Alternative splicing (AS) is tightly coupled to transcription for the majority of human genes. However, how these two processes are linked is not well understood. Here, we unveil a direct role for the transcription factor FBI-1 in the regulation of AS. FBI-1 interacts with the splicing factor SAM68 and reduces its binding to BCL-X mRNA. This, in turn, results in the selection of the proximal 5′ splice site in BCL-X exon 2, thereby favoring the anti-apoptotic BCL-XL variant and counteracting SAM68-mediated apoptosis. Conversely, depletion of FBI-1, or expression of a SAM68 mutant lacking the FBI-1 binding region, restores the ability of SAM68 to induce BCL-XS splicing and apoptosis. FBI-1's role in splicing requires the activity of histone deacetylases, whose pharmacological inhibition recapitulates the effects of FBI-1 knockdown. Our study reveals an unexpected function for FBI-1 in splicing modulation with a direct impact on cell survival.
The oncogenic transcription factor FBI-1 regulates alternative splicing by preventing the splicing regulator SAM68 from binding to BCL-X mRNA. This increases anti-apoptotic BCL-X isoform expression and cell survival.
The transcription factor FBI-1 directly interacts with the splicing regulator SAM68.
FBI-1 impairs binding of SAM68 to BCL-X mRNA and thus promotes production of the anti-apoptotic BCL-X isoform and cell survival.
FBI-1's effect on BCL-X splicing requires histone deacetylase activity.
Alternative splicing (AS) amplifies genome complexity in eukaryotes by allowing different combinations of exons in the pre-mRNAs transcribed from each gene. The large spectrum of protein variants produced by AS greatly contributes to fine-tune many complex biological processes . Splicing of pre-mRNAs is operated by a ribonucleoprotein complex named spliceosome . Due to the lack of a strict consensus signature at the exon–intron junctions in the majority of eukaryotic genes, spliceosome recruitment requires the assistance of RNA-binding proteins (RBPs) that recognize specific elements in the vicinity of the splice sites (ss) . Thus, the relative balance of the activity or expression of these splicing factors modulates AS in each given cell and determines changes in mRNA variants that affect several biological processes, including cell proliferation, survival, and differentiation .
Programmed cell death, or apoptosis, participates to cell homeostasis in development and disease and represents a typical example of biological process controlled by AS . The majority of apoptotic genes produce multiple splice variants that often function antagonistically in the cell . BCL-X represents a striking example of apoptotic protein whose function is tightly regulated by AS. Alternative usage of two 5′ ss in exon 2 of the BCL-X gene yields splice variants that play antagonistic effects on cell survival. Selection of the proximal 5′ ss at the end of the exon leads to expression of a long isoform (BCL-XL) with anti-apoptotic function, whereas selection of the distal 5′ ss within the exon yields a short variant that promotes cell death (BCL-XS) . Thus, cells can respond to external cues and opt for survival or death by fine-tuning the regulation of BCL-X splicing. Importantly, cancer cells frequently overexpress BCL-XL, and treatments with antisense oligonucleotides that promote splicing of the pro-apoptotic BCL-XS variant sensitize them to chemotherapy, suggesting that modulation of BCL-X splicing has potential therapeutic value .
A splicing regulator involved in BCL-X AS and apoptosis is SAM68 , a member of the evolutionary conserved signal transduction and activator of RNA (STAR) family of RBPs, which regulate cell proliferation, survival, and differentiation in various organisms . In prostate cancer (PCa), SAM68 supports cell proliferation and survival . Intriguingly, whereas up-regulation of SAM68 generally causes apoptosis by promoting BCL-XS splicing , its depletion in PCa cells negatively affects the expression of the anti-apoptotic BCL-XL isoform . These observations suggest that the splicing activity of SAM68 is modulated by cell-specific factors.
Herein, we identified the transcription factor FBI-1 as a novel regulator of the splicing activity of SAM68 by two-hybrid screen. FBI-1 is an essential protein  that is overexpressed in human cancers , like SAM68 . We show that FBI-1 physically interacts with SAM68 and affects its interaction with the BCL-X mRNA, thereby promoting splicing of the anti-apoptotic BCL-XL variant and cell survival. These findings identify an unpredicted, transcription-independent function for FBI-1 in splicing regulation through the interaction with a splicing factor.
