The RNA-Binding Protein TIA-1 Is a Novel Mammalian Splicing Regulator Acting through Intron Sequences Adjacent to a 5′ Splice Site

Plasmids. (i) Splicing constructs.

β-Tropomyosin constructs for in vitro synthesis of pre-mRNA substrates were derived from the previously described Tropo 6A-7 clone (1) using standard techniques. Constructs with two competing 5′ss were derived from a truncated adenovirus E1A gene (Sp1), in which the 13S 5′ss (D1 site) is replaced with a polylinker allowing insertion of a region containing two 5′ss (5).

(ii) Expression vectors.

A mouse TIA-1 cDNA clone was obtained as an I.M.A.G.E. consortium clone (identification no. 1261161) containing the TIA-1 coding sequence lacking alternative exon 5. It was used to make pTIA-1, an expression vector for an N-terminal FLAG-tagged TIA-1, and pTIA-coat, an expression vector for an N-terminal FLAG-tagged TIA-1–coat fusion. The coat expression vector pcoat (or pCI-MS2) and the hnRNP A1-coat fusion vector have been described elsewhere (13). The entire coding sequence of hnRNP C1 was introduced into the StuI site of pCI-MS2-NLS-FLAG (13) to make an expression vector for an N-terminal FLAG-tagged hnRNP C1-coat fusion (in which coat sequences are C terminal). The hnRNP C1 expression vector phnRNP C1 was obtained from it by eliminating coat sequences by BamHI digestion and religation.

(iii) Minigenes.

Rat preprotachykinin minigene pBPSVpA+2-7 (37) was a gift from P. J. Grabowski. The CD44 minigene was obtained by cloning a 6.9-kb ClaI-SmaI fragment of the human CD44 gene containing exons v8 to v10 and flanking introns (41) between the KpnI and HindIII sites of pRK3. pRK3 and pRK20 have been described elsewhere (10). RK-MS2 was made by inserting a 137-bp SpeI-EcoRI fragment of pIII/MS2-2 carrying MS2 coat protein binding sites (42) between nucleotides 15 and 505 of the 1,220-bp intron downstream of the K-SAM exon, in an RK3 derivative missing the BEK exon (deletion 1156–1412 of Fig. 1 in reference 12). pRK97 and pRK98 were made from pRK3 and pRK20, respectively, by deletion of the BEK exon's 3′ss and associated polypyrimidine sequence (deletion 1156–1233 of Fig. 1 in reference 12). pRK99 was made from pRK20 by deleting the BEK exon (deletion 1156–1412 of Fig. 1 in reference 12) and then replacing nucleotides 213 to 505 of the intron downstream of the K-SAM exon with the fragment carrying coat binding sites.

Extract preparations and in vitro splicing assays.

HeLa nuclear extract and cytoplasmic S100 extract were prepared as described previously (40). For preparation of whole-cell extracts (WCE) from 293-EBNA cells, cells resuspended in lysis buffer (20 mM Tris-HCl [pH 7.6], 400 mM KCl, 20% glycerol, 1 mM dithiothreitol, 0.2% NP-40, and a cocktail of protease inhibitors) were sonicated and then centrifuged at 10,000 rpm for 10 min in an SS-34 rotor. The supernatant (WCE) was dialyzed against buffer D for 5 h. For preparation of WCE from 293-EBNA cells overexpressing TIA-1 (WCE/TIA-1), cells were transfected with 6 μg of TIA-1-expressing vector and 14 μg of pBluescript SK(+) (Stratagene) and collected 48 h later, and WCE were prepared from them as described above.

Capped ³²P-labeled pre-mRNA substrates were made by runoff in vitro transcription with SP6 RNA polymerase as described in reference 8. For the tropomyosin-derived transcripts, in vitro splicing was performed as described in reference 5 (25-μl final volume, using 12 μl of nuclear extract, in 60 mM KCl–1.3 mM MgCl₂). Splicing of the E1A-derived transcripts was performed under a variety of conditions as indicated in the figure legends. Reaction mixtures were incubated at 30°C for 90 to 120 min, and splicing products were resolved on denaturing 5 to 6% polyacrylamide gels, followed by autoradiography.

UV cross-linking and immunoprecipitation assays.

