INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Alternative splicing of pre-mRNA is responsible for the production of multiple mRNA and protein products from individual genes. In many cases, different protein isoforms have unique functions, and their production is tightly regulated at the splicing level. Although a common form of alternative splicing involves the omission or skipping of specific exons during splicing, there are many examples of alternative splicing in which two or more alternative 5' splice sites (5'SSs) compete for joining to a single 3' splice site. In such cases, both sites may be used ubiquitously, or their use may be stringently regulated. Similarly, the deliberate introduction of a duplicate 5'SS may result in use of either site or both sites, depending on the precise context, including the sequence of the sites, their relative 5'-3' order, separation, adjacent sequences, or secondary structures (9, 28, 31, 33, 46, 57, 58, 66, 73). Some of these influences reflect the activity of trans-acting factors.

One trans-acting factor is the U1 snRNP, the RNA component of which forms base pairs across the 5'SSs. The strength of base pairing correlates with the choice of 5'SSs in vivo, at least in some circumstances (39, 77, 95), suggesting that a low probability of binding by U1 snRNP and different affinities for various 5'SSs can dictate 5'SS preferences. This idea is consistent with the observation in Saccharomyces cerevisiae that interactions by components of the U1 snRNP with the cap-binding complex or adjacent pre-mRNA sequences can influence 5'SS selection (34, 71). However, this simple explanation is contradicted in mammals by two sets of results: selective splicing with duplicated sites and a poor correlation of splicing preferences with U1 snRNP binding in vitro (24, 58, 65, 66, 92). Ribonuclease H protection assays have shown that two alternative consensus 5'SSs are occupied simultaneously on each molecule of pre-mRNA by U1-dependent complexes at the initial stages of a splicing reaction, even though only one site (the downstream or intron-proximal site) is used for splicing (32). Thus occupation of one site does not affect binding to another. This finding suggests a model in which U1 snRNP particles bind independently to all alternative 5' splice sites, but the splicing outcome depends on the numbers of sites occupied on each molecule; when the level of binding is low, then whichever site is occupied at a critical point will be used, and the ratio of use depends primarily on the affinity or probability of occupancy at each site (which will depend on the sequence of the site and also other factors, such as the flanking sequences, cap proximity, etc.); at the opposite extreme, when multiple sites are occupied simultaneously by U1 snRNP particles, the downstream site will be used (32, 68). This model offers an explanation for some of the strong polarity effects seen in 5'SS selection, without invoking any inherent polarity in 5'SS complex formation.

After U1 snRNP binding, a 5'SS is incorporated into a commitment complex (47, 60, 61, 78, 93). It is not known whether it is at this stage that the downstream site is selected when U1 snRNP particles are bound to several alternative sites on a molecule of pre-mRNA. After formation of the spliceosome, U1 snRNA is displaced by U6 snRNA (64). Surprisingly, the extent of U6 base pairing to close alternative 5'SSs can determine preferences if there is a sequence nearby to which U1 snRNA can form base pairs (39). Similarly, there are circumstances in which base pairing of U5 snRNA can affect alternative 5'SS preferences (27, 67). However, in neither case is it clear whether the effect is of any significance for the selection of competent alternative sites.

Other factors appear to be able to affect the choice of sites on the basis of the position of the sites rather than their sequences. The first proteins shown to affect splice site selection were the SR proteins (36, 37, 43, 91). These proteins have one or two RNA recognition motif (RRM)-type RNA-binding domains and a characteristic C-terminal domain rich in SR or RS dipeptides (7). They generally stimulate splicing of either constitutive or optional exons (reviewed in references 15, 52, 86), and at least some of their functions are mediated by recognition of specific sequences in the exons (25, 45, 49-51, 59, 72, 74, 75, 82, 84, 85). The SR proteins affect the choice of alternative 5' splice sites as well, and generally they stimulate use of the downstream alternative site both in vitro (14, 22, 32, 37, 43, 54, 91, 96) and in vivo (16, 17, 62, 76, 87, 88, 94). Only p54, which is the most divergent family member, has never been reported to cause a shift in splicing to the downstream 5'SS but only to the upstream one (94).

The effect of most SR proteins on the selection of alternative 5'SSs raises two important questions: do they, as with the inclusion of exons, require specific recognition of sequences in the pre-mRNA, and do they act via U1 snRNP redistribution on the alternative sites? Although only a few substrates have been tested, some results indicate that the effects of SRp20 (16, 76), SRp40, and SRp55 (91, 92) vary with the substrate. Interestingly, exon enhancer sequences have been identified that do affect the choice of alternative 5'SSs in caldesmon and adenovirus E1a pre-mRNA (11, 31). Based on the model for selection of 5'SSs by U1 snRNP, described above, we predicted that SF2/ASF shifted splicing to the downstream site because it tightened U1 snRNP binding indiscriminately, increasing the proportion of pre-mRNA in which the alternative sites were bound simultaneously. This prediction was confirmed by showing that the protein did increase the number of pre-mRNA molecules in which two alternative 5'SSs were occupied simultaneously in nuclear extract (32). Further support for this model came from experiments showing that SF2/ASF directly promoted U1 snRNP binding to a single 5'SS to form a ternary complex (41, 42). This activity requires the C-terminal RS domain of SF2/ASF (41, 42). However, the domain is not required for alternative splicing activity (14, 16, 87, 96). The implication is that the alternative splicing activity is not mediated by effects on U1 snRNP binding. However, no other mechanisms have been proposed, and it is important to reexamine the possibility that the result depended on the native gel electrophoresis assay used.

