Heterogeneous Nuclear Ribonucleoprotein (hnRNP) K Is a Component of an Intronic Splicing Enhancer Complex That Activates the Splicing of the Alternative Exon 6A from Chicken β-Tropomyosin Pre-mRNA*

Splicing of the chicken β-tropomyosin exon 6A is stimulated, both in vivo and in vitro, by an intronic pyrimidine-rich element (S4) located 37 nucleotides downstream of exon 6A. Several pyrimidine-rich sequences are able to substitute for the natural S4 enhancer with various stimulatory effects. We show that the different enhancer sequences recruit U1 small nuclear ribonucleoprotein (SnRNP) to the exon 6A 5′ splice site, with an efficiency that correlates with the splicing activation. By using RNA affinity and two-dimensional gel electrophoresis, we characterized several proteins that bind to the different enhancer sequences. Heterogeneous nuclear ribonucleoprotein (hnRNP) K and hnRNP I (polypyrimidine track-binding protein, PTB) exhibit a higher level of interaction with the strong enhancer sequences (S4) than with the weakest enhancers. Functional analysis shows that hnRNP K is a component of the enhancer complex that promotes exon 6A splicing through the wild-type S4 sequence. The addition of recombinant hnRNP K to nuclear extracts preincubated with poly(rC) RNA competitor completely restores splicing efficiency to the original level. hnRNP I (PTB) was also found associated with the strong enhancer sequences. Its function in the splicing of exon 6A is discussed.

Alternative splicing of pre-mRNAs generates different mRNAs from the same primary transcript and as such contributes to protein diversity (1)(2)(3). This process is thought to be important for regulation of the gene expression and has been enlightened by the discovery that human genome contains only 30,000 -40,000 genes. From these studies, it has been suggested that more than 50% of the genes are alternatively spliced (4). In many cases, alternative splicing is regulated in a tissue developmental stage or sex-specific manner and responds differently to metabolic stimuli (2,3). Studies of different splicing events show that the bona fide regulation is orchestrated by multicomponent complexes that promote or repress the use of alternative splice sites through binding to regulatory sequences (1,5,6). The SR protein family is a well characterized class of proteins that are involved in both consti-tutive and alternative splicing (7,8). It has been shown that the binding of members of the SR proteins to purine enhancer sequences promotes the inclusion of the corresponding exon by facilitating the recognition of weak 3Ј splice sites (9). Another group of proteins that affect splice site selection are the hnRNPs. 1 They constitute a large group of RNA-binding proteins that associate with nascent transcripts, and they can hinder or assist the splicing machinery (10,11). Early experiments have shown that hnRNP A1 is able to antagonize the activity of ASF/SF2 by promoting a shift toward the selection of distal 5Ј splice sites (12,13). More recently, it has been found that hnRNP A1 interacts with silencer elements to repress splicing of several pre-mRNAs (14 -18). In the case of human immunodeficiency virus type 1 tat splicing, it has been proposed that the binding of hnRNP A1 to intron silencer sequences blocks the U2 snRNP interaction (19). hnRNP I or PTB is also implicated in the negative control of alternative splicing of several pre-mRNAs, including ␣ and ␤ tropomyosin from rat, ␥-GABA A receptor ␥2, fibroblast growth factor receptors, c-src, and caspase-2 (6, 20 -29). In some cases, it has been suggested that PTB mediates repression by interfering with the binding of U2AF65 on the polypyrimidine tract (22). However, in other models in which PTB is implicated, it has been shown that multiple binding sites for PTB are required for the repression, suggesting a more complex role for PTB (28,30). Several other hnRNP have been implicated in the tissue-specific regulation of pre-mRNAs. KSRP, an hnRNP that contains KH motifs, has been identified in an intronic enhancer complex (downstream control sequence, DCS) that is required for inclusion of the N1 exon of c-src pre-mRNA (31). In contrast, PSI is another K-H protein that prevents the splicing of the Drosophila P element third intron, through binding to an exonic silencer element (32). hnRNP H has been also implicated in the regulation of splicing. Depending on which pre-mRNA it is bound, hnRNP H behaves either as an activator or a repressor of splicing (33,34). From studies of several alternative splicing models, it is clear that splice site selection depends on the intricate combined control of both enhancer and silencer elements. In addition, several of the regulatory sequences are made of individual motifs that have opposite effects on splice site selection, making it difficult to precisely dissect the role of the motifs and the proteins bound to them. This is well exemplified in the case of the N1 exon from c-src pre-mRNA in which PTB binding motifs that participate in the repression of the N1 exon in nonneuro-nal cells have been found interspersed within the DCS sequence (35).
