Molecular Evidence for Epithelial Origin of Mixed Ovarian Epithelial-Germ Cell Neoplasms: Report of 2 Cases and Review of Literature.

Patient A is 58-year-old woman who presented with abdominal bloating and pain; computed tomography (CT) scans of brain, chest, abdomen, and pelvis were significant for a 13 cm pelvic mass without evidence of metastatic disease. Pre-operative CA125 was elevated at 77 U/L (0–35 U/L); pre-operative beta-HCG was not obtained. Intraoperatively, the mass arose from the left ovary and involved the rectosigmoid colon and pelvic sidewalls, requiring aggressive surgical debulking including hysterectomy, bilateral salpingo-oophorectomy, omentectomy, pelvic and para-aortic lymph node dissection, rectosigmoid resection, and re-anastomosis. Surgical pathology showed FIGO stage IIIA1(ii) mixed non-gestational ChCA (95%) and mucinous adenocarcinoma (5%; MAC); nodal metastases were noted in the left pelvic, left iliac vein, and para-aortic beds, with both tumor morphologies seen in sites of nodal disease. Additional sites of extraovarian disease included the left pelvic peritoneum, rectal serosa, and pericolonic soft tissue. A postoperative beta-HCG was obtained in light of the diagnosis and measured 15,135 IU/L (0–5 IU/L). Patient A’s 15.5 cm left ovarian mass was predominantly solid with areas of hemorrhage, necrosis, and a small multiloculated cystic area containing thin serous fluid. Microscopic sections demonstrated tumor cells with two morphologies. The majority of the lesion contained a biphasic population composed of mononuclear cells with vesicular chromatin, prominent nucleoli, and brisk mitotic activity admixed with multinucleated cells with eosinophilic cytoplasm and hyperchromatic, smudgy chromatin, with the minor component showing overt gland formation with tumor cells containing intracellular mucin droplets, vesicular nuclei, and occasional bizarre nuclear atypia. The MAC and ChCA portions demonstrated disparate reactivity with immunohistochemical (IHC) stains. The MAC stained positively for CK20 and CDX2, displayed a wild-type p53 pattern, and was negative for SALL4 and beta-HCG. Conversely, the ChCA showed positive staining with SALL4, beta-HCG, and CK20 (weak), with aberrant overexpression of p53 (see Table 2 and Fig 1A – F ). In both components of the tumor, identical missense or small insertion and/or deletion (indel) mutations in CDKN2A, PIK3CA, and TP53 were identified ( Table 3 , Fig 3 ). Of note, the PIK3CA p.H1047Y is a well-documented recurrent mutation site (hotspot). Prior to initiating adjuvant therapy, the patient developed pulmonary and hepatic metastases and beta-HCG was 37,000 IU/L at that time. She completed 4 cycles of bleomycin, etoposide and cisplatin (BEP) with complete response by imaging and both CA125 12 U/L and beta-HCG 7 IU/L. One of the hepatic lesions was biopsied after completion of BEP which showed non-viable tumor consistent with treatment effect. Eight months after competing BEP, she had recurrence of biopsy-proven choriocarcinoma in the para-aortic lymph nodes. Expression of programed cell death ligand 1 (PD-L1) has been demonstrated in gestational trophoblastic disease, and allows the use of pembrolizumab to target these tumors. 43 , 44 Patient A subsequently completed 4 cycles of pembrolizumab in the setting of her first recurrence, though she experienced further progression of the choriocarcinoma on this regimen, with beta-HCG rising to 326 IU/L; progression of non-gestational choriocarcinoma has been reported in patients undergoing pembrolizumab treatment. 