Role of anticytokeratin antibodies in distinction between primary and secondary ovarian adenocarcinoma

The morphological distinction between a primary ovarian adenocarcinoma, usually of mucinous or endometrioid subtype, and a metastatic adenocarcinoma, especially of colorectal origin, can be extremely difficult. Especially with mucinous neoplasms, metastatic adenocarcinomas can contain areas resembling benign or borderline tumour. Histological features have been described that facilitate this distinction between primary and secondary tumour,^1–3 but in some cases it might be difficult or impossible to decide with certainty. So called specific tumour associated antigens (discussed later) might be of some value but these are often not, in fact, specific to a particular site of origin and their value is limited. Epithelia from different sites differ in their pattern of keratin expression and this pattern is often, but importantly not always, maintained in the corresponding carcinomas. Recent studies have investigated the value of anticytokeratin (anti-CK) antibodies in the distinction between an ovarian primary and a metastatic carcinoma, especially of colorectal origin.^4–8 Primary ovarian adenocarcinomas, including those of mucinous and endometrioid types, are usually CK7 positive (fig 1A) and might be CK20 positive or negative (usually negative) (fig 1B), whereas colorectal adenocarcinomas are generally CK7 negative and CK20 positive (fig 2). However, caution should be exercised because occasional colorectal adenocarcinomas are CK7 positive and some primary ovarian adenocarcinomas, especially of mucinous type, are CK7 negative. It is stressed that when using anti-CK antibodies to assist in the distinction between a primary and secondary adenocarcinoma there should always be close correlation with the morphological features. Be that as it may, the pattern of staining with antibodies to CK7 and CK20 might be of value in the distinction. The value of these antibodies is probably greater in the distinction between a primary ovarian endometrioid carcinoma and a secondary than in the case of a primary ovarian mucinous carcinoma because the latter might be CK7 negative and/or CK20 positive.⁷ The value of anti-CK antibodies might be increased when used as part of a larger panel, which may include antibodies to carcinoembryonic antigen (CEA), CA125, and CA19.9. Table 1 is a suggested panel that might facilitate the distinction between a primary ovarian adenocarcinoma and a metastatic tumour of colorectal origin. In general, anti-CK antibodies are of no value in distinguishing between a primary ovarian adenocarcinoma and a metastasis from the breast, stomach, pancreas, or endometrium because these neoplasms usually also exhibit CK7 positivity. This is important because metastasis from these neoplasms might also closely mimic a primary ovarian adenocarcinoma, especially of mucinous type.

Role of immunohistochemistry in determining origin of pseudomyxoma peritonei

Pseudomyxoma peritonei (PMP) is a condition characterised by mucinous ascites and the presence of multifocal mucinous tumours within the peritoneal cavity. Women with PMP may have mucinous tumours of the ovary and appendix and there has been debate regarding the origin of PMP in such cases, with conflicting opinions based on gross and microscopic pathological features.^9–13 Some have concluded that the ovarian tumour is usually a metastasis from the appendiceal neoplasm, whereas others consider the ovarian and appendiceal lesions to be independent primaries. The ovarian tumour is often interpreted as a primary mucinous cystadenocarcinoma or as a primary mucinous tumour of low malignant potential (borderline tumour). As stated earlier, most primary ovarian mucinous tumours are CK7 positive, whereas most colorectal adenocarcinomas are CK7 negative and CK20 positive. Recent studies using these anti-CK antibodies have found that most appendiceal mucinous tumours are also CK7 negative and CK20 positive, as would be expected, although a small number might be CK7 positive.^14,15 These studies have suggested that in most cases of PMP with concomitant ovarian and appendiceal neoplasms, the PMP is of appendiceal origin (CK7 negative, CK20 positive). Furthermore, in most of these cases, the ovarian neoplasm has an identical immunophenotype, suggesting that it represents a metastasis from the appendix. Similarly, other studies using molecular genetic techniques have demonstrated that most, but not all, synchronous mucinous tumours of the ovary and appendix are of appendiceal origin.^16,17 I have recently seen a case of a presumed primary ovarian mucinous borderline tumour associated with PMP where a primary appendiceal mucinous tumour only became manifest one year later. It is recommended that in all cases of PMP associated with an ovarian mucinous tumour (even a presumed primary ovarian benign or borderline mucinous tumour), CK7 staining should be performed and, if negative, a primary appendiceal tumour should be strongly suspected. In such cases, the appendix can be macroscopically normal or can be obliterated by the fibrosis that often accompanies PMP, and a mucinous tumour might only be identified after surgical removal and histological examination. The finding of CK7 positivity within the PMP in such cases is less useful, and correlation with clinical, pathological, and other immunohistochemical findings is necessary to determine the origin of the PMP.

