Far upstream element-binding protein 1 and RNA secondary structure both mediate second-step splicing repression

Significance

Splicing of mRNA precursors occurs in two sequential transesterification steps. We characterize a highly unusual inhibition of splicing in which the reaction is blocked between the two steps. We demonstrate that RNA secondary structure and an exonic splicing silencer element (ESS) can independently mediate second-step splicing repression in vitro and cause exon exclusion in vivo. Importantly, we provide evidence that far upstream element-binding protein 1, a single-stranded DNA- and RNA-binding protein initially identified as a regulator of MYC transcription and recently implicated in several cancers, binds the ESS and functions as a splicing regulator in vitro and in vivo.

Abstract

Splicing of mRNA precursors consists of two steps that are almost invariably tightly coupled to facilitate efficient generation of spliced mRNA. However, we described previously a splicing substrate that is completely blocked after the first step. We have now investigated the basis for this unusual second-step inhibition and unexpectedly elucidated two independent mechanisms. One involves a stem–loop structure located downstream of the 3′splice site, and the other involves an exonic splicing silencer (ESS) situated 3′ to the structure. Both elements contribute to the second-step block in vitro and also cause exon skipping in vivo. Importantly, we identified far upstream element-binding protein 1 (FUBP1), a single-stranded DNA- and RNA-binding protein not previously implicated in splicing, as a strong ESS binding protein, and several assays implicate it in ESS function. We demonstrate using depletion/add-back experiments that FUBP1 acts as a second-step repressor in vitro and show by siRNA-mediated knockdown and overexpression assays that it modulates exon inclusion in vivo. Together, our results provide additional insights into splicing control, and identify FUBP1 as a splicing regulator.

Removal of introns from pre-mRNAs by splicing is a precise process required for the expression of nearly all genes in human cells. Splicing takes place via two sequential transesterification reactions in the spliceosome, a large complex consisting of several hundred proteins and five small nuclear RNAs (1, 2). During the first step, an adenosine residue designated the branch point, attacks the 5′ splice site (5′SS) to generate the splicing intermediates (free exon1 and lariat-exon2). In the second step, the first exon attacks the 3′ splice site (3′SS), yielding ligated exons and a lariat intron.

Splicing, like numerous other cellular processes, must be under strict regulatory control. Indeed, aberrant splicing is involved in a wide range of human diseases (3, 4). Besides the core splicing signals, additional cis-regulatory elements play important roles in splicing and its control. These elements exist within exons and introns and can function both positively and negatively (5–7). Two major classes of exonic elements have been identified, known as exonic splicing enhancers (ESEs) and exonic splicing silencers (ESSs), and both of these contribute to regulation of alternative splicing (AS). ESEs are usually bound by members of the serine/arginine-rich (SR) protein family to enhance recognition of adjacent splice sites. By contrast, ESSs function to inhibit the use of adjacent splice sites, and are generally dependent on the interaction with members of the heterogeneous nuclear ribonucleoprotein(hnRNP) family, as well as other splicing regulators such as NOVA1/2 and RBFOX1/2 (8, 9).

A good deal is now known about the structure and function of ESSs. For example, several large-scale strategies have been developed to identify and characterize ESSs (10–12). The sequences of known and selected ESSs share little similarity with each other, perhaps reflecting the diversity of their cognate binding proteins. It has been thought that ESSs inhibit splicing either by antagonizing the functions of ESEs or by recruiting factors that interfere directly with the splicing machinery. Upon binding of splicing repressors to ESSs, splicing is most often blocked at the stages of splice-site recognition and/or early spliceosome assembly (5, 6). However, inhibition can also occur at later stages of spliceosome assembly or activation. For example, in Drosophila, the sex-lethal protein (Sxl), an hnRNP-like splicing repressor, autoregulates its own splicing by binding to the pre-mRNA and causing exon exclusion by blocking splicing at the second step via interference with a splicing factor, SPF45, specifically required for the second step (13). However, examples of splicing regulation resulting from second-step inhibition have not been described in vertebrates.

RNA secondary structures can also contribute to splicing control. For example, RNA structure has been shown to affect splicing efficiency by masking splice sites (14) or binding sites for splicing factors (15). Indeed, an increasing amount of genomic data in recent years indicate that RNA structures might be a common method to regulate AS by repressing exon inclusion (16, 17). Additionally, there is also evidence, although currently only in yeast, that secondary structure can function to inhibit the second step of splicing. A stem–loop inserted directly after the branch point of the ACT intron inhibits the second step in vitro (18). Likewise, an exonic stem–loop structure adjacent to the 3′SS in an ACT-LacZ fusion gene primarily blocks the second step in vivo (19).

We previously characterized the role of the serine/arginine-rich splicing factor 10 (SRSF10) in regulation of AS in murine embryonic hearts (20). In that study, we showed that SRSF10 controls AS of the pre-mRNA encoding the cardiac protein triadin, by binding to and activating inclusion of exon 9. During the course of these studies, we also examined splicing in vitro of triadin exon 10 in the context of a chimeric β-globin substrate in which exon 10 was placed in the downstream exon (β-E10). We had used this assay previously with other exons to characterize the role of SRSF10 as a sequence-specific splicing activator (21). Unexpectedly, we found that the β-E10 substrate underwent the first step of splicing efficiently but was completely blocked before the second step (20). This led us to suggest that exon 10 contains sequences that can act as a potent ESS, but functioning to block splicing at the second step.

