Papillary tubal hyperplasia: the putative precursor of ovarian atypical proliferative (borderline) serous tumors, noninvasive implants, and endosalpingiosis

Considerable controversy has surrounded the terminology and behavior of noninvasive low-grade ovarian serous tumors [atypical proliferative serous tumors (APSTs) and serous borderline tumors (SBTs)] but little attention has been directed towards determining their origin. The most contentious issues concerning their terminology were resolved in 2003 by the Borderline Ovarian Tumor conference, under the auspices of the National Cancer Institute, which deemed that the terms APST and SBT are synonymous 17 . Accordingly, APST and SBT are used interchangeably in this report. The recent revelation that intraepithelial carcinoma in the fallopian tube is the likely precursor of most, if not all, ovarian high-grade serous carcinomas prompted us to reconsider the fallopian tube as a possible site of origin of low-grade serous tumors. Previously, we reported that mucosal and luminal calcifications frequently surrounded by bland epithelium in the fallopian tube, so-called “salpingoliths”, were significantly more frequently associated with APSTs than all other ovarian neoplasms (p<0.001) 16 . In that study, salpingoliths were found in 24% of stage I and in 51% of stage II and III APSTs. We were unable to determine whether these lesions arose in the fallopian tube and implanted on the ovary and peritoneum, arose in the ovary and implanted in the tube or that the tubal and ovarian lesions developed independently. During the course of a large population-based study of ovarian low-grade serous tumors from Denmark, aimed at evaluating their clinicopathologic and epidemiologic features as well as identifying biomarkers to assist in the distinction of noninvasive from invasive implants, we noted the frequent association and close morphologic resemblance of a particular type of tubal hyperplasia designated “papillary tubal hyperplasia (PTH)” with APSTs. Papillary tubal hyperplasia is characterized by small rounded clusters of tubal epithelial cells and small papillae, with or without associated psammoma bodies, floating within the tubal lumen. The present study describes the morphologic features of this lesion and its association with ovarian APSTs, peritoneal implants and endosalpingiosis.

Under low magnification the plical architecture has an abnormal, complex, “busy” appearance compared to the normal architecture ( Figs. 1A and 1B ). Under higher magnification the detached clusters of cells and small papillae floating in the tubal lumen are more clearly evident ( Figs. 2 – 5 ). They are composed of a single layer of epithelium containing ciliated, secretory and intraepithelial lymphocytes, identical to the cellular composition of the normal tubal mucosa. The intraepithelial lymphocytes are located just above the basement membrane and are smaller than the ciliated and secretory cells. They contain a small, hyperchromatic, round or convoluted nucleus that is surrounded by a clear space or halo ( Figs. 6 , 7 ). These cells have previously been termed “indifferent” cell and were thought to possibly be somatic stem cells of fallopian tube epithelium 22 but more recent studies have shown that they are lymphocytes which represent a form of mucosal-associated lymphoid tissue (MALT) 10 . Rare clusters of epithelial cells are flat and some contain a vacuole similar to that seen in ovarian serous tumors. Other clusters contain slightly larger cells with eosinophilic cytoplasm similar to eosinophilic metaplastic cells in the endometrium. Mitotic figures are rarely seen. Both the secretory and ciliated cells are bland and bear no resemblance to serous tubal intraepithelial carcinoma. As observed in our previous study 16 psammoma bodies can be found isolated in the tubal lumen or surrounded by a mantle of bland tubal epithelium in which case they qualify as “salpingoliths” ( Figs. 5 , 6 ). In addition to the tubal lumen, psammoma bodies are also present in the epithelium and lamina propria of the tube ( Fig. 6 ). A lesser degree of hyperplasia is characterized by small, raised tufts of epithelium that emanate from the tubal epithelium ( Fig. 8 ). As the hyperplasia evolves the tufts enlarge and form round clusters of cells identical to the clusters and papillae floating in the lumen but still attached to the tubal mucosa. Their connection to the epithelium is tenuous and gives the impression that they are about to be pinched off. ( Figs. 7 , 9 ). In some instances these attached small papillae contain a psammoma body in the core (salpingolith) identical to those papillae floating in the lumen. Due