Ovarian dysgerminomas are characterised by frequent KIT mutations and abundant expression of pluripotency markers

Background Ovarian germ cell tumours (OGCTs) typically arise in young females and their pathogenesis remains poorly understood. We investigated the origin of malignant OGCTs and underlying molecular events in the development of the various histological subtypes of this neoplasia. Results We examined in situ expression of stem cell-related (NANOG, OCT-3/4, KIT, AP-2γ) and germ cell-specific proteins (MAGE-A4, NY-ESO-1, TSPY) using a tissue microarray consisting of 60 OGCT tissue samples and eight ovarian small cell carcinoma samples. Developmental pattern of expression of NANOG, TSPY, NY-ESO-1 and MAGE-A4 was determined in foetal ovaries (gestational weeks 13–40). The molecular genetic part of our study included search for the presence of Y-chromosome material by fluorescence in situ hybridisation (FISH), and mutational analysis of the KIT oncogene (exon 17, codon 816), which is often mutated in testicular GCTs, in a subset of tumour DNA samples. We detected a high expression of transcription factors related to the embryonic stem cell-like pluripotency and undifferentiated state in OGCTs, but not in small cell carcinomas, supporting the view that the latter do not arise from a germ cell progenitor. Bilateral OGCTs expressed more stem cell markers than unilateral cases. However, KIT was mutated in 5/13 unilateral dysgerminomas, whereas all bilateral dysgerminomas (n = 4) and all other histological types (n = 22) showed a wild type sequence. Furthermore, tissue from five phenotypic female patients harbouring combined dysgerminoma/gonadoblastoma expressed TSPY and contained Y-chromosome material as confirmed by FISH. Conclusion This study provides new data supporting two distinct but overlapping pathways in OGCT development; one involving spontaneous KIT mutation(s) leading to increased survival and proliferation of undifferentiated oogonia, the other related to presence of Y chromosome material and ensuing gonadal dysgenesis in phenotypic females.


Background
The three main categories of ovarian tumours are: surface epithelium-stromal tumours, sex-cord tumours and germ cell tumours (GCTs; benign and malignant). Malignant ovarian (OGCTs) have a median age at onset of 18 years and represent approximately 3% of all ovarian cancers in Western countries [1]. The following histological subtypes exist: dysgerminoma, yolk sac tumour, embryonal carcinoma, polyembryoma, choriocarcinoma, immature teratoma, and mixed GCTs. Bilateral tumours occur in up to 10% of cases [1,2]. The correct differential diagnosis is imperative since the prognosis and choice of therapy remain different among the various ovarian cancer types. A treatment consisting of a combination of surgical resection and platinum-based chemotherapy cures the majority of malignant OGCT patients [2]. The fact that OGCTs often affect women in their reproductive years further imply the importance of optimal therapy in order to maximize the number of women in which ovarian function can be conserved.
Many similarities exist between GCTs of the ovary and testis, including a morphological resemblance and a similar pattern of chromosomal alterations [3,4]. Furthermore, families with both ovarian and testicular GCTs have been reported, suggesting a possible association/common genetic aetiology [5,6]. Ovaries and testes develop similarly until approximately 2 months of embryonic life, which is also consistent with a common origin of, at least some cases of, ovarian and testicular GCTs. In this study, recent knowledge of underlying mechanisms in development of testicular GCTs was used as a guide to investigate patterns of in situ protein expression in OGCTs. As a close resemblance between testicular GCTs and embryonic stem cells has been shown [7] particular focus was on stem cell-related factors including KIT (also known as c-Kit, tyrosine kinase receptor for stem cell factor (SCF)), OCT-3/4 (POU5F1, a POU-family transcription factor), NANOG, and AP-2γ (TFAP2C, transcription factor activator protein-2), and focus was also on germ cell-specific proteins (including MAGE-A4 and NY-ESO-1 belonging to the cancer/testis gene family) with a cell differentiation related biological function or a developmentally regulated expression pattern [8]. KIT is involved in the migration of primordial germ cells (PGCs) [9] and there has been reported a frequent presence of KIT mutations in GCTs and in particular bilateral testicular GCTs [10][11][12][13][14][15][16][17]. Therefore KIT mutation status was also determined in the same samples and we established the expression pattern of proteins not previously studied in the ovary during foetal ovarian development. Finally, patients with intersex disorders and dysgenetic ovaries have an increased risk of harbouring a GCT, and the ovarian malignancies were therefore investigated for the presence of Y-chromosome material. Gonadoblastomas are rare neoplasms com-posed of germ cells and immature granulosa/Sertoli cells that develop nearly exclusively in males and phenotypic females harbouring Y-chromosome material. We chose TSPY as a marker, because this gene has been mapped to a smallest region of the Y chromosome consistently present in females with gonadoblastoma, and TSPY has been proposed to be responsible for the origin of this tumour [18]. We attempted to analyse the various steps from the hypothesised development of some dysgerminomas from gonadoblastoma, considering why these neoplasias develop and progress.