The other major class of proteins shown to affect 5'SS selection comprises the hnRNP A/B proteins. hnRNP A1, the best characterized member of the group, has activities including nucleocytoplasmic shuttling (69), a possible role in mRNA export (40, 81), and accelerating RNA annealing (21, 63, 70). The proteins have two RRM-type RNA-binding domains and a C-terminal region rich in arginine and glycine. Like other hnRNP proteins, they bind nascent RNA in vivo (30) and are components of the nonspecific H complex that assembles on exogenous RNA in nuclear extracts (6). The intrinsic specificity with which hnRNP A1 binds RNA is controversial (1, 2, 12); purified hnRNP A1 binds RNA cooperatively (21).

In pre-mRNA splicing, the hnRNP A/B proteins generally promote the use of the 5'-most of two alternative 5'SSs and exon skipping both in vitro and in vivo, counteracting SF2/ASF (17, 53-55, 79, 82, 90). Whether these actions depend in general on the recognition of specific sequences is an open question. In certain cases, it has been shown that hnRNP A1 recognizes specific sites within an exon (19, 29) or in flanking intron sequences (10, 23), as a prelude to exon skipping. However, although the specific sites enhanced the effect of exogenous hnRNP A1 on the use of alternative 5'SSs derived from the hnRNP A1 pre-mRNA itself, they were not essential (10). An extensive analysis of the effect of mutations in hnRNP A1 has shown that the strengths of sequence-specific binding and annealing activities are correlated with the ability of mutants to affect alternative splicing of a -globin-derived substrate but that the alternative splicing activity is much more sensitive to the mutations, i.e., that these activities of hnRNP A1 are insufficient for modulation of 5'SS selection (55, 56).

In terms of the model based on U1 snRNP binding described above, the effects of hnRNP A1 on 5'SS selection suggest that it might reduce binding of U1 snRNP and thus the likelihood of simultaneous occupancy of alternative 5'SSs. As a result, the use of specific sites would depend on the individual probabilities that they are occupied, i.e., on their affinity for U1 snRNP, rather than their position. Consequently, we have tested the prediction that hnRNP A1 reduces U1 snRNP binding, and we have determined the mechanisms of this effect. We conclude that the effects of both SF2/ASF and hnRNP A1 are consistent with our model and involve modulations of U1 snRNP binding to alternative 5'SSs, shifting the balance between simultaneous and single occupancy of the alternative 5'SSs on each molecule of pre-mRNA.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Proteins and RNA. hnRNP A1, SF2/ASF, and their mutants were all prepared as recombinant proteins from Escherichia coli as previously described (14, 26, 44, 54) and dialyzed against buffer D. The proteins had no extraneous tags, except for oligohistidine on the RRM1/RS and RRM2/RS mutant forms of SF2/ASF. U1 snRNP was generously supplied by B. Kastner, C. L. Will, and R. Luhrmann (4). HeLa nuclear extracts were supplied by 4C (Mons, Belgium). C175G and AdML WW transcripts have been described (32, 68); CE1a transcripts were derived from pS10 (23); human -globin sequences comprising exon 1, intron 1 (IVS-1), and part of exon 2 were transcribed from plasmid 3'D-205 (73). Mutants of AdML WW containing potential target sites for hnRNP A1 were prepared by incorporating the sequence TAGGGCAGGC, from the K-SAM exon of the fibroblast growth factor receptor 2 gene (29), between nucleotides 40 and 41 from the transcription start site (producing Ad 40), between nucleotides 132 and 133 (producing Ad 132), and at both positions (Ad 40/132). Ad 40/132 contains an additional difference at position 125, which is T rather than C.

Splicing and psoralen cross-linking. AdML transcripts were prepared with m⁷G caps as previously described (68). The final concentrations of components in the splicing reaction mixtures were 0.4 mM ATP, 17 mM phosphocreatine, 2.7 mM MgCl₂, 1.8% (wt/vol) low-molecular-weight polyvinyl alcohol (PVA), 33% nuclear extract, 2.5% (vol/vol) RNasin (Promega), additional 17 mM HEPES-KOH (pH 7.5), 0.005% Tween 20, 33% (vol/vol) buffer D or protein supplements, and AMT-psoralen (HRI) at 0.016 µg/µl. The protein concentrations were as noted in the figure legends. Mixtures were preincubated at 30°C for an hour to permit phosphorylation of SF2/ASF before pre-mRNA was added (38). Cross-linking and quantification were performed as previously described (68), using a PhosphorImager (Molecular Dynamics) and a Cyclone (Canberra Packard) for measurements.

