Abstract
The diagnostic distinction between endometrial serous carcinoma (ESC) and endometrial clear-cell carcinoma (CCC) may occasionally be problematic, and one potentially contributing factor is the finding of clear cells in otherwise classic cases of ESC. This study aimed to define the frequency of this finding and comparatively assessed the immunophenotype of the clear cells. A review of 56 cases of ESC identified 8 (14.28%) with clear cells, representing 1% to 20% (median 7.5) of tumoral volume in these cases. In only 3 cases were clear cells discernible at low (×20) magnification. There was no significant difference in stage distribution or age between ESC patients with and without clear cells. The immunophenotypes of ESC-associated clear cells (group 1) were compared with foci of conventional ESC on another tissue block within the same case (group 2; n = 8) as well as a randomly selected cohort of CCC cases (group 3; n = 8). Groups 1 and 2 showed no significant differences regarding p53, ER, PR, Napsin-A, p504S, and hepatocyte nuclear factor 1β (HNF1β) expression, or regarding mitotic indices or Ki67 proliferation rate. In contrast, group 1 cases showed an immunophenotypic profile that was notably different from that of group 3 cases, with the former showing statistically significantly higher/more frequent expression of ER, PR, Ki67, and p53 and lower/less frequent expression of Napsin-A, p504S, and HNF1β. We conclude that clear-cell change is seen in 14% of ESCs and is discernible at low magnification in only 5%; these areas show an immunophenotype that is essentially identical to the associated background conventional ESC and are phenotypically dissimilar to CCC.
Keywords
Introduction
The diagnostic reproducibility associated with the histotyping of high-grade endometrial carcinomas is no more than moderate,1,2 and the occurrence of clear cells in histotypes other than clear-cell carcinoma (CCC) is likely to be one of the contributing factors to the phenomenon.1,3,4 Endometrial serous carcinoma (ESC) represents approximately 10% of endometrial carcinomas, 5 whereas CCCs are much more uncommon. 6 Although both are considered to be high-risk histotypes,7,8 it is recommended that they be distinguished whenever possible because there may be differences between them regarding patterns of tumor spread, prognosis, risk of subsequent malignancy, and the most optimal therapeutic approach.7-11 For example, investigators from the Canadian high-risk endometrial cancer consortium reported that adjuvant radiation was associated with improved overall survival for CCC, whereas adjuvant chemotherapy was associated with improved overall survival for ESC. 8 Additionally, there is evolving evidence, albeit not entirely congruent, that patients with CCC have a higher risk of thromboembolic events than their counterparts with other high-risk histotypes of endometrial carcinoma.12-14
It is well recognized that a subset of high-grade endometrial carcinoma cannot be readily assigned a histotype.15,16 However, cognizance of the full pathological spectrum for the established histotypes should minimize the erroneous classification as ambiguous those tumors that show variant morphology in one of those histotypes. In the current study, we aim to determine the frequency with which clear cells are identified in otherwise classic cases of ESC. We also aim to determine whether the clear-cell areas are closer in phenotype to the background, non–clear-cell ESC areas (which would suggest that they represent a morphological variant) than they are to CCC (which would suggest that they represent an area of histotypical ambiguity or a CCC component).
Materials and Methods
Case Selection
The institutional database at University of California San Diego was queried for cases of ESC. After review, the final data set comprised 56 cases of ESC, classified as such using applicable criteria of the World Health Organization. 17 The review excluded any cases that the authors considered to be mixed ESC/CCC, were histotypically ambiguous, or that were otherwise considered to be a non-ESC tumor based on the same criteria. 17 All slides for each case were reviewed (range = 7-59 slides, average 17) at low (×20, 1.1 mm field diameter) and high (×40, 0.55 mm field diameter) magnifications using an Olympus BX43 microscope (Olympus Corporation, Tokyo, Japan). The presence of any clear cells was noted, as also whether they were discernible at low or high magnification and whether the foci were at intrauterine and/or extrauterine locations. The morphological features of these foci were described in detail. ESC with clear cells formed our analytic group 1.
