hnRNP A1 and the SR Proteins ASF/SF2 and SC35 Have Antagonistic Functions in Splicing of (cid:1) -Tropomyosin Exon 6B*

Mutually exclusive splicing of exons 6A and 6B from the chicken (cid:1) -tropomyosin gene involves numerous regulatory sequences. Previously, we identified a G-rich intronic sequence (S3) downstream of exon 6B. This element consists of six G-rich motifs, mutations of which abolish splicing of exon 6B. In this paper, we investigated the cellular factors that bind to this G-rich element. By using RNA affinity chromatography, we identified heterogeneous nuclear ribonucleoprotein (hnRNP) A1, the SR proteins ASF/SF2 and SC35, and hnRNP F/H as specific components that are assembled onto the G-rich element. By using hnRNP A1-depleted HeLa nuclear extract and add-back experiments, we show that hnRNP A1 has a negative effect on splicing of exon 6B. In agreement with in vitro data, artificial recruitment of hnRNP A1, as a fusion with the MS2

Splicing is the process by which introns from premessenger RNAs are removed in eukaryotes. Pre-mRNA splicing takes place within the spliceosome, which is a large, highly dynamic complex composed of four small ribonucleoprotein particles (snRNP 1 U1, U2, U4/U6, and U5) and multiple non-snRNP factors (1,2). One of the most intriguing questions that remains in RNA splicing is how the 5Ј and 3Ј splice sites are selected and paired together within large RNA sequences (3). This question takes on particular importance in alternative splicing, where the selection of certain splice sites is modulated depending on the developmental stage, on tissue differentiation, or on metabolic changes of the cells (3). Numerous studies have dem-onstrated that regulatory sequences within the pre-mRNA that lie outside the splicing signals play a crucial role in controlling the choice of splicing sites in a given cellular context (reviewed in Refs. 4 and 5).
Among these sequences are the splicing enhancers. These elements are found in a wide variety of metazoan pre-mRNAs, either within exons or introns. Purine-rich splicing enhancers (known as ESE) are a well characterized class of exonic splicing enhancers that mostly interact with specific subsets of SR proteins (reviewed in Refs. 6 and 7). SR proteins belong to a family of essential splicing factors that are highly conserved between Drosophila and mammals and that are involved in both constitutive and regulated splicing events (reviewed in Refs. 8 and 9). It has been proposed that the function of SR proteins is to stimulate the recognition of weak upstream 3Ј splice sites, by recruiting U2AF 65/35 , or to facilitate U1 snRNP binding to the 5Ј splice site (reviewed in Ref. 4). It also has been proposed that ASF/SF2 bound to an ESE element antagonizes the function of hnRNP A1 bound to a juxtaposed exonic silencer element by impairing the propagation of cooperative binding of hnRNP A1 along the exon (10).
Intronic splicing enhancers are another class of elements that have been identified in numerous pre-mRNAs (reviewed in Ref. 4). Analyses of intronic splicing enhancers reveal that most of them have a multipartite structure composed of several distinct motifs that are shared by different intronic splicing enhancers. It was first noticed that there is an enrichment in G triplets and quartets downstream of 5Ј splice sites by computer analysis of human intronic sequences, and it was suggested that they play a role in the splicing process (11). Subsequent studies demonstrated that G-rich motifs are involved in the regulation of alternative splicing of numerous pre-mRNAs. Depending on their position within the pre-mRNA and on the proteins that bind to them, G-rich elements have been shown to function as enhancer or silencer elements. In the case of the cardiac troponin T pre-mRNA, it has been shown that multiple GGGGCUG intronic motifs are required to promote the recognition of the 6-nucleotide microexon 17 (12). It has been suggested that the binding of SF1 to these motifs extends the domain of the microexon, which facilitates exon recognition during early events of spliceosome assembly (13). The inclusion of the c-src N1 exon in neuronal cells requires an intronic regulatory element, named the downstream control element, that is made up of an array of individual motifs that bind to different proteins of the hnRNP family (14,15). Within the downstream control element, a GGGGGCUG motif has been shown to activate splicing of the N1 exon through its interactions with hnRNP H and F (14,16). In contrast, the GUGGG motif at the 5Ј end of the skeletal exon of the rat ␤-tropomyosin pre-mRNA acts as a negative regulator through its interaction with hnRNP H (17). Several G tracts have been found in an intronic enhancer element that stimulate splicing of the thyroid hormone receptor gene (18). This element interacts with hnRNP H (18). Another class of G-rich motifs, which bears homology with the high affinity hnRNP A1 binding site UAGGGU, also is involved in the inhibition of splicing of various pre-mRNAs (19 -31). In those cases, it has been shown that the repression is mediated through its interaction with hnRNP A1. Recent surveys of genomic human sequences that inhibit splicing have also identified several sequences enriched for G-triplets that are capable of promoting exon skipping (32,33).
In a previous study, we identified an intronic G-rich element downstream of exon 6B from the chicken ␤-tropomyosin gene that is thought to be essential for activating splicing between exons 6B and 7 (34). This sequence, called S3, contains several G triplets and quartets, which we have arbitrary divided into three elements. The RG1 element contains a GGGGCUG and an overlapping UGGGC motif. The RG2 element contains two AGGGU motifs flanking a GGGGA motif, and the RG3 element contains a UGGGGU motif (Fig. 1). Mutations of the six G motifs abolish the splicing of the intron downstream of exon 6B and inhibit spliceosomal assembly, suggesting that the G-rich element acts as an enhancer (34). In the present work, we analyze the proteins that are assembled onto the S3 G-rich sequence by RNA affinity chromatography. We identify the proteins hnRNP A/B, hnRNP F/H, ASF/SF2, and SC35 as specific proteins associated with the S3 element. Depletion and add-back experiments combined with ex vivo analysis show that hnRNP A1 inhibits splicing of exon 6B. In contrast, ASF/ SF2 and SC35 activate in vitro splicing of exon 6B. We demonstrate that hnRNP A1 counteracts the stimulatory effect of ASF/SF2 and SC35. In addition, we show that both ASF/SF2 and SC35 are able to displace bound hnRNP A1 on the S3 element. These data suggest that the G-rich intronic element shares overlapping or juxtaposed motifs that act as enhancer or silencer depending on the interacting proteins.