Value of antibodies against tumour specific antigens in determining ovarian origin of adenocarcinoma

Monoclonal antibodies raised against tumour associated antigens have been investigated as a means of diagnosing cancer, including ovarian carcinoma. In general, in the case of ovarian cancer these have not proved to be sufficiently specific, and their value is limited in a diagnostic sense. Human alveolar macrophage 56 (HAM56), a monoclonal antibody reactive against macrophages and endothelial cells, has been proposed as a useful marker in distinguishing between primary ovarian adenocarcinoma and metastatic adenocarcinoma, especially of colorectal origin. Primary ovarian adenocarcinomas are usually HAM56 positive whereas colorectal adenocarcinomas are usually negative.^18–22 However, although the results of studies are somewhat conflicting, it is clear that positivity for HAM56 may be present in a large proportion of colorectal adenocarcinomas, as well as primary adenocarcinomas arising at other sites. Although positivity is found in a higher proportion of ovarian adenocarcinomas than those arising primarily at other sites, this antibody is not specific for an ovarian primary, and in individual cases is of limited value when confronted with an adenocarcinoma of unknown origin. However, HAM56 may be used as part of a larger panel.

Similarly, other studies have investigated the use of antibodies such as those directed against CEA and CA125 in distinguishing between a primary ovarian adenocarcinoma and one of extraovarian origin.^5,23–28 These antibodies are not specific to a particular site of origin and are of limited diagnostic value. CA125 is a useful biochemical marker of ovarian carcinoma,²⁹ and there is diffuse strong positive immunohistochemical staining in most non-mucinous ovarian adenocarcinomas (fig 3), but often no staining in primary ovarian mucinous tumours. In addition, primary colorectal adenocarcinomas and adenocarcinomas at other sites might be positive, albeit often focally, limiting the value of CA125 as a diagnostic discriminator. CA125 is also positive in most mesotheliomas,³⁰ which can be confused with ovarian type serous adenocarcinomas, either of ovarian or peritoneal origin, when associated with multifocal peritoneal tumour deposits. A recent study using six monoclonal antibodies, including anti-CA125, in an attempt to distinguish between a primary ovarian and colorectal adenocarcinoma concluded that most ovarian carcinomas, including those of mucinous type, can be discriminated with high probability from a colorectal carcinoma using a panel of three antibodies directed against CEA, CK7, and vimentin.⁵

The search for a monoclonal antibody specific to ovarian adenocarcinoma continues and, as yet, no specific marker has been found. Antibodies such as MX35, OV-TL3, and MOV18 have been developed but have not yet become established in a diagnostic setting.^31,32 A recently developed monoclonal antibody SMO47 reacts with most ovarian adenocarcinomas, but positivity can also be found in primary adenocarcinomas arising at other sites, although in these cases usually with a more limited distribution. It is stressed that in problematic cases, careful morphological examination combined with immunostaining using a panel of antibodies that will vary, depending on the differential diagnosis under consideration, should allow for a correct diagnosis in most instances.

Immunohistochemistry in the diagnosis of ovarian sex cord stromal tumours

Sex cord stromal tumours are relatively rare, accounting for approximately 4% of benign ovarian neoplasms and 7% of primary ovarian malignancies.³³ They have a varied histological appearance and can mimic a wide range of other ovarian neoplasms, both benign and malignant. Sertoli-Leydig cell tumours and granulosa cell tumours can closely resemble ovarian endometrioid adenocarcinomas, and the juvenile variant of granulosa cell tumour might be difficult to distinguish from small cell carcinoma of hypercalcaemic type. Other neoplasms that might resemble sex cord stromal tumours include endometrial stromal sarcoma, undifferentiated carcinoma, carcinoid tumour, desmoplastic small round cell tumour, Brenner tumour, and germ cell neoplasms.³⁴ Although antibodies reactive against cytoplasmic intermediate filaments such as CKs and vimentin might be useful in some situations, these lack sufficient diagnostic specificity, and until recently there have been no specific immunohistochemical markers of sex cord stromal differentiation. However, in the past few years a large number of studies have investigated the value of several monoclonal antibodies in the diagnosis of ovarian sex cord stromal tumours and in their distinction from histological mimics. These antibodies are reviewed in the next few sections.

