PPP1R1B-STARD3 Chimeric Fusion Transcript in Human Gastric Cancer Promotes Tumorigenesis Through Activation of PI3K/AKT Signaling

ORIGINAL ARTICLE PPP1R1B-STARD3 chimeric fusion transcript in human gastric cancer promotes tumorigenesis through activation of PI3K/AKT signaling

SM Yun1,2,7, K Yoon1,7, S Lee3,4,7, E Kim1,7, S-H Kong5,6, J Choe1, JM Kang1, T-S Han5, P Kim1, Y Choi1, S Jho3,HYoo1, J Bhak3,4, H-K Yang5,6 and S-J Kim1,2

Fusion genes act as potent oncogenes, resulting from chromosomal rearrangements or abnormal transcription in many human cancers. Although multiple gastric cancer genomes have been sequenced, the driving recurrent gene fusions have not been well characterized. Here, we used paired-end transcriptome sequencing to identify novel gene fusions in 18 human gastric cancer cell lines and 18 pairs of primary human gastric cancer tissues and their adjacent normal tissues. Multiple samples revealed expression of PPP1R1B-STARD3 fusion transcript. The presence of PPP1R1B-STARD3 correlated with elevated levels of PPP1R1B mRNA. PPP1R1B- STARD3 fusion transcript was detected in 21.3% of primary human gastric cancers but not in adjacent matched normal gastric tissues. Based on reverse transcription PCR analysis of DNA, unlike other fusions described in gastric cancer, the PPP1R1B-STARD3 appears to be generated by RNA processing without chromosomal rearrangement. Overexpression of PPP1R1B-STARD3 in MKN-28 signiﬁcantly increased cell proliferation and colony formation. This increased proliferation was mediated by activation of phosphatidylinositol-3-kinase (PI3K)/AKT signaling. Furthermore, expression of PPP1R1B-STARD3 fusion transcript enhanced the tumor growth of MKN-28 cells in athymic nude mice. These ﬁndings show that PPP1R1B-STARD3 fusion transcript has a key role in subsets of gastric cancers through the activation of PI3K/AKT signaling.

Oncogene (2014) 33, 5341–5347; doi:10.1038/onc.2013.472; published online 25 November 2013 Keywords: fusion transcript; gastric cancer; transcriptome sequencing

INTRODUCTION chimeric transcript that starts in the first exon of the upstream Gastric cancer was the fourth most common cancer diagnosis gene and stops at the termination of the downstream gene with worldwide in men and fifth most common in women in 2011.1 It is the splicing region between two genes. This mechanism can be a leading cause of global cancer mortality, with a 5-year survival explained as a ‘run-off’ transcription of the upstream gene or rate of B27%. Approximately, 70% of new cases occur in ‘leakage’ accidence of the transcriptional machinery. And these developing countries, including Eastern Asia and many parts of events create a frameshift, causing a premature stop of the South America.2 The main risk factors of gastric cancer are fusion gene.12,13 The other mechanism is trans-splicing, which Helicobacter pylori infection, dietary factors and smoking. In appears between the separate pre-mRNA of tandem genes. addition, the result of multistep conversion of normal cell to Chimeric transcripts give rise to a diverse range of functional malignant cell is due to genetic and epigenetic alterations of cell protein by generating chimeric protein and altered regulation of adhesion molecules, oncogenes, cytokines and tumor-suppressor mRNA in normal human tissues.14 However, in carcinogenesis, the genes such as CDH1,3 ERBB2 and KRAS,4 TLR4,5 p536 and RUNX3,7 accumulation of aberrant expression of chimeric transcript causes respectively. dysfunction of protein by dysregulation of RNA processing.15,16 Fusion genes, caused by genomic aberrations, such as PPP1R1B was initially identified in the brain and has been translocation, amplification and rearrangement, are known to implicated in physiologic processes, including neuronal disease trigger human cancers.8–10 Fusion transcript is a chimeric RNA and dopaminergic abnormality.17–19 PPP1R1B has also been from two different genes. The fusion transcript unrelated to involved in common carcinomas,20 including gastric cancer.21,22 chromosomal rearrangements often occurs by transcription- PPP1R1B-STARD3 fusion transcript was first reported in breast induced chimeras. Transcription-induced chimeras can be cancer.23 However, the function of PPP1R1B-STARD3 fusion in generated from two genes, which are located up to 50 kb from cancers has not been investigated. each other, and are more likely to occur between genes of a short Here, we identified PPP1R1B-STARD3 fusion in gastric cancer distance.11 The two possible mechanisms of transcription-induced using next-generation sequencing. The PPP1R1B-STARD3 expres- chimeras in the formation of the gene fusions contribute to their sion was validated in gastric cell lines and human gastric patient biological significance. One mechanism is read-through, which is a tissues by using reverse transcription PCR and quantitative