UV cross-linking was performed as described previously (7) with minor modifications. RNA probes were synthesized in vitro from pBluescript SK(+)-based plasmids containing appropriate sequences (5′ss-IAS1, 5′ss-RAN, IAS1, or RAN) downstream of the T7 promoter. They were uniformly labeled at high specific activity using [α-³²P]UTP, and 50 fmol was used per assay. Cross-linking assay mixtures (15 μl) containing 3 μl of cell extracts, supplemented or not with 150 ng of TIA-1 or hnRNP C1, were incubated with RNA probes in 0.6× buffer D containing 2.6% polyvinyl alcohol and 0.5 μg of Escherichia coli tRNA. After a 20-min incubation at 30°C, reaction mixtures were exposed to UV light for 15 min at 4°C and then treated with a mixture of RNases A (750 ng) and T₁ (250 U) for 30 min at 37°C. For direct analysis, samples were diluted with a 2× sodium dodecyl sulfate (SDS) protein loading buffer and resolved by SDS-polyacrylamide gel electrophoresis (PAGE) on 12% polyacrylamide gels. For UV cross-linking and immunoprecipitation assays, the RNase-treated samples were diluted to 60 μl with IPP buffer (50 mM Tris-HCl [pH 7.5], 150 mM NaCl, 0.1% NP-40), and 5 μg of anti-TIA-1 polyclonal antibody (Santa Cruz Biotech) or 10 μg of anti-hnRNP C1 monoclonal antibody (a generous gift from G. Dreyfuss) was added. After incubation at 4°C for 3 h, 10 μl of protein G-Sepharose beads was added and incubation was continued overnight. After three washes of the beads with 200 μl of IPP buffer, bound proteins were eluted in 20 μl of SDS loading buffer at 100°C for 5 min and loaded on an SDS–12% polyacrylamide gel.

To analyze the role of U1 snRNP in TIA-1 binding to the 5′ss-IAS1 probe, 3-μl aliquots of extracts (WCE or WCE/TIA-1) were pretreated with 0.5 μg of a 14-nucleotide oligodeoxynucleotide complementary to the 5′ end of U1 snRNA, or with a nonrelated probe complementary to the T7 promoter, in the presence of 0.5 U of RNase H, for 30 min at 30°C before cross-linking to the 5′ss-IAS1 probe and immunoprecipitation as described above.

Recombinant proteins and total SR preparation.

Total SR proteins from HeLa cells were purified as described in reference 51. To produce recombinant TIA-1 and hnRNP C1 in E. coli, the corresponding coding sequences from the second (TIA-1) or first (hnRNP C1) codon up to the stop codon were inserted in frame between the BamHI and EcoRI sites of pET28-b (Novagen). Resulting plasmids were used for production of six-His-tagged proteins. They were expressed and purified under nondenaturing conditions as recommended by the manufacturer and dialyzed against buffer D for 1 h.

Transfections and RT-PCR.

Transfection of SVK14 (46) and 293-EBNA cells (Invitrogen) was performed as described previously (10, 13). For cotransfections, 2 μg of minigene was cotransfected with 18 μg of the appropriate expression vector. Forty-eight hours later, RNA was harvested and analyzed by reverse transcription-PCR (RT-PCR) using reporter-specific primers. Preprotachykinin primers P1 and P2 were as follows: P1, GGAAATCGGTGCCAACG; P2, GAGAGATCTGACCATGCC. CD44 and FGFR-2 primers were as follows: P3, ATCCAGTGGATCAAGCAC, and P4, GGCAACCTAGAAGGCACAG. Twenty cycles of amplification were used so as to remain in the range of exponential amplification. PCR products were caused to migrate on agarose gels, transferred to nylon filters (Hybond N+; Amersham), and hybridized with different probes. Experiments were carried out at least in triplicate, and representative results are shown here. ³²P-labeled RNA probes used were obtained by in vitro transcription of DNA fragments corresponding to (i) nucleotides 3 to 134 of the 148-nucleotide K-SAM exon; (ii) linked C1 and C2 exon sequences; and (iii) linked exons 2, 3, and 5 of the preprotachykinin gene.

IAS1 activation of heterologous exons.

