Cystic Granulosa Cell Tumors of the Ovary: An Analysis of 80 Cases of an Often Diagnostically Challenging Entity

Eighty cystic GCTs were identified from our consultation files and the archives of our institution between 1960 and 2020, including 8 reported in a prior study.²  All cases were reviewed by at least 2 senior gynecologic pathologists. The clinical and gross features were obtained from the consultation correspondence or the in-house pathology reports. Age and androgenic manifestations, if present, were recorded.

Tumors were included in this study only when the cyst lining was smooth or, at most, slightly granular; tumors with any intracystic solid component were excluded. From 2 to 28 (average, 8) hematoxylin and eosin–stained slides of tumor were available for review. Pathologic features evaluated included size; type of cyst (unilocular or multilocular); type of GCT (AGCT or JGCT); extent of denudation of cyst lining; number of cell layers when present; presence and extent of architectural patterns and cytologic features of granulosa cells lining cysts and within cyst walls; presence/appearance of theca cell component; and presence and extent of luteinization of uninvolved ovarian stroma.

Five AGCTs with available material underwent FOXL2 mutation analysis. A hematoxylin and eosin–stained slide cut from the paraffin block was reviewed by pathologists involved in the study, and tumor tissue was selected for analysis. Corresponding tumor tissue from the block was obtained with coring (1 mm diameter) with a Uni-Core (Sigma-Aldrich, St Louis, Missouri) or slide scraping, depending on the amount of tumor material. Formalin-fixed, paraffin-embedded samples were extracted after an overnight proteinase K digestion step at 70°C. DNA extractions were performed with a Maxwell RSC DNA FFPE kit according to the manufacturer's protocols on a Maxwell RSC device from Promega (Madison, Wisconsin). DNA was quantified using QuantiFluor ONE dsDNA System (Promega). The detection of mutations within exon 1 of the FOXL2 gene was performed using Sanger sequencing. A part of FOXL2 exon 1 (the amplicon size was 182 base pairs) was amplified using the following primers: FOXL2 forward primer 5′-CACAACCTCAGCCTCAACGAGTGC-3′ and FOXL2 reverse primer 5′-TTGCCGGGCTGGAAGTGCG-3′. Briefly, 50 ng of DNA was amplified with the thermal cycling profile of 95° C for 10 minutes, then 35 cycles of 95° C for 30 seconds, 68°C for 45 seconds, and 72° C for 45 seconds, with a final extension at 72° C for 10 minutes. Direct sequencing was performed using the above primers with the Big-Dye DyeDeoxy terminator cycle sequencing kit (Applied Biosystems, Foster City, California). Sequencing reactions were carried out on the ABI Prism 3100 Genetic Analyzer (Applied Biosystems). As a reference for the FOXL2 gene, NM_023067.3 aligned on a sequence of chromosome 3 (hg19) was used.

Twenty-four cystic AGCTs and 56 cystic JGCTs were included in the study. The patients with AGCTs ranged in age from 6 to 83 years (mean, 35 years; median, 29 years) whereas those with JGCTs ranged in age from 3 to 51 years (mean, 22 years; median, 20 years). Among the 43 patients for whom clinical history was available (12 AGCT and 31 JGCT), 44% (3 AGCT and 16 JGCT) had androgenic manifestations. For cases received in consultation, the diagnoses offered by the submitting pathologist included nonneoplastic lesions such as follicle cyst (n = 11) and luteinized follicle cyst of pregnancy or puerperium (n = 2), as well as neoplastic lesions including sex cord–stromal tumor of mixed forms (gynandroblastoma, n = 4; Sertoli-Leydig tumor, n = 1), Brenner tumor (n = 2), immature teratoma (n = 1), and struma ovarii (n = 1).