Immunohistochemistry pattern of expression in OGCTs and foetal ovaries
The expression pattern in OGCTs of a panel of markers for testicular GCTs is summarised in Table 1 and illustrated in Fig. 1 and 2. Placental alkaline phosphatase (PLAP), a classical marker of GCTs, was expressed in 100% of dysgerminomas, gonadoblastomas, embryonal carcinomas, and in 46% of yolk sac tumours. Expression of the stem cell-related markers OCT-3/4 and KIT were present in 80% of dysgerminomas, 75-100% of tumours containing gonadoblastoma and in two yolk sac tumours, whereas NANOG and AP-2γ were present in approximately half of the dysgerminomas and gonadoblastomas. All bilateral (n = 6) OGCTs were KIT positive. Overall the tumours containing dysgerminoma and/or gonadoblastoma showed expression of 0-4 stem cell markers (mean 2.5). Only one dysgerminoma and one dysgerminoma with gonadoblastoma did not express any of the stem cell markers. Bilateral cases expressed a mean of 3.4 stem cell markers, whereas unilateral cases expressed a mean of 2.1 markers. MAGE-A4 staining was present heterogeneously in 40% of dysgerminomas and in 13% of dysgerminomas with gonadoblastoma. No OGCTs showed expression of NY-ESO-1 (CTAG1B/LAGE) or AMH (anti-Müllerian hormone). As the OGCTs may develop in dysgenetic ovaries with some testicular differentiation, we examined the expression of AMH (anti-Müllerian hormone), a glycoprotein involved in involution of Müllerian ducts in the foetus leading to male differentiation [19], but did not detect expression in any OGCTs. Analysed cores of tumour tissues from mixed OGCTs did not differ in expression pattern from cores from pure OGCTs. Most of the markers expressed by dysgerminoma and gonadoblastoma were negative in non-dysgerminomas and none of the ovarian small cell carcinomas expressed any of the analysed markers. In foetal ovaries of all analysed ages, strong expression of MAGE-A4 was detected in the majority of oogonia (Fig. 2) and NY-ESO-1 in a subset of oogonia. However, neither of the two was expressed in developing follicles. NANOG and TSPY were not expressed in foetal ovary of GW 13-40.

Analysis of tumour DNA for KIT gene mutations
PCR amplification was performed on 52 tumour DNA samples from 43 patients (Table 2 and Fig. 3). KIT was mutated in codon 816 in 5/14 unilateral dysgerminomas, whereas all dysgerminomas from bilateral cases (n = 4) showed a wild type sequence, as did all OGCTs of other histotypes (n = 22). A mutation of G to C in the first base of the codon (GAC to CAC) was found once, whereas a G to T (GAC to TAC) was detected in two cases. The second base exchange GAC to GTC was also detected in two cases. The mutations in KIT were not linked to KIT protein overexpression, as 2/5 samples containing mutations were negative for KIT. On the other hand, all cores from bilateral cases were KIT positive by immunohistochemistry.