1CHA Cancer Institute, CHA University, Seoul, Korea; 2Department of Biomedical Science, College of Life Science, CHA University, 605 Yeoksam-dong, Gangnam-gu, Seoul, Korea; 3Personal Genomics Institute, Genome Research Foundation, Suwon, Korea; 4TheragenEtex Bio Institute, TheragenEtex Inc., Suwon, Korea; 5Cancer Research Institute, Seoul National University, Seoul, Korea and 6Department of Surgery, Seoul National University College of Medicine, Seoul, Korea. Correspondence: Professor S-J Kim, CHA Cancer Institute, CHA University, 605 Yeoksam 1-Dong, Seoul 135-081, Korea. E-mail: [email protected] 7These authors contributed equally to this work. Received 30 May 2013; revised 24 September 2013; accepted 4 October 2013; published online 25 November 2013 PPP1R1B-STARD3 chimeric fusion transcript promotes tumorigenesis SM Yun et al 5342 real-time PCR (Q-PCR). PPP1R1B-STARD3 increased cell prolifera- of the genomes (Supplementary Figure S1d). PPP1R1B gene was tion through the phosphatidylinositol-3-kinase (PI3K)/AKT path- highly expressed without relation to fusion junction, whereas way in vitro. And PPP1R1B-STARD3 affected the tumorigenic STARD3 expression increased from exon 2 in comparison to exon capacity of gastric cancer cells in vivo. 1, which was breakpoint junction of the fusion transcript. This suggested that abnormal overexpression of STARD3 was rendered by PPP1RB-STARD3 fusion transcript. RESULTS Discovery of the PPP1R1B-STARD3 fusion transcript in gastric cancer Expression of PPP1R1B-STARD3 fusion gene in primary gastric To investigate fusion genes in gastric cancer, we conducted cancer tissues and gastric cancer cell lines paired-end transcriptome sequencing with 18 gastric cancer cell To confirm whether PPP1R1B-STARD3 was expressed in other lines and 18 pairs of primary human gastric cancer tissues gastric cell lines and clinical specimens, we performed reverse and their adjacent normal tissues. The deFuse algorithm (version transcription PCR and Q-PCR screening of PPP1R1B-STARD3 with 0.4.1) was used with default parameters to identify fusion 18 gastric cancer cell lines and 47 pairs of primary human gastric transcripts. PPP1R1B-STARD3 fusion gene was identified in cancer tissues and their adjacent normal tissues. PPP1R1B-STARD3 NCI-N87 gastric cancer cell line. The sequence reads were aligned was expressed in about 33% of gastric cancer cell lines (6/18), with the reference transcriptome to determine the fusion junction which correlated to fragments per kilobase per million read (Figures 1a and b, Supplementary Figures S2a and b). PPP1R1B and (FPKM) values estimated by RNA sequencing (Figures 2a and b, STARD3 lay adjacent to each other on chromosome 17, separated Supplementary Figures S2a and d, and Table 1). This fusion by 455 bp. PPP1R1B-STARD3 fusion transcript is made of PPP1R1B transcript was generated by RNA processing without DNA-level exons 1–6 and STARD3 exons 2–15. A fusion of the complete exon rearrangement (Supplementary Figure S2b). Interestingly, the 6ofPPP1R1B with the beginning of exon 2 of STARD3 resulted in a PPP1R1B-STARD3 fusion transcript was differentially expressed in read-through fusion protein (Figure 1c). It is predicted that tumors compared with normal samples. PPP1R1B-STARD3 was translation initiation from an ATG site in PPP1R1B exon 1 could expressed in about 21.3% of gastric tumor tissues (10/47), whereas produce a 245-amino-acid protein, comprising of 188 amino acids expression was not detected in their adjacent normal tissues of PPP1R1B and 57 amino acids from STARD3. The latter resulted (Figure 2d, Supplementary Figures S2c–e, and Table 1). from an out-of-frame of STARD3. Translation from this fusion PPP1R1B and STARD3 genes are located within PPP1R1B-STARD3- apparently produced loss of c-terminal sequences from the full ERBB2-GRB7 amplicon on human chromosome 17q 12. It was 204 amino acid of PPP1R1B protein but retained all functionally reported that PPP1R1B, STARD3, ERBB2 and GRB7 are frequently relevant domains (Supplementary Figure S1c). We estimated the co-amplified in human gastric cancer.24 To confirm if PPP1R1B- RNA expressions of PPP1R1B and STARD3 by read distributions. STARD3 transcript was to be generated by this amplicon, we Read sequences were annotated to mRNAs (Figure 1d) and exons analyzed expression of these genes in gastric cancer cell lines.

Figure 1. Discovery of PPP1R1B-STARD3 fusion transcript in gastric cancer. (a) Identiﬁcation of PPP1R1B-STARD3 fusion transcript by RNA sequencing. The sequences of spanning reads are displayed across the fusion junction. Black and blue indicate PPP1R1B exon 6–7 and STARD3 exon 2, respectively. (b) The number of paired-end reads whose 50 and 30 end sequences are successfully aligned to each of fused genes across the exon junction of PPP1R1B and STARD3 and spanning reads are shown. (c) Schematic representations of the wild-type PPP1R1B and STARD3 genes with breakpoints of respective exon numbers and mRNA positions for each gene. The fusion occurred between the sixth exon of PPP1R1B and the second exon of STARD3. Red line indicates the position of start and stop codon of the predicted fusion protein. The chromatogram by Sanger sequencing presents the fusion junction. (d) RNA expression levels of PPP1R1B and STARD3 are estimated through read distributions, which were annotated to their mRNAs.