The chicken ␤-tropomyosin pre-mRNA provides an example of the complexity of alternative splicing regulation. This transcript contains two mutually exclusive exons that are selected in a tissue-specific fashion (36). Exon 6A is recognized in nonmuscle cells and myoblasts, whereas exon 6B is recognized in skeletal muscle and myotubes. This regulation is maintained in vitro with HeLa cell nuclear extracts, where exon 6A is included and exon 6B is excluded (37). The recognition of exon 6A requires an intronic enhancer element (S4) lying 37 nt downstream of the exon 6A 5Ј splice site (38,39). This enhancer sequence also exerts a negative control on exon 6B recognition by a mechanism that is presently unknown (38). We have previously shown that a sequence derived from the pyrimidine tract upstream of exon 3 of the rat ␣-tropomyosine (P3S) is able to substitute for S4 and to activate exon 6A recognition (40). We have shown that splicing enhancement by S4 and P3S is mediated through their interaction with ASF/SF2 and that SC35 has an antagonistic effect (40). Another group has identified other pyrimidine-rich sequences that activate exon 6A splicing (41). One is located in the intron downstream of exon 5 (PyI5), and the other (S5) lies immediately downstream of S4. The two sequences are able to stimulate the splicing of exon 6A in myoblasts but at a level that is lower than with the P3S and the natural S4 sequences. Based on these results, we reasoned that the activation of exon 6A splicing by the different pyrimidinerich sequences was mediated either through the interaction of a common set of proteins, whose level of interaction determines the efficiency of stimulation, or by different proteins that operate through the same activation mechanism. Regardless of the mechanism, the analysis of the proteins assembled onto the different enhancer sequences might provide an approach to identify the activating proteins that bind to the natural S4 enhancer.
In this paper, we purified the protein complexes that bind to the different enhancer sequences by RNA affinity chromatography. The protein pattern analyzed by two-dimensional gel electrophoresis shows that a common set of hnRNPs interacts with the different enhancers, which correlates with splicing activation. hnRNP K and hnRNP I exhibit stronger interaction with the strong enhancers S4 and P3S than with the weaker PyI5 and S5. We show that hnRNP K is involved in the recognition of exon 6A through its interaction with the S4 sequence. Inhibition of exon 6A splicing by competition with poly(rC) is fully relieved by the addition of recombinant hnRNP K. In addition, we show that the activating sequences act through a common mechanism to recruit U1 snRNP to the exon 6A 5Ј splice site.

EXPERIMENTAL PROCEDURES
RNA Affinity Chromatography-Streptavidin-agarose beads were preblocked by adding acetylated bovine serum albumin (500 g/ml) and Escherichia coli tRNA (100 g/ml) and rotating 15 min at 4°C in WB50 buffer (50 mM Tris-HCl, pH 7.5, 50 mM NaCl, 0.01% Nonidet P-40). Beads were washed once with WB 100 (as WB 50, but with 100 mM NaCl). Biotinylated RNA was added in the same buffer and gently rotated for 45 min at 4°C. The beads were collected and washed three times in buffer D (12 mM Hepes, pH 8, 60 mM KCl, 1 mM dithiothreitol, and 12% glycerol), snap frozen in liquid nitrogen, and stored at Ϫ20°C or used immediately. The efficiency of fixation was routinely above 80% of the 32 P-labeled RNA input and was carefully controlled for a strict comparison of the different RNA samples.
For analytical gels, 125 or 250 pmol of RNA was fixed to 50 l of beads and mixed with 1 ml of 30% HeLa nuclear extracts under splicing conditions without polyvinylalcohol by rotating for 5 min at 30°C. Beads and splicing mixture were separately prewarmed at 30°C for 5 min before mixing. Beads were collected by centrifugation 1 min at 3000 rpm and washed two times with ice-cold WB 100. About 75% of the RNA was recovered. The equivalent of 60 pmol of RNA was sufficient for analysis by silver staining on a two-dimensional polyacrylamide gel. For preparative interactions, RNAs were at 4 -4.4 pmol/l of beads for a total of 500 l of beads and incubated in 2 ml of 30% HeLa cell nuclear extracts under splicing conditions. Preparation of Proteins from the RNA Bead Complexes-The 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. 300 l were used for 60 l of beads. After centrifugation for 2 min at 3000 rpm, the supernatants were precipitated with 5 volumes of acetone for 16 h at Ϫ20°C, using 40 g of glycogen as carrier. The proteins were centrifuged in a swinging bucket rotor at 4000 rpm for 5 min. The pellets were washed three times with acetone/H 2 O (5:1) and stored dried at Ϫ20°C or dissolved in 9 M urea, 2% Nonidet P-40, 0.1 M dithiotreitol, and 2% ampholyne, pH 3.5-9.5, and stored at Ϫ20°C.
Two-dimensional Gel Electrophoresis-The first dimension was either an isoelectofocusing (IEF) or a nonequilibrium pH gradient gel electrophoresis (NEPHGE) as described (42) using ampholyne, pH 3.5-9.5 (Amersham Biosciences) or pH 3-10 (Amersham Biosciences). The second dimension was usually a 6.5% polyacrylamide gel electrophoresis as described (43). The proteins were visualized by silver staining or by Coomassie Brilliant Blue G-250. For a strict comparison of the protein interactions, the gels were processed together in the same batch, and overstaining was avoided.