INHIBIN

Inhibin is a dimeric 32 kDa peptide hormone composed of an α subunit and a β subunit.³⁵ The β subunit may be of the A or B type, giving rise to two forms of inhibin (A and B) of apparently identical biological activity. The hormone is produced by ovarian granulosa cells and inhibits the release of follicle stimulating hormone from the pituitary gland, thus acting as a modulator of folliculogenesis.³⁶ Inhibin is also produced by testicular Sertoli cells, and extragonadal expression has been demonstrated in the placenta, adrenal gland, pituitary gland, and liver.^37–42 As well as being produced normally by granulosa cells, inhibin is also produced by granulosa cell tumours, and serum measurements can be used to monitor these neoplasms. Raised serum inhibin values can indicate recurrent or metastatic disease before it becomes clinically apparent.^43,44

In recent years, immunohistochemical staining with anti-inhibin antibodies has been performed on tissue sections. Several studies have found that antibodies against inhibin, especially against the α subunit, are useful immunohistochemical markers of ovarian sex cord stromal tumours, especially granulosa cell tumours.^45–58 Most ovarian sex cord stromal tumours of all histological types stain positively with anti-α inhibin, and this might be of value in their distinction from other neoplasms that might mimic them, especially endometrioid type adenocarcinoma with a sex cord stromal pattern, undifferentiated carcinoma, and small cell carcinoma of hypercalcaemic type. Positive staining for α inhibin in these tumours, and in other primary and secondary ovarian adenocarcinomas, is usually confined to non-neoplastic ovarian stromal theca cells.⁵⁹ This might account for the modest rise in serum inhibin concentrations sometimes seen in these patients, especially those with mucinous adenocarcinomas.⁶⁰ However, caution should be exercised in that primary ovarian adenocarcinomas, as well as extraovarian adenocarcinomas, might occasionally exhibit positivity for α inhibin. A recent study found 20 of 78 adenocarcinomas from various sites to be positive for α inhibin.⁶¹ However, in these cases positive staining is generally weak, contrasting with the usual strong positivity in ovarian sex cord stromal tumours (fig 4). In attempting to distinguish between an ovarian sex cord stromal neoplasm and an epithelial tumour, especially ovarian endometrioid adenocarcinoma, α inhibin is extremely useful when performed as part of a larger panel. This panel should include an antibody to epithelial membrane antigen (EMA) because this is almost invariably negative in ovarian sex cord stromal tumours but positive in most cases of adenocarcinoma.⁶² The detection of α inhibin immunoreactivity in an ovarian tumour that is EMA negative provides both sensitive and specific support for the diagnosis of a sex cord stromal neoplasm.⁵⁰ It should be noted that anti-CK antibodies are of limited value in the distinction between a sex cord stromal and epithelial neoplasm because antibodies against broad spectrum CKs might be positive in sex cord stromal tumours.⁶³ However, staining with anti-CK7 might be of value because a recent study has found this antibody to be positive in ovarian endometrioid adenocarcinomas but negative in sex cord stromal tumours, except for the retiform areas in Sertoli-Leydig cell tumours.⁵⁴ Table 2 is a suggested panel that might assist in the distinction between an ovarian sex cord stromal tumour and an endometrioid adenocarcinoma with a sex cord like pattern. A recent case study found that α inhibin staining was of value in confirming that a malignant spindle cell tumour within the ovary and intestine was a malignant ovarian fibrothecomatous tumour with intestinal metastasis.⁶⁴ Positive staining with α inhibin has also been found in uterine tumours resembling ovarian sex cord tumours and in endometrial stromal tumours with sex cord like elements, supporting the hypothesis that these tumours might show true sex cord differentiation.^65–67

Ovarian granulosa cell tumours are malignant neoplasms with a tendency to late recurrence or metastasis. When recurrent or metastatic disease develops, the history of removal of an ovarian granulosa cell tumour might not always be available. In such instances, immunostaining with anti-α inhibin might provide good evidence of a metastatic ovarian granulosa cell tumour. Positivity is found even in the sarcomatous variant, which might be confused with haemangiopericytoma and other mesenchymal neoplasms. Positive staining for α inhibin is also present in most juvenile granulosa cell tumours, and this might aid in the distinction of this neoplasm from its histological mimics, especially small cell carcinoma of hypercalcaemic type, which is invariably negative.⁵⁰

Staining for α inhibin has also been performed on cytological preparations and has been found to be of value in ovarian fine needle aspirates in the separation of functional cysts (α inhibin positive) from epithelial lined cysts (α inhibin negative).⁶⁸ This separation is of some importance in that functional cysts can be observed, whereas epithelial lined cysts, even when benign, should probably be surgically removed.

