|
|
Background Granulosa-cell tumors (GCTs) are the most common type of malignant ovarian sex cord–stromal tumor (SCST). The pathogenesis of these tumors is unknown. Moreover, their histopathological diagnosis can be challenging, and there is no curative treatment beyond surgery.
Methods We analyzed four adult-type GCTs using whole-transcriptome paired-end RNA sequencing. We identified putative GCT-specific mutations that were present in at least three of these samples but were absent from the transcriptomes of 11 epithelial ovarian tumors, published human genomes, and databases of single-nucleotide polymorphisms. We confirmed these variants by direct sequencing of complementary DNA and genomic DNA. We then analyzed additional tumors and matched normal genomic DNA, using a combination of direct sequencing, analyses of restriction-fragment–length polymorphisms, and TaqMan assays.
Results All four index GCTs had a missense point mutation, 402C
Conclusions Whole-transcriptome sequencing of four GCTs identified a single, recurrent somatic mutation (402C
G (C134W), in FOXL2, a gene encoding a transcription factor known to be critical for granulosa-cell development. The FOXL2 mutation was present in 86 of 89 additional adult-type GCTs (97%), in 3 of 14 thecomas (21%), and in 1 of 10 juvenile-type GCTs (10%). The mutation was absent in 49 SCSTs of other types and in 329 unrelated ovarian or breast tumors. G) in FOXL2 that was present in almost all morphologically identified adult-type GCTs. Mutant FOXL2 is a potential driver in the pathogenesis of adult-type GCTs.
|
Methods for rapidly sequencing whole cancer genomes have recently been developed,11,12,13,14 and initial studies have suggested that hundreds of samples of any cancer type would have to be studied to derive clinically useful information.15 Because GCTs are genomically stable, we hypothesized that common, GCT-specific molecular abnormalities in the transcriptomes of GCTs could be discovered through the analysis of a small number of samples. We thus subjected four GCTs to whole-transcriptome paired-end RNA sequencing (also called RNA-seq) (see Glossary).
Methods
Patients and Samples
We selected 4 ovarian adult-type GCTs (1 primary and 3 recurrent tumors) obtained from index patients, along with 10 ovarian carcinomas and 1 cell line derived from a serous borderline tumor provided by the OvCaRe (Ovarian Cancer Research) frozen tumor bank. Patients provided written informed consent for research using these tumor samples before undergoing surgery, and the consent form acknowledged that a loss of confidentiality could occur through the use of samples for research. Separate approval from the hospital's institutional review board was obtained to permit the use of these samples for RNA-sequencing experiments.
For primary validation, we obtained 74 formalin-fixed, paraffin-embedded blocks of frozen samples from additional putative GCTs, along with 48 matched samples of normal tissue. We obtained two additional blocks (GCT78 and GCT59) that were recurrences of samples GCT29 and GCT76, respectively, which we also genotyped. In addition, we analyzed frozen tissue samples from 149 epithelial ovarian tumors and 180 breast carcinomas. We obtained another series of formalin-fixed, paraffin-embedded ovarian sex cord–stromal tumors (SCSTs), some of which have been described previously16 (see Glossary). This second series consisted of 95 tumor samples: 27 adult-type GCTs, 8 juvenile-type GCTs, 23 fibromas, 14 Sertoli–Leydig cell tumors, 13 thecomas, and 10 steroid-cell tumors. (Details regarding the numbers and types of GCTs that were provided by each institution are listed in Table 1 in the Supplementary Appendix, available with the full text of this article at NEJM.org.)
Pathological Review
All tumor samples that are included in this study were independently reviewed by a gynecologic pathologist before mutational analysis. In cases in which the review diagnosis differed from the source diagnosis, the samples were further reviewed by another gynecologic pathologist, who acted as arbiter. Both pathologists were unaware of the results of genomics studies. Immunohistochemical staining for calretinin, epithelial membrane antigen, and inhibin were performed. (For details, see the Methods section in the Supplementary Appendix.) Histologic images of all GCTs were obtained with the use of a ScanScope XT digital scanning system (Aperio Technologies) and are available on request.