Mass Spectrometry Analysis-Proteins analyzed by mass spectrometry were prepared from a scaled-up reaction in which 1000 pmol of RNA bound to streptavidin-agarose were added to 8 ml of prewarmed solution containing 2.4 ml of nuclear extracts under splicing conditions and then treated as indicated above. The proteins were separated on two-dimensional gel electrophoresis and visualized by Coomassie Brilliant Blue. The protein spots were excised from the gel and digested within the gel with trypsin. Mass spectra were recorded in positive reflector mode with a matrix-assisted laser desorpton-ionization timeof-flight mass spectrometer (Voyager Elite, Perseptive Biosystems, Inc., Framingham, MA) equipped with a delayed extraction device. The proteins were identified by data base searching using Peptide Search and MS Fit.
RNA Synthesis-Capped pre-mRNAs were synthesized in vitro as described previously using [␣-32 P]UTP and SP6 RNA polymerase (45). Transcripts were purified on polyacrylamide/urea gels. RNAs used for affinity chromatography and S4b and S5b competitor RNAs were transcribed using a SP6 MEGAshortscript Kit (Ambion), without the cap analogue, and labeled at 10 -20 cpm/pmol with [␣-32 P]UTP to facilitate quantification. RNAs were purified on a microbiospin column (Bio-Rad) and 3Ј-biotinylated as described (40). Poly(rC) competitor was purchased from Amersham Biosciences. To compare poly(rC) with other competitor RNAs, 20 ng of poly(rC) was taken as the equivalent of 1 pmol of a 65-nucleotide-long RNA.
Splicing Reactions-Splicing reactions were performed as described with 25 pmol of labeled pre-mRNA in 12.5-l reactions with 40% HeLa nuclear extract (39). Competitor RNAs and recombinant hnRNP K were added to the splicing reaction on ice prior to adding the labeled pre mRNA. Reactions were analyzed by 7% (19:1) polyacrylamide gel electrophoresis. RNA products were quantified using a PhosphoImager (Molecular Dynamics, Inc., Sunnyvale, CA). Splicing efficiency was calculated as the ratio between the mRNA value or final lariat and the sum of pre-mRNA, mRNA, and lariat values. For each RNA, the number of uracil residues was taken into account for the calculated splicing efficiency.
Preparation of Recombinant hnRNP K-E. coli BL21(DE3) was transformed by the plasmid pET 28aK (kindly provided by G. Dreyfuss). Transformants were grown at 37°C in LB medium with 25 g/ml kanamycin at 0.7 at 600 nm and induced by 1 mM isopropyl-1-thio-␤-D-galactopyranoside for 2.5 h. The pellet was resuspended in 10 ml of buffer 1 (50 mM sodium phosphate, pH 7.5, 1 mM dithiothreitol) containing 50 mM NaCl and a mixture of protease inhibitors and then broken with a French press. The supernatant was adjusted to 0.5 M NaCl and loaded at 0.2 ml/min on a 1-ml poly(rC) column containing 100 g of poly(rC). The column was washed with 8 ml of buffer 1 plus 2 M NaCl. A first step elution was in buffer 1 with 8 M urea and a second step in buffer 1 with 4 M guanidine HCl. The eluted proteins were dialyzed against 10 mM Hepes-HCl, pH 7.4, 100 mM KCl, 1 mM dithiothreitol, and 20% glycerol. The protein from the guanidine HCl elution step was used in the complementation experiments. hnRNP K was more than 90% pure as judged from a control SDS gel and Western blot.
Interaction of Poly(rC) with HeLa Nuclear Extract-Random length poly(rC) was 5Ј-labeled with [␥-32 P]ATP, 3Ј-biotinylated, and fixed on streptavidin-agarose beads (as described above). The yield was generally from 2.0 to 2.5 g for 50 l of beads. Taking 13,200 g/mol for the molecular weight of an RNA of 65 nucleotides in length, 250 pmol of poly(rC) equivalent was mixed with 1 ml of 20 or 40% nuclear extract. The proteins were prepared and analyzed on SDS and two-dimensional acrylamide gel as above.
UV Cross-linking Experiments-Nuclear extracts were either mocktreated or treated with an oligodeoxynucleotide complementary to the 5Ј-end of U1 snRNA and then treated with RNase H as described by Black et al. (46). For binding experiments, 20 or 50 fmol of 32 P-labeled 5Ј-RNAs were incubated in 12.5 l (at 40% nuclear extract under splicing conditions with 4% polyvinyl alcohol) for 10 min at 30°C and then irradiated on ice for 15 min in a UV (254 nm) Stratalinker. The samples were treated with 50 l of proteinase K buffer (100 mM Tris-HCl, pH 7.5, 10 mM EDTA, 2% lauryl sarkosyl) and 20 g of proteinase K for 45 min at 37°C, phenol-extracted, and analyzed by electrophoresis on a 6% (19:1) polyacrylamide gel. The level of the cross-linked species was quantified in a PhosphorImager.