Efficient recognition of the chicken β-tropomyosin gene's exon 6A (1) normally requires a 33-nucleotide pyrimidine-rich sequence (S4) starting 37 nucleotides downstream from its 5′ss. Can IAS1 replace S4 in this system? Various pre-mRNA substrates (Fig. 1A) containing exons 6A and 7 were used for in vitro splicing assays in HeLa cell nuclear extract. As expected from previous work (17), splicing (Fig. 1B) of a pre-mRNA lacking S4 (6A-Δ4-7, lane 7) was inefficient compared to splicing of pre-mRNAs with S4 (Tropo 6A-7, lane 2) or a purine-rich sequence (6A-P3AS-7, lane 3). When S4 was replaced with IAS1 (IAS1 down, Fig. 1A), splicing dropped to levels below those observed in the absence of S4 (compare lanes 6 and 7, Fig. 1B). In the IAS1 down pre-mRNA, IAS1 lies 43 nucleotides downstream of the exon 6A-intron junction. In vivo, IAS1 activates splicing only if it lies immediately downstream from the K-SAM exon's 5′ss (F. Del Gatto-Konczak, unpublished data). A further pre-mRNA was made (IAS1 up, Fig. 1A) which lacked S4 but contained IAS1 positioned immediately downstream from the exon 6A 5′ss. Splicing of IAS1 up was at least as efficient as splicing of substrates containing S4 or the purine-rich sequence (compare lanes 2, 3, and 4, Fig. 1B). Pre-mRNA RAN (Fig. 1A) is a version of IAS1 up in which IAS1 has been replaced with a random sequence incapable of activating K-SAM exon splicing in vivo (10). Splicing of pre-mRNA RAN was significantly less efficient than splicing of substrate IAS1 up (Fig. 1B, compare lanes 4 and 5) while being slightly more efficient than splicing of pre-mRNA 6A-Δ4-7 (compare lanes 5 and 7). These results show that IAS1 can activate splicing of a heterologous exon in vitro, provided that it is positioned immediately downstream of the exon's 5′ss. They also show that the random sequence does not act as a repressor of an adjacent 5′ss in vitro.

In the FGFR-2 pre-mRNA, IAS1 is involved in a competitive splicing choice. This encouraged us to test if IAS1 can function when two 5′ss are in competition. We used pre-mRNAs derived from the adenovirus E1A gene (Fig. 2), in which the unique 13S 5′ss (D1) has been replaced with different pairs of competing splice sites (5). Pre-mRNA D2/D2-wt contains two identical copies of the E1A 12S 5′ss D2. When this pre-mRNA is spliced in nuclear extract, the distal D2 5′ss is markedly preferred to the proximal D2 5′ss (Fig. 3A, lane 1). This preference for the distal 5′ss is probably due to involvement of the nuclear cap binding complex (CBC), which acts to favor use of the 5′ss closer to the pre-mRNA's cap (31). However, placing IAS1 immediately downstream from the proximal site (pre-mRNA D2/D2-IAS1, Fig. 2) induces a very strong shift of splicing toward use of this latter site (Fig. 3A, lane 2).

TIA-1 activation of 5′ss.

Having established that IAS1 can activate splicing of an adjacent 5′ss in vitro, we searched for possible effects of TIA-1 on IAS1 activity. To stand a chance of observing an effect on splicing in vitro of an increase in TIA-1 levels, it is necessary to start with splicing extracts in which the concentration of TIA-1 is suboptimal. For this reason, we used S100 extract (with added SR proteins) for further experiments, as TIA-1 is less abundant in cytoplasmic S100 extracts than in nuclear extracts (data not shown). In the absence of added exogenous TIA-1, only use of the proximal D2 site linked to IAS1 was detected in S100 extract (Fig. 3B, lane 1). Splicing was less efficient than that in nuclear extract (compare with Fig. 3A, lane 2), most probably because several factors, including TIA-1, are limiting in the S100 extract. Addition of recombinant TIA-1 (600 ng) led to a strong stimulation of splicing using the D2-IAS1 site (Fig. 3B, lane 2). In contrast, addition of hnRNP C1, a protein which, like TIA-1, binds to U-rich sequences (20), actually decreased use of the D2-IAS1 site (lane 3). While no comparable stimulation by TIA-1 of use of a D2 site linked to the random sequence was observed (compare lanes 4 and 5), addition of 600 ng of TIA-1 did weakly stimulate use of both copies of the D2 5′ss in the D2/D2-wt substrate (compare lanes 4 and 5 in Fig. 3B). This suggests that TIA-1 can also activate 5′ss linked to sequences other than IAS1, at least in S100 extract supplemented with SR proteins, which, compared to nuclear extract, is suboptimal for splicing. Note that both D2 5′ss copies are stimulated to approximately the same extent, and so TIA-1 is not preferentially activating either the proximal or the distal 5′ss here. Importantly, activation by TIA-1 of the D2 5′ss not linked to IAS1 cannot be detected on the pre-mRNA containing a competing D2 5′ss adjacent to IAS1 (compare lanes 1 and 2, dist. D2/A mRNA). TIA-1 is thus showing a preference for the 5′ss adjacent to IAS1.