All tumors except 1 were unilateral. The neoplasms from a patient with bilateral cystic JGCTs were illustrated in a prior study.⁵  The details of gross appearance and size were known for 56 tumors. They ranged from 8.2 to 55 cm (average, 17 cm); 40 of 56 (72%) were unilocular (Figure 1, a), including 8% with small secondary cysts in the wall of a dominant cyst, and 16 (28%) were multilocular. The cyst content was typically described as clear, thin fluid that was occasionally blood tinged. The cyst linings were mostly smooth and tan-yellow and showed occasional slight granularity. When noted (44 tumors), the wall thickness ranged from 0.2 to 1.7 cm but was typically thin (Figure 1, b). On sectioning of the wall, ill-defined small nodules were occasionally observed (Figure 1, b).

The great majority of the cyst linings in the 24 AGCTs (Figures 2, a through d, and 3, a through c) had areas of denudation ranging from minimal to moderate to extensive (>90% denudation in 7; Figure 2, a and c). In areas with preserved granulosa cells (Figure 2, b), the average number of cell layers was 22 (range, 5–45), with basal palisading in 9 (Figure 2, d). Only 1 tumor had a small nodular protrusion of granulosa cells (Figure 2, b). Microfollicles (Call-Exner bodies) were present within the cyst lining in 6 tumors, and 2 showed focal papillae. Invagination of the lining granulosa cells into the cyst wall was noted in 17 tumors (Figure 3, b and c); this was typically focal and consisted of trabeculae (n = 9), cords (n = 8; Figures 2, c, and 3, c), nests with microfollicles (n = 13; Figure 3, b), and nests with macrofollicles (n = 5; Figure 3, a). The granulosa cells were uniformly bland, with oval pale nuclei with occasional small nucleoli and easily identified grooves. The cytoplasm ranged from scant (n = 16) to moderately abundant (n = 8), with occasional vacuolization (n = 4). No appreciable mitotic activity was observed in 15 tumors, but it ranged from 2 to 20 per 10 high-power fields (HPFs) (average, 8/10 HPFs) in the remainder; no atypical forms were noted.

Among the 56 JGCTs (Figures 4, a through e, and 5, a and b), there was great variation in the extent of denudation of the cyst lining epithelium, ranging from minimal to approximately 90% in 8 tumors. All tumors had a predominantly flat lining, but 20 showed focal papillae (Figure 4, a) and 5 showed small polypoid nodules (Figure 4, b). The number of cell layers ranged in thickness from 3 to 55 cells (average, 14 cells), with basal palisading seen in 11. The lining granulosa cells invaginated into the cyst wall in 45 tumors (Figures 4, c and d, and 5, b), in the form of follicles (n = 42), trabeculae (n = 10), cords (n = 8), and nests that occasionally aggregated to form solid nodules (Figure 4, c; n = 8). The predominant pattern observed within the cyst wall was variably sized and shaped follicles (Figures 4, c and e, and 5, b), which were extensive in 24 (Figures 4, e, and 5, b) and rare to occasional in 18 tumors (Figure 4, c). The follicles contained basophilic (n = 25; Figure 4, e) or eosinophilic (n = 15) secretions. Myxoid matrix was occasionally noted within the background stroma and was extensive in 6 tumors. The granulosa cells had scant (n = 5; Figure 4, b) to moderate (n = 14, Figure 4, d) to abundant (n = 37; Figure 5, b) eosinophilic cytoplasm. Prominent cytoplasmic vacuolization was seen in 24 tumors. Nuclei were typically round and hyperchromatic without grooves. Cytologic atypia was absent or mild in 42 tumors and moderate to severe in 14, and when present was of the bizarre type (Figure 5, a). Twenty tumors had minimal mitotic activity (up to 3 mitoses/10 HPFs) but it ranged from 3 to 50 per 10 HPFs (average, 12/10 HPFs) in the remainder, with 5 showing occasional atypical forms. Apoptotic bodies were seen in 36 tumors, being easily identified in 14.

