Recurrent BCAM-AKT2 fusion gene leads to a constitutively activated AKT2 fusion kinase in high-grade serous ovarian carcinoma Kalpana Kannana,b,1, Cristian Coarfab,c,1, Pei-Wen Chaoa,b, Liming Luoa,b, Yan Wanga,b, Amy E. Brinegara, Shannon M. Hawkinsc,d, Aleksandar Milosavljevicc,e, Martin M. Matzuka,b,c,e,f,g,2, and Laising Yena,b,c,g,2 Departments of aPathology & Immunology, bMolecular and Cellular Biology, dObstetrics and Gynecology, eMolecular and Human Genetics, and f Pharmacology, cDan L. Duncan Cancer Center, and gCenter for Drug Discovery, Baylor College of Medicine, Houston, TX 77030 Contributed by Martin M. Matzuk, January 30, 2015 (sent for review December 22, 2014; reviewed by Stephen Hillier and Barbara Vanderhyden)
High-grade serous ovarian cancer (HGSC) is among the most lethal forms of cancer in women. Excessive genomic rearrangements, which are expected to create fusion oncogenes, are the hallmark of this cancer. Here we report a cancer-specific gene fusion between BCAM, a membrane adhesion molecule, and AKT2, a key kinase in the PI3K signaling pathway. This fusion is present in 7% of the 60 patient cancers tested, a significant frequency considering the highly heterogeneous nature of this malignancy. Further, we provide direct evidence that BCAM-AKT2 is translated into an in-frame fusion protein in the patient’s tumor. The resulting AKT2 fusion kinase is membrane-associated, constitutively phosphorylated, and activated as a functional kinase in cells. Unlike endogenous AKT2, whose activity is tightly regulated by external stimuli, BCAM-AKT2 escapes the regulation from external stimuli. Moreover, a BCAM-AKT2 fusion gene generated via chromosomal translocation using the CRISPR/Cas9 system leads to focus formation in both OVCAR8 and HEK-293T cell lines, suggesting that BCAM-AKT2 is oncogenic. Together, the results indicate that BCAM-AKT2 expression is a new mechanism of AKT2 kinase activation in HGSC. BCAM-AKT2 is the only fusion gene in HGSC that is proven to translate an aberrant yet functional kinase fusion protein with oncogenic properties. This recurrent genomic alteration is a potential therapeutic target and marker of a clinically relevant subtype for tailored therapy of HGSC. cancer-specific fusion gene BCAM fusion kinase
|
| high-grade serous ovarian cancer | AKT2 |
varian cancer is responsible for the death of ∼140,200 women yearly, and 70% of these deaths are due to the high-grade serous ovarian cancer (HGSC) subtype (1), which is typically detected only after it has metastasized. Genomic characterization of HGSC tumors shows that TP53 mutations are present in 96% and BRCA1/2 in 22% of HGSC (2). These mutations are thought to occur early in pathogenesis (3) and promote genomic instability, thus giving rise to the high degree of heterogeneity and genomic rearrangements that are characteristic of these cancers. The characteristic genome rearrangements in HGSC imply that recombination events such as gene fusions should be common. Moreover, if the activity of a fusion gene represents a common oncogenic mechanism, the same fusion gene is likely to occur in many patients. Recurrent fusion genes are important for understanding cancer mechanisms and developing useful clinical biomarkers and anticancer therapies. For instance, the BCR-ABL fusion gene in chronic myeloid leukemia is known to initiate oncogenesis through formation of a misregulated BCR-ABL fusion kinase. The BCR-ABL fusion gene is also a clinical biomarker of high diagnostic and prognostic utility in chronic myeloid leukemia. In addition, this fusion protein serves as a therapeutic target for the successful drug imatinib (4). In solid human tumors, fusion genes such as the TMPRSS2-ERG fusion in prostate cancer (5), FGFR-TACC in glioblastoma (6), and DNAJB1-PRKACA in liver cancer (7) are important molecular signatures for understanding cancer mechanisms and stratifying patient groups.
Aberrations of the phosphatidylinositol 3-kinase (PI3K) signaling pathway, which occur in >40% of HGSC, are important for tumor progression in HGSC (2, 8). In particular, alterations in AKT2, a key member of the PI3K pathway that regulates cellular metabolism, proliferation, and migration (9, 10), have been implicated by several studies. AKT2 gene copy number is amplified in 12–27% of epithelial ovarian cancers (11, 12). Phosphorylated AKT2, which represents the activated form of the AKT2 kinase, is observed at high levels in HGSC (13, 14). The characteristic genomic instability of HGSC could give rise to gene fusion events involving PI3K pathway members and lead to tumor progression. However, such gene fusion events have previously not been found in HGSC. In this report, we identify BCAM-AKT2, a cancer-specific gene fusion in HGSC that displays a new mechanism of AKT2 activation by fusion with a membrane protein, BCAM. We show that this fusion gene was expressed in 7% of patients with HGSC, and that in a patient tumor it was translated into an in-frame BCAMAKT2 fusion protein retaining the transmembrane domain of BCAM and the kinase domain of AKT2. The resulting AKT2 fusion kinase is membrane-associated, constitutively phosphorylated, and activated as a functional kinase in cells. BCAM-AKT2 is the only high-occurrence fusion transcript proven to translate into a fusion protein in the patient’s tumor. The significance of this fusion is further indicated by its ability to lead to focus formation, suggesting that BCAM-AKT2 is oncogenic. This fusion
O
E1272–E1277 | PNAS | Published online March 2, 2015
Significance High-grade serous ovarian cancer (HGSC) is the most common subtype of ovarian cancer and is typically detected only at advanced stages due to lack of effective early screening tools. Fusion genes are among the most cancer-specific signatures known and, when highly recurrent, they have the potential to serve as screening tools. Here we identified BCAM-AKT2 as a cancer-specific fusion gene present in 7% of HGSC tumors, a significant frequency in this highly heterogeneous disease. This fusion results in an aberrant kinase whose constant activity contributes to cancer formation. Thus, the BCAM-AKT2 fusion gene could be important for understanding and identifying clinically relevant subtypes of HGSC, and could be a novel therapeutic target for developing small-molecule drugs. Author contributions: K.K. and L.Y. designed research; K.K., P.-W.C., L.L., Y.W., and A.E.B. performed research; S.M.H. and M.M.M. contributed new reagents/analytic tools; K.K., C.C., A.M., and L.Y. analyzed data; and K.K., M.M.M., and L.Y. wrote the paper. Reviewers: S.H., The University of Edinburgh; and B.V., University of Ottawa. The authors declare no conflict of interest. 1
K.K. and C.C. contributed equally to this work.
2
To whom correspondence may be addressed. Email: [email protected] or yen@bcm. edu.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. 1073/pnas.1501735112/-/DCSupplemental.
