Identification of Anaplastic Lymphoma Kinase as a Potential Therapeutic Target in Ovarian Cancer

Ovarian cancer is the fifth leading cause of death from cancer in women, following lung, breast, colon, and pancreatic cancers. In 2011, a total of 21,990 new cases and 15,460 deaths of ovarian cancer were projected in the United States (The SEER Program of NCI). More than 90% ovarian cancer cases are epithelial cancer, which represents a series of molecularly and etiologically distinct diseases (1). Epithelial ovarian cancer can be grouped into 2 types. Type I tumors are genetically stable and confined to the ovary at presentation. It includes low-grade serous, low-grade endometrioid, clear cell, mucinous, and Brenner carcinomas, 90% of which are curable. In contrast, type II tumors, including high-grade serous carcinoma, undifferentiated carcinoma, and malignant mixed mesodermal tumors are highly aggressive and usually present in advanced stages (2). Most type II tumors bear TP53 mutations and share no genetic alterations found in type I tumors (2, 3). High-grade serous carcinoma stands out from other subtypes because it accounts for 60% to 80% of ovarian cancer cases and most deaths from ovarian cancer. Recent studies have shown that high-grade serous carcinoma may arise from fallopian tube epithelium other than ovary itself (1, 4, 5). Because of the subtle nature of symptoms and inadequate screening tools, most serous ovarian carcinoma patients present at advanced stages with poor prognosis (2, 6). Although initial response rates to standard surgery and chemotherapy are high, the disease recurs in the majority of patients and eventually becomes chemoresistant (7). The fact that the overall survival rate for women with ovarian cancer has not changed over the last 30 years demands new therapeutic approaches. With advances in understanding the genetics and molecular biology of ovarian cancer, targeted therapy will likely have a significant impact on overcoming chemoresistant disease and improvement of patient outcome (1, 6, 8).

Oncogenic receptor tyrosine kinases (RTK) have been implicated in many types of solid tumors, including ovarian cancer. A recent report by Sheng and colleagues showed that an NRG1-Her3 autocrine loop could promote ovarian cancer cell proliferation (9). Overexpression of other RTKs, such as EGF receptor (EGFR), Her2, Her3, platelet-derived growth factor receptor, and EphA2, was reported in ovarian cancer (10–13). However, it remains to be determined whether these RTKs play any roles in tumor initiation and progression in ovarian cancer.

In this article, we used an immunoaffinity profiling approach (14–16) to identify activated tyrosine kinases in 69 primary ovarian tumors. Our results indicate that many RTKs are aberrantly phosphorylated in tumors. Most interestingly, phosphorylation of anaplastic lymphoma kinase (ALK) is identified in 3 high-grade serous carcinoma patients and one stromal tumor patient. When activated by mutations or translocations, ALK can drive oncogenesis in non-Hodgkin lymphoma (17), neuroblastoma (18–21), non–small cell lung cancer (NSCLC) (15, 22), inflammatory myofibroblastic tumor (23), and renal cell carcinoma (24, 25). Our biochemical and functional analysis suggests that, as it does in other malignancies, ALK is likely driving a subset of ovarian cancers and could be a new drug target in this devastating disease.

Ovarian tissues were collected randomly from 69 ovarian tumor patients and 19 patients with benign gynecologic conditions who needed oophorectomy in the Xiangya Hospital (Changsha, Hunan, China) from February 2009 to February 2010 with written consent from the patients (Institutional Review Board approval). Tissue specimens were taken intraoperatively and were either formalin-fixed and paraffin-embedded (FFPE) or snap frozen in liquid nitrogen. Histologic typing and grading were carried out according to International Federation of Gynecology and Obstetrics (FIGO) guidelines. General pathologic information of the 69 ovarian tumor patients is listed in Supplementary Table S1. FFPE ovarian tumor tissue microarray (TMA) slides were purchased from Folio Biosciences and Biochain Institute, Inc.

Ovarian cancer cell lines Ovsaho, Ovmana, and Ovmiu are purchased from Japanese Collection of Research Bioresources/Health Science Research Resources Bank. Antibodies against ALK (D5F3 XP), phospho-ALK (Y1278/1282/1283), SHC, phospho-SHC (Y234/240), Erk, phospho-Erk (T202/Y204), Stat3, phospho-Stat3(Y705), phospho-AKT (S473), phospho-S6 (S235/236), CDC2 (CDK1), phospho-CDC2 (Y15), and β-actin are from Cell Signaling Technology, Inc.