In this study, we investigated the mechanism underlying the second-step inhibition observed with the β-E10 substrate. Remarkably, we found that two independent mechanisms contribute to the block. One is mediated by a serendipitously created stem–loop structure downstream of the 3′SS that involves β-globin and triadin exon 10 sequences. This structure contributes to the second-step block in vitro and also induces exon skipping in vivo. The second mechanism involves an ESS in exon 10, which also induces a second-step block in vitro and exon skipping in vivo. Furthermore, we show that FUBP1, a single-stranded DNA- and RNA-binding protein that was implicated previously in regulation of MYC transcription (22) and aspects of RNA metabolism other than splicing (23, 24) binds the exon 10 ESS. Depletion/add-back experiments show that FUBP1 can induce the second-step block in vitro, whereas siRNA-mediated knockdown or overexpression of FUBP1 causes exon skipping and AS changes in vivo. Our results thus indicate that both RNA secondary structure and an ESS can repress splicing at the second step and that FUBP1 can function as an unusual second-step splicing repressor.

Results

Characterization of a Second-Step Splicing Block.

We previously constructed and analyzed chimeric β-globin substrates (β-E9 and β-E10) in which triadin exon 9 and exon 10 were inserted 16 nt downstream of the 3′SS, replacing the remaining sequences in the second exon of the β-globin gene (20). Whereas the β-E9 RNA was efficiently spliced in HeLa nuclear extract (NE) (Fig. 1A, lanes 1–3), the β-E10 RNA only produced what appeared to be the 5′ exon cleavage product (Fig. 1A, lanes 4–6; see also ref. 20), consistent with the splicing reaction stalling after completion of the first step.

Fig. 1.

Characterization of the second-step splicing block. (A) In vitro splicing was performed in HeLa NE using β-E9 or β-E10 RNAs for the indicated time. Products of splicing were analyzed by denaturing 6% (vol/vol) PAGE and autoradiography. Splicing products are indicated schematically. (B) The β-E10 plasmid was linearized with the enzymes indicated at the bottom, and the two RNA substrates were analyzed for splicing in NE as in A. (C) β-E9 or β-E10 RNAs were incubated in NE in the presence (lanes 1–3 and 5–7) or absence (lanes 4 and 8) of ATP for the indicated time. Splicing products were resolved on a 12% (vol/vol) denaturing PAGE. (D) His-SRSF1, His-SRSF2, or His-SRSF10 (100 ng) was incubated with β-E9 or β-E10 RNAs in HeLa S100 alone or supplemented with NF40-60. Products were analyzed as in A.

We first characterized the apparent second-step inhibition in greater detail. We initially asked whether inhibition indeed required exon 10 sequences. To test this, we analyzed splicing of an RNA substrate that was prepared from β-E10 linearized using AccI, as indicated in Fig. 1B. The resultant RNA lacks the E10 sequences, containing only a 12-nt β-globin–derived second exon. This RNA was spliced efficiently, in contrast to the second-step inhibition observed with the β-E10 RNA (Fig. 1B, lanes 1–3, compare with lanes 4–6), indicating that the E10 sequences are required for inhibition. Given that only one of the two splicing intermediates, the 5′ exon, was apparent, we next wished to confirm that this was indeed a splicing-related reaction and that this RNA was not simply a degradation product that was fortuitously the same size as 5′ exon. To this end, we performed in vitro-splicing assays as a function of time and in the presence or absence of ATP and resolved the products on a 12% (vol/vol) polyacrylamide denaturing gel instead of the 6% (vol/vol) polyacrylamide gel used above. As shown in Fig. 1C, incubation of the β-E10 RNA under splicing conditions resulted in time- and ATP-dependent accumulation of the lariat-3′ exon intermediate, which migrated more slowly than the precursor RNA, as well as the free 5′ exon (lanes 1–4, compare with lanes 5–8). As expected, first-step splicing of the β-E10 RNA was also Mg²⁺-dependent (Fig. S1). Additionally, using an S100-based splicing assay (21), first-step splicing of the β-E10 RNA was enhanced by SR proteins, such as SRSF1 and SRSF2 (Fig. 1D, lanes 13–16). Together, these results demonstrate that the β-E10 RNA undergoes the first step of splicing efficiently in vitro but is unable to proceed to the second step, reflecting the presence of inhibitory sequences in E10 that specifically block the second step.

E10 Sequences Induce Exon Skipping in Vivo.

We next wished to determine whether the second-step inhibition observed in vitro could affect exon use in vivo. To address this, we used minigene constructs containing E9 or E10 sequences in a middle exon flanked by β-globin genomic sequences, driven by a CMV promoter (Fig. 2A; see Materials and Methods for details). Similar substrates have been used previously to analyze ESE-dependent exon inclusion (25). All four splice sites, as well as immediately adjacent sequences, were the same in all constructs. Splicing was assayed following transient expression in HeLa cells. Total RNA was extracted from transfected cells and exon inclusion/skipping was analyzed by RT-PCR (Fig. 2B). Strikingly, whereas the E9-containing exon was fully included, the corresponding E10 exon was completely excluded. Thus, the strong second-step block brought about by E10 sequences in vitro correlates with complete exon exclusion in vivo.

Fig. 2.

E10 sequences induce exon skipping in vivo. (A) Diagram of a splicing minigene construct containing E9 or E10 in the context of β-globin genomic sequences and the two alternatively spliced products. Sequences derived from β-globin exon 1 or exon 2 are boxed in white and gray, respectively, and blue boxes represent E9 and E10 sequences. Primer pairs used in the RT-PCR in B are shown as two reverse arrows. (B) RT-PCR was performed with RNAs extracted from HeLa cells transiently transfected with the indicated plasmids, and products were analyzed directly on a 1.5% (vol/vol) agarose gel and visualized with ethidium bromide. The empty vector was used as a control.

A Secondary Structure Downstream the 3′SS Contributes to Exon Skipping in Vivo.

