|Home | About | Journals | Submit | Contact Us | Français|
In the past 50 years, the concept of serous ovarian cancer has been progressively refined, with the distinction of the borderline serous tumour, identification of a smaller subset of well-differentiated serous malignancies and, recently, closer attention to the pathogenesis of high-grade serous malignancies. High-grade serous carcinoma, traditionally presumed to arise within Müllerian inclusion cysts of the ovarian surface, cortex and peritoneum, has recently been linked to the distal fallopian tube. This review addresses the disparate forms of serous neoplasia, which reflect both different genetic abnormalities and stages of differentiation of Müllerian epithelium. The significance of these different origins is addressed in the context of ovarian cancer prevention.
Epithelial carcinoma of the ovary is a common malignancy in women and, after lung, breast, colorectal and pancreatic carcinomas, is the fifth most frequent cause of cancer in women.1,2 Approximately 15% of epithelial ovarian tumours are borderline tumours (low malignant potential). The malignant group consists of a diverse histological profile, including tumours resembling endocervix or gastrointestinal mucosa (mucinous), endometrium (endometrioid), fallopian tube (serous) and unspecified glycogenated epithelium (clear cell tumours). The proposed origin for all of these tumours is Müllerian epithelium that has found its way into the ovarian cortex. The most common explanations for serous/tubal differentiation include ovarian surface epithelium that has undergone Müllerian metaplasia, endometrial tissue (endometriosis) derived from such metaplastic foci or transported directly from the endometrial cavity to the ovary, and fallopian tube epithelial cells that have implanted on the ovarian surface or become adhesed to the ovary and incorporated into the parenchyma (endosalpingiosis).3,4 Repeated ovulation presumably assisting this process by producing defects in the ovarian surface and encouraging the incorporation of such epithelium.5
Endometrioid tumours have traditionally been linked to ovarian endometriosis, which can be demonstrated in approximately 50% of cases. Mucinous tumours theoretically arise from either endometriosis (Müllerian mucinous) or Müllerian inclusions (intestinal mucinous). Low-grade serous tumours presumably arise from Müllerian inclusions, commonly found in the ovarian cortex and mesothelial surfaces.6 The origin of high-grade serous carcinomas, however, remains controversial. This controversy has been based primarily on the lack of concrete morphological data linking these tumours to cortical inclusions or the ovarian surface epithelium. Superimposed on this conundrum is the fact that not all ovarian carcinomas are readily classified into one of these types. In particular, the distinction between high-grade endometrioid and serous carcinomas in the female genital tract is not always clear-cut.7
This review will address recent advances in the histopathological and molecular classification of pelvic serous neoplasia, from the perspective of both biological behaviour and cell of origin. The discussion will address low- and high-grade serous malignancies.
Low-grade serous tumours are composed of cells that closely mimic the fallopian tube, including the presence of both ciliated and secretory differentiation. They are the most common epithelial tumours of the ovary, accounting for about 30% of ovarian neoplasms. They are less frequently reported in Asia. Overall, approximately 70% are benign or borderline when detected. Although most women with serous malignancies are in their sixth or seventh decades when diagnosed, the borderline tumours present at an average age of 45 years and can be encountered in the second and third decades. Seventy per cent are confined to one or both ovaries when diagnosed and disease-free survival is > 98%. Only 15–30% of higher stage borderline tumours recur, signifying that pelvic spread is generally not an adverse prognostic indicator. The most common and most benign form of spread is the non-invasive implant. Other features of borderline tumours that have been increasingly scrutinized are micropapillary epithelial growth pattern, microinvasion in the stroma, lymph node spread and invasive implants.
By the 1990s the ‘invasive’ peritoneal implant emerged as an important prognostic indicator, distinguishing borderline tumours with a good versus poor outcome. The 7-year disease-free survival in patients with non-invasive implants is 95.3%; with invasive implants it is 66%.8–10
In addition to the category of implant, two other indices that conceivably would impact on prognosis – the presence of microinvasion in the supporting stroma and lymph node spread, have been described.11,12 Taken together, these features have resulted in a progressive refinement in the classification of low-grade serous tumours to include conventional borderline tumours (also termed proliferative serous cystadenomas) and those with complex microarchitecture, variously termed borderline serous tumours with micropapillary features and low-grade micropapillary serous carcinomas. Reasons for the disagreements in terminology are legion, but in practice stem from the fact that irrespective of the term used to classify the primary tumour, the presence or absence of invasive implants is the paramount question, determining the likelihood of extra-ovarian tumour growth, recurrence and death.8–10
The origin of the majority of low-grade serous tumours is presumed to be ectopic Müllerian epithelium. This epithelium exhibits the same immunophenotype as epithelium derived from the salpinx, although its origin is disputed. The most easily understood is the transfer of salpingeal epithelium from the tube to the ovarian surface or sites of ovulation, leading to cortical inclusion cysts (CICs) or endosalpingiosis.13 Alternatively, retrograde regurgitation of endometrial epithelium (and stroma) could produce the same result, giving rise to similar inclusions, including endometriosis. A third mechanism would be through the presence of a specialized ovarian surface epithelium (OSE) that is capable of Müllerian metaplasia.14 Irrespective of the mechanism by which this epithelium becomes entrapped within the ovarian cortex, the fundamental requirement for most (but not all) low-grade serous tumours appears to be ovarian stroma and a resulting ovarian stromal–epithelial relationship that plays some role in neoplastic transformation.