An average of 15 milligrams of peptides were prepared from 0.2 to 0.5 grams of resected frozen ovarian tissues by homogenization, trypsin digestion, and Sep-pak C18 column purification as described previously (14–16). Peptides containing phosphotyrosine were isolated by immunoprecipitation with a monoclonal antibody (mAb) against phosphotyrosine (pY100), concentrated on reverse-phase micro tips and analyzed by liquid chromatography tandem mass spectrometry (LC/MS–MS). Briefly, samples were collected with an LTQ-Orbitrap mass spectrometer, using a top-ten method, a dynamic exclusion repeat count of 1, and a repeat duration of 30 seconds. MS and MS/MS spectra were collected in the Orbitrap and LTQ component of the mass spectrometer, respectively. SORCERER-SEQUEST (TM, v4.0.3 (c) 2008; Sage-N Research, Inc.) searches were done against the NCBI human RefPept database downloaded on January 6, 2009 (containing 37,742 proteins) or March 1, 2010 (containing 36,500 proteins), allowing for serine, threonine, and tyrosine phosphorylation (STY+80) and methionine oxidation (M+16) as differential modifications. The PeptideProphet probability threshold was chosen to give a false positive rate of 5% for the peptide identification (26).

To assess potentially aberrant tyrosine phosphorylation of RTK in tumor tissues, spectral counts per RTK were summed and normalized to the amount of peptide subjected to phosphotyrosine immunoprecipitation (15 mg), elevated spectral count in each tumor sample was calculated by subtracting average spectral count in 19 normal ovarian tissues. Elevated phosphotyrosine spectral counts of RTKs, representing elevated tyrosine phosphorylation, observed in 60 tumor samples were used as basis for hierarchical Clustering using the Pearson correlation distance metric and average linkage (MultiExperiment Viewer version 4.4). RTKs with elevated phosphorylation in tumors are ranked based on the average value of elevated spectral count in corresponding tumors.

Frozen ovarian tumor tissue were minced in liquid nitrogen and resuspended in 1× cell lysis buffer. Tissue or cell suspension was then sonicated and cleared by centrifugation. Immunoblot analysis was carried out according to the manufacturer's standard protocol (www.cellsignal.com).

RNeasy Mini Kit (Qiagen) was used to extract RNA from human tumor samples. Rapid amplification of cDNA ends (RACE) was done with the use of 5′-RACE system (Invitrogen) with primer 5′-GCAGTAGTTGGGGTTGTAGTC for cDNA synthesis and 5′-GCGGAGCTTGCTCAGCTTGT and 5′-TGCAGCTCCTGGTGCTTCC for a nested PCR reaction, followed by cloning and sequencing the PCR products.

RNA was extracted from ovarian tumor tissue or cell lines using RNeasy Mini Kit (Qiagen). For reverse transcription PCR (RT-PCR), first-strand cDNA was synthesized from 2.5 mg of total RNA using SuperScript III first-strand synthesis system (Invitrogen) with oligo (dT)₂₀. Primer pairs used to amplify wild-type ALK and FN1-ALK transcripts were 5′-GAGGATATATAGGCGGCAAT, 5′-TGCAGCTCCTGGTGCTTCC, and 5′-TAAGCTGGGTGTACGACCAA, 5′TGCAGCTCCTGGTGCTTCC, respectively. Transcripts from the control gene glyceraldehyde-3-phosphate dehydrogenase (GAPDH) were amplified using primer pairs 5′-GATTCCACCCATGGCAAATTCC and 5′-GATTCCACCCATGGCAAATTCC.