A variable theca cell component was observed in all tumors of both types, either underlying the cyst lining (Figure 5, a) or between the various architectural patterns of granulosa cells within the cyst wall (Figure 3, a and c). This component was limited in 36 (16 AGCTs, 20 JGCTs) and extensive in 44 (8 AGCTs, 36 JGCTs; Figure 3, a) tumors. The associated tumor stroma lacked admixed preexisting follicles and ranged from loose to dense connective tissue, showing fibromatous features in 9 and focal hyalinization in 8. There was edema in 3 tumors, and 2 showed prominent vasculature within the cyst walls. Ovarian stroma outside the tumor was evaluated when present in the submitted slides (n = 45). In the patients with known androgenic manifestations, luteinization of the ovarian stroma was variable (none, n = 4; mild, n = 6; moderate, n = 3; extensive, n = 1). In patients in whom androgenic manifestations were either absent or unknown, luteinization of the background ovarian stroma was either absent (n = 13), mild (n = 11), or moderate (n = 7).

Immunohistochemical studies were performed in 19 tumors (Figure 6, a through e), and results reported by the submitting pathologist showed that granulosa cells were positive for inhibin (17 of 17; 4 focal), calretinin (8 of 8), vimentin (4 of 4), FOXL2 (2 of 2), WT1 (2 of 2), CD56 (1 of 1), and AE1/AE3 (8 of 14; all patchy/focal). The tumors were negative for EMA (0 of 6), PAX8 (0 of 4), thyroglobulin (0 of 2), PLAP (0 of 2), glypican-3 (0 of 1), OCT3/4 (0 of 1), keratin 7 (0 of 1), AFP (0 of 1), CD117 (0 of 1), napsin A (0 of 1), synaptophysin (0 of 1), chromogranin (0 of 1), and p53 (wild-type pattern, 2 of 2). A reticulin stain was performed in 13 tumors and showed staining surrounding groups of granulosa cells.

Sequencing was successful in 4 of 5 tumors; 1 AGCT yielded low-quality DNA for testing. FOXL2 c.402C>G mutation was identified in 1 cystic AGCT, and the other 3 AGCTs lacked FOXL2 mutation (1 of 4; 25%; Figure 7, a and b).

GCTs of the ovary of both adult and juvenile types are associated with many diagnostic problems for the pathologist because of their wide morphologic spectrum, which has been discussed at length in the literature.⁷  The purpose of this study was to focus on the cystic nature of a subset of these neoplasms, as they may be confused with nonneoplastic lesions, specifically follicle cysts, as well as other cystic neoplasms. Our interest in the subject was in part stimulated by recent studies on follicle cysts^3,5  and an older study on large solitary luteinized follicle cysts of pregnancy and puerperium.⁴  Denudation of cyst linings in cystic GCTs often causes a diagnostic challenge and has resulted in our having available a large number of tumors sent in consultation. In this report, we highlight the gross and microscopic characteristics of these tumors, including those that enable them to be correctly diagnosed with thorough sampling, rigorous histopathologic examination, and, when needed, immunohistochemical and/or molecular methods.

