Ovarian borderline tumors in the 2014 WHO classification: evolving concepts and diagnostic criteria

Approximately 50–55% of BOT belong to this subtype (synonymous “atypical proliferative serous tumor”) [5]. Molecular analyses have demonstrated that serous borderline tumors (SBTs) harbor similar molecular and genetic alterations as low-grade serous carcinomas (LGSC) [14‐17]. In some cases, a continuous tumor progression from cystadenomas and BOT to low-grade carcinomas may exist, and co-existing areas of SBT were observed in 30 out of 50 LGSC in a series by Malpica et al. [18]. KRAS and BRAF mutations are each present in about 30% of SBT, usually in a mutually exclusive fashion [14, 19‐21]. In contrast, p53 mutations are almost exclusively found in high-grade serous ovarian carcinomas [15, 22]. While the fallopian tube has been established as the site of origin for some high-grade serous carcinomas, SBTs were historically presumed to originate in the ovarian cortex or peritoneal surface. Recent reports showing a higher frequency of PAX2-negative secretory cell outgrowths (SCOUT) in the fallopian tubes of women with SBT may warrant further study [23].

On gross pathologic examination, SBTs are unilocular or multilocular cystic tumors with or without epithelial proliferations on the outer tumor surface. About one third of SBTs are bilateral. The histology of BOT is characterized by hierarchically branching papillae and pseudopapillae with paucicellular, edematous, or hyalinized fibrous stroma, lined by architecturally complex epithelial proliferations (Fig. 1a). The epithelial cells are typically columnar, resembling secretory cells of the fallopian tube, admixed with variable numbers of ciliated cells. There is mild to moderate nuclear atypia, hyperchromasia, epithelial multilayering, and cell detachment (“tufting”) into the lumen [6, 24] (Fig. 1b). Most authors, including the WHO 2014 classification, agree that >10% borderline histology within a cystadenoma or cystadenofibroma qualifies as BOT. In contrast, serous cystadenomas with foci qualifying as SBT in <10% of the epithelial volume are designated “cystadenoma/fibroma with focal epithelial proliferation” [1, 25] [5]. By immunohistochemistry, SBTs are characterized by expression of WT1, PAX8, Bcl-2, estrogen and progesterone receptor [26‐28]

.

A recent retrospective study by the authors demonstrated that overdiagnosis of cystadenoma/fibroma as SBT has been a relevant clinical problem in the past, albeit in a low percentage of cases. Out of 81 consecutive cases diagnosed as BOT over a 10-year period at a single tertiary center (1998–2008), the diagnosis of SBT was rejected due to a diagnosis of serous cystadenoma/fibroma in 7 (9%) patients [29]. This cohort of 81 patients was part of a larger multicenter study, which confirmed that an overdiagnosis of borderline tumor had been made in 11.5% (92/803) of patients [30].

Mucinous borderline tumors (synonymous atypical proliferative mucinous tumors) are the second most common type and account for about 35 to 45% of ovarian borderline tumors [5]. These tumors are usually large, unilateral, and cystic with a smooth ovarian surface, composed of multiple cystic spaces with variable diameter. The cysts are lined by columnar mucinous epithelium of gastric or intestinal differentiation, with papillary or pseudopapillary infoldings, and admixed goblet cells and neuroendocrine cells [57] (Fig. 2a). The nuclei are basally located, isomorphic, and with evenly distributed chromatin [58, 59] (Fig. 2b). Mucinous cystadenomas are characterized by a similar mucinous epithelium but lack papillary infoldings. At least 10% of the epithelial volume must demonstrate increased proliferation with papillary infoldings or pseudostratification and mild to moderate nuclear atypia to qualify as mucinous borderline tumor (MBT)). Immunohistochemically, MBTs are characterized by their non-Mullerian differentiation with absence of WT1, estrogen and progesterone receptor expression [60, 61]. Most tumors demonstrate diffuse expression of cytokeratin 7 with patchy co-expression of cytokeratin 20 and variable (usually weak) expression of CDX2 in approximately 40% of cases [57, 62, 63]. While limited previous studies had only assessed PAX8 expression in mucinous carcinomas (10/25; 40% expression) [64], a recent study confirmed expression of PAX8 in MBT (14/23; 61%) as well as ovarian mucinous carcinomas (11/24; 46%). HER2 overexpression or amplification has been reported in up to 20% of MBT [65] and may sometimes be useful in distinguishing primary from metastatic ovarian involvement. The proliferative rate is usually low, and Ki67 demonstrates predominant expression in cells at the base of the papillary structures with decreasing expression toward their tips.