EXPERIMENTAL PROCEDURES
Plasmid Constructions-All ␤-tropomyosin constructs were derived from a 1.7-kb chicken genomic clone spanning exons 4 -7 (35). The pSma and pPmac constructs have been described previously (34,36). The pPmac mtA1 construct was generated by PCR from the pPmac construct with sense primer 5Ј-AGAGCGCGCCGGGTTGACGGGAGC-ACGGTCCTTCACT-3Ј and antisense primer 5Ј-CTAGAAGCTTGATT-TCCT-3Ј. The PCR product was cleaved with BssHII and HindIII and inserted into the pPmac construct, which was cleaved with BssHII and HindIII. The pPmac tri-RG2 was made by cloning a synthetically prepared hybrid oligonucleotide 5Ј-CGCGCAGTGTTGAGTGGAGCAGT-3Ј into the pPmac construct, which was cleaved with BssHII and PpuMI. The transcription template for 5-6A was made by replacing a PstI 6A fragment from 6B⌬4 (37) by a PstI 6A fragment from pSVK6A⌬4 in which the 5Ј splice site of exon 6A has been mutated (38). This mutation introduced a BamHI restriction site. Linearization of the construct by BamHI gave rise to the 5-6A transcript. The transcription template for S3 was obtained by cloning a synthetically prepared hybrid oligonucleotide 5Ј-GCTGGGGCTGGGCAGAGCGCGCAGGGTTGAGGGGAG-CAGGGTCCTTCACTGGGGT-3Ј flanked by EcoRI and HindIII restriction sites into a pGEM-3Z vector, which was cleaved with EcoRI and HindIII. The transcription template for mut S3 was made by cloning a synthetically prepared hybrid oligonucleotide 5Ј-GCTTCATCTCA-CCAGAGCGCGCTCACTTGAGTTCAGCTCTCTCCTTCACTCACCT-3Ј flanked by EcoRI and HindIII restriction sites into pGEM 3Z vector previously cleaved with EcoRI and HindIII. The transcription templates for RG2 and A1 "winner" (19) were made by hybridizing two complementary oligonucleotides containing the SP6 RNA polymerase promoter (see the sequences in Fig. 3). The transcription templates for S3 mtA1 and S3 tri-RG2 were made by PCR from pPmac mtA1 and pPmac tri-RG2 constructs using a sense primer containing the SP6 RNA polymerase (5Ј-ATTTAGGTGACACTATAGAATTCTGGGG-CTGGGCA-3Ј) and an antisense primer (5Ј-CACCCCAGTGAAGGA-3Ј). For the transfection experiments, a ␤-tropomyosin minigene extending from exon 5 to exon 7 was cloned in the pcDNA 3 (ϩ) vector to give pcDNA 980. Plasmid pcDNA 980 -2MS2 was made by replacing a BssHII-PpuMI RG2 fragment of pcDNA 980 by a BssHII-PpuMI PCR fragment of pIII/MS2-1 containing the coat binding sites (gift from Dr. R. Breathnach). The coat expression vector pCI-MS2 and the hnRNP A1-coat fusion vector (gift from Dr. R. Breathnach) have been described (23).
In Vitro Transcription and Splicing Experiments-Capped pre-mRNAs were synthesized in vitro using SP6 RNA polymerase and [␣-32 P]UTP as previously described (39). Transcripts were purified by electrophoresis on polyacrylamide/urea gels. Capped RNAs used as competitors in the splicing experiments were 3Ј-biotinylated as described elsewhere (40). S3 and mut S3 RNAs for affinity selection were transcribed without capping using a SP6 MEGAshortscript kit (Ambion) according to the manufacturer's recommendations. RNAs were then biotinylated as described above.
HeLa cell nuclear extracts were purchased from A. Miller (Computer Cell Center, Mons, Belgium) and prepared as described (41,42). Splicing reactions were performed with 20 fmol of 32 P-labeled pre-mRNA in 10-l reactions with 30% HeLa nuclear extract, except when indicated (36). Competitor RNA and recombinant proteins were added to the splicing reaction on ice prior to adding the labeled pre-mRNA. Reaction products were analyzed on 5% denaturing polyacrylamide gels except when indicated. RNA bands were quantified using a PhosphorImager (Amersham Biosciences). Splicing efficiency was calculated as the ratio between the mRNA value and the sum of the pre-mRNA value plus the mRNA value. For each RNA, the number of uracil residues was taken into account for the calculated splicing efficiency.
RNA Affinity Chromatography, Mass Spectrometry Analysis, and Western Blotting-Coupling to streptavidin-agarose and affinity purification of protein complexes were performed as described elsewhere (45). Proteins were separated on two-dimensional gel electrophoresis. The first dimension was a nonequilibrium pH gradient gel electrophoresis (NEPHGE) using ampholytes 3.5-9.5 (46). The second dimension was a 7.5% PAGE (47). The selected proteins were visualized by silver staining or Coomassie Brilliant Blue G-250. The protein spots were excised from the gel, digested in gel with trypsin, and identified by mass spectrometry (45).
For Western blot analysis of affinity-selected proteins, 100 pmol of RNA bound to 25 l of streptavidin-agarose beads were incubated with 500 l of 30% HeLa nuclear extracts under splicing conditions without polyvinylalcohol and rotated 5 min at 30°C. After washing, proteins were released from the beads by mixing twice in 66% glacial acetic acid and 33 mM MgCl 2 for 30 min at 4°C and then proceeded as described elsewhere (45). The proteins were dissolved in 250 l of SDS-sample buffer. Proteins were separated on a 12% SDS-polyacrylamide gel, electroblotted onto a nitrocellulose membrane, and probed with the following antibodies: anti-hnRNP A1 monoclonal antibodies 9H10 (a gift from Dr. G. Dreyfuss), anti-hnRNP F/H monoclonal antibodies 1G11 (a gift from Dr. G. Dreyfuss), rabbit polyclonal antibodies against the C-terminal peptide of hnRNP F (a gift from Dr. Black), monoclonal antibodies against the N-terminal peptide of ASF/SF2 (a gift from Dr. J. Stévenin), monoclonal antibodies against the C-terminal peptide of SC35 (a gift from Dr. J. Stévenin), and monoclonal antibody mAb 104 monoclonal antibodies (a gift from Dr. Roth). The bands were detected with the SuperSignal West Pico detection kit (Pierce).