Paired-End RNA Sequencing and Analysis
A detailed description of RNA sequencing and subsequent data analysis are provided in the Methods section of the Supplementary Appendix. Briefly, double-stranded complementary DNA (cDNA) was synthesized from polyadenylated RNA and sheared. The fraction from 190 to 210 bp was isolated and amplified with 10 cycles of a polymerase-chain-reaction (PCR) assay, according to the paired-end protocol for the Genome Analyzer II (GAII) (Illumina). The resulting libraries were then sequenced on the GAII. Short DNA sequences (reads) obtained from the GAII were mapped to the reference human genome (National Center for Biotechnology Information build 36.1, hg18) and a database of known exon junctions13 with the use of MAQ software17 in paired-end mode. Putative point mutations and small insertions and deletions were identified (for details, see the Methods section in the Supplementary Appendix). These mutations were cross-referenced against human genome databases to eliminate previously described germ-line variants. Genome instability of the index samples was determined with the use of Affymetrix 6.0 genotyping arrays interpreted for copy number.
Mutation Validation
We selected variants present in at least three of the four GCT RNA-sequencing libraries and examined RNA-sequencing libraries from 10 non-GCT ovarian tumors and from a serous borderline-tumor–derived cell line for the presence of the GCT-derived variants. Transcriptome variants that were absent in the 11 non-GCT ovarian tumors were classified as GCT-specific variants and subjected to further analysis. We confirmed the identification of the GCT-specific variants using Sanger sequencing of PCR amplicons of cDNA and genomic DNA (gDNA) obtained from the index samples (both tumor and matched normal tissue when available).
We used a combination of direct sequencing, analyses of restriction-fragment–length polymorphisms (RFLPs), and a TaqMan real-time PCR–based allelic discrimination assay (Applied Biosystems) to genotype the FOXL2 402CG mutation in additional cancers (see Glossary). In the primary validation studies, samples were scored as positive or negative for this mutation only if there were clear and concordant results from at least two assays. The use of assays with different primers was intended to minimize PCR artifacts caused by amplifying poor-quality DNA templates from formalin-fixed, paraffin-embedded tissue blocks.18 The TaqMan and RFLP assays produced clear and concordant results for all 149 non-GCT tumors, as well as for the 4 index GCT samples, and these results matched the results from sequencing of the 74 samples with interpretable results from all assays. In cases in which the PCR–RFLP analysis did not produce a PCR product of sufficient quality and quantity for further analysis (presumably because of poor-quality template DNA), it was replaced by the TaqMan assay, which was also used in primary screening for the additional series of SCSTs studied to determine the frequency of mutations in related tumors.
Results
Paired-End Whole-Transcriptome RNA Sequencing
The genomic stability of the four index GCTs was confirmed by inferring DNA copy-number alterations from high-density genotyping arrays (Fig. 1 in the Supplementary Appendix). The samples were then subjected to RNA sequencing. The total number of mapped DNA sequence reads and the total number of reads (which included the fraction of exonic, intronic, and intergenic sequences), the number of base pairs covered, and details regarding the variants detected in the 4 GCT samples and in the other 11 ovarian neoplasm-derived comparator libraries are shown in Table 1. Sequence reads from these samples showed expected subtype-specific differential gene expression (Fig. 2 in the Supplementary Appendix).
|
The RNA-sequencing analysis generates short reads that map to a reference sequence. Figure 2 depicts typical RNA-sequencing data for a portion of the FOXL2 gene and shows how nucleotides for expressed genes can be represented by multiple reads, as opposed to an average read generated by standard sequencing. We predicted the identification of mutations using SNVmix, a probabilistic model that is used to infer such a result on the basis of a binomial mixture model (see the Supplementary Appendix for details). Two putative point mutations that were implicated by SNVmix and a base-pair insertion met our criteria for candidate tumorigenic mutations, since they were present in three or more GCTs and absent in the comparator sequence libraries (i.e., sequence libraries generated from non-GCT tumors, in addition to sequences in existing data banks). All four GCTs contained a CG change in the FOXL2 gene at position 402 (genomic position 140147853 on chromosome 3; National Center for Biotechnology Information [NCBI] human genome build 36.1), which is predicted to result in the substitution of a tryptophan residue for a highly conserved cysteine residue at amino acid position 134 (C134W) (Figure 2). The Grantham distance (see Glossary) of this amino acid change was 215, the highest value we observed in the 6410 nonsynonymous predicted variants derived from the analysis of all 15 ovarian cancers (Table 1). Of the 6410 mutations, only 15 had a Grantham distance of 215. All predicted variants from the four index GCTs are available from the European Genotyping Archive (accession number, EGA00000000040).
