Splicing modulator FR901464 is a potential agent for colorectal cancer in combination therapy

FR901464 (FR) was first described as an anticancer drug and later identified as a modulator of splicing factor 3B subunit 1 (SF3B1). Although the effectiveness of splicing modulators has been investigated in colorectal cancer (CRC) cells, their usefulness in animal experiments has not been confirmed. The association of SF3B1 with CRC progression and the influence of FR on transcriptional activity in CRC has not been fully elucidated. FR showed strong cytotoxicity against CRC cell lines, SF3B1-mutated cancer cell lines, and human fibroblasts with IC50 values less than 1 ng/ml. FR-resistant clones derived from HCT116, DLD1, Lovo, and CT26 cells showed IC50 values greater than 100 ng/ml. SF3B1 sequencing demonstrated low frequencies of SF3B1 mutations in CRC and mutations in codon 1074 of exon 22 in all FR-resistant clones. Unlike hematological malignancies, SF3B1 expression was not associated with CRC progression. Although FR showed significant growth inhibition in a xenograft model of RKO cells, severe toxicity was also induced. These data indicated CRC might be a suitable target of FR unless toxicity occurs. Microarray analysis and real-time quantitative PCR demonstrated downregulation of genes associated with Fanconi anemia (BRCA1 and BRCA2) and 28 driver oncogenes. These data suggested combination treatment of FR with other anticancer drugs whose sensitivity is associated with genes affected by FR treatment. Combination treatment with PARP1 inhibitor olaparib, whose sensitivity was enhanced by BRCA 1/2 deficiency, showed synergistic effects in CRC cells. Our data indicates the potential of FR in combination therapy rather than monotherapy for CRC treatment.


INTRODUCTION
FR901464 (FR) and FR901228 were discovered as anticancer drugs by Nakajima et al. [1][2][3] and were subsequently identified as a SF3B1 inhibitor and a histone deacetylase inhibitor (HDACI), respectively [4,5]. FR901228 and other HDACIs with different chemical structures have similar biological effects on cancer cells, including induction of cell cycle arrest, differentiation, and apoptosis, by influencing transcriptional activity [6,7]. Because FR was also shown to enhance the transcriptional activity of SV40 promoter [1] we postulated that FR might also have biological effects in cancer cells by influencing transcriptional activity.
Colorectal cancer (CRC) is one of the most common malignancies in Western countries and Japan [37]. Advanced metastatic CRC is still a lethal disease despite the recent application of combination chemotherapy involving 5-fluorouracil (5-FU) and oxaliplatin (OHP) or irinotecan with molecular targeted therapy against vascular endothelial growth factor (VEGF) and epidermal growth factor receptor (EGFR) [38].
Yokoi et al. and Teng et al. reported SF3B1 mutations at codon 1074 (exon 22) in CRC cells resistant to Pladienolides [16,23]. CRC cell lines have been widely used for the study of splicing modulators [11, 13-15, 18, 22]. However, the development of splicing modulators for CRC treatment has been limited and the influence of SF3B1 on CRC progression was not determined.
In the present study, we assessed the possible application of the splicing modulator FR for CRC treatment. First, we evaluated the cytotoxic effect of FR on CRC in vitro and in vivo. Next, we analyzed SF3B1 mutations and expression in CRC cell lines and CRC tumors to assess the association between SF3B1 and CRC progression. Then, the influence of FR on transcriptional activity and alternative splicing was evaluated. Finally, based on the influence of FR on BRCA1 and BRCA2, we evaluated combination treatment with FR and the PARP1 inhibitor olaparib in CRC cells to explore the possible application of FR with other anticancer drugs.

