2 were suboptimal for the function of the heterodimer in the complementation assay. One difference between wild-type IgM and 3'exon AdML 3'ss AG/C-IgM is that the first nucleotide of the AdML exon 2 is a cytidine, and therefore the chimeric 3' splice site diverges from the consensus AG/G, which is present in IgM. To test whether the presence of a consensus 3' splice site was important for complementation, the first nucleotide of AdML exon 2 was changed from C to G to re-create a consensus 3' splice site (designated 3'exon AdML-3'ss AG/G-IgM; Fig. 1B). This RNA was spliced in nuclear extracts, and addition of U2AF⁶⁵ and U2AF³⁵ could restore splicing in depleted extracts (Fig. 6A, lanes 5 to 8). Phosphorimager-based quantification of the signals revealed that the levels of activation [ratio between products of splicing and unspliced pre-mRNA compared to the background in oligo(dT)-depleted extracts] obtained by addition of U2AF⁶⁵ and U2AF³⁵ to depleted extracts were comparable for IgM and 3'exon AdML-3'ss AG/G-IgM substrates (between 3- and 3.5-fold in three independent experiments), whereas U2AF⁶⁵ alone failed to activate splicing (Fig. 2B and 6A). We conclude that the presence of a consensus 3' splice site is important for the activity of U2AF³⁵ in the complementation assay.

U2AF³⁵ RRM is sufficient to activate IgM splicing. To further investigate a possible role for ESEs in U2AF³⁵-dependent splicing, an experiment was designed to test a critical tenet of the recruitment model for exon enhancer function: SR proteins bound to the enhancer sequence establish protein-protein interactions with U2AF³⁵ that increase the local concentration of the U2AF heterodimer and facilitate U2AF⁶⁵ binding to the Py tract (13). Evidence has accumulated that these protein-protein interactions involve the RS domains present in both SR protein family members and U2AF³⁵ (1, 15, 21, 25, 49). Recently, Zhu and Krainer have shown that RS domains are particularly critical for the activity of enhancer-bound SR proteins when the substrate analyzed has a weak Py tract and is U2AF³⁵ dependent, as is the case for IgM (54). U2AF³⁵ consists of a carboxy-terminal RS domain, two zinc fingers, and a region that, although significantly homologous to RRMs, shows some unusual sequence features and has therefore been dubbed RRM (3). The result presented in Fig. 7 indicates that a U2AF heterodimer consisting of U2AF⁶⁵ and only the RRM of U2AF³⁵ can restore splicing of IgM in oligo(dT)-depleted extracts (Fig. 7, lane 4). This result indicates that the RS domain of U2AF³⁵ is not essential for the biochemical activity tested in our assays and therefore that RS-mediated protein-protein interactions with SR proteins bound to the IgM ESE are not indispensable for the function of U2AF³⁵ in our experimental system. Consistent with the proposal that U2AF³⁵ needs to recognize the 3' splice site to restore activity to oligo(dT)-depleted extracts, preliminary results indicate that the RRM domain can be cross-linked to a pre-mRNA site-specifically labeled at the 3' splice site (S. Guth and E. Kellenberger, unpublished data).

View larger version (36K): [in this window] [in a new window]   FIG. 7.   The U2AF35 RRM is sufficient to provide U2AF35 activity for splicing of IgM in oligo(dT)-depleted nuclear extracts. Radioactively labeled IgM RNA was incubated in NE or odtNE supplemented with either U2AF65 alone, with additional U2AF35 RRM, or with full-length U2AF35 and was analyzed as described for Fig. 2. A schematic representation of the U2AF35 domain structure and the truncation mutant are shown on the right.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