DNeasy Blood & Tissue Kit (Qiagen) was used to extract genomic DNA from frozen ovarian tumor samples. Quantitative PCR (qPCR) was done using iQ SYBR Green Supermix (Bio-Rad). Each qPCR reaction contains 10 μL of SYBR Green Master Mix, 1 μL of 10 nmol/L forward primer, 1 μL of 10 nmol/L reverse primer, 6 μL of nuclease free water, and 2 μL of genomic DNA at 10 ng/mL. qPCR reaction was carried out with the following parameters: 95°C for 10 minutes followed by 40 cycles of 95°C for 15 seconds and 65°C for 15 seconds. A dissociation run from 55°C to 95°C was added to the end of qPCR reaction for melting curve analysis. Nonspecific amplification was not detected with the primers used in the study. PCR reactions were set up in duplicates for each sample tested. Data were exported from CFX96 Real-Time PCR Detection System and analyzed using Microsoft Excel. We used an Ultra Conservative Element (UCE) located on human Chromosome 7 (Chr7: 1,234,865–1,235,026) that has been identified to have a copy number of 2 as reference DNA segment and calculated the ΔC_t by subtracting the C_t value of the UCE reactions from that of the ALK reactions in each sample. This normalizes the variation in the amount of templates between samples. We then used ΔC_t values from each individual to calculate relative copy number against calibrator DNA isolated from NA10851 (Coriell Cell Repository). Calculations were carried out to normalize a copy number of 2 to a value of 1 in the bar graph. SDs were based on normalized C_t values of duplicates. The sequences of the primers that are used to amplify ALK gene are 5′-ACAAGGTCCACGGATCCAGAAACA and 5′-AGTCTCCCAGTTGCAACGTTAGGT.

The coding sequences of wild-type ALK and FN1-ALK were cloned from primary tumor samples OC26 and OC19, respectively, into retroviral vector MSCV-Neo. The expression plasmids were transfected into 293T cells by FuGene 6 (Roche Diagnositcs), and retrovirus was harvested 48 hours later. NIH3T3 cells were infected with MSCV-neo/ALK or MSCV-neo/FN1ALK and selected in G418-containing media for 7 days. The cells were then cultured in soft agar or injected subcutaneously into nude mice.

A total of 1.5 × 10⁶ to 2 × 10⁶ transduced 3T3 cells expressing Src, FN1-ALK, or ALK were resuspended in Matrigel (BD Biosciences) and injected subcutaneously into the upper or lower right flank of 6 to 8 weeks old female NCR NU-F mice purchased from Taconic. Mice were randomized to 3 treatment groups (n = 6 per group) once the tumors became palpable (∼50 mm³) and gavaged orally with vehicle (10% 1-methy-2-pyrrolidinone/90% PEG-300) alone, 100 mg/kg/d crizotinib or 10 mg/kg/d TAE684 (NVP-TAE684) for 7 to 11 days. Crizotinib and TAE684 were purchased from ChemieTek. Tumors were measured every other day using calipers, and tumor volume was calculated as 0.52 × width² × length. Mice were monitored daily for general conditions. The experiment was terminated when the mean size of the treated or control mice reached 1,500 mm³. Animal protocol is in accordance with Cell Signaling Technology Animal Care and Use Committee with approval ID 650.

To identify oncogenic kinases in ovarian tumors, we carried out immunoaffinity profiling of tyrosine phosphorylation in 69 tumor and 19 normal tissues with PhosphoScan (14). The characteristics of the 69 ovarian cancer patients are summarized in Table 1. Most of the 43 serous type tissues were high-grade serous carcinoma characterized by cancerous spreading in the pelvic area. LC/MS–MS identified 2,484 tyrosine phosphorylation sites in 1,349 proteins, with global false discovery rate to be less than 5% using peptide prophet (26). These data are available at PhosphoSitePlus (www.phosphosite.org). Spectral counts for all phosphotyrosine sites identified in tyrosine kinases across the 88 ovarian tissues are listed in Supplementary Table S2.

Previous studies have shown that elevated tyrosine phosphorylation of RTKs may represent aberrant kinase activity in tumor tissues, which could drive oncogenesis (15). Using RTK tyrosine phosphorylation in 19 normal ovarian tissues as basal phosphorylation level, we assessed potentially aberrant tyrosine phosphorylation of RTKs in tumor tissues. As shown in Fig. 1A, we observed elevated tyrosine phosphorylation in many RTKs across the 60 tumor tissues, including discoidin domain receptors (DDR), ephrin receptor kinases (EphA, EphB), as well as EGFR and FGFR family kinases. Interestingly, ALK was ranked as the top kinase with the highest elevated tyrosine phosphorylation among other RTKs (Fig. 1B).