Ribonuclease H cleavage assays. For kinetic assays, incubations were done routinely in volumes of 15 µl, comprising U1 snRNP, hnRNP A1, or mutant proteins, at concentrations given in the figure legends, C175G pre-mRNA, approximately 66% (vol/vol) buffer D [with 0.1 M KCl or 0.08 M potassium glutamate (8)], 3.2 mM MgCl₂, 20 mM phosphocreatine, 1.7 mM ATP, 1.7% (wt/vol) PVA, and 2.7% (vol/vol) Rnasin. In some experiments, 0.05% Nonidet P-40 (NP40) was included (80). These were standard in vitro-binding conditions. Cleavage was initiated typically by addition to all reaction mixtures simultaneously of 5 to 8 µl containing 300 pmol of 14-mer oligodeoxyribonucleotide, 0.05% NP40, and approximately 1 U of RNase H (Pharmacia) in buffer D. Samples of 1 to 1.5 µl were withdrawn simultaneously from the reactions and mixed with 50 µl of proteinase K digestion mixture on ice before incubation, precipitation with ethanol, and gel electrophoresis. Quantification of the extent of cleavage was done with a PhosphorImager, correcting the signals for the numbers of labeled nucleotides in the RNA. The simulations and fitting of reaction kinetics were done with the program KfitSim.

Solid-phase assays. For immobilization of U1 snRNP in microtiter plates, 10-µl aliquots containing 1 µg of U1 snRNP in the standard in vitro-binding buffer were incubated at 4°C in each well of a high-protein-binding vinyl microtiter plate for 2 to 3 h. The wells were blocked for several hours with 2 mg of acetylated bovine serum albumin (BSA) (Sigma) per ml in buffer D-0.05% NP40, washed with the same buffer, and incubated at 30°C for 30 min in standard in vitro-binding buffer with labeled pre-mRNA, SF2/ASF, or hnRNP A1 at concentrations given in the figure legends and 2 mg of acetylated BSA per ml. The unbound RNA was removed by washing with buffer D-0.05% NP40 (and BSA in some experiments). For the experiment summarized in Fig. 2B (see below), one set of wells was washed three times in 1 min, a second set was washed five times in 5 min, and a third set was washed seven times in 10 min. The residual RNA was measured by scintillation counting. For RNase H cleavage of the 5' end of U1 snRNA (see Fig. 7A below), the unbound U1 snRNP was removed after blocking, and then 10-µl portions of standard in vitro-binding buffer were added to each well, containing 43 pmol of specific or arbitrary oligonucleotide, 0.6 U of RNase H, and 2 mg of acetylated BSA per ml. After 30 min at 30°C, wells were washed three times, and the RNA and recombinant-protein components were added as above.

Protein cross-linking. The binding reactions in Fig. 6B and C (see below) were done in 5 µl of standard in vitro-binding buffer (without ATP or phosphocreatine). Each sample was irradiated for 30 s with a broad-wavelength UV source (SpotCure, UVP). After RNase treatment, the samples were run on SDS-PAGE and transferred electrophoretically onto nitrocellulose for phosphorimage analysis.

Other methods. Incubation mixtures for nitrocellulose-binding assays were set up with various volumes rather than masses of macromolecules, and the active proportion of RNA was estimated. The analysis of binding to partially hydrolyzed, end-labeled RNA was done as previously described (5), with the difference that the complexes were recovered on nitrocellulose filters, and the RNA was eluted for analysis by gel electrophoresis.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

SF2/ASF and hnRNP A1 alter 5'SS preferences, but neither causes selective redistribution of U1 snRNP binding to 5'SS. To determine whether the effect of hnRNP A1 on alternative 5'SS selection is associated with a general reduction in U1 snRNP binding, splicing reactions were done with an AdML substrate (AdML WW) that has two copies of the wild-type 5'SS (68). Splicing of this substrate in nuclear extract responded as expected to both SF2/ASF and hnRNP A1, with approximately threefold changes in the ratio of use of the upstream and downstream sites (Fig. 1A, lanes 4, 8, and 12, and B).

View larger version (50K): [in this window] [in a new window]   FIG. 1.   Effects of hnRNP A1 and SF2/ASF on selection of alternative 5'SSs and U1 snRNP binding. (A) Splicing in vitro of AdML WW in nuclear extract (NE) supplemented with SF2/ASF (SF2 [1.2 µM]) or hnRNP A1 (A1 [5.5 µM]) or both (SF2+A1). The splicing reaction mixtures contained AMT-psoralen and were incubated for the times shown (in minutes) above each lane before irradiation and electrophoresis of a portion of each reaction on an 8% denaturing polyacrylamide gel. (B) Ratios of splicing efficiency at the upstream (u/s) and downstream (d/s) 5'SSs. The signals from u/s and d/s mRNA in panel A at 60 min were measured, corrected for label incorporation, and expressed as the ratios of the two isoforms of mRNA produced in each reaction. (C) Analysis of U snRNA cross-links formed in the reactions. Portions of the reactions in panel A were analyzed by electrophoresis on a 5% denaturing polyacrylamide gel. (D) Abundance of the cross-linked adducts. The intensities of the cross-linked U1 snRNA bands at 15 min in panel C are shown as percentages of the pre-mRNA intensities and as a ratio for the two sites; the ratio of the U6 cross-links is plotted also. (E) Correlation between U1 snRNP binding and splicing at the downstream (d/s) site. The graph shows on the ordinate the fraction of splicing to the d/s site at 60 min (d/s mRNA/[u/s mRNA + d/s mRNA]) versus the values for d/s U1 cross-links at 15 min (as in the bar chart). Four points are derived from the experiment shown in panels A and C; four others are derived from an independent experiment (not shown).