The most challenging differential diagnosis of a cystic GCT is the benign follicle cyst. Both the standard solitary type of follicle cyst and the 2 forms associated with pregnancy, hyperreactio luteinalis (multiple luteinized follicle cysts) and the large solitary follicle cyst of pregnancy and the puerperium, may enter this differential diagnosis, depending on the clinical context. A recent study on follicle cysts³  showed that a number of those not associated with pregnancy (7 of 30) were greater than 8 cm, a cutoff previously thought to be generally reliable in separating a follicle cyst from cystic GCT. Thus, although our current series supports that cystic GCTs are nearly always greater than 8 cm, size alone is not sufficient in excluding a follicle cyst. Other gross features, including the generally more delicate appearance of follicle cysts, are also unreliable in distinguishing between the 2, although in some cases nodularity within the wall (Figure 1, b) strongly suggests a cystic GCT. Therefore, ample sampling and diligent microscopic examination are essential to rendering the correct diagnosis. In both forms of cystic GCT, the typical architectural patterns of GCTs are usually seen within the cyst wall, including cords, trabeculae, nests, and variably sized follicles. Variably sized follicles are generally absent in follicle cyst walls, with the exception of very minimal satellite cystic follicles reported in 12 of 30 cases in one study.³  Such adjacent cysts may be at different stages of development (eg, atretic follicles in background), a clue that the multicystic nature of the lesion may be hormonally driven rather than neoplastic in nature.³  Thus, for any large cyst (>8 cm) where the diagnosis of follicle cyst is entertained, extensive sampling can help to identify invagination of the cyst lining into the wall in the form of various architectural patterns of GCTs. Other key features aiding in diagnosis of cystic GCT (summarized in the Table) include the presence of nuclear grooves (AGCT), the presence of Call-Exner bodies (AGCT), marked thickening of the granulosa cell layer (>10 cells thick), or the presence of marked atypia, features absent in follicle cysts.³  Exceptions include large solitary luteinized follicle cysts of pregnancy and the puerperium, as they may display marked degenerative atypia.⁸  Clinical history may also assist in distinguishing follicle cysts from GCTs, especially if the cysts are bilateral in a pregnant patient (strongly suggests hyperreactio luteinalis over GCT) or if there are associated androgenic manifestations (strongly suggests cystic GCT).

Many ovarian neoplasms, particularly surface epithelial tumors, may have gross characteristics that overlap with those seen in GCTs, including serous carcinomas (especially those with transitional cell features) and, to a lesser degree, endometrioid tumors. When preparing this manuscript, by happenstance, we came across some old file cases of “transitional cell carcinoma” included in the series authored by Eichhorn and Young.⁹  We were struck by the gross descriptions of markedly cystic tumors often having a conspicuous smooth lining, which prompted us to review the gross pathology section of that study (16% of the tumors were predominantly cystic; Figure 8, a and b). However, on microscopic examination, slitlike spaces, marked pleomorphism, and, if needed, a distinct immunohistochemical profile establish the diagnosis. Endometrioid cystic tumors may be in the differential diagnosis, as they often have some degree of hemorrhage (consistent with a frequent origin from endometriosis) and may be predominantly cystic. On microscopic examination, endometrioid tumors may also display trabeculae, cords, and small acini mimicking Call-Exner bodies, as well as grooved nuclei mimicking the morphologic appearance of AGCTs. However, finding well-formed endometrioid glands with intraluminal mucin and/or squamous differentiation and associated endometriosis almost always makes this diagnosis straightforward. Brenner tumor can also be markedly cystic, usually because of overgrowth of the mucinous cystic epithelium. However, there are rare cystic Brenner tumors in which the lining epithelium is not overtly mucinous, and in such cases a potential pitfall is mistaking the longitudinal grooves of the transitional cells for the nuclear grooves of neoplastic granulosa cells. In the vast majority of cystic Brenner tumors, one will still see some classic Brenner nests within the fibromatous stroma between cysts, and although historically Brenner tumor and insular GCT were confused with each other, currently this is a mistake rarely made.

Other tumors in the sex cord–stromal family are almost never as cystic as exhibited in the cohort reported herein. The greatest degree of cystic change in Sertoli-Leydig cell tumors is seen in those with mucinous heterologous elements or with a retiform component, features that make the differential diagnosis with a GCT moot, although a rare AGCT has been associated with mucinous epithelium.¹⁰  We have seen occasional Sertoli-Leydig cell tumors without heterologous elements/a retiform component with a unilocular or multilocular appearance on imaging and to a significant degree on gross inspection. They may have soft, almost gelatinous, cyst contents, and microscopic examination shows more intraluminal growth than in our cystic GCTs. We have also seen 2 cystic unilocular Sertoli cell tumors; however, there was a tubular pattern and the cells were columnar, had more voluminous pale cytoplasm, and lacked nuclear grooves, in contrast to GCTs (Figure 8, c). Detection of DICER1 hotspot mutations would further aid in the diagnosis of challenging cases, except in well-differentiated Sertoli-Leydig cell tumors.¹¹  Occasionally, sex cord tumors with annular tubules (non–Peutz-Jeghers related) may be strikingly cystic. Two of the 47 tumors in one large study¹²  and rare others published in the literature^13,14  were “predominantly cystic,” including at least 1 that was unilocular.¹⁵  One pure stromal tumor that may rarely be massively cystic is the sclerosing stromal tumor, as was noted in 8 of 100 cases in a large study¹⁶ ; however, on microscopic examination its typical features are seen (Figure 8, d). Fibromas, and much less often thecomas, may show cystic degeneration, which is rarely conspicuous, but the cysts lack any epithelial lining as in cystic GCTs.

