Role of the splicing factor SRSF4 in cisplatin-induced modifications of pre-mRNA splicing and apoptosis

MCF7, MDA-MB-231, HT1080, BT549, RD, HDF1 and HDF2, MG-63, MSU and AT5BIVA (deficient in ATM, Coriell Cell Repository, Camden, NJ, USA) cells were cultured in Dulbecco’s Modified Eagle’s Medium (DMEM, Lonza, Verviers, Belgium) supplemented with non-essential amino-acids (NEAA) (1%), penicillin and streptomycine (1%), gentamycin (0.1%), fungizone (0.1%) and 10% FCS (Lonza). Ishikawa cells (human endometrial adenocarcinoma cell line) were cultured in RPMI 1640-glutamax (Lonza) supplemented with NEAA (1%), sodium pyruvate (1%), penicillin and streptomycine (1%), fungizone (0.1%) and 10% FCS, GM09607 cells (deficient in ATM, Coriell Cell Repository) in EMEM (Lonza) supplemented with 10% FCS and 1% NEAA, and MO59J cells (glioblastoma cell line, deficient in the catalytic subunit of DNA-PK) in DMEM/F12 supplemented as DMEM.

The study conforms to the principles outlined in the Declaration of Helsinki and was approved by the ethic committee of Liège University Hospital (B707201110973).

MCF7 and Ishikawa cells were authenticated by DSMZ (Braunschweig, Germany). Although no authentication of the other cell lines was made, the deficiency in ATM of GM09607 and AT5BIVA was ascertained by western blotting, and that of p53 in MG-63 was confirmed by RT-PCR.

Cisplatin (cis-diamine platinum (II) dichloride), wortmannin, caffein, and triciribine were from Sigma-Aldrich (St-Louis, MO, USA), oxaliplatin from Santa Cruz Biotechnology (Santa Cruz, CA, USA), ATM kinase inhibitor from Calbiochem EMD biosciences (La Jolla, CA, USA), NU7026 from Merck Millipore (Darmstadt, Germany), TGX221, IC87114 and MK2206 from Selleckchem (Munich, Germany) and PX866 from LC Laboratories (Woburn, MA, USA).

Cell survival and apoptosis/necrosis were measured, respectively, by trypan blue exclusion in blind tests and by FACS analysis as described in [17].

RNAs were purified from cultured cells using the High Pure RNA isolation kit (Roche, Mannheim, Germany) and quantified by spectrometry. Gene expression was measured by RT-qPCR. Details according to the Minimum Information for Quantitative RT-PCR Experiment (MIQE) guidelines [18], are given in Additional files 1 and 2. For analysis of exon inclusion/exclusion, primers were chosen on exons surrounding the sequences potentially alternatively spliced. Primers, protocols and amplification products sizes are detailed in Additional file 2. Splice variants were discriminated by electrophoresis as described [17].

RNA libraries and sequencing were performed on total RNA samples at the GIGA Genomics facility, University of Liège, Belgium. The quality of RNA was checked with BioAnalyser 2100 (Agilent technologies, CA, USA) that indicated a RQI score >8. The libraries were prepared with Truseq® mRNA Sample Prep kit (Illumina, CA, USA) from 1 microgram of total RNA following manufacturer’s instructions. mRNAs were isolated by poly-A selection and fragmented (8 minutes at 94°C). Fragmented mRNAs (around 170 nucleotide-long in average) were used for reverse-transcription in the presence of Superscript II (Invitrogen, Oregon, USA) and random primers. After second strand synthesis, end-repair, A-tailing and purification, the double strand cDNA fragments were ligated to Truseq® adapters containing the index sequences. Fifteen cycles of PCR in the presence of dedicated PCR primers and PCR master mix were applied to generate the final libraries. Libraries were sequenced in pair-end sequencing runs on the Illumina GAIIx in multiplexed 2 × 76 base protocols. The raw data was generated through CASAVA 1.6 suite (Illumina, CA, USA). TopHat (http://​ccb.​jhu.​edu/​software/​tophat/​index.​shtml) software was used to align RNA-Seq reads to the reference genome (hg19, UCSA) and discover transcript splice sites. Cufflinks (http://​cole-trapnelllab.​github.​io/​cufflinks/​) used the resulting alignment files to quantify the gene expression levels, identify up- and down-regulated transcripts and find the alternative splice junctions.