Immunohistochemistry
Immunohistochemical studies were performed on all cases. For each case with a clear-cell change, we selected another block showing conventional ESC (ie, without clear cells) from the same case. These formed group 2. Randomly selected cases of CCC (group 3) were also selected for comparison, and 4 of the group 3 cases were from a previously reported data set. 18 With one exception, all were performed on 4 µm-thick, unstained slides of formalin-fixed, paraffin-embedded target tissue, using the Ventana Benchmark automation and the Ultra View detection kit (Ventana Medical Systems, Tucson, AZ) per manufacturer protocol. The following primary antibodies were used: estrogen receptor (ER; Clone SP1; prediluted, Ventana), progesterone receptor (PR; clone IE2; prediluted; Ventana), p53 (clone DO-7; dilution 1:40; Ventana), polyclonal Napsin-A (Cell Marque, Rocklin, CA), α-methylacyl-CoA racemase (P504S or AMACR; polyclonal; prediluted; Biocare Medical, Concord, CA), and Ki67 (clone MIB1; dilution 1:150; Dako, Carpinteria, CA). Hepatocyte nuclear factor 1β (HNF1β; dilution 1:200) was performed on the Leica Bond Max autostainer (Leica Microsystems, Buffalo Grove, IL), using heat-induced antigen retrieval and the Leica Epitope Retrieval 2 solution for 30 minutes. The Bond Polymer Refine detection system was used for visualization. p53, HNF1β, Napsin-A, Ki67, p504S, ER, and PR were performed on all cases. Additionally, Wilms tumor 1 protein (WT-1; Clone 6F-H2; dilution 1:100; Dako) was performed on the extrauterine tumors in 2 cases. Expected signal localization was nuclear for HNF1β, WT-1, Ki67, ER, p53, and PR and was cytoplasmic for Napsin-A and p504S. For each case and area of interest, staining intensity (on a 0 to 3+ scale) was multiplied by the extent and distribution of immunoreactivity (on a 0 to 4+ scale) for potential scores that ranged from 0 to a maximum of 12. For the latter scale (extent and distribution of staining), 0, 1+, 2+, 3+, and 4+ scores were assigned for cases with 0%, 1% to 33%, 34% to 66%, and 67% to 100% of tumor cells being immunoreactive, respectively. Ki67 was scored differently: for each case, the final score that was used represented the percentage of positive nuclei within 3 high-powered fields (×40 magnification), inclusive of 1 hot spot and 2 other randomly selected fields. Proportional extrapolations were performed as was deemed necessary for small foci. This represents a significant modification of recently reported criteria. 19 p53 was scored as aberrant or wild type. A case was classified as p53 aberrant if there was strong nuclear staining in at least 75% of the target cells or if there was complete absence of staining (null phenotype) in the context of wild-type staining in background (nonneoplastic) tissues. 17 Any staining pattern other than the aforementioned aberrant pattern was classified as wild type. 17
Statistics
The groups were compared with regard to the various immunohistochemical scores using the Mann-Whitney U and Student t tests for continuous variables and Fisher exact test for categorical variables. All tests were performed using the computational program at http://vassarstats.net/. A 2-tailed P value <.05 was considered to be statistically significant in all analyses.
This study was approved by the Human Research Protections Program at the University of California San Diego (Project Number: 161379CX).
Results
General Features
Of 56 cases, 8 (14%) had clear cells and accordingly formed group 1. The group 1 patients ranged in age from 62 to 83 years (mean = 72 years), which was not significantly different from the remaining 48 patients (mean age = 67.2 years; P = .64). None of the group 1 patients received neoadjuvant chemotherapy or radiotherapy. One patient received hormone replacement therapy more than 5 years before her diagnosis of endometrial cancer; this history was not available in the remaining patients. All 8 patients were surgically staged, including a total hysterectomy, bilateral salpingo-oophorectomy, regional lymph node dissection, peritoneal washings, and in 6 patients, omentectomy. Of the 8 patients, 3 (37.5%) showed extrauterine disease (1 involved the fallopian tube and ovary; 1 involved the omentum, bilateral ovaries, and pelvic lymph nodes; and 1 involved the omentum, bilateral adnexa, and pelvic and paraaortic lymph nodes), which was not significantly different than the rate of advanced stage disease (50% [24/48]) in patients whose ESCs were devoid of clear cells (P = .71).