Results and Discussion
The transcription factor FBI-1 is a novel SAM68-interacting protein
Earlier observations indicated that over-expression of SAM68 in non-transformed cells caused G1 arrest and triggered apoptosis . Notably, SAM68-dependent apoptosis required its splicing activity in HEK293T cells . However, in cancer cells, the pro-apoptotic activity of SAM68 is substantially repressed . Thus, we hypothesized that interaction with specific cofactors could modulate the splicing activity of SAM68 in cancer cells. To search for such functional partners, we carried out a yeast two-hybrid screen using a library amplified from the PCa cell line LNCaP. Among several novel SAM68-interacting proteins, we focused our attention on FBI-1, a transcription factor previously described as a proto-oncogene in mouse and humans . Since FBI-1 was shown to regulate transcription of genes involved in cell proliferation and survival , we first verified that expression of FBI-1 alone did not activate reporter genes. Yeast cells were co-transformed with the plasmid recovered from the original clone (# 16) and either pGBKT7 alone or pGBKT7-SAM68. Eight independent transformants confirmed that the interaction between SAM68 and FBI-1 was required to support yeast growth under high-stringency conditions (Fig 1A).
FBI-1 contains an N-terminal POZ/BTB domain and a C-terminal region characterized by four Kruppel-like zinc finger motifs (Fig 1B). The clone retrieved from our screen encoded the C-terminal 306 amino acids (aa 278–584). In line with the result of the two-hybrid assay, GST pull-down assays showed that Myc-SAM68 interacts with full-length GST-FBI-1 and with GST-FBI-1CT (aa 274–584), but not with the POZ/BTB domain in GST-FBI-1NT (aa 1–276) nor with GST alone (Fig 1B). This interaction was also confirmed by using the CheckMate™ Mammalian System reporter (Promega Corporation, Madison, WI, USA) assay in live HEK293T cells (Supplementary Fig S1A). Moreover, we found that the endogenous FBI-1 and SAM68 proteins were mainly localized in the nucleus of PCa cells (Fig 1C), co-sedimented in sucrose gradient fractionations of nuclear extracts (Supplementary Fig S1B and C), and were co-immunoprecipitated from LNCaP cell extracts (Fig 1D). Lastly, endogenous SAM68 associated with Flag-FBI-1 in transfected HEK293T cells (Supplementary Fig S1D). Collectively, these results demonstrate that FBI-1 interacts with SAM68.
FBI-1 represses the ability of SAM68 to induce apoptosis
To test whether FBI-1 modulates the function of SAM68, we co-expressed the two proteins in HEK293T cells. As expected , expression of GFP-SAM68 augmented the number of apoptotic cells, but co-expression of Flag-FBI-1 completely suppressed this effect (Fig 2A and B). To investigate whether physical interaction between FBI-1 and SAM68 is required for this inhibition, we set out to identify the minimal region of SAM68 involved in the association. SAM68 is characterized by a GSG domain required for RNA binding and homo-dimerization, flanked by regulatory regions that are sites of protein–protein interactions and post-translational modifications affecting SAM68 localization and function (Supplementary Fig S2A) . GST pull-down assays showed that FBI-1 interacted with the full-length SAM68 and with its C-terminal region (aa 321–443), but not with the N-terminal region, including the GSG domain (aa 1–277) (Supplementary Fig S2B). Progressive deletions of the C-terminal region from both ends restricted the surface of interaction to 35 residues in SAM68 (aa 400–434) (Supplementary Fig S2B). Since the nuclear localization signal (NLS) in SAM68 partially overlaps with this region , we constructed a deletion mutant lacking residues 400–420 to avoid its mislocalization. SAM68Δ400–420 localized in the nucleus and interacted with poly-U and poly-A synthetic RNA similarly to full-length SAM68 (Supplementary Fig S2C and D), indicating that it is a functional protein. The two-hybrid CheckMate™ Mammalian System and pull-down assays showed that the interaction of SAM68Δ400–420 with FBI-1 is strongly reduced (Supplementary Fig S2E and F). Notably, while Sam68Δ400–420 triggered apoptosis like the full-length protein (Fig 2B), co-expression of FBI-1 did not suppress this activity (Fig 2B), indicating that FBI-1 inhibits SAM68-dependent apoptosis through physical interaction with it.