to the complex arrangement of tubal plicae, particularly in younger women, it is often difficult to be certain whether some of these papillary tufts without psammoma bodies are normal constituents of the tubal architecture cut tangentially or whether they are the earliest form of hyperplasia. In order to be conservative in determining the association of these lesions with APSTs, as indicated in the Materials and Methods section, we included only those in which there were at least 3 clusters/papillae per section. The presence of associated psammoma bodies was a useful feature in classifying a specimen as PTH as psammoma bodies are not a feature of normal fallopian tube epithelium. The morphologic similarity of PTH, APSTs and noninvasive implants is striking ( Fig. 10 ). Patients ranged in age from 22–83 years with a mean of 42 years and a median of 41.5 years. A more detailed description of the demographic findings will be presented in a future publication that describes the clinical and pathologic features of the entire cohort. The women in the current study had a mean age approximately 10 years younger than the remainder of the cohort. In all other aspects they did not differ significantly. ( Table 1 ) As noted in the Materials and Methods section, in the Danish study a subset of cases with tubal implants were selected because adjacent tubal tissue was available for review. Papillary tubal hyperplasia was found in 20 (91%) of 22 cases and was not detected in 2 (9%). The 2 cases in which PTH was not detected had only one section in each case available for review. In 15 (75%) of the 20 cases with PTH, papillae were found in all the sections of tube examined whereas in 5 (25%) papillae were absent in some of the sections indicating that the process may not diffusely involve the tube. It appeared as if the ampullary portion of the tube was more frequently and more extensively involved compared to the infundibulum and isthmus, however, in this study the ampulla was more frequently sampled than the other parts of the tube. The number of intraluminal papillae varied from as few as 3 per case (two cases) to “too numerous to count” (6 cases). In the remaining cases the number of papillae ranged from 5–42. Psammoma bodies were found in 11 (50%) cases ranging from a few (2 per case) to many (too numerous to count). They were located in the lamina propria of the tubal plicae, epithelium and lumen. A chronic inflammatory infiltrate, confined to the lamina propria of the tube, was present in 8 (36%) of the 22 cases. One of these cases also demonstrated acute inflammation. Evidence of previous episodes of pelvic inflammatory disease based on agglutination of plicae, bridging of plicae across the tubal lumen or loss of plicae resulting in hydrosalpinyx was found in 8 (36%) of the 22 cases. As mentioned in the Materials and Methods section all of the cases selected for study had implants involving the fallopian tube. They were noninvasive epithelial implants in 15 (68%) of the 22 cases, noninvasive desmoplastic implants in 4 (18%) and invasive implants (metastatic low-grade serous carcinoma) in 3 (14%). It should be noted that some noninvasive epithelial implants were surrounded by fibrotic tissue but they did not have the typical appearance of desmoplastic implants in which the epithelial component merges with the stromal component and has a mesothelial appearance. The 3 invasive implants were all associated with invasive low-grade serous carcinomas of the ovary. ( Table 2 ) The associated ovarian tumor was an APST in 19 cases and an APST with an invasive low-grade serous carcinoma in 3. In addition to the fallopian tube, noninvasive implants were identified in the omentum in 2 cases and on the uterine serosa in one. Endosalpingiosis involving the omentum and tube was found in 6 (27%) of cases. The clinical and pathologic findings are summarized in Table 3 . A detailed description of the 7 cases is included in an appendix . Briefly, among the 7 JHH consultation cases the number of sections from the tubes ranged from 2–21. Papillary tubal hyperplasia was found in all sections in only two cases indicating that the lesion can be focal. Psammoma bodies were detected in 4 (57%) of the 7 cases. Chronic salpingitis was present in 3 (43%) of the cases, one of which was severe and also acute. In the latter case there was pyosalpinx and marked destruction of the plicae and in 2 other cases there was distortion of the plical architecture but no evidence of inflammation. Thus, there was evidence of pelvic inflammatory disease in 5 (71%) of the 7 cases. Endosalpingiosis was present in 4 (57%) cases and a noninvasive epithelial implant in another.