Sensitivity of cancer cell lines to FR901464
The IC 50 values for FR in cancer cell lines and human fibroblasts were less than 1 ng/ml (0.18 to 0.71 ng/ml) ( Figure 1A). The IC 50 of DLD1 (0.71 ng/ml) was highest among the cell lines tested and was significantly higher than that of all other cell lines except for HCC38.
The IC 50 for HCT116 (0.31 ng/ml) was similar to that of other cells except for DLD1 and HCC38. The IC 50 for human fibroblasts (0.18 ng/ml) was the lowest but statistically similar to that of other cell lines except for DLD1, HCC38, and COLO829. The melanoma cell line COLO829 and the breast cancer cell line HCC38, which were reported to have SF3B1 missense mutations in exons 12 and 15, respectively, are also sensitive to FR [26].
The cytotoxicity of FR depended on the time and dose of treatment ( Figure 1B and Figure 1C). Treatment with FR for 3 h and 6 h induced similar cytotoxicity to 12 h of treatment in HCT116 and DLD1 cells, respectively, using Dunnett's test.

Effect of FR on tumor growth
In mouse subcutaneous xenograft models, the antitumor effect of FR was accompanied by high toxicity. Three of seven mice receiving 0.5 mg/kg FR ( Figure 1G) and four of nine mice receiving 0.75 mg/kg FR ( Figure  1H) died following intraperitoneal administration of FR. However, there was significant inhibition of tumor growth in a xenograft model of RKO at day 14 (P = 0.038), day 18 (P = 0.04), and day 22 (P = 0.0086) ( Figure 1H). The lack of a significant effect of FR on HCT116-derived xenografts ( Figure 1G) was consistent with results of a previous study using FR901464 analogues [13].
All human FR-resistant clones were derived from human CRC cells with MSI, which are known to carry many gene mutations [38]. We sequenced all of the SF3B1 exons in 50 patients with MSS (45 CRC and five other tumors) and in 31 patients with MSI (30 with CRC and one with small bowel gastrointestinal stromal tumor) but no mutations were detected in these 81 patients ( Table 2).
All human FR-resistant clones and their parental cells had silent mutations in SF3B1 exon 8 (codon 320 in LoVo and codon 371 in DLD1 and HCT116), whereas the other samples had no exon 8 mutations. Therefore, we postulated that mutations in exon 8 could be used to predict the possibility of SF3B1 exon 10 or exon 22 mutations. We further assessed samples from 171 CRC tumors and one metastatic renal cancer in the colon and found no SF3B1 mutation in exons 8, 10, and 22 (Table 2). Of the 171 CRC tumors, we found 66 CRC tumors bore no SF3B1 mutation in exons 12−15, which were recognized as mutational hotspots in other malignancies [26].
We found no SF3B1 mutations in these 253 tumors including 246 CRC tumors ( Table 2). These data suggest that SF3B1 mutations are rare in malignancies of the lower gastrointestinal tract irrespective of MSI or MSS status.

Influence of FR, SF3B1 mutation, and CRC progression on SF3B1 expression
FR treatment had a limited effect on SF3B1 expression in CRC cell lines, with a 2-fold change at most (Figure 2A). FR did not affect SF3B1 expression in DLD1 cells. However, in HCT116 cells, SF3B1 expression was enhanced by 1-3 h of FR treatment and inhibited by 24 h of FR treatment. SF3B1 expression in FR-resistant clones was unaffected by FR treatment, similar to their parental cells ( Figure 2B).    We also assessed SF3B1 expression in cancer tissues compared with normal mucosa (relative SF3B1 expression) in 80 CRC patients. Relative SF3B1 expression varied from −2.66 to 1.5, with a median of −0.35 and mean of −0.39 ( Figure 2C). There was no significant difference in relative SF3B1 expression among patients according to clinical stage (I, 10 patients; II, 28 patients; III, 25 patients; IV, 17 patients by TNM classification version 7) [41].
These data suggested that SF3B1 expression was not greatly influenced by FR treatment, SF3B1 mutation, or CRC progression.