Immunostaining with antibodies against β inhibin is of less diagnostic value than α inhibin because positivity might be found in a high proportion of primary ovarian adenocarcinomas.^{61 69} Immunoreactivity might also be present in adenocarcinomas of diverse origins, and β inhibin staining is of no value in determining the origin of an adenocarcinoma of unknown primary site.⁶¹

CD99

Monoclonal antibodies such as 013, which react with the CD99 antigen (MIC2 gene product), are sensitive immunohistochemical markers of Ewing's sarcoma and peripheral primitive neuroectodermal tumour, and are in widespread use.^70,71 These antibodies also react with ovarian granulosa cells,⁷² and positivity has been demonstrated recently in ovarian sex cord stromal tumours of a variety of histological types (fig 5).^{46,50,72–74} However, the proportion of tumours exhibiting positivity is less than for α inhibin, and some epithelial tumours may also be positive. In one study, 58% of small cell carcinomas of hypercalcaemic type were positive.⁵⁰ Therefore, it seems that CD99 is a less sensitive and specific marker than α inhibin in the confirmation of an ovarian sex cord stromal neoplasm, although it might be of value when used as part of a larger panel. It is stressed that, as in Ewing's sarcoma, only membranous staining for CD99 should be regarded as true positivity, and a cytoplasmic reaction should be considered negative for diagnostic purposes. Interestingly, a case of peripheral primitive neuroectodermal tumour of the ovary has been described that, as expected, was CD99 positive.⁷⁵ Because the differential diagnosis in such a case might include small cell carcinoma of hypercalcaemic type and adult or juvenile granulosa cell tumour, positive staining for CD99 is of limited value. It is possible that, in the past, peripheral primitive neuroectodermal tumours that have arisen in the ovary have been misdiagnosed as small cell carcinoma.⁷⁵ Similar to α inhibin, CD99 positivity has also been found in sex cord like elements in endometrial stromal tumours and in uterine tumours resembling ovarian sex cord tumours, supporting true sex cord differentiation.^65–67

MELAN-A

The antibody A103, which reacts against the melan-A antigen, is widely used in the diagnosis of melanocytic disorders, positivity being found in most benign and malignant lesions.⁷⁶ A103 immunoreactivity has been demonstrated recently in a small number of ovarian sex cord stromal tumours, positivity being found in three of four Sertoli-Leydig cell tumours.⁷⁷ Clearly, the value of this antibody in ovarian sex cord stromal tumour diagnosis requires further investigation of a larger number of cases.

MULLERIAN INHIBITING SUBSTANCE

Mullerian inhibiting substance (MIS) is a homodimeric glycoprotein secreted by ovarian granulosa cells and testicular Sertoli cells.⁷⁸ One of the functions of MIS is the regression of the Mullerian ducts during male fetal development.⁷⁹ MIS can be demonstrated immunohistochemically in ovarian granulosa cells and raised serum concentrations have been found in patients with ovarian sex cord stromal tumours.^80,81 Recently, an immunohistochemical study has shown the presence of MIS in all ovarian sex cord stromal tumours, although staining was not as strong or as diffuse as with anti-α inhibin.⁴⁶ There was no staining of seven endometrioid adenocarcinomas resembling sex cord stromal tumours. This study found that two of the endometrioid adenocarcinomas were positive for CD99. Clearly, the value of immunohistochemical staining for MIS in confirmation of an ovarian sex cord stromal tumour has not been investigated as extensively as for α inhibin, and further studies are required before anti-MIS becomes accepted in routine diagnosis practice.

RELAXIN LIKE FACTOR

The relaxin like factor (RLF), also known as the Leydig cell insulin like peptide, is a hormone that is strongly expressed by testicular Leydig cells.⁸² A recent immunohistochemical study has shown expression of RLF in ovarian hilar Leydig cell hyperplasia and in ovarian sex cord stromal tumours with a component of Leydig or luteinised cells.⁸³ RLF was not found in granulosa cell tumours, and neoplasms other than those of sex cord stromal type were not included in the study. The authors suggested that RLF might be a useful diagnostic tool in Leydig cell tumours but clearly this requires confirmation by further studies. Furthermore, Leydig cell tumours generally have a characteristic histological appearance and their distinction from other ovarian sex cord stromal tumours is usually not problematic.