We next set out to investigate the molecular basis for the observed E10-induced splicing inhibition. One possibility we considered is that an inhibitory secondary structure forms in the β-E10 RNA. We initially used RNAfold (26) for RNA secondary structure prediction, which revealed a potential stem–loop structure immediately downstream of the 3′SS, involving β-globin exon 2 sequences and the first 10 nt of E10 (Fig. 3A). To test whether the putative secondary structure contributes to exon exclusion, we first made two 7-mer mutations in the E10-based construct, both of which disrupt the putative stem: one mutated UGCUGGU to CCGACGC (E10M1), and the other mutated the complementary sequence AUCAGUA to GCGUUGG (E10M2) (Fig. 3A, Upper). Following transfection and RT-PCR analysis, we found that both mutations strongly activated E10 inclusion (Fig. 3A, Lower, lanes 2–3, compare with lane 1). Furthermore, when the secondary structure was reconstituted by combining the two mutations (E10M1/E10M2), E10 containing exon was again totally skipped (Fig. 3A, Lower, lane 4).

Fig. 3.

A secondary structure downstream the 3′SS contributes to exon skipping in vivo. (A, Upper) Putative stem–loop structure located downstream of the 3′SS involving β-globin exon 2 (black) and E10 (red) sequences. Two mutants (E10M1 and E10M2) disrupt the putative stem, which is reconstituted in the double mutant (E10M1/M2). The 3′SS is indicated by an arrow. (A, Lower) HeLa cells were transiently transfected with each of the E10 mutant constructs, and RT-PCR was performed and products analyzed as described in Fig. 2B. (B, Upper) Diagram of several E10 truncations. (B, Lower) In vivo-splicing analysis of E10 truncations. (C, Upper) Diagram of a spacer inserted downstream the 3′SS of the F19 construct. (C, Lower) F19 constructs with different length spacer were analyzed by RT-PCR as above. (D, Left) Diagram of a stem–loop structure in the E9/stem minigene. Mutations (red) were introduced in the E9 construct as indicated. (D, Right) HeLa cells were transfected with E9 or E9/stem constructs, and RT-PCR analysis was performed as above.

To provide additional support for the importance of the stem in inducing exon exclusion, we analyzed several mutants in which a truncated version of E10 was substituted for the intact E10, as shown in Fig. 3B. The stem–loop structure is predicted to remain intact in the F19 and F30 constructs, whereas it is disrupted in L24 and M39. Consistent with the importance of the structure for exon skipping, complete exon exclusion was detected with both F19 and F30 constructs (Fig. 3B, Lower, lanes 2 and 4), whereas almost complete (M39) or partial (L24) inclusion was observed with the L24 and M39 (Fig. 3B, Lower, lanes 3 and 5). In addition, we analyzed another four sets of more subtle mutations, made in the F19-based construct and also disrupting/reconstituting the stem. Although, in some cases, the effects observed were partial, the results obtained with all four sets were entirely consistent with the existence/importance of the stem–loop structure (Fig. S2). Together, these findings provide strong evidence that exonic RNA secondary structure can block exon inclusion.

We also investigated how proximity of the secondary structure element to the 3′ SS affects its inhibitory activity. For this, we used the F19 construct, which allowed straightforward insertion of increasing length spacers downstream of the 3′SS (Fig. 3C, Upper). If the secondary structure directly blocks recognition of the 3′SS, insertion of longer spacers would perhaps disrupt its inhibitory effect and increase inclusion of the internal exon. In agreement with this, whereas insertion of 2 nt had, at most, a minimal effect, 5- and 8-nt spacers significantly increased exon inclusion, to nearly 40% and 60%, respectively (Fig. 3C, Lower, lanes 3–4, compare with lane 1).

We next asked whether introduction of a comparable structure in a naturally included exon would be sufficient to induce skipping. To this end, mutations were made in the E9 construct to allow for formation of a stem just downstream of the 3′SS in the middle exon (Fig. 3D, Left). When RNA from transfected cells was analyzed by RT-PCR, ∼50% of the mutated exon was excluded, compared with the total inclusion observed with the wild-type E9 construct (Fig. 3D, Right). These results suggest that a 3′SS-proximal secondary structure is sufficient to induce exon skipping. However, although the secondary structure was almost identical to that in E10, exclusion was only partial, suggesting that other elements contribute to the complete exclusion observed with the E10 construct (see below).

Secondary Structure Contributes to but Is Insufficient for the Second-Step Block in Vitro.

We next wished to investigate whether any of the mutations that led to exon inclusion in vivo could also prevent the second-step block and allowed splicing in vitro. To this end, we analyzed several of these mutations in the context of the β-globin–based splicing substrates used in our initial in vitro assays (Fig. 1). For the β-F30 and β-E10M1/M2 RNAs, second-step splicing inhibition was observed, in both NE and S100 plus NF40-60 [Fig. 4A, lanes 1–2 and 9–10; NF40-60 is a nuclear fraction that contains sufficient SR proteins to allow splicing, albeit weak, of β-globin–based substrates such as β-E9 RNA (Fig. 1D, lanes 1 and 2, and ref. 21)]. This is consistent with the fact that these two mutants displayed exon exclusion in vivo. In contrast, when incubated in NE, efficient splicing was detected with β-M39, β-E10M1 and β-E10M2 substrates (Fig. 4A, lanes 4, 6, and 8), consistent with their ability to promote exon inclusion in vivo. Unexpectedly, however, when these three mutants were analyzed in NF40-60 plus S100, splicing remained blocked after the first step (Fig. 4A, lanes 3, 5, and 7). This inhibition suggests the presence of a second inhibitory element in exon 10, which can function independently of the secondary structure. It is possible, for example, that when secondary structure was disrupted, binding of an unknown factor to an ESS was sufficient to result in a second-step block in S100 plus NF40-60, perhaps reflecting limiting amounts of SR proteins (21) and/or a relative excess of the putative second-step repressor.

Fig. 4.

