Jerantinine A induces tumor-specific cell death through modulation of splicing factor 3b subunit 1 (SF3B1)

Precursor mRNA (pre-mRNA) splicing is catalyzed by a large ribonucleoprotein complex known as the spliceosome. Numerous studies have indicated that aberrant splicing patterns or mutations in spliceosome components, including the splicing factor 3b subunit 1 (SF3B1), are associated with hallmark cancer phenotypes. This has led to the identification and development of small molecules with spliceosome-modulating activity as potential anticancer agents. Jerantinine A (JA) is a novel indole alkaloid which displays potent anti-proliferative activities against human cancer cell lines by inhibiting tubulin polymerization and inducing G2/M cell cycle arrest. Using a combined pooled-genome wide shRNA library screen and global proteomic profiling, we showed that JA targets the spliceosome by up-regulating SF3B1 and SF3B3 protein in breast cancer cells. Notably, JA induced significant tumor-specific cell death and a significant increase in unspliced pre-mRNAs. In contrast, depletion of endogenous SF3B1 abrogated the apoptotic effects, but not the G2/M cell cycle arrest induced by JA. Further analyses showed that JA stabilizes endogenous SF3B1 protein in breast cancer cells and induced dissociation of the protein from the nucleosome complex. Together, these results demonstrate that JA exerts its antitumor activity by targeting SF3B1 and SF3B3 in addition to its reported targeting of tubulin polymerization.

receptor (EGFR) 8 , p53 9 , vascular endothelial growth factor (VEGF) 10 , and E-cadherin 11 reportedly promoted cancer-associated pathways, including the evasion of apoptosis, increased cell proliferation, angiogenesis, and invasion. Mutations in SF3B1 have also been reported in myelodysplastic syndromes (MDS) as well as numerous cancers, including acute myeloid leukemia, primary myelofibrosis, chronic myelomonocytic leukemia (CML) 12 , chronic lymphocytic leukemia (CLL) 13,14 , multiple myeloma, uveal melanoma [15][16][17][18] and breast cancers [19][20][21] . While it is currently unclear as to how SF3B1 mutations might alter its function, previous studies have shown that the dysregulation of spliceosomal components can alter splicing patterns, causing intron retention or exon skipping, and affect protein isoform balances leading to abnormal cell proliferation or differentiation 2,22 . As such, the spliceosome has emerged as an attractive target for anticancer treatment. Several spliceosome modulators have already been identified, including natural products derived from bacterial fermentation (e.g. pladienolides, GEX1, FR901463, etc.) and their synthetic analogues (spliceostatin A, meayamycin and E7107) as well as natural plant products (e.g. isoginkgetin) 23 .
Indole alkaloids represent a large and highly structurally diverse group of secondary metabolites with remarkable bioactivities against the different targets in cancer. The importance of this group of compounds is best represented by the Vinca alkaloid vinblastine, which is currently among the foremost drugs used in cancer chemotherapy 24 . Previously, we have described the potent and selective antitumor activity of seven new Aspidosperma indole alkaloids, jerantinines A-G, isolated from the leaf extracts of the Malayan plant Tabernaemontana corymbosa (Fig. 1A) 25 . Jerantinines A-E were found to display pronounced in vitro anti-proliferative activities against human cancer cell lines in the nanomolar range [26][27][28] . Furthermore, we have recently demonstrated that jerantinine A and B and the acetate derivative inhibited tubulin polymerization, polo-like kinase 1 (PLK1) activity and induced G2/M cell cycle arrest in a panel of human cancer cell lines consisting of vincristine-resistant nasopharyngeal carcinoma cells 25 , as well as breast, colorectal, lung and pancreatic carcinoma cells 27,28 . Similarly, jerantinine E was also shown to disrupt microtubules, and displayed significant antitumor activity against human cervical carcinoma cells 29 . Importantly, no cross-resistance to jerantinines was observed in vincristine-resistant HCT-116 cells, suggesting that jerantinines overcome p-glycoprotein-mediated multidrug resistance and might affect other cancer-relevant targets besides tubulin 25,27,28 .
Using a pooled-genome wide shRNA library screen and global proteomic profiling, we demonstrated that jerantinine A (JA) targets the cancer spliceosome through the upregulation of SF3B1 and SF3B3 proteins in breast cancer cells. Importantly, ectopic expression of SF3B1, SF3B3 or JA treatment induced significant tumor-specific cell death accompanied by the accumulation of unspliced pre-mRNAs. In contrast, the depletion of endogenous SF3B1 or SF3B3 abrogated the apoptotic effects induced by JA, but not the G2/M cell cycle arrest. Further analyses revealed that JA stabilizes endogenous SF3B1 protein and disrupts the binding of the protein to the nucleosome complex in breast cancer cells. Together, our results suggest that JA exerts its antitumor activity by targeting SF3B1 in addition to its reported targeting of tubulin polymerization.