Down-regulation of FBI-1 in prostate cancer cells restores SAM68-mediated apoptosis
We reasoned that depletion of FBI-1 should enhance SAM68-dependent apoptosis. Indeed, expression of suboptimal amounts of SAM68 did not trigger cell death in control cells (si-CTRL), but knockdown of FBI-1 restored SAM68-dependent apoptosis (si-FBI-1 in Fig 2C). In line with the requirement for physical interaction, Sam68Δ400–420 induced apoptosis at this lower dose also in control cells and depletion of FBI-1 did not affect its activity (Fig 2C).
Previous reports documented that FBI-1 and SAM68 are up-regulated in various cancer cells . Consistently, we found that both proteins were expressed at higher levels in PCa cells (LNCaP and PC3) with respect to benign prostate epithelial cells (BPH1) (Fig 2D). LNCaP cells derive from androgen-sensitive PCa lymph node metastases, whereas PC3 cells derive from androgen-insensitive PCa bone metastases and represent cells at advanced stage of the disease. Noteworthy, dose-dependent overexpression of SAM68 in PC3 cells did not trigger apoptosis (si-CTRL in Fig 2E), whereas a small effect was observed in LNCaP cells (si-CTRL in Supplementary Fig S3). Strikingly, depletion of endogenous FBI-1 restored the pro-apoptotic activity of SAM68 in PC3 cells (Fig 2E) and strongly enhanced its effect in LNCaP cells (Supplementary Fig S3). These results indicate that high levels of FBI-1 are required to mask the pro-apoptotic activity of SAM68 in PCa cells.
FBI-1 modulates BCL-X splicing through direct physical interaction with SAM68
The ability of SAM68 to induce apoptosis correlates with its effect on BCL-X AS . The BCL-X gene encodes two splice variants that play antagonistic effects on cell survival (Fig 3A): the anti-apoptotic BCL-XL or the pro-apoptotic BCL-XS proteins . Moderate knockdown of FBI-1 in HEK293T promoted the expression of the pro-apoptotic BCL-XS variant, which was not detectable in control cells by conventional PCR (Fig 3A). Quantitative real-time PCR (qPCR) using exon junction-specific primers confirmed that knockdown of FBI-1 enhanced splicing of BCL-XS in HEK293T and PC3 cells (Fig 3B). This result suggests that splicing of the endogenous BCL-X pre-mRNA is modulated by FBI-1 in live cells.
To gain more insight into the regulation of BCL-X AS by FBI-1, we employed a minigene that recapitulates the splicing of the endogenous gene . Up-regulation of FBI-1, or FBI-1CT, favored splicing of the anti-apoptotic BCL-XL variant (Fig 3C). This effect required SAM68 expression, as its down-regulation by RNAi in HEK293T (Fig 3D), or by stable transfection of a different siRNA in PC3 cells (Supplementary Fig S4A), reduced or abolished the ability of FBI-1 to promote BCL-XL splicing. Moreover, FBI-1 was unable to modulate BCL-X splicing when the splicing-defective SAM68V229F  was co-transfected in HEK293T cells (Supplementary Fig S4B), indicating that the splicing activity of SAM68 was necessary.
We then asked whether a functional interaction with SAM68 was required for the effect of FBI-1 on BCL-X splicing. As previously reported , SAM68 induced splicing of the pro-apoptotic BCL-XS when it was expressed alone (Supplementary Fig S4C, lane 3). However, this effect was suppressed by co-expression of FBI-1 or its C-terminal domain (Supplementary Fig S4C, lane 4 and 6). Importantly, SAM68Δ400–420 strongly enhanced splicing of the BCL-XS variant, like full-length SAM68, but co-expression of FBI-1 did not affect its splicing activity (Supplementary Fig S4D), proving that physical interaction with SAM68 is required for the effect of FBI-1 on BCL-X AS.