Oncogene (2014) 5341 – 5347 & 2014 Macmillan Publishers Limited PPP1R1B-STARD3 chimeric fusion transcript promotes tumorigenesis SM Yun et al 5343

Figure 2. PPP1R1B-STARD3 expression in gastric cancer cell lines and primary gastric cancer tissues. (a) RNA transcriptional levels of PPP1R1B- STARD3, PPP1R1B and STARD3 in 18 gastric cancer cell lines were measured by Q-PCR. (b) The levels of RNA expression of PPP1R1B and STARD3 were estimated by RNA sequencing. (c and e) Comparison of PPP1R1B and STARD3 expressions in 18 gastric cancer cell lines (c) and 47 gastric cancer patient tumor tissues (e). *Po0.001 and **Po0.0006, which is expression of fusion transcript versus PPP1R1B in gastric cancer cell lines and gastric tumor tissues, respectively. NS indicates not signiﬁcant. (d) The representative Q-PCR result of PPP1R1B-STARD3 and each parental gene expressions in 47 pairs of primary human gastric cancer tissues and their adjacent normal tissues was shown.

lines such as breast cancer, endometrial cancer and ovarian cancer Table 1. The frequency of fusion gene in 18 gastric cancer cell lines (Supplementary Figure S4). We observed that a positive correlation and in 47 pairs of primary human gastric cancer tissues and their between PPP1R1B-STARD3 chimeric and PPP1R1B expressions in adjacent normal tissues several cancer cell lines was consistent with that in gastric cancer. The expression of PPP1R1B-STARD3 was speciﬁc not only in gastric Fusion Fusion Total Frequency positive negative (n) (%) cancers but also in other cancers, implying that this fusion transcript (n) (n) may be involved in cancer development.

Gastric cancer cell lines 6 12 18 33.3 Induction of cancer cell proliferation and colony formation by Gastric cancer Normal 0 47 47 0 PPP1R1B-STARD3 fusion gene Patient tissues Tumor 10 37 47 21.3 As demonstrated in Figure 2a, we observed PPP1R1B-STARD3 in six human gastric cancer cell lines. We confirmed endogenous protein expression of PPP1R1B-STARD3 fusion gene in the cell lines, which showed a high level of fusion expression in NCI-N87, Cell lines with high expression of PPP1R1B-STARD3, excluding KATOIII and MKN-45 (Figure 3a). Given that the dysregulation of NCI-N87, did not show high expression of the amplicon proliferation is an important step for cancer cell development, we (Supplementary Figures S3a and b). We also analyzed copy next investigated whether the ectopic expression of the fusion number variations of PPP1R1B-STARD3-ERBB2-GRB7 amplicon by gene influenced cell proliferation. Using retroviral constructs, we whole-genome sequencing.25 However, copy number gain or loss stably expressed empty vector, PPP1R1B or PPP1R1B-STARD3 in was not observed in the PPP1R1B-STARD3-ERBB2-GRB7 amplicon MKN-28, a fusion negative cell line. As shown in Figure 3b, the (Supplementary Figure S3c). Therefore, it was not possible that expression of the fusion gene at the protein level was confirmed fusion transcripts arise from cryptic genomic rearrangement by by immunoblotting. MKN-28 expressing PPP1R1B-STARD3 had the amplicon. significantly increased in proliferation rate when compared with Moreover, cells expressing high levels of PPP1R1B wild-type vector controls (Figure 3c). In addition, PPP1R1B-STARD3-expres- transcript had a tendency to be PPP1R1B-STARD3 fusion positive. As sing MKN-28 cells grew at a higher density and formed more shown in Figures 2c and e, a highly significant correlation was found foci during the focus-forming assay (Figure 3d). We obtained between expression of PPP1R1B parental gene and fusion transcript. similar results from AGS cells expressing PPP1R1B-STARD3 This fusion transcript was identified in several different cancer cell (Supplementary Figure S5). Overall, these results suggested that

& 2014 Macmillan Publishers Limited Oncogene (2014) 5341 – 5347 PPP1R1B-STARD3 chimeric fusion transcript promotes tumorigenesis SM Yun et al 5344

Figure 3. Characterization of PPP1R1B-STARD3 fusion gene. (a) Endogenous expressions of PPP1R1B wild-type (WT; lower bands) and PPP1R1B-STARD3 (upper bands) in fusion-positive and -negative gastric cancer cell lines. (b) Western blot analysis showing the levels of PPP1R1B and PPP1R1B-STARD3 in empty vector, PPP1R1B- or PPP1R1B-STARD3-expressing MKN-28 stable cells. (c) Proliferation assay of three stable cell lines was performed using Cell Titer-Glo Luminescent assay kit (*Po0.05; **Po0.005). (d) Focus-forming assay demonstrates signiﬁcant increase in relative cell survival in MKN-28 cells stably expressing PPP1R1B-STARD3 (top). Quantitative result of focus formation assays was displayed (bottom). (e) MKN-28 cells stably expressing empty vector, PPP1R1B or PPP1R1B-STARD3 were injected subcutaneously in to nude mice. (f) Tumor weights were signiﬁcantly higher in the PPP1R1B-STARD3 group. *Po0.05, which is fusion versus PPP1R1B WT expressing tumors. **Po0.0001, which is PPP1R1B WT versus empty vector expressing tumors. (g) Tumor volume was measured at the indicated times. Each data point represents the mean (±) standard deviation for six xenografts (*P ¼ 0.028; **P ¼ 0.0044; ***P ¼ 0.0001).