SpliceSeq(1.2) (http://​bioinformatics.​mdanderson.​org/​main/​SpliceSeq:​Overview) was used for a focused AS analysis. Using alignment database and Bowtie, SpliceSeq aligns reads from RNA-Seq data to a reference collection of splice variants [19,20].

Lists of genes modulated in term of expression and splicing were imported in the ToppGene Suite for analysis [21].

Antibodies directed against Akt, phospho-Akt (ser473) and β-actin were purchased from Cell Signalling (Beverly, MA, USA). Cells were lysed in Laemmli buffer containing 50 mM DTT. Lysates were briefly sonicated, incubated at 65°C for 15 min and analyzed by SDS-PAGE. Proteins were electroblotted and detected as described in [17]. Probing of β-actin was performed as a control of protein loading.

SMARTpool siGENOME (Dharmacon by Thermo Fisher Scientific, Lafayette, CO, USA), consisting of four siRNA duplexes, were used to target SRSF4 and SRSF6 mRNA. siRNA targeting ATR were from Ambion (Life technologies). The 5′-UUGCAUACAGGACUCGUUATT-3′ and 5′-UAACGAGUCCUGUAUGCAATT-3′ oligoribonucleotides were used as control siRNA (siSCR) that does not target any known human transcript [22]. Cells were transfected by siRNAs as previously described [23].

The means and standard deviation were calculated from three or four independent experiments. The significance of differences was determined using t-test or ratio paired t-test of Student.

In vitro treatment of cells by cisplatin induces alterations of splicing in various transcripts [15,24,25]. Following treatment of MCF7 and Ishikawa cell lines with cisplatin (Figure 1A-B), the RT-PCR analysis of MDM2, a negative regulator of p53, showed a reduction in the full length product and the appearance of smaller splicing variants. The smallest variant had the expected size of MDM2-ALT1 splice variant. This splicing shift was maximal at 50 – 100 μM of cisplatin (Figure 1A-B) and after 24 – 48 hours (Figure 1E). A similar dose-dependent shift was obtained with VEGF where cisplatin decreased the expression of VEGF-165 and concomitantly increased the production of the VEGF-111 splice variant (Figure 1C-D). Similar alterations of splicing of MDM2 were also observed in MDA-MB-231 (breast adenosarcoma), BT549 (breast carcinoma), HT1080 (fibrosarcoma), RD (rhabdomyosarcoma), MG-63 (osteosarcoma), MSU (fibrobastic cell line) and HDF1 and 2 (primary dermal fibroblasts) cells treated with 50 μM cisplatin for 24 hours (Figure 1F). The cisplatin analog oxaliplatin induced similar effects on MDM2 splicing, suggesting that this splicing alteration is generalized to platinum-based agents (Figure 1G).

Poly A+ RNA from MCF7 cells untreated or treated with 50 μM cisplatin for 24 hours was isolated and prepared for next-generation sequencing analysis. No significant cell death over untreated samples was noted in these conditions (as measured by trypan blue exclusion). The average number of reads approached or exceeded 20 millions in both samples. Alignment of transcripts to the genome indicated that 16733 and 16969 genes were expressed in the control and cisplatin-treated samples, respectively. Sequencing data are available in the ArrayExpress database (www.​ebi.​ac.​uk/​arrayexpress) under accession number E-MTAB-2663. The global gene expression in the two conditions was highly correlated, with a Pearson correlation coefficient = 0.835 (p = 0.000000).