Results
Jerantinine A induces tumor-specific cell death in breast cancer cell lines. To test the selective antitumor activity of JA, we compared its anti-proliferative activities in a panel of breast cancer cell lines consisting of estrogen receptor (ER)-positive (MCF-7 and T47D) and triple-negative cells (MDA-MB-468). As shown in Fig. 1B and Supplementary Table S1, JA elicits activity against all the breast cancer cell lines being tested (approx. IC 50 = 1 μ M), while the non-transformed MCF-10A cells were relatively resistant (IC 50 > 10 μ M). Similarly, apoptotic morphological changes were also observed in JA-treated cancer cells, but no such changes were observed in non-transformed cells (Fig. 1C). The effects of JA on cell cycle arrest and cell death were further investigated by propidium iodide staining and annexin V/7-AAD flow cytometry analysis, respectively. Consistent with our previous reports, JA induced significant G2/M arrest in MCF-7 cells (19% in control cells vs 64% in JA-treated cells, Supplementary Fig. S1A) 27,28 . The percentage of apoptotic cells in MCF-7 and MDA-MB-468 cells after JA treatment was also significantly higher than in the control cells (P < 0.01, Student's t-test; Fig. 1D). However, no significant induction of caspase 3/7, 8 and 9 activities were detected in response to JA treatment, suggesting that JA might induce caspase-independent apoptosis ( Supplementary Fig. S1B). Of note, the induction of apoptosis was observed as early as 6 h after treatment, suggesting that JA might target endogenous proteins rather than transcriptional activation.

Proteomic profiling of MCF-7 breast cancer cells following JA treatment.
We have previously demonstrated that jerantinine A and B inhibited tubulin polymerization, PLK1 activity and induced G2/M cell cycle arrest in a panel of human cancer cell lines 27,28 . Importantly, no cross-resistance to jerantinine A and B was observed in vincristine-resistant cancer cells, suggesting that jerantinine A and B might affect other cancer-relevant targets in addition to tubulin. To systematically identify molecular targets involved in modulating the antitumor activity of JA, we carried out a combined proteomic-RNAi screen. We hypothesized that direct targets mediating the antitumor activities of JA should have two properties: 1) they should be differentially expressed following JA treatment, and 2) they should be functionally required to sustain the antitumor activity of JA.
To identify proteins that are differentially regulated in JA-treated MCF-7 cells, we conducted a global proteomic analysis using multidimensional protein identification technology (MudPIT) ( Fig. 2A and Supplementary Fig. S2A). We identified 509 up-regulated proteins (Supplementary Table S2) and 429 down-regulated proteins (Supplementary Table S3). Among them were proteins that are involved in assembly of the spliceosome, cellular metabolism, endocytosis and axon guidance ( Supplementary Fig. S2B). Of note, several proteins that are associated with mitosis (e.g. ANAPC7, MCM3, MCM7, SMC2, SMC3, TUBG1 and TUBG2) and apoptosis (e.g. FADD and DIABLO) were also found to be up-regulated, consistent with the phenotype we observed previously 27,28 .
Scientific RepoRts | 7:42504 | DOI: 10.1038/srep42504 Genome-wide shRNA functional screen identifies a compendium of genes affecting sensitivity to JA. In a parallel RNAi screen, we used the same cell line used for the proteomic profiling to identify genes that were functionally required for JA activity. We treated the pooled genome-wide shRNA transduced cells for 72 h with JA, isolated the genomic DNA and analyzed the abundance of shRNA by next-generation sequencing (NGS). Conceptually, the shRNAs that are "enriched" (Z-scores > 2) in the surviving cell population are genes that are essential for eliciting the killing response caused by JA since silencing the gene prevents the cell-death signal from propagating. Thus, analysis of the "enriched" shRNAs indicates which genes are necessary for the  lethal activity of JA. Vice versa, shRNAs that make cells more susceptible to JA treatment would be "depleted" in the surviving cell population, indicating that these genes might help cells survive in the presence of JA. Hence, inhibiting the function of these genes might sensitize the antitumor effects of JA.