www.pnas.org/cgi/doi/10.1073/pnas.1501735112
Results BCAM-AKT2 Is a Recurrent Fusion in HGSC. To examine whether gene fusions involving PI3K members are present in HGSC, we carried out fusion transcript analyses (SI Materials and Methods) on transcriptome sequencing data from high-throughput sequencing of seven HGSC patient samples (15). These cancer samples were primary tumors from patients who did not receive neoadjuvant chemotherapy before removal of the cancerous ovary. Our analyses uncovered a new fusion transcript between the genes BCAM and AKT2 in one HGSC cancer sample (patient S4; Table S1). This fusion transcript is novel and has not been reported (16–19). The presence of this fusion transcript in patient S4 is supported by eight paired “chimeric” reads, with one read mapping to BCAM and the other to AKT2 (Fig. 1A). This fusion transcript is absent in the transcriptome sequencing data from pooled noncancerous ovaries and pooled noncancerous fallopian tubes (SI Materials and Methods). To experimentally validate the presence of this fusion transcript, we performed nested RT-PCR (see Table S2 for primers) using a cohort of 60 HGSC patient samples. As shown in Fig. 1B, 4 out of 60 HGSC patient samples gave a distinct band of expected size, indicating an occurrence frequency of ∼7% among this patient population. In contrast, neither the 25 noncancerous ovary samples nor the 9 noncancerous fallopian tube samples were positive by nested RT-PCR, suggesting that BCAM-AKT2 is cancer-specific. All positive RT-PCR bands were gel-purified, subjected to Sanger sequencing, and confirmed to harbor the identical BCAM-AKT2 RNA fusion junction, with the 3′ end of BCAM exon 13 (uc002ozu.4) joined to the 5′ end of AKT2 exon 5 (uc002onf.3) by annotated splice sites (Fig. 1C). This indicates that the RNA junction of this fusion transcript is the result of splicing. The identified RNA junction sequence enabled us to search among previously unmapped reads for paired “junction-spanning” reads
BCAM-AKT2 Results from Chromosomal Rearrangement. BCAM and AKT2 are both located on chromosome (chr) 19 and are separated by ∼4.6 Mb. To check whether the fused BCAM-AKT2 transcript indeed results from chromosomal rearrangement, we used a primer on BCAM exon 13 paired with another primer on AKT2 exon 5 to amplify from patient S4 genomic DNA. This yielded a single band of ∼3.5 kb, which was absent in other patient samples that do not express the fusion transcript (Fig. 2A). Gel purification and Sanger sequencing of this genomic DNA band revealed that the genomic breakpoint is located in the intron 548 bp downstream of exon 13 of BCAM (chr19: 44820273, GRCh38/hg38) and in the intron 2,743 bp upstream of exon 5 of AKT2 (chr19: 40245429, GRCh38/hg38) (Fig. 2B and Table S3). This directly proves that the BCAM-AKT2 transcript results from a gene fusion (Fig. 2C). Further sequence analysis of the breakpoint revealed no obvious sequence homology except a 4-nt ATTT microhomology at the breakpoint, suggesting that the gene fusion may be mediated by nonhomologous break repair mechanisms (20, 21). BCAM-AKT2 Is Translated into an In-Frame Fusion Protein in Patient Tumors. Based on the RNA junction sequence of the fusion
transcript, we predicted that translation of the BCAM-AKT2 transcript would result in a fusion protein with both BCAM and AKT2 coding regions translated in-frame. The BCAM gene normally produces two different alternatively spliced RNA isoforms. This fusion retains all domains that define the short protein isoform called BCAM but lacks the last 40 amino acids that define
Fig. 1. BCAM-AKT2 is a frequent fusion transcript in HGSC cancer samples. (A) Schematic showing the position of eight paired chimeric reads aligning to both BCAM and AKT2 genes identified in patient S4. (B) The results of nested RT-PCR for BCAM-AKT2 in 60 HGSC patient samples (S), 25 noncancerous ovary samples (OV), and 9 noncancerous fallopian tube samples (FT). NTC, no cDNA control. S4 is the sample in which the fusion transcript was initially identified. It serves as the positive control. (C, Top) Sanger sequencing chromatogram of the RT-PCR bands revealed that the BCAM-AKT2 RNA junction joins the 3′ end of BCAM exon 13 to the 5′ end of AKT2 exon 5 (as demarcated by the black line). (Middle) Protein reading frame at the RNA junction (green, BCAM; blue, AKT2). (Bottom) Eight junction-spanning reads indicated by black lines identified from transcriptome sequencing are shown.
Kannan et al.
PNAS | Published online March 2, 2015 | E1273
PNAS PLUS
in which one read spans the RNA junction and another read maps to one of the fusion gene partners. We were able to identify eight corresponding paired junction-spanning reads in patient S4, which further supports the presence of the BCAM-AKT2 fusion transcript (Fig. 1C and Table S1). No other type of RNA junction was observed in our transcriptome sequencing data or from RT-PCR assays, suggesting that the RNA junction is precisely specified and would lead to a single specified fusion protein.
GENETICS
gene could be important for understanding and identifying clinically relevant subtypes, and serve as a novel therapeutic target for developing small-molecule drugs.
the 110-kDa band in patient S4 is the BCAM-AKT2 fusion protein and that it is translated in-frame as predicted. Further, Western blot analysis using an antibody that specifically detects phosphorylation at serine 474 of AKT2 showed that the AKT2 kinase domain in the fusion protein is phosphorylated (Fig. 3C, lane 3). Together, the results strongly indicate that BCAM-AKT2 is translated and phosphorylated in patient tumors. BCAM-AKT2 Results in a Membrane-Associated, Constitutively Phosphorylated AKT2 Fusion Kinase. The fact that the AKT2 fu-
Fig. 2. BCAM-AKT2 fusion is the result of genomic rearrangement. (A) Longrange PCR using primers on exon 13 of BCAM and exon 5 of AKT2 on genomic DNA resulted in a band of about 3.5 kb in patient S4 but not in other patients lacking the fusion transcript. (B) Schematic of the BCAM-AKT2 genomic breakpoint (*) identified by Sanger sequencing of the genomic PCR product from patient S4. Complete Sanger sequencing results are listed in Table S3. (C) Putative mechanism of genomic translocation leading to BCAM-AKT2 fusion.
the longer isoform known as Lutheran or Lu (Fig. 3A). The Lu protein isoform is known to localize on the basolateral membrane characteristic of differentiated epithelial cells, whereas the BCAM isoform, which is overexpressed in ovarian cancers, is localized in a nonpolarized manner (22, 23). The fusion protein is predicted to encode most of AKT2, including its kinase domain (Fig. 3A). However, it lacks 97 amino acids from the N terminus of AKT2 that constitutes the pleckstrin homology (PH) domain necessary for targeting AKT2 to the cell membrane, where it normally gets activated (24). Because the BCAM product is a membrane protein, we predict that the BCAM-AKT2 fusion protein will localize to the membrane. To confirm the translation of the predicted BCAM-AKT2 fusion protein in HGSC patients, we performed Western blot analysis of patient tumor samples and noncancerous ovary and fallopian tube samples. As shown in Fig. 3B, the AKT2 antibody recognizes a band in patient S4 corresponding to the predicted fusion protein (∼110 kDa), which is larger than endogenous AKT2 (∼60 kDa). This 110-kDa band was absent in patient S27, who did not express the fusion transcript, and also lacking in noncancerous ovary and fallopian tube (Fig. 3B, Top). To confirm that the 110-kDa band is indeed the BCAM-AKT2 fusion protein, we performed a coimmunoprecipitation assay first by precipitating the target proteins using an anti-AKT2 antibody, followed by Western blot analysis using an anti-BCAM antibody. As shown in Fig. 3B (Bottom), the BCAM antibody recognizes the 110-kDa band that was precipitated by the AKT2 antibody but does not recognize the precipitated endogenous AKT2 (∼60 kDa). This confirmed that E1274 | www.pnas.org/cgi/doi/10.1073/pnas.1501735112
sion kinase is phosphorylated at serine 474 (Fig. 3C) suggests that loss of the PH domain is compensated by fusion to BCAM, a membrane protein that could relocate AKT2 to the cell membrane, where it gets activated by phosphorylation (24). To confirm this, we set out to generate epitope-specific antibodies that could specifically recognize the junction peptide of the fusion protein but not parental BCAM or AKT2. However, several monoclonal antibodies that we produced unfortunately failed to recognize the fusion protein specifically. This prevented the most direct confirmation of membrane localization of BCAM-AKT2 fusion protein in the patient’s tumor, although the fusion protein was clearly translated (Fig. 3B). To circumvent this difficulty, we cloned BCAM-AKT2 from patient S4, FLAG-tagged it, and expressed it ectopically in OVCAR8 cells. Western blot analysis confirmed the translation and phosphorylation of BCAM-AKT2 at both serine 474 (Fig. 3D) and threonine 309 of AKT2 (Fig. S1), suggesting that the fusion protein is in the active state. Immunocytochemical imaging showed that the fusion protein is mainly located at the cell membrane (Fig. 3E). Thus, BCAM-AKT2 fusion leads to a fusion protein that relocates the AKT2 kinase to the cell membrane, where it is phosphorylated. The phosphorylation of AKT2 and thus its activation are known to respond swiftly to extracellular growth stimuli such as insulin-like growth factor 1 (IGF1) (25). To answer the question of whether phosphorylation of BCAM-AKT2 fusion protein is also regulated by IGF1, we serum-deprived transfected OVCAR8 cells overnight so that AKT2 became unphosphorylated (26). Cells were then treated with IGF1 for 30 min to activate AKT2 followed by immediate protein extraction. Western blot analysis using an antibody against phosphorylated AKT2 (serine 474) shows that the endogenous AKT2 responds to IGF1 treatment swiftly through phosphorylation (Fig. 4A, Top, lane 3 versus 4). In contrast, BCAMAKT2 remains phosphorylated regardless of the presence or absence of IGF1 (Fig. 4A, Top, lane 1 versus 2). Further, there is no unphosphorylated BCAM-AKT2 detectable (Fig. 4A, Middle, lanes 1 and 2). This result suggests that BCAM-AKT2 is constitutively phosphorylated, presumably as a consequence of its membrane localization. To investigate whether the constitutively phosphorylated BCAM-AKT2 still possesses functional kinase activity, we immunoprecipitated BCAM-AKT2 from transfected OVCAR8 cells using an anti-FLAG antibody and incubated the fusion protein with GSK-3, a kinase substrate. As shown in Fig. 4B, incubation with immunoprecipitated BCAM-AKT2 led to efficient phosphorylation of GSK-3, indicating that BCAM-AKT2 is a functional and active kinase. Recreation of the BCAM-AKT2 Fusion Gene Using CRISPR/Cas9 Leads to Focus Formation. To investigate the functional consequences of
the BCAM-AKT2 fusion gene, we recreated the fusion gene in cells by means of chromosomal translocation induced using the CRISPR/Cas9 system. Unlike commonly used overexpression assays that use a CMV promoter to express cloned cDNA, the strategy we used generates a fusion gene with all genomic sequence elements intact, including the endogenous BCAM promoter and the appropriate UTRs and introns. The resulting fusion gene thus would mirror closely the BCAM-AKT2 fusion gene found in patients. To achieve this, we designed three guide RNAs targeting BCAM intron 13 to pair with three guide RNAs Kannan et al.