RESULTS
Previous studies identified an S4 enhancer that is required for the inclusion of exon 6A in myoblasts and HeLa nuclear extracts (38,40). It has been shown that several pyrimidinerich sequences can substitute for the S4 enhancer element with different efficiencies (41). These sequences are a pyrimidinerich sequence derived from the pyrimidine tract upstream of exon 3 of the rat ␣-tropomyosine (P3S), the pyrimidine tract downstream of exon 5 of the chicken ␤-tropomyosine (pyI5), and the sequence S5, which lies just downstream of S4 (Fig.  1A). To be able to compare the proteins assembled onto the different enhancers, we first tested in vitro splicing activity of pre-mRNA 6A-7 containing the wild-type S4 enhancer sequence and the other pyrimidine-rich sequences P3S, PyI5, and S5 using the same batch of HeLa cell nuclear extract. As shown in Fig. 1B, P3S was as efficient as S4 in promoting splicing of exon 6A. Depending on nuclear extracts, splicing activation by P3S was even higher than with S4 (40). As expected the S5 and PYI5 sequences were less potent activators than the S4 sequence. Splicing efficiencies decreased 2-fold for pre-mRNA 6A-S5-7 and around 1.5-fold for the pre-mRNA 6A-PyI5-7 as compared with that of the wild-type pre-mRNA (Fig. 1B). We conclude that the hierarchy of activation promoted by the various pyrimidine sequences was maintained both in vivo and in vitro, making it possible to identify the proteins that mediate the activation of exon 6A.
The Level of Interaction between snRNP U1 and Exon 6A 5Ј Splice Site Depends on the Type of Intronic Sequence-Splicing decisions are thought to be determined early in spliceosome assembly, at the step of commitment complex E formation that is characterized by U1 snRNP binding to the 5Ј splice site (47). To test whether the enhancer sequences S4, P3S, S5, and PYI5 promote exon 6A splicing at an early stage of spliceosome assembly, we used UV cross-linking experiments. 32 P-Labeled substrate RNAs shown in Fig. 2 were incubated under splicing conditions for 10 min at 30°C, irradiated at 254 nm, and treated extensively with proteinase K to digest any crosslinked proteins. As a control for cross-linking efficiency, we used a short RNA of 33 nucleotides that was complementary to the 5Ј terminus of snRNA U1 (data not shown). Upon incubation of the labeled RNAs with HeLa cell nuclear extracts, two specific cross-linked species were observed (Fig. 3). The intensity of these bands strongly decreased in HeLa nuclear extracts in which U1 snRNA had been degraded by an oligonucleotidedirected RNase H cleavage (Fig. 3, lanes ⌬). Note that a third cross-linked band was also detected with the 5Ј S4 RNA that did not disappear in U1-snRNA-degraded nuclear extracts. This is most likely an intramolecular cross-link, since it was detected after UV irradiation of the RNA in splicing buffer alone (data not shown). RNase H-directed cleavage of the crosslinked RNA after UV irradiation with an oligonucleotide complementary to a region of U1 snRNA immediately downstream of the 5Ј arm led to the disappearance of the two bands (data not shown). These results indicate that the cross-linked bands correspond to a functional interaction between the U1 snRNA and the 5Ј-splice site. Interestingly, the level of U1 snRNP interaction was strongly dependent on the downstream intronic sequence. Results from three independent experiments FIG. 1. Efficiency of the S4, S5, PyI5, and P3S enhancer to promote in vitro splicing of exon 6A to exon 7 in HeLa nuclear extracts. A, schematic representation of the pre-mRNA 6A-7 and of the enhancer sequences. The sequences of the S4 enhancer and the pyrimidine sequences P3S, S5, and pyI5 are shown below. Note that 6A-S5-7 contains two successive repeats of the S5 sequence. B, in vitro splicing of 6A-7 pre-mRNAs containing the various enhancer sequences. 32 P-Labeled pre-mRNAs were spliced in 40% HeLa nuclear extract at 30°C for the times indicated. The products of the splicing reaction were separated by 7% acrylamide denaturing gel electrophoresis and quantified with a PhosphorImager (Molecular Dynamics). A schematic representation of unspliced and spliced RNAs is shown on the right.
hnRNP K, PTB, and hnRNP L Preferentially Interact with the Strong Intronic Enhancer Sequences-The next step was to determine which proteins interact with the different enhancer sequences. We used affinity chromatography with RNA as bait (Fig. 2). In a first set of experiments, we included the 5Ј splice site of exon 6A with the different enhancer sequences. The rationale was that the presence of the 5Ј splice site might stabilize the enhancer complexes. RNA corresponding to the intronic enhancer sequences P3S, S4, S5, and PYI5 were 3Јbiotinylated and coupled to streptavidin-agarose beads (see "Experimental Procedures"). The RNA concentration was chosen to be far less than that required for saturation in the in vitro splicing assay. Under these conditions, it was expected that the interactions of the proteins that promoted the splicing of exon 6A might respond proportionally to their binding affinity for the different RNAs. The RNAs were then incubated with 40% HeLa cell nuclear extracts for 5 min at 30°C under splicing conditions. After washing the beads, the interacting proteins were extracted and analyzed by two-dimensional gel electrophoresis with either NEPHGE (Fig. 4A) or IEF (Fig. 4B) in the first dimension. As shown in Fig. 4, a large number of proteins are common to all four RNAs, and they differ only by minor variations in the staining intensity. This is particularly obvious in Fig. 4A, in which the affinity-purified proteins were analyzed on NEPHGE in the first dimension. Some protein spots were common to both types of gels (e.g. spots of protein L, pI Ͻ7 for NEPHGE and Ͼ7 for IEF). Three protein spots systematically showed larger differences between the four activating RNAs. The identification of the three proteins was done by mass spectrometry. hnRNP K, hnRNP L, and hnRNP I (PTB) were identified with no ambiguity and confirmed by Western blots for hnRNP K and PTB (data not shown). hnRNP K and hnRNP I migrated as several spots, which might correspond to different isoforms and/or different levels of phosphorylation as previously reported (48). Experiments were repeated with different batches and different concentrations of nuclear extracts (20 and 60%) with no change in the profile and level of interacting proteins. The hierarchy of interaction was P3S Ͼ S4 Ͼ PyI5 Ͼ S5 for hnRNP K and P3S Ͼ S4 Ͼ S5 Ͼ PyI5 for PTB. The differences were less obvious for hnRNP L with P3S Ϸ S4 Ϸ PyI5 Ͼ S5. These data suggest that these proteins could be involved in the activation of 6A. Note that additional proteins were specifically associated with the P3S enhancer sequences (Fig. 4B, white stars). They were absent on S4, S5, and PYI5 sequences, suggesting that these proteins might be functionally associated with the more potent P3S enhancer to promote splicing of exon 6A. In addition, proteins of high molecular weight were common to P3S and S4 (Fig. 4b, white  arrows).