hnRNP A1 directly reduces U1 snRNP binding to 5'SS. To test whether the effects of hnRNP A1 are mediated directly, we used an RNase H cleavage assay to measure U1 snRNP binding to consensus 5' splice sites. The substrate for these assays was C175G, a -globin pre-mRNA derivative with a consensus 5'SS 175 nucleotides upstream of a natural 5'SS (32). Binding of U1 snRNP at the consensus 5'SS protects the site against cleavage directed by an oligonucleotide complementary to that site. The proportion of uncut (protected) RNA in a reaction was measured. The consensus 5'SS was protected by purified U1 snRNP (Fig. 2A, lane 1 versus lane 5), unlike a control site (lane 9 versus lane 13). The presence of hnRNP A1 caused a marked reduction in protection (cf. lane 3 with lane 1; 62% of the U1-dependent signal was lost), consistent with a loss of U1 snRNP binding or an enhanced rate of dissociation. SF2/ASF reversed this effect (cf. lane 4 with lane 1), suggesting that this simplified system might contain a sufficient number of components to explain the known antagonistic effects of the two proteins. Time-course analyses of RNase H cleavage confirmed that the effect of hnRNP A1 on U1 snRNP-dependent protection could not be caused by enhanced RNase H activity (data not shown; see below).

View larger version (31K): [in this window] [in a new window]   FIG. 2.   hnRNP A1 reduces binding of U1 snRNP to 5'SSs. (A) Effects of hnRNP A1 and SF2/ASF on U1 snRNP-dependent protection of a consensus 5'SS against RNase H cleavage. C175G pre-mRNA was incubated with RNase H and purified U1 snRNP (0.035 µM), recombinant hnRNP A1 (0.59 µM), or recombinant SF2/ASF (0.24 µM). The components in each reaction are shown by shaded boxes above each lane. Portions of each mixture were incubated for 15 min with one of two cleavage oligonucleotides (sites of cleavage shown by arrows). Reaction mixtures 1 through 8 were incubated with an oligonucleotide complementary to a consensus 5'SS, which is protected by U1 snRNP binding; reaction mixtures 9 through 16 were incubated with an oligonucleotide complementary to an unprotected site. The diagrams at the sides show the substrate, with black bars indicating the structure of the RNA fragment in the corresponding position of the gel. (B) Effects of hnRNP A1 and SF2/ASF on binding of RNA to and dissociation from immobilized U1 snRNP. The labeled RNA retained (cpm) is plotted against the approximate time of washing. Purified U1 snRNP was immobilized in microtiter plate wells (ca. 1 µg per well) and incubated in splicing buffer with 32P-labeled C175G RNA and either hnRNP A1 or SF2/ASF at 0.5 µM. Parallel wells were treated identically but without U1 snRNP [curves labeled ()].

Mutations in hnRNP A1 that blocked alternative splicing activity also blocked its effects on U1 binding. The preceding results showed that hnRNP A1 can reduce the occupancy of 5'SSs by U1 snRNP. It was not clear whether this effect caused the change in splicing patterns, although it occurred when splicing was affected and was predicted by our model. To test whether this property is sufficient for hnRNP A1 to affect alternative 5'SS selection, mutant proteins that lack alternative splicing activity were assayed to determine whether any had retained the ability to reduce U1 snRNP binding. The same mutants have been used to show that the activities of hnRNP A1 in general RNA binding and annealing are not sufficient and may or may not be necessary for alternative splicing function (55). The results of multiple RNase H cleavage assays with purified U1 snRNP are summarized in Fig. 3. All of the hnRNP A1 mutants lost much of their ability to reduce U1 snRNP binding; only one (A1/RS, a chimera of the two RNA-binding domains of hnRNP A1 with the RS domain of SF2/ASF) had any effect, but the reduction was very small compared with that caused by wild-type hnRNP A1. In addition, two other recombinant hnRNP A/B proteins were tested: hnRNP A1^B, an alternatively spliced isoform of hnRNP A1, and hnRNP A2, which is encoded by a separate gene and is 70% identical to hnRNP A1. These proteins have effects on 5'SS selection like that of hnRNP A1, although the effect of hnRNP A1^B is severalfold weaker (55). Like hnRNP A1, both proteins allowed nearly complete cleavage (cf. absence of U1 snRNP), but separate time course experiments showed that hnRNP A1^B and hnRNP B1 (an isoform of hnRNP A2) affected U1 snRNP binding to a lesser extent than did hnRNP A1 and A2 (data not shown). Thus, the results with both the mutant forms of hnRNP A1 and the related proteins were consistent with the hypothesis that the inhibition of U1 snRNP binding may be necessary and sufficient for the effects of hnRNP A1 on alternative splicing.