Morphological Features
Morphological analyses were carried out on resection (rather than sampling) specimens for all cases. The clear cells were discernible at low magnification (×20) in 3 (37.5%) of the 8 group 1 cases and required a high magnification (×40) review to conclusively discern in the others (Figures 1 and 2). The clear cells represented 1% to 20% (median 7.5) of overall tumoral volume. In 1 case, clear cells were restricted to foci of serous endometrial intraepithelial carcinoma (serous EIC; Figures 1E and 1F). Foci of serous EIC without clear-cell change were present in 4 additional cases. Clear cells were present at extrauterine tumoral deposits in 2 of the 3 cases with advanced stage disease. In both, the extrauterine tumors showed a higher proportion of clear cells than their intrauterine counterparts (Figure 3). In all cases, the architectural patterns in the clear-cell areas were solid, papillary, and glandular and were generally identical to the background (nonclear) areas of the same case. The interface between the clear and nonclear areas was abrupt in 4 cases and gradual in the others. The average mitotic index of the group 1 cases (26.25 mitotic figures [MF] per 10 high-power fields [HPF]) was similar to that of the group 2 cases (27.125 MF/10 HPF; P = .875) and was notably higher than that of the group 3 cases (2.375 MF/10 HPF; P < .0001). Other morphological differences between the clear-cell areas and the background ESC included a greater prominence of cell membranes (8/8) and less-prominent nucleolomegaly (4/8) in the former. The clear-cell foci and the background tumors were otherwise morphologically identical.
Two cases with clear cells that were discernible only at high magnification: Morphology of ESC (A and B) with focal cytoplasmic clarity (arrow in B). C. In this case, Napsin-A expression was mostly restricted to clear-cell areas. D. HNF1β expression was present in the area, inclusive of clear and non–clear cells. Focal cytoplasmic clarity (arrow) in an area of serous endometrial intraepithelial carcinoma (E and F).
An example of ESC with clear cells that were discernible at low magnification (2A-2F); Figures 2A to 2C show morphological features; notice the well-defined cell membranes, the transitions between the clear-cell and non–clear-cell areas, and the maintenance of the serous-like architecture, even in the clear-cell areas. D. Some diminution of PR expression in clear-cell areas (arrows). E. p53 Overexpression in clear-cell areas (arrows) as well as background. F. No expression of Napsin-A in clear-cell areas (arrow).
An example of ESC with prominent clear-cell change at an extrauterine site. Fallopian tube showing tumor involvement in an intraepithelial carcinoma-like pattern (A and B). Tumor cells show p53 overexpression (C) and lack of expression of PR (D), Napsin-A (E), and WT-1 (F).
Immunophenotypical Features
Immunophenotypical features are outlined in Tables 1 and 2. Groups 1 and 2 showed no significant differences regarding p53, ER, PR, Napsin A, p504S, and HNF1β expression or Ki67 proliferation rate. In contrast, the group 1 cases showed an immunoprofile that was notably different from that of the group 3 cases, with the following respective average immunohistochemical scores for groups 1 and 3: ER (4.125 vs 1.125, P = .054); PR (1 vs 0.5, P = .415); Napsin-A (0.25 vs 7, P = .00015); p504S (0.75 vs 7, P = .002); HNF1β (1 vs 9.25, P < .0001); The average Ki67 (72.5 in group 1 vs 42.5 in group 3, P = .004) and p53 aberrant (100% in group 1 vs 12.5% in group 3, P = .0014) rates were similarly disparate between these 2 groups. The significance (or lack thereof) associated with the aforementioned P values were generally concordant with our morphological impression for all markers except PR. For PR, our morphological impression was that there was diminution of PR expression in the clear-cell areas of ESC. Indeed, the average PR score for the clear-cell areas (1) was only half of the scores for the non–clear-cell areas of the same case (2), although the difference was not found to be statistically significant (P = .277). For the 2 cases with clear cells at an extrauterine site, these deposits were essentially identical in phenotype to their intrauterine counterparts. WT-1 was negative in both cases.
Raw Immunohistochemical Scores.
Abbreviations: PR, progesterone receptor; ER, estrogen receptor; HNF1β, hepatocyte nuclear factor 1β.
Comparison of the 3 Groups.
Abbreviations: ER, progesterone receptor; PR, progesterone receptor; HNF1β, hepatocyte nuclear factor 1β; MI, mitotic index; MF, mitotic figures; HPF, high-power fields.
Discussion
The morphological spectrum of ESC has undergone a relatively modest level of expansion in the approximately 4 decades since original reports of it appeared in the literature.20-22 Most prominent among these are the recognition of solid 23 and glandular 24 morphological variations, a distinctive “gaping gland” pattern of myometrial invasion,17,25 and the description of an intraepithelial form (serous EIC) that has been postulated to represent the immediate precursor lesion to the conventionally invasive malignancy.26-29 Clear-cell change appears to be within the morphological spectrum of ESC. In their report on a series of endometrial carcinomas with clear cells, Han et al 4 included cases that they classified as ESC. In our own series, similar cases were included. 3 In 1 recent review, Gatius and Matias-Guiu 25 asserted that clear-cell change may be encountered in a significant subset of ESCs, although the current study is the first formal examination of the issue.