Can this effect of TIA-1 be reproduced using other pairs of splice sites, particularly when one of them is the K-SAM exon D_SAM 5′ss? We have analyzed splicing of pre-mRNA from other constructs, which contain as competing splice sites the strong E1A 13S D1 5′ss and the weaker D_SAM 5′ss. In pre-mRNA D1/D_SAM-IAS1 (Fig. 2), the D_SAM site is followed by IAS1. In pre-mRNA D1/D_SAM-RAN, IAS1 has been replaced with the random sequence described above. When the D1/D_SAM-IAS1 pre-mRNA was spliced in S100 extract (with added SR proteins), the D1 site was scarcely used for splicing (Fig. 4A, lane 2), while the D_SAM 5′ss was preferred (Fig. 4A, lane 2). However, addition of TIA-1 (300 or 600 ng) strongly activated use of the D_SAM 5′ss (approximately fourfold), without affecting the very weak use of the D1 5′ss (lanes 3 and 4). As observed previously (Fig. 3B), addition of hnRNP C1 decreased use of the IAS1-linked (D_SAM) 5′ss (Fig. 4A, lane 5). No comparable TIA-1-induced activation was observed when the D_SAM 5′ss was linked to the random sequence (compare lanes 6 and 7). However, addition of 600 ng of TIA-1 did stimulate use of the D1 5′ss in the D1/D_SAM-RAN substrate (compare lanes 6 and 7 in Fig. 4A). Yet despite this, no activation of the D1 5′ss by TIA-1 could be observed when it was in competition with an IAS1-linked 5′ss (the D_SAM-IAS1 5′ss) on D1/D_SAM-IAS1 pre-mRNA (compare lane 2 to lanes 3 and 4). TIA-1 is once again showing a preference for the 5′ss adjacent to IAS1.

If TIA-1 does really have a preference for the IAS1-linked 5′ss, it should be able to provoke a significant switch in splice site use under appropriate circumstances. To attempt to visualize such a switch, we set out to increase the use of the D1 site for splicing of the D1/D_SAM-IAS1 pre-mRNA. On this pre-mRNA, the D1 site is distal, and use of a distal 5′ss close to a cap site is known to be facilitated by the CBC (31). Cap binding protein is less abundant in S100 extract than in nuclear extract, and its concentration in S100 extract may be suboptimal (25). We therefore performed splicing assays using the D1/D_SAM-IAS1 pre-mRNA in the presence of a mixture of S100 extract and nuclear extract, trying in this way to limit starting concentrations of TIA-1 while benefiting from significantly higher levels of CBC. Under these conditions, the distal D1 site was indeed efficiently used (Fig. 4B, lane 2). The competing (proximal) D_SAM 5′ss was also used for splicing, though at a lower level (Fig. 4B, lane 2). However, addition of TIA-1 (200, 400, or 600 ng) strongly activated use of the D_SAM 5′ss (approximately fourfold), whereas splicing using the D1 5′ss concomitantly decreased (lanes 3 to 5). These results show that TIA-1 can effectively provoke a significant switch in splice site use, from predominant use of the D1 site (a strong 5′ss) to approximately equal use of the D1 and the IAS1-linked D_SAM 5′ss. Taken together, our results show that TIA-1 activates 5′ss use in a manner dependent on the downstream intron sequence, with a preference for a downstream U-rich sequence.

Preferential binding of TIA-1 to IAS1 adjacent to a 5′ss.