In 1969 Pauerstein and Woodruff wrote in the preface of their textbook on the fallopian tube “The era of indifference to the fallopian tube has passed” 22 . This statement was more prophetic than timely because it took almost another half a century for indifference to turn to intense interest as investigators proposed that intraepithelial carcinoma in the fallopian tube was the long sought for precursor of ovarian high-grade serous carcinoma 4 , 5 , 8 , 12 , 13 , 19 . As part of this renaissance in the role of the fallopian tube in ovarian carcinogenesis we renewed our interest in exploring the link between tubal hyperplasia and ovarian low-grade serous tumors. Proliferative lesions in the fallopian tube other than carcinoma have received relatively little attention. There is general agreement that tubal hyperplasia exists but the criteria for its definition, including what constitutes the minimal degree of proliferation, have not been clarified 3 , 9 , 11 , 14 , 22 , 23 . Similarly, investigators have debated whether tubal hyperplasia is significantly associated with APSTs. Two studies that investigated this relationship arrived at diametrically opposite conclusions. Robey and Silva retrospectively reviewed representative sections of fallopian tubes from 99 women with SBTs and compared them to the fallopian tubes of 58 patients with cervical carcinoma and high-grade ovarian serous carcinoma and found tubal hyperplasia in 68.7% of the SBT cases compared to 25.9% in the control group (p< 0.01) 14 . In contrast, Yanai-Inbar and Silverberg analyzed fallopian tubes from 49 women with SBTs compared to a control group, which included other borderline tumors, a variety of invasive genital tract carcinomas, benign ovarian and uterine tumors and portions of tubes in women undergoing tubal ligation and found no significant association 23 . These authors graded tubal hyperplasia as mild, moderate and marked based on the degree of stratification and cellular atypia. They concluded that since mild hyperplasia was often found in women undergoing tubal ligation this degree of proliferation was within normal limits. Nonetheless, even moderate and marked hyperplasia showed no significant difference between the SBT cases and controls. Based on reviewing the criteria for tubal hyperplasia and the photomicrographs in these papers we believe that the different conclusions that were reached are due to the use of different criteria for the diagnosis of tubal hyperplasia. Our analysis confirms that there are varying degrees of tubal hyperplasia. The earliest recognizable form begins with focal epithelial stratification followed by a slight uplifting of the tubal epithelium that forms a small tuft. With increasing proliferation the tuft expands and extends into the tubal lumen to form a small epithelial bud. The buds expand and form rounded papillae composed of cells identical to those in the tubal mucosa. The papillae can be solid in which case they are classified as “clusters” or they can contain a central stromal core in which case they are termed “papillae”. Since they appear to be variations on the same theme, we have used the terms interchangeably. As the tubal buds enlarge, they become more distinct, rounded and appear to be pinched off and extruded into the tubal lumen. Psammoma bodies can be present in the center of the papillae or are "naked" without associated epithelium. Besides the papillae in the tubal lumen, psammoma bodies are located in the lamina propria and the tubal mucosa. It is these papillae and psammoma bodies that are the key diagnostic features of PTH. The presence and location of the psammoma bodies is an intriguing aspect of this lesion as they are often the dominant feature. Their location in the lumen, epithelium and lamina propria suggest that they develop in the epithelium during the formation of an epithelial bud. It appears that the buds may then be either extruded into the tubal lumen or invaginate into the underlying stroma. The epithelium subsequently degenerates leaving a “naked” psammoma body in the tubal lumen or in the lamina propria. A similar process can occur in extratubal sites as well. In this study and in our experience it is not unusual to find a myriad of psammoma bodies coating peritoneal surfaces, notably the uterine serosa, ovaries and bowel serosa often with very little associated epithelium. We speculate that initially epithelium surrounded the psammoma bodies but then degenerated leaving only “naked” psammoma bodies which act as an irritant in the peritoneal cavity and induce fibrosis that can lead to adhesions and