The pro-apoptotic activity of SAM68 was suppressed by FBI-1 in PC3 cells (Fig 2E). Consistently, overexpression of SAM68 in PC3 cells only slightly affected BCL-X AS, whereas it strongly enhanced splicing of the pro-apoptotic BCL-XS in the benign BPH1 cells (Fig 3E), which express low levels of FBI-1 (Fig 2D) similarly to HEK293T cells (data not shown). On the other hand, knockdown of endogenous FBI-1 significantly restored SAM68-dependent splicing of BCL-XS in PC3 cells (Fig 3F) and strongly enhanced this splicing event in HEK293T cells (Supplementary Fig S4E). These results suggest that elevated expression of FBI-1 promotes cell survival through direct interaction with SAM68 and modulation of BCL-X splicing.
BCL-X splicing is involved in FBI-1-dependent suppression of apoptosis
Next, we further evaluated whether the pro-apoptotic effect mediated by FBI-1 depletion was due to the splicing activity of SAM68. Expression of the RNA binding-defective SAM68V229F mutant exerts dominant-negative effect on the splicing activity of endogenous SAM68 , while retaining the transcriptional activity of the protein . We infected LNCaP cells, whose apoptosis was strongly induced upon FBI-1 depletion (GFP samples in Fig 4A), with a retrovirus encoding GFP-SAM68V229F, or GFP as a control. Expression of the dominant-negative GFP-SAM68V229F almost completely abolished apoptosis caused by FBI-1 knockdown in GFP-infected cells (Fig 4A), suggesting that FBI-1 counteracts SAM68-mediated apoptosis by modulating its splicing activity toward BCL-X. Accordingly, overexpression of the anti-apoptotic BCL-XL also blocked cell death induced by SAM68 overexpression in PC3 cells depleted of FBI-1 (Fig 4B). Thus, in our conditions, the anti-apoptotic effect of FBI-1 correlates with SAM68-dependent regulation of BCL-X splicing. Since FBI-1 was shown to promote cell survival in mouse erythroblasts by directly repressing transcription of the pro-apoptotic BIM gene products , it is likely that this protein can promote cell survival through at least two distinct mechanisms: direct repression of transcription of target genes or indirect modulation of their AS through SAM68.
FBI-1 modulates SAM68 binding to BCL-X mRNA through histone deacetylases
FBI-1 interacts with residues 400–420 in the C-terminal region of SAM68, which was previously shown to bind other proteins involved in splicing . Moreover, FYN-dependent tyrosine phosphorylation of this region modulates SAM68 homodimerization and protein–protein interactions . However, we found that overexpression of FBI-1 did not perturb the ability of SAM68 to homodimerize and to interact with hnRNP A1 (Supplementary Fig S5A). Furthermore, FYN-dependent phosphorylation did not impair the interaction of SAM68 with FBI-1 (Supplementary Fig S5B). Thus, FBI-1 does not appear to modulate the splicing activity of SAM68 by interfering with its interaction with cofactors.
To investigate the mechanism by which FBI-1 modulates SAM68-dependent AS, we first analyzed the binding of SAM68 to BCL-X RNA. Cross-link immunoprecipitation (CLIP) experiments showed that SAM68 binds BCL-X in exon 2, while no binding was observed in the distal part of intron 2 (Supplementary Fig S5C). Notably, overexpression of FBI-1 strongly reduced binding of endogenous SAM68 or Myc-SAM68 to BCL-X mRNA (Fig 5A). However, FBI-1 did not generally reduce the affinity of SAM68 for RNA, as determined by binding to synthetic poly-U RNA in vitro (Supplementary Fig S5D).