CALRETININ

Calretinin is a 29 kDa calcium binding protein that is a useful immunohistochemical marker of mesothelial cells and mesothelioma.⁸⁴ Calretinin positivity has been demonstrated recently in all cases of a small number of ovarian sex cord stromal tumours.⁸⁴ However, because the number of cases studied is small, the value of this antibody in a diagnostic setting, as with anti-MIS and anti-RLF, needs to be confirmed by further investigations of larger numbers of cases.

Miscellaneous

This last section reviews several other recent developments in the immunohistochemistry of ovarian neoplasms that might be of use to the practising surgical pathologist.

Serum α fetoprotein (αFP) measurements and the demonstration of this protein immunohistochemically in tissue sections are useful in the diagnosis of an ovarian yolk sac tumour. However, positive staining and/or raised serum concentrations can also be found in ovarian sex cord stromal neoplasms, especially of Sertoli-Leydig origin, but also of other types.^85–88 In these cases, positivity might involve heterologous hepatoid cells, Sertoli cells, or Leydig cells. A recent report has described four ovarian granulosa cell tumours with foci of hepatic cell differentiation that were confirmed by positive immunohistochemical staining for αFP.⁸⁹ Because sex cord stromal neoplasms might bear some morphological resemblance to yolk sac tumours, this raises obvious diagnostic problems, especially in a young patient, with implications for future management and treatment regimens. In such cases, staining for α inhibin might be useful because this generally appears to be negative in yolk sac tumours, although the number of cases studied is small. Positive immunohistochemical staining for αFP (and raised serum concentrations) has also been demonstrated in ovarian endometrioid adenocarcinoma, a tumour that can also morphologically resemble a yolk sac tumour.⁹⁰

There are two distinct types of ovarian small cell carcinoma. The pulmonary type usually occurs in middle aged to elderly patients, and is morphologically and immunohistochemically identical to its pulmonary counterpart, exhibiting neuroendocrine differentiation. The hypercalcaemic type generally occurs in young women and is likely to be confused morphologically with granulosa cell tumour, especially the juvenile variant. The origin of small cell carcinoma of hypercalcaemic type has been debated, with an epithelial, stromal, or germ cell origin being proposed.⁹¹ A recent study showed EMA positivity in most cases, with all being α inhibin negative.⁵⁰ The authors concluded that small cell carcinoma of hypercalcaemic type is probably a variant of surface epithelial tumour. Table 3 is a suggested immunohistochemical panel to facilitate the distinction between ovarian juvenile granulosa cell tumour and small cell carcinoma of hypercalcaemic type.

The so called female adnexal tumour of probable wolffian origin (FATWO) (which is most commonly situated within the broad ligament but which might occasionally occur within the ovary) is a tumour of uncertain histogenesis. An origin from wolffian remnants is suggested by its usual location in areas where these remnants are abundant, and this is supported by ultrastructural analysis.⁹² A recent imunohistochemical study provides support for the derivation of this tumour from wolffian remnants, the pattern of staining being nearly identical to that of the normal rete ovarii.⁹³ Normal rete ovarii and FATWO coexpressed cytokeratin and vimentin and usually lacked immunoreactivity for EMA. Interestingly, most cases exhibited positive staining for α inhibin and calretinin. Inhibin positivity has also been noted in another study,⁵⁶ suggesting that anti-α inhibin might be useful in the diagnosis of FATWO. Because confusion can arise with a sex cord stromal neoplasm, α inhibin positivity does not allow distinction between these two types of tumour, although in FATWO staining is usually weak in contrast to the diffuse strong positivity in sex cord stromal neoplasms. Another lesion that can morphologically resemble FATWO is the adenomatoid tumour. This neoplasm is of mesothelial derivation and clearly staining for calretinin is of no value in this distinction.

Transitional cell lesions of the ovary include benign, proliferating, low malignant potential and malignant Brenner tumour, and primary transitional cell carcinoma. Normal urothelium and urothelial neoplasms of the urinary tract express CD20. A recent study has found that most transitional cell lesions of the ovary fail to express CD20.⁹⁴ The authors concluded that despite their morphological similarity to transitional cell carcinoma of the urinary tract, transitional cell lesions of the ovary are immunohistochemically distinct. They thought that the identification of pronounced CD20 expression in a malignant ovarian tumour with transitional features provides support for a metastasis from the urinary tract, rather than a primary ovarian surface epithelial neoplasm.

Pure sarcomas of the ovary are rare and a recent report described four cases of primary angiosarcoma.⁹⁵ In these cases, the vascular nature of the tumour was not immediately apparent, but all were immunoreactive for vascular markers.