bowel obstruction. The different conclusions reached by Robey and Silva and Yanai-Inbar and Silverberg regarding the association of ovarian serous borderline tumors with tubal hyperplasia probably relates to the different criteria that were used to diagnose tubal hyperplasia. Specifically, PTH is frequently associated with APSTs but lesser degrees of hyperplasia are not. Yanai-Inbar and Silverberg illustrate the very early budding stage (their Fig 3 ) but not the fully developed free floating epithelial clusters 21 . On the other hand Robey and Silva mention and illustrate the early budding stage as well as the detached luminal clusters of epithelial cells (their Fig 1 ) but they do not specifically indicate how often this latter change was found in their cases 14 . The findings in the current study strongly suggest that PTH is the most advanced stage of tubal hyperplasia and it is this lesion that is associated with APSTs. We speculate that lesser degrees of tubal hyperplasia may not necessarily progress to PTH and therefore are not significantly associated with APSTs. For that matter, not all PTH progresses to APSTs as illustrated by the 7 JHH cases in this study which were not associated with an ovarian tumor. It is conceivable that if the tubes in these cases were not removed PTH may have resulted in the development of an APST. The fallopian tube has also been implicated in the development of endosalpingiosis. Sampson, who first described endosalpingiosis, believed that it arose from epithelium that was sloughed from the fallopian tubes 15 . Subsequently, Zinssner and Wheeler in a study of 128 omenta removed at autopsy and surgery reported 16 (14.8%) examples of endosalpingiosis 24 . Of further interest in their study was the presence of an associated serous cystadenoma in one case and SBTs in 3 others. One of these latter cases was reported as a serous carcinoma with metastases but the primary tumor and metastases were described as having numerous cilia and therefore based on this observation and their two photomicrographs (their Figs. 8 and 9 ) we interpret this as an SBT with noninvasive epithelial implants and endosalpingiosis. These authors found no transition of mesothelium to endosalpingiosis. They also noted that there were no examples of endosalpingiosis in the male patients and therefore concluded that endosalpingiosis resulted from sloughing and implantation of tubal epithelium. We postulate that PTH is responsible for the development of both endosalpingiosis and noninvasive epithelial implants even in the absence of an APSTs. Endosalpingiosis is composed of glands lined by a single layer of tubal-type epithelium. When endosalpingiosis exhibits papillary tufting or detachment of cell clusters or cribriform patterns and varying degrees of cytologic atypia it has been classified as “atypical endosalpingiosis” 2 . As the latter becomes more florid it merges into what is classified as a “noninvasive epithelial implant”. Endosalpingiosis, frequently accompanies SBTs and has been reported as high as 40% in some series 18 . In our Danish population-based study of low-grade serous tumors the prevalence of endosalpingiogsis was 10% (unpublished data). Accordingly, it appears that endosalpingiosis and noninvasive epithelial implants are part of a continuum of benign low-grade serous proliferations involving the peritoneum that result from implantation of tubal epithelium. Although sometimes noninvasive epithelial implants are embedded in fibrous tissue, they differ from noninvasive desmoplastic implants, morphologically and immunohistochemically (unpublished data). The latter typically display a biphasic epithelial and spindle cell growth pattern often associated with a chronic inflammatory infiltrate. It is conceivable that some of these may develop in situ. In contrast, invasive implants are metastatic low-grade serous carcinoma from either a noninvasive low-grade ovarian serous carcinoma (micropapillary serous) or an invasive low-grade serous carcinoma 19 . In the current study the three cases of invasive implants were all associated with an ovarian low-grade serous carcinoma. Another interesting observation in this study, as in our previous study on salpingoliths 16 , was the frequent association of chronic salpingitis and distorted tubal architecture consistent with previous pelvic inflammatory disease with PTH and endosalpingiosis. Zinsser and Wheeler in their autopsy study reported 14 (88%) of 16 cases of endosalpingiosis had evidence of chronic salpingitis either in the form of chronic inflammation or distorted tubal architecture 