The commonest monodermal teratoma, struma ovarii, may be remarkably cystic,⁶  but grossly, a minor component of green to brown-tinged, almost jellylike material is present in many tumors. On microscopic examination, cystic struma, like cystic GCT, may show conspicuous denudation of the lining.¹⁷  Thus, immunohistochemistry can be helpful in this differential diagnosis, as cystic struma ovarii is typically PAX8, TTF-1, and thyroglobulin positive¹⁸  and negative for sex cord markers. Other rare monodermal teratomas, such as so-called endodermal teratomas,¹⁹  might also be predominantly cystic but should be readily distinguished on microscopic examination from GCTs. Primitive germ cell tumors can also be prominently cystic but always have a solid component, except for the polyvesicular vitelline variant of yolk sac tumor, which may have a distinct honeycomb appearance.²⁰  In difficult cases, AFP, glypican-3, and SALL-4 would aid in this differential diagnosis.²¹  For the sake of completeness, cystic change is common in metastatic tumors in the ovary, and the so-called maturation phenomenon that often occurs may produce deceptively benign characteristics subject to misinterpretation. However, this is most notorious with mucinous cystic tumors, which are not likely to be in the differential diagnosis of GCTs. Overall, in our experience, the degree of denudation in GCTs is typically greater than that seen with most other cystic ovarian tumors.

Although diligent gross inspection, good sampling, and thorough microscopic examination will help to distinguish cystic GCTs from lesions in the differential, there will be circumstances in which immunohistochemistry or molecular studies will be of assistance. Indeed, in the setting of particularly massive denudation, immunohistochemical findings may be crucial in establishing the diagnosis (Figure 6, a through e). A number of antibodies can be used, although most are also expressed in other sex cord–stromal tumors as well as follicle cysts, the main differential diagnostic consideration. Inhibin is likely the most widely used antibody and is positive with rare exceptions,^22–24  although calretinin has been shown to be more sensitive though less specific.^25–27  Steroidogenic factor 1 (SF-1) has been more recently reported to be the most sensitive sex cord–stromal marker.²⁵  FOXL2 is positive in both adult and juvenile GCTs, but other sex cord and stromal tumors are often also positive.^28,29  However, it should be emphasized that immunohistochemical results should be interpreted in the context of the morphologic findings. We were strongly reminded of this scenario when recently consulted on a striking cystic JGCT in which the diagnosis was evident on a routine morphologic basis but inhibin and calretinin were at most weakly positive, making the initial pathologist reluctant to render such a diagnosis.