No unequivocal cases of peritoneal implants associated with MBT have been reported in the literature, and their occurrence should prompt exclusion of secondary ovarian involvement (see below discussion on metastatic disease). The vast majority of MBTs have an excellent prognosis with overall survival approaching 95–100% [5]

. A recent single-center study including 254 patients with stage I BOT found a higher incidence of invasive recurrences in MBT compared to SBT following fertility-conserving surgery. However, possible insufficient sampling is a concern in large MBT, and this finding needs to be confirmed in further studies [66]. Histogenetically, development of MBT from mature teratoma or via metaplastic Brenner tumors has been discussed, and ovarian mucinous and Brenner tumors may share differentiation from a common stem cell [67, 68, 69, 70]. The most common molecular genetic aberrations are KRAS mutations previously observed in approximately 60% of MBT, with detection frequencies of up to 92% by newer targeted deep sequencing approaches [71‐73]. Ovarian mucinous tumors are markedly heterogeneous, with frequent co-occurrence of adenomatous, borderline, and carcinomatous components, suggesting a stepwise progression in at least part of the cases. Therefore, careful gross examination and sampling is mandatory and at least one section per centimeter largest tumor diameter should be examined, increasing to two blocks per centimeter diameter in mucinous tumors >10 cm [5]

.

Transitional cell/Brenner tumors of the ovary are generally rare, and only less than 3–5% are of borderline or invasive type. To date, approximately 30 borderline Brenner tumors have been reported in the literature [5, 97].

Borderline Brenner tumors are thought to arise from benign Brenner tumors, as both components often occur together. Commonly found Walthard cell nests have been suggested as possible histogenetic origin based on their shared immunohistochemical expression profile, but are only infrequently associated with Brenner tumors [68, 98]. Limited data on their molecular characteristics suggested that loss of CDK2A (gene encoding p16) and somatic mutations in KRAS and PIK3CA may be involved in the progression from benign to borderline Brenner tumors [99, 100]. Data on their biologic behavior is limited due to the small number of reported cases. Despite a generally favorable prognosis, rare recurrences and deaths have been reported, including one uterine recurrence of a borderline Brenner tumor harboring an exon 9 PIK3CA mutation [97].

Compared to benign Brenner tumors which are commonly small (<2 cm) and predominantly solid/fibromatous (Fig. 4a), borderline Brenner tumors are usually larger than 10 cm (mean 18 cm) with a predominating epithelial proliferation (Fig. 4e and f). Cystic areas demonstrate papillary or polypoid infoldings, covered by a thick layer of transitional-type cells, resembling non-invasive papillary urothelial carcinomas of the urinary tract. Mitotic figures may be numerous.

The immunohistochemical expression profile of Brenner tumors, including benign, borderline, and malignant, overlaps with urothelial differentiation, with expression of GATA3 (Fig. 4c), cytokeratin 7, and p63; variable expression of uroplakin and thrombomodulin; and absence of WT1 and estrogen and progesterone receptors. Calretinin may highlight luteinized stromal cells. (Fig. 4d) In contrast to urothelial differentiation, cytokeratin 20 is usually absent [98, 101].

An important differential diagnosis of borderline Brenner tumor is (benign) metaplastic Brenner tumors, where the solid transitional cell nests are replaced by mucinous differentiation (often expressing cytokeratin 20) with a central cystic cavity (Fig. 4b) and sometimes associated with complex glandular proliferations [102]. As mentioned above, it is possible that such biphasic mucinous-transitional cell neoplasms may represent divergent differentiation from a common stem cell. Malignant Brenner tumors of the ovary are by definition associated with a benign or borderline Brenner tumor and distinguished by an invasive component usually resembling high-grade invasive urothelial carcinoma. In the differential diagnosis with metastatic urothelial carcinoma of the urinary tract, the detection of an associated benign Brenner tumor component, as well as expression of CA125 and absence of cytokeratin 20, may be helpful to confirm a borderline Brenner tumor of primary ovarian origin.