ALK phosphorylation was found in 4 patients, including 3 serous carcinoma patients OC07, OC16, and OC26 and a stromal sarcoma patient OC19 (Fig. 1C). Hyperphosphorylation of ALK in OC19 is indicated by multiple phosphotyrosine sites in the juxtamembrane, kinase, and C-terminal regulatory domains, many of which were seen previously in other type of cancers (data not shown). Phosphorylation of Y1507, which is equivalent to Y567 in NPM-ALK found in anaplastic large cell lymphoma, was detected in all 4 patients (Fig. 1C). This phospho site was also observed in primary tumor tissues and cell lines of NSCLC bearing EML-ALK fusions, as well as neuroblastoma cell lines bearing ALK-activating mutations (data not shown). Extracted ion chromatograms, MS2 spectra of 2 tryptic peptides containing phosho-Y1507 indicated the detection of ALK peptides in these patients (Supplementary Fig. S1A and B).

Previous studies have shown that Y1507 (Y527 in NPM-ALK) located at the carboxyl terminal region of ALK is a direct docking site for SH2 domain-containing transforming protein (SHC1). Recruitment and phosphorylation of SHC1 activate Ras/extracellular signal–regulated kinase (ERK) pathway and promotes tumorigenesis (27, 28). Consistent with these observations, using label-free quantification analysis, we found hyperphosphorylation of SHC1, mitogen-activated protein kinase 14 (MAPK14), and CDK1 at specific sites (Supplementary Fig. S2 and Table S3) in the 3 serous carcinoma patients bearing ALK phosphorylation. In addition, phosphorylation of other ALK downstream signaling molecules, including those that are involved in Jak/Stat3, PI3K/AKT pathways, are upregulated in these tumors (Supplementary Fig. S2 and Table S3). These results suggested that in serous carcinoma patients OC07, OC16, and OC26, ALK might contribute to activation of multiple signaling pathways that promote tumor cell survival, growth, and proliferation.

Taken together, we found 4 of 69 (5.8%) ovarian cancer patients carrying phosphorylated ALK at sites that correlate with ALK oncogenesis, including 3 of 43 (7.0%) serous carcinoma and one stromal sarcoma patients. To our knowledge, this is the first time that ALK tyrosine phosphorylation is described in ovarian cancer.

Using an antibody directed against the ALK intracellular domain, we examined ALK protein localization and expression by immunohistochemical and Western blot analysis with patient tissue (Fig. 2). Similar to neuroblastoma cells and tumor tissues that overexpress ALK (21), we observed a diffused ALK staining at the cytoplasm of serous carcinoma OC26, but not in other serous carcinoma tissues such as OC29a (Fig. 2A). In the stromal tumor OC19, strong ALK signal is present at both the plasma membrane and the cytoplasm, with membrane accentuation in some cells (Fig. 2A). Western blot analysis detected mature full-length ALK at 220 kDa in OC16 and OC26, and the potential proteolytically cleaved 140 kDa form (29) in OC26, which might be too weak to be detected in OC16. We identified multiple ALK signals of variable sizes in OC19, with the most prominent signal at approximately 78 kDa, suggesting the presence of a novel form of ALK different from the EML4-ALK Variant 3 fusion protein in the lung cancer cell line H2228 (Fig. 2B). These results suggested that OC19 might carry a different gene translocation involving ALK.

As a neuronal receptor kinase, ALK is not normally expressed in the ovary. Expression and phosphorylation of ALK in 4 ovarian cancer patients prompted us to test whether the ALK gene in these patients may have undergone genetic alterations. As the activating mutations previously reported in neuroblastoma (18–20) were not detected in the cytoplasmic region of ALK in OC07, OC16, OC26, and OC19 (data not shown), we set out to examine other possible genetic alterations of ALK.