FUBP-1 is the major F30 RNA-binding protein. (A) In vitro-splicing assays with each of E10 derivative RNAs. Splicing was performed either in S100 plus NF40-60 or in NE alone with the indicated E10-derivative RNAs and analyzed as in Fig. 1. (B) ³²P-labeled 30-nt F30 RNA was incubated with NE followed by UV crosslinking and RNaseA treatment. Products were resolved on a 10% (vol/vol) polyacrylamide SDS gel. Position of molecular mass standard in kilodaltons is indicated at right. (C) Biotin-labeled 30-nt F30 RNA was incubated with NE, and bound proteins were resolved on a 10% (vol/vol) SDS/PAGE gel. Two major species were identified by MS as FUBP1 and hnRNPD (indicated by arrows).

FUBP1 Binds to a Potential ESS in E10.

To identify a possible ESS in E10, we first examined E10 and active derivatives for sequences that match previously described ESSs. We found a sequence, AUAUAUGAU, located downstream of the stem-forming region (Fig. 5A) that is similar to several ESS motifs identified by a SELEX approach, which likely constitute binding sites for inhibitory protein(s) (11). Next, we sought to identify nuclear proteins that might bind to the AU-rich sequence. To this end, we prepared an [α-³²P]UTP-labeled 30-nt RNA spanning the stem-forming region and AU-rich consensus of E10 (F30) and performed UV-crosslinking assays with HeLa NE. After RNase treatment and separation by SDS/PAGE, a prominent product of ∼70 kDa was detected (Fig. 4B). To identify the bound protein, we performed RNA-affinity chromatography using a 5′ biotin-labeled version of the RNA used in crosslinking. After SDS/PAGE and Coomassie staining, one of two specifically bound proteins closely matched the protein observed after UV crosslinking (Fig. 4C). The indicated protein was excised and identified by mass spectrometry (MS) as the FUBP1 (also known as FBP). FUBP1 is a largely nuclear protein that appears to possess both RNA- and DNA-binding activity. It has been shown, on the one hand, to regulate transcription of the MYC protooncogene (22, 27) and, on the other, to bind to AU-rich sequences in the 3′ UTRs of several transcripts (23, 28, 29). However, it has not been implicated previously in splicing. The other bound protein, of ∼35 kDa, was identified by MS as hnRNPD (Fig. 4C). Given that hnRNPD was not detected by UV crosslinking (Fig. 4B), and also that the other mammalian FBP family member (KH-type splicing regulatory protein, KSRP) (30) and the Drosophila homolog of the FBPs (P-element somatic inhibitor, PSI) have been previously linked to splicing regulation (31), we decided to investigate further the possible role of FUBP1 in splicing repression.

Fig. 5.

FUBP1 binding correlates with ESS activity. (A) Sequence of F30 RNA and mutant derivatives. The stem-forming region (5′-aucaguaugc-3′) and putative FUBP1-binding sites (5′-auauaugau-3′) are underlined. Mutations were made within the FUBP1-binding sites as indicated. Note that the stem-forming nucleotides were unchanged. (B) The indicated ³²P-labeled RNAs were incubated with increasing amounts of recombinant His-FUBP1 (100 and 300 ng), and complexes were resolved by nondenaturing PAGE. (C) Each of the indicated RNA probes was incubated with 150 ng of recombinant FUBP1, subjected to UV crosslinking, and analyzed as in Fig. 4B. (D) HeLa cells were transfected with constructs containing F30 or each of the indicated mutant derivatives. RNA was isolated and analyzed by RT-PCR as in Fig. 2B. (E) Splicing assays in NE with RNA substrates containing F30 or the indicated mutant derivatives were performed and analyzed as in Fig. 1.

Binding of FUBP1 Correlates with ESS Activity.

We next set out to examine in more detail the role of the AU-rich element in binding of FUBP1 and in splicing inhibition. To this end, we first analyzed the effects of a number of mutations in the AU sequence on FUBP1 binding to F30 RNA. As shown in Fig. 5A, one mutant RNA (F30M1) contained continuous replacement of AU with GC, the second mutant (F30M2) had four U-to-C transitions, and the third (F30M3) had four A-to-C substitutions. Fig. 5B shows the results of a gel-shift experiment in which increasing amounts of purified baculovirus-expressed His-tagged FUBP1 (Fig. S3) were incubated with F30 RNA and with each of three mutated derivatives. Whereas FUBP1 binding to F30 was readily detected, F30M1 and F30M2 RNAs were bound much less efficiently (Fig. 5B, lanes 4–6 and 7–9, compare with lanes 1–3). In contrast, the F30M3 mutant displayed only slightly weakened binding (Fig. 5B, lanes 10–12). We also estimated the K_d of FUBP1 for F30 RNA by gels shift, which suggested a value of ∼50 nM (Fig. S4). Although this is typical of other sequence-specific RNA-binding proteins (21), it is significantly higher than values reported for single-stranded DNA (32, 33). Whether this reflects differences in affinity of FUBP1 for DNA versus RNA, or that the F30 RNA site is not the optimal sequence, remains to be determined. To extend the above analysis, we examined the interaction of purified FUBP1 and RNA using UV-crosslinking assays. Results shown in Fig. 5C revealed a sequence preference for FUBP1 binding essentially identical to that observed in the gel-shift assays. Taken together, these data indicate that FUBP1 preferably binds to sequences highly enriched for U residues and that the AU-rich sequence is critical for efficient binding of FUBP1 to the F30 RNA.