The few studies that have addressed prominent cystic change in GCTs have largely focused on the peculiar propensity of cystic AGCTs to be associated with androgenic manifestations, as was exhibited in 3 of 12 (25%) of our patients. Although prior studies of JGCTs, typically uniformly solid or solid and cystic, have occasionally included sporadic tumors with androgenic manifestations,³⁰  it is a noteworthy aspect of this series that a surprisingly large number of our cystic JGCTs (16 of 31; 52%) were associated with androgenic manifestations. Although the pathophysiology of the androgen production is not entirely understood, the so-called pressure theory of cystic tumors, in which expanding cystic lesions put pressure on surrounding ovarian stroma and thereby induce luteinization of the ovarian parenchyma, has been proposed as a possible explanation. Indeed, in the series of Nakashima et al,²  11 of 17 GCTs with androgenic manifestations demonstrated notable stromal luteinization. We investigated the surrounding stroma in the current cases and identified luteinized cells in 10 of 14 tumors (71%) in patients with androgenic manifestations. In comparison, 58% of tumors from patients with unknown or absent androgenic manifestations demonstrated luteinization. Although we are limited by the available clinical information and sampling of the surrounding ovarian parenchyma, our findings are in keeping with those of Nakashima et al²  and again suggest a possible association between cyst formation, stromal luteinization, and subsequent androgenic manifestations.

It is worth noting that, in the initial seminal contribution of Norris and Taylor¹  on cystic GCTs, all their patients were young, ranging from 12 to 31 years. The average age of patients with AGCTs in our series, 35 years, supports what their study had first indicated, namely that AGCTs that are massively cystic tend to be disproportionately common in younger patients in contrast to typical AGCTs, which generally peak between 45 and 60 years. This finding emphasizes a point made elsewhere,³¹  namely that the designations adult and juvenile GCT are terms of convenience. Tumors with the morphology that falls under the adult designation typically occur in the middle to perimenopausal and postmenopausal years and those with the juvenile designation typically occur in the first 3 decades, but there is considerable overlap, and one series focusing on GCTs in the young³²  had 3 AGCTs among a total of 32 GCTs.

A final remark on FOXL2 mutations is warranted, as a small subset of our AGCTs were examined for missense mutations in this gene. FOXL2 is a transcription factor involved in normal development of granulosa cells, and hotspot missense mutations in the FOXL2 gene are present in most AGCTs and reported only rarely in JGCTs and other sex cord–stromal tumors.^33–35  In this study, we identified FOXL2 hotspot mutation in 1 of 4 AGCTs (Figure 7, a and b). We suspect that the absence of FOXL2 hotspot mutations in the remaining cystic AGCTs was due to the low number of tumor cells available for sequencing, as the lining was mostly denuded and the assay requires at least 15% tumor cellularity to detect mutations. Similarly, Nolan et al³⁶  detected FOXL2 mutations in only 50% of AGCTs with prominent fibrothecomatous background, noting that mutations were likely detected only in tumors with higher numbers of granulosa cells. In their series, tumors with typically greater than 40% granulosa cell cellularity and those with large lobules of granulosa cells showed the FOXL2 mutation. The 3 cystic AGCTs in our series lacking FOXL2 mutations displayed typical patterns of GCTs, and one patient showed androgenic manifestations. Additionally, in a recent case report on a large follicle cyst (18.5 cm) lacking these features, the cyst lacked FOXL2 mutations, consistent with its nonneoplastic nature.³⁷  We conclude that molecular findings, like immunohistochemical findings, are not entirely sensitive and thus should be interpreted in the context of routine morphologic findings. In addition to FOXL2 mutations, other genetic alterations have been described, including TERT promoter mutations in recurrent AGCTs³⁸  and TP53 mutations in AGCTs with high-grade transformation.³⁹  JGCTs show GNAS mutations and duplications,⁴⁰ AKT1 mutations,⁴¹  and rare DICER1 mutations.^42,43 

In conclusion, there appears to be a tendency for cystic AGCT to occur in younger patients than its typical counterpart, and cystic JGCTs are more often androgenic than the prior literature has indicated. Adult or juvenile GCTs may be strikingly cystic, and on gross examination a variety of nonneoplastic and neoplastic lesions may be in the differential diagnosis. Despite extensive denudation, good sampling will almost invariably disclose diagnostic features of GCT, which can be confirmed in particularly difficult cases by immunohistochemical or molecular findings. However, it is important to be aware that FOXL2 mutations may be absent in cystic AGCT because of a limited number of neoplastic granulosa cells, and this result may be misleading.