PNAS PLUS GENETICS
Fig. 3. BCAM-AKT2 gives rise to an in-frame fusion protein in patient tumors. (A) Protein domains of BCAM/Lu, AKT2, and the predicted BCAM-AKT2 fusion protein. As a result of the gene fusion, BCAM loses the Lu-specific region but retains all domains, including the transmembrane domain. AKT2 loses the PH domain but retains the kinase domains. Antibodies used against AKT2 and BCAM domains are shown. (B, Upper) Western blot using AKT2 antibody. (Lower) Western blot using BCAM antibody (after immunoprecipitation with AKT2 antibody). In both cases, a band corresponding to the predicted size of the fusion protein (∼110 kDa) was observed in patient S4 but absent in patient S27 or in noncancerous ovary and fallopian tube. IB, immunoblotting; IP, immunoprecipitation. (C) Western blot using phospho-AKT2 antibody indicates that BCAM-AKT2 is phosphorylated at serine 474 (lane 3). Controls include Jurkat cells treated with LY294002 (an inhibitor of AKT phosphorylation) or with Calyculin A (to preserve AKT in a phosphorylated state). (D) Western blot indicates that the cDNA cloned from patient S4 is translated into a full-length in-frame fusion protein in OVCAR8 cells as visualized by FLAG antibody. It is also phosphorylated as indicated by phospho-AKT2 (serine 474) antibody. (E) Immunocytochemistry of OVCAR8 cells transfected with the BCAM-AKT2-FLAG expression construct. The fusion protein was seen mainly on the cell membrane (arrows) as well as on the protruding filopodia (arrowheads) that form focal adhesions with the extracellular matrix. The deconvolution microscopy image was visualized by FLAG antibody. The image is a projection of seven stacks, each with 0.35-μm thickness.
targeting AKT2 intron 4 (Table S4) to induce double-stranded breaks at defined sites (Fig. 5A, Upper). As shown in Fig. 5A (Lower), transfection of nine combinations of paired guide RNAs with Cas9 in HEK-293T cells induced fusion transcript expression to various degrees depending on the combination used. Sanger sequencing of the RT-PCR band from the pair of BCAM-g3/ AKT2-g1 confirmed that the induced band is indeed the expected BCAM-AKT2 fusion transcript (Fig. S2). To assay for an oncogenic phenotype, OVCAR8 and HEK293T cells were transfected with BCAM-g3/AKT2-g1 along with Cas9 and then plated at a low density to allow individual cells to develop into foci. After 2 wk, foci were observed sporadically among the transfected population but not in the untransfected Kannan et al.
parental cells. To test whether the expression of BCAM-AKT2 is associated with focus formation, 44 prominent foci from transfected OVCAR8 cells and 26 from transfected HEK-293T cells were isolated and cultured. RT-PCR performed on these foci showed that ∼60% of the isolated foci from OVCAR8 (Fig. 5B) and 80% of the isolated foci from HEK-293T (Fig. 5C) were positive for the BCAM-AKT2 fusion transcript as confirmed by Sanger sequencing of the excised bands (an example from an OVCAR8 focus is shown in Fig. 5D). Thus, 60–80% of the isolated foci carry the BCAM-AKT2 translocation, and this is far greater than the expected low rate of long-range chromosome translocation that can be induced by the CRISPR/Cas9 system (1–4%) in cells (27). The observed focus formation, therefore, is PNAS | Published online March 2, 2015 | E1275
BCAM-AKT2 also differs from endogenous AKT2 in that the expression of the fusion AKT2 kinase is under the control of the BCAM promoter, not the AKT2 promoter, and phosphorylation of BCAM-AKT2 no longer responds to external stimuli. These observations suggest that the fusion AKT2 kinase escapes the tight control of normal cellular regulatory pathways, providing a new property that could significantly contribute to oncogenesis. This feature differs markedly from the effect of simple gene copy number increase in AKT2, which still could be tightly regulated by cells (3, 12). Excessive genome rearrangements that occur in HGSC make the analysis of gene fusion occurrence rate particularly challenging. An HGSC tumor could exist as a large mass composed of many mutational lineages. Depending on the part of the tumor that is available for experimental analysis, a mutational event could be completely missed even though it is present in other areas of the tumor. Further, different laboratories that obtain different parts of the same tumor may come to different conclusions because of the high degree of heterogeneity. Differences in methodology, such as reports that rely entirely on transcriptome sequencing and bioinformatics versus reports that rely on more sensitive experimental RT-PCR validation, could also contribute to the discrepancy observed by different studies
Fig. 4. BCAM-AKT2 is constitutively phosphorylated in transfected OVCAR8 cells. (A, Top) Western blot using phospho-AKT2 (serine 474) antibody shows that BCAM-AKT2 remains phosphorylated regardless of the presence or absence of IGF1 (lane 1 vs. 2). In contrast, the endogenous AKT2 remains largely unphosphorylated and responds to IGF1 treatment swiftly through phosphorylation (lanes 1 vs. 2, and 3 vs. 4). (Middle) No unphosphorylated BCAM-AKT2 was detected. (Bottom) Actin controls for protein loading. (B) The expressed fusion protein was immunoprecipitated by FLAG antibody and incubated with GSK-3 substrate. Western blot using anti–phospho-GSK-3 antibody showed that immunoprecipitated BCAM-AKT2 efficiently phosphorylated GSK-3.
not a random event but largely associated with BCAM-AKT2 expression. Discussion Fusion genes are among the most cancer-specific signatures known. In this report, we identified a novel recurrent fusion kinase, BCAMAKT2, that is specific to HGSC. Several key features are associated with the BCAM-AKT2 fusion. First, the exact same RNA junction is observed in the fusion transcript from different patients, suggesting the gain of a precisely defined fusion protein sequence. Second, it is an in-frame fusion, with both BCAM and AKT2 protein sequences remaining largely intact. Third, as demonstrated in Fig. 3, the fusion protein is clearly translated in HGSC patient tumors. Fourth, the fusion leads to a membrane-associated and constitutively activated AKT2 kinase, a key member in the PI3K/AKT pathway known to play a role in oncogenesis in HGSC (10). Fifth, BCAM-AKT2 is expressed in 7% of all HGSC tumors, a significant frequency given the highly heterogeneous nature of this malignancy. Because HGSC has increasingly been viewed as many diseases instead of a single disease, BCAM-AKT2 could represent an important genomic alteration for identifying a clinically relevant subtype of HGSC. Last, recapitulation of the BCAM-AKT2 fusion results in cellular transformation. E1276 | www.pnas.org/cgi/doi/10.1073/pnas.1501735112
Fig. 5. Generation of BCAM-AKT2 fusion in cells via chromosomal translocation leads to focus formation. (A) Schematic showing target sites of designed guide RNAs in BCAM intron 13 and AKT2 intron 4. Dashed red lines indicate genomic breakpoints identified in patient S4. Nine combinations of guide RNA pairs were transfected into HEK-293T cells. RT-PCR results show that the BCAM-AKT2 fusion transcript was detected in varying degrees depending on the combination of guide RNAs used. (B and C) (Upper) Examples of foci induced after transfection of BCAM-g3/AKT2-g1 in OVCAR8 and HEK-293T cells. Numbers indicate the ID number of the foci. (Lower) Presence of the BCAM-AKT2 fusion transcript as revealed by RT-PCR of foci isolated from transfected OVCAR8 and HEK-293T populations. (D) Representative Sanger sequencing of RT-PCR bands (an example from focus OVCAR8-46) confirmed that it contains the expected BCAM-AKT2 fusion junction.
Kannan et al.