Modulation of alternative splicing of MDM2 and CRNDE by FR treatment or SF3B1 mutation
We next evaluated the effect of FR treatment or SF3B1 mutation on alternative splicing of MDM2 ( Figure  2D, 2E) and CRNDE ( Figure 2F, 2G), as previously described [14,42].
As shown in Figure 2D, FR treatment (10 ng/ml and 100 ng/ml) for 24 h resulted in PCR products of 1019 bp and 785 bp instead of a 1604-bp product as a result of modulation of alternative splicing of MDM2 in HCT116 cells but not in the FR-resistant clones HCT/FR1 and Oncotarget 357 www.oncotarget.com HCT/FR2 [29]. Modulation of alternative splicing was induced by 1 h of FR treatment, resulting in PCR products of sizes other than 1604 bp ( Figure 2E).
CRNDE, which possesses 10 transcript variants (a, b, c, d, e, f, g, h, i, j) through alternative splicing, is highly expressed in CRC cell lines and patients [30]. FR treatment significantly reduced the expression of variants a, d, f, and j in HCT116, but not in FR-resistant clones HCT/FR1 and HCT/FR2 ( Figure 2F). FR significantly reduced the expression of variants a, b, f, g, h, and j in DLD1 cells, but not in FR-resistant clones DLD/FR1 and DLD/FR2 ( Figure 2G). The fact that FR had a different effect on alternative splicing between HCT1116 and DLD1 cells suggests that the influence of FR on alternative splicing would vary among CRCs.

Pathways affected by SF3B1 inhibition through FR treatment or SF3B1 mutation
To evaluate the pathways affected FR inhibition of SF3B1 in CRC, we used the Agilent expression array. A total of 22,125 probes were differentially expressed on the basis of signal fold-changes of ≥4 and signal differences of ≥100. Hierarchical cluster analysis indicated significant differences in gene expression patterns by cell type (between DLD1 and HCT116 cells, and between HCT116 and HCT/FR1 cells) and FR treatment (between FR (−) and FR10 in both DLD1 and HCT116 cells) ( Figure 3A).
PC analysis also showed significant differences in gene expression by cell type (between DLD1 and HCT116 cells, and between HCT116 and HCT/FR1 cells) and FR Principal component analysis showed that PC1 was associated with FR treatment in both DLD1 and HCT cells, and PC2 was associated with cell type. (C, D) Combination treatment of DLD1 and HCT116 with FR and olaparib. In DLD1 cells, pretreatment with FR enhanced cytotoxicity of olaparib (0.5 μM and 10 μM) at an FR concentration of 2.5 ng/ml but not at 1 ng/ml (C). In HCT116 cells, pretreatment with FR enhanced cytotoxicity of olaparib (0.1 μM to 5 μM) at a concentration of 2.5 ng/ml but not at 1 ng/ml (D).
Oncotarget 358 www.oncotarget.com treatment (between FR (−) and FR10 in both DLD1 and HCT116 cells) ( Figure 3B). In contrast, gene expression in DLD/FR1 and HCT/FR1 cells was not affected by FR treatment. PC1, but not PC2, was associated with FR treatment in both DLD1 and HCT116 cells. PC2 was associated with cell type between DLD1 and HCT116 or between HCT116 and HCT/FR1, although the difference between HCT116 and HCT/FR1 was much smaller than that between DLD1 and HCT116. The cumulative proportion of variance was 47.8% in PC1 and 30.3% in PC2.
We further examined the KEGG pathways according to the results of principal component analysis. A total of 2,691 probes were upregulated ≥2-fold and 4,896 probes were downregulated ≥2-fold in both DLD1 and HCT116 cells. Gene symbols corresponding to these probes showed that 2,474 genes were upregulated, 4,549 were downregulated, and 132 were either upregulated or downregulated depending on the probe. A total of 4,341 genes were differentially expressed in DLD/FR1 or HCT/ FR1 compared with their parental cells; of these, 149 genes were upregulated and 30 were downregulated in both DLD/FR1 and HCT/FR1. Therefore, the influence of FR treatment on changes in gene expression was much more consistent in CRC cell lines than in FR-resistant clones.
Pathway analysis predicted the enrichment of several significant pathways. The pathways that were significantly influenced in PC1, which represented differences in FR treatment, were different from those in PC2, which represented differences in cell type (Table  3). These results indicated that FR treatment significantly influenced pathways associated with spliceosomes, the cell cycle, Fanconi anemia, homologous recombination, ubiquitin-mediated proteolysis, nucleotide excision repair, oocyte meiosis, basal transcription factors, and endocytosis. Cyclin A2, Cyclin B1, Cyclin E2, CDK1, CDK6, and CDK8 were included among the cell cycle pathway genes affected by FR. BRCA1, BRCA2, and PALB2 were among the genes associated with the Fanconi anemia pathway. BRCA1 and BRCA2 were also included in the ubiquitin-mediated proteolysis and homologous recombination pathways, respectively.