Following this screen, a total of 381 candidate genes that were enriched in JA-treated cells and 121 candidate genes that were depleted in JA-treated cells were identified ( Fig. 2A and Supplementary Fig. S3A,B). Among the shRNAs that were highly enriched in JA-treated cells are genes that are involved in the human spliceosome, cell metabolic pathways, the neurotrophin pathway and the MAPK signaling pathway (Supplementary Fig. S3C  and Supplementary Table S4). Interestingly, shRNAs targeting genes that are involved in calcium signaling (e.g. P2RX1, PDE1C, TACR1, NOS3 and ITPR2), autophagy (e.g. ATG4A and ATG4C) and the estrogen signaling pathway (e.g. ITPR2, CREB3L2, NOS3 and SOS2) were found to be depleted in the surviving cells, suggesting that the inhibition of these pathways in combination with JA treatment may evoke a synergistic response (Supplementary Fig. S3C and Supplementary Table S5).
Combined functional genomic and proteomic approaches identify the cancer spliceosome as a putative target of JA. Next we 'intersected' the proteomic data set with the RNAi screen data set. In the intersection, we found 28 targets that were required for JA activity and were induced in JA-treated cells ( Fig. 2A and Supplementary Table S6). Among these targets was the Fas-associated via death domain (FADD) which is known to regulate apoptosis and necroptosis 30 , and glycogen synthase kinase 3 beta (GSK3B), which has been shown to function in a wide range of cellular processes and signaling pathways including suppression of the Wnt/ beta-catenin pathway 31 . Functional enrichment analysis was carried out using the Kyoto Encyclopedia of Genes and Genomes (KEGG) database to classify the candidate targets into statistically significant over-represented functional categories. The genes involved in the spliceosome were significantly over-represented (P < 0.01, Fig. 2B). To further investigate the mechanism of action of JA, the 28 candidate targets were mapped onto protein-protein interaction networks using STRING 32 . Interestingly, the analysis revealed that 7 targets (SF3B1, SF3B3, SNRPA, SNRPE, SNRPF, HNRNPA3 and EFTUD2) are physically interacting proteins within the spliceosome complex, suggesting that JA might regulate the cancer spliceosome to induce tumor-specific cell death (Fig. 2C).

JA triggers dysregulation of SF3B1 and SF3B3 proteins.
To validate whether JA regulates the cancer spliceosome, we carried out immunoblotting to confirm the expression of SF3B1 and SF3B3 following JA treatment in MCF-7 and MDA-MB-468 cells. As shown in Fig. 3A, JA induced a time-dependent increase in cellular SF3B1 and SF3B3 proteins, consistent with the results obtained from the MudPIT analysis. Of note, the induction of SF3B1 and SF3B3 was observed as early as 6 h after treatment, which corroborated with the induction of apoptosis (Fig. 1D). Interestingly, no significant changes in mRNA levels of the SF3B family in response to JA were detected ( Supplementary Fig. S4A), suggesting that the induction of protein expression is independent of transcriptional activity. Further analyses after the cycloheximide chase assay revealed that JA markedly prolonged the life-span of the SF3B1 protein in both MCF-7 and MDA-MB-468 cells, suggesting that JA regulates SF3B1 protein stability rather than transcriptional activation ( Supplementary Fig. S4B,C). In contrast to the breast cancer cells, both SF3B1 and SF3B3 were found to be down-regulated by JA in the non-transformed MCF-10A cells ( Supplementary Fig. S5A).
JA induces pre-mRNA splicing errors. As previous studies have demonstrated that anticancer spliceosome-targeting compounds such as pladienolides and sudemycin inhibit splicing activity [33][34][35] , we sought to evaluate the effect of JA on splicing activity by quantifying the amounts of unspliced pre-mRNA of RIOK3, CDKN1B, DNAJB1, and BRD2 by qPCR using primer sets specific for the intronic regions as described previously 33,35 . We observed that JA treatment resulted in a significant increase in unspliced pre-mRNA in MCF-7 and MDA-MB-468 cells, indicating that JA significantly impairs splicing activity ( Fig. 3B and Supplementary Fig. S6A). In contrast, the amounts of unspliced pre-mRNA of RIOK3 and CDKN1B were significantly lower in MCF-10A cells treated with JA ( Supplementary Fig. S5B), corroborating the down-regulation of SF3B1 and SF3B3 in these cells (Supplementary Fig. S5A). Of note, no significant changes in the levels of spliced mRNA was observed in cells following JA treatment (Supplemental Figs S5B and S6B).