1. Jemal A, et al. (2011) Global cancer statistics. CA Cancer J Clin 61(2):69–90. 2. Bell D, et al.; Cancer Genome Atlas Research Network (2011) Integrated genomic analyses of ovarian carcinoma. Nature 474(7353):609–615. 3. Bast RC, Jr, Mills GB (2012) Dissecting “PI3Kness”: The complexity of personalized therapy for ovarian cancer. Cancer Discov 2(1):16–18. 4. Mitelman F, Johansson B, Mertens F (2007) The impact of translocations and gene fusions on cancer causation. Nat Rev Cancer 7(4):233–245. 5. Tomlins SA, et al. (2005) Recurrent fusion of TMPRSS2 and ETS transcription factor genes in prostate cancer. Science 310(5748):644–648. 6. Singh D, et al. (2012) Transforming fusions of FGFR and TACC genes in human glioblastoma. Science 337(6099):1231–1235. 7. Honeyman JN, et al. (2014) Detection of a recurrent DNAJB1-PRKACA chimeric transcript in fibrolamellar hepatocellular carcinoma. Science 343(6174):1010–1014. 8. Romero I, Bast RC, Jr (2012) Minireview: Human ovarian cancer: Biology, current management, and paths to personalizing therapy. Endocrinology 153(4):1593–1602. 9. Garofalo RS, et al. (2003) Severe diabetes, age-dependent loss of adipose tissue, and mild growth deficiency in mice lacking Akt2/PKB beta. J Clin Invest 112(2):197–208. 10. Hers I, Vincent EE, Tavaré JM (2011) Akt signalling in health and disease. Cell Signal 23(10):1515–1527. 11. Bellacosa A, et al. (1995) Molecular alterations of the AKT2 oncogene in ovarian and breast carcinomas. Int J Cancer 64(4):280–285. 12. Cheng JQ, et al. (1992) AKT2, a putative oncogene encoding a member of a subfamily of protein-serine/threonine kinases, is amplified in human ovarian carcinomas. Proc Natl Acad Sci USA 89(19):9267–9271. 13. Altomare DA, et al. (2004) AKT and mTOR phosphorylation is frequently detected in ovarian cancer and can be targeted to disrupt ovarian tumor cell growth. Oncogene 23(34):5853–5857. 14. Yuan ZQ, et al. (2000) Frequent activation of AKT2 and induction of apoptosis by inhibition of phosphoinositide-3-OH kinase/Akt pathway in human ovarian cancer. Oncogene 19(19):2324–2330. 15. Kannan K, et al. (2014) CDKN2D-WDFY2 is a cancer-specific fusion gene recurrent in high-grade serous ovarian carcinoma. PLoS Genet 10(3):e1004216. 16. Kim P, et al. (2010) ChimerDB 2.0—A knowledgebase for fusion genes updated. Nucleic Acids Res 38(Database issue):D81–D85. 17. McPherson A, et al. (2011) deFuse: An algorithm for gene fusion discovery in tumor RNA-seq data. PLOS Comput Biol 7(5):e1001138.
Kannan et al.
PNAS PLUS
for leukemia (4) and TMPRSS2-ERG for prostate cancer (5). Because the fusion involves both BCAM and AKT2, the use of the currently available BCAM antibody and/or competitive AKT2 inhibitors should be evaluated for treatment of BCAM-AKT2– positive patients given that limited therapeutic options exist beyond standard chemotherapy. Last, this fusion protein may fold differently from the parental proteins. A small-molecule drug preferentially targeting this fusion protein can be developed, as has been demonstrated in the case of imatinib targeting BCR-ABL. Materials and Methods Anonymous ovarian cancer tissue samples were obtained from the Tissue Acquisition and Distribution Core of the Dan L. Duncan Cancer Center, Department of Pathology & Immunology, and Gynecologic Oncology Group under an approved Baylor College of Medicine Institutional Review Board protocol. Informed consent was obtained from all patients. All tumor samples were confirmed to have greater than 80% serous adenocarcinoma before processing. RNA was extracted from cancer samples and noncancerous samples using a RiboPure Kit (Ambion) and processed for RT-PCR or transcriptome sequencing as described earlier (15). Total RNA samples were processed for transcriptome sequencing using the Illumina mRNA-seq protocol. Bioinformatic identification of chimeric reads and corresponding junction reads is detailed in SI Materials and Methods. For RT-PCR, RT was performed using SuperScript III (Invitrogen), and PCR was performed using the primers listed in Table S2. The OVCAR8 cancer cell line was maintained using the media RPMI 1640, 10% (vol/vol) FBS, and 1% penicillin/streptomycin. Detailed materials and methods for immunoassays, genomic breakpoint identification, and CRISPR/Cas9 are included in SI Materials and Methods. ACKNOWLEDGMENTS. We thank Sriram Ayyaswamy and Thuy Phung for providing some of the antibodies, and Michael Ittmann and Patricia Castro for assisting with the cancer samples. We thank Radhika Dandekar, Fabio Stossi, and Michael Mancini for assisting with microscopy. S.M.H., M.M.M., and L.Y. have been supported by Ovarian Cancer Research Foundation OCRF PPD/BCM/01.12. L.Y. has been supported by Department of Defense W81XWH10-10327 and NIH R01EB013584. K.K. has been supported by a Prostate Cancer Foundation Young Investigator Award. A.M. was supported by a grant from the Weathervane Foundation. This project was also supported in part by the Genomic and RNA Profiling Core at Baylor College of Medicine with funding from an NIH National Cancer Institute Grant (P30CA125123).
18. Salzman J, et al. (2011) ESRRA-C11orf20 is a recurrent gene fusion in serous ovarian carcinoma. PLoS Biol 9(9):e1001156. 19. Stransky N, Cerami E, Schalm S, Kim JL, Lengauer C (2014) The landscape of kinase fusions in cancer. Nat Commun 5:4846. 20. Lieber MR (2010) The mechanism of double-strand DNA break repair by the nonhomologous DNA end-joining pathway. Annu Rev Biochem 79:181–211. 21. Zhang F, Carvalho CM, Lupski JR (2009) Complex human chromosomal and genomic rearrangements. Trends Genet 25(7):298–307. 22. El Nemer W, et al. (1999) Isoforms of the Lutheran/basal cell adhesion molecule glycoprotein are differentially delivered in polarized epithelial cells. Mapping of the basolateral sorting signal to a cytoplasmic di-leucine motif. J Biol Chem 274(45):31903–31908. 23. Garinchesa P, Sanzmoncasi M, Campbell I, Rettig W (1994) Non-polarized expression of basal-cell adhesion molecule B-CAM in epithelial ovarian cancers. Int J Oncol 5(6): 1261–1266. 24. Franke TF, et al. (1995) The protein kinase encoded by the Akt proto-oncogene is a target of the PDGF-activated phosphatidylinositol 3-kinase. Cell 81(5):727–736. 25. Liu AX, et al. (1998) AKT2, a member of the protein kinase B family, is activated by growth factors, v-Ha-ras, and v-src through phosphatidylinositol 3-kinase in human ovarian epithelial cancer cells. Cancer Res 58(14):2973–2977. 26. Watton SJ, Downward J (1999) Akt/PKB localisation and 3′ phosphoinositide generation at sites of epithelial cell-matrix and cell-cell interaction. Curr Biol 9(8):433–436. 27. Torres R, et al. (2014) Engineering human tumour-associated chromosomal translocations with the RNA-guided CRISPR-Cas9 system. Nat Commun 5:3964. 28. Micci F, et al. (2014) Low frequency of ESRRA-C11orf20 fusion gene in ovarian carcinomas. PLoS Biol 12(2):e1001784. 29. Andjelkovic M, et al. (1997) Role of translocation in the activation and function of protein kinase B. J Biol Chem 272(50):31515–31524. 30. Meier R, Alessi DR, Cron P, Andjelkovic M, Hemmings BA (1997) Mitogenic activation, phosphorylation, and nuclear translocation of protein kinase Bbeta. J Biol Chem 272(48):30491–30497. 31. Mende I, Malstrom S, Tsichlis PN, Vogt PK, Aoki M (2001) Oncogenic transformation induced by membrane-targeted Akt2 and Akt3. Oncogene 20(32):4419–4423. 32. Carpten JD, et al. (2007) A transforming mutation in the pleckstrin homology domain of AKT1 in cancer. Nature 448(7152):439–444. 33. Banerji S, et al. (2012) Sequence analysis of mutations and translocations across breast cancer subtypes. Nature 486(7403):405–409.