In this study, we aimed to define the frequency of clear-cell change in ESCs and to compare the immunophenotypical features of ESC-associated clear cells with those of conventional ESCs and CCCs. We found that 14% of ESCs show clear cells, although only 5% were discernible at low magnification. The cases with clear-cell change showed no significant differences from the rest of the cohort (ie, ESC without clear-cell change) regarding patient age and stage distribution. Furthermore, the clear cells showed an immunophenotype and proliferation rate that was essentially identical to the background ESC (without clear cells) and that was very dissimilar to the immunophenotype displayed by CCCs. Some relative decrease in PR expression was observed in the clear-cell areas (Figure 2D), although the average difference between them and the background ESC within the same case was not found to be statistically significant. Diminution of hormone receptor expression in foci of clear-cell change has previously been reported in endometrioid carcinomas, 30 and the same phenomenon may be occurring in ESCs. Our findings affirmed the nonspecificity of HNF1β for CCCs 31 because clear-cell areas in 4 of 8 ESC cases showed some expression HNF1β. Napsin-A expression proved to be more specific for CCCs, 32 but it was also expressed in the clear-cell areas of ESCs in 1 of 8 cases. For both HNF1β and Napsin-A, the pattern of expression was relatively restricted (focal or patchy), which is in contrast to the patterns of expression for these markers that were observed in CCCs (Figures 1C and 1D).
The diagnostic distinction between ESCs and CCCs may occasionally be problematic. Both histotypes may display hobnail cells; psammoma bodies; stromal hyalinization; solid, glandular, and papillary architectures; eosinophilic cells; foci of nuclear anaplasia; and as the current analysis demonstrated, clear cells. There have been reported analyses of “CCC with serous features” to assess cases whose features are at this interface, 33 and molecular analyses of mixed ESC/CCC suggests that at least some of these tumors are biologically ESC with a component that morphologically mimics CCC. 34 Indeed, the distinction between CCC and ESC was specifically recognized in 1 analysis as a problematic area that contributes to suboptimal reproducibility in the histotyping of high-grade endometrial carcinomas by pathologists. 1 In the current analysis, our original analytic data set was reviewed such that the final case list only included fairly prototypical examples of ESCs. However, this potential for diagnostic variability is an intrinsic limitation to all studies of ESCs, past and present.
The underlying basis for the clear-cell change in ESCs is unclear. In endometrioid carcinomas with secretory change, the latter is thought to be related to a transient response to endogenous or exogenous progestins in a subset of cases. Accordingly, secretory changes may be present in the biopsy and not in the resection. However, in general, in only a small minority of secretory carcinomas can the findings be plausibly related to endogenous or exogenous hormones.35,36 In our group 1 data set, 1 of the 8 patients received hormone replacement therapy years before her presentation. There was no information on whether the remaining 7 patients took any hormonal regimens years before presentation, but none of the 8 patients was on any such regimen in the period that immediately preceded their presentation. Ultimately, numerous medications, including nonhormonal regimens, may theoretically be associated with clear-cell change in a hormone-responsive organ. For example, antipsychotic and antihypertensive medications have been associated with clear-cell and lactational-like changes in the breast in nonlactating, nonpregnant women. 37 Although the underlying causes of the clear-cell change are unknown in ESCs, the frequency with which they are identifiable indicates that it is best considered to be within the morphological spectrum of the histotype. A much larger data set would be required to conclusively assert that the clear change observed herein is entirely devoid of any significance. However, comparable findings in ovarian high-grade serous carcinoma suggests that clear-cell change is simply a morphological variation that is devoid of prominent significance. 38
In summary, we have defined the frequency and immunophenotype of clear cells in ESCs. Our findings suggest that clear-cell change in a tumor whose other cytological and architectural features are consistent with ESC is not a valid reason to classify the tumor as displaying a CCC component or to report it descriptively as histotypically ambiguous. Consistent with World Health Organization criteria, any areas of putative CCC should display a set of pathological features—architectural, cytological, immunophenotypical—that if viewed in isolation, would be fully diagnostic of that histotype. 18 Minor clear-cell changes are within the morphological spectrum of ESC.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.