|
G variant in the cDNA and gDNA of the four index samples and determined that it was somatic in the two samples from patients for whom normal tissue was available. The two other variants that met our criteria were one at genomic position 357452 on chromosome 4 in ZNF141 and a base-pair insertion at position 24488600 on chromosome 16 in RBBP6. We could not confirm either of these variants by PCR validation and considered them to be systematic artifacts generated by misalignment of the reads. We therefore focused on the FOXL2 402CG variant (Figure 3A, and Fig. 3 in the Supplementary Appendix).
|
G
We used a combination of RFLP, TaqMan (Figure 4A), and direct-sequencing (Figure 4B) analyses to determine whether the FOXL2 variant was present in 69 samples of pathologically confirmed GCTs (67 adult-type and 2 juvenile-type GCTs) and three other SCSTs (Figure 3B). We obtained unequivocal results in 64 of 69 GCTs and in all 3 SCSTs (Fig. 4 in the Supplementary Appendix). FOXL2 402CG was present in 59 of 64 GCTs. Two additional paired recurrences also carried the 402CG variant. The mutation was absent in DNA extracted from paired normal tissue from 48 mutation-positive patients. Five GCTs were mutation-negative, including the two juvenile-type GCTs (GCT24 and GCT45) (Fig. 5A and 5B in the Supplementary Appendix). Two mutation-negative tumors (GCT11 and GCT22) did not express inhibin or calretinin, which were expressed in 100% and 86% of the mutation-positive GCTs, respectively (Fig. 5C and 5D in the Supplementary Appendix). The fifth mutation-negative tumor (GCT33) had a prominent fibrothecomatous component, also seen in two mutation-positive samples (GCT44 and GCT71) (Fig. 5E in the Supplementary Appendix). The FOXL2 mutation was absent in 149 other epithelial ovarian tumors and in 180 breast cancers (Figure 3D).
|
G (C134W) mutation, as determined by sequencing and the TaqMan assay profile, or had a loss of heterozygosity of the normal allele. A single GCT (GCT44) had both the 402CG (C134W) mutation and a second FOXL2 somatic variant, 404AG (E135G) (Fig. 7 in the Supplementary Appendix). By sequencing cloned PCR products, we observed that these variants were in cis. This sample tested positive for the 402CG variant on RFLP analysis, but repeated TaqMan assays did not produce results. We did not detect the 404AG variant in any of the 4 index samples, in 37 other GCTs, or in 21 non-GCTs for which we had sequencing data.
We carried out fluorescence in situ hybridization to assess potential amplification of FOXL2 on 32 GCTs and 5 epithelial ovarian cancers and obtained negative results (Fig. 8 and Table 2 in the Supplementary Appendix). Immunohistochemical analysis showed that FOXL2 was expressed in the nuclei of normal granulosa cells as well as in GCTs that were heterozygous or appeared to be hemizygous or homozygous for the 402CG mutation (Figure 4C, 4D, and 4E).
Test for Replication
We obtained a second series of 95 clinical samples to determine the specificity of this variant within SCSTs of the ovary (Figure 3C). We found the FOXL2 mutation in all 27 adult-type GCTs that we tested. Among the other ovarian SCSTs, 1 of 8 juvenile-type GCTs and 2 of 13 thecomas carried the FOXL2 variant; all other tumors (14 Sertoli–Leydig cell tumors, 23 fibromas, and 10 steroid-cell tumors) were negative for the mutation.