Effect of SF3B1 inhibition and SF3B1 mutations on driver genes
Of 138 driver genes reported by Vogelstein [43], 52 oncogenes and 60 tumor suppressor genes were affected by FR treatment in either HCT116 or DLD1 cells (Tables  4, 5). Quantitative RT-PCR confirmed the inhibitory effect of FR treatment on 27 oncogenes including BCL2, CTNNB1, EGFR, ERBB2, KRAS, and MET driver genes.
In FR-resistant clones, 36 genes were affected in either DLD/FR1 or HCT/FR1, of which the expression of two genes, RET and SETBP1, was increased in both clones by microarray (Table 6). In contrast to FR treatment, quantitative RT-PCR did not confirm the changes in gene expression in HCT/FR1 and DLD/FR1 cells bearing SF3B1 gene mutations ( Table 6).

Effect of SF3B1 inhibition on Fanconi anemiaassociated genes and cancer-associated genes
qRT-PCR data confirmed that the expression of Fanconi anemia-associated genes including BRCA1, BRCA2, PALB2, and RAD51 and cancer-associated genes (ALDH1A3 and IGF1R) was strongly inhibited by FR treatment, as shown by microarray analysis ( Table  3). Inhibition of BRCA1 and BRCA2 gene expression suggested a synergistic cytotoxic effect with PARP1 inhibitor based on previous findings in breast cancer with mutation of BRCA1 or BRCA2 [44]. PALB2, ALDH1A3, and IGF1R were reported to show associations with sensitivity to mitomycin, stem cell markers, and cancer progression, respectively [45][46][47].

Combination treatment with FR and olaparib
Based on the above findings of inhibition of BRCA1 and BRCA2 expression by FR, we assessed the effect of combination treatment with FR and olaparib. In DLD1 cells, pretreatment with FR enhanced cytotoxicity of olaparib (0.5 μM and 10 μM: P < 0.05) at a FR concentration of 2.5 ng/ml but not at 1 ng/ml ( Figure  3C). In HCT116 cells, pretreatment with FR enhanced the cytotoxicity of olaparib (0.1 μM to 5 μM: P < 0.005) at 2.5 ng/ml but not at 1 ng/ml. (Figure 3D). These findings are consistent with the data of combination index (CI). In DLD1 cells, CI was 0.84 and 1.08 at a FR concentration of 2.5 ng/ml and 1 ng/ml, respectively. In HCT116 cells, CI was 0.93 and 1.37 at a FR concentration of 2.5 ng/ml and 1 ng/ml, respectively. Synergistic effects were shown at a FR concentration of 2.5 ng/m in both DLD1 and HCT116.