We next tested whether binding of hnRNP K, hnRNP I, and hnRNP L was dependent on the presence of the 5Ј splice site of exon 6A. RNA chromatography was performed with enhancer sequences without the 5Ј splice site (Fig. 5). Only the proteins interacting with the S4b and S5b sequences with IEF in the first dimension are presented. As shown in Fig. 5, hnRNP K and L interact more strongly with S4 than with S5. The interaction of PTB was also much higher with the S4 sequence than with S5 (data not shown). From these experiments, we con-FIG. 2. Sequences of the different RNAs used for RNA-protein interactions. RNAs 5Ј-S4, 5Ј-S5, 5Ј-pyI5, and 5Ј-P3S have in common seven nucleotides from exon 6A, the 5Ј splice site of exon 6A, and 15 nucleotides derived from the pSP65 polylinker (shown in boldface type). Restriction sites at the 3Ј-end of the RNAs are indicated in italic type. Sequence S6, which has no functional effect on 6A splicing, was deleted (38,40). Sequences S4b and S5b have in common 15 nucleotides derived from the pSP65 polylinker. Note that S5b contains additional intronic nucleotides (not present in 5Ј-S5) to have the same length as S4b.
FIG. 3. The different enhancers promote U1 snRNP binding to exon 6A 5 splice site with various efficiencies. HeLa nuclear extracts were either mock-preincubated (M) or preincubated with an oligonucleotide complementary to the 5Ј-end of U1 snRNA and RNase H degraded (⌬) before cross-linking to the different 5Ј-RNAs shown in Fig.  2. After treatment with proteinase K, the RNAs were purified and resolved by 6% polyacrylamide gel electrophoresis. In lanes 1-8, 5Ј splice site RNAs containing the different enhancer sequences. Crosslinks of 5Ј single-stranded RNA with U1 snRNA are indicated by arrows (left). Note that 5Ј-RNA S4 has an additional characteristic product, which corresponds to an auto-cross-link (lanes 1 and 2). The crosslinking efficiency for each RNA was quantified with a PhosphorImager.
clude that the interactions of hnRNP K and hnRNP I are independent of the 5Ј splice site of exon 6A.
In Vitro Splicing Inhibition of 6A-S4-7 Induced by Poly(rC) Can Be Relieved by hnRNP K-The previous results raised the question of which of these proteins was involved in activating exon 6A splicing. In several documented models of alternative splicing, PTB was found to act as a negative regulator of splicing (for a review, see Ref. 28). In previous studies, it has been shown that the S4 sequence is also involved in the repression of exon 6B (38). One possibility is that the binding of PTB on S4 is related to its negative role in splicing regulation (see "Discussion"). We therefore decided to focus on hnRNP K as a possible activator. hnRNP K is considered to be the major poly(C)-binding protein (49). We first analyzed the proteins of the nuclear extract that interact with poly(rC). In this experiment, poly(rC) was 3Ј-biotinylated, fixed to streptavidin-agarose beads, and processed under the same conditions as those used for the other RNAs (see "Experimental Procedures"). As shown in Fig. 6, only two proteins were affinity-purified. hnRNP K that migrates as several spots was formally identified by Western blot analysis using specific monoclonal antibodies (Fig. 6B). One other basic protein also was detected as two spots (Fig. 6C). According to the molecular weight, isoelectric point, and affinity to poly(rC), it might correspond to protein PCB1 (50). The affinity of hnRNP K to poly(rC) was exploited in in vitro splicing competition assays. As controls, competitor RNAs S4b and S5b were also tested. To be able to compare the effects of poly(rC) (which is a random length polymer) with S4b and S5b, we expressed the amount of poly(rC) in pmol as being equivalent to a polymer of 65 nucleotides. This comparison is certainly an approximation, because it supposes that the different sequences are uniformly covered with proteins, which is probably not the case for S4b and S5b. However, this permitted a qualitative comparison. As shown in Fig. 7, the addition of poly(rC) to the splicing reaction moderately inhibits splicing of the pre-mRNA 6A-S4-7. The inhibition curve reaches a plateau around 2 pmol. Splicing of pSma and adenovirus E1A pre-mRNAs, which do not contain S4, were  unaffected by the addition of poly(rC) (Fig. 8). Furthermore, as shown in Fig. 7, the inhibition by poly(rC) was different from the two other competitor RNAs S4b and S5b. This first suggests that the splicing inhibition induced by poly(rC) was actually specific and, second, suggests that the trapping of hnRNP K by poly(rC) affected only a fraction of the enhancing effect mediated by the S4 sequence. In agreement with these results, competition experiments with the S4b RNA at the same concentration as poly(rC) dramatically inhibited splicing of pre-mRNA 6A-S4-7. The addition of 5 pmol of S4 RNA competitor resulted in a 50% decrease in splicing efficiency. Complete inhibition was obtained with 10 pmol of RNA competitor. In comparison, RNA S5b, which was a weak activator, reduced splicing efficiency by 25% (Fig. 7B, compare the curves with S4b and S5b).