45 Her second recurrence was treated with etoposide, methotrexate, actinomycin D, cyclophosphamide, and vincristine chemotherapy, which has been shown to provide sustained remissions of tumors containing a choriocarcinoma component, and has been useful in treating gestational trophoblastic disease in the salvage setting. 41 , 42 With her third recurrence, an ovarian epithelial regimen including carboplatin and paclitaxel is being planned to broadly treat this mixed tumor. At 26 months, she remains alive with disease. See Table 1 for a clinical summary. Patient B is a 43-year-old woman who initially presented with acute onset abdominal pain. CT demonstrated a 10 cm right adnexal mass and peritoneal thickening, with CA125 elevated at 89 U/L. Other tumor markers were also elevated, including CA19–9 132 U/mL (0–35 U/mL), LDH 680 U/L (313–618 U/L), and AFP 2152 ng/mL (0–9 ng/mL). She underwent open hysterectomy, bilateral salpingo-oophorectomy, omentectomy, bilateral pelvic and para-aortic lymph node dissection, and appendectomy. Surgical pathology showed FIGO stage IIIB mixed YST (95%) and endometrioid carcinoma (5%; EC). The only site of extraovarian disease was involvement of the right peritoneum above the pelvic brim by the YST component. Patient B’s 10 cm left ovarian mass revealed tan cystic and solid components with abundant necrosis on cut surface. Microscopic sections showed a YST component with glandular, micro- and macrocystic, tubulopapillary, and solid growth patterns. Tumor cells exhibited increased nuclear to cytoplasmic ratios, primitive nuclei, and brisk mitoses. The minor EC component demonstrated well-developed glands with intervening squamous metaplasia and bland nuclear features without significant mitotic activity. The EC and YST areas similarly showed inverse staining patterns, with the EC expressing ER and lacking SALL4 or AFP, while the YST showed the converse (see Table 2 and Fig 2A – F ). Again, in both components of the tumor, identical missense or indel mutations were identified in CTNNB1, PIK3R1, and PTEN (see Table 3 , Fig. 4 ). The CTNNB1 p.S37F variant is noted to be a mutational hotspot. She completed 4 cycles of BEP with complete response by imaging and tumor markers. She remains without evidence of disease at 22 months. See Table 1 for a clinical summary. All reported variants demonstrated a median variant allele frequency of 0.40 (range 0.09 to 0.62) and a median read depth of 462 (range 121 to 1566). There were no unshared mutations identified in the tumor components from the same patient. A summary of clinicopathologic features are listed in Table 4 . The median age of presentation was 58 (range 23–82) often presenting with abdominal pain, bloating, and discomfort. Most patients (94%) presented with unilateral disease with a slight right-sided predominance (53% vs. 41%) and a median tumor size of 13 cm. As expected, YST was more commonly reported than ChCA (80% vs. 20%), and only two cases reported more than one germ cell component, both of which were YST with immature teratoma. 14 , 15 The most common epithelial components reported included EC (43%), high-grade serous carcinoma (22%; HGSC), clear cell carcinoma (13%; CCC), MAC (7%), and neuroendocrine tumor (7%). Up to 20% of patients were reported to have multiple epithelial components. Of note, 17 cases (31%) were reported to have endometriosis, an endometrioid cyst, or atypical endometriosis. 5 , 6 , 11 , 13 , 16 – 26 EC and CCC were associated in 70% and 11% of cases with endometriosis, respectively. Tumor markers corresponding to the GCT components were elevated in 87% of cases. Forty-four reports provided a FIGO tumor stage with the following frequencies: Stage I (16 patients, 36%), Stage II (3 patients, 7%), Stage III (18 patients, 41%), and Stage IV (7 patients, 16%). 