Protein Purification-Total SR proteins from HeLa cells were prepared as described elsewhere (48). GST-hnRNP A1 (a gift from Dr. B. Chabot) was expressed in Escherichia coli BL21 (DE3) and purified using glutathione-Sepharose beads (Amersham Biosciences) according to the protocol described by the manufacturer. The untagged hnRNP A1 expression vector (pET9d) was expressed in Escherichia coli BL21 (DE3), and the protein was purified as described (49). Recombinant proteins ASF/SF2 and SC35 expressed in baculovirus and purified from Sf9 cells were a gift from Dr. J. Stévenin (40). Recombinant GST-ASF/ SF2 ⌬RS and GST-SC35 ⌬RS proteins cloned into the pGEX vector were a gift from Dr. J. Stévenin (50).
Affinity Depletion of hnRNP A1-HnRNP A1 was depleted from HeLa nuclear extracts using a biotinylated oligonucleotide (TTAGGGT-

FIG. 1. The G-rich S3 element interacts with cellular factors.
A, schematic representation of the pre-mRNA pSma containing exon 6B (76 nt), exon 7 (41 nt), and the intron between them (125 nt). The S3 sequence is represented with the G motifs in boldface type. The three elements, RG1, RG2, and RG3, are indicated. The mut S3 RNA, in which the G motifs have been mutated, is noted below the wild-type sequence. B, in vitro splicing of pSma pre-mRNA in the presence of an excess of unlabeled RNA competitor. Labeled pre-mRNA was incubated under splicing conditions with 30% HeLa nuclear extract for the indicated times in the absence or in the presence of 8 and 16 pmol of S3 or mut S3 RNA competitors. The products of the splicing reaction were separated by 5% acrylamide, denaturing gel electrophoresis and quantified with a PhosphorImager. The identity of the splicing products is shown at the left. C, gel mobility assays of S3 and mut S3 RNAs with HeLa nuclear extract. Labeled RNAs were incubated under splicing conditions with 30% nuclear extract. Lanes 1 and 4, S3 and mut S3 RNA alone without extract; lanes 2 and 5, S3 and mut S3 RNA incubated with extract at 0°C; lanes 3 and 6, S3 and mut S3 RNA incubated with extract at 30°C for 15 min. D, competition experiments with an excess of S3 and mut S3 RNA. The S3 sequence was incubated with 30% nuclear extract (lane 2) and with either 8 pmol (lanes 3 and 5) or 16 pmol (lanes 4 and 6) of the indicated RNA competitor. Lane 1, S3 RNA alone without nuclear extract. The brackets indicate RNA-protein complexes.
FIG. 2. Proteins are specifically associated with the S3 sequence. Affinity-selected proteins purified from S3 (left) and mut S3 RNA (right) were separated by two-dimensional gel electrophoresis and silver-stained. The black arrows and the inset represent specific proteins assembled on the S3 RNA. The open arrow represents specific proteins associated with mut S3 RNA. Sizes of molecular mass markers are shown. TAGGGGTTAGGGTTAGGG) coupled to streptavidin-agarose beads (50 mol/100 l of beads). The interaction between the oligonucleotide and the beads was for 2 h at 4°C. After washing the beads with buffer D, 100 l of nuclear extract were added (v/v), and the mix was rotated for 2 h at 4°C. The beads were pelleted, and the supernatants were used as hnRNP A1-depleted extract. Mock-depleted nuclear extract was made in the same conditions but with the beads alone. Proteins were separated on a 10% SDS-polyacrylamide gel, transferred to a nitrocellulose membrane, and probed with monoclonal antibodies as described above.
UV Cross-linking Experiments-32 P-Uniformly labeled RNAs were synthesized with [␣-32 P]UTP and [␣-32 P]GTP as described above. 12.5 fmol of RNAs were incubated in a 12.5-l reaction mixture containing 5% HeLa nuclear extracts under splicing conditions without polyvinylalcohol and the indicated amount of recombinant proteins. In competition experiments without nuclear extracts, 0.1 g of tRNA were added as a nonspecific competitor. After a 15-min incubation at 30°C, the samples were irradiated on ice for 15 min in a UV (254 nm) Stratalinker and treated with 1 mg/ml RNase A and 10 units of RNase T1 at 37°C for 30 min. The cross-linked proteins were resolved on a 12% SDSpolyacrylamide gel and visualized by a PhosphorImager.
Transfections and RNA Analysis-QM7 quail myogenic cells were plated at a density of 5 ϫ 10 5 cells/60-mm diameter tissue culture dish in Dulbecco's modified Eagle's medium with 4500 mg/liter D-glucose supplemented with 10% fetal bovine serum and 2% chick serum. The next day, the cells were transfected with 1 g of minigene DNA and 0.5 and 1 g of coat or coat fusion hnRNP A1 expression vectors using FuGene 6 (Roche Applied Science) according to the manufacturer's recommendations. Forty-eight hours later, RNA was harvested from cells using Tri Reagent TM (Sigma), treated by RQ1 DNase I, and analyzed by RT-PCR. Reverse transcription was performed with oligo(dT) and superscript II reverse transcriptase (Invitrogen) for 60 min at 42°C. PCR was carried out with one-twentieth of the RT reaction in . PCR products were resolved on 6% denaturing polyacrylamide gels and quantified using a PhosphorImager.