Induction of SF3B1 impairs pre-mRNA splicing and induces apoptosis in breast cancer cells.
Since SF3B1 and SF3B3 have been shown to be a core component of the U2 snRNP of the spliceosome and is essential in regulating pre-mRNA splicing in cancer cells 4 , we hypothesize that the dysregulation of SF3B1 and/or SF3B3 (activation and inhibition) might result in the loss of critical proteins and the expression of variants with aberrant function that influence the survival of cancer cells.
To test whether SF3B1 and SF3B3 upregulation alone would trigger apoptosis and perturb splicing activity, SF3B1 and SF3B3 was transiently overexpressed in MCF-7 and MDA-MB-468 (Fig. 4A,B). Indeed, we observed that SF3B1 and SF3B3 overexpression induced significant apoptosis (Fig. 4C,D). Similarly, overexpression of SF3B1 also inhibited splicing activity in both cell lines, analogous to the phenotype induced by JA (Fig. 4E). In contrast, the splicing of RIOK3, CDKN1B, DNAJB1, and BRD2 remained unchanged in SF3B3 overexpressed cells (Fig. 4F). SF3B1 and SF3B3 are required for the antitumor effects of JA. Next, we investigated whether SF3B1 or SF3B3 mediate the antitumor activity of JA. We generated a stable pool of isogenic cells depleted for SF3B1 or SF3B3 by the transduction of two independent lentiviral shRNA constructs targeting the endogenous proteins in MCF-7 and MDA-MB-468 cells, followed by brief drug selection with puromycin. Indeed, depletion of SF3B1 or SF3B3 partially rescued the apoptotic cell death induced by JA in both MCF-7 and MDA-MB-468 cells (Fig. 5A-D). Furthermore, double knock-down of SF3B1 and SF3B3 completely abrogated the apoptotic effects of JA, suggesting that both SF3B1 and SF3B3 are required for the antitumor effects of JA (Fig. 5E,F). Also, knock-down of SF3B1, but not SF3B3, significantly reduced the accumulation of unspliced pre-mRNA induced by JA (Supplementary Fig. S7), suggesting that the mis-splicing of RIOK3, CDKN1B, BRD2 or DNAJB1 following JA treatment is dependent on SF3B1 but not SF3B3. SF3B1 depletion, however, did not rescue the G2/M cell cycle arrest induced by JA treatment (Supplementary Fig. S8), indicating that SF3B1 is dispensable for the cytostatic effects of JA, but more critical for apoptosis execution.

JA disrupts the interactions of SF3B1 and SF3B3 in the nucleosome complexes. A recent study
has shown that SF3B1 is associated with the nucleosome complexes and affects splicing outcomes 36 . To evaluate whether the association of SF3B1 and SF3B3 in the nucleosome complexes might be affected by JA, we performed co-immunoprecipitation (co-IP) assays in MCF-7 cells treated with JA for 24 h. Our results indicated a strong association of SF3B1 and SF3B3 with H2B and H3 in the nucleosome (Fig. 6). However, such associations were

Discussion
The spliceosome represents an under-explored and unusual target in cancer cells. The recent identification of small molecules (e.g. Spliceostatin, Sudemycins and Pladienolides) that interact with SF3B1, and the identification of mutant SF3B1 in tumor samples indicate that this complex can be a viable target for chemotherapeutic intervention 2,3 . While the exact mechanism of cytotoxicity induced by these spliceosome modulators is currently unclear, evidence indicates that interaction with SF3B1 is crucial for their antitumor activity 2,3 .
JA is an anticancer compound that has been previously reported by our group to possess remarkable antitumor activities 27,28 . Using combined proteomic and functional genomics analyses, we identified SF3B1 and SF3B3 as putative targets of JA. However, unlike other spliceosome modulators which have been shown to inhibit SF3B1, JA was found to induce SF3B1 and SF3B3 protein expression through increased protein stability. Intriguingly, the induction of SF3B1 by JA or ectopic expression of SF3B1 in MCF-7 and MDA-MB-468 cells impairs mRNA splicing activity as evidenced by the accumulation of unspliced RIOK3, CDKN1B, DNAJB1 and BRD2 pre-mRNA. In contrast, the amounts of unspliced pre-mRNA remained unchanged following ectopic expression of SF3B3 in MCF-7 and MDA-MB-468 cells, despite observation of significant apoptotic cell death. Of note, depletion of the endogenous SF3B1 or SF3B3 partially inhibited the apoptosis induced by JA in breast cancer cells, while the simultaneous knock-down of SF3B1 and SF3B3 completely abrogated the apoptotic effects of JA, suggesting that both SF3B1 and SF3B3 are required for the antitumor activity of JA in breast cancer cells. However, only knock-down of SF3B1, but not SF3B3, significantly reduced the accumulation of unspliced pre-mRNA following JA treatment indicating that the mis-splicing of RIOK3, CDKN1B, BRD2 or DNAJB1 following JA treatment is dependent on SF3B1 but not SF3B3. These results suggest that SF3B1 and SF3B3 might induce tumor specific cell death through distinct mechanisms. Indeed, SF3B3 has been recently shown to regulate alternative splicing of EZH2 pre-mRNA in renal cancer and mediates ubiquitination through modulation of the NEDD8 pathway 37,38 .