The tight association of U2AF³⁵ with the essential splicing factor U2AF⁶⁵, the significant phylogenetic conservation of U2AF³⁵ across metazoa, and the essential nature of the gene in Drosophila melanogaster and Caenorhabditis elegans argue that U2AF³⁵ plays an important function in the splicing process (38, 59). Early biochemical results, however, failed to show an absolute requirement for this factor in the splicing of model substrates in vitro (16, 44, 46, 53) (Fig. 2). More recent work indicated that, although not essential to complement U2AF-depleted nuclear extracts, U2AF³⁵ did enhance splicing (14, 60). This enhancement was found to be particularly critical for achieving significant levels of splicing of pre-mRNAs like mouse IgM, suggesting substrate-specific differences in the requirement of U2AF³⁵ (16). It is also important to point out that the apparent degree of U2AF³⁵ dependence is affected by the precise conditions of the biochemical assays. Thus, U2AF³⁵ requirement is more stringent when extracts are depleted of U2AF by chromatography in oligo(dT)-cellulose than when they are immunodepleted by using antibodies against U2AF⁶⁵ (16, 20). It is possible that chromatographic depletion of factors in addition to U2AF [e.g., PUF60 (32)] makes U2AF³⁵ function more rate limiting in oligo(dT)-depleted extracts. Therefore it seems likely that discrepancies in U2AF³⁵ requirement observed in different laboratories (14, 16, 20, 60) result from differences in the sources of depleted extracts and recombinant proteins as well as on the particular pre-mRNA tested. Codepletion of other factors could also explain why some of the substrates used in this manuscript (e.g., 3' exon AdML-IgM and 3'ss AG/C-IgM) failed to be spliced in depleted extracts complemented by U2AF⁶⁵ and U2AF³⁵ but were processed in complete extracts. Differences in assays and specific RNA constructs could also be at the basis of discrepancies on whether or not ESE sequences are required for IgM splicing, depending on the presence or absence of other exonic sequences, including a downstream inhibitory sequence element (14, 20).

The data presented in this report indicate that U2AF³⁵ dependence correlates with AG dependence (Fig. 2, 4A, and 5). In addition, the nature of the first nucleotide in exon 2 has a strong influence on U2AF³⁵ function, with guanosine being significantly more efficient than cytidine (Fig. 6). This result is consistent with in vitro selection studies by Wu et al. showing that while only uridine-rich sequences were selected by U2AF⁶⁵, an adjacent AG/G sequence was selected when the process was carried out using the U2AF^65/35 heterodimer (51). Interaction between U2AF³⁵ and the 3' splice site AG/G stabilizes U2AF⁶⁵ binding (30, 51, 58), and the cross-linking data in Fig. 5 are certainly consistent with this. The question, however, is whether U2AF⁶⁵ binding is rate limiting for splicing under the conditions of a specific substrate and assay. The combined results of Fig. 2B and 5, together with those of Fig. 6, suggest that U2AF⁶⁵ binding is not rate limiting for splicing under the conditions used in our assays. While similar levels of cross-linking were observed in the presence of U2AF⁶⁵ (90 nM) and U2AF³⁵ or in the presence of higher concentrations (270 nM) of U2AF⁶⁵ alone (Fig. 5), significant levels of splicing were observed only in the presence of U2AF³⁵ (Fig. 2B). This argues for a function of U2AF³⁵ in addition to assisting U2AF⁶⁵ binding. The fact that providing a Py tract with higher affinity for U2AF⁶⁵ results in U2AF³⁵ independence (Fig. 3C) or ESE independence (14, 28, 43) does not necessarily mean that the exclusive role of U2AF³⁵ is to improve U2AF³⁵ binding. Given the highly cooperative nature of spliceosome assembly, the additional role provided by U2AF³⁵ in our experimental conditions could promote a set of interactions whose final effect can also be achieved by U2AF⁶⁵-Py tract interactions of higher affinity. It is worth pointing out in this context that the stimulation of U2AF⁶⁵ cross-linking by U2AF³⁵ reported here using the 3'-half RNAs is stronger than that observed using a full-length IgM RNA specifically labeled at the Py tract region (16). A possible explanation for these observations is that factors bound to the 5' splice site stabilize U2AF⁶⁵ binding in a U2AF³⁵-independent manner, thus attenuating the apparent contribution of U2AF³⁵ to U2AF⁶⁵ binding.