RESULTS
The G-rich Regulatory Sequence Interacts with Specific Proteins-In a previous paper (34), we showed that the splicing of the intron downstream of exon 6B is stimulated by a G-rich intronic element (named S3). To test whether this element interacts with specific cellular factors, we performed splicing competition experiments with a pre-mRNA containing exon 6B/intron/exon 7 (Fig. 1A, pSma). The RNA used as a competitor contained the wild-type sequence from 21 to 73 nt downstream of exon 6B (Fig. 1A). As a control, we used an RNA (named mut S3) in which the G-rich motifs had been mutated (Fig. 1A). The addition of 16 pmol of unlabeled S3 RNA dramatically decreased exon 6B to 7 splicing (Fig. 1B). After 2 h of incubation, splicing efficiency was reduced 3-fold as compared with that without an RNA competitor (Fig. 1B). In contrast, the  (36). The wild-type RG2 sequence and the two mutated sequences (mtA1 and tri-RG2) are shown below with mutated nucleotides in boldface type. B, in vitro splicing assays in 40% HeLa nuclear extracts. Labeled pre-mRNAs were incubated under splicing conditions for 30 min (lanes 1, 4, and 7), 60 min (lanes 2, 5, and 8), and 90 min (lanes 3, 6, and 9). The products of the splicing reaction were separated by 5% acrylamide, denaturing gel electrophoresis and quantified with a PhosphorImager (Amersham Biosciences). The identity of the splicing products is shown on the left. hnRNP A1 and SR Proteins Regulate Splicing of ␤-TM Exon 6B addition of the same amount of the mut S3 RNA had no effect on or even slightly stimulated the splicing of exon 6B to 7 (Fig.  1B). Splicing inhibition was sequence-specific, because splicing of Adeno pre-mRNA was not affected by the addition of the S3 RNA competitor (data not shown). These results suggested that the stimulatory effect on exon 6B-7 splicing was mediated by cellular factors that specifically interacted with the G-rich element. This assumption was confirmed by gel retardation assays. Incubation of the S3 RNA under splicing conditions with a 30% nuclear extract resulted in a specific high molecular weight complex that did not form with the mut S3 RNA (Fig.  1C). Assembly of this specific complex was very rapid, since it was formed in the first minutes of incubation on ice (Fig. 1C). A faster migrating complex was also observed, whose mobility was approximately the same as the complex formed with the mut S3 RNA (Fig. 1C, lanes 5 and 6). Raising the nuclear extract concentration to 60% did not modify the pattern of the protein complexes; nor did it alter their specificity (data not shown). In agreement with in vitro splicing experiments, the addition of an excess of unlabeled S3 RNA significantly reduced complex formation, whereas the addition of the same molar excess of unlabeled mut S3 RNA had no effect (Fig. 1D, compare lanes 3 and 4 with lanes 5 and 6). Altogether, these results suggested that the G-rich element interacted with cellular factors in a sequence-specific manner.
HnRNP A1 Is a Component of Regulatory Complexes That Binds to the S3 Element-To identify proteins that bind to S3, we used an RNA chromatography method. Biotinylated S3 RNA or its mutated version mut S3, were bound to streptavidin-agarose beads and incubated with 30% HeLa nuclear extracts under standard splicing conditions (45). Since specific complexes assemble very rapidly, the incubation time was for 5 min at 30°C. After washing, the purified complexes were eluted, and the proteins were separated by two-dimensional gel electrophoresis with a nonequilibrium pH gradient in the first dimension, and the gel was then analyzed by silver staining. Fig. 2 shows that both the S3 and mut S3 RNAs were associated with various proteins. A comparison of the protein pattern revealed proteins that were specifically associated with the S3 element. Remarkably, there was a group of basic proteins with molecular mass ranging from 33 to 46 kDa (indicated by an inset in Fig. 2, left panel) that had a two-dimensional gel electrophoresis pattern that closely resembled the one reported for the hnRNP A/B group (51). Mass spectrometry analysis and immunoblotting with antibodies against the hnRNP A1 confirmed that these proteins belonged to hnRNP from A/B group. Northwestern analysis revealed that only hnRNP A1, hnRNP A2, and a third protein of 70 kDa directly contacted the S3 element (data not shown). To determine the minimal RNA motif required to bind to hnRNP A1, gel shift experiments were performed with shortened versions of the S3 sequence. The central region, named RG2, binds to hnRNP A1 to the same extent as S3 (data not shown). The RG2 motif contains two AGGGU that closely resemble the hnRNP A1 "winner" sequence identified by the SELEX approach (19). In agreement with this observation, hnRNP A1 interacted with RG2 with a high affinity, albeit lower than did the A1 "winner" sequence (Fig. 3, A and B). In addition, a protein complex that migrated more slowly was formed at the highest hnRNP A1 concentrations (Fig. 3A, RNA-protein Complex II). This complex was not formed with the A1 "winner" sequence, which excluded the possibility that the formation of this complex was due to aggregations of the hnRNP A1 molecules (Fig. 3, A and B). Altogether, these results showed that the hnRNP A1 was a specific component of the G-rich complex that assembled onto the S3 regulatory sequence.
The hnRNP A1 Binding Site Overlaps or Closely Juxtaposes the ASF/SF2, SC35, and hnRNP F/H Binding Sites-HnRNP A1 was shown to act as a splicing repressor in several alternative splicing models. Therefore, we tested whether mutations in the RG2 hnRNP A1 binding site had an effect upon splicing of exon 6B. Several pre-mRNAs were tested for in vitro splicing assays. The wild-type pPmac construct contained exon 6B to exon 7 and part of the intronic sequence upstream of exon 6B (36). The two mutant constructs were pPmac tri-RG2 and pPmac mt A1, in which three punctual mutations were made in the RG2 motif (Fig. 4A). Compared with pPmac, the two mutant constructs showed decreased splicing of exon 6B (Fig. 4B). Mutations in pPmac mtA1 reduced splicing of exon 6B by 35%, whereas that of pPmac tri-RG2 was reduced by 25%. The experiments were repeated with different preparations of HeLa nuclear extracts, and the splicing of pPmac mtA1 was always less efficient than that of pPmac tri-RG2. One possible explanation for our results was that mutations introduced to disrupt the RG2 hnRNP A1 binding site affected the binding for proteins that were required for splicing of exon 6B. SR proteins are good candidates. Together with hnRNP A1, these proteins are key regulators in numerous alternative splicing models. Consistent with this hypothesis, ESEfinder identified several putative binding sites for ASF/SF2 and SC35 (52). Four sequences were revealed for ASF/SF2. Two juxtaposed CAGAGCG and CGCAGGG motifs were at the beginning of the RG2 motif. An AGCAGGG sequence was found at the end of the RG2 motif. Interestingly, the putative SC35 binding site, CTTCACTG, overlapped the last CACTGGG ASF/SF2 binding site. Whereas mutations aimed to disrupt hnRNP A1 did not affect the last FIG. 5. SR proteins, hnRNP A1, and hnRNP F/H bind to the S3 element. A, sequences of the different RNAs used for the RNA affinity chromatography. RNAs S3, S3 mtA1, and S3 tri-RG2 have a common GAATT sequence that is derived from transcription initiation. B, biotinylated RNAs coupled to streptavidin-agarose beads that were incubated in 30% HeLa nuclear extracts under splicing conditions. After elution, proteins were separated on a 12% SDS-PAGE and immunoblotted with antibodies against the indicated proteins. Note that membranes probed with anti-hnRNP A1 antibodies contain smaller amounts of proteins due to the high titer of antibodies. Lanes 1 and 2 contain 5 and 2.5 g of HeLa nuclear extracts, respectively, except for membrane probed with hnRNP A1, where the amount of nuclear extracts was reduced by two. Lanes 3, 5, and 7 contain 5 l of eluted proteins, except for membrane probed with hnRNP A1, where 2 l were loaded. Lanes 4, 6, and 8 contain 10 l of eluted proteins except for membrane probed with hnRNP A1, where 4 l were loaded.