View larger version (38K): [in this window] [in a new window]   FIG. 3.   hnRNP A1 mutants that are unable to affect alternative splicing do not block U1 snRNP binding. C175G pre-mRNA was subjected to RNase H digestion at the consensus 5'SS for a fixed time after incubation in the presence of U1 snRNP (0.04 µM) and either hnRNP A1, mutant proteins, or other hnRNP A/B proteins (1.0 µM). The proportion of uncut (protected) RNA in each case is expressed relative to the high proportion protected by U1 snRNP in the absence of hnRNP A1. The values shown are the means of 12 determinations in each case, and the error bars represent the standard deviation. The proteins tested are represented to the right of the chart. The two RRMs (boxed) and the Gly-rich C-terminal domain (hexagon) are shown. X indicates mutations in the conserved RNP-1 submotif of an RRM; A1/RS contains the C-terminal RS domain of SF2/ASF (hatched star) instead of the Gly-rich domain; A1-B is an alternatively spliced variant of A1 with an insertion within the C-terminal domain.

hnRNP A1 blocked U1 snRNP association with the pre-mRNA. We considered three mechanisms for the effects of hnRNP A1 on U1 snRNP interaction with 5'SSs: (i) the protein might bind to the RNA and prevent U1 snRNP binding, (ii) it might bind U1 snRNP and prevent it from binding, or (iii) it might interact with U1 snRNP-RNA complexes and stimulate U1 snRNP dissociation.

1

1

View larger version (35K): [in this window] [in a new window]   FIG. 4.   HnRNP A1 does not facilitate U1 snRNP dissociation from 5'SSs or form stable complexes with U1 snRNP or pre-mRNA. Time courses for oligonucleotide-directed RNase H cleavage at the consensus 5'SS in C175G are shown, with the proportion of uncut RNA plotted against the time of RNase H digestion. Components were incubated for some or all of a 20-min period, as shown at the right of the time courses, before RNase H cleavage was initiated by addition of the oligonucleotide complementary to the 5'SS. (A) Effect of hnRNP A1 on U1 snRNP dissociation rates. U1 snRNP was present at 0.04 µM, and hnRNP A1 was present at 1 µM. HnRNP A1 was added with the U1 snRNP (curve 1), after 10 min (curve 2) or just before the cleavage oligonucleotide was added (curve 3). U1 snRNP was added with the oligonucleotide in one sample (curve 5). In this case, RNase H cleavage was unaffected by the addition, indicating that the dissociation of U1 snRNP was irreversible once the oligonucleotides were added. (B) Addition of unlabeled RNA to test whether hnRNP A1 forms stable complexes with pre-mRNA or U1 snRNP. The components were added for the times shown in the diagram. "RNA" is pre-mRNA; "tRNA" is yeast tRNA. The final concentrations, before addition of the oligonucleotides, were as follows: U1 snRNP, 0.04 µM; hnRNP A1, 0.5 µM; yeast RNA, 20 ng/µl.

hnRNP A1 did not block U1 snRNP binding by forming stable complexes. Chase experiments were done to test whether hnRNP A1 formed stable complexes with either the pre-mRNA (RNA) or U1 snRNP. Pure hnRNP A1 was preincubated with either of these components, and then the missing component was added with yeast tRNA for a short time before RNase H cleavage was initiated. The yeast tRNA had no effect on U1 snRNP binding (Fig. 4B, curve 4; cf. curve 3) but blocked the effect of hnRNP A1 (Fig. 4B, curve 2 versus curve 1). If hnRNP A1 formed stable complexes with either component, before the free hnRNP A1 was sequestered by the tRNA, then the hnRNP A1 effect would persist. When hnRNP A1 was preincubated with U1 snRNP, the carrier tRNA eliminated the hnRNP A1 activity (curve 2), and we inferred that the effect of hnRNP A1 is not mediated via stable complexes with U1 snRNP. A small effect was seen in some experiments when hnRNP A1 was preincubated with C175G RNA (Fig. 4B, curve 6), but it was not consistently observed. The reason for this variability may be that the interactions of hnRNP A1 with RNA are very labile (UV laser cross-linking showed that the bulk of hnRNP A1 dissociates from RNA in seconds [I. C. Eperon and O. Makarova, unpublished data).

hnRNP A1 competed with U1 snRNP for binding to pre-mRNA. Definitive evidence in favor of mechanism i, direct competition between hnRNP A1 and U1 snRNP for binding to the pre-mRNA, could be achieved by demonstrating that the effect of hnRNP A1 depends on the length of the RNA. If hnRNP A1 formed an inactive complex with the snRNP, the length of the target RNA would be irrelevant; if hnRNP A1 acted by binding to the RNA, then, because binding to RNA is highly cooperative (21), it ought to be much less effective at interfering with U1 snRNP binding to 5'SSs on short RNA sequences. This prediction was tested using as substrates C175G and a short RNA of 15 nucleotides containing a consensus 5'SS. Preliminary experiments with both psoralen cross-linking and RNase H assays showed that U1 binding to the shorter RNA was scarcely affected by hnRNP A1. However, to eliminate variables it was necessary to test both substrates in the same reaction and analyze the outcomes simultaneously. Several different RNA sequences were incubated simultaneously with U1 snRNP immobilized on nitrocellulose in the presence or absence of hnRNP A1. The bound RNA was eluted and analyzed by electrophoresis on a discontinuous denaturing gel. One of the RNA sequences contained no high-affinity binding sites for U1 snRNP (AdML MM), and thus the recovery of the other sequences could be expressed relative to it. Results are shown in Fig. 5A and quantified in Fig. 5B. U1 snRNP immobilized on nitrocellulose retained preferentially the three RNA sequences containing 5'SSs (Fig. 5A, lanes 4 through 6 [cf. lanes 1 through 3]). The inclusion of hnRNP A1 blocked the U1-dependent retention of the two longer RNA sequences, C175G and AdML CC, but the binding of the short RNA was unaffected (lanes 10 through 12 [cf. lanes 4 through 6]). The UP1 fragment of hnRNP A1 was used as a control for nonspecific effects of RNA-binding proteins. UP1 has the RNA-binding domains of hnRNP A1 but no C-terminal domain, and it was shown earlier to have no effect on U1 snRNP binding to 5'SS (Fig. 3). It had a smaller effect than hnRNP A1 in this assay (Fig. 5A, lanes 16 through 18 versus 10 through 12). Thus, hnRNP A1 acts more effectively to block U1 snRNP binding to 5'SSs in long RNA substrates; this observation excludes mechanism ii, in which hnRNP A1 would bind directly to U1 snRNP and sequester it, and it also shows that hnRNP A1 does not act primarily by recognition of and binding to 5'SSs. Instead, we infer that hnRNP A1 blocks U1 snRNP binding as a consequence of binding to RNA either indiscriminately or at specific sites away from the 5'SSs.