DISCUSSION
In this study, we assessed the possible application of the splicing modulator FR for CRC treatment. FR showed effective cytotoxicity against CRC cells in vitro and in vivo. This is the first report to show the efficacy of FR analogues in solid tumors, although the antitumor effect was dependent on the cell line and was associated with severe toxicity in vivo. The instability of FR under serumcontaining conditions has been described [5,9,11]. FR analogues including SSA and Meayamycin appear to be more stable than FR in serum [5,18]; however, sufficient efficacy of FR analogues for solid tumors has not yet been demonstrated [18,21,22]. Although we considered that intraperitoneal administration of FR might reduce both toxicity and efficacy in vivo, FR via this route maintained both toxicity and efficacy. Therefore, improvement of Oncotarget 359 www.oncotarget.com All of the FR-resistant clones carried mutations in codon 1074 of the SF3B1 gene, which is consistent with a previous report on another SF3B1 inhibitor, Pladienolide [16,23]. In contrast, DLD1, COLO829, and HCC38 cells, which contained mutations at other sites in the SF3B1 gene, were sensitive to FR. Therefore, mutation sites associated with FR sensitivity are different from those associated with cancer progression. Our study also showed low frequencies of SF3B1 gene mutations and no mutations in codon 1074 of the SF3B1 gene in CRC patient samples. Thus, these results indicate that CRC cells, and possibly all cancer cells, are intrinsically sensitive to splicing modulators.
The influence of FR on alternative splicing of MDM2 and CRNDE as a result of splicing modulation was abrogated in these FR-resistant clones. Moreover, the effect of FR on gene expression, including suppression of oncogenes and cancer-associated genes, was also abolished in FR-resistant clones. These results indicated that the cytotoxicity of FR was associated with its influence on gene splicing and gene expression through inhibition of SF3B1 function.
Splicing modulators have been reported to inhibit expression of VEGF, c-MYC, and MCL-1 [12,22,48]. A previous study reported downregulation of VEGF and MCL-1 and upregulation of c-MYC by a splicing modulator. In our microarray study, downregulation of VEGFA in HCT116 and upregulation of MYC in both DLD1 and HCT116 were the same as the previous study. However, MCL1 was upregulated in DLD1 and does not reflect the findings of the previous report. Differences from previous studies may depend on variations in cell types and evaluation methods.
Our study suggested that splicing modulation might downregulate many oncogenes and cancer-associated genes. Moreover, splicing modulators could influence gene expression associated with drug sensitivity. Inhibition of BRCA1 and BRCA2 expression by FR suggested a synergistic cytotoxic effect with olaparib based on the fact that this PARP1 inhibitor is effective in breast cancer with genetic deficiencies in BRCA1 or BRCA2 [44]. The enhanced cytotoxicity of olaparib after pretreatment with FR in this study supports this hypothesis. Combination therapy with splicing modulators to overcome drug resistance has been reported in lung cancer cells and melanoma [17,21].   Our study showed low frequencies of SF3B1 mutations and no mutations in codon 1074 of the SF3B1 gene in CRC patient samples. These results indicate that CRC cells, and possibly all cancer cells, are intrinsically sensitive to splicing modulators. SF3B1 expression levels in the CRC patients of our study were not associated with tumor progression, in contrast to previous findings [48,49]. This discrepancy may reflect differences in sample conditions; we stored samples for RNA isolation in RNAlater at −80°C whereas the other studies collected RNA from samples in which the RNA might be fragmented.
Our study has several limitations. First, the severe toxicity of FR in the xenograft model prevented us from further examining the efficacy of FR in vivo. However, the lower IC 50 value of FR in RKO cells in vitro seemed to be associated with efficacy in vivo. Improved FR delivery in vivo is critical for sufficient efficacy and reduced toxicity.
Oncotarget 363 www.oncotarget.com Second, we did not show that mutations in codon 1074 of SF3B1 definitively induced FR resistance, although the presence of codon 1074 mutations in all the FRresistant clones, in pladienolide-resistant clones, and in clones derived from CT26 cells with MSS strongly implicates a role for mutations in codon 1074 of SF3B1 in FR resistance. Third, the influence of FR on driver oncogenes and cancer-associated genes was not evaluated in other types of cancer and with other splicing modulators. A consistent effect on driver oncogenes and cancer-associated genes, regardless of cancer type and splicing modulator, will promote further study of splicing modulators. Finally, the effect of combination therapy was not confirmed in vivo because of a shortage of FR.
In conclusion, our study indicates the possible application of splicing modulators in combination therapy for the treatment of cancers including CRC through their influence on gene expression associated with drug sensitivity.