hnRNP A1 and SR Proteins Regulate Splicing of ␤-TM Exon 6B
putative ASF/SF2 and SC35 motifs, they could have reduced the binding of these proteins at the other motifs or at additional weaker binding sites. To test this possibility, proteins assembled onto S3, S3 mtA1, and S3 tri-RG2 RNAs were separated on an SDS-polyacrylamide gel and then analyzed by Western blotting with monoclonal antibodies against hnRNP A1, ASF/SF2, and SC35 (Fig. 5, A and B). We also probed the membrane with monoclonal antibodies against hnRNP F/H, because a GGGA motif, which was identical to the sequence recognized by all hnRNP H family, was mutated in the construct pPmac tri-RG2 (53). In addition to hnRNP A1, we found ASF/SF2, SC35, and hnRNP F/H in complexes assembled on the wild-type S3 RNA FIG. 6. Depletion of hnRNP A1 from nuclear extracts activates splicing of exon 6B. A, Western blot analysis of hnRNP A1-depleted nuclear extract. hnRNP A1 was depleted from HeLa nuclear extract by using streptavidin beads coupled to a biotinylated oligonucleotide containing four hnRNP A1 binding sites. Proteins were analyzed by Western blotting with monoclonal antibodies against hnRNP A1 (9H10) and hnRNP F/H (1G11). Lanes 1 and 2, mock-depleted nuclear extracts (0.5 and 1 l, respectively). Lanes 3 and 4, hnRNP A1-depleted extracts (0.5 and 1 l, respectively). B, 1 l of the same samples were also probed with monoclonal antibodies against SR proteins (monoclonal antibody 104). C, in vitro splicing of pPmac, 5-6A, and ␤-globin transcripts in various extracts. Labeled pre-mRNAs were incubated with 40% of mock-depleted (mock) for 30, 60, and 90 min (lanes 1-3, 7-9, and 13-15) or 40% of hnRNP A1-depleted HeLa nuclear extracts (⌬A1) for 30, 60, and 90 min ( lanes  4 -6, 10 -12, and 16 -18). The splicing products were run on a 5% polyacrylamide denaturing gel and quantified using a PhosphorImager. RNA precursors and splicing products are indicated by the schematics on both sides. D, the addition of recombinant hnRNP A1 restores splicing repression of pPmac. Labeled pre-mRNAs were incubated for 90 min with 40% of mock-depleted (lanes 1, 6, and 11) or 40% of hnRNP A1-depleted HeLa nuclear extracts (lanes 2-5, 7-10, and 12-15) in the absence (lanes 2, 7, and 12) or presence of 0.5, 0.75, and 1 g of recombinant GST-hnRNP A1 (lanes 3-5, 8 -10, and 13-15). Note that a high background is observed in D as compared with C, although the same preparation of nuclear extracts was used. This is due to freezing/unfreezing of nuclear extracts. hnRNP A1 and SR Proteins Regulate Splicing of ␤-TM Exon 6B (Fig. 5). As expected, the mutations in the RG2 motif decreased the interaction of hnRNP A1 to the S3 mtA1 and S3 tri-RG2 RNAs relative to the wild-type sequence (Fig. 5). Furthermore, these mutations also had an effect on binding of ASF/SF2 and SC35. As shown in Fig. 5, the S3 mtA1 and S3 tri-RG2 RNAs exhibited a reduced binding affinity for ASF/SF2 and SC35, which was more dramatic for S3 mtA1 than for S3 tri-RG2 (Fig.  5, compare lanes 5 and 6 with lanes 7 and 8). In contrast, binding of the hnRNP F/H protein family, as probed either with 1G11 monoclonal antibodies, which recognized hnRNP F/H proteins, or with polyclonal antibodies against hnRNP F, did not seem to be affected by the mutations. Altogether, these results indicated that mutations in RG2 reduced the ability of ASF/SF2 and SC35 to bind to the S3 element, which might explain the decreased splicing efficiency of the mutant transcripts relative to the wild-type transcript. Furthermore, these data suggested that the binding sites for hnRNP A1, ASF/SF2, and SC35 overlapped or were in close proximity, thus making it difficult to disrupt the binding site for one protein without altering the binding sites for the other proteins.