View larger version (59K): [in this window] [in a new window]   FIG. 5.   HnRNP A1 does not reduce binding of short 5'SS RNA to immobilized U1 snRNP. Pieces of nitrocellulose were incubated with or without 0.5 µg of U1 snRNP. After blocking and washing, the filters were incubated in splicing buffer with four RNA sequences and either hnRNP A1, UP1 (both at 0.5 µM), or no protein. Reactions were done in triplicate. After washing, the RNA was eluted. (A) Analysis of the RNA on biphasic denaturing gels of 5 and 20% polyacrylamide. AdML MM and AdML CC were derived from AdML WW by mutation of the 5'SSs to respectively prevent recognition or produce consensus sequences (68). (B) Quantification of the results in panel A. The band intensities were expressed relative to the intensity of the AdML MM RNA in each lane, and the mean was determined for each set of triplicates. The standard deviations are shown.

Molecular mechanisms for the antagonistic effects of SF2/ASF and hnRNP A1 on alternative 5'SS selection. SF2/ASF enhances U1 snRNP binding, and its effects are antagonized by hnRNP A1. The enhancement may require initial interactions with the pre-mRNA and then with U1 snRNP (41, 42). Since hnRNP A1 does not interact with SF2/ASF (20) or U1 snRNP (I. C. Eperon, unpublished observations), it is probable that its antagonistic effects arise from its binding to pre-mRNA. The mechanism of action proposed in the previous section for hnRNP A1 suggests two simple possibilities: SF2/ASF might facilitate the binding of U1 snRNP by binding to the RNA at a small number of specific sites, without serious disruption of the hnRNP A1 complex, or it might bind more generally, displacing hnRNP A1 and perhaps opening up any interfering secondary or protein-mediated structures in the RNA. These alternatives were tested with purified proteins.

8

View larger version (40K): [in this window] [in a new window]   FIG. 6.   SF2/ASF and hnRNP A1 compete for binding to pre-mRNA. (A) Cross-linking at equilibrium of proteins mixed in various proportions to labeled C175G pre-mRNA. Each 10-µl reaction contained hnRNP A1 at 0.5 µM and SF2/ASF as indicated. After UV irradiation and RNase digestion, samples were analyzed by SDS-PAGE. The right panel shows the ratio of the signals from the cross-linked proteins (y axis) at the different SF2/ASF concentrations, compared with the molar ratios of the proteins added to the RNA. (B) Lack of effect of a high-affinity site for hnRNP A1 on competition for binding. Cross-linking was done with labeled CE1a RNA, containing either a high-affinity site (UAGAGU) or one of two mutant sites (UAGCGU and UAGCU). Reactions were in 5 µl with hnRNP A1 at 0.5 µM. (C) Reactions were done as in panel B but with mutant SF2RS.

The RS domain of SF2/ASF was not required for the enhancement of U1 snRNP binding to 5'SSs. If SF2/ASF affects the polarity of 5'SS selection by strengthening U1 snRNP binding to both sites, then mutants of SF2/ASF that affect alternative splicing should still affect U1 snRNP binding. Gel mobility assays indicated that this prediction may not hold for proteins lacking the RS domain (41, 42). Because binding assays based on gel electrophoresis may be affected by the net charge of the protein, this issue was reexamined with different assays. Purified U1 snRNP was immobilized in microtiter plate wells and incubated with ³²P-labeled human -globin pre-mRNA, comprising the first intron and flanking sequences. This substrate was used because the level of U1 snRNP binding to the weak 5'SS was low, which increased the proportional response to SF2/ASF. The results (Fig. 7A) showed that the levels of RNA binding to the U1 snRNP were enhanced considerably by the presence of SF2/ASF and even more so by mutant derivatives that are active in alternative splicing: FF-DD, in which two phenylalanine groups in the RNP-1 motif in the first RNA-binding domain were changed to aspartate, and RS, in which the C-terminal RS domain was deleted (14). It was possible that this effect was caused by RNA binding to the protein which, in turn, had bound to U1 snRNP, and that it did not result from the enhancement of U1 snRNA base-pairing interactions with the -globin RNA. To test this possibility, immobilized U1 snRNP was treated with oligonucleotides and RNase H before incubation with the labeled RNA. As shown in Fig. 7B, the level of RNA bound was halved when the oligonucleotide was complementary to the 5' end of U1 snRNA. It is probable that at least half of the stimulation in binding seen in the presence of SF2/ASF and the mutant proteins was produced by an enhancement of U1-5'SS interactions. A further test was done by binding biotinylated C175G RNA to streptavidin-coated beads, followed by incubation with U1 snRNP, recovery of the bound material, and detection of the U1 snRNP polypeptides by Western blotting. The results show that binding of U1 snRNP was enhanced by SF2/ASF and mutant proteins (Fig. 7B). Most of the binding was eliminated if the U1 snRNP was treated beforehand with RNase H and an oligonucleotide complementary to the U1 snRNA but not if it was treated with a control oligonucleotide. Additional controls were done with two further mutants of SF2/ASF (RRM1/RS and RRM2/RS) in which the first or second RNA-binding domains of SF2/ASF had been deleted (14, 16). The mutants bind RNA weakly but lack alternative splicing activity in vitro (14). As predicted, RNase H cleavage assays showed that neither protein affected U1 snRNP binding, even when present at concentrations up to 15 µM (results not shown). We conclude that there is a good correlation between the abilities of SF2/ASF mutants to promote alternative splicing and their abilities to enhance U1 snRNP binding.