Presence of Y-chromosome material in dysgerminoma/ gonadoblastoma samples
TSPY protein was present in six tumour tissue cores from five patients, all harbouring dysgerminoma and/or gonadoblastoma elements. Four of five patients had bilateral neoplasia. Immunohistochemistry for TSPY was then repeated on whole tumour sections with identical results. Karyotypes of normal tissues (e.g. blood cells) of the OGCT patients were not available in the majority of cases, In situ expression pattern of germ cell-related markers (MAGE-A4 and TSPY) in OGCTs and sex centromere material in inter-phase nuclei from OGCTs

Discussion
In this study, we provide novel evidence for distinct, partly overlapping mechanisms acting in the pathogenesis of OGCTs. To our knowledge, this study is the first comprehensive analysis of an in situ protein expression profile of pluripotency genes and germ cell differentiation markers in these relatively rare malignant tumours. The expression of the transcription factors OCT-3/4 and NANOG were of particular interest, as they are key regulators of selfrenewal and pluripotency of embryonic stem cells. In addition to the role in early embryonic development, these proteins are only present in PGCs, oogonia and gonocytes and in the testis only in malignant GCTs [20][21][22]. In the present study, OCT-3/4 expression was abundant in all gonadoblastomas and most dysgerminomas, in accordance with previously published data [23]. NANOG has not previously been evaluated in the ovary. Approximately half of the dysgerminomas and gonadoblastomas expressed NANOG, whereas tumours with somatic differentiation were largely negative. We did not detect NANOG expression in foetal ovaries, but cannot  exclude expression in first trimester. Murine foetal ovary expresses Nanog in early foetal development with earlier downregulation than in male gonads [24]. Our results indicate that in analogy to the mouse model, NANOG could to be earlier down-regulated in human foetal ovary than testis, where gonocytes abundantly express NANOG until GW 20 [22].
The tissue specific adult stem cell-related factors AP-2γ and KIT are highly expressed in immature gonocytes, testicular carcinoma in situ (the pre-invasive stage of testicular GCTs) and in seminomas, and AP-2γ, but not KIT, is expressed in embryonal carcinoma [25][26][27]. AP-2γ expression was previously reported in ovarian carcinomas [28] and in 5/5 dysgerminomas [29]. In contrast to the male pattern, AP-2γ was only expressed in 33% of dysgerminomas and 38% of dysgerminomas containing gonadoblastoma. Higher oestrogen levels in the ovaries may in part explain the observed difference, as AP-2γ has been sug-gested involved in oestrogen signalling [27]. The role of AP-2γ in the gonads has not yet been elucidated but disruption of the gene in mice leads to complete loss of germ cells, probably because of failure of KIT induction, which is hypothesised to be a main target for AP-2γ [30].
The KIT/SCF system is of particular interest in regard to the origin of GCTs, due to a role in PGC proliferation and survival in the developing human gonad [9]. KIT expression was previously described in oogonia and oocytes [25,31,32] and in some OGCTs, predominantly dysgerminomas [33][34][35]. In accordance with these reports, we found KIT expressed in 80-100% of neoplastic cells in dysgerminomas and gonadoblastomas, and in 15% of yolk sac tumours. The high expression of KIT in dysgerminoma prompted us to investigate the possibility of a gene mutation. A gain-of-function mutation in the KIT gene was first reported in human GCTs, including one ovarian dysgerminoma/yolk sac tumour by Tian et al. [10] and later studies have shown KIT mutations in a varying proportion of phenotypically indistinguishable testicular seminomas [12,14,15,36], primary mediastinal seminomas [11] and intracranial germinomas, both in males and females [13,17]. Among 11 OGCTs previously analysed for KIT mutations two (one unilateral case -for the other such info was not given) had a GAC to CAC change in codon 816 [10,14,16]. We detected five KIT mutations in the same codon, but found three different base changes in five OGCTs (four pure dysgerminomas and one dysgerminoma with gonadoblastoma). Strikingly, all mutations were found in unilateral cases and none among four analysed bilateral cases. This could suggest that KIT mutations occur after completion of PGC migration to the gonadal ridges in females, and may be in contrast to what has been suggested for males, where some reports have shown KIT mutations more frequently in bilateral seminomas [12,15]. We hypothesise that activating KIT mutation(s) in early oogonia may stimulate their proliferation and delay meiotic entry, which in normal ovaries occurs gradually in foetal life, thus increasing the pool of immature germ cells.