Discussion
We found a recurrent somatic mutation (402CG) in FOXL2 in tumor samples from four patients with GCTs, using RNA sequencing to study the transcriptomes of the samples. FOXL2 was not expressed in the other RNA-sequencing libraries derived from ovarian cancers, and the mutation was absent in the gDNA of these cancers. We analyzed two additional series of GCTs for this mutation. The combined results from both series showed that the mutation was present in 86 of 89 morphologically identified adult-type GCTs (97%), in 1 of 10 juvenile-type GCTs (10%), and in 3 of 14 thecomas (21%); the mutation was absent in 49 other ovarian SCSTs, in 149 epithelial ovarian tumors, and in 180 breast cancers.
Although juvenile-type GCTs share some features with adult-type GCTs and have a similar biomarker-expression profile, they differ with respect to clinical presentation (in prepubertal children vs. young adults) and histopathological features.1 These features, along with those of a naturally occurring mouse model of juvenile-type GCT,20 suggest that juvenile-type GCT is a distinct disease from the adult type; our data support this hypothesis. The finding of a mutation that is characteristic of adult-type GCTs in thecomas is not surprising because the distinction between thecomas and GCTs can be arbitrary and mixtures of the two types of tumor cells may be seen in individual tumors. The diagnostic category of granulosa–theca-cell tumor has been used for such cases in the past.21 The presence of the FOXL2 mutation in a subgroup of thecomas suggests a cut point on this morphologic continuum between thecoma and GCT. Studies of additional patients with follow-up will be required to determine whether thecomas with a FOXL2 mutation are biologically more similar to GCTs than to mutation-negative thecomas.
The three mutation-negative adult-type GCTs were diagnostically challenging. Two of these tumors had immunohistochemical profiles that were distinct from those of most GCTs, suggesting that they were not true GCTs but, rather, morphologic mimics. We selected GCTs for this study on the basis of light microscopy (reflecting current diagnostic practice), not on the basis of immunostaining profiles; had we taken immunostaining into consideration, we would have excluded these two tumors.
FOXL2 is a member of the forkhead–winged-helix family of transcription factors containing a highly conserved DNA-binding forkhead domain. It is one of the earliest markers of ovarian differentiation, and its expression persists into adulthood. FOXL2 is required for the normal development of granulosa cells22,23 and shows strong expression in granulosa cells and moderate expression in stromal cells; no expression has been detected in oocytes.22,24 Few targets of FOXL2 have been described; it has been shown to have a role, as part of an AP1–SMAD3–SMAD4 complex, in activating the transcription of GNRHR (encoding the gonadotropin-releasing hormone receptor) in pituitary cells and repressing the transcription of STAR (encoding steroidogenic acute regulatory protein) in the adult ovary.25,26 Granulosa-cell proliferation and differentiation are governed, at least in part, by transforming growth factor β–receptor signaling through SMAD2 and SMAD3.26
To date, all mutations that have been described in FOXL2 are germ-line loss-of-function mutations and are associated with the blepharophimosis–ptosis–epicanthus inversus syndrome27,28 with primary ovarian failure (in particular, granulosa-cell failure29). Like JAK2 mutations in the case of polycythemia vera,30 mutations in FOXL2 are present in a large majority of adult-type granulosa-cell tumors and involve a single base substitution. Cys 134 is located on the surface of the forkhead DNA-binding domain; modeling suggests that the substitution of tryptophan for cysteine does not disrupt the folding of this domain or its interactions with DNA (Fig. 9 in the Supplementary Appendix). We therefore speculate that its pathogenicity is imparted through other mechanisms, such as the altering of one or more interactions between FOXL2 and other proteins. We have ruled out mislocalization of the mutant protein as a probable cause of its pathogenicity by showing that FOXL2 is present in the nuclei of GCTs (Figure 4) that are either heterozygous or appear to be hemizygous or homozygous for the 402CG mutation (Table 2 in the Supplementary Appendix).