View larger version (41K): [in this window] [in a new window]   FIG. 7.   Absence of the RS domain of SF2/ASF does not compromise the enhancement of U1 snRNP binding to 5'SSs. (A) Effects of SF2/ASF mutant proteins on binding of human -globin IVS-1 pre-mRNA to immobilized U1 snRNP. (Left panel) U1 snRNP was immobilized in microtiter plate wells and, after blocking and washing, incubated with 32P-labeled RNA in splicing buffer in the presence of SF2/ASF proteins at 0.5 µM; reactions were done in triplicate, and the mean value for labeled RNA retained was plotted, with the standard deviations shown; control wells were treated identically, but U1 snRNP was omitted during immobilization. (Right panel) Involvement of the 5' end of U1 snRNA in the enhancement of RNA binding by SF2/ASF was determined; reactions were done in triplicate, as above, but before addition of the labeled RNA, the wells were treated with RNase H and either an oligonucleotide complementary to the 5' end of U1 snRNA (-U1 5' oligo) or an arbitrary control oligonucleotide. (B) Binding of U1 snRNP to immobilized RNA was examined. (Left panel) U1 snRNP (0.04 µM) and SF2/ASF proteins (0.5 µM) were incubated with streptavidin-coated beads that had or had not been previously bound by biotinylated C175G RNA; bound U1 snRNPs were eluted and detected after SDS-PAGE by Western blotting with anti-Sm antibodies. (Right panel) U1 snRNP was incubated with -U1 5' or control oligonucleotides and RNase H before addition to the binding mixtures; U1 snRNP was detected by both anti-Sm and anti-U1-A antibodies; the U1 snRNP proteins detected are labeled; the light additional bands are produced by cross-reactivity with the SF2/ASF proteins.

hnRNP A1 high-affinity sites flanking a splice site redirected splicing without a corresponding change in U1 snRNP binding. The binding studies with partially hydrolyzed CE1a RNA had shown that high-affinity sites were not required for efficient binding of hnRNP A1. Likewise, the same site did not reduce the displacement of hnRNP A1 by SF2/ASF (Fig. 6B). This was surprising in view of the effects of such sites on splicing (10, 19, 23, 29). To test whether, in nuclear extracts, such sites would enhance recruitment of hnRNP A1 and magnify the effects of hnRNP A1 on 5'SS selection, we prepared derivatives of AdML WW pre-mRNA in which the 10-nucleotide exon repressor and hnRNP A1 target from the K-SAM exon of fibroblast growth factor receptor 2 (29) were inserted into either or both of two positions: between the cap and the upstream 5'SS (Ad 40) and/or between the two 5'SSs (Ad 132). After UV irradiation of psoralen-containing splicing reactions, splicing and cross-linking in each reaction were measured as in Fig. 1. The results of UV irradiation after incubation in splicing conditions for various times are shown in Fig. 8A, and the effects of exogenous hnRNP A1 on splicing preferences are shown in Fig. 8B.

View larger version (66K): [in this window] [in a new window]   FIG. 8.   Effects of high-affinity binding sites for hnRNP A1 on splicing and U1 snRNP binding. (A) Derivatives of AdML WW containing 10-nucleotide hnRNP A1 target sites from the fibroblast growth factor receptor 2 gene at position 40 or 132 or both positions were incubated in splicing reaction mixtures supplemented where shown (shaded) with SF2/ASF (1.2 µM) or hnRNP A1 (5.5 µM) or both. The reactions contained AMT-psoralen and were irradiated after incubation for approximately 2, 5, 15, or 60 min (left to right in each block of four lanes). Samples were analyzed by electrophoresis on 5 and 8% denaturing polyacrylamide gels. The portions of the 5% gel that resolved the cross-linked adducts are shown. The asterisks above the diagrams of the pre-mRNA show the sites of insertion of the target sites. (B) Effects of hnRNP A1 on splicing efficiencies at the upstream (u/s) and downstream (d/s) 5'SS. The signals from u/s and d/s mRNA in the 60-min reactions (lanes 4, 12, 20, 28, etc., in panel A) were measured and corrected for label incorporation, and they are shown both separately, as percentages of the total RNA (pre-mRNA and splicing intermediates and products) in the reaction and as the ratios of the two. (C) Abundance of the cross-linked adducts of AdML WW and Ad 40/132. The intensities of the cross-linked U1 snRNA bands at 5 min in panel A (lanes 2, 10, 50, and 58) are shown as percentages of the pre-mRNA intensities and as a ratio for the two sites.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