Drugs and chemicals
FR, a gift from Astellas Pharma Inc. (Tokyo, Japan), and olaparib (Funakoshi, Tokyo, Japan) were dissolved in DMSO at a concentration of 100 μg/ml and stored at −20°C. OHP (Yakult Inc., Tokyo, Japan) was stored at 4°C. 5-FU (Sigma-Aldrich Japan, Tokyo, Japan) was dissolved in phosphate-buffered saline at a concentration of 10 mM and stored at 4°C. All drugs were diluted in culture medium immediately before use.

Cell lines and cloning of drug-resistant cells
The human melanoma cell line COLO829, human breast cancer cell line HCC38, and murine colon tumor cell line CT26 were purchased from ATCC (Manassas, VA, USA). Human CRC cell lines Caco2, DLD1, HCT116, LoVo, HT29, and RKO were gifts from Dr. Hirofumi Yamamoto (Osaka University Graduate School of Medicine, Japan). All human cell lines were authenticated by ATCC using DNA profiling. Cells were maintained in appropriate media as follows: DMEM (DLD1 and HCT116), RPMI (HT29, RKO, COLO829, CT26, and HCC38), and Hanks-F12K (LoVo) supplemented with 10% fetal bovine serum, 10,000 units penicillin, 10 mg/ ml streptomycin, and 25 μg/ml amphotericin B. Culture media and fetal bovine serum were obtained from Life Technologies Japan (Tokyo, Japan). All cells were grown at 37°C in a humidified incubator with 5% CO 2 . To obtain Oncotarget 364 www.oncotarget.com FR-resistant cells, cells were cultured in the presence of high concentrations of FR (20 ng/ml for DLD1 and 10 ng/ ml for all other cell lines).

Patients
Specimens were collected from 318 patients who underwent surgery at our department. All protocols were approved by the ethics committee of Hyogo College of Medicine as RINHI-0120 and all patients provided written informed consent. The specimens consisted of 312 CRC samples (299 primary tumors and 13 metastatic tumors), two small bowel gastrointestinal stromal tumors (GISTs), and one case each of small intestinal adenocarcinoma, rectal carcinoid, retroperitoneal liposarcoma, and metastatic renal cancer in the colon. The specimens were obtained with adjacent normal tissues for comparison and stored at −80°C before use. For RNA analysis, samples were stored with RNAlater (Qiagen, Venlo, Netherlands).

Drug treatment in cell culture
Cells were seeded in 200 μl medium in 96-well flat-bottom plates at a density of 2×10 3 cells per well. The next day, the medium was removed and various concentrations of FR (0.1−500 ng/ml), OHP (0.1−500 μM), and 5-FU (0.1−500 μM) were added. To assess the time dependency, cells were incubated with FR for 1 h, 3 h, 6 h, 12 h, and 96 h. To assess the effect of combination treatment with FR and olaparib, cells were treated with FR (0, 1 ng/ml, 2.5 ng/ml) for 1 h followed by treatment with olaparib at various concentrations for 96 h. After drug treatment, cells were counted using a cell counting kit (Dojindo Laboratories, Kumamoto, Japan) according to the manufacturer's instructions. Half maximal inhibitory concentration (IC 50 ) was calculated as the concentration that corresponded to a 50% reduction in cellular proliferation compared with untreated cells. Experiments were independently performed at least three times and data are shown as means ± standard deviations.

Mutation analysis of SF3B1 by Sanger sequencing
Genomic DNA was extracted from eight cancer cell lines (six CRC cell lines, COLO829, and HCC38), nine FR-resistant clones, and 253 patient-derived malignant tumors (including 246 CRC patients) using the QIAamp DNA mini kit (Qiagen) ( Table 2). We analyzed all 25 exons of SF3B1 in all cancer cell lines and FR-resistant clones in addition to 81 tumors, including 75 CRC patients, by Sanger sequencing as previously described [26]. Based on these findings, we further examined specific SF3B1 exons in the remaining 172 tumors, including 171 CRCs (Table 2).