expression of exogenous PPP1R1B-STARD3 fusion gene conferred activation is due to a shorter PPP1R1B protein, or if the additional a growth and a proliferative advantage in gastric cancer. amino acids derived from STARD3 sequences have any role, we constructed truncated form of PPP1R1B (PPP1R1B truncated), Promotion of tumor growth in vivo by PPP1R1B-STARD3 fusion which deleted the STARD3 part in the PPP1R1B-STARD3 fusion protein sequences. Activation of AKT phosphorylation by truncated form of PPP1R1B was similar to that of PPP1R1B wild type Overexpression of PPP1R1B showed an oncogenic potential in NIH (Supplementary Figure S7b). PPP1R1B wild type and truncated 3T3 cells and induced tumorigenesis in gastric cancer.22 To form of PPP1R1B induced proliferation of MKN-28 similarly determine if overexpression of the fusion gene increased growth compared with vector control, whereas PPP1R1B-STARD3 further in vivo, we used the xenograft mouse model. MKN-28 cells stably increased proliferation (Supplementary Figure S7c). This correlated expressing empty vector, PPP1R1B or PPP1R1B-STARD3 were well with AKT activation levels induced by these proteins. As AKT injected subcutaneously (5 Â 106 cells per site) in to nude mice, activation level induced by PPP1R1B-STARD3 fusion protein is and the growth of the tumors was monitored (Figure 3e). These greater than those by PPP1R1B wild type and truncated form of recurrent tumors displayed high-level expression of the PPP1R1B PPP1R1B, it is likely that the sequences derived from STARD3 part and PPP1R1B-STARD3 (Supplementary Figure S6). As shown in of PPP1R1B-STARD3 fusion sequences may potentiate the ability Figures 3f and g, MKN-28 cells with PPP1R1B-STARD3 expression of PPP1R1B to activate AKT. Because elevated levels of PPP1R1B- generated significantly larger tumors, further confirming that STARD3 transcript were associated with elevated levels of PPP1R1B PPP1R1B-STARD3 reinforced tumorigenicity in vivo. wild-type transcript, we compared AKT activation in MKN-28 cells stably expressing PPP1R1B-STARD3 alone or PPP1R1B wild type Regulation of the PI3K/AKT survival pathway by PPP1R1B-STARD3 and PPP1R1B-STARD3 together to test whether expression of both fusion gene in gastric cancer cells PPP1R1B wild type and PPP1R1B-STARD3 further enhances AKT It has been previously demonstrated that PPP1R1B activates AKT phosphorylation. Both wild-type and fusion overexpressing cells phosphorylation.26 Therefore, we examined whether PPP1R1B- showed increased AKT activation greater than wild type or fusion STARD3 also activates AKT phosphorylation. The activation of the alone overexpressing cells (Supplementary Figure S7d). PI3K/AKT signaling cascades regulates cell proliferation, survival, To further test whether induction of AKT phosphorlyation was angiogenesis, invasion and metastasis.27,28 To test whether involved in the PI3K/AKT pathway, we treated cells with PI3K PPP1R1B-STARD3 is involved in the PI3K/AKT signaling pathway, inhibitor, LY294002. The result showed that AKT activation was we assessed the activation of AKT in three cell lines. Intriguingly, completely suppressed following LY294002 treatment (Figure 4b AKT phosphorlyation was induced in PPP1R1B and even and Supplementary Figure S7e). To confirm that the repression of more significantly so in PPP1R1B-STARD3 stable cells (Figure 4a AKT activation was due to PIK3CA inhibition and not an off-target and Supplementary Figure S7a). To clarify if the enhanced AKT effect, we treated the stable cells with another AKT inhibitor,

Oncogene (2014) 5341 – 5347 & 2014 Macmillan Publishers Limited PPP1R1B-STARD3 chimeric fusion transcript promotes tumorigenesis SM Yun et al 5345

Figure 4. Induction of the PI3K-AKT activation by PPP1R1B-STARD3 in gastric cancer cells. (a) Western blot analysis of AKT and phospho-AKT (S473) in empty vector, PPP1R1B- or PPP1R1B-STARD3-expressing MKN-28 stable cells. (b) AKT activations with or without PI3K inhibitor, LY294002, treatment (50 mM) in three stable cell lines were examined by western blotting. (c) Overexpression of PPP1R1B-STARD3 in MKN-28 cells signiﬁcantly reduced cell proliferation with (red lines) or without (black lines) LY294002 (10 mM) treatment for 4 days (*P ¼ 0.0036; **P ¼ 0.0022; ***Po0.0001). (d) Focus-forming assays demonstrate signiﬁcant reduction in relative cell survival in the MKN-28 cells stably expressing PPP1R1B-STARD3 after LY294002 (25 mM) treatment for 14 days (left panel). Quantitative result of focus formation assays was shown (right panel; *P ¼ 0.0023; **P ¼ 0.0001; ***Po0.0001). DMSO, dimethyl sulfoxide; WT, wild type.