We reasoned that if binding of FUBP1 to the AU sequence is responsible for the observed inhibitory effects on splicing, then the mutations that disrupt FUBP1 binding should activate splicing. To test this, we first introduced each of the three above mutations into the minigene construct containing F30 and tested the effects on exon skipping in vivo. Fully in line with the above RNA-binding results, the F30M1 and F30M2 mutations that disrupted FUBP1 binding strongly impaired exon skipping, leading to more than 90% F30 inclusion (Fig. 5D, compare lanes 2–3 with lane 1). In contrast, the F30M3 mutant displayed only a slight effect on exon skipping (lane 4), consistent with its weak effect on FUBP1 binding. We next introduced the same mutations into the β-F30–splicing substrate and tested their effects on in vitro splicing (Fig. 5E). Strikingly, and consistent with both the RNA-binding and in vivo-splicing results, efficient splicing was observed with the β-F30M1 and β-F30M2 RNAs when incubated in NE (Fig. 5E, lanes 3–4 and 5–6), whereas β-F30M3 splicing was still blocked at the second step (lanes 7–8). Taken together, the results provide strong evidence that the AUAUAUGAU consensus binds FUBP1 and functions as an ESS capable of blocking splicing at the second step in vitro and inducing exon skipping in vivo.

FUBP1 Inhibits Pre-mRNA Splicing at the Second Step.

We next wished to analyze directly the function of FUBP1 in splicing, to determine if it indeed contributes to the observed second-step inhibition. To this end, we first added increasing amounts of purified recombinant FUBP1 to splicing reactions in NE containing each of the E10-derived substrates. As shown in Fig. 6A, with two mutant substrates that retained the AU-rich element, β-E10M1 and β-M39, addition of increasing amounts of FUBP1 caused 50–80% inhibition of the second step, estimated from three independent experiments (lanes 1–3 and 4–6). However, only a very modest effect on splicing of the β-F30M2 RNA, which lacks an intact ESS, was observed (Fig. 6A, lanes 7–9). Furthermore, FUBP1 had no effect on the stalled first-step splicing reaction observed with the β-E10 RNA itself (Fig. 6A, lanes 10–12). Together, these results provide evidence that FUBP1 functions as a sequence-dependent second-step repressor.

Fig. 6.

FUBP1 inhibits pre-mRNA splicing at the second step in vitro and modulates exon skipping in vivo. (A) In vitro splicing was performed in NE in the presence of increasing amounts of recombinant FUBP1 (100 and 300 ng) with indicated RNAs. The presence or absence of the intact ESS in the RNAs is indicated above. (B) A biotinylated RNA (Bio-AU) 20-mer containing the E10 AU-rich element was used to deplete FUBP1 from NE. Mock depletion was done with beads alone. The resultant extracts were analyzed by Western blotting with anti-FUBP1 and anti-actin antibodies. (C) Extracts from B were used for splicing with β-F30 RNA (lanes 2 and 3), and standard NE was used for comparison (lane 1). FUBP1-depleted extracts were supplemented with increasing amount of FUBP1 (0, 100, and 300 ng) and analyzed for splicing (lanes 4–6). (D) In vitro splicing with the β-M18 RNA was performed essentially as above. The 18-nt sequence is shown at the bottom, with the FUBP1-binding site underlined. (E) Whole-cell lysates were prepared from FUBP1 siRNA- and control siRNA-transfected cells and analyzed by Western blotting using an anti-FUBP1 antibody. (F) The M18-containing minigene plasmid and each siRNA were cotransfected into HeLa cells, and total cellular RNA was analyzed by RT-PCR as above (top gel). Alternatively spliced isoforms of endogenous ACLY, Caspase9, ENAH, and PTBP2 RNAs from each of the FUBP1 siRNA- and control siRNA-transfected cells were analyzed by RT-PCR (bottom gels). Quantification of their RNA products was measured as inclusion/exclusion (In/Ex) ratio with SD and indicated below each gel. RNA products are indicated schematically on the right.

To provide additional evidence that FUBP1 functions in splicing repression of β-E10 RNA, we wished to deplete the protein from NE to determine whether such a depleted extract becomes competent for the second-step of splicing. Our initial attempts to do this by using anti-FUBP1 antibodies were unsuccessful, because we were unable to deplete the protein efficiently. We, therefore, decided to attempt to deplete FUBP1 by another approach, RNA affinity. The AU-rich sequence shown to bind to FUBP1 efficiently was used to produce a 30-nt biotinylated RNA containing three copies of the consensus motif (Bio-AU), and this was used to deplete FUBP1 from NE with streptavidin beads. RNA affinity reduced the levels of FUBP1 to less than 10% of the amount in mock-depleted extracts (Fig. 6B, compare lane 2 with lane 1), with no effect on KSRP protein levels, as demonstrated by Western blotting using KSRP-specific antibodies (Fig. S5A). We then compared the splicing activity of the mock- and Bio-AU–depleted extracts with untreated NE using the β-F30 substrate. Whereas the mock and untreated extracts were inactive for the second step, the Bio-AU–depleted extract indeed displayed activity, albeit weak, in facilitating the exon ligation step (Fig. 6C, compare lane 3 with lanes 1–2). Furthermore, addition of purified FUBP1 specifically inhibited appearance of fully spliced RNA (Fig. 6C, lanes 4–6). We reasoned that the weak effect observed could reflect the presence of the inhibitory secondary structure within the β-F30 RNA. To test this possibility, we constructed a substrate, β-M18, which contained 18 nt spanning the AUAUAUGAU consensus but lacked the stem-forming region (Fig. 6D, Lower). Indeed, this sequence was sufficient to induce the second-step block (Fig. 6D, lane 1). However, most importantly, the Bio-AU–depleted extract now displayed significant second-step activity (Fig. 6D, compare lane 3 with lanes 1–2), and addition of increasing amounts of FUBP1 again specifically restored second-step inhibition (Fig. 6D, lanes 4–6).

FUBP1 Causes Exon Skipping in Vivo.