In neuroblastoma, ALK was identified as a recurrent target of copy number gain and gene amplification (18, 19, 21). Recent study conducted by The Cancer Genome Atlas (TCGA) research network indicated a remarkable degree of genomic disarray in serous carcinoma (30). To examine whether there is copy number variation (CNV) of ALK in ovarian tumors in which ALK phosphorylation was detected, we used real-time quantitative PCR with genomic DNA isolated from serous carcinoma OC07, OC16, OC26, and stromal sarcoma OC19, as well as serous carcinoma OC08, in which ALK phosphorylation is not detected. Although no obvious CNV of ALK was detected in OC08 and OC19, ALK copy number gains of 1.5- and 1.7-fold were detected in OC16 and OC26, respectively (Fig. 3A). The result is consistent with our high-resolution single-nucleotide polymorphism array analysis, which indicated a copy number of 3 in the ALK region in patient OC26 (data not shown). The ALK copy number gain in OC16 and OC26 may account for the aberrant expression of ALK in these serous carcinoma patients. Surprisingly, we observed a loss of ALK gene copies in serous carcinoma OC07 (Fig. 3A), in which ALK phosphorylation (Fig. 1A and C, Supplementary Fig. S1) and mRNA expression (see below) were detected, suggesting that aberrant ALK expression and phosphorylation is independent of ALK copy number in this patient.

Fusion of ALK kinase domain to a partner protein, resulting in constitutive activation of ALK and oncogenesis in non-Hodgkin lymphoma (17), NSCLC (15, 22, 31), inflammatory myofibroblastic tumor (23), and renal cell carcinoma (24, 25) has been reported previously. Given the presence of highly phosphorylated ALK with differing sizes in patient OC19 and the fact that no ALK mutation or copy number gain was detected in this patient, we suspected the presence of an ALK gene translocation. To test this possibility, we carried out 5′-RACE with RNA isolated from OC19 tumor. Because the majority of previously reported ALK fusions have a common breakpoint between exons 19 and 20 (15, 17, 22–25, 31), we searched the sequences upstream of this common breakpoint using a primer annealing to the 5′ end of exon 20. Cloning and sequence determination of the RACE product revealed that ALK is fused in frame to fibronectin 1 (FN1), encoding a ubiquitous component of extracellular matrix (ECM) and plasma (32), at a novel breakpoint between ALK exons 18 and 19. The resultant fusion protein FN1-ALK contains the amino-terminal 1,201 amino acids of FN1 and the carboxyl-terminal 598 amino acids of ALK, which include the transmembrane and cytoplasmic regions (Fig. 3B). Encoded by the first 23 exons, the FN1 portion of FN1-ALK retains a diverse set of binding domains involved in fibronectin self-association and interaction with other ECM components, which could potentially provide strong activating signal to ALK (Fig. 3B). Unlike previous ALK fusions, the novel breakpoint in FN1-ALK allows the ALK transmembrane region to be retained in the fusion protein (Fig. 3B). The predicted protein sequence of full-length FN1-ALK is shown in Supplementary Fig. S3B.

To test mRNA expression of FN1-ALK or full-length ALK in tumors with ALK phosphorylation, we carried out RT-PCR analysis with primers that specifically amplify fragments from FN1-ALK or full-length ALK cDNA (Fig. 3C). ALK mRNA expression was observed in the 3 serous carcinoma patients, OC07, OC16, and OC26, as well as in 2 ovarian cell lines, OVMANA and OVSAHO (Fig. 3C). In contrast, only FN1-ALK mRNA, but not full-length ALK mRNA, was detected in OC19 (Fig. 3C). As expected, neither full-length ALK nor FN1-ALK mRNA was detected in OC08 and the ovarian cancer cell line Ovmiu. These results indicated that FN1-ALK fusion is present in the stromal sarcoma patient OC19, whereas full-length ALK expression resides in serous carcinoma patients OC07, OC16, and OC26.

To examine the FN1-ALK gene fusion at genomic level, we used PCR primers annealing to the 3′ end of FN1 exon 23 and the 5′ end of ALK exon 19 to amplify DNA sequence spanning the joint of these 2 genes. Sequences of the PCR product revealed that the first 946 base pairs of FN1 intron 23 were fused to the last 14 base pairs of ALK intron 18 (Supplementary Fig. S3A). Given that ALK and FN1 map to the short and long arm of chromosome 2 (2p23 and 2q34), respectively, in the same orientation, it remains to be determined whether the formation of the FN1-ALK fusion was caused by a large interstitial deletion spanning the centromere or other chromosomal changes.