AKT-IV. The result also demonstrated that AKT activation was mechanisms were generated by two genes. The biological completely suppressed by AKT-IV treatment (Supplementary significance of chimeric RNAs would also be similar to that of Figure S7f). We next examined proliferation rate of three stable gene fusions.15 And, chimeric RNAs, like gene fusions, are cell lines with or without LY294002 stimulation to explore a expected to increase the diversity of proteins in normal cells.14 growth advantage of PPP1R1B-STARD3 through the PI3K/AKT On the other hand, abnormal expression of chimeric transcripts pathway. Even though proliferation of empty vector and PPP1R1B provided cell growth and survival in human cancer.30 PPP1R1B- wild-type cells slightly increased, the proliferation reduction in the STARD3 chimeric fusion transcripts were expressed far higher rate fusion was observed at a concentration of 10 mM of LY294002 after than a few gene fusion in gastric cancer cell lines and human 96 h (Figure 4c). To further confirm the enhanced proliferative gastric patient cancer tissues at a relatively high percentage of response of PPP1R1B-STARD3 in MKN-28, we performed focus- 33.3% and 21.3%, respectively.31,32 Therefore, a high occurrence forming assay after LY294002 treatment for 14 days. Consistently, rate of PPP1R1B-STARD3 may be a more useful target in cancer inhibition of AKT signaling significantly decreased cell foci diagnosis and therapy compared with that of low occurrence formation (Figure 4d), indicating that PPP1R1B-STARD3 has an fusion genes. important role in gastric cancer development through the PI3K/ Our experiments investigated the mechanism and function AKT pathway. of PPP1R1B-STARD3 in gastric cancer. Intriguingly, the Q-PCR results showed that PPP1R1B parental gene involved in the expression of chimeric transcript was significantly overexpressed in gastric DISCUSSION cancers, whereas the expression level of STARD3 did not show This study provides a new insight into chimeric RNAs expressed in any correlation with the fusion transcript (Figures 2c and e). This human gastric cancer. We identified PPP1R1B-STARD3 as a novel expression pattern in other cancer cell lines was consistent with chimeric transcript in gastric cancer cell lines and tumors using that of gastric cancer cell lines (Supplementary Figure S4). Random paired-end transcriptome sequencing. PPP1R1B was found to be transcriptional leakage is likely to generate more events in highly overexpressed in several human cancers including gastric expressed genes. Furthermore, given that PPP1R1B parental gene cancer,26,29 which implied the involvement of PPP1R1B-STARD3 was differentially expressed in gastric cancers in comparison to in carcinogenesis. Even though PPP1R1B-STARD3 fusion gene was matched normal samples, it was speculated that PPP1R1B first identified in HCC1569 breast cancer cell line,23 the biological regulatory factors modulated both parental and fusion genes. The function of PPP1R1B-STARD3 has not yet been examined in cancer tumor-associated aberrant regulation of cis-ortrans-elements in development. transcriptional regulation level potentially affects their gene A fusion of the complete exon 6 of PPP1R1B with the beginning expression and biological functions.30 These elements such as of exon 2 of STARD3 resulted in a read-through fusion protein. suppressors or regulators may influence the promoter or PPP1R1B-STARD3 chimeric transcript was not generated by the termination of PPP1R1B in transcription and RNA splicing. PPP1R1B-STARD3-ERBB2-GRB7 amplicon (Supplementary Expression of PPP1R1B-STARD3 markedly enhanced proliferation Figure S3). Moreover, RNA processing not DNA rearrangement as well as tumorigenicity of gastric cancer cells compared with produced this transcript (Supplementary Figure S2b). Chimeric PPP1R1B wild type. In this study, we found that PPP1R1B-STARD3 RNAs were similar to gene fusions in the sense that the activates the PI3K/AKT signaling pathway (Figure 4). As PPP1R1B