Endometriosis affects 6–10% of reproductive-age women and has been associated with specific molecular and developmental abnormalities that allow endometrial cells to grow outside of the uterus following menstrual efflux, epithelial metaplasia, or differentiation of stem cells [103, 104]. Endometriosis can be separated into superficial peritoneal and deep infiltrating subgroups. Ovarian endometriosis and in particular ovarian endometriotic cysts represent examples of the second group and have been shown to be clonal lesions by many but not all investigators [105]. Endometriotic cysts have also been associated with increased levels of oxidative stress and reactive oxygen detoxification pathway intermediates, specific cytogenetic abnormalities, loss of PTEN and ARID1A expression, microsatellite instability, let-7 microRNA-dysregulated overexpression of KRAS, and significant epithelial atypia (seen in 8% of cases) [105‐109]. The last feature, known as atypical endometriosis, is frequently seen adjacent to endometriosis-associated carcinomas and BOT as detailed below [110].

Seromucinous borderline tumor (SMBT), also known as atypical proliferative seromucinous tumor, formerly endocervical-type mucinous BOT, or mullerian mucinous BOT, accounts for approximately 5–7% of all BOT. Although long recognized under a variety of names, these tumors are now formally established as a separate category in the revised 2014 WHO Classification of Tumours of the Female Genital Organs. Of note, the term mullerian mucinous BOT or mixed mullerian BOT is more recently being favored by some pathologists as it reflects the mullerian differentiation of these tumors [111].

Unlike the other endometriosis-associated BOT (discussed below), SMBT is the most common type of tumor within its category far exceeding benign seromucinous tumors (cystadenoma and adenofibroma) and seromucinous carcinomas [112]. SMBTs usually arise in young women (34–44 years old) and present as unilocular or paucilocular cysts averaging 8–10 cm in diameter often with intracystic papillae [113, 114]. Bilateral involvement is seen in 40%, and up to 20% have peritoneal implants or lymph node involvement. Associated endometriosis is seen in 30–70% of cases, and SMBT often co-exists with endometriosis-related cancers, most commonly of endometrioid histology, so extensive sectioning is indicated. It has been proposed that SMBTs usually arise within atypical endometriotic cysts that undergo mucinous differentiation. As in other endometriosis-related BOT, loss of ARID1A expression is common and mutations have been documented in 33% of cases [115, 116]. Similar to other mucinous female genital tract tumors, KRAS mutations are frequent, being detected in 69% of cases [117]. Histologically, SMBTs have architectural features similar to SBTs (Fig. 5a). However, the branching papillae in SMBTs are lined by varying proportions of endocervical-type mucinous, tubal-type serous, endometrioid, and indeterminate cells with dense eosinophilic cytoplasm. Hobnail or clear cells and prominent squamous metaplasia may also be seen. The presence of at least two different types of mullerian differentiation is required for the diagnosis [5]. Virtually all cases are associated with a significant amount of acute inflammation, edema, and occasional eosinophils which can be diagnostically useful (Fig. 5b). Unlike most other BOT, the majority of cases do stain positively for vimentin. Estrogen and progesterone receptors are generally positive (Fig. 5c), and WT1 (Fig. 5d), CK20, and CDX2 are negative. Diffuse cytoplasmic expression of MUC5AC is seen in the endocervical-type mucinous component [59, 60, 118]. Microinvasion and intraepithelial carcinoma may be seen and are defined in the same way as in other BOT (see above). Peritoneal implants associated with SMBT have been infrequently reported. Most cases have a good outcome, even in the presence of extraovarian disease. Destructive stromal invasion of 5 mm or more, complex expansile growth, and/or invasive extraovarian disease are the standard criteria defining progression to seromucinous carcinoma [112].