For differential gene expression, the following filters were applied: absolute fold change >2 and q-value < 0.05. Five hundred fifty-three genes were regulated (111 up and 442 down, Additional file 3). The top 20 up-regulated and down-regulated genes are listed in Table 1 with SERPINB5 (126×) and GPHN (159×) being first in each category, respectively. The expression of a panel of genes commonly used as calibrators was not significantly affected (GAPDH: 1.20; ACTB: 0.79; ACTG: 0.93; PPIA (cyclophylin A): 0.94; PPIB (cyclophylin B): 1.07). RT-qPCR was performed to confirm the expression level of 9 up-, down- or non-regulated genes on the samples used for RNA-Seq. Selected genes were either conventional calibrators (GAPDH, ACTB, β2M) or encoded proto-oncogenes (MYB, SERPINB5, JAK2), anti-oncogenes (BRCA1, RB1) and a factor regulating apoptosis (FAS). RT-qPCR analysis correlated with RNA-seq data with a Pearson coefficient of 0.98 (p = 0.000004), validating the RNA-seq data. Moreover, these changes in gene expression noted by RT-qPCR were confirmed in three independent experiments using MCF7 cells, and in two independent experiments in Ishikawa cells, indicating that the changes are reproducible and not restricted to MCF7 cells.

Gene ontology analysis was performed using the ToppFun Suite software. Significantly (p < 0.05) affected biological processes were identified (Table 2). Many genes regulated by cisplatin belong to two main groups: cell cycle and proliferation. Surprisingly, neurogenesis also appeared as a regulated category.

ToppFun identified genes matching annotations for transcription factors PITX2 (38 genes), E2F (18 genes, Table 3) and FOXF2 (19 genes). For genes matching with PITX2 and FOX2, no significant difference in the proportion of up- and down-regulated genes (versus total numbers of up- and down-regulated genes, respectively) was observed. In contrast, the 18 genes matching with E2F were all down-regulated. As the expression of E2F transcription factors themselves was not significantly changed, this suggests that cisplatin may affect their activity.

The list of genes regulated by cisplatin was compared to lists of oncogenes (http://​www.​uniprot.​org/​uniprot/​?​query=​keyword:​KW-0656) and tumor suppressors (http://​www.​uniprot.​org/​uniprot/​?​query=​keyword:​KW-0043). Data showed that cisplatin reduced the level of some tumor suppressor genes and of oncogenes while inducing others (Table 4). Strikingly, the expression of the two AP-1 members FOS and JUN was found to be increased.

Potential modifications of splicing by cisplatin were investigated from the RNA-seq data. The SpliceSeq software identified 717 AS events occurring in 619 primary transcripts (Additional file 4). Only 5 genes (UGDH, SLC38A1, RETSAT, PDE8A, NASP) (0.44%) were affected simultaneously at transcriptional and post-transcriptional levels. Changes in splicing were grouped based on the type of events being affected: 79 changes involved cassette exon inclusion events, 166 were cassette exon exclusion events (of which 49% were not annotated as alternative exons in NCBI), 243 changes affected alternative 5′ or 3′ splice site selection events, 144 involved alternative promoters, 83 indicated alternative terminations and 2 were splicing changes attributed to mutually exclusive exon. Significantly affected biological processes identified by ToppFun Suite software on affected genes were “RNA splicing and processing” and “cell cycle” (Table 2).

For validation purpose, 16 splicing events identified by SpliceSeq as affected by cisplatin were evaluated by RT-PCR. These events were chosen such as to cover the range of AltSplice RPKM (reads per kilobase per million reads) values from 0 to 40. RT-PCR confirmed the alternative splicing of the ten exon skipping events and four of the six exon inclusion events in MCF7 and Ishikawa cells, indicating a good concordance with RefSeq data (drawings of nine splicing events in MCF-7 cells are shown in Figure 2).

As cisplatin induces DNA damage, the contribution of three main actors of the DDR pathway (ATM, ATR and DNA-PK) in cisplatin-induced exon inclusion/exclusion was investigated. Cisplatin affected the splicing in ATM-deficient AT5BIVA cells (illustrated for HNRNPDL exon 6 exclusion and exon 8 inclusion in Figure 3A) and GM9607 cells (not illustrated). Similarly to its effect in MCF7 cells (Figure 2), identical data were also found in MO59J cells that lack the catalytic subunit of DNA-PK (Figure 3B), and in MCF7 cells transfected with a siRNA targeting ATR (not shown). Moreover, specific inhibition of ATM (using ATM kinase inhibitor) or of DNA-PK (using NU7026) did not reverse the splicing induced by cisplatin in MCF7 cells (Figure 3C-D). Combined inhibition experiments were performed to evaluate whether the three DDR members might functionally compensate each other. Inhibition of ATM and ATR (using caffeine) and DNA-PK (using NU7026) failed to reverse the splicing induced by cisplatin (Figure 3E-F). The same results were obtained in MO59J cells (deficient in DNA-PK activity) treated with caffeine (not shown). Together, these data strongly suggest that the DDR does not participate in the splicing change of HNRNPDLe6 and AMZ2 induced by cisplatin. The DNA damage-activated protein p53 is similarly not involved since cisplatin induces an alteration of MDM2 splicing in the p53-deficient cell line MG-63 (Figure 1F).