& 2014 Macmillan Publishers Limited Oncogene (2014) 5341 – 5347 PPP1R1B-STARD3 chimeric fusion transcript promotes tumorigenesis SM Yun et al 5346 wild type was known to activate AKT signaling by promoting Reagents and antibodies 26 interaction between EGFR and ERBB3, it is likely that the function LY294002, PI3K/AKT inhibitor was purchased from SelleckBio (Houston, TX, of PPP1R1B-STARD3 retained all functionally relevant domain of USA). Antibodies were obtained from Abcam (Cambridge, MA, USA) (anti- PPP1R1B is similar to PPP1R1B. Interestingly, the effect of PPP1R1B), Cell Signaling Technology (Cell signaling, Inc, Danvers, MA, USA) PPP1R1B-STARD3 on cell proliferation and tumorigenesis was (anti-pAKT Ser473 and anti-AKT) and Sigma-Aldrich (Milwaukee, WI, USA) slightly greater than that of PPP1R1B. Therefore, we cannot (anti-a-tubulin). exclude the possibility that 57 amino acids of novel sequences from STARD3 may also contribute to gastric tumorigenesis. Retroviral production and infection Taken together, PPP1R1B-STARD3 fusion transcript promotes The Platinum-GP (Plat-GP) cells were transfected with pMX-puro empty tumorigenesis in gastric cancer. Moreover, the fusion gene vector, pMX-puro-PPP1R1B wild-type or pMX-puro-PPP1R1B-STARD3-expres- increased cell proliferation through activation of the PI3K/AKT sing vector using FuGENE 6 (Roche Applied Science, Indianapolis, IN, USA). pathway. We suggest that PPP1R1B-STARD3 may be a candidate After 48 h, viral supernatants were collected for infection of MKN-28 cells with for a new genetic and epigenetic marker in gastric cancer. polybrene (10 mg/ml; Sigma-Aldrich). Puromycin-resistant transfectants were selected and maintained in RPMI with puromycin (4 mg/ml).

MATERIALS AND METHODS Cell proliferation assay Primary tissues and cell lines Proliferation studies were performed using cell counts and Cell Titer-Glo Luminescent Cell Viability Assay (G7572; Promega, Madison, WI, USA). For the Eighteen pairs of gastric cancer and normal matched control tissues for cell count experiment, MKN-28 cells stably expressing empty vector, PPP1R1B RNA sequencing were obtained from the Gastrointestinal Division in or PPP1R1B-STARD3 were plated with 1 Â 104 cells per well in a 12-well plate Department of Surgery at Seoul National University Hospital. The usage of containing RPMI medium with 10% fetal bovine serum and incubated the patient tissues and related clinicopathologic data, and the protocols overnight. Subsequently, cells were incubated with or without LY294002 for and informed consent documents for this study were approved by the an additional 72 h. After trypsinization, cells were stained with trypan blue institutional review board of the Seoul National University Hospital (IRB no. and counted using a Neubauer chamber (Celeromics, Valencia, Spain). The H-0806-072-248). The clinical information of the gastric cancer tissues is mean±s.d. of at least three independent experiments was displayed. provided in Supplementary Table S1. Eighteen human gastric cancer cell For Cell Titer-Glo Luminescent Assay, MKN-28 cells stably expressing lines, purchased from Korean Cell Line Bank, were used in this study (SNU- empty vector, PPP1R1B or PPP1R1B-STARD3 were plated with 3 Â 103 cells 1, SNU-5, SNU-16, SNU-216, SNU-484, SNU-520, SNU-601, SNU-620, SNU- per well in a 96-well plate in RPMI medium with 10% fetal bovine serum 638, SNU-668, SNU-719, MKN-1, MKN-28, MKN-25, MKN-74, KATOIII, AGS and incubated overnight. And then, cell proliferation was measured every and NCI-N87). These gastric cancer cell lines were maintained with RPMI day according to the manufacturer’s instructions. All experiments were (WelGENE, Daegu, Republic of Korea) containing 25 mM HEPES, 10% fetal carried out in triplicates. bovine serum (WelGENE) and 1% penicillin/streptomycin (WelGENE).

Focus-forming assay RNA sequencing Five hundred cells were seeded per well in six-well plates. After incubation Total RNA for RNA sequencing was isolated by TRIzol Reagent (Invitrogen, for 2 weeks, colonies were rinsed with PBS and stained with 2% methylene Carlsbad, CA, USA) according to the manufacturing instructions. And then, blue. All experiments are carried out in triplicates. miRNeasy (Qiagen, Hilden, Germany) were used for the purification of total RNA after DNase I treatment following their specific protocol. In vivo experiments Illumina platform for analyzing fusion genes of gastric cancer MKN-28 cell lines stably expressing empty vector, PPP1R1B or PPP1R1B- transcriptomes with 90 bp paired-end library were used according to the 6 manufacturer’s instructions (Illumina, San Diego, CA, USA). STARD3 fusion gene were injected subcutaneously (5 Â 10 cells per site) in to nude mice. Mice were killed 4 weeks after injection, and tumors were isolated. Tumor volume (V) was calculated by using the formula (LXSXS) Â 0.5, where L and S were the long and short dimensions, Reads alignment and fusion gene detection from RNA sequencing respectively. All animals were maintained according to the CHA Hospital data Animal Care and Use Committee guidelines under protocol number Paired-end sequence reads were aligned to hg19 human reference IACUC120026. genome (NCBI build 37) with a TOPHAT algorithm (ver. 2.0.4).33 Transcript assembly and gene expression level estimation were 34 conducted by using Cufflinks ver. 2.0.2, which calculates the fragments CONFLICT OF INTEREST per kilobase per million reads. Gene model information from Ensembl release 62 (http://asia.ensembl.org/info/data/ftp/index.html) was used in The authors declare competing financial interests. Some authors are present both Tophat and Cufflinks algorithms. employees of TheragenEtex, and some of them have personal financial interests as Fusion genes were detected using RNA-seq data by the deFuse shareholders in TheragenEtex. algorithm35 ver.0.4.1 with a human genome sequence and gene model from Ensembl release 62, UCSC hg19 repeat data, and NCBI unigene data. Supporting reads, that is, split reads and spanning reads, of a gene fusion ACKNOWLEDGEMENTS point were selected by a script ‘get_reads.pl’ of the defuse package. This work was supported in part by the National Research Foundation of Korea (NRF) grants (2009-0081756 and 2012M3A9C4048736 to S-JK). JB and SL were supported by TheragenEtex and Genome Research Foundation internal funds. JB was supported by Validation of fusion gene expression the Industrial Strategic Technology Development Program (10040231) funded by the Expression levels of PPP1R1B, STARD3 and PPP1R1B-STARD3 were Ministry of Knowledge Economy (MKE, Korea). determined by reverse transcription PCR, and Q-PCR using SYBR Green assay. All oligonucleotide primers, listed in Supplementary Table S2, were synthesized by Bioneer Company. Samples were normalized against 18S AUTHOR CONTRIBUTIONS rRNA, and each target was measured in triplicates. Total RNA was extracted SMY and KY performed research, analyzed data and wrote the manuscript; from human gastric patient tissues with Tri reagent (Molecular Research Center, Inc., Cincinnati, OH, USA) according to the manufacturer’s SL analyzed and interpreted sequencing data; EK and PK performed research instructions. SuperScript II (Invitrogen) was used for first-strand cDNA and collected data; JMK performed an animal study; JC, YC and HY contributed synthesis following the manufacturing instructions. Genomic DNAs of 18 to the manuscript; T-SH, S-HK and H-KY collected clinical information and gastric cancer cell lines were isolated by QIAmp DNA mini kit (Qiagen) prepared clinical samples; SJ and JB analyzed and collected sequencing data; according to the manufacturer’s instructions. S-JK designed research, interpreted data and drafted the manuscript.