Graveley et al. have recently reported that splicing of an IgM-derived construct, in which an ESE was replaced by a binding site for the bacteriophage RNA binding protein MS2, was strictly dependent upon addition of a fusion protein between MS2 and an SR protein RS domain (14). Efficient splicing reconstitution with this substrate in oligo(dT)-depleted extracts required, in addition to MS2-RS, both U2AF⁶⁵ and U2AF³⁵, which is consistent with our observations using ESE-containing IgM. These observations supported the notion that RS-mediated interactions are important for splicing activation and required U2AF³⁵. As MS2-RS also increased cross-linking of U2AF⁶⁵ and U2AF³⁵ to the pre-mRNA under conditions of limited amounts of nuclear extract, these observations were in agreement with the idea that RS domain-mediated interactions facilitate U2AF recruitment (14), possibly through interactions with the RS domain of U2AF³⁵ (49, 60). The levels of U2AF⁶⁵ cross-linking in the reconstituted reaction mixtures, however, were not analyzed in these experiments. As a direct correlation between splicing and U2AF⁶⁵ cross-linking in parallel reconstitution reactions was not established, it is possible that recruitment of U2AF⁶⁵ is not the rate-limiting step in oligo(dT)-depleted extracts supplemented with MS2-RS, U2AF⁶⁵, and U2AF³⁵ using the IgM-MS2 substrate, as was the case in our experiments using wild-type IgM. These arguments do not detract from the fact that the experiments of Graveley et al. persuasively argue that SR proteins bound to the IgM ESE play an important role in IgM splicing. This function could be related to antagonizing silencing effects of other sequences (20) and/or facilitating other steps in spliceosome assembly, including U1 snRNP binding to the 5' splice site through interactions with the splicing coactivator SRm160/300 (4, 9, 13, 26). Indeed, recent results strongly argue that ESEs can act through different mechanisms depending on the specific sequence configuration of the 3' splice site (54). This is consistent with the widespread role that ESEs play in both constitutive and regulated splicing, as highlighted by the high incidence of ESE mutations in disease (reviewed in reference 18).

Alternatively (or additionally), an important role of ESEs in normal extracts could indeed be to facilitate U2AF recruitment, and results from a number of laboratories and different pre-mRNA substrates are certainly consistent with this (5, 14, 37, 47). Also consistent are the results presented in Fig. 4 and 5, showing that replacement of the 3' splice site AG/G by GA/C, a mutation that abolishes U2AF³⁵ binding (30, 51, 58), decreased endogenous U2AF⁶⁵ cross-linking (Fig. 5, lanes 9 and 19) concomitantly with decreased spliceosome assembly (Fig. 4B) and failure to undergo splicing (Fig. 4A).

Reconstitution of depleted extracts requires amounts of recombinant subunits apparently in excess of the levels of the endogenous factors present in undepleted nuclear extracts (data not shown). It is conceivable that this excess, combined with other effects derived from the chromatographic depletion procedure, can change the rate-limiting step for splicing. Although this could, in principle, question the significance of results obtained by using depleted nuclear extracts, we believe that the experimental conditions described here have been useful to unravel a function for U2AF³⁵ in splicesome assembly in addition to its role in promoting U2AF⁶⁵ binding to the Py tract. This function is influenced by the nature of the first nucleotide after the 3' splice site. Proper interaction of U2AF³⁵ with this residue could result in a particular conformation of U2AF³⁵ (or even U2AF⁶⁵) that exposes domains required, directly or indirectly, for U2 snRNP recruitment (e.g., interactions with the 17S U2 snRNP component SAP155 [12]). Another possible scenario is that U2AF³⁵ facilitates binding of SR proteins to the downstream exon enhancer, which in turn could lead to the recruitment of other factors and complexes that activate splicingincluding SRm160/300 (4)or that antagonize inhibitory signals (20). Intriguingly, the result with the U2AF³⁵ deletion mutant of Fig. 7 suggests that whatever the nature of this function, it can be provided by the RRM alone.

In summary, we propose that U2AF³⁵ plays a dual function in splicing, separable under different experimental conditions, and that the precise sequence context of the 3' splice site strongly influences the effects of U2AF³⁵. The two subunits of U2AF work in concert to recognize the 3' splice site region and promote U2 snRNP binding. Depending on the given sequence context of different pre-mRNAs, the role of U2AF³⁵ may be critical for either one or both of its functions. The multiple interactions and their different relative contributions can set the stage for versatile substrate-specific regulation of splice site recognition at the earliest steps of spliceosome assembly.