24 . It is well known that inflammation may stimulate proliferation of tubal epithelium and therefore it is plausible that chronic salpingitis may play a role in the pathogenesis of PTH and endosalpingiosis. In the current study evidence of pelvic inflammatory disease in the form of chronic salpingitis or distorted tubal architecture was found in 35–70% of the cases (Danish and JHH cases respectively). Three possible mechanisms can explain the association of PTH and APSTs. One is that the tubal and ovarian lesions arise independently, a process that has been termed a “field effect”. A second is that the ovarian tumor is primary and that the clusters of cells and papillae in the tubal lumen are a secondary phenomenon. The third possibility is that the tubal lesion is primary and that the ovarian and extraovarian lesions (APST, noninvasive epithelial implants and endosalpingiosis) are secondary. Robey and Silva concluded that tubal hyperplasia and SBTs developed independently as a result of a so-called "field effect" 14 as did McCaughey et al. in their study of SBTs and implants 7 . This implies that SBTs develop from peritoneal mesothelium, through a process of metaplasia resulting in their müllerian phenotype. The concept of müllerian metaplasia of peritoneal mesothelium, although difficult to completely disprove has, on the other hand, no evidence to support it. In the current study, the morphologic similarity of PTH, noninvasive epithelial implants and APSTs is striking. All of these lesions, in addition to being composed of ciliated and secretory cells identical to those in normal fallopian tube epithelium, also contain intraepithelial lymphocytes which are also found in normal tubal epithelium. Invoking metaplasia of the peritoneum (mesothelium) to account for the development of this complex assortment of cells seems highly unlikely. In addition, in our experience PAX 8, a müllerian marker, is expressed in fallopian tube epithelium and APSTs whereas calretinin, a mesothelial marker, is not. In contrast, the ovarian surface epithelium strongly expresses calretinin but not PAX 8 (unpublished data). Furthermore, psammoma bodies are an integral part of PTH as they are found in the majority of cases and psammoma bodies are frequently associated with APSTs. In contrast, psammoma bodies are typically not associated with mesothelial proliferations and when they are, they usually are few in number. Finally, noninvasive epithelial implants, like endosalpingiosis, have not been reported in male patients. Not withstanding rare reports of endometriosis in the prostate of men treated with estrogen 1 there is no support for the argument that female sex steroid hormones are responsible for müllerian metaplasia of the peritoneum leading to the development of endosalpingiosis and noninvasive epithelial implants in women. We believe the most plausible explanation is that PTH is the precursor of all of these lesions since it was identified in 7 cases in the current study in the absence of a primary ovarian tumor. Moreover, in one of these cases, a noninvasive epithelial implant was identified and in 4 cases PTH was associated with endosalpingiosis. Based on the findings in this study we propose the following model for the origin and development of the entire spectrum of pelvic low-grade serous proliferations. Chronic inflammation induces a proliferation of tubal epithelium that can progress to PTH in some women. Epithelial clusters and papillae are shed and implant on ovarian and peritoneal surfaces. On the ovary the lesion is termed “cortical inclusion cyst” and when it involves extraovarian sites it is termed “endosalpingiosis”. In contrast to cortical inclusion cysts lined by tubal-type epithelium, some cysts are lined by flat epithelium and represent invaginated clefts from the surface of the ovary that have been tangentially cut. Accordingly, there are two types of cortical inclusion cysts, one derived from invaginated surface epithelium and therefore of mesothelial origin and another resulting from implanted tubal epithelium and therefore of müllerian origin. It is also conceivable that normal tubal epithelium can implant directly on the ovary at the time of ovulation when the ovarian surface epithelium is disrupted and the fimbria are in direct contact with the denuded ovarian surface. When endosalpingiosis exhibits a complex glandular arrangement with varying degrees of cytologic atypia it qualifies for the diagnosis of “atypical endosalpingiosis” which merges imperceptibly into noninvasive epithelial implants. If mutation of KRAS or BRAF occurs in any