Oncogene (2014) 5341 – 5347 & 2014 Macmillan Publishers Limited PPP1R1B-STARD3 chimeric fusion transcript promotes tumorigenesis SM Yun et al 5347 REFERENCES 19 Fienberg AA, Hiroi N, Mermelstein PG, Song W, Snyder GL, Nishi A et al. DARPP-32: 1 Jemal A, Bray F, Center MM, Ferlay J, Ward E, Forman D. Global cancer statistics. regulator of the efficacy of dopaminergic neurotransmission. Science 1998; 281: CA Cancer J Clin 2011; 61: 69–90. 838–842. 2 Guggenheim DE, Shah MA. Gastric cancer epidemiology and risk factors. JSurg 20 Beckler A, Moskaluk CA, Zaika A, Hampton GM, Powell SM, Frierson HF Jr et al. Oncol 2013; 107: 230–236. Overexpression of the 32-kilodalton dopamine and cyclic adenosine 3 Guilford P, Hopkins J, Harraway J, McLeod M, McLeod N, Harawira P et al. E- 30,50-monophosphate-regulated phosphoprotein in common adenocarcinomas. cadherin germline mutations in familial gastric cancer. Nature 1998; 392: 402–405. Cancer 2003; 98: 1547–1551. 4 Deng N, Goh LK, Wang H, Das K, Tao J, Tan IB et al. A comprehensive survey of 21 Belkhiri A, Zaika A, Pidkovka N, Knuutila S, Moskaluk C, El-Rifai W. Darpp-32: genomic alterations in gastric cancer reveals systematic patterns of molecular a novel antiapoptotic gene in upper gastrointestinal carcinomas. Cancer Res 2005; exclusivity and co-occurrence among distinct therapeutic targets. Gut 2012; 61: 65: 6583–6592. 673–684. 22 Mukherjee K, Peng D, Brifkani Z, Belkhiri A, Pera M, Koyama T et al. Dopamine and 5 Hold GL, Rabkin CS, Chow WH, Smith MG, Gammon MD, Risch HA et al. A func- cAMP regulated phosphoprotein MW 32 kDa is overexpressed in early stages of tional polymorphism of toll-like receptor 4 gene increases risk of gastric carci- gastric tumorigenesis. Surgery 2010; 148: 354–363. noma and its precursors. Gastroenterology 2007; 132: 905–912. 23 Robinson DR, Kalyana-Sundaram S, Wu YM, Shankar S, Cao X, Ateeq B et al. 6 Wei J, Noto J, Zaika E, Romero-Gallo J, Correa P, El-Rifai W et al. Pathogenic Functionally recurrent rearrangements of the MAST kinase and Notch gene bacterium Helicobacter pylori alters the expression profile of p53 protein isoforms families in breast cancer. Nat Med 2011; 17: 1646–1651. and p53 response to cellular stresses. Proc Natl Acad Sci USA 2012; 109: 24 Katoh M. Evolutionary recombination hotspot around GSDML-GSDM locus is E2543–E2550. closely linked to the oncogenomic recombination hotspot around the PPP1R1B- 7 Ito K. [Tumor suppressive functions of RUNX3 in gastric carcinogenesis]. Seikagaku ERBB2-GRB7 amplicon. Int J Oncol 2004; 24: 757–763. 2012; 84: 278–282. 25 Yoon K, Lee S, Han TS, Moon SY, Yun SM, Kong SH et al. Comprehensive 8 Wang XS, Prensner JR, Chen G, Cao Q, Han B, Dhanasekaran SM et al. An inte- genome- and transcriptome-wide analyses of mutations associated with grative approach to reveal driver gene fusions from paired-end sequencing data microsatellite instability in Korean gastric cancers. Genome Res 2013; 23: in cancer. Nat Biotechnol 2009; 27: 1005–1011. 1109–1117. 9 Prensner JR, Chinnaiyan AM. Oncogenic gene fusions in epithelial carcinomas. 26 Zhu S, Belkhiri A, El-Rifai W. DARPP-32 increases interactions between epidermal Curr Opin Genet Dev 2009; 19: 82–91. growth factor receptor and ERBB3 to promote tumor resistance to gefitinib. 