We used UV cross-linking analysis to test whether TIA-1 can bind to IAS1. The first RNA probe used (5′ss-IAS1) contained a strong 5′ss sequence (AG/GUAAGU) linked to IAS1. The major protein cross-linked to this probe in HeLa cell nuclear extract was an approximately 60-kDa protein (which might be PTB or U2AF, both known to bind to pyrimidine-rich sequences); several smaller proteins were also detected (Fig. 5A, lane 1). A number of proteins cross-linked weakly to the 5′ss-IAS1 probe in the S100 extract, which contains only low amounts of TIA-1 and hnRNP C (Fig. 5A, lane 2). When the latter extract was enriched with recombinant TIA-1 or hnRNP C1, the probe cross-linked to major new proteins with net molecular masses of ≈45 (corresponding in size to recombinant TIA-1, lane 3) and ≈40 (corresponding in size to recombinant hnRNP C1, lane 4) kDa, respectively. We also tested S100 extract enriched with recombinant ASF/SF2 lacking its RS domain. This protein, which binds with high affinity to purine-rich sequences, was not detected with the 5′ss-IAS1 probe (data not shown).

To test if TIA-1 expressed in mammalian cells can also bind to IAS1, we performed similar experiments using WCE prepared from human 293-EBNA cells, either transfected or not with the mouse TIA-1 expression vector pTIA-1. Transfection led to an approximately 10-fold increase in TIA-1 levels (data not shown). The 5′ss-IAS1 probe cross-linked to several proteins in the untransfected, control WCE (Fig. 5A, lane 5). In WCE from cells transfected with pTIA-1, a major new protein corresponding in size to TIA-1 was detected with the 5′ss-IAS1 probe (lane 6). However, when the 5′ss-RAN (the random sequence linked to a 5′ss) probe was used, no difference between WCE from untransfected (lane 8) and transfected (lane 9) cells could be observed. Thus, these results show that both TIA-1 produced bacterially and TIA-1 produced in a mammalian cell can bind specifically and efficiently to IAS1.

Somewhat different results were obtained when the probe used was IAS1 without an adjacent 5′ss (Fig. 5B). This probe detects a major ≈40-kDa protein in HeLa cell nuclear extract (lane 1) which is hnRNP C (Fig. 5D), and a major ≈80-kDa protein in S100 extract (lane 2). While we detected cross-linking of the IAS1 probe with recombinant TIA-1 (lane 3) or hnRNP C1 (lane 4) in S100 extract supplemented with these proteins, interaction of TIA-1 with the IAS1 probe appeared weaker than that with the 5′ss-IAS1 probe (compare lane 3, Fig. 5B, with lane 3, Fig. 5A). This difference between the two probes was also observed using WCE from cells overexpressing TIA-1 following transfection (compare Fig. 5A and B, lane 6). As expected, we did not detect any cross-linking of TIA-1 to the random sequence probe RAN (Fig. 5B, compare lanes 8 and 9).

Further experiments were carried out to demonstrate that the ≈45-kDa protein detected in extracts with the 5′ss-IAS1 probe really was TIA-1. A variety of extracts were incubated with either the 5′ss-IAS1 probe or the IAS1 probe before UV cross-linking. Aliquots were analyzed either before immunoprecipitation (total) or after immunoprecipitation with antibodies recognizing either TIA-1 or hnRNP C1. Both TIA-1 and hnRNP C1 cross-linked to the 5′ss-IAS1 probe in HeLa cell nuclear extract (Fig. 5C, lanes 2 and 3), albeit weakly. This cross-linking appears to be weak because of competition for the probe by the major 60-kDa cross-linking protein (lane 1). (Note that this competition does not necessarily take place during in vitro splicing assays where bona fide splicing substrates are used.) Thus, when the IAS1 probe is used (Fig. 5D), cross-linking to the 60-kDa protein is much less marked (lane 1), while cross-linking to hnRNP C1 concomitantly increases greatly (lane 3). However, despite the apparently greater probe availability, no cross-linking of the IAS1 sequence to TIA-1 is seen (lane 2). This result is in agreement with those shown in Fig. 5A and B, which demonstrated that TIA-1 binds very weakly to IAS1 which is not adjacent to a 5′ss.