To determine whether hnRNP K was involved in promoting the splicing of exon 6A, hnRNP K was expressed in E. coli and added back in an in vitro splicing competition assay. As shown in Fig. 9, the addition of increasing amounts of recombinant hnRNP K completely relieves the splicing inhibition induced by poly(rC). From this result, we conclude that hnRNP K is one of the components of the enhancer complex that mediates activation of exon 6A splicing.
Of the Three Other Pyrimidine Sequences, Only Pre-mRNA 6A-7 Containing pyI5 Is Stimulated by hnRNP K after Inhibition by Poly(rC)-In vitro splicing competition assays were performed to determine whether splicing activation of exon 6A through the enhancer sequences P3S, pyI5, and S5 was also mediated by hnRNP K. To our surprise, in vitro splicing of 6A-P3S-7 was almost unaffected by the addition of poly(rC), even at higher concentrations than that required to inhibit splicing of pre-mRNA 6A-S4-7. Depending on the nuclear extract batches, the inhibition of in vitro splicing did not exceed 10% (data not shown). These results were unexpected,since the P3S enhancer sequence interacted strongly with hnRNP K. hnRNP K has been shown to form heteodimeric association with other hnRNPs (51). Thus, one possibility might be that a fraction of hnRNP K is in a complex that is not accessible to competition by poly(rC). Another possibility might be that other proteins might play a major role in the enhancing effect of P3S. In contrast, the splicing of 6A-pyI5-7 pre-mRNA was strongly affected by the addition of poly(rC) RNA competitor. The addition of 20 ng (1 pmol) of poly(rC) led to an 80% reduction of splicing efficiency (Fig. 10). The inhibition curve did not exhibit a plateau, suggesting that, in contrast to S4, the activation of exon 6A by pyI5 might involve a limited number of proteins (compare Figs. 7 and 10). In agreement with this, the addition of recombinant hnRNP K completely relieved the splicing inhibition promoted by poly(rC) (Fig. 10B). DISCUSSION Spliceosome assembly at alternative splice sites involves the recruitment of protein complexes that bind to various regulatory sequences in order to ensure efficient recognition of the FIG. 6. Poly(rC) interacts essentially with hnRNP K under splicing conditions. 3Ј-Biotinylated poly(rC) bound to streptavidinagarose beads were incubated with 40% nuclear extract under splicing conditions for 10 min at 30°C. The interacting proteins were eluted and analyzed on one-dimensional SDS gel electrophoresis (A) and twodimensional gel electrophoresis (B and C). B, IEF in the first dimension shows isoforms of hnRNP K. C, NEPHGE in the first dimension shows two spots that are probably two forms of PCB1 protein. Gels were stained with Coomassie Blue G-250. FIG. 7. Inhibition of pre mRNA 6A-S4 -7 in vitro splicing by competition with poly(rC) and RNA S4b and S5b. A, labeled 6A-S4-7 pre-mRNA was spliced in 40% nuclear extract for 75 min without or with increasing amounts of RNA competitors. HeLa nuclear extract was preincubated for 10 min at 0°C before the addition of the labeled pre-mRNA in the absence (lanes 1 and 2, duplicate) or in the presence of 20, 30, 50, and 70 ng of poly(rC) RNA competitor (lanes 3-6); 1, 2, 5, and 7 pmol of S4b (lanes 7-10); or 1, 2, 5, and 7 pmol of S5b (lanes [11][12][13][14]. Lane 0, unspliced pre-mRNA. The products of the reaction were analyzed by 7% acrylamide gel electrophoresis. B, the percentage of splicing is shown as a function of increasing amount of RNA competitors. To compare on the same graph the inhibitory effect of poly(rC) with the S4b and S5b RNAs, we express the amount of poly(rC) in pmol equivalent to a 65-nucleotide length, taking 20 ng for 1 pmol. Black squares, poly(rC); white squares, S4b; circles, S5b. splicing signals. The challenge is to identify the proteins that mediate the regulation through their interaction with the regulatory sequences. We took advantage of the fact that diverse pyrimidine-rich sequences can replace the natural S4 enhancer downstream of exon 6A to identify the proteins that activate splicing of exon 6A in nonmuscle cells.