A model for development of ovarian and testicular germ cell tumours
It has been proposed that dysgerminoma may evolve from gonadoblastoma [37,38]. We analysed the stages of tumour development, from pure gonadoblastoma, through dysgerminoma with gonadoblastoma, to pure dysgerminoma, using the TSPY gene, a proposed GBY gene, as a marker. Besides gonadoblastoma, TSPY is expressed in male pre-meiotic germ cells and testicular GCTs that retain germ cell phenotype [18,39,40]. We found abundant TSPY in 5/7 cases of gonadoblastoma or dysgerminoma with gonadoblastoma, but not in pure dysgerminoma (n = 11). Consistent results were detected with TSPY protein and FISH Y-centromere analyses, suggesting that TSPY is included in the sub-region of the Ychromosome involved. In one case of combined dysgerminoma/gonadoblastoma we detected a KIT mutation. Taken together, these results indicate that development of dysgerminoma with gonadoblastoma versus pure dysgerminoma may follow different molecular routes, with a role for one or more Y-chromosome genes in the development of gonadoblastoma in phenotypic females. Sexchromosome alteration leads to gonadal dysgenesis and partial masculinisation in at least some cells of the developing female ovary. This may lead to a delay of the normal differentiation of germ cells in foetal ovaries and a failure of the onset of meiotic prophase, thus increasing the chance of mutational events (e.g. in KIT) and neoplastic transformation in the mitotically dividing germ cells, perhaps stimulated by pleiotrophic action of TSPY.
As the OGCTs may develop in dysgenetic ovaries with some testicular differentiation, we examined the expression of AMH, a glycoprotein involved in involution of Müllerian ducts in the foetus leading to male differentiation [19], but did not detect expression in any OGCTs. We furthermore examined MAGE-A4 and NY-ESO-1 [41], two representatives of the germ cell-specific "cancer/testis" gene family on the X-chromosome, which appear in gonocytes from around gestational week 17-18 at differentiation into pre-spermatogonia, and are expressed in adult spermatogonia and primary spermatocytes, as well as in some testicular tumours and a broad range of somatic cancers [42,43]. In accordance with previous studies [43,44] we detected expression of MAGE-A4 and NY-ESO-1 in foetal oogonia/oocytes of 13-40 weeks of gestational age. MAGE-A4 and NY-ESO-1 have been reported in ovarian epithelial and serous carcinomas [45], but this is the first study regarding OGCTs. We found a heterogeneous expression of MAGE-A4 but not NY-ESO-1 in dysgerminomas, analogous to the pattern in testicular seminomas.
In this study, we included also some small cell carcinomas of the hypercalcaemic type that occasionally develop in the ovaries of young patients. None of the applied markers were positive, and therefore it is unlikely that small cell carcinomas have a germ cell progenitor, as has been suggested [46,47]. With regard to the immature teratomas, they were also negative for all analysed markers, which may support the observation that this tumour, which is usually diploid and does not have the isochromosome (12p) imbalance, probably develops in a different manner than other malignant OGCTs [3,4]. No KIT mutations or presence of Y-chromosome material was detected in any of the small cell carcinomas or immature teratomas.
We detected many similarities, but also some notable differences in the protein expression profiles of ovarian and testicular GCTs, but nevertheless propose a comparable developmental model, as depicted in Fig. 3. Genetic alterations and biological mechanisms may be identical, with abnormal persistence of undifferentiated PGCs in both sexes, with the large difference in incidence reflecting the much lower number of susceptible cells in females at the time of puberty [48]. The proposed stepwise progression from gonadoblastoma to dysgerminoma in dysgenetic ovaries [38], and the similarities of gonadoblastoma and testicular carcinoma in situ [49], could indicate that gonadoblastoma -analogous to carcinoma in situ in the testes -is a precursor stage, which probably only occurs in phenotypic females with Y-chromosome material. As the stem cell-related markers were abundantly expressed, our hypothesis is that in 46, XX females, the OGCTs probably are derived from PGCs or oogonia. We speculate that despite the lack of the Y-chromosome material, some degree of gonadal dysgenesis leading to a delay in the differentiation of PGC/oogonia and initiation of gonadoblastoma-like lesions, may be due to changes in hormonal environment, perhaps caused by endocrine disrupters or by lifestyle changes, e.g. obesity. However, the low incidence of OGCTs complicates establishment of epidemiological associations.