Using WCE from untransfected 293-EBNA cells (Fig. 5C, lanes 4 to 6), we detected efficient cross-linking of both TIA-1 (lane 5) and hnRNP C1 (lane 6) to the 5′ss-IAS1 probe. Furthermore, a strong increase of TIA-1 cross-linking was observed with WCE from cells transfected with pTIA-1 (compare lanes 5 and 8). Interestingly, when the IAS1 probe was used, TIA-1 was detected, and only weakly, solely in WCE from the transfected cells (Fig. 5D, compare lanes 5 and 8), while hnRNP C1 was detected in WCE from both untransfected (lane 6) and transfected (lane 9) cells. Taken together, the results in Fig. 5 demonstrate that TIA-1 interacts preferentially with the IAS1 motif when adjacent to a 5′ss. This conclusion was confirmed by competition experiments using WCE from transfected cells. Interaction of TIA-1 with the labeled 5′ss-IAS1 probe was strongly reduced (four- to fivefold) in the presence of an 80-fold excess of unlabeled 5′ss-IAS1 probe, while the presence of a 320-fold excess of unlabeled IAS1 probe had no significant effect (data not shown).

TIA-1 binding is U1 snRNP dependent.

The preferred binding site for TIA-1 has been identified as a U-rich sequence by experiments analyzing TIA-1–RNA interaction in the absence of any other cellular proteins (14). Under these conditions, there was no indication that the U-rich sequence need be adjacent to a sequence resembling a 5′ss. It seemed possible to us that the preferred binding of TIA-1 to IAS1 adjacent to a 5′ss in cell extracts might reflect the presence, in these extracts, of U1 snRNP and its binding to the 5′ss. Is the binding of TIA-1 to IAS1 adjacent to a 5′ss in fact dependent on U1 snRNP binding to the 5′ss? To address this question, we incubated WCE from TIA-1-transfected cells with either an oligonucleotide complementary to the 5′ 14 nucleotides of U1 snRNA or an “irrelevant” oligonucleotide (corresponding to one strand of the T7 promoter), in the presence of RNase H. The former oligonucleotide provokes degradation of the 5′ nucleotides of U1 snRNA necessary for U1 snRNP binding to the 5′ss (and so abolishes this binding). The latter oligonucleotide has no such effect. When WCE from TIA-1-transfected cells were subjected to either a mock preincubation or preincubation with the T7 oligonucleotide before cross-linking to the 5′ss-IAS1 probe, probe cross-linking to both TIA-1 and hnRNP C1 was readily observed (Fig. 6, lanes 1 and 3). However, when preincubation was done with the antisense U1 snRNA oligonucleotide, cross-linking to TIA-1 was selectively eliminated (lane 2). This result was confirmed when aliquots of cross-linked material were subjected to immunoprecipitation using anti-TIA-1 antibodies (compare lanes 4 and 5). Similar results were observed when WCE from untransfected cells were analyzed (lanes 6 to 10), although the amount of cross-linked TIA-1 was, as expected, lower. The U1 snRNP-dependent cross-linking of TIA-1 to the probe could nevertheless be visualized after immunoprecipitation of samples with anti-TIA-1 antibodies (compare lanes 9 and 10). We conclude that TIA-1 binding to IAS1 in cell extracts is dramatically enhanced by U1 snRNP binding to an adjacent 5′ss.

TIA-1 enhances K-SAM exon splicing in vivo.

As described previously (10), the RK3 minigene (Fig. 7) contains a wild-type FGFR-2 gene fragment carrying the very similarly sized alternative K-SAM and BEK exons, together with flanking intron sequences and the upstream and downstream constitutive exons C1 and C2. Transfection of epithelial SVK14 cells with RK3 leads to preferential splicing of the K-SAM exon, while transfection of 293-EBNA cells leads to preferential splicing of the BEK exon. Skipping of both exons is also observed at a low level. Cotransfection of 293-EBNA cells with RK3 and pcoat (a bacteriophage MS2 coat protein expression vector) resulted in BEK exon splicing (Fig. 8A and B, lanes 1), reflected by major RT-PCR product C1BC2 (Fig. 8E). For RK3 and derivatives, two probes were used to identify RT-PCR products. The C1C2 probe detects all products, while the K-SAM probe detects only products containing the K-SAM exon. Cotransfection of RK3 with pTIA-1 resulted in splicing of the K-SAM exon to the BEK exon (Fig. 8A and B, lanes 2), reflected by the major product C1SBC2, though some C1SC2 product is also detected (Fig. 8E). Importantly, cotransfection of RK3 with the hnRNP C1 expression vector did not detectably activate K-SAM exon splicing (lanes 3), suggesting that the TIA-1 effect on the K-SAM exon cannot be reproduced by just any U-rich sequence binding protein. RK20 (Fig. 7) is a version of RK3 in which IAS1 has been replaced with the random sequence (10) unable to bind TIA-1. Cotransfection of RK20 with pTIA-1 did not lead to K-SAM exon splicing, as RT-PCR products of type C1BC2 (Fig. 8E) were obtained regardless of whether RK20 was cotransfected with pcoat (Fig. 7A and B, lanes 4), pTIA-1 (lanes 5), or phnRNP C1 (lanes 6). TIA-1 activation of K-SAM exon splicing is thus IAS1 dependent. Note that, in the absence of IAS1, TIA-1 activates splicing of C1 to C2 somewhat (Fig. 8A, lane 5) but that this effect cannot be detected when IAS1 is present (Fig. 8A, lane 2). This is in agreement with our in vitro splicing results, which show that, while TIA-1 can activate a variety of 5′ss, when a 5′ss linked to IAS1 is in competition with one not so linked, it is use of the former which is favored by TIA-1.