PNAS | Published online March 2, 2015 | E1277
GENETICS
(17–19, 28). It is possible that each experimentally validated fusion gene in HGSC may have a higher frequency of occurrence than is currently estimated, as it is not possible to sample each HGSC tumor in its entirety. The oncogenic consequences of membrane-associated AKT kinases have been documented by several studies. For instance, when AKT2 is myristoylated––a posttranslational modification that makes AKT2 membrane-associated––it leads to constitutively activated AKT2 in transfected cells (29, 30). Furthermore, when myristoylated AKT2 is expressed in cells, it results in loss of contact inhibition and leads to focus formation, an indicator of oncogenicity (31). Similarly, AKT1 mutation (E17K) in the PH domain found in breast and ovarian cancer patients leads to membrane-bound and constitutively activated AKT1 with increased focus formation (32). Perhaps the most intriguing example parallel to our discovery is the fusion gene MAGI3-AKT3 found in 7% of triple-negative breast cancer patients (33). In this case, AKT3 is fused in-frame to the membrane protein MAGI3. Although the translation of this fusion protein was not confirmed in patient tumor tissues, it was observed that in transfected cells, MAGI3-AKT3 was constitutively phosphorylated in the absence of growth factors and also oncogenic, as evidenced by focus formation (33). Consistent with the above reports, our experiments also demonstrate that the BCAM-AKT2 fusion kinase is indeed membrane-associated (Fig. 3E). In addition, by using the CRISPR/Cas9 system, we generated the BCAM-AKT2 fusion gene that mirrors closely the fusion gene found in patients, and this leads to focus formation (Fig. 5). Although other recurrent fusion genes, namely CDKN2DWDFY2 (15) and ESRRA-C11orf20 (18), are present in HGSC, the BCAM-AKT2 fusion reported here is the most clinically relevant and functionally significant and is the only cancer-specific fusion gene proven to translate a fusion protein in HGSC patients. Further, unlike TP53 and BRCA1/2 mutations that are present in many types of cancer, BCAM-AKT2 is “unique” to high-grade serous ovarian cancer, and therefore may represent a marker specific for HGSC in a manner similar to that of BCR-ABL
High-grade serous ovarian cancer (HGSC) is among the most lethal forms of cancer in women. Excessive genomic rearrangements, which are expected to create fusion oncogenes, are the hallmark of this cancer. Here we report a cancer-specific gene fusion between BCAM, a membrane adhesion molecule, and AKT2, a key kinase in the PI3K signaling pathway. This fusion is present in 7% of the 60 patient cancers tested, a significant frequency considering the highly heterogeneous nature of this malignancy. Further, we provide direct evidence that BCAM-AKT2 is translated into an in-frame fusion protein in the patient’s tumor. The resulting AKT2 fusion kinase is membrane-associated, constitutively phosphorylated, and activated as a functional kinase in cells. Unlike endogenous AKT2, whose activity is tightly regulated by external stimuli, BCAM-AKT2 escapes the regulation from external stimuli. Moreover, a BCAM-AKT2 fusion gene generated via chromosomal translocation using the CRISPR/Cas9 system leads to focus formation in both OVCAR8 and HEK-293T cell lines, suggesting that BCAM-AKT2 is oncogenic. Together, the results indicate that BCAM-AKT2 expression is a new mechanism of AKT2 kinase activation in HGSC. BCAM-AKT2 is the only fusion gene in HGSC that is proven to translate an aberrant yet functional kinase fusion protein with oncogenic properties. This recurrent genomic alteration is a potential therapeutic target and marker of a clinically relevant subtype for tailored therapy of HGSC. cancer-specific fusion gene BCAM fusion kinase
|
| high-grade serous ovarian cancer | AKT2 |
varian cancer is responsible for the death of ∼140,200 women yearly, and 70% of these deaths are due to the high-grade serous ovarian cancer (HGSC) subtype (1), which is typically detected only after it has metastasized. Genomic characterization of HGSC tumors shows that TP53 mutations are present in 96% and BRCA1/2 in 22% of HGSC (2). These mutations are thought to occur early in pathogenesis (3) and promote genomic instability, thus giving rise to the high degree of heterogeneity and genomic rearrangements that are characteristic of these cancers. The characteristic genome rearrangements in HGSC imply that recombination events such as gene fusions should be common. Moreover, if the activity of a fusion gene represents a common oncogenic mechanism, the same fusion gene is likely to occur in many patients. Recurrent fusion genes are important for understanding cancer mechanisms and developing useful clinical biomarkers and anticancer therapies. For instance, the BCR-ABL fusion gene in chronic myeloid leukemia is known to initiate oncogenesis through formation of a misregulated BCR-ABL fusion kinase. The BCR-ABL fusion gene is also a clinical biomarker of high diagnostic and prognostic utility in chronic myeloid leukemia. In addition, this fusion protein serves as a therapeutic target for the successful drug imatinib (4). In solid human tumors, fusion genes such as the TMPRSS2-ERG fusion in prostate cancer (5), FGFR-TACC in glioblastoma (6), and DNAJB1-PRKACA in liver cancer (7) are important molecular signatures for understanding cancer mechanisms and stratifying patient groups.
Aberrations of the phosphatidylinositol 3-kinase (PI3K) signaling pathway, which occur in >40% of HGSC, are important for tumor progression in HGSC (2, 8). In particular, alterations in AKT2, a key member of the PI3K pathway that regulates cellular metabolism, proliferation, and migration (9, 10), have been implicated by several studies. AKT2 gene copy number is amplified in 12–27% of epithelial ovarian cancers (11, 12). Phosphorylated AKT2, which represents the activated form of the AKT2 kinase, is observed at high levels in HGSC (13, 14). The characteristic genomic instability of HGSC could give rise to gene fusion events involving PI3K pathway members and lead to tumor progression. However, such gene fusion events have previously not been found in HGSC. In this report, we identify BCAM-AKT2, a cancer-specific gene fusion in HGSC that displays a new mechanism of AKT2 activation by fusion with a membrane protein, BCAM. We show that this fusion gene was expressed in 7% of patients with HGSC, and that in a patient tumor it was translated into an in-frame BCAMAKT2 fusion protein retaining the transmembrane domain of BCAM and the kinase domain of AKT2. The resulting AKT2 fusion kinase is membrane-associated, constitutively phosphorylated, and activated as a functional kinase in cells. BCAM-AKT2 is the only high-occurrence fusion transcript proven to translate into a fusion protein in the patient’s tumor. The significance of this fusion is further indicated by its ability to lead to focus formation, suggesting that BCAM-AKT2 is oncogenic. This fusion
O
E1272–E1277 | PNAS | Published online March 2, 2015
Significance High-grade serous ovarian cancer (HGSC) is the most common subtype of ovarian cancer and is typically detected only at advanced stages due to lack of effective early screening tools. Fusion genes are among the most cancer-specific signatures known and, when highly recurrent, they have the potential to serve as screening tools. Here we identified BCAM-AKT2 as a cancer-specific fusion gene present in 7% of HGSC tumors, a significant frequency in this highly heterogeneous disease. This fusion results in an aberrant kinase whose constant activity contributes to cancer formation. Thus, the BCAM-AKT2 fusion gene could be important for understanding and identifying clinically relevant subtypes of HGSC, and could be a novel therapeutic target for developing small-molecule drugs. Author contributions: K.K. and L.Y. designed research; K.K., P.-W.C., L.L., Y.W., and A.E.B. performed research; S.M.H. and M.M.M. contributed new reagents/analytic tools; K.K., C.C., A.M., and L.Y. analyzed data; and K.K., M.M.M., and L.Y. wrote the paper. Reviewers: S.H., The University of Edinburgh; and B.V., University of Ottawa. The authors declare no conflict of interest. 1
K.K. and C.C. contributed equally to this work.
2
To whom correspondence may be addressed. Email: [email protected] or yen@bcm. edu.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. 1073/pnas.1501735112/-/DCSupplemental.
www.pnas.org/cgi/doi/10.1073/pnas.1501735112
Results BCAM-AKT2 Is a Recurrent Fusion in HGSC. To examine whether gene fusions involving PI3K members are present in HGSC, we carried out fusion transcript analyses (SI Materials and Methods) on transcriptome sequencing data from high-throughput sequencing of seven HGSC patient samples (15). These cancer samples were primary tumors from patients who did not receive neoadjuvant chemotherapy before removal of the cancerous ovary. Our analyses uncovered a new fusion transcript between the genes BCAM and AKT2 in one HGSC cancer sample (patient S4; Table S1). This fusion transcript is novel and has not been reported (16–19). The presence of this fusion transcript in patient S4 is supported by eight paired “chimeric” reads, with one read mapping to BCAM and the other to AKT2 (Fig. 1A). This fusion transcript is absent in the transcriptome sequencing data from pooled noncancerous ovaries and pooled noncancerous fallopian tubes (SI Materials and Methods). To experimentally validate the presence of this fusion transcript, we performed nested RT-PCR (see Table S2 for primers) using a cohort of 60 HGSC patient samples. As shown in Fig. 1B, 4 out of 60 HGSC patient samples gave a distinct band of expected size, indicating an occurrence frequency of ∼7% among this patient population. In contrast, neither the 25 noncancerous ovary samples nor the 9 noncancerous fallopian tube samples were positive by nested RT-PCR, suggesting that BCAM-AKT2 is cancer-specific. All positive RT-PCR bands were gel-purified, subjected to Sanger sequencing, and confirmed to harbor the identical BCAM-AKT2 RNA fusion junction, with the 3′ end of BCAM exon 13 (uc002ozu.4) joined to the 5′ end of AKT2 exon 5 (uc002onf.3) by annotated splice sites (Fig. 1C). This indicates that the RNA junction of this fusion transcript is the result of splicing. The identified RNA junction sequence enabled us to search among previously unmapped reads for paired “junction-spanning” reads
BCAM-AKT2 Results from Chromosomal Rearrangement. BCAM and AKT2 are both located on chromosome (chr) 19 and are separated by ∼4.6 Mb. To check whether the fused BCAM-AKT2 transcript indeed results from chromosomal rearrangement, we used a primer on BCAM exon 13 paired with another primer on AKT2 exon 5 to amplify from patient S4 genomic DNA. This yielded a single band of ∼3.5 kb, which was absent in other patient samples that do not express the fusion transcript (Fig. 2A). Gel purification and Sanger sequencing of this genomic DNA band revealed that the genomic breakpoint is located in the intron 548 bp downstream of exon 13 of BCAM (chr19: 44820273, GRCh38/hg38) and in the intron 2,743 bp upstream of exon 5 of AKT2 (chr19: 40245429, GRCh38/hg38) (Fig. 2B and Table S3). This directly proves that the BCAM-AKT2 transcript results from a gene fusion (Fig. 2C). Further sequence analysis of the breakpoint revealed no obvious sequence homology except a 4-nt ATTT microhomology at the breakpoint, suggesting that the gene fusion may be mediated by nonhomologous break repair mechanisms (20, 21). BCAM-AKT2 Is Translated into an In-Frame Fusion Protein in Patient Tumors. Based on the RNA junction sequence of the fusion
transcript, we predicted that translation of the BCAM-AKT2 transcript would result in a fusion protein with both BCAM and AKT2 coding regions translated in-frame. The BCAM gene normally produces two different alternatively spliced RNA isoforms. This fusion retains all domains that define the short protein isoform called BCAM but lacks the last 40 amino acids that define
Fig. 1. BCAM-AKT2 is a frequent fusion transcript in HGSC cancer samples. (A) Schematic showing the position of eight paired chimeric reads aligning to both BCAM and AKT2 genes identified in patient S4. (B) The results of nested RT-PCR for BCAM-AKT2 in 60 HGSC patient samples (S), 25 noncancerous ovary samples (OV), and 9 noncancerous fallopian tube samples (FT). NTC, no cDNA control. S4 is the sample in which the fusion transcript was initially identified. It serves as the positive control. (C, Top) Sanger sequencing chromatogram of the RT-PCR bands revealed that the BCAM-AKT2 RNA junction joins the 3′ end of BCAM exon 13 to the 5′ end of AKT2 exon 5 (as demarcated by the black line). (Middle) Protein reading frame at the RNA junction (green, BCAM; blue, AKT2). (Bottom) Eight junction-spanning reads indicated by black lines identified from transcriptome sequencing are shown.