hnRNP A1 Acts as a Repressor of Splicing of the Intron between Exons 6B and 7 Both in Vitro and ex Vivo-To get a clearer picture of the role of hnRNP A1 in splicing of exon 6B, we depleted hnRNP A1 from HeLa cell nuclear extract using a biotinylated oligonucleotide containing four motifs of the hnRNP A1 binding site (see "Experimental Procedures"). Western blotting experiments showed that ϳ75% of the hnRNP A1 was removed, whereas the mock-depleted extract was not affected (Fig. 6A). We did not go further in the depletion procedure because we observed a decrease in the splicing activity in the mock-depleted extract. Neither hnRNP F/H proteins nor the SR proteins were affected by the depletion procedure (Fig.  6, A and B). Mock-depleted and hnRNP A1 nuclear extracts were then tested in in vitro splicing assays with a pPmac transcript. As controls, we used a ␤-globin transcript, which extended from exon 1 to exon 2, and a 5-6A transcript, which contained exons 5 to 6A from the ␤-tropomyosin gene. In hnRNP A1-depleted nuclear extracts, splicing of the pPmac transcript was increased, as compared with mock-depleted extracts (Fig. 6C, compare lanes 4 -6 with lanes 1-3). The quantified results from several independent experiments showed that splicing of pPmac in depleted A1 nuclear extracts was activated between 2-and 3-fold. In contrast, splicing of exons 5-6A was not affected by the depletion, whereas that of the globin intron was slightly decreased (Fig. 6B, compare lanes  10 -12 with lanes 7-9 for 5-6A and lanes 16 -18 with lanes  13-15 for ␤-globin). We then added a recombinant hnRNP A1 to the depleted nuclear extracts. As shown in Fig. 6D, the addition of increasing amounts of recombinant hnRNP A1 restored the splicing repression of pPmac, whereas splicing of 5-6A and ␤-globin pre-mRNAs were not affected (Fig. 6D). These results suggested that hnRNP A1 specifically inhibited splicing of exon 6B. However, mutants pPmac mtA1 and pPmac tri-RG2 were still activated in hnRNP A1-depleted nuclear extracts, albeit to a lower extent than the wild-type transcript, which suggested that additional hnRNP A1 binding sites were present elsewhere on the transcript (data not shown). To confirm the role of hnRNP A1 in vivo, we used a tethered strategy that allowed artificial recruitment of hnRNP A1. To do this, we introduced a tandem copy of bacteriophage MS2 coat protein binding sites in place of the RG2 motif of a pcDNA 980 minigene extending from exon 5 to exon 7 to generate pcDNA 980 -2MS2. Reporter minigenes were cotransfected into QM7 with expression vectors encoding the coat protein alone or the hnRNP A1-coat fusion protein. RNAs were analyzed by RT-PCR with the primer 5Ј T7 and primers specific for exons 6A and 6B (Fig. 7A).
We also investigated splicing of exons 6B-7 using a sense primer against exon 6B and the bovine growth hormone primer poly(A) derived from the vector. As expected, transfections of QM7 myoblasts with the wild-type minigene resulted mainly in 6A splicing, as shown by the major 221-bp product (Fig. 7B,  lane 1). Two other faint products were also observed. One reflected the splicing of exon 6A to exon 6B, as indicated by a 262-bp product. The other corresponded to an mRNA in which exon 5 was spliced to exon 6B, as shown by a 186-bp product. Consistent with this result, splicing of exon 6B to 7 also gave rise to a faint product of 160 bp. Cotransfection of pcDNA 980 with hnRNP A1 coat fusion protein had no effect upon the splicing pattern (Fig. 7B, compare lanes 1 and 2). When pcDNA 980 -2MS2 was transfected with the expression vector encoding the coat protein alone, no change in the splicing pattern  2 and 6, 1 g). Spliced products were analyzed by RT-PCR using primers shown in A and separated on a 6% acrylamide denaturing gel. Quantification of the 186-bp product is indicated. ␤-Actin expression analyzed by RT-PCR is shown below each lane. hnRNP A1 and SR Proteins Regulate Splicing of ␤-TM Exon 6B was observed despite the replacement of RG2 by 2MS2. This might be explained by the fact that the replacement also eliminated binding sites for ASF/SF2, SC35. However, when pcDNA 980 -2MS2 was transfected with the hnRNP A1-coat fusion protein, splicing of exon 6B was repressed, which was reflected by the strong decrease of RT-PCR products containing exon 6B (186 and 160 bp). RT-PCR analysis performed with an exon 6B sense primer and the bovine growth hormone primer showed that splicing of exons 6B-7 was strongly affected, reflected by the disappearance of the 160-bp product (Fig. 7B,  compare lanes 3 and 5 with lanes 4 and 6). This indicates that the binding of hnRNP A1 to the intronic RG2 motif as a coat fusion protein severely compromised the removal of the intron between exons 6B and 7. In contrast, splicing of exon 6A was unaffected. Analysis of mRNA products with sense primer T7 and antisense primer against exon 7, after digestion with restriction enzymes specific for exon 6A and 6B, gave the same results, corresponding to a decrease of mRNA 5, 6B, and 7 (data not shown). These data showed that artificial recruitment of hnRNP A1 via the intronic MS2 binding sites inhibited splicing of exon 6B and together with the in vitro data demonstrated that hnRNP A1 acted as a repressor.
HnRNP A1 Counteracts the Stimulation Induced by ASF/ SF2 and SC35-The preceding data showed that two members of the SR protein family bound to RG2, which raised the possibility that these proteins were involved in splicing of exon 6B. To determine whether this was the case, we first tested the activity of total SR proteins on in vitro splicing of pPmac. 