TIA-1 can activate K-SAM exon splicing not only to the BEK exon but also to the C2 exon. In RK97 (Fig. 7), the BEK exon's 3′ss and associated polypyrimidine sequence have been deleted, blocking its splicing. Cotransfection of RK97 with pcoat in 293-EBNA cells results mainly in skipping of the K-SAM exon. The major RT-PCR product obtained (Fig. 8C and D, lanes 1) is C1C2 (Fig. 8E); only a little splicing of the K-SAM exon is observed (product C1SC2). Cotransfection of RK97 with pTIA-1 leads to a marked increase of K-SAM exon splicing, as evidenced by an increase in the levels of the C1SC2 product (Fig. 8C and D, lanes 2). Once again, this activation of K-SAM splicing is IAS1 dependent, as cotransfection of pTIA-1 with RK98, a version of RK97 in which IAS1 has been replaced with the random sequence (Fig. 7), has no similar effect (compare lanes 3 and 4 of Fig. 8C and D).

Artificial recruitment of TIA-1 obviates the IAS1 requirement.

Our results suggest that IAS1 serves as a binding site to recruit TIA-1 close to the 5′ss. If so, artificial recruitment of TIA-1 downstream from the 5′ss might activate splicing of a K-SAM exon not linked to IAS1. In RK-MS2 (Fig. 7), which has the BEK exon deleted, nucleotides 15 to 505 (which include IAS1) of the intron downstream from the K-SAM exon have been replaced with a tandem copy of the bacteriophage MS2 coat protein operator. Proteins can thus be recruited to RK-MS2 pre-mRNA downstream from the K-SAM exon as fusions with coat protein. When RK-MS2 is transfected together with the empty expression vector pCI-neo or pcoat, the K-SAM exon is skipped. RT-PCR products detected with the C1C2 probe (Fig. 9A, lanes 1 and 3) reflect splicing of exon C1 to exon C2 (product C1C2). This result is expected, as IAS1 is required for efficient K-SAM exon splicing. Cotransfection of pTIA-1 with RK-MS2 does not detectably induce K-SAM exon inclusion: note that none of the RT-PCR products detected with the C1C2 probe (Fig. 9A, lane 2) hybridize with the K-SAM exon probe (Fig. 9B, lane 2). However, when RK-MS2 is cotransfected with an expression vector for a TIA-1–coat fusion protein, some inclusion of the K-SAM exon is induced: RT-PCR products which hybridize to the K-SAM probe and correspond to C1SC2 can be detected (Fig. 9A and B, lanes 4). No activation of K-SAM exon splicing was observed when RK-MS2 was cotransfected with expression vectors for hnRNP C1-coat fusions (Fig. 9A and B, lanes 5) or hnRNP A1-coat fusions (lanes 6).

The K-SAM exon inclusion induced by TIA-coat with RK-MS2 pre-mRNA is less than that induced by TIA-1 with pre-mRNAs containing IAS1 (compare Fig. 8 and 9), although approximately equal amounts of the two proteins are made following transfection of their expression vectors (data not shown). The TIA-coat fusion bound to the MS2 operator may be presented to the splicing apparatus suboptimally compared to TIA-1 in its natural position. Activation by the TIA-coat fusion is, as expected, position dependent. In RK99 (Fig. 7), the operator is placed 213 bp downstream from the 5′ss, and not 15 bp downstream as in RK-MS2. (Note that RK99 does not contain IAS1, which has been replaced with the random sequence of RK20.) When RK99 is transfected together with the TIA-coat expression vector, no activation of K-SAM exon splicing is observed (Fig. 9C and D). In fact, TIA-coat expression now represses K-SAM exon splicing (compare lanes 1 and 2).