Kannan et al.
PNAS | Published online March 2, 2015 | E1273
PNAS PLUS
in which one read spans the RNA junction and another read maps to one of the fusion gene partners. We were able to identify eight corresponding paired junction-spanning reads in patient S4, which further supports the presence of the BCAM-AKT2 fusion transcript (Fig. 1C and Table S1). No other type of RNA junction was observed in our transcriptome sequencing data or from RT-PCR assays, suggesting that the RNA junction is precisely specified and would lead to a single specified fusion protein.
GENETICS
gene could be important for understanding and identifying clinically relevant subtypes, and serve as a novel therapeutic target for developing small-molecule drugs.
the 110-kDa band in patient S4 is the BCAM-AKT2 fusion protein and that it is translated in-frame as predicted. Further, Western blot analysis using an antibody that specifically detects phosphorylation at serine 474 of AKT2 showed that the AKT2 kinase domain in the fusion protein is phosphorylated (Fig. 3C, lane 3). Together, the results strongly indicate that BCAM-AKT2 is translated and phosphorylated in patient tumors. BCAM-AKT2 Results in a Membrane-Associated, Constitutively Phosphorylated AKT2 Fusion Kinase. The fact that the AKT2 fu-
Fig. 2. BCAM-AKT2 fusion is the result of genomic rearrangement. (A) Longrange PCR using primers on exon 13 of BCAM and exon 5 of AKT2 on genomic DNA resulted in a band of about 3.5 kb in patient S4 but not in other patients lacking the fusion transcript. (B) Schematic of the BCAM-AKT2 genomic breakpoint (*) identified by Sanger sequencing of the genomic PCR product from patient S4. Complete Sanger sequencing results are listed in Table S3. (C) Putative mechanism of genomic translocation leading to BCAM-AKT2 fusion.
the longer isoform known as Lutheran or Lu (Fig. 3A). The Lu protein isoform is known to localize on the basolateral membrane characteristic of differentiated epithelial cells, whereas the BCAM isoform, which is overexpressed in ovarian cancers, is localized in a nonpolarized manner (22, 23). The fusion protein is predicted to encode most of AKT2, including its kinase domain (Fig. 3A). However, it lacks 97 amino acids from the N terminus of AKT2 that constitutes the pleckstrin homology (PH) domain necessary for targeting AKT2 to the cell membrane, where it normally gets activated (24). Because the BCAM product is a membrane protein, we predict that the BCAM-AKT2 fusion protein will localize to the membrane. To confirm the translation of the predicted BCAM-AKT2 fusion protein in HGSC patients, we performed Western blot analysis of patient tumor samples and noncancerous ovary and fallopian tube samples. As shown in Fig. 3B, the AKT2 antibody recognizes a band in patient S4 corresponding to the predicted fusion protein (∼110 kDa), which is larger than endogenous AKT2 (∼60 kDa). This 110-kDa band was absent in patient S27, who did not express the fusion transcript, and also lacking in noncancerous ovary and fallopian tube (Fig. 3B, Top). To confirm that the 110-kDa band is indeed the BCAM-AKT2 fusion protein, we performed a coimmunoprecipitation assay first by precipitating the target proteins using an anti-AKT2 antibody, followed by Western blot analysis using an anti-BCAM antibody. As shown in Fig. 3B (Bottom), the BCAM antibody recognizes the 110-kDa band that was precipitated by the AKT2 antibody but does not recognize the precipitated endogenous AKT2 (∼60 kDa). This confirmed that E1274 | www.pnas.org/cgi/doi/10.1073/pnas.1501735112
sion kinase is phosphorylated at serine 474 (Fig. 3C) suggests that loss of the PH domain is compensated by fusion to BCAM, a membrane protein that could relocate AKT2 to the cell membrane, where it gets activated by phosphorylation (24). To confirm this, we set out to generate epitope-specific antibodies that could specifically recognize the junction peptide of the fusion protein but not parental BCAM or AKT2. However, several monoclonal antibodies that we produced unfortunately failed to recognize the fusion protein specifically. This prevented the most direct confirmation of membrane localization of BCAM-AKT2 fusion protein in the patient’s tumor, although the fusion protein was clearly translated (Fig. 3B). To circumvent this difficulty, we cloned BCAM-AKT2 from patient S4, FLAG-tagged it, and expressed it ectopically in OVCAR8 cells. Western blot analysis confirmed the translation and phosphorylation of BCAM-AKT2 at both serine 474 (Fig. 3D) and threonine 309 of AKT2 (Fig. S1), suggesting that the fusion protein is in the active state. Immunocytochemical imaging showed that the fusion protein is mainly located at the cell membrane (Fig. 3E). Thus, BCAM-AKT2 fusion leads to a fusion protein that relocates the AKT2 kinase to the cell membrane, where it is phosphorylated. The phosphorylation of AKT2 and thus its activation are known to respond swiftly to extracellular growth stimuli such as insulin-like growth factor 1 (IGF1) (25). To answer the question of whether phosphorylation of BCAM-AKT2 fusion protein is also regulated by IGF1, we serum-deprived transfected OVCAR8 cells overnight so that AKT2 became unphosphorylated (26). Cells were then treated with IGF1 for 30 min to activate AKT2 followed by immediate protein extraction. Western blot analysis using an antibody against phosphorylated AKT2 (serine 474) shows that the endogenous AKT2 responds to IGF1 treatment swiftly through phosphorylation (Fig. 4A, Top, lane 3 versus 4). In contrast, BCAMAKT2 remains phosphorylated regardless of the presence or absence of IGF1 (Fig. 4A, Top, lane 1 versus 2). Further, there is no unphosphorylated BCAM-AKT2 detectable (Fig. 4A, Middle, lanes 1 and 2). This result suggests that BCAM-AKT2 is constitutively phosphorylated, presumably as a consequence of its membrane localization. To investigate whether the constitutively phosphorylated BCAM-AKT2 still possesses functional kinase activity, we immunoprecipitated BCAM-AKT2 from transfected OVCAR8 cells using an anti-FLAG antibody and incubated the fusion protein with GSK-3, a kinase substrate. As shown in Fig. 4B, incubation with immunoprecipitated BCAM-AKT2 led to efficient phosphorylation of GSK-3, indicating that BCAM-AKT2 is a functional and active kinase. Recreation of the BCAM-AKT2 Fusion Gene Using CRISPR/Cas9 Leads to Focus Formation. To investigate the functional consequences of
the BCAM-AKT2 fusion gene, we recreated the fusion gene in cells by means of chromosomal translocation induced using the CRISPR/Cas9 system. Unlike commonly used overexpression assays that use a CMV promoter to express cloned cDNA, the strategy we used generates a fusion gene with all genomic sequence elements intact, including the endogenous BCAM promoter and the appropriate UTRs and introns. The resulting fusion gene thus would mirror closely the BCAM-AKT2 fusion gene found in patients. To achieve this, we designed three guide RNAs targeting BCAM intron 13 to pair with three guide RNAs Kannan et al.