We next wished to determine whether FUBP1 plays a role in exclusion of the E10 exon in vivo. To address this, siRNAs were used to reduce the levels of FUBP1 in HeLa cells. As shown in Fig. 6E, both #1 and #3 siRNA resulted in 70–80% decrease in FUBP1 accumulation compared with the control (lanes 2 and 4, compare with lane 1), whereas no effects on FUBP1 expression were observed with siRNA #2 (compare lane 3 with lane 1). KSRP protein levels were not affected by any of the siRNAs (Fig. S5B). To examine the effect of FUBP1 depletion independent of the E10 secondary structure, we prepared a minigene construct analogous to those used above, except containing the 18-nt M18 sequence in the middle exon. Cells transiently cotransfected with this construct and the control siRNA gave rise to predominantly exon-skipped RNA, as determined by RT-PCR, although exclusion was not complete, perhaps reflecting the absence of the secondary structure (Fig. 6F, lane1). Strikingly, knockdown of FUBP1 by either siRNA #1 or #3 significantly increased exon inclusion compared with the control or siRNA #2 (Fig. 6F, compare lanes 2 and 4 with lanes 1 and 3), indicating that FUBP1 indeed induces exon exclusion.

FUBP1 Regulates Alternative Splicing of Endogenous Pre-mRNAs.

We next investigated whether FUBP1 can regulate splicing of endogenous transcripts. To this end, we designed primer pairs that detect alternative exons in 10 transcripts regulated by other AU-rich binding proteins, such as HuR and TIA-1/TIAR (34–37), as well as a number of other potential target transcripts, chosen based on their cancer relevance, reflecting the emerging role of FUBP1 in cancer (see below), and used these to analyze RNA isolated from control and each of the FUBP1 siRNA-treated cells. Although there were no obvious changes in the splicing patterns of 51 transcripts examined, 4 revealed significant differences. Specifically, inclusion of exon 14 of the pre-mRNA encoding ATP citrate lyase (ACLY) and exons 4–7 of the caspase 9 pre-mRNA were increased upon FUBP1 knockdown (Fig. 6F, compare lanes 2 and 4 with lanes 1 and 3). On the other hand, increased exclusion of PTBP2 exon 10 and ENAH/MENA exon 11 were observed in the FUBP1-depleted cells (Fig. 6F, compare lanes 2 and 4 with lanes 1 and 3). Significantly, overexpression of Flag-tagged FUBP1 switched splicing in the opposite direction (Fig. S6). The effects of overexpression were relatively modest, likely reflecting the fact that FUBP1 is highly abundant in HeLa cells (∼3 × 10⁶ molecules per cell; Fig. S7). These data provide evidence that FUBP1 functions in control of AS of endogenous transcripts, and, as discussed below, can function both positively and negatively in regulating exon inclusion.

Discussion

In this paper, using a chimeric β-globin–triadin exon 10 RNA substrate (β-E10), we have shown that a previously described second-step inhibition results from two separate inhibitory elements. The first is a serendipitously created stem–loop structure downstream of the 3′SS, and the other is an ESS situated 3′ to the structure that depends on the RNA/DNA-binding protein FUBP1 (Fig. 7). Our data suggest that these two independent mechanisms operate simultaneously to block splicing at the second step, because disrupting either of them rescues the second-step defect (Fig. 7). We also demonstrated that these two elements, and FUBP1, contribute to exon skipping in vivo. Below, we discuss the features of FUBP1-regulated splicing compared with previously reported examples of splicing regulation, how this function of FUBP1 might be relevant to its role in disease and the function of RNA secondary structure in splicing regulation. It is striking that both of these mechanisms result in an unusual second-step inhibition, and we discuss the basis for this and how this type of inhibition may relate to exon exclusion in vivo.

Fig. 7.

Model for splicing regulation by FUBP1 binding to the ESS and by RNA secondary structure. (A) Schematic representation of the β-E10 pre-mRNA in which the intron can be processed only through the first catalytic step of splicing. The positions of 5′SS, branch point (BP), and 3′SS, the stem–loop downstream of the 3′SS, and the FUBP1-dependent (blue oval) ESS are shown. The green arrow represents first-step splicing. (B) RNA secondary structure and binding of FUBP1 to the ESS both block the second step, leading to accumulation of free exon 1, and exon 2-intron in a lariat configuration. The green arrow represents the second-step, and X indicates blockage of this step. These two elements contribute to exon skipping in vivo. (C) Disruption of the secondary structure can relieve the inhibitory effect caused by binding of FUBP1 to the ESS and thus activate the second step, leading to exon inclusion in vivo. (D) Disruption of the ESS or depletion of FUBP1 binding enhances the second step even in the presence of RNA secondary structure, resulting in partial exon inclusion in vivo.

Our data have shown that an AU-rich sequence in triadin exon 10 acts as an FUBP1-dependent ESS. Although FUBP1 was originally identified as a ssDNA-binding protein that modulates MYC expression (22, 38, 39), the protein was later demonstrated to be a member of the AU-rich element (ARE)-binding protein family (40) and has subsequently been found to be involved in several aspects of RNA metabolism other than splicing. For example, FUBP1 has been reported to interact with the poly(U) tract of the hepatitis C virus 3′ UTR and is required for its efficient replication (28). FUBP1 may also function in stabilization of certain ARE-containing transcripts (29, 40) and was recently shown to bind the 3′ UTR of nucleophosmin mRNA to repress translation (23) and to the 5′ UTR of CDKN1B mRNA to enhance translation (24). These findings suggest that FUBP1 can function in biological processes occurring in both the cytoplasm and the nucleus.