The mechanisms by which neoplastic transformation occurs in Müllerian inclusions are not well understood. A common but not universal component of this process is a mutation in K-Ras.15 More obvious, if equally mysterious from a mechanistic standpoint, is the shift in differentiation that takes place in the target Müllerian epithelium. Studies have implicated homeobox regulatory (HOX) genes in this process, suggesting that the expression of HOX 11 mediates the salpingeal phenotype. In this model, other cell-type-specific transitions from CIC to neoplasia are regulated by the respective HOX gene, with mucinous tumours expressing HOX 9 and endometrioid tumours expressing HOX 10. In the former, a more rapid transition likely takes place from CIC to an intestinal ovarian mucinous tumour, given the relative lack of benign-appearing CICs with an intestinal mucinous phenotype. In this model, the activation of HOX 9 would occur concurrently with neoplastic transformation. Endometrioid (and derived Müllerian mucinous tumours) would be predated by a stable endometrioid (HOX 10) phenotype on which subsequent events would be superimposed to develop their respective neoplasms.16i
Central to the low-grade serous neoplasm – and all serous tumours – is the target epithelium and its differentiation programme. Salpingeal epithelium has two predominant cell types, secretory and ciliated cells. A third cell type, and one which is most conspicuous in the region of the fimbria, is a sub-columnar reserve cell similar to what has been described in the cervix and endometrium. These reserve cells appear to arise from the epithelium near the distal portion of the tube and probably gives rise to some of the metaplastic epithelia (Walthard cell rests) that are often seen there. The ciliated and secretory cells are largely independent, both immunohistochemically and functionally, the latter verified by cell culture studies.17 However, there is circumstantial evidence that the two may be related to a common progenitor cell. Low-grade serous neoplasms, which are presumed to be clonal in origin, typically exhibit both ciliated and secretory differentiation, the former evidenced by staining for LHS28 and p73, and the latter by localization of bcl-2, HMFG2 and PAX-8 (Parast M, Crum C, Hirsch M, unpublished observation). Thus, a fundamental characteristic of low-grade serous tumours, including those classified as borderline and low-grade serous carcinoma, is a molecular pathway that permits expansion of a tumour cell population with the ability to execute the transition to both ciliated and secretory cells (Figs 1 and and22).
Cell cycle proteins are identified almost exclusively in the non-ciliated cell population, suggesting on the surface that ciliated cells emerge from a less differentiated precursor pool.18 While this may be true, it would be premature to declare the ciliated population terminally differentiated. Cell culture studies of bovine oviduct have shown that ciliated cells can be plated, at which time they lose their cilia and begin to grow. This suggests that ciliated cells possess considerable independence and are capable of reentering the proliferative compartment, at least in vitro.19ii Whether this proliferative compartment is secretory per se is unclear.