Next-generation sequencing has been used to characterize mutations and gene expression in cell lines and to identify mutations in patients with acute myeloid leukemia.12,13,31 On the basis of these studies and others, it has been suggested that the genomic complexity of cancers is so extreme that whole-genome–sequencing studies must be performed at great depth and must include several hundred cancers of any type to yield data that can be interpreted clinically or biologically.11,14,15 However, our data suggest that for some tumors (most likely the cytogenetically simple and clinicopathologically homogeneous ones, such as GCTs), the pattern of somatic mutations is recurrent and constrained and thus can be analyzed by studying a small number of samples. Since transcriptomes are less complex and 50 to 100 times smaller than genomes, sequencing transcriptomes, as opposed to genomes, has been proposed as a more efficient method of finding mutations expressed through RNA.12 Our study supports this proposal. This approach cannot detect certain types of mutations, such as noncoding changes and mutations that are subject to nonsense-mediated decay.
The diagnosis of GCTs can be difficult, and treatment for this cancer is nonspecific and often unsuccessful. We speculate that testing for the presence of the FOXL2 402CG mutation may improve diagnosis in problematic cases and lead to more targeted therapies. In addition, new insights into the cause of this disease may be derived from future studies of the function of this mutation.
Glossary
Bit: A measure of the purity of the distribution of alleles at each nucleotide position, with 2 representing a homozygous position.
Grantham distance: A measure of amino acid dissimilarity developed by Grantham in 1974 on the basis of composition, polarity, and molecular volume of amino acids.
Paired-end sequencing: Sequencing of fragmented and size-selected complementary DNA from both ends, resulting in two short sequences with an unsequenced insertion between the sequences.
Restriction-fragment–length polymorphism: A variation in the DNA sequence that can be determined by an assay that takes advantage of a restriction site that is either introduced or removed by a polymorphism; the assay verifies the presence of the polymorphism with the use of restriction-enzyme digestion, followed by gel electrophoresis.
RNA sequencing: A standard term for whole-transcriptome sequencing with the use of high-throughput parallel processing technology to sequence randomly fragmented, size-selected DNA (see the Supplementary Appendix for details).
Sequencing read: A short string of decoded nucleotides that is obtained from a single fragment of DNA or RNA. Millions of reads are compiled through analytic software to determine allelic frequency at each position.
Sex cord–stromal tumor: Any tumor that is derived from ovarian stroma, including sex cords, forerunners of the endocrine apparatus of the postnatal ovary.
TaqMan: A rapid fluorophore-based real-time polymerase-chain-reaction assay that uses specific primers and labeled probes to detect sequence variants.
Transcriptome: The set of all messenger-RNA transcripts that are present in a cell or population of cells.
Supported by grants from the British Columbia Cancer Foundation and the Vancouver General Hospital Foundation (to OvCaRe in British Columbia), Genome Canada (to the Genome Sciences Centre), the Michael Smith Foundation for Health Research (to OvCaRe and the Genome Sciences Centre and to Drs. Shah, Marra, and Huntsman), Eli Lilly Canada (to Dr. Köbel), and the Canadian Institutes of Health Research's Training Program on Clinician-Scientists in Molecular Oncologic Pathology (to Dr. Turashvili); nondirected research funds from Sanofi-Aventis (to the Genetic Pathology Evaluation Centre); and funding for a Canada Research Chair in Molecular Oncology (to Dr. Aparicio). The contributing tumor banks were supported by OvCaRe and Ovarian Cancer Canada (Vancouver General Hospital), Fonds de la Recherche en Santé du Québec (Montreal), and Ovarian Cancer Canada (Ottawa). The Westmead Gynaecological Oncology Tissue Bank is a member of the Australasian Biospecimens Network-Oncology group, which is funded by the Australian National Health and Medical Research Council; by Cancer Research UK, the University of Cambridge, Hutchinson Whampoa, the Experimental Cancer Medicine Centre, and the National Institute for Health Research's Cambridge Biomedical Research Centre; by a grant (DAMD17-01-1-0729) from the U.S. Army Medical Research and Materiel Command; and by grants from the Cancer Council Tasmania, the Cancer Foundation of Western Australia, and the National Health and Medical Research Council of Australia to the Australian Ovarian Cancer Study.