The involvement of FUBP1 in splicing regulation was unanticipated. However, several other ARE-binding proteins have previously been implicated in regulation of AS, in addition to their roles in RNA metabolism events occurring in the cytoplasm (34, 36, 41). HuR, well characterized for its role in posttranslational regulation of AU- and U-rich mRNAs (42), was recently found to enhance skipping of alternative exons in ZNF207 and PTBP2 pre-mRNAs, likely via interactions with intronic AU-rich elements (34). Likewise, TIA-1/TIAR has been shown to regulate translation of various mRNAs by binding to AU-rich elements located in the 3′ UTR. In the nucleus, these proteins act as splicing regulators of several alternatively spliced pre-mRNAs (36, 43, 44). Our data have demonstrated that FUBP1, like these ARE-binding proteins, functions as an AS regulator. Significantly, FUBP1 can either increase or decrease exon inclusion. This property likely depends on the location of its binding site within the pre-mRNA; such position-dependent activity has been shown previously for a number of splicing regulators (5, 7).

FUBP1, like HuR, regulates skipping of exon 10 of the pre-mRNA encoding the splicing regulator PTBP2 (also known as nPTB). It is intriguing although that HuR contributes to exon 10 exclusion, whereas FUBP1 facilitates inclusion. Skipping of this exon causes nonsense-mediated decay, and is thought to be used in the cross-regulation of expression of PTB itself (45). Our results suggest a possible interplay involving FUBP1 and these splicing regulators in AS control. Reduced levels of FUBP1 increase skipping of nPTB exon 10, resulting in decreased nPTB expression, which, in turn, up-regulates PTB expression. A switch in expression of nPTB/PTB has been proposed to be critical for cell-fate determination (46), and it will, thus, be of interest to determine how the cross-regulatory network involving these factors, including FUBP1, functions in development. For example, inclusion of triadin exon 10 varies during cardiac development in the mouse (20), and it will be important to determine whether FUBP1 functions in this process.

Deregulation of FUBP1 expression has recently been implicated in several cancers. For example, two groups observed elevated expression of FUBP1 in more than 70% of human hepatocellular carcinoma (HCC) samples and also provided evidence that FUBP1 plays an important role in tumor growth (47, 48). Notably, they also revealed no significant correlations between FUBP1 and MYC expression in HCC cells, suggesting that FUBP1-dependent regulation of MYC plays a minor role in hepatocarcinogenesis, although evidence for a role in renal carcinomas has been presented (49). In addition, FUBP1 target genes other than MYC such as those encoding the microtubule-destabilizing protein stathmin (48, 50) and the cell-cycle regulator CDKN1A (47), were found to be involved in tumor-relevant functions. More recently, mutations in FUBP1 were shown to contribute to another type of cancer, oligodendroglioma (51). All these findings indicate that multiple functions of FUBP1 may be critical for its role in tumorigenesis, and our data suggest that deregulation of cancer-related AS events, such as Caspase9 (52) and ENAH/MENA (53), by FUBP1 might be involved in this process. Aberrant AS events have been implicated in many types of cancer, including HCC and gliomas (54, 55). Thus, it will be important to determine whether alterations in these AS events caused by deregulated expression of FUBP1 contribute to tumor cell proliferation.

There are now a number of examples of RNA secondary structures that modulate AS. Such structures often function by blocking or interfering with recognition of the core splicing signals (16). In addition, several genome-wide studies investigating the potential of RNA secondary structure formation near such core elements supports the idea that these structures can indeed be involved in splicing regulation (56, 57). Our findings that an exonic stem–loop close to the 3′SS induced a splicing block in vitro and exon skipping in vivo provide further support for this theory. A unique feature of this regulation in our case is that the second step is specifically affected, as opposed to a more typical first-step inhibition (16).

Traditional models of splicing repression target early stages of splice-site recognition and/or early spliceosome assembly, and indeed many examples of this type of regulation, have been described (5, 6). However, several studies have shown that inhibition can also occur at later stages of spliceosome assembly and even during conformational changes between the two transesterification steps (13, 58, 59). This is further exemplified by the unique functions of FUBP1 as a second-step repressor described in our study. Similar to the role of Drosophila Sxl in splicing autoregulation (13), we speculate that binding of FUBP1 interferes with functions of protein factors required for the second step, thus preventing exon ligation by the splicing machinery. It is intriguing that second-step inhibition was also caused by secondary structure close to the 3′SS with the β-E10 RNA used here. It is possible that the structure might block a helicase activity required for the exon–exon ligation step, as was shown previously with a yeast ACT-LacZ fusion transcript with a structure similar to that of β-E10 RNA at the 3′SS (19). How both RNA secondary structure and the FUBP1-dependent ESS result in an unusual second-step block with the β-E10 RNA is currently unclear and will be an important focus of future studies.

In summary, the results presented here have provided additional insights into mechanisms of splicing control. FUBP1 represents a distinct type of splicing regulatory protein in that it binds to what appears to be a typical ESS but then inhibits splicing at the second step, which presumably underlies its ability to induce ESS-dependent exon skipping in vivo. A nearby exonic RNA structure was also shown to induce exon skipping in vivo and to block splicing at the second step in vitro. It will be of interest in the future to learn how widely these two mechanisms are used to regulate AS in human cells, as well as to investigate further the role of FUBP1 as a splicing regulator in development and disease.

Materials and Methods

Plasmids Constructions.

β-E9 and β-E10 plasmids were described previously (20). All of the β-E10 truncations were constructed by placing indicated sequences (F30, L24, M39, and M18) between AccI and BamHI sites to substitute for E10 sequences. Mutations on the β-E10 and β-F30 plasmids were created by site-directed mutagenesis, with specific sequences shown in the figures. Plasmids were linearized by BamHI or by AccI as indicated and used for making RNAs for in vitro-splicing assays. For transcribing RNAs for gel-shift and UV-crosslinking assays, corresponding sequences (F30, F30M1, F30M2, and F30M3) were inserted between HindIII and BamHI sites in the β-globin construct. In vivo minigene plasmids were constructed by subcloning modified β-globin genomic sequences into pcDNA3.1(+). Specifically, 4 bp of the first exon of β-globin adjacent to the 5′ splice site, the first intron and the second exon were inserted downstream of the β-E9 or β-E10 plasmid, generating a construct containing three exons with E9 or E10 in the middle exon flanked by two identical introns. Indicated sequences were inserted in the internal exon for splicing assays in vivo. The length of E10 was 54 nt; F30 represents the first 30 nt within E10, and L24 represents the last 24 nt of E10. M39 and M18, containing 39 and 18 nt, respectively, located in the middle region of E10. The underlined ATATATGAT consensus is the putative FUBP1-binding site: M39, 5′-GCATTCTGTCGATATATGATTGACATGTTTGTCCATGGG; M18, 5′- TCTGTCGATATATGATTG.