We first show that diverse pyrimidine sequences are able to promote U1 snRNP binding to the exon 6A 5Ј splice site with efficiencies that correlate with splicing activation. U1 binds more strongly with the strong S4 and P3S enhancers than with the weaker PyI5 and S5 activators. Intronic sequences that facilitate 5Ј splice site recognition have been already reported (33,52,53). In a recent study, it has been shown that TIA-1 binding to the U-rich enhancer, named IAS1, activates the 5Ј splice site of the fibroblast growth factor receptor 2 (FGR-2) alternative K-SAM exon (54). Interestingly, the authors show that IAS1 can replace S4 in the pre-mRNA 6A-7, only if it is positioned immediately downstream of the exon 6A 5Ј splice site, which is analogous to its natural position in the FGR-2 pre-mRNA. In contrast, positioning IAS1 at the location of S4 did not stimulate exon 6A splicing (54). A similar observation was made for the ␤-tropomyosin pre-mRNA. Interestingly, moving the S4 sequence 100 nt downstream of the 5Ј splice site also abolishes the activation of exon 6A splicing (38). All of these data indicate that different proteins can promote U1 binding through their binding to different enhancers, provided that the regulatory sequences are positioned correctly downstream of the 5Ј splice sites.  5-9, 300, 400, 500, 600 and 700 ng, respectively). Lane 0, unspliced pre-mRNA. B, percentage of splicing is shown as a function of the amount of recombinant hnRNP K added, in the presence of 10 ng of poly(rC). Note that splicing inhibition by poly(rC) is stronger than that shown in Fig. 7 due to the use of another batch of nuclear extract. RNA affinity chromatography allowed us to identify several hnRNPs that bind to the diverse S4, P3S, S5, and PyI5 sequences. Three proteins emerge from this study. They were formally identified by mass spectrometry and Western blot analyses as hnRNP K, hnRNP I, or PTB and hnRNP L. It has been shown that hnRNP K is involved in a broad range of regulatory functions, including transcription regulation, transport, signal transduction, and mRNA stability (55). Because of its interaction with some of the SR proteins, hnRNP K has been suspected to play a role in RNA processing (56). hnRNP K is the prototype of the K-H motif-containing hnRNP family, members of which are involved in the regulation of splicing (31,32,57).
Here, we show that the founder of this hnRNP family, hnRNP K, is involved in the activation of exon 6A. Two-dimensional gel electrophoresis shows that hnRNP K interacts strongly with enhancers S4 and P3S, which promote the highest level of splicing activation, followed by PyI5 and then S5, which interacts poorly. Recently, SELEX experiments for RNA bound to hnRNP K have identified an AUC 3/4 (U/A)(A/U) consensus motif (58). All of the diverse enhancers have stretches of poly(rC). However, based on the consensus sequence, no clear correlation can be established between these motifs and the level of hnRNP K binding. It is likely that hnRNP K binding depends not only on the intrinsic affinity of the protein for each site but also on its interactions with other proteins. Experiments using two-hybrid systems and in vitro co-precipitation assays have shown that hnRNP K, hnRNP L, and PTB interact with each other (51). Thus, it is possible that protein-protein interactions determine their affinity and binding specificity. Functional analysis shows that hnRNP K is a component of the enhancer complex that promotes exon 6A splicing through the wild-type S4 sequence. The evidence for this is that, first, poly(rC), which mainly interacts with hnRNP K, inhibits the splicing of the pre-mRNA 6A-S4-7. Second, the addition of recombinant hnRNP K to nuclear extracts preincubated with poly(rC) RNA competitor completely restores splicing efficiency to the original level. However, it is clear from the partial inhibition induced by the addition of poly(rC) that other proteins are required to mediate the activation (see Fig. 7). Consistent with this observation, we have previously shown that ASF/SF2 promotes splicing of exon 6A (40). Based on the various biological functions of hnRNP K, it has been proposed that hnRNP K might serve as a docking platform that promotes the assembly of effector molecules. Thus, it is tempting to speculate that hnRNP K might recruit ASF/SF2 to the activating S4 sequence.
In contrast to the S4 enhancer, splicing of pre-mRNA containing P3S is weakly affected by the addition of poly(rC). This was unexpected, since splicing efficiency and the strength of the hnRNP K interaction are in the same order of magnitude for both pre-mRNAs. We have no simple explanation for this. If the only reason is a higher affinity of hnRNP K for P3S, then splicing of 6A-P3S-7 should be competed by a higher poly(rC) concentration than that needed for 6A-S4-7. This was not the case, which implies that the fraction of hnRNP K that is not sequestered by poly(rC) could be recruited by the other proteins of the enhancer complex. However, we could not totally exclude the possibility that other proteins might play a major role in the enhancing effect of P3S. The additional proteins that are associated with P3S (Fig. 4B, white stars and white arrows) might be good candidates. In the case of pre-mRNA 6A-PyI5-7, the addition of poly(rC) to nuclear extract almost completely inhibits the splicing reaction, suggesting that hnRNP K is the major protein through which activation is mediated. Consistent with this result, the addition of recombinant hnRNP K completely relieves the inhibition of splicing by poly(rC).