Conclusion
We provide in this study new evidence for two partially overlapping mechanisms that may contribute to the pathogenesis of ovarian germ cell neoplasms. One mechanism is related to the presence of Y-chromosome material and ensuing gonadal dysgenesis. The other mechanism is primarily genetic and involves spontaneous mutation(s) in the KIT gene, leading to increased survival and proliferation of undifferentiated oogonia. We expect this mechanism not to be limited to KIT, but demonstrate that mutations in KIT are responsible for a sizeable number (5/17) of cases containing dysgerminoma. The neoplastic transformation must occur before oocytes enter meiosis, most likely after the migration to the gonadal ridges, as the mutations were detected in unilateral tumours.

Tissue samples
The samples originated from patients diagnosed between 1983-2001, who were admitted to 18 Norwegian hospitals for surgery and subsequently referred to the Norwegian Radium Hospital for adjuvant treatment. The tissue microarray (TMA) consisted of 60 cores from tumours of 50 OGCTs patients as described in Table 3. The TMA also contained tissue cores of ovarian small cell carcinoma (n = 8) and various non-ovarian normal and neoplastic tissues, see Table 1. For studies of the ontogeny of expression of selected proteins, we used normal foetal ovarian specimens, which were obtained after induced or spontaneous abortions and stillbirths, mainly due to placental or maternal problems (n = 12, from Rigshospitalet, Copenhagen University Hospital). Developmental age was calculated according to menstrual bleeding and foetal foot size [25]. A representative series of paraffin-embedded testicular GCT samples from the tissue bank at Rigshospitalet was used for positive controls. The Norwegian ethical guidelines were followed and the biobank is registered at the Norwegian Institute of Public Health. The Regional Committee for Medical Research Ethics in Denmark approved the use of Danish tissue samples.

Immunohistochemistry
The  To define the pattern of expression of the analysed proteins and to control for specificity of all antibodies, we analysed expression patterns in various normal tissues, testicular GCTs, and non-gonadal malignancies and results were in accordance with previously reported expression patterns (Table 1).

Fluorescence in situ hybridisation (FISH)
FISH was performed to analyse the presence of X and Ychromosome material. Preparation of the sections was similar to that described above for immunohistochemistry. Sections were deparaffinised, rehydrated and pretreated with proteinase K solution/PBS (20 μg/ml, Invitrogen, Carlsbad, CA). Dehydration was applied before denaturation of slides together with the probes specific for centromeres of the X-chromosome (Texas red labelled) and the Y-chromosome (FITC labelled, green fluorescence) (probes from DakoCytomation). Sections were hybridised for 2 h, washed, dehydrated and mounted in Antifade with DAPI II (125 ng/ml, Vysis Inc., IL), followed by examination using an epifluorescence microscope (Leica, Wetzlar, Germany) with 100× magnification and the CytoVysion software (Applied Imaging, Newcastle, UK).

Declaration of competing interests
The author(s) declare that they have no competing interests.

Authors' contributions
C.E.H.H. performed and co-analysed most immunohistochemical stainings, supervised FISH analysis, performed part of the data analysis and drafted the manuscript. S.M.K. coordinated and designed the tissue microarray, performed part of the mutation analysis and data analysis, and contributed to writing of the manuscript. V.M.A. performed pathology evaluation of the tumours and guided tissue selection for microarray preparation, co-analysed immunohistochemical stainings, and contributed to writing of the manuscript. J.K. was responsible for inclusion of the clinical data and performed the clinical follow-up of many of the patients. E.R.M. supervised immunohistochemical stainings contributed to the planning, data analysis, and writing of the manuscript. R.A.L. initiated the project, supervised KIT mutation analysis, contributed to the planning, data analysis, and writing of the manuscript. All authors read and approved the final manuscript.  [1], and the most optimal tissue blocks were chosen. The TMA was constructed by transferring core biopsies of 0.6 μm in diameter from each donor block into an acceptor block (Beecher Instruments, Sun Prairie, USA). Abbreviations: GCT = germ cell tumour; No. = number; DG = dysgerminoma; GB = gonadoblastoma; YST = yolk sac tumour; IT = immature teratoma; EC = embryonal carcinoma; CC = choriocarcinoma; SCC = small cell carcinoma.