To determine whether full-length ALK and FN1-ALK have transforming potential, we generated NIH3T3 cells stably expressing full-length ALK or FN1-ALK (Fig. 4A and B). As expected, we observed both full-length ALK (220 kDa) and the 140 kDa cleaved ALK (29, 33) in 3T3/ALK cells. Interestingly, in addition to full-length FN1-ALK proteins (predicted molecular weight: 198.82 kDa), which migrates at approximately 225 kDa, we observed a prominent ALK signal of approximately 78 kDa in 3T3/FN1-ALK cells, which is a reminiscence of the strong ALK signal in patient OC19 tumor (Fig. 4B, Fig. 2B). We suspected that this approximately 78 kDa protein is a fragment derived from full-length FN1-ALK. Both ALK and FN1-ALK showed intracellular reticulum/Golgi localization (Fig. 4C), as indicated by immunofluorescence analysis, probably due to a defect of glycosylation of these overexpressed proteins in 3T3 cells (33). When 3T3/Neo, 3T3/ALK, or 3T3/FN1-ALK cells were injected subcutaneously into nude mice, we observed tumor growth induced by both 3T3/ALK and 3T3/FN1-ALK cells, with the FN1-ALK tumors growing more aggressively than the ALK tumors (Fig. 4C). The average tumor sizes 12 days after injection were around 2,000 mm³ and 100 mm³ for FN1-ALK and ALK tumors, respectively. If allowed to grow, ALK tumors could reach approximately 2,000 mm³ 20 days after injection (data not shown). These results indicated that both the full-length ALK and the FN1-ALK fusion isolated from the ovarian cancer patients have transforming potential and that FN1-ALK is more oncogenic than the full-length ALK.

To investigate downstream signaling pathways activated by full-length ALK and FN1-ALK, we analyzed endogenous signaling molecules of 293T cells when they overexpress full-length ALK and FN1-ALK. Both full-length ALK and FN1-ALK showed constitutive activation as indicated by the phosphorylation of tyrosine sites in their kinase domain (Supplementary Fig. S4). Interestingly, the 140 kDa cleaved ALK and the approximately 78 kDa FN1-ALK fragment had stronger phosphorylation than their full-length forms. Consistent with ALK activation, phosphorylation of Stat3, SHC, and Erk was upregulated, with FN1-ALK cells showing stronger upregulation of these molecules than ALK cells (Supplementary Fig. S4). ALK inhibitor crizotinib abolished phosphorylation of both ALK and FN1-ALK. As expected, phosphorylation of Stat3, SHC, and Erk was also diminished upon crizotinib treatment (Supplementary Fig. S4). Another ALK inhibitor TAE684 showed similar results (data not shown). These results indicated that, similar to what we observed in patient tumors using phosphoproteomic approach, expression of both ALK and FN1-ALK in 293T cells leads to activation of multiple downstream signaling molecules, which can be inhibited by ALK inhibitors. In addition, the stronger oncogenic activity of FN1-ALK than full-length ALK might be attributed to its stronger potential in activating downstream molecules, such as Stat3, SHC, and Erk, whose activation contributes to cell growth and proliferation.

Both PF-02341066 (crizotinib) and TAE684 are potent and selective ALK inhibitors that have been shown to suppress tumor cell growth in several types of cancers expressing different ALK fusion proteins (34–38). We tested the effects of these inhibitors on the growth of ALK and FN1-ALK tumors in nude mice. As shown in Fig. 5A, administration of crizotinib (100 mg/kg/d) drastically inhibited the growth of both ALK and FN1-ALK tumors, but not the 3T3/SRC tumors. Consistent with these observations, Western blot analysis using antibodies specific for ALK and phospho-ALK revealed that crizotinib abolished phosphorylation of both full-length FN1-ALK and the approximately 78 kDa ALK variant 24 hours after treatment (Fig. 5B). We also observed an ALK signal with lower molecular weight next to the full-length FN1-ALK in crizotinib-treated tumor, which is likely to be a signal of nonphophorylated FN1-ALK (Fig. 5B). Similar tumor growth-inhibiting effects were observed with TAE684 (Supplementary Fig. S5). These results suggested that ALK and FN1-ALK tumors are highly sensitive to ALK inhibitors.