Figures 2–4 illustrate three scenarios that might be encountered in low-grade serous tumours. The first is a typical borderline tumour with some epithelial complexity, illustrating the different cell types involved (Fig. 2). The second is a biphasic borderline serous tumour, undergoing increasing complexity with stromal invasion. The latter is mirrored on the surface of the tumour and in the peritoneum, where the disease is sufficiently extensive to be classified as a low-grade serous carcinoma (Fig. 3). The ‘morphological gap’ between the polar ends of this tumour is clear, indicating that at least two events, initiation and progression, took place. In the third, a low-grade serous tumour with florid micropapillary architecture is associated with non-invasive and invasive peritoneal implants, the latter features presumably reflecting the higher risk of adverse outcome expected from such tumours (Fig. 4).20
Serous carcinomas account for up to 60% of all epithelial carcinomas and are the most lethal.21 Over 80% are stage III or greater when diagnosed and this group carries a 5-year survival of 20–30%. In contrast, endometrioid and mucinous carcinomas carry a 60–80% 5-year survival due to their much higher likelihood of being discovered when at least apparently confined to one or both ovaries. These unfortunate statistics for serous carcinomas apply principally to the high-grade category, which are by far the most common and make up a significant proportion of the 12 000–15 000 women who die of ovarian carcinomas each year. Serous carcinomas are particularly common in women with a heritable heterozygous mutation in BRCA1 or BRCA2, who may also rarely develop endometrioid tumours.22,23
The gross presentation of pelvic serous malignancies varies widely, but the rarest is the unilateral (stage I) tumour confined within a cyst. The more common presentations include either bilateral ovarian involvement with surface nodules or tumours that predominate in the omentum with relative sparing of the ovaries, except for surface implants. Rarely, serous carcinomas present as a dominant mass in the fallopian tube, typically with fimbrial adhesions that seal the tumour within the endosalpinx, resulting in a sausage-like tumour mass replacing the entire tube. Assigning the primary site of origin is typically difficult, in as much as the entire tube and ovary can be obliterated. Although this review will use the generic term ‘pelvic serous carcinoma’, thus avoiding any commitment (and controversy) in the general sense, standards have been proposed for assigning site of origin. They are as follows:
Despite the fact that pathologists have relied on the above algorithm or a similar approach to assign primary site, several observations have prompted investigators to revisit the concept of tumour origin in serous tumours. The most obvious is that approximately one-half of ovarian carcinomas are not associated with an obvious pre-existing conditions in the ovarian parenchyma.25 This has supported the theory of a more rapid evolution for many serous malignancies. Three sources in the microenvironment are as follows:
Recent studies have shown that a high percentage of early carcinomas in BRCA+ women can be traced to the distal fallopian tube rather than the ovary and in many the only manifestation is an early carcinoma (TIC, Fig. 7). In the study by Finch et al, six of seven carcinomas involved the distal tube.42 Callahan et al presented data supporting an origin in the distal tube in all of their seven cases. Since that study an additional three cases have been diagnosed at Brigham and Women’s Hospital. In all, only 1 of 10 appeared to arise within the ovary, within a large MIC (Fig. 5). These studies imply that a very high percentage of carcinomas arising in BRCA+ women occur in the distal tube. Based on the literature, the great majority are serous in type. However, it is important to emphasize that two of the cases we have seen were distinctly endometrioid in appearance and were p53 immunonegative, suggesting that the BRCA mutation places these women at risk for both types.
Although the majority of pelvic serous carcinomas in BRCA+ women appear to arise in the distal fallopian tube, the origin(s) of serous carcinomas in women without a documented BRCA mutation – 85% of all serous carcinomas – has not been scrutinized in detail. In prior reports, tubal involvement has been estimated to occur in < 20% and intraepithelial carcinomas (TIC) have been the subject of case reports, presumably signifying a field effect rather than a causal event. In a recent study, we surveyed a consecutive series of pelvic serous carcinomas and examined the fallopian tubes in detail using the SEE-FIM protocol to determine the frequency of TIC. In all, approximately three of four cases exhibited endosalpingeal involvement. When cases fulfilling the criteria for peritoneal or tubal primary neoplasms were excluded, one-half of the remaining cases contained evidence of TIC in the fimbria (Fig. 8). One-quarter contained endosalpingeal involvement without documented TIC and another quarter contained circumstantial evidence (cystadenoma, dominant ovarian mass, endometriosis) of an ovarian origin.44 In five cases with both TIC and ovarian carcinoma, a discrete p53 mutation was shared by both tumours (Fig. 8). Although this finding does not confirm that the TICs were the initial event in tumourigenesis, they are identical in appearance to their counterparts in BRCA+ women and are logical candidates for an early phase in serous carcinogenesis.
In a follow-up study, we examined a series of women with tumours classified as primary peritoneal in origin. This group of neoplasms traditionally has been presumed to arise from the peritoneal surface, within MICs or endometriosis. Consecutive cases were selected which met the 2001 WHO criteria for primary peritoneal serous carcinoma (PPSC) and fallopian tube involvement was assessed in eight consecutive, prospectively accrued samples with complete tubal exam, using the SEE-FIM protocol. Five of the eight (63%) cases with both tubes available for analysis contained endosalpingeal involvement and four (50%) had TIC (Fig. 9). Based on this, we concluded that TIC is frequently present in PPSC and that the fimbria is a candidate site of origin for many of these tumours.45 What is evident in both studies, however, is that not all primary peritoneal serous carcinomas can be attributed to the distal fallopian tube.