4 – 9 , 13 – 15 , 17 – 21 , 23 , 25 – 31 , 33 , 34 , 36 When considering only tumors with YST component, FIGO Stages were as follows: Stage I (16, 42%), Stage II (3, 8%), Stage III (17, 45%), and Stage IV (2, 5%). 4 – 7 , 9 , 13 – 15 , 17 – 20 , 23 , 25 – 31 , 36 FIGO Stages for tumors with ChCA were as follows: Stage I/II (0), Stage III (1, 17%) and Stage IV (5, 83%). 8 , 21 , 33 – 34 Omentum and lung were the most common metastatic sites for YST and ChCA tumors, respectively, and while the composition of metastatic foci was not consistently reported, presence of the germ cell component was described both with and without the epithelial component. 7 , 16 , 20 , 23 , 25 , 35 A post-surgical follow-up summary is listed in Table 5 . Forty-four reports provided post-surgical follow-up. 4 – 12 , 14 – 19 , 20 , 22 – 26 , 28 – 37 Sixteen patients (36%) were reported to be disease free without recurrence with follow-up periods ranging from 1 to 48 months (median, 12 months). 4 , 5 , 9 , 14 , 15 , 17 , 20 , 22 , 23 , 26 , 29 , 31 , 33 , 37 Twenty patients (45%) were dead of disease with a survival period ranging from 7 days to 29 months after surgery (median, 12 months). 6 – 10 , 12 , 14 , 16 , 18 , 19 , 23 , 25 , 29 , 34 Thirteen patients (30%) were reported to have recurrences which occurred from 2 to 24 months post-operatively (median, 8 months). 10 , 11 , 15 , 16 , 23 , 24 , 35 ,- 37 Of the patients who recurred, 12 (92%) recurred within twelve months from the time of surgery. 11 , 15 , 16 , 23 , 24 , 35 ,- 37 Additional long-term follow up was provided for 7 patients. Five patients succumbed to disease 5 to 12 months (median, 6 months) after the initial recurrence, and two patients were reported alive with disease, both at 23 months, with initial recurrences at 10 and 8 months. 9 , 10 , 15 , 16 , 23 , 28 As expected, patients with earlier stage disease were less likely to die of disease, but if outcomes were compared stage for stage to their epithelial counterparts, the overall prognosis of OE-GCT was worse. For stage I/II tumors, six of the fifteen (40%) patients were dead of disease within 21 months of surgery (median, 15 months). 6 , 14 , 19 , 23 , 25 , 28 Again, epithelial histotypes that are expected to be less aggressive did not impact survival, even in early stage tumors. For example, of the ten patients with early stage tumors containing EC, four were dead of disease within 14 months of surgery. 6 , 15 , 19 , 23 Not surprisingly, patients with stage III/IV tumors had worse outcomes with a greater number of patients dead of disease (68%) and a survival time ranging from 7 days to 20 months (median, 9 months), with one patient succumbing to disease 7 days after surgery due to hemorrhagic brain metastases. 8 It is clear that predictions of tumor behavior and survivorship data from usual EOTs does not translate to these mixed tumors. 1 , 15 , 28 Additionally, when stage was compared between YST and ChCA tumors, the incidence of early (stage I/II) and late stage (stage III/IV) disease was distributed equally in those with YST; however, all patients with ChCA tumors presented with late stage disease. There was no relationship between the degree of tumor marker elevation and patient outcome. Of the five patients without tumor marker elevations, four were disease free at the time of follow-up, with follow-up periods ranging from 12 to 30 months (median, 21 months) and one had recurrent disease at 12 months. 20 , 23 , 26 , 32 , 34 Of note, this was the only clinicopathologic characteristic correlating with disease free status.