Numerous studies have also demonstrated that overexpression and mutations of key spliceosome components are associated with malignancies in various tissues, resulting in aberrant splicing activity and the emergence of hallmark cancer phenotypes 3,6,23,39-41 . Alterations in SF3B1 have been reported in myelodysplastic syndromes (~20%) 19,42,43 , acute myeloid leukemia (~5%) 19 , chronic lymphocytic leukemia (~5%) 12,19 and other solid tumors including breast cancer (~1%) [19][20][21] , renal cancer (~3%) 19 and adenoid cystic carcinoma (~4%) 19 . In breast cancer, in particular, SF3B1 mutations were significantly associated with ER-positive disease, AKT1 mutations, and distinct copy number alterations 20 . While the potential function of SF3B1 mutants has yet to be fully elucidated, the spliceosome modulator, spliceostatin A, was found to be more potent in cell lines harboring SF3B1 mutations compared to SF3B1 wild-type cells 20 . This, however, was not observed in the JA-treated cells, as the SF3B1 wild-type MDA-MB-468, T47D and MCF7 cells are equally sensitive to JA as compared to the HCC38 cells, which harbors a heterozygous SF3B1 Q534P missense mutation (data not shown) 44 . This result suggests that the antitumor effect of JA is independent of SF3B1 mutations.
It is intriguing to investigate how the dysregulation of SF3B1, via either activation or inhibition, could lead to tumor-specific cell death. It is known that the maintenance of high-fidelity mRNA splicing is important, because the translation of mis-spliced mRNAs into proteins with aberrant function could be deleterious to all cells. Therefore, it is not surprising that eukaryotic cells have evolved a protective mechanism, known as nonsense-mediated decay (NMD), to target and eliminate the inappropriately spliced mRNAs 45 . Interestingly, several pathways involved in NMD were recently found to be inactivated in many cancers including breast, colon, liver, lung, thyroid and esophagus cancer 46 . Although it is currently unclear whether JA sensitivity is dependent on NMD deficiency, a recent study showed that NMD efficiency in MCF-7 cells, which are sensitive to JA, was significantly lower compared to HeLa cells 47 . Similarly, siRNA-mediated depletion of SNRPE or SNRPD1, components of the core spliceosomal heteroheptameric Sm complex, also led to a marked reduction of cell viability in MDA-MB-468 and SKBR3 breast cancer cells, but not in the non-tumoral MCF-10A cells 48 . Thus, it is plausible that the dysregulation of SF3B1 and spliceosome activity by JA might lead to an imbalance in the splicing program in susceptible cancer cells, which may induce apoptosis by the accumulation of mis-spliced mRNAs into deleterious proteins.
While most spliceosome modulators vary greatly in terms of chemical structure, they have all been shown to have almost identical effects on spliceosome assembly 49,50 . A recent study has demonstrated that spliceostatin A, pladienolide B and herboxidiene all interact with the same site on SF3B1, and likely exhibit the same mechanism of action of inhibiting spliceosome assembly, possibly by interfering with SF3B1 conformational states 50 . Also, unlike the splicing modulator sudemycin which was recently shown to induce a specific antitumor response in chronic lymphocytic leukemia through alternative splicing of MCL1 to generate its proapoptotic isoform 51 , no such effect was observed in MCF-7 and MDA-MB-468 treated with JA ( Supplementary Fig. S9). Whether JA interacts with SF3B1 in a similar manner to that of the aforementioned spliceosome modulators remains to be further investigated. However, based on the distinct mechanism of JA in stabilizing SF3B1, rather than inhibition as demonstrated by other spliceosome modulators, it is likely that JA might engage a novel mechanism to regulate the cancer spliceosome activities. Hence, further understanding of how JA might interact with SF3B1 will be invaluable for the design and further development of new drug candidates.