PNAS PLUS GENETICS
Fig. 3. BCAM-AKT2 gives rise to an in-frame fusion protein in patient tumors. (A) Protein domains of BCAM/Lu, AKT2, and the predicted BCAM-AKT2 fusion protein. As a result of the gene fusion, BCAM loses the Lu-specific region but retains all domains, including the transmembrane domain. AKT2 loses the PH domain but retains the kinase domains. Antibodies used against AKT2 and BCAM domains are shown. (B, Upper) Western blot using AKT2 antibody. (Lower) Western blot using BCAM antibody (after immunoprecipitation with AKT2 antibody). In both cases, a band corresponding to the predicted size of the fusion protein (∼110 kDa) was observed in patient S4 but absent in patient S27 or in noncancerous ovary and fallopian tube. IB, immunoblotting; IP, immunoprecipitation. (C) Western blot using phospho-AKT2 antibody indicates that BCAM-AKT2 is phosphorylated at serine 474 (lane 3). Controls include Jurkat cells treated with LY294002 (an inhibitor of AKT phosphorylation) or with Calyculin A (to preserve AKT in a phosphorylated state). (D) Western blot indicates that the cDNA cloned from patient S4 is translated into a full-length in-frame fusion protein in OVCAR8 cells as visualized by FLAG antibody. It is also phosphorylated as indicated by phospho-AKT2 (serine 474) antibody. (E) Immunocytochemistry of OVCAR8 cells transfected with the BCAM-AKT2-FLAG expression construct. The fusion protein was seen mainly on the cell membrane (arrows) as well as on the protruding filopodia (arrowheads) that form focal adhesions with the extracellular matrix. The deconvolution microscopy image was visualized by FLAG antibody. The image is a projection of seven stacks, each with 0.35-μm thickness.
targeting AKT2 intron 4 (Table S4) to induce double-stranded breaks at defined sites (Fig. 5A, Upper). As shown in Fig. 5A (Lower), transfection of nine combinations of paired guide RNAs with Cas9 in HEK-293T cells induced fusion transcript expression to various degrees depending on the combination used. Sanger sequencing of the RT-PCR band from the pair of BCAM-g3/ AKT2-g1 confirmed that the induced band is indeed the expected BCAM-AKT2 fusion transcript (Fig. S2). To assay for an oncogenic phenotype, OVCAR8 and HEK293T cells were transfected with BCAM-g3/AKT2-g1 along with Cas9 and then plated at a low density to allow individual cells to develop into foci. After 2 wk, foci were observed sporadically among the transfected population but not in the untransfected Kannan et al.
parental cells. To test whether the expression of BCAM-AKT2 is associated with focus formation, 44 prominent foci from transfected OVCAR8 cells and 26 from transfected HEK-293T cells were isolated and cultured. RT-PCR performed on these foci showed that ∼60% of the isolated foci from OVCAR8 (Fig. 5B) and 80% of the isolated foci from HEK-293T (Fig. 5C) were positive for the BCAM-AKT2 fusion transcript as confirmed by Sanger sequencing of the excised bands (an example from an OVCAR8 focus is shown in Fig. 5D). Thus, 60–80% of the isolated foci carry the BCAM-AKT2 translocation, and this is far greater than the expected low rate of long-range chromosome translocation that can be induced by the CRISPR/Cas9 system (1–4%) in cells (27). The observed focus formation, therefore, is PNAS | Published online March 2, 2015 | E1275
BCAM-AKT2 also differs from endogenous AKT2 in that the expression of the fusion AKT2 kinase is under the control of the BCAM promoter, not the AKT2 promoter, and phosphorylation of BCAM-AKT2 no longer responds to external stimuli. These observations suggest that the fusion AKT2 kinase escapes the tight control of normal cellular regulatory pathways, providing a new property that could significantly contribute to oncogenesis. This feature differs markedly from the effect of simple gene copy number increase in AKT2, which still could be tightly regulated by cells (3, 12). Excessive genome rearrangements that occur in HGSC make the analysis of gene fusion occurrence rate particularly challenging. An HGSC tumor could exist as a large mass composed of many mutational lineages. Depending on the part of the tumor that is available for experimental analysis, a mutational event could be completely missed even though it is present in other areas of the tumor. Further, different laboratories that obtain different parts of the same tumor may come to different conclusions because of the high degree of heterogeneity. Differences in methodology, such as reports that rely entirely on transcriptome sequencing and bioinformatics versus reports that rely on more sensitive experimental RT-PCR validation, could also contribute to the discrepancy observed by different studies
Fig. 4. BCAM-AKT2 is constitutively phosphorylated in transfected OVCAR8 cells. (A, Top) Western blot using phospho-AKT2 (serine 474) antibody shows that BCAM-AKT2 remains phosphorylated regardless of the presence or absence of IGF1 (lane 1 vs. 2). In contrast, the endogenous AKT2 remains largely unphosphorylated and responds to IGF1 treatment swiftly through phosphorylation (lanes 1 vs. 2, and 3 vs. 4). (Middle) No unphosphorylated BCAM-AKT2 was detected. (Bottom) Actin controls for protein loading. (B) The expressed fusion protein was immunoprecipitated by FLAG antibody and incubated with GSK-3 substrate. Western blot using anti–phospho-GSK-3 antibody showed that immunoprecipitated BCAM-AKT2 efficiently phosphorylated GSK-3.
not a random event but largely associated with BCAM-AKT2 expression. Discussion Fusion genes are among the most cancer-specific signatures known. In this report, we identified a novel recurrent fusion kinase, BCAMAKT2, that is specific to HGSC. Several key features are associated with the BCAM-AKT2 fusion. First, the exact same RNA junction is observed in the fusion transcript from different patients, suggesting the gain of a precisely defined fusion protein sequence. Second, it is an in-frame fusion, with both BCAM and AKT2 protein sequences remaining largely intact. Third, as demonstrated in Fig. 3, the fusion protein is clearly translated in HGSC patient tumors. Fourth, the fusion leads to a membrane-associated and constitutively activated AKT2 kinase, a key member in the PI3K/AKT pathway known to play a role in oncogenesis in HGSC (10). Fifth, BCAM-AKT2 is expressed in 7% of all HGSC tumors, a significant frequency given the highly heterogeneous nature of this malignancy. Because HGSC has increasingly been viewed as many diseases instead of a single disease, BCAM-AKT2 could represent an important genomic alteration for identifying a clinically relevant subtype of HGSC. Last, recapitulation of the BCAM-AKT2 fusion results in cellular transformation. E1276 | www.pnas.org/cgi/doi/10.1073/pnas.1501735112
Fig. 5. Generation of BCAM-AKT2 fusion in cells via chromosomal translocation leads to focus formation. (A) Schematic showing target sites of designed guide RNAs in BCAM intron 13 and AKT2 intron 4. Dashed red lines indicate genomic breakpoints identified in patient S4. Nine combinations of guide RNA pairs were transfected into HEK-293T cells. RT-PCR results show that the BCAM-AKT2 fusion transcript was detected in varying degrees depending on the combination of guide RNAs used. (B and C) (Upper) Examples of foci induced after transfection of BCAM-g3/AKT2-g1 in OVCAR8 and HEK-293T cells. Numbers indicate the ID number of the foci. (Lower) Presence of the BCAM-AKT2 fusion transcript as revealed by RT-PCR of foci isolated from transfected OVCAR8 and HEK-293T populations. (D) Representative Sanger sequencing of RT-PCR bands (an example from focus OVCAR8-46) confirmed that it contains the expected BCAM-AKT2 fusion junction.
Kannan et al.
1. Jemal A, et al. (2011) Global cancer statistics. CA Cancer J Clin 61(2):69–90. 2. Bell D, et al.; Cancer Genome Atlas Research Network (2011) Integrated genomic analyses of ovarian carcinoma. Nature 474(7353):609–615. 3. Bast RC, Jr, Mills GB (2012) Dissecting “PI3Kness”: The complexity of personalized therapy for ovarian cancer. Cancer Discov 2(1):16–18. 4. Mitelman F, Johansson B, Mertens F (2007) The impact of translocations and gene fusions on cancer causation. Nat Rev Cancer 7(4):233–245. 5. Tomlins SA, et al. (2005) Recurrent fusion of TMPRSS2 and ETS transcription factor genes in prostate cancer. Science 310(5748):644–648. 6. Singh D, et al. (2012) Transforming fusions of FGFR and TACC genes in human glioblastoma. Science 337(6099):1231–1235. 7. Honeyman JN, et al. (2014) Detection of a recurrent DNAJB1-PRKACA chimeric transcript in fibrolamellar hepatocellular carcinoma. Science 343(6174):1010–1014. 8. Romero I, Bast RC, Jr (2012) Minireview: Human ovarian cancer: Biology, current management, and paths to personalizing therapy. Endocrinology 153(4):1593–1602. 9. Garofalo RS, et al. (2003) Severe diabetes, age-dependent loss of adipose tissue, and mild growth deficiency in mice lacking Akt2/PKB beta. J Clin Invest 112(2):197–208. 10. Hers I, Vincent EE, Tavaré JM (2011) Akt signalling in health and disease. Cell Signal 23(10):1515–1527. 11. Bellacosa A, et al. (1995) Molecular alterations of the AKT2 oncogene in ovarian and breast carcinomas. Int J Cancer 64(4):280–285. 12. Cheng JQ, et al. (1992) AKT2, a putative oncogene encoding a member of a subfamily of protein-serine/threonine kinases, is amplified in human ovarian carcinomas. Proc Natl Acad Sci USA 89(19):9267–9271. 13. Altomare DA, et al. (2004) AKT and mTOR phosphorylation is frequently detected in ovarian cancer and can be targeted to disrupt ovarian tumor cell growth. Oncogene 23(34):5853–5857. 14. Yuan ZQ, et al. (2000) Frequent activation of AKT2 and induction of apoptosis by inhibition of phosphoinositide-3-OH kinase/Akt pathway in human ovarian cancer. Oncogene 19(19):2324–2330. 15. Kannan K, et al. (2014) CDKN2D-WDFY2 is a cancer-specific fusion gene recurrent in high-grade serous ovarian carcinoma. PLoS Genet 10(3):e1004216. 16. Kim P, et al. (2010) ChimerDB 2.0—A knowledgebase for fusion genes updated. Nucleic Acids Res 38(Database issue):D81–D85. 17. McPherson A, et al. (2011) deFuse: An algorithm for gene fusion discovery in tumor RNA-seq data. PLOS Comput Biol 7(5):e1001138.