TIA-1's effect on other exons in vivo.

To test whether TIA-1 can influence alternative splicing of other exons in vivo, we used several minigenes reflecting well-documented cases of alternative exon splicing. Minigenes were cotransfected into cells with different expression vectors, and splicing patterns were investigated by RT-PCR analysis of transfected cell RNA using minigene-specific primers. Normally, splicing of preprotachykinin pre-mRNA involves preferential skipping of optional exon 4 (23, 37). In 293-EBNA cells cotransfected with a preprotachykinin minigene (containing exons 2 to 7) and the empty expression vector pCI-neo, inclusion of exon 4 is inefficient (34%) as judged by RT-PCR analysis (Fig. 10A, lane 1). When the minigene is cotransfected with the expression vector pTIA-1, exon 4 inclusion increases to 63% (lane 2). However, a similar increase also occurs (to 51%, lane 3) following cotransfection with the hnRNP C1 expression vector phnRNP C1.

Exons v8, v9, and v10 of the CD44 gene's pre-mRNA are spliced to generate mRNA for the epithelial cell form of CD44 (9). A minigene was made in which these exons and their flanking introns were placed between two constitutively spliced exons, C1 and C2, of the human FGFR-2 gene (Fig. 10B). When this minigene was transfected into the epithelial cell line SVK14, exons v8, v9, and v10 were efficiently spliced to the flanking exons C1 and C2 (data not shown). However, when the minigene was transfected into 293-EBNA cells together with the control vector pCI-neo, RT-PCR analysis revealed a number of products (Fig. 10B, lane 1). The identity of some of these products was investigated by hybridization to a probe composed of v8, v9, and v10 sequences and by sequencing of subcloned fragments (data not shown). Unlike in SVK14 cells, product C1V8V9V10C2 is not the major product, representing only 12% of all products. More abundant products correspond either to skipping of all CD44 exons (product C1C2) or to splicing of exons v8, v9, and v10, with use of an alternative 5′ss for exon v9 and inclusion of an additional exon (represented by a black box) between v9 and v10 (product C1V8V9*V10C2). (This splicing possibility has already been described for the mouse CD44 gene [50]). The identity of other products was not established. When the CD44 minigene was transfected into 293-EBNA cells together with pTIA-1, the C1C2 product disappeared and the RT-PCR products (lane 2) shifted in favor of the C1V8V9V10C2 product, which now represents 45% of all products. A similar increase of this product (to 34% of all products, lane 3) also occurs following cotransfection with the hnRNP C1 expression vector.

In summary, for both the preprotachykinin and CD44 pre-mRNAs, TIA-1 overexpression and hnRNP C1 overexpression have qualitatively similar effects. The smaller effect of phnRNP C1 transfection could be explained if the transfection-induced increase in hnRNP C1 levels is lower than that in TIA-1 levels. On the other hand, our in vitro analysis has shown that TIA-1 can activate 5′ss not linked to IAS1, albeit less effectively than those so linked. This could also contribute to the increased effect of TIA-1 relative to that of hnRNP C1. TIA-1 overexpression does not lead to the activation of splicing of all exons, however, as TIA-1 had no detectable effect (data not shown) on splicing of another alternative exon, the poorly spliced EIIIb exon (24) of the rat fibronectin gene.

The similar effects on preprotachykinin and CD44 pre-mRNAs of TIA-1 and hnRNP C1 overexpression suggest that increasing the level of any protein which binds to U-rich sequences may suffice to perturb splicing in these cases. One possible mechanism could involve competition for pyrimidine-rich binding sites with a protein such as PTB. Insofar as PTB is known to repress splicing of a variety of exons (48a), limiting its access to the pre-mRNA could favor exon inclusion. It should be recalled that, in contrast to the preprotachykinin and CD44 exons, for the K-SAM exon, TIA-1 overexpression markedly stimulates splicing, while hnRNP C1 overexpression has no detectable effect. This suggests that TIA-1's stimulation of K-SAM exon splicing cannot be attributed to the same type of effect as that exerted on either preprotachykinin or CD44 exon splicing.