In conclusion, our findings indicate that JA evokes tumor-specific activity by targeting SF3B1 and causing abnormal splicing patterns, in addition to its previously reported activity of inhibiting microtubule polymerization. While the precise mechanism by which JA inhibits splicing activity has yet to be elucidated, the fact that it is multi-targeted as well as tumor-specific makes it an attractive candidate for further development. More comprehensive understanding of these mechanisms, coupled with knowledge of the spliceosome functions in tumors are important for the further development of this novel class of experimental antitumor agents.

Chemicals and reagents. Jerantinine A was isolated from the leaf ethanolic extract of the Malayan
Tabernaemontana corymbosa Roxb. ex Wall and characterized as reported previously 25 .
Cell lines and cell culture. The human breast cancer cell lines (MDA-MB-468, T47D and MCF7) were obtained from American Type Culture Collection (ATCC) and were maintained in RPMI-1640 medium containing 10% fetal bovine serum (FBS), 100 IU/ml of penicillin, and 100 μ g/ml of streptomycin (Sigma-Aldrich, St. Louis, MO, USA). The non-transformed human mammary epithelial cell line MCF-10A was cultured in DMEM/ F12 (Invitrogen, Carlsbad, CA, USA) supplemented with 5% horse serum, 20 ng/ml epidermal growth factor (EGF), 0.5 μ g/ml hydrocortisone, 10 μ g/ml insulin, 100 IU/ml penicillin, and 100 μ g/ml streptomycin. All cells were maintained at 37 °C in 5% CO 2 . Cells were passaged for less than 6 months and no further authentication was performed by the authors.
Cell proliferation assay. Dose-response curves and IC 50 values were determined using the 3-(4,5-dime thylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) cell proliferation assay, as described previously 52,53 . Pooled genome-wide shRNA library screen. MCF-7 cells were transduced with MISSION LentiPlex Human Pooled shRNA Library, targeting more than 15000 human genes (Sigma-Aldrich, USA). Tranduced cells were selected with 1 μ g/mL puromycin for 4 days and treated with vehicle or 1 μ M of JA for 72 h. Genomic DNA was isolated using the GenElute Mammalian Genomic DNA Miniprep Kit (Sigma, USA) according to the manufacturer's protocol. shRNA sequences present in each sample were determined by next-generation sequencing (NGS). Massively parallel sequencing for shRNA retrieval and screen data analysis were described in Supplementary Methods. Quantitative real-time PCR analysis. mRNA expression levels were measured by real-time quantitative PCR (qPCR) as described previously 54,55 . Briefly, total cellular RNA was extracted using the Qiagen RNA Isolation Kit (Qiagen, Valencia, CA, USA). First-strand cDNA was synthesized using the High Capacity RNA-to-cDNA Master Mix (Applied Biosystems, Foster City, CA, USA) according to the manufacturer's instructions. Gene expression levels were measured by qPCR using the FastStart Universal SYBR Green Master (Roche Diagnostics, Indianapolis, IN, USA) and a Biorad iQ5 real-time PCR detector system (Bio-Rad, Hercules, CA, USA). Spliceosome activity was estimated by amplification of the intronic regions of DNAJB1, CDKN1B, RIOK3 and BRD2, as described previously 51,56,57 . qPCR conditions for all amplifications are as follows: 3 min at 94 °C followed by 40 cycles of 94 °C for 40 sec, 60 °C for 40 sec, and 72 °C for 25 sec. Gene expression data were normalized to GAPDH as the house-keeping gene. The specific primers are shown in Supplementary Table S7.
Lentivirus production and transduction. Lentiviral non-targeting shRNA (NS) and shRNA constructs targeting SF3B1 and SF3B3 were purchased from Sigma-Aldrich (St. Louis, MO, USA). The shRNAs target sequences were presented in Supplementary Table S8. Lentivirus production and transduction were performed as described previously 52,54 . Briefly, high-titer lentiviral stocks were generated by co-transfection with packaging plasmids psPAX2 (Addgene plasmid #12260) and envelope plasmids pMD2.G (Addgene plasmid #12259) into HEK-293T using CalPhos Transfection Kits (Clontech, Mountain View, CA, USA). Supernatants containing lentiviral stocks were supplemented with polybrene (Sigma, USA) and used for transduction of target cells. Stable pools were generated by transduction of two independent lentiviral shRNA constructs targeting SF3B1 followed by brief selection with puromycin (Sigma, USA).