In Vitro Transcription and Splicing.

In vitro transcription and splicing assays in HeLa NE was performed essentially as described previously (20, 21).

RNA Gel-Shift and UV-Crosslinking Assay.

³²P-labeled F30 RNA and each of its derivatives were incubated with recombinant His-tagged human FUBP1 (100 and 300 ng) as described previously (4, 60).

RNA Affinity and FUBP1 Depletion.

For RNA-affinity assays, 1 nmol of 5′ biotin-labeled F30 RNA was mixed with 100 μL of HeLa NE in 500 μL of binding buffer [10 mM Hepes (pH7.9), 100 mM KCl, 10% glycerol, 2 mM MgCl₂, 0.75 mM ATP, 25 mM creatine phosphate, 30 μg/mL tRNA]. After incubation at 30 °C for 40 min, reaction mixtures were briefly centrifuged and 50 μL of streptavidin agarose (Sigma) was added to the supernatant. Reaction mixtures were then incubated at 4 °C on a rotating wheel for 1 h. Proteins were eluted from the beads with SDS loading buffer, resolved by SDS/PAGE, and stained with Coomassie blue. Protein bands were excised and analyzed by MS as described previously (61).

FUBP1 depletion was conducted using an RNA-affinity assay. A 5′ biotin-labeled RNA (Bio-AU) was synthesized containing three copies of the putative FUBP1-binding sequence (5′-AUAUAUGAU); 5 nmol of Bio-AU RNA bound to 60 μL of streptavidin agarose beads was pelleted, and 200 μL of NE was added and rotated at 4 °C overnight. Mock depletion was done with beads alone. Depletion efficiency was determined by Western blot.

Splicing Extracts and Recombinant Proteins.

NE, S100, NF40-60, His-SRp38, SRSF1, and SRSF2 were prepared essentially as described (20, 21). Human FUBP1 cDNA was first cloned into pQE80L vector (Qiagen) containing a His₆ tag sequence and then subcloned into pFastBac^TM HT B (Invitrogen) for His-FUBP1-baculovirus production. Sf9 and High-Five cells were used for baculovirus production and His-FUBP1 expression, respectively. His-FUBP1 was purified by Ni²⁺ agarose under native conditions and dialyzed against buffer D.

Cell Culture, Transfection, RT-PCR, and Western Blot.

HeLa cells were cultured as adherent cells in Dulbecco’s modified Eagle medium with 10% FBS. Minigene constructs were transfected by using lipofectamine (Invitrogen) following the manufacturer’s instructions. RNA extraction and reverse transcription were all carried out as described previously (20). For PCR analysis of minigene splicing, the forward (F) primer on exon 1 was 5′-ACTTAAGCTTGCTTACATTTGC, and the reverse (R) primer on exon 3 was 5′-ACTCAAAGAACCTCTGGGTC. Sequences for primer sets used in the study were as follows: ACLY-F, 5′-CAAACTTCCTCCTCAACGC; ACLY-R, 5′-GAGGGTGGTGCTCTTTCC; Caspase9-F, 5′-GACCAGTGGACATTGGTTC; Caspase9-R, 5′-GGTCCCTCCAGGAAACAA; ENAH-F, 5′-TGCTGGCCAGGAGGAGAAG; ENAH-R, 5′-ACTGGGCTGTGATAAGGGTG; PTBP2-F, 5′-GGCAATACAGTCCTGTTGGT; and PTBP2-R, 5′-ATGGCAAGTTGTGATTGGTT.

Sequences for other primer pairs will be provided upon request. Western blot was carried out as described previously (20). Primary antibodies used, mouse anti-FUBP1 (catalog no. sc-136137) and mouse anti-actin, were purchased from Santa Cruz Biotechnology. The monoclonal anti-KSRP antibody (62) was a kind gift from D. Black (University of California, Los Angeles, CA).

RNA Interference.

siRNA sequences were designed using Clontech RNAi Target Sequence Selector. All siRNAs were synthesized at GenePharma, and the following sequences were sense strands of chosen siRNA duplexes: siRNA-FUBP1 (#1), GGAGGAGUUAACGACGCUUTT; siRNA-FUBP1 (#2), GCAGCAAAGCAGAUCUGUATT; siRNA-FUBP1 (#3), CUGGAACACCUGAAUCUGUTT; and siRNA control, UUCUCCGAACGUGUCACGUTT. siRNA duplexes were transfected at 120 pmol per well (12-well plate) by using Lipofectamine 2000 (Invitrogen) following manufacturer’s instructions. Transfected cells were harvested for RNA isolation or protein extraction 48 h after siRNA transfection. siRNA and minigene coupled transfection assay was carried out as described previously (63). In brief, siRNA transfection (120 pmol per well; 12-well plate) was performed when HeLa cells reached 30–40% confluence. Minigene plasmids (0.6 μg per well; 12-well plate) were transfected 24 h after siRNA transfection. After another 24 h, cells were collected for further analysis.

RNA Secondary Structure Prediction.

RNA secondary structure prediction was conducted using the Vienna RNAfold Web server (26).