Since FBI-1 is known to associate with histone deacetylase (HDAC) 1, 2 and 3  and HDAC inhibitors affect a wide spectrum of AS events , we asked whether HDAC activity influences BCL-X splicing. First, we confirmed the interaction of FBI-1 with HDAC1 by co-immunoprecipitation (Fig 5B). Next, we tested whether HDAC activity influences splicing of the endogenous BCL-X gene by using the pan-HDAC inhibitor LBH589 . Treatment of HEK293T cells with LBH589 mimicked the effect of FBI-1 knockdown on AS of BCL-X mRNA, causing a dose-dependent accumulation of the BCL-XS variant (Fig 5C). To directly test whether HDAC activity is implicated in the effect of FBI-1 on BCL-X AS, we co-transfected FBI-1 and the BCL-X minigene in the presence or absence of LBH589. Notably, the increase in BCL-XL splicing elicited by FBI-1 was completely abolished when cellular HDAC activity was suppressed (Fig 5D). Moreover, the effect on splicing could be explained with modulation of SAM68 recruitment on the BCL-X mRNA. In fact, LBH589 enhanced binding of SAM68 to the BCL-X transcript and completely suppressed the ability of FBI-1 to inhibit its recruitment on this target mRNA (Fig 5E). Collectively, these results strongly suggest that FBI-1 affects recruitment of SAM68 on the BCL-X transcript through an HDAC-dependent mechanism, thereby modulating BCL-X AS and apoptosis in live cells.
Our studies suggest that FBI-1 modulates BCL-X splicing by directly interacting with SAM68 and inhibiting its recruitment on the BCL-X mRNA. Thus, FBI-1 appears to function by lowering the abundance on the pre-mRNA of a splicing factor that promotes BCL-XS splicing. Mechanistically, several observations point to an involvement of HDACs in FBI-1-mediated BCL-X splicing, as we found that an increase in the BCL-XS variant can be similarly obtained either by lowering the levels of FBI-1 or by blocking the activity of HDACs, which form a complex with this transcription factor . These results, together with the notion that FBI-1 is a transcriptional regulator, suggest that the effect of FBI-1 on SAM68-dependent BCL-X splicing occurs co-transcriptionally. Whether HDACs influence FBI-1-mediated AS by altering chromatin structure or by affecting the SAM68/RNA polymerase II complex  (or both) remains to be elucidated.
In conclusion, our results highlight a new, unexpected function of FBI-1 in splicing regulation and unveil a new layer in the regulation of gene expression by this essential transcription factor through the physical interaction with a splicing factor.
Materials and Methods
Yeast two-hybrid screen
The yeast two-hybrid screen was performed according to the manufacturer's instruction (Clontech). Briefly, the Gal4 activation domain-fusion library was generated by using 1 μg polyA+ RNA isolated from LNCaP cells. The ds-cDNA was generated by SMART technology using a modified oligo(dT) primer (CDSIII primer) and cloned in pGADT7-rec vector (see also Supplementary information).
Cell cultures, transfections, and cell extract preparation
Cell cultures, transfections, and sample preparation were carried out by standard methods as described  (see also Supplementary information).
Co-immunoprecipitation and GST pull-down assays
GST pull-down assays and co-immunoprecipitation (co-IP) experiments were carried out in presence of RNase and DNase using either total cell extracts or nuclear extracts from LNCaP or HEK293T cells transfected with the indicated plasmids as previously described  (see also Supplementary information).
Cells were washed with PBS, UV-irradiated (400 mJ/cm2), and used for CLIP experiments as described  (see also Supplementary information).
Statistical significance was calculated by Student's t-test on at least three independent experiments (*P < 0.05; **P < 0.01; ***P < 0.001; n.s., not significant).
The authors wish to thank Drs D. Del Bufalo, M.P. Paronetto, D. Farini, S. Pedrotti, M. Cappellari, and C. Verri for generation of reagents used in this study and for helpful suggestions. This work was supported by the Association for International Cancer Research (AICR), the Associazione Italiana Ricerca sul Cancro (AIRC) and the Ministry of Health Ricerca Corrente to the Fondazione Santa Lucia.
PB designed the study, performed most of the experiments, and wrote the manuscript; RB and FB performed immunofluorescence analyses; MJM, ARK, and SMDS provided critical reagents and contributed to writing the manuscript; CS designed the study and wrote the manuscript.
Conflict of interest
The authors declare that they have no conflict of interest.