Dr. Köbel reports receiving grant support from Eli Lilly; and Dr. Huntsman, receiving grant support from Sanofi-Aventis and Novartis and being listed as an inventor on a provisional patent application filed by the British Columbia Cancer Agency on the diagnostic use of the FOXL2 402C
We thank Ying Ng and Katie Chow for performing the array comparative-genomic-hybridization experiments; Krista Marcon for performing the immunohistochemical analysis and preparing tissue sections for DNA extraction; Samuel Leung for establishing the online image archive; Mona Mazgani and Salvador Saldivar for helping in the acquisition of informed consent and specimens for the gynecologic tumor bank at Vancouver General Hospital; Catherine Kennedy for helping in the acquisition of specimens from the Westmead Gynaecological Oncology Tissue Bank; Julie Irving for helping with accessing specimens for the secondary validation; the members of the Australian Ovarian Cancer Study (AOCS) management group: D. Bowtell, G. Chenevix-Trench, A. Green, P. Webb, A. deFazio, and D. Gertig; the AOCS study nurses and research assistants; and all the patients who participated in the AOCS study.
G mutation and its development as a therapeutic target. No other potential conflict of interest relevant to this article was reported.
Source Information
The authors' affiliations are listed in the Appendix.
This article (10.1056/NEJMoa0902542) was published on June 10, 2009, at NEJM.org.
Address reprint requests to Dr. Huntsman at the Center for Translational and Applied Genomics, British Columbia Cancer Agency, 600 W. 10th Ave., Vancouver, BC V5Z 4E6, Canada, or at dhuntsma{at}bccancer.bc.ca.
References
The authors' affiliations are as follows: the Centre for Translational and Applied Genomics (S.P.S., J.S., K.C.W., G.L., A.Z., S.E.K., M.S., R.G., E.Y., N.B., D.G.H.), the Genetic Pathology Evaluation Centre (M.K., E.M., P.B.C., C.B.G., D.G.H.), the Genome Sciences Centre (R.D.M., S.J., R.V., Y.Z., T.Z., M.H., M.M.), the Cheryl Brown Ovarian Cancer Outcomes Unit (K.D.S.), the Department of Gynecology (D.M.), and the Department of Molecular Oncology (J.K., G.T., S.A.A.), British Columbia Cancer Agency; Vancouver General Hospital (M.K., E.M., P.B.C., D.M., C.B.G., D.G.H.); and the University of British Columbia (M.K., E.M., P.B.C., P.L., C.B.G., D.G.H.) — all in Vancouver; the Department of Pathology, University of Toronto and University Health Network, Toronto (B.A.C.); the Departments of Pathology and Laboratory Medicine (C.C.) and Cellular and Molecular Medicine (B.V.), University of Ottawa, Ottawa; the Centre de Recherche du Centre Hospitalier de l'Université de Montréal, Montreal (J.M., D.P., A.-M.M.-M.); and the Department of Biochemistry, University of Alberta, Edmonton (J.N.M.G.) — all in Canada; the Department of Gynaecological Oncology, Westmead Hospital and Westmead Institute for Cancer Research, University of Sydney at Westmead Millennium Institute, Westmead (A.D., D.D.L.B.); Peter MacCallum Cancer Centre, Melbourne (D.D.L.B.); Queensland Institute for Medical Research, Brisbane (D.D.L.B.); and Victorian Cervical Cytology Registry, Melbourne (D.D.L.B.) — all in Australia; the Department of Pathology, Magee–Women's Hospital, University of Pittsburgh Medical Center, Pittsburgh (C.Z.); and the Functional Genomics of Ovarian Cancer Laboratory, Cancer Research UK Cambridge Research Institute, Li Ka Shing Centre, Cambridge, United Kingdom (C.A.P., M.J.-L., J.D.B.).
| |||||||||||||||||||||||||||||||||||||||||||||||||||||
This article has been cited by other articles:
HOME | SUBSCRIBE | SEARCH | CURRENT ISSUE | PAST ISSUES | COLLECTIONS | PRIVACY | TERMS OF USE | HELP | beta.nejm.org Comments and questions? Please contact us. The New England Journal of Medicine is owned, published, and copyrighted © 2010 Massachusetts Medical Society. All rights reserved. |
Previous