5-6A and ␤ globin pre-mRNAs were used as controls. As shown in Fig. 8A, the addition of increasing amounts of HeLa SR pro-teins stimulated the splicing of pPmac and ␤-globin transcripts by 2-fold. In contrast, splicing of 5-6A was slightly decreased. The addition of recombinant ASF/SF2 significantly increased the splicing of pPmac as compared with 5-6A and ␤-globin transcripts (Fig. 8, compare lanes 5 and 6 with lanes 13 and 14  and lanes 21 and 22). The stimulation of pPmac by ASF/SF2 was between 3-and 4-fold, whereas that of ␤-globin and 5-6A was around 1.5-fold. SC35 also increased splicing of pPmac, but to a lower extent than ASF/SF2 (Fig. 8, lanes 7 and 8). However, splicing activation of pPmac by SC35 was always higher than that observed with the ␤-globin transcript. In contrast, splicing of 5-6A was decreased at higher SC35 concentrations (Fig. 8, lanes 15 and 16). We have shown that mutations made to disrupt the hnRNP A1 binding site also affected the binding sites for ASF/SF2 and SC35 (see Fig. 5). Therefore, we tested whether these mutations have an effect on the splicing of exon 6B. As expected from affinity selection experiments, stimulation by SC35 was abolished in pPmac mtA1 and pPmac tri-RG2, which suggested that SC35 bound to RG2 was specifically involved in splicing activation of exon 6B (Fig. 8B, compare lanes 4 and 5 with lane 1 for pPmac mtA1 and lanes 9 and 10 with lane 6 for pPmac tri-RG2). In contrast, pPmac mtA1 and pPmac tri-RG2 were still activated by ASF/SF2, but about 2 times less than for the wild-type transcript (Fig. 8B, compare  lanes 2 and 3 with lane 1 and lanes 7 and 8 with lane 6). In several alternative splicing models, it has been shown that hnRNP A1 antagonized the activation mediated by SR proteins (10,(27)(28)(29). To determine whether this was the case for exon 6B splicing, we added increasing amounts of hnRNP A1 to pPmac and 5-6A transcripts that were stimulated by ASF/ SF2   FIG. 8. SR proteins activate in vitro splicing of pPmac transcript. A, labeled pPmac (lanes 1-8), 5-6A (lanes 9 -16), and ␤-globin (lanes 17-24) transcripts were incubated in 40% HeLa nuclear extracts in the absence (lanes 1, 9, and 17) or in the presence of 100, 200, and 300 ng of total HeLa cell SR proteins (lanes 2-4, 10 -12, and 18 -20) or 200 and 300 ng of recombinant ASF/SF2 (lanes 5 and 6, lanes 13 and 16, and lanes  21 and 22), or 150 and 225 ng of recombinant SC35 (lanes 7 and 8, lanes 15 and 16, and lanes 23 and 24). hnRNP A1 and SR Proteins Regulate Splicing of ␤-TM Exon 6B or SC35. As shown in Fig. 9, the addition of increasing amounts of hnRNP A1 repressed splicing of pPmac that was stimulated by both the ASF/SF2 and the SC35 proteins. In contrast, hnRNP A1 had no effect on splicing of 5-6A that was activated by ASF/SF2 and SC35. These results suggested that competition between hnRNP A1 and the SR proteins ASF/SF2 and SC35 could modulate splicing of exon 6B. To test this hypothesis, we did competition binding experiments involving UV cross-linking on labeled S3 RNA with recombinant hnRNP A1 and with ASF/SF2 and SC35 in the presence of HeLa nuclear extracts. To distinguish between the mobility of recombinant proteins, GST-⌬RS proteins were used. In the absence of recombinant proteins, a prominent band of 55 kDa (Fig. 10A) was observed that was identified as the hnRNP F/H proteins by immunoprecipitation experiments (data not shown). As expected, hnRNP A1, ASF/SF2⌬RS and SC35⌬RS alone crosslinked to the S3 probe (Fig. 10A). Interestingly, the addition of ⌬RS proteins was associated with a decrease of hnRNP F/H binding. This confirms previous experiments showing that the binding sites of these proteins were overlapping (Fig. 10A). When increasing amounts of SC35⌬RS or ASF/SF2⌬RS were added, hnRNP A1 cross-linking to the S3 probe was diminished as did hnRNP F/H (Fig. 10B, compare lanes 3-5 with lane 2 and lanes 6 -8 with lane 2). The same results were observed in the absence of nuclear extracts. The addition of SC35⌬RS or ASF/ SF2⌬RS was able to displace the amount of hnRNP A1 that cross-linked to the probe (Fig. 10C). In agreement with splicing experiments, these results indicated that ASF/SF2 and SC35 antagonized the repressing effect of hnRNP A1 by competing for overlapping binding sites. DISCUSSION RNA affinity chromatography identified hnRNP A1 as a specific component of regulatory complexes bound to S3. By using a shortened version of S3, we defined a central region, named RG2, as the minimal element for hnRNP A1 binding. In nearly all of the cases examined, RNA-binding motifs for hnRNP A1 are closely related to the A1 "winner" sequence, and they contain one or several motifs with UAG at their core (19 -25, 27, 28, 30, 31). Whereas the RG2 motif does not contain the UAG motif that appears to be the hallmark of hnRNP A1 binding site, hnRNP A1 nonetheless binds to RG2 with a high affinity. One possible explanation for this might be a particular structural arrangement of RG2 with two AGGGU motifs flanking an AGGGGA motif. Consistent with this hypothesis, RG2 exhibits an additional complex of low mobility upon binding to hnRNP A1 that is not observed with the A1 "winner" RNA ( Fig.  3). Another example of an hnRNP A1 binding site that bypasses the UAG rule is provided by the H-Ras silencer element (29). Instead of the UAG core motif, the ISS1 sequence contains three AGGG motifs that resemble the S3 G motifs. It is clear from those studies that, in addition to sequences related to the A1 "winner" sequence, hnRNP A1 has the ability to interact with a broad spectrum of RNA motifs. By using hnRNP A1depleted extracts and add-back experiments, we show that hnRNP A1 inhibits splicing of exon 6B. The splicing of unrelated pre-mRNAs, such as ␤-globin and 5-6A, was unaffected by the addition of hnRNP A1, which shows that the repression of exon 6B was specific (Fig. 6). Moreover, artificial recruitment of hnRNP A1, as a fusion with the MS2 coat protein, reduces recognition of exon 6B in myoblasts, which provides strong evidence that hnRNP A1 acts as a negative regulator of exon 6B in vivo. However, splicing is still activated in hnRNP A1depleted extracts in a pre-mRNA in which the RG2 hnRNP A1 binding site is disrupted, suggesting that either RG2 is not essential for the repression or that additional hnRNP A1 binding sites are present along the transcript. We favor the second hypothesis. Indeed, the RG2 motif is able to functionally replace the natural silencer elements in the hnRNP A1 pre-mRNA, which strongly suggests that splicing repression of exon 6B is mediated at least through binding of hnRNP A1 to RG2. 2 Furthermore, these results are not without precedence. In the case of HIV-1 tat exon 3 splicing, it was observed that the deletion of the exonic silencer element 3 high affinity hnRNP A1-binding site does not abolish further stimulation of splicing in hnRNP A1-depleted extracts (10).