of these lesions an APST develops 20 . This model can explain the development not only of ovarian APSTs but also the development of noninvasive implants and APSTs involving pelvic and abdominal sites in the absence of an ovarian tumor. As previously noted, one of the aims of our ongoing population-based study of ovarian low-grade serous tumors in Denmark is to identify biomarkers to assist in the distinction of noninvasive and invasive implants and therefore the presence of PTH was not recorded in the data collection. We are, therefore, unable to determine the frequency of PTH in the study population. In an effort to address this issue we are currently performing a study comparing the frequency of PTH in women with APSTs to a control group of patients in whom the entire fallopian tube has been processed using the SEE-FIM technique 8 to determine more accurately the frequency of this lesion in women with APSTs and in the normal population. In conclusion, we propose a model for the development of all low-grade serous proliferations involving ovarian and extraovarian sites (APSTs, noninvasive epithelial implants and endosalpingiosis). The process begins with chronic inflammation leading to tubal hyperplasia, which if it progresses to PTH can shed and implant tubal epithelium on ovarian and peritoneal surfaces resulting in a variety of low-grade serous proliferations. If this hypothesis is confirmed it would indicate that all ovarian serous tumors, low- and high-grade, originate from tubal epithelium and involve the ovary secondarily.

Twenty-two cases from a population-based study of approximately 1000 ovarian low-grade serous tumors in Denmark and 7 recent cases of PTH seen in consultation at The Johns Hopkins Hospital (JHH) Department of Pathology, Division of Gynecologic Pathology constitute this descriptive analysis. Clinical information was obtained from a review of medical records and pathology reports from the Danish cases and from consultation letters, pathology reports and communication with the referring pathologists or clinicians for the JHH consultation cases. From 1 to 4 sections from the fallopian tubes were examined in the Danish study and 2–21 sections from the JHH consult cases. It should be noted that in all cases the sections were representative sections only. The aims of the Danish study were to evaluate the clinicopathologic and epidemiologic features of ovarian low-grade serous tumors in the female population of Denmark. One specific aim was to determine if immunohistochemical and molecular genetic analysis could distinguish noninvasive from invasive implants. This will be the subject of a future report. From among the 1000 tumors that qualified for the diagnosis of APST a subset of cases with implants in the mesosalpinx in which fallopian tubes were available for microscopic review were selected for further study. It should be noted that although one of us (RJK) was aware of these tubal proliferations and their association with APSTs for some time, it was not the intent at the start of the Danish study to evaluate them. It was only during the course of selecting cases for immunohistochemical and molecular genetic analysis of the implants that we were again struck by the resemblance of the tubal lesions to APSTs and noninvasive implants and their high frequency in these tubes that led to the current investigation. The 7 consultation cases were sent to us around the same time as the analysis of the Danish cases was being performed. Although we have observed a number of such cases over the years, an exhaustive attempt to search for similar cases in our files was not undertaken. Because of the difficulty in the past in agreeing on what constitutes tubal hyperplasia 3 , 9 , 14 , 23 we restricted our definition to include only those tubal proliferations that exhibited papillary tufting and detached clusters of bland epithelium frequently, but not always, associated with psammona bodies. These clusters of epithelial cells and small papillae were found floating in the lumen or protruding from the tubal mucosa into the lumen and hence the lesion was termed “papillary tubal hyperplasia (PTH)”. As the number of papillae can vary considerably, for purposes of this study, we arbitrarily included only those cases in which, a minimum of 3 papillae/section were identified. Stratification of tubal cells alone was not sufficient for inclusion as this finding, which in some instances is probably due to tangential sectioning, is often found in fallopian tubes without any other evidence of hyperplasia and is therefore considered a normal variant.