Endometrioid borderline tumor (EBT), also known as atypical proliferative endometrioid tumor, accounts for 2–3% of BOT [119, 120]. Mean age at diagnosis is 57 years. Co-existing endometriosis is seen in 63% of cases, and 39% have synchronous hyperplasia or carcinoma in the endometrium. Therefore, endometrial curettage is recommended in case of fertility preserving therapy. ARID1A mutations are detected in 40% of cases [115]. KRAS mutations are found in 29% of cases with associated endometriosis, but are rare in the remainder [121]. Abnormalities in the Wnt/beta-catenin and PI3K/mTOR pathways are also common. Average tumor diameter is 9 cm with two thirds of cases having a cystic component and the remainder being predominantly solid. An adjacent benign endometrioid adenofibroma is observed in 50% of cases. Bilaterality is seen in 4% of cases. Histologically, tumors can show two patterns, adenofibromatous or villoglandular. Both subtypes may show focal areas of cribiform growth, squamous morules, and intermixed necrotic debris. In the adenofibromatous subgroup, distinction from adenofibroma is made on the basis of a complex hyperplastic epithelial growth pattern with or without mild to moderate nuclear atypia (Fig. 6a and b). Villoglandular EBT may occasionally demonstrate an architecture similar to SBT but differ in their cytologic features (Fig. 6h). The cells are more cylindrical with oval nuclei orientated perpendicular to the basement membrane. Sometimes, this can be ambiguous and there may be overlap with serous differentiation. Immunohistochemical stains (Fig. 6c to f) are not usually necessary to differentiate EBT from SBT. However, WT1 is usually negative in contrast to SBT, while focal p16 staining may be seen in up to 50% of EBT [122]. Intraepithelial carcinoma (severe nuclear atypia) and microinvasion follow the same criteria as in MBT, but have little or no prognostic significance. Peritoneal implants are exceedingly rare [119]. Progression to carcinoma is defined by a confluent expansile growth and/or one or more foci of destructive invasion of ≥5 mm. In view of its rarity, absence of extraovarian disease, and the lack of any well-documented malignant behavior, the diagnosis of EBT currently has few if any implications for clinical management [123]

.

Mimics of EBT include metastases from gastrointestinal, endocervical, or endometrial adenocarcinomas. Positivity of PAX8 and estrogen receptor distinguish EBT from gastrointestinal carcinoma. Diffuse and strong nuclear p16 expression and detection of HPV-DNA may be helpful for the differentiation between EBT and metastasis of HPV-related endocervical adenocarcinoma. MUC4 and HIK1083 can be useful for distinguishing HPV-negative endocervical adenocarcinomas of gastric type. Metastases from well-differentiated endometrial carcinomas demonstrate a similar immunohistochemical expression pattern as EBT.

Clear cell borderline tumor (CCBT), also known as atypical proliferative clear cell tumor, represents less than 1% of all BOT [5, 124, 125]. It usually occurs between 59 and 68 years of age. Tumors average 6 cm in diameter and are generally unilateral. The majority have a solid appearance with small cysts (“swiss cheese” pattern), similar to clear cell adenofibroma but with some softer, fleshier areas. A minority arise within atypical endometriotic cysts. CCBTs are highly likely to be associated with foci of frank clear cell carcinoma, so extensive sectioning, preferably submitting the entire specimen, is recommended. Given its rarity, few specific molecular studies have been conducted, but there is a strong association with endometriosis, often atypical [110]. When seen adjacent to clear cell carcinoma, a shared loss of ARID1A expression suggests a clonal relationship [126]. The frequency of PIK3CA mutations, an early event in clear cell carcinoma, has not been studied in adenofibromatous-type CCBT. Histologically, CCBTs are characterized by round to oval evenly spaced glands embedded in an adenofibromatous-type stroma (Fig. 7a). Compared with clear cell adenofibroma, there is more glandular crowding and proliferation. Glands are lined by flat, cuboidal, or hobnail cells with moderate nuclear atypia (Fig. 7b), small nucleoli, and coarse chromatin clumping. Mitotic rate should be less than four per high power field. The key discriminating feature is the degree of nuclear atypia. Low-grade atypia characterizes adenomas whereas an intermediate nuclear grade defines the borderline category. Clear cell carcinoma is characterized by at least focal high-grade nuclear atypia with prominent nucleoli (Fig. 7d), commonly in a background of intermediate nuclear atypia, and often associated with hyalinized stroma [127]. However, nuclear grading is highly subjective and diagnosis should err on the side of malignancy, particularly in the tubulocystic variant which can often appear deceptively bland. Similarly, a purely cystic growth pattern is very unusual in CCBT and should raise suspicion for the intracystic variant of clear cell carcinoma (Fig. 7c). Immunostains for HNF1 and Napsin A (usually positive) and WT1 and estrogen and progesterone receptors (usually negative) may be helpful to establish clear cell differentiation in some cases. Microinvasion is defined similarly to other BOT (<5 mm). Although it has been stated that CCBT may show intraepithelial carcinoma, this diagnosis should be made with great discretion in view of the frequency of co-existent clear cell carcinoma. All CCBTs followed thus far have acted in a benign fashion. However, there is no data regarding the prognosis of cases with either intraepithelial carcinoma or microinvasion.