We recently looked for potential differences between tumours that appear to arise from the ovary – by virtue of a dominant ovarian mass – and those that are characterized by a similar distribution of tumour in both. A significantly higher percentage of the latter were associated with involvement of the endosalpinx, suggesting that a higher percentage of such tumours could have originated from the distal fallopian tube.46 However, differences in expression of tubal secretory cell markers (PAX8) and p53 were virtually identical, implying that both cell type and mechanism of tumour development will not distinguish the two origins. This is indirect evidence that a common cell of origin is responsible for all serous carcinomas, and supports prior assertions that the tube (or its equivalent) is responsible for all high-grade pelvic serous carcinomas.
Three components of serous cancer, TIC, invasive and metastatic serous carcinoma, are well enough defined to construct a model of progression that includes non-invasive neoplasms. However, a non-malignant precursor to this process has not been defined. Wells and colleagues described p53-positive MICs in the ovarian cortices of women with serous cancer. Boyd, in unpublished reports, described similar changes, and confirmed the presence of p53 mutations. In the fallopian tube, Piek et al described tubal ‘dysplasias’, defined as non-cancerous but probably precancerous. Recently, in our efforts to detect early serous carcinomas in BRCA+ women, we immunostained selected cases for the tumour suppressor protein p53. p53 Protein is over-accumulated in serous carcinomas and can be appreciated as a strongly positive intranuclear signal in approximately 85% of cases. We found that during our review of these cases, normal-appearing, non-dysplastic mucosa occasionally exhibited strong nuclear staining. These foci, which we term ‘p53 signatures’, shared several features in common with early fallopian tube carcinomas detected in BRCA+ women (Fig. 10).47 These are summarized as follows:
The p53 signatures are distinct from intraepithelial carcinoma in two respects. First, they demonstrate minimal proliferative activity and second, minimal or no nuclear atypia. However, in cases of tubal carcinoma, we have occasionally identified a second form of signature that exhibits increased proliferative activity, is discontinuous to the carcinoma and does not fulfil all morphological criteria for intraepithelial carcinoma. These non-proliferative and proliferative p53 signatures suggest that the transition from a non-proliferative p53 signature to intraepithelial carcinoma is not always abrupt.51
We have made two morphological observations suggesting that p53 signatures are capable of transition to TIC. One was a case in which a non-proliferative p53 signature was in direct continuity with a TIC; the transition was marked by a conspicuous increase in proliferative (MiB1) activity cyclin E staining. The second was spatially separate lesions that were intermediate between p53 signatures and TICs, containing increased nuclear stratification, intermediate levels of MiB1 and cyclin E activity and, often, p53 mutations. The latter entity suggests a more gradual transition from p53 signature to TIC.
Parallels can be made between the fimbrial model and traditional models of ovarian carcinogenesis: because the fimbriated end of the fallopian tube is in immediate proximity to the ovarian surface, mechanisms proposed for carcinogenesis of the ovarian surface epithelium (OSE) could be applied to the fimbrial mucosa. One of the more intriguing proposes a scenario in which ovulation leads to the release of oxidants, which in turn cause DNA damage to adjacent epithelial cells. The risk of damage diminishes geometrically as a function of increasing distance from the point of follicle rupture. This mechanism would explain the high frequency of p53 signatures in the fimbriae, diminishing rapidly in prevalence in the more proximal tube. DNA damage can result from, or be encouraged by, a wide range of factors, including DNA proliferation, hormonal stimuli and defects in both BRCA and p53 function. Nevertheless, given the relationship between oxidative injury and DNA damage, p53 stabilization and the role of p53 as both a component of the DNA damage response and as an intracellular antioxidant, the p53 signature emerges as an attractive early marker of serous carcinoma risk.
Taken together, the above observations suggest that several variables influence the actual or perceived origin of pelvic serous carcinoma:
Both low- and high-grade pelvic serous carcinomas appear to arise from Müllerian epithelium, and the target epithelium is similar, irrespective of location, be it tube, ovary or peritoneum. However, the presence of ovarian stroma appears particularly important to the lower grade serous tumours, suggesting that epithelial–stromal interactions influence the early events leading to tumour development. Central to this process is the frequent preservation of both a secretory and ciliated cell phenotype. Higher grade tumours, most of which exhibit p53 mutations, are strongly associated with not only the ovary but the distal fallopian tube, and early on (p53 signatures and TICs) appear to arise from cells of secretory differentiation. Although these tumours may express antigens associated with ciliated cells (such as p73), they are best viewed as either more poorly differentiated or following a pathway of differentiation more typical of the secretory cell. The choice depends in part on whether the secretory cell is viewed as a direct precursor to the ciliated cell or whether the two diverge from a common progenitor. A more complete understanding of these differentiation pathways, their site specificity and their vulnerability to early genotoxic events will be vital to understanding the pathogenesis of pelvic epithelial cancer and proposing approaches to cancer prevention.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.