As cisplatin is also known to activate PI3K in several cell types [26], the implication of the PI3Ks in the AS changes induced by cisplatin was evaluated. MCF7 cells were pre-treated with wortmannin three hours prior to adding cisplatin. At the concentrations used, wortmannin inhibits class I and III PI3Ks as well as PI3KC2b, but not ATM, ATR, DNA-PK. Dose-dependent inhibition of the effect of cisplatin on the splicing events was observed (Figure 3G-J). A similar impact was observed with the wortmannin derivative PX866 (Figure 3K-L). To gain further insights into the identity of PI3Ks involved, inhibitors specifically targeting the class I PI3Ks (TGX211, IC87114) were used (Figure 4A-D). As observed with wortmannin, these inhibitors significantly reversed the cisplatin-induced splicing changes, while no effect was observed in the absence of cisplatin. These results suggest that the class I PI3Ks are involved in the cisplatin-mediated response. RNA-Seq data indicate that p110α and p110β, but not p110γ and p110δ, are expressed in MCF7 cells. P110α, but not p110β, is activated by insulin. At the concentration used, TGX221 and IC87114 did not reduce the phosphorylation of Akt induced by insulin (not shown). Cisplatin treatment did not induce Akt phosphorylation on Ser473 under our experimental conditions (not shown). Moreover, the Akt inhibitors triciribine and MK2206, while efficiently reducing the insulin-induced phosphorylation of Akt, did not affect the cisplatin-induced changes in splicing in AMZ2 and HNRNPDL-E6 (Figure 4E-H). Finally, insulin did not induce a change in AS that was similar to cisplatin in conditions that increase Akt phosphorylation (not shown). Together, these observations strongly suggest that the splicing alterations elicited by cisplatin require p110β, but are independent of Akt.

Using a siRNA screen targeting 57 splicing factors, we identified SRSF4 as a regulator of hnRNPDL exon 6 splicing in MCF7 cells in basal growth conditions. We evaluated the role of SRSF4 in mediating the effect of cisplatin on AS by using a siRNA targeting SRSF4 and a control siRNA targeting the splicing factor SRSF6. Reduced levels of SRSF4 mRNA (81 ± 7% reduction, n = 3) and SRSF6 mRNA (80 ± 10%, n = 3) were confirmed by end-point RT-PCR (Figure 5A-B). siSRSF4 alone or when combined to siSRSF6 (siSRSF4/6) partly abrogated the splicing changes induced by cisplatin in the events tested (Figure 5C-F). The control siSCR and siSRSF6 alone had no effect. A similar reduction of the cisplatin-induced HNRNPDL exon 6 exclusion and exon 8 inclusion after knock-down of SRSF4 was observed in the breast cancer cell line BT549 (not illustrated).

To address the potential involvement of SRSF4-dependent splicing events induced by cisplatin in cell death or cell survival, MCF7 cells transfected with the control siRNA siSCR or with siSRSF4 were treated or not with cisplatin for 48 hours. Apoptosis and necrosis were measured by FACS (Figure 6A-D). Although the downregulation of SRSF4 had no effect on growth when cisplatin was absent, it strongly reduced cell death observed in the presence of cisplatin (15 ± 4% for siSRSF4 versus 6 ± 2% for siSCR; p = 0.02), which represents a 62 ± 9% reduction of cisplatin-induced cell death by SRSF4 repression (Figure 6E). These data were further confirmed by using the trypan blue exclusion assay (53 ± 3% reduction, p ≤ 0.01, n = 3, Figure 6E).