10 Mitelman F, Johansson B, Mertens F. The impact of translocations and gene Gastroenterology 2011; 141: 1738–1748. fusions on cancer causation. Nat Rev Cancer 2007; 7: 233–245. 27 Hennessy BT, Smith DL, Ram PT, Lu Y, Mills GB. Exploiting the PI3K/AKT pathway 11 Akiva P, Toporik A, Edelheit S, Peretz Y, Diber A, Shemesh R et al. Transcription- for cancer drug discovery. Nat Rev Drug Discov 2005; 4: 988–1004. mediated gene fusion in the human genome. Genome Res 2006; 16:30–36. 28 Courtney KD, Corcoran RB, Engelman JA. The PI3K pathway as drug target in 12 Zayed A, Couban S, Hayne O, Sparavalo N, Shawwa A, Sadek I et al. Acute human cancer. J Clin Oncol 2010; 28: 1075–1083. promyelocytic leukemia: a novel PML/RARalpha fusion that generates a frameshift 29 El-Rifai W, Smith MF Jr, Li G, Beckler A, Carl VS, Montgomery E et al. Gastric in the RARalpha transcript and ATRA resistance. Leuk Lymphoma 2007; 48: cancers overexpress DARPP-32 and a novel isoform, t-DARPP. Cancer Res 2002; 62: 489–496. 4061–4064. 13 Poulin F, Brueschke A, Sonenberg N. Gene fusion and overlapping reading 30 Wang K, Ubriaco G, Sutherland LC. RBM6-RBM5 transcription-induced chimeras frames in the mammalian genes for 4E-BP3 and MASK. J Biol Chem 2003; 278: are differentially expressed in tumours. BMC Genomics 2007; 8: 348. 52290–52297. 31 Palanisamy N, Ateeq B, Kalyana-Sundaram S, Pflueger D, Ramnarayanan K, 14 Li H, Wang J, Mor G, Sklar J. A neoplastic gene fusion mimics trans-splicing of Shankar S et al. Rearrangements of the RAF kinase pathway in prostate cancer, RNAs in normal human cells. Science 2008; 321: 1357–1361. gastric cancer and melanoma. Nat Med 2010; 16: 793–798. 15 Kannan K, Wang L, Wang J, Ittmann MM, Li W, Yen L. Recurrent chimeric RNAs 32 Tao J, Deng NT, Ramnarayanan K, Huang B, Oh HK, Leong SH et al. CD44-SLC1A2 enriched in human prostate cancer identified by deep sequencing. Proc Natl Acad gene fusions in gastric cancer. Sci Transl Med 2011; 3: 77ra30. Sci USA 2011; 108: 9172–9177. 33 Trapnell C, Pachter L, Salzberg SL. TopHat: discovering splice junctions with RNA- 16 Persson M, Andren Y, Mark J, Horlings HM, Persson F, Stenman G. Recurrent fusion Seq. Bioinformatics 2009; 25: 1105–1111. of MYB and NFIB transcription factor genes in carcinomas of the breast and head 34 Trapnell C, Williams BA, Pertea G, Mortazavi A, Kwan G, van Baren MJ et al. and neck. Proc Natl Acad Sci USA 2009; 106: 18740–18744. Transcript assembly and quantification by RNA-Seq reveals unannotated tran- 17 Brene S, Lindefors N, Ehrlich M, Taubes T, Horiuchi A, Kopp J et al. Expression of scripts and isoform switching during cell differentiation. Nat Biotechnol 2010; 28: mRNAs encoding ARPP-16/19, ARPP-21, and DARPP-32 in human brain tissue. 511–515. J Neurosci 1994; 14: 985–998. 35 McPherson A, Hormozdiari F, Zayed A, Giuliany R, Ha G, Sun MG et al. deFuse: 18 Greengard P, Allen PB, Nairn AC. Beyond the dopamine receptor: the DARPP-32/ an algorithm for gene fusion discovery in tumor RNA-Seq data. PLoS Comput Biol protein phosphatase-1 cascade. Neuron 1999; 23: 435–447. 2011; 7: e1001138.

Supplementary Information accompanies this paper on the Oncogene website (http://www.nature.com/onc)

& 2014 Macmillan Publishers Limited Oncogene (2014) 5341 – 5347