The roles of U1 snRNP in selection of alternative 5'SSs. We have proposed that spliceosome assembly generally incorporates the U1 snRNP-5'SS complex furthest downstream; when U1 snRNP-binding levels are low, then the affinity of binding will determine the likelihood that a site is occupied and used; when binding levels are higher and more than one site is occupied, then the relative positions of the sites determine the outcome (32). Proteins that affect splicing patterns might do so, according to this scheme, without any intrinsic specificity or polarity by affecting the balance between these two conditions. We have shown previously that U1 snRNP-dependent complexes can form simultaneously on two alternative consensus 5'SSs on each molecule of pre-mRNA and that the levels of these complexes were increased by exogenous SF2/ASF, which shifts splicing to the downstream site (32). However, the RNase H cleavage assay used for those experiments does not detect initial (U2-independent) complexes at nonconsensus sites, and it is only an indirect assay of U1 snRNP interactions. Thus, to test whether the model was informative for modulation of splicing by hnRNP A1, we used psoralen-mediated cross-linking in nuclear extracts instead.

Mechanisms of action of hnRNP A1 and SF2/ASF. We demonstrated in Fig. 2 to 4 that purified hnRNP A1 can reduce U1 snRNP binding to 5'SS and that it does so by preventing binding. The results in Fig. 5 and other data (not shown) indicate that long pre-mRNA substrates are more susceptible, which tends to argue against the possibility that hnRNP A1 blocks by forming stable binary complexes with free U1 snRNP. Although far-Western blots did suggest that hnRNP A1 could interact with some U1 snRNP proteins, such interactions were not detected by kinetic assays (Fig. 4B) or by a variety of other techniques. U2 and U4 (but not U1) snRNPs have been found to coimmunoprecipitate with hnRNP A1 in vitro (13), but the interactions involved were not identified. Thus, we assume that the primary cause of the reduction in U1 snRNP binding is competition for binding to the RNA.

View larger version (34K): [in this window] [in a new window]   FIG. 9.   Models for the molecular mechanisms by which hnRNP A1 and SR proteins affect 5'SS selection. (A) The top panel shows hypothetical binding sites of various affinities for two proteins on a pre-mRNA containing two alternative 5'SSs. The distributions take account of the fact that the proteins have marked sequence preferences and yet appear to be able to occupy all available sites on the RNA at concentrations used in alternative splicing assays in vitro. hnRNP A1 is shown to bind in a more regular fashion because of its cooperativity. The affinity of interaction at each site is indicated by a range from + to +++++. The middle panel depicts a possible binding pattern in nuclear extract at a high concentration of SF2/ASF. The high-, middle-, and low-affinity sites for SF2/ASF are occupied by SF2/ASF protein (small circles), although hnRNP A1 (among other proteins) competes for binding (crescent shape) and excludes SF2/ASF from the latter's lowest-affinity sites. Thus, both splice sites are close to a sequence bound by SF2/ASF, and U1 snRNP binding is enhanced by interactions (double-headed arrow) at both sites. Double occupancy leads to use of the downstream site. Double occupancy also results even at normal concentrations of SF2/ASF if the 5'SSs have the consensus sequence, because U1 snRNP binding outcompetes hnRNP A1. The lower panel illustrates the effect of elevated hnRNP A1 concentrations. Only at the highest-affinity sites can SF2/ASF or other SR proteins bind in preference to hnRNP A1. U1 snRNP binding is reduced by competition with hnRNP A1, and so double occupancy of the 5'SSs is unlikely. Thus, splicing reflects the probability of occupancy of the individual sites. In this case, the upstream site is favored because it is close to a high-affinity site for SF2/ASF, permitting some U1 snRNP binding. The equilibrium between hnRNP A1, SF2/ASF, and U1 snRNP at the downstream site is dynamic, and, on molecules in which U1 snRNP bound initially at the upstream 5'SS, it would be expected that U1 snRNP would also bind eventually to the downstream site. By this time, the upstream site might be committed to splicing. This may account for the observation of a slower rate of nonproductive and cap-independent U1 snRNP binding to unused sites (68). (B) The introduction of two high-affinity binding sites (+++++) for hnRNP A1 resulted in an inversion of the normal effects of hnRNP A1 on alternative splicing. Although the shift to the downstream 5'SS was akin to the effect of SF2/ASF, the reductions in U1 snRNP cross-linking at both 5'SSs and the results with the individual high-affinity sites showed that the mechanism is quite different. We propose that the two sites flanking the upstream 5' splice site allow it to be inactivated when hnRNP A1 is elevated, either because they mark out a domain for unusually stable cooperative hnRNP A1 binding (patterned hnRNP A1 proteins) or because two bound sites form a stable interacting complex that mitigates against splicing. U1 snRNP occupancy at the downstream site would be low, because of competition with hnRNP A1, but this site would be used eventually. U1 snRNP may bind also to the upstream site, but the sequestered site is unable to interact with other components. It is not known whether the cooperative binding of hnRNP A1 is propagated in both directions along the RNA or is unidirectional.