Microsatellite instability
Genomic DNA from patients was used to evaluate their microsatellite instability (MSI) status. The mononucleotide microsatellite markers BAT-25, BAT-26, NR21, NR22, and NR24 were used for evaluation as previously described [50].

Human tumor xenografts
Five-week-old female BALB/cAJcl-/nu/nu mice were purchased from Japan Clea Inc. (Tokyo, Japan). Mouse care and experiments were performed under specific pathogen-free conditions. All animal protocols were approved by the Institutional Animal Care and Use Committee of Hyogo College of Medicine and performed at the Institute of Experimental Animal Sciences of Hyogo College of Medicine.
A total of 5×10 6 HCT116 or RKO cells were subcutaneously injected into mice and the resultant tumors were allowed to grow larger than 5 mm in diameter. FR (0.5 mg/kg or 0.75 mg/kg) was injected intraperitoneally every fourth day into seven mice that received HCT116 cells and nine mice that received RKO cells for a total of six treatments. Tumor volume was calculated as a × b 2 , where a represents the tumor length and b represents its width.

Microarray
A total of 1×10 6 DLD1, HCT116, DLD/FR1, or HCT/FR1 cells were seeded into P10 tissue culture plates. After overnight culture, the cells were incubated for a further 24 h with or without 10 ng/ml FR and RNA was extracted using the TRIzol RNA Purification kit (Invitrogen, Waltham, MA, USA). Gene expression profiles were analyzed using the Agilent SurePrint G3 Human GE 8x60K v2 Microarray kit (Agilent Technologies, Santa Clara, CA, USA) according to the manufacturers' instructions at Dragon Genomic Center (Mie, Japan). The data set is available at Gene Expression Ominibus under accession number GSE58241.

Microarray data analysis
Signal data were imported into GeneSpring (Agilent) for analysis. Among the 40,380 probes evaluated to be suitable, 22,125 were extracted as differentially expressed genes based on signal fold-changes of ≥4 and signal differences of ≥100. Hierarchical analysis and principal component (PC) analysis were performed to explore similarities and differences in gene expression following FR treatment and in the presence of SF3B1 mutations. Based on the principal analysis, probes in the first and second principal component were used for further Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis.
Oncotarget 365 www.oncotarget.com PCR and quantitative real-time reverse transcription PCR (qRT-PCR) Total RNA was extracted from CRC cell lines and samples from 80 CRC patients using an RNeasy Mini kit (Qiagen). CRC samples stored with RNAlater were used for qRT-PCR. After generation of cDNA using Superscript III reverse transcriptase (Invitrogen), PCR was performed using specific oligonucleotides (Supplementary Table 1) to quantify gene expression and for RNA sequencing of SF3B1. Primers for SF3B1 sequencing have been reported previously [26].
To assess MDM2 RNA splicing, products were analyzed by agarose gel electrophoresis and ethidium bromide staining [14]. qRT-PCR was performed using the Power SYBR Green Master Mix and appropriate primer pairs (Life Technologies, Supplementary Table  1). Data were collected from wells in triplicate and each experiment was performed at least twice. ΔΔ CT was calculated by evaluating the difference in ∆ CT between the target gene and GAPDH as a control. The methods used to assess variants of CRNDE were reported previously [42].

Statistical analysis
Tukey-Kramer test was used to evaluate the IC 50 of FR in cell lines. Dunnett's test was used to evaluate timeand dose-dependency of FR with 12 h treatment as the control. Dunnett's test was also used to assess the IC 50 of FR, OHP, and 5-FU in FR-resistant clones relative to the parental cells. The efficacy of combination treatment of FR and olaparib was assessed by Dunnett's test relative to treatment without FR. Student's t-test was used to evaluate tumor volume in xenograft models. P-values less than 0.05 were considered statistically significant. To evaluate the effect of combination therapy of FR and olaparib, combination index (CI) was calculated [51]. A CI of less than 1 was considered to be a synergistic effect.