Our database was searched for primary ovarian GCTs with a malignant epithelial component in patients over 35 years of age, from 2006 to 2021. Patients under the age of 35 years were excluded to avoid true ovarian GCTs. Two cases were identified and reviewed by two gynecologic pathologists (AB, MDP) to confirm the diagnosis. Slides were annotated for microdissection to include discrete areas of germ cell and epithelial components. Microdissection was carried out on 10 micron cut slides counterstained by hematoxylin under a microdissecting microscope by trained individuals. The gene mutation NGS assay was clinically validated and performed in the Colorado Molecular Correlates Laboratory in the Department of Pathology at the University of Colorado – Anschutz Medical Campus. Total nucleic acid (TNA) was extracted from formalin-fixed paraffin-embedded (FFPE) processed material via the Agencourt FormaPure Kit (Beckman Coulter, Brea, CA). TNA was processed by a customized version of the Archer VariantPlex Solid Tumor library preparation kit (ArcherDx, Boulder, CO). The resulting libraries were sequenced on either Illumina MiSeq or Illumina NextSeq instruments (Illumina, San Diego, CA). Raw sequence data was analyzed using the ArcherDx Analysis software package (version 5.1.2, ArcherDx, Boulder, CO). Bioinformatically identified mutations were manually inspected by highly trained personnel. A PubMed search was performed for articles containing combinations of the following key words: “yolk sac tumor”, “choriocarcinoma”, “endometrioid adenocarcinoma”, “mucinous adenocarcinoma, “mixed ovarian epithelial-germ cell tumor”, “somatic derivation”, and “postmenopausal”. Case reports, case series, review articles, and abstracts were reviewed for relevancy and included in the literature review if histologic evidence of a mixed OE-GCT was described; this included case series with reported molecular findings. Cases of epithelial carcinomas arising from teratomas were excluded. Additionally, references cited in these articles were reviewed and included if appropriate. Approximately 300 articles were retrieved; however, only 36 articles contained cases that met the above criteria.

As of 2018, ovarian cancer was the eighth most common cancer diagnosis and cancer-related death in women worldwide. The majority of malignant tumors are epithelial and occur in the peri- or post-menopausal age range. 1 In Western countries, germ cell tumors (GCT) account for 2–3% of ovarian malignancies and often occur within the first two decades of life. The most common GCT subtype is dysgerminoma, followed by yolk sac tumor (YST), while the rarest is choriocarcinoma (ChCA). Similar to GCTs in the testis, malignant ovarian germ cell tumors (mOGCT) can occur in pure or mixed form; however, the co-occurrence of a malignant epithelial component is rare. 2 , 3 In the cases reported, it is recognized that these tumors occur in the 5 th or 6 th decades of life, have an association with elevated tumors markers (i.e. beta-HCG, AFP), and are less chemosensitive compared to GCTs, with a poor prognosis. 2 – 11 It is speculated that these tumors are of somatic origin secondary to the process of neometaplasia or aberrant differentiation; however, few case reports and small case series have been published with supporting molecular evidence. 10 , 12 , 13 Herein, we present two patients, ages 58 and 43, both with mixed ovarian epithelial-germ cell tumors (OE-GCTs) and molecular evidence of a common somatic derivation, based on targeted Next Generation Sequencing (NGS).

Both patients presented with mixed epithelial and germ cell neoplasms of the ovary, the germ cell component of which was overwhelming. Due to the clinical, prognostic, and possible therapeutic implications, identifying a germ cell component in ovarian tumors is critical with potential and reported pitfalls including misinterpreting the germ cell component for clear cell carcinoma, carcinosarcoma, and other carcinomas with syncytiotrophoblastic or osteoclastic giant cells. 4 , 9 While adenocarcinomas of the ovary can contain scattered trophoblastic cells, this does not represent a choriocarcinomatous component; a biphasic population of malignant cytotrophoblast and syncytiotrophoblast is required to diagnose ChCA, as was seen in Patient A. 34 Furthermore, an epithelial endometrioid component of a YST could be misidentified as an endometrioid-like yolk sac tumor (ELYST). ELYST is a separate entity and is a variant of YST resembling endometrial glands, usually occurring in younger women. 