To test possible aberrant ALK expression in a larger patient cohort, we carried out immunohistochemical analysis on ovarian TMAs. Similar to what we observed in the Chinese cohort (69 tumors) using PhosphoScan, although no ALK staining was detected in normal tissues, granulosa-theca cell tumors, or clear cell carcinomas, ALK staining was detected in 14 of 353 (4%) serous carcinoma and 3 of 37 endometrioid carcinoma specimens, with half of the specimens having strong ALK signal (Fig. 6A and B). We observed mostly weak ALK staining in some mucinous carcinoma specimens (Fig. 6A, images not shown). Consistent with what we found in patient OC26, most of the ALK staining observed in serous carcinomas in TMA was diffusedly localized in the cytoplasm (Fig. 6B). This result suggested that ALK might be activated and signaling in these tumors, similar to what has been observed in neuroblastomas, in which full-length ALK is overexpressed (21). Considering that a significant number of high-grade serous carcinoma has been misclassified as endometrioid carcinoma (1, 39), our data suggested that strong ALK expression resides mainly in a portion (∼2%) of serous carcinoma.

Oncogenic role of ALK has been well characterized in many types of hematopoietic and solid tumors, in which ALK is deregulated by gene translocations, activating mutations, and amplification (15, 17–25, 31, 40, 41). Although ALK transcript was reported to be present in epithelial ovarian cancer (42), ALK protein expression and phosphorylation in ovarian cancer has not been reported. Using a phosphoproteomic approach, we identified ALK tyrosine phosphorylation in 4 of 69 primary ovarian tumor tissues. Further investigation revealed ALK gene copy number gain in 2 serous carcinoma patients and a novel ALK gene translocation (FN1-ALK) in a stromal sarcoma patient. Mouse 3T3 fibroblasts overexpressing full-length ALK or FN1-ALK generated subcutaneous tumor in nude mice. In addition, we showed that ALK inhibitors dramatically suppressed the growth of both 3T3 ALK and 3T3 FN1-ALK tumors. This study provides new insight into the molecular pathogenesis of ovarian cancer and compelling rationale for extending targeted therapy against ALK to this new type of solid tumor.

Oncogenic role of full-length ALK has been implicated in neuroblastoma (21), glioblastoma (43), and breast cancer (44). Consistent with a recent report from TCGA Research Network, in which 489 high-grade serous ovarian adenocarcinomas were analyzed (30), we did not find any activating mutation or significant gene amplification of ALK in 41 serous carcinoma patients. However, our data showed phosphorylation of full-length ALK at Y1507 in 3 serous carcinoma patients, 2 of which have ALK gene copy number gain. The phosphorylation of multiple downstream signaling molecules such as SHC1, STAT3, tyrosine kinases (SYK, FRK, and FYN), SHIP 2 (INPPL1), MAPK14, CDK1, and phosphoinositide 3-kinase (PI3K), are upregulated compared with other serous carcinoma patients. These results indicate that previously reported Ras/ERK, Jak3/Stat3, and PI3K/Akt pathways, which mediate the effect of ALK activity and promote tumor cell proliferation and survival (27, 28), are activated in ALK-bearing serous carcinoma patients. Our proteomic approach also identified many other molecules such as NEDD9 and PKCδ (PRKCD) that are involved in transformation (45, 46) in these patients, although their relevance to ALK activity has yet to be determined. Aberrant expression, phosphorylation, cytoplasmic localization, as well as activation of ALK downstream signaling pathways suggest an oncogenic role of full-length ALK in serous ovarian carcinoma. Interestingly, unlike 3T3/FN1-ALK cells, 3T3/ALK cells did not generate transformed colonies in soft agar assay (data not shown). However, they generated tumors in nude mice. We suspect that full-length ALK could be activated through paracrine signaling under in vivo conditions. Functional validation using more pathologically relevant models such as primary tumor xenografts or fallopian tube secretory epithelial cell (4) models will be helpful in assessing the transforming effects of ALK and determine how full-length ALK is activated in serous carcinomas.