Kannan et al.
PNAS PLUS
for leukemia (4) and TMPRSS2-ERG for prostate cancer (5). Because the fusion involves both BCAM and AKT2, the use of the currently available BCAM antibody and/or competitive AKT2 inhibitors should be evaluated for treatment of BCAM-AKT2– positive patients given that limited therapeutic options exist beyond standard chemotherapy. Last, this fusion protein may fold differently from the parental proteins. A small-molecule drug preferentially targeting this fusion protein can be developed, as has been demonstrated in the case of imatinib targeting BCR-ABL. Materials and Methods Anonymous ovarian cancer tissue samples were obtained from the Tissue Acquisition and Distribution Core of the Dan L. Duncan Cancer Center, Department of Pathology & Immunology, and Gynecologic Oncology Group under an approved Baylor College of Medicine Institutional Review Board protocol. Informed consent was obtained from all patients. All tumor samples were confirmed to have greater than 80% serous adenocarcinoma before processing. RNA was extracted from cancer samples and noncancerous samples using a RiboPure Kit (Ambion) and processed for RT-PCR or transcriptome sequencing as described earlier (15). Total RNA samples were processed for transcriptome sequencing using the Illumina mRNA-seq protocol. Bioinformatic identification of chimeric reads and corresponding junction reads is detailed in SI Materials and Methods. For RT-PCR, RT was performed using SuperScript III (Invitrogen), and PCR was performed using the primers listed in Table S2. The OVCAR8 cancer cell line was maintained using the media RPMI 1640, 10% (vol/vol) FBS, and 1% penicillin/streptomycin. Detailed materials and methods for immunoassays, genomic breakpoint identification, and CRISPR/Cas9 are included in SI Materials and Methods. ACKNOWLEDGMENTS. We thank Sriram Ayyaswamy and Thuy Phung for providing some of the antibodies, and Michael Ittmann and Patricia Castro for assisting with the cancer samples. We thank Radhika Dandekar, Fabio Stossi, and Michael Mancini for assisting with microscopy. S.M.H., M.M.M., and L.Y. have been supported by Ovarian Cancer Research Foundation OCRF PPD/BCM/01.12. L.Y. has been supported by Department of Defense W81XWH10-10327 and NIH R01EB013584. K.K. has been supported by a Prostate Cancer Foundation Young Investigator Award. A.M. was supported by a grant from the Weathervane Foundation. This project was also supported in part by the Genomic and RNA Profiling Core at Baylor College of Medicine with funding from an NIH National Cancer Institute Grant (P30CA125123).
18. Salzman J, et al. (2011) ESRRA-C11orf20 is a recurrent gene fusion in serous ovarian carcinoma. PLoS Biol 9(9):e1001156. 19. Stransky N, Cerami E, Schalm S, Kim JL, Lengauer C (2014) The landscape of kinase fusions in cancer. Nat Commun 5:4846. 20. Lieber MR (2010) The mechanism of double-strand DNA break repair by the nonhomologous DNA end-joining pathway. Annu Rev Biochem 79:181–211. 21. Zhang F, Carvalho CM, Lupski JR (2009) Complex human chromosomal and genomic rearrangements. Trends Genet 25(7):298–307. 22. El Nemer W, et al. (1999) Isoforms of the Lutheran/basal cell adhesion molecule glycoprotein are differentially delivered in polarized epithelial cells. Mapping of the basolateral sorting signal to a cytoplasmic di-leucine motif. J Biol Chem 274(45):31903–31908. 23. Garinchesa P, Sanzmoncasi M, Campbell I, Rettig W (1994) Non-polarized expression of basal-cell adhesion molecule B-CAM in epithelial ovarian cancers. Int J Oncol 5(6): 1261–1266. 24. Franke TF, et al. (1995) The protein kinase encoded by the Akt proto-oncogene is a target of the PDGF-activated phosphatidylinositol 3-kinase. Cell 81(5):727–736. 25. Liu AX, et al. (1998) AKT2, a member of the protein kinase B family, is activated by growth factors, v-Ha-ras, and v-src through phosphatidylinositol 3-kinase in human ovarian epithelial cancer cells. Cancer Res 58(14):2973–2977. 26. Watton SJ, Downward J (1999) Akt/PKB localisation and 3′ phosphoinositide generation at sites of epithelial cell-matrix and cell-cell interaction. Curr Biol 9(8):433–436. 27. Torres R, et al. (2014) Engineering human tumour-associated chromosomal translocations with the RNA-guided CRISPR-Cas9 system. Nat Commun 5:3964. 28. Micci F, et al. (2014) Low frequency of ESRRA-C11orf20 fusion gene in ovarian carcinomas. PLoS Biol 12(2):e1001784. 29. Andjelkovic M, et al. (1997) Role of translocation in the activation and function of protein kinase B. J Biol Chem 272(50):31515–31524. 30. Meier R, Alessi DR, Cron P, Andjelkovic M, Hemmings BA (1997) Mitogenic activation, phosphorylation, and nuclear translocation of protein kinase Bbeta. J Biol Chem 272(48):30491–30497. 31. Mende I, Malstrom S, Tsichlis PN, Vogt PK, Aoki M (2001) Oncogenic transformation induced by membrane-targeted Akt2 and Akt3. Oncogene 20(32):4419–4423. 32. Carpten JD, et al. (2007) A transforming mutation in the pleckstrin homology domain of AKT1 in cancer. Nature 448(7152):439–444. 33. Banerji S, et al. (2012) Sequence analysis of mutations and translocations across breast cancer subtypes. Nature 486(7403):405–409.
PNAS | Published online March 2, 2015 | E1277
GENETICS
(17–19, 28). It is possible that each experimentally validated fusion gene in HGSC may have a higher frequency of occurrence than is currently estimated, as it is not possible to sample each HGSC tumor in its entirety. The oncogenic consequences of membrane-associated AKT kinases have been documented by several studies. For instance, when AKT2 is myristoylated––a posttranslational modification that makes AKT2 membrane-associated––it leads to constitutively activated AKT2 in transfected cells (29, 30). Furthermore, when myristoylated AKT2 is expressed in cells, it results in loss of contact inhibition and leads to focus formation, an indicator of oncogenicity (31). Similarly, AKT1 mutation (E17K) in the PH domain found in breast and ovarian cancer patients leads to membrane-bound and constitutively activated AKT1 with increased focus formation (32). Perhaps the most intriguing example parallel to our discovery is the fusion gene MAGI3-AKT3 found in 7% of triple-negative breast cancer patients (33). In this case, AKT3 is fused in-frame to the membrane protein MAGI3. Although the translation of this fusion protein was not confirmed in patient tumor tissues, it was observed that in transfected cells, MAGI3-AKT3 was constitutively phosphorylated in the absence of growth factors and also oncogenic, as evidenced by focus formation (33). Consistent with the above reports, our experiments also demonstrate that the BCAM-AKT2 fusion kinase is indeed membrane-associated (Fig. 3E). In addition, by using the CRISPR/Cas9 system, we generated the BCAM-AKT2 fusion gene that mirrors closely the fusion gene found in patients, and this leads to focus formation (Fig. 5). Although other recurrent fusion genes, namely CDKN2DWDFY2 (15) and ESRRA-C11orf20 (18), are present in HGSC, the BCAM-AKT2 fusion reported here is the most clinically relevant and functionally significant and is the only cancer-specific fusion gene proven to translate a fusion protein in HGSC patients. Further, unlike TP53 and BRCA1/2 mutations that are present in many types of cancer, BCAM-AKT2 is “unique” to high-grade serous ovarian cancer, and therefore may represent a marker specific for HGSC in a manner similar to that of BCR-ABL