5 , 23 Although the distinction can be challenging, careful review of morphology may be helpful as ELYST can be more primitive in appearance and lack the squamous metaplasia that is frequently seen in ECs. As was seen in Patient B, the use of IHC is useful to delineate a true endometrioid tumor from an ELYST which have the same immunoprofile as other YSTs. 4 , 5 , 22 , 23 , 32 All cases identified in the literature, as well as the two presented here, reported a disparate staining pattern. This should be taken into consideration prior to classifying a tumor as having a germ cell tumor component. Overall, the correct diagnoses can be made by first remembering to consider germ cell tumors in older women as well as correlating morphology with IHC and tumor markers, if available. While it has been well-documented that all GCTs have chromosome 12 abnormalities, specifically the gain of isochromosome 12p, the mutational landscape of OGCTs has not been studied as widely as epithelial ovarian tumors (EOTs). 1 , 38 , 39 Van Nieuwenhuysen et al. used whole-exome sequencing and deep targeted resequencing to identify non-germline mutations, such as missense mutations or indels, in 24 OGCTs. 38 They ascertained that despite the lower mutational rate of mOGCTs, recurrent mutations could be identified, specifically in KIT and KRAS. While KIT mutations were exclusively seen in dysgerminomas, the KRAS mutations were associated with an embryonal carcinoma and a mixed yolk sac-dysgerminoma. Other genes with variants identified in this study included AFF3 , AKT1 , BIB1B, CASC5, CHEK2, FIP1L1, NF1, PIK3CD, TBL1XR1 , TPR, and UBR5 . 39 Conversely, EOTs often have somatic mutations with a landscape similar to that seen in primary endometrial tumors. 1 Genes with recurrent alterations and/or notable associations with malignant EOTs include ARID1A, BCOR, BRAF, CDKN1B, CHD4, CTNNB1, EIF1AX, FBXW7, KRAS, MBD3, NRAS, PIK3CA, PPP2R1A, PTEN, TERT promoter, and TP53. 1 , 40 While additional studies are needed to further characterize the mutational landscape of OGCTs, there appears to be little overlap between OGCTs and EOTs. Xing et al. and Skala et al. have utilized NGS to study gynecologic GCTs co-existing with epithelial tumors, with findings suggestive of a common clonal and somatic origin. 11 , 13 Xing et al compared the NGS findings of 5 tumors (3 endometrial, 1 cervical, and 1 ovarian) with choriocarcinomatous and epithelial components to 6 pure ovarian ChCAs. The single ovarian tumor demonstrated that all three components (ChCA, EC, and CCC) contained mutations in ARID1A , BRD3 , PIK3CA , and STAG2 genes, and clonal evolution was demonstrated with additional variants in ARID1A, BRD3, CDKN2A, SRSF2 being found only in the CCC and ChCA components. Of the five mixed tumors, 24 genes with mutational variants were identified within the choriocarcinomatous components while only two genes, CDH2 and IDH2, with mutational variants were identified in the six pure gestational ChCAs. There was no overlap in genes identified between the two groups. Although this study largely contains endometrial tumors, it is notable that this study demonstrates a considerably lower mutational burden in GCTs in comparison to gynecologic epithelial tumors, as well as a lack of overlap in somatic mutations. 11 Skala et al. described six OGCTs, of which three were directly associated with an epithelial tumor. In the three associated with an EOT, variants in CTNNB1, FBXW7 , PIK3CA , PTEN , and TP53 were identified, including the same CTNNB1 and PIK3CA hotspot variants identified in our patients. The other three OGCTs without an EOT were associated with or suspicious for endometriosis involvement and variants were identified in FBXW7 , FGFR2 , RB1, PTEN , PIK3CA , and TP53 . Both groups concluded that these mixed tumors were likely of somatic, epithelial origin. 11 , 13

Therapy guidelines are not in place for mixed OE-GCTs, and choosing an appropriate treatment regimen can be challenging, as many patients have an aggressive course with poor response to chemotherapy. 2 , 4 – 6 , 8 – 11 Many patients were treated with a combination platinum-based approach targeting both components, but larger studies with longer follow up periods are needed to fully delineate an optimal approach in either a first-line or salvage setting. 10 , 22 , 28 In summary, while the genes with mutational variants identified in our study are not specific to malignant EOTs, all genes have been reported in other OE-GCTs and to our knowledge have not been reported in pure GCTs. 1 , 11 , 40 While the studies to date are small, there are notable commonalities that support the conclusion that the epithelial and germ cell components are clonally related, and are of somatic, epithelial derivation. 11 , 13 , 39 The implications of these findings on therapeutics remains under further investigation.