The novel fusion protein FN1-ALK identified in a stromal sarcoma patient represents the first ALK fusion protein bearing the transmembrane domain of ALK itself. Fusion to the FN1 gene that encodes a ubiquitously expressed ECM protein likely causes high-level expression of the ALK kinase. Unlike most ALK fusion partner proteins that render ALK constitutive activity through homodimerization, FN1 may do so through its strong interactions with ECM structures and components such as fibrin, heparin, collagen, and gelatin. Because fibronectin is subjected to proteolytic breakdown by a number of neutral proteases (47), the approximately 78 kDa fragment of FN1-ALK that we observed in both the patient tumor and transfected 3T3 or 293T cells is likely a proteolytically cleaved product of full-length FN1-ALK (∼220 kDa). Strong expression, hyperphosphorylation, and membrane/cytoplasmic localizations of ALK suggest constitutive internalization and activation of FN1-ALK, which may lead to its strong oncogenic potential, as evidenced by the strong activation of downstream molecules in 293T cells and the in vivo tumorigenecity assay. Considering the sensitivity of 3T3/FN1-ALK tumors to ALK inhibitors and the recent promising results of treating inflammatory myofibroblastic tumor (IMT) with ALK inhibitor crizotinib (38), we expect optimistic response of stromal sarcoma patients carrying FN1-ALK to targeted therapy against ALK. Our discovery broadens the scope of ALK fusion oncoproteins in human cancer and identifies a compelling therapeutic target in a new type of mesenchymal neoplasm.

To overcome chemoresistance and to improve the overall survival of ovarian cancer patients, targeted therapy has come to the forefront of investigation. Among reported oncogenes (8) associated with ovarian cancer identified to date, EGFR received substantial attention as a therapeutic target because of its overexpression in 70% ovarian cancer patients (48). However, inhibitors and monoclonal antibodies against EGFR only resulted in marginal efficacy due to toxic side effects and lack of inhibitor-sensitizing EGFR mutations in ovarian cancer patients (49). Unlike EGFR, which is ubiquitously expressed in normal tissue, ALK expression is restricted to certain parts of the brain (50). This might explain why inhibition of ALK has little toxic effects in patients (28). As shown recently in NSCLC and IMT patients, ALK inhibitor crizotinib only showed grade 2 side effects at a dose of 250 mg twice daily (37, 38). Compared with 10% response rate to second-line chemotherapy, the response rate of NSCLC patients carrying the EML4-ALK fusion to crizotinib is 57% (37). Given the presence of an ALK-driven subgroup of ovarian cancer, especially serous carcinoma patients, and the fact that crizotinib inhibited ALK or FN1-ALK tumor growth and activation of downstream signaling molecules, we speculate that crizotinib or other ALK-targeting agent may become effective treatment in these patients.

Besides ALK inhibitors, other therapeutic options for ALK bearing ovarian cancer patients might also be considered. First, coexpression/activation of other RTKs in ALK-bearing serous carcinoma patients, such as Ret (in OC26) (21), suggests that combinational therapy against multiple RTK could provide better outcome. Second, as ALK expression is tumor specific in ovarian cancer, antibody-mediated immunotherapy against ALK could be highly efficient with little damage to normal tissue. Third, as previously shown in anaplastic large cell lymphoma (28), ALK is likely to be highly immunogenic in ovarian cancer patients and thus could be an ideal target for antitumor vaccination in these patients.

Combining our results from the Chinese patient cohort and the immunohistochemical analysis of TMA containing more than 400 ovarian tumors of a distinct patient cohort, we estimate that 2% to 4% high-grade serous ovarian carcinoma patients have aberrant ALK expression. Given the annual approximately 140,000 worldwide new cases and approximately 7,000 deaths from high-grade serous ovarian carcinoma, the number of patients who might benefit from targeted therapy against ALK could be substantial.

No potential conflicts of interest were disclosed.

The authors thank Michael Lewis, Roy Scialdone, Curtis Desilets, Joan MacMeill, Julie Nardone, Cindy Reeves, and Yi Wang for technical support and Yu Li, Roberto Polakiewicz, and Long Ma for critical comments.

This study was partially supported by a grant from National Natural Science Foundation of China, 81101475, to Z.-P. Tan, X. Zhu, and Y.-F. Yang.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.