hnRNP L also associates with the different enhancer sequences. In a recent study, an RNA binding site for hnRNP L, which confers an increased stability has been identified in the 3Ј-untranslated region of human vascular endothelial growth factor mRNA (59). The 5Ј-CACCCACCCACAUACAUACAU-3Ј sequence contains three repetitions of the ACAU motif that has been suggested to be essential for hnRNP L binding. No such motif can be found in the P3S, S4, PYI5, and S5 sequences. Moreover, the binding of hnRNP L does not correlate with the enhancing activity. Knowing that hnRNP L forms a heterodimeric association with hnRNP K and PTB, one possibility is that hnRNP L binds indirectly to the different enhancers through protein-protein interactions with those hnRNPs (51).  1 and 2), with 20 ng of poly(rC) (lanes 3 and 4 are duplicates), with 20 ng of poly(rC), and with increasing amounts of hnRNP K (lanes 5-9, 300, 600, 600, 900, and 1200 ng, respectively). Lane 0, unspliced pre-mRNA. B, percentage of splicing is shown as a function of the amount of recombinant hnRNP K added, in the presence of 20 ng of poly(rC).
The third protein for which binding to the enhancer is correlated with splicing activation is PTB. The binding of PTB is stronger with the strong enhancer S4 and P3S than with the weaker PyI5 and S5 enhancers. Consistent with PTB interaction, CUCUCU and UCUU motifs that have been identified as PTB binding sites in several alternative pre-mRNAs are present within the S4 and P3S sequences (22,30,60). In the case of P3S, this observation was expected, because P3S is derived from the B3P3 intronic region of the ␣-tropomyosin pre-mRNA from which PTB was originally identified (61). In the case of PyI5 and S5, no clear PTB binding motifs can be found, which is consistent with the weak PTB interaction. What might be the function of PTB in the splicing regulation of exon 6A? In a long list of alternative splicing models, PTB is shown to act as a repressor (22, 24 -27, 30). In the case of the alternative N1 exon from c-src, it has been shown that PTB binding to a CUCUCU motif interspersed within the DCS-activating sequence collaborates with PTB binding to the upstream intron to repress the N1 exon in nonneuronal cells (30). The S4 sequence exhibits similar features to the DCS element, where PTB binding sites are inserted within the activating element. Examination of intronic sequences flanking exon 6A reveals several putative PTB binding sites that are located just upstream of the branch point of exon 6B and within the long pyrimidine stretch upstream of exon 6B. Thus, it is tempting to speculate that PTB binding sites across the intron collaborate to regulate the mutually exclusive use of exon 6A and 6B. Consistent with this hypothesis, deletion of sequence S4 or mutations within the long polypyrimidine stretch allows recognition of exon 6B in both myoblasts and in HeLa cell nuclear extracts (37)(38)(39)62). Another possibility is that PTB binding to the intron upstream of exon 6B might impair the interaction with U2AF and consequently U2 snRNP binding, as has been shown for the short alternative exon of the GABA A receptor ␥2 pre-mRNA (22). However, if this is the case, what is the significance of PTB binding to the S4 sequence? The S4 element contains two putative PTB binding sites: a CUCUCU motif that is juxtaposed just downstream of the C stretch and a UCUU motif that is localized at the 3Ј-end (Fig. 1). Several punctual mutations have been introduced in the CUCUCU motif and tested in vitro with the 6A-7 pre-mRNA. While a mutation at the 5Ј-end of the motif improves splicing of exon 6A in HeLa nuclear extracts, mutations in the middle or at the end of the motif, respectively, decrease or have no effect on splicing (data not shown). Without knowing more about the interaction of hnRNP K and PTB on the mutated RNA molecules, these results are difficult to interpret. Binding sites for hnRNP K and PTB are contiguous, which raises the question of whether the two proteins bind to the same RNA molecule. A possibility is that PTB facilitates the binding of hnRNP K due to protein-protein interactions. Heterodimeric formation between hnRNP K and PTB has already been reported (51). The importance of protein-protein contacts has been well documented in the case of the N1 exon from c-src, in which cooperative assembly of hnRNP complex on the DCS element requires nPTB and/or PTB (35). Thus, it is possible that PTB binding sites on S4 help to recruit hnRNP K. The use of recombinant proteins might address this question.
Juxtaposition of positive and negative elements within the same regulatory sequence seems to be a characteristic of several alternatively spliced pre-mRNAs. Classes of such elements involve PTB as a regulator (22,35). In the case of the S4 sequence, the regulated element is involved in two opposite splicing events: activation of exon 6A and repression of exon 6B in nonmuscle cells and myoblasts. A similar feature has been observed for the IAS3 regulatory sequence (or ISAR in the rat gene) of the FGFR-2 pre-mRNA (63,64). IAS3 is required for the activation of the K-SAM exon and for the repression of the BEK exon. Interestingly, several putative PTB binding sites have been reported within the intron separating the two mutually exclusive exons from the rat gene and in the BEK exon (24,27). Thus, it might be possible that a similar mechanism is used to repress the BEK exon (IIIc in the rat) and exon 6B. Clearly, further experiments are required to elucidate what role PTB plays in the regulation of tissue-specific expression of exon 6A and exon 6B.