The existence of several hnRNP A1 binding sites seems to be a common feature of numerous splicing events regulated by hnRNP A1 (21,24,26,27,54). In the case of the alternative splicing of hnRNP A1 pre-mRNA, it has been proposed that interactions between multiple hnRNP A1 molecules bound to introns flanking the alternative exon allow loop formation that prevents internal 5Ј splice site selection and promotes the skipping of the alternative exon (21,55). Splicing repression of exon 6B might conform to this general rule. Consistent with this view, examination of the RNA sequence revealed a UA-GAGG motif in the beginning of exon 6B that strongly interacts with hnRNP A1 (data not shown). We are currently testing whether hnRNP A1 bound to exon 6B is involved in splicing repression. The mechanism by which hnRNP A1 mediates silencing is presently unknown. In the case of splicing of hnRNP FIG. 9. Splicing activation of pPmac by ASF/SF2 and SC35 is counteracted by hnRNP A1. In vitro splicing of pPmac and 5-6A in 40% HeLa nuclear extracts supplemented with 200 ng of ASF/SF2 (lanes 2-4 and 9 -11) or 150 ng of SC35 (lanes 5-7 and 12-14) in the absence (lanes 2, 5, 9, and 12) or in the presence of 500 and 750 ng of recombinant GST-A1 (lanes 3 and 4, lanes 6 and 7, lanes 10 and 11, and  lanes 13 and 14). Lanes 1 and 8, without additional factors. The products of the splicing reaction were separated by 7% acrylamide, denaturing gel electrophoresis and quantified with a PhosphorImager. hnRNP A1 and SR Proteins Regulate Splicing of ␤-TM Exon 6B A1 pre-mRNA, the shift toward distal 5Ј selection does not seem to be associated with modifications in the U1 snRNP occupancy at the competing 5Ј splice sites (20). However, in our system, previous studies have shown that the repression of the pPmac transcript is associated with a decrease of A complex formation and a reduction of U1 snRNP interaction on exon 6B 5Ј splice site (56). Interestingly, this repression is dependent on an RNA secondary structure around exon 6B (36,57). Thus, it may be possible that the folding of RNA is stabilized by hnRNP A1 interactions, which in turn interferes with U1 snRNP binding. In addition to its role in 5Ј splice site selection, hnRNP A1 also is involved in the regulation of 3Ј splice site usage (10,22,24,27,30,31,58). In the well characterized 3Ј splice site regulation of HIV tat exon 3, it has been proposed that a repressive wave is initiated at a high affinity hnRNP A1 binding site that propagates along the exon to inhibit splicing of the upstream intron (10). In that model, it was shown that binding of ASF/SF2 to an ESE prevents the propagation of hnRNP A1 along the exon and blocks splicing silencing (10). During the course of this study, we identified ASF/SF2 and SC35 as components of the protein complex that is bound to the RG2 motif. Each of the SR proteins activates splicing of exon 6B. This effect is mediated at the least through their interactions with the RG2 motif, because mutations mtA1 and tri-RG2 strongly decrease splicing activation (Fig. 6). However, additional SR binding sites cannot be totally excluded. Two putative ASF/SF2 binding sites (TAGAGG and AGGAGGA), which overlap with hnRNP A1 binding site on exon 6B, can be identified by ES-Efinder (52). Mutations of these motifs decrease in vitro splic-ing of the pPmac construct (data not shown). Further experiments are required to elucidate the role of these motifs in splicing of exon 6B. We also reported, as it was observed in other systems, that splicing activation by SR proteins was counterbalanced by hnRNP A1. In addition, by using UV-crosslinking experiments, we demonstrated that SR proteins were able to displace hnRNP A1 bound to the S3 element. Interestingly, ASF/SF2 and SC35 that lack the RS domain are still able to block binding of hnRNP A1 to the S3 element, which suggests that binding sites for all of these proteins are closely linked (Fig. 10). Consistent with this hypothesis, mutational analyses were unsuccessful in separating the binding sites for ASF/SF2, SC35, and hnRNP A1. Overlapping binding sites for the SR proteins and hnRNP A1 seem to be a common feature encountered in several alternative splicing models (27,28,58). In the case of HIV, 3Ј splice site selection of A7 involved the AUA(GAA) 3 Janus regulatory element, which is made up of overlapping binding sites for ASF/SF2 and hnRNP A1. The authors proposed that the binding of ASF/SF2 to this element disrupted hnRNP A1 protein-protein interactions, and this leads to the decompaction of the RNA structure (27). Our system presents some resemblance to HIV; it contains an RNA secondary structure and several hnRNP A1 binding sites that are overlapping with ASF/SF2 and SC35. Thus, it is tempting to speculate that the SR proteins might function similarly to destabilize the RNA structure.
Western blots of affinity-selected proteins have identified hnRNP F/H as a component of the complexes assembled onto the S3 element (Fig. 5B). The S3 element contains several hnRNP A1 and SR Proteins Regulate Splicing of ␤-TM Exon 6B G-triplets, and, in particular, it contains in the middle of the RG2 region a GGGA motif that has been identified as a core sequence needed for binding to the hnRNP H family (53). The mutation of tri-RG2 that interrupts the G-triplet does not impair the ability of hnRNP F/H to bind to the S3 element (Fig. 5). However, disruption of the consecutive G residues does abolish the hnRNP F/H interactions, which suggests that the RG2 motif is a true binding site for the hnRNP F/H proteins (data not shown). Additional hnRNP F/H binding sites also have been identified in the RG1 motif (data not shown). Inside this motif is the sequence GGGGCUG that closely resembles the hnRNP F/H binding site present in the downstream control element of c-src (15). Cross-linking experiments show that, in addition to hnRNP A1, ASF/SF2 and SC35 also compete with hnRNP F/H. This finding suggests that these proteins might play an important role in the regulation of exon 6B (Fig. 10). In several systems, these proteins were shown to promote or to repress splicing (14, 16, 17, 59 -61). It has been shown that hnRNP H activates the inclusion of the tev-specific exon 6D of HIV-1 (59). Interestingly, a SC35 binding site in close proximity was proposed to help in the recruitment of hnRNP H. Another example where binding sites for hnRNP F/H and SR proteins are in close proximity was provided by the exonic enhancer of the N1 exon of c-src (28). It is interesting to note that hnRNP A1 has been also found associated with the exonic enhancer. Thus, it appears that the peculiar arrangement of overlapping binding sites having positive or negative effects depends on which proteins bind to them, and this plays an important role in the regulation of numerous splicing events. This creates an exquisite system of regulation of splicing events that can respond to small variations in the relative concentration of regulatory factors. Further experiments will be needed to elucidate the role of the hnRNP F/H protein family and to understand how all these regulatory proteins communicate with each other to regulate splicing of exon 6B.