Department of Pathology and Laboratory Medicine, David Geffen School of Medicine, University of California at Los Angeles, Los Angeles, California
Department of Pathology and Laboratory Medicine, David Geffen School of Medicine at UCLA, University of California, Los Angeles, 10833 Le Conte Ave, Center for the Health Sciences, Box 951747, Los Angeles, CA 90095-1747
Endometrial cancer (EC) is a common malignancy worldwide. It is often preceded by endometrial hyperplasia, whose management and risk of neoplastic progression vary. Previously, the authors have shown that the tetraspan protein epithelial membrane protein-2 (EMP2) is a prognostic indicator for EC aggressiveness and survival. Here the authors validate the expression of EMP2 in EC, and further examine whether EMP2 expression within preneoplastic lesions is an early prognostic biomarker for EC development.
A tissue microarray (TMA) was constructed with a wide representation of benign and malignant endometrial samples. The TMA contains a metachronous cohort of cases from individuals who either developed or did not develop EC. Intensity and frequency of EMP2 expression were assessed using immunohistochemistry.
There was a stepwise, statistically significant increase in the average EMP2 expression from benign to hyperplasia to atypia to EC. Furthermore, detailed analysis of EMP2 expression in potentially premalignant cases demonstrated that EMP2 positivity was a strong predictor for EC development.
Endometrial cancer (EC) is the most common gynecologic malignancy diagnosed among women in developed countries.1-4 The incidence of developing EC has steadily increased over the past 20 years, with nearly 1 of every 35 women developing the disease over their lifetimes.3, 5 Often, women initially present with hyperplastic lesions, which are classified by the World Health Organization (WHO) according to the severity of glandular crowding (simple vs complex hyperplasia) along with the presence or absence of nuclear atypia.6, 7 Currently, there is no reliable method to determine which hyperplasia will progress to cancer, due in part to the lack of reliable and consistent pathological indicators of hyperplastic subvariants.8-10 As such, consistently distinguishing low-risk hyperplastic lesions from higher-risk lesions can be difficult.
Conservative management of hyperplastic endometrium includes progesterone therapy along with regular follow-up.6, 11 However, serious side effects are associated with progesterone therapy, such as increased risk of atherogenesis, severe weight gain, and mood changes.12 Moreover, treatments need to be long-term or lesions reappear and progress.6, 7, 13 Although hysterectomy is usually indicated for women with atypical hyperplasia, the management of simple or complex hyperplasia lacks consensus.6, 7 Moreover, the incidence of both endometrial hyperplasia and EC in women of childbearing age creates additional challenges in diagnostic and therapeutic strategies.14, 15 Therefore, effective algorithms that identify the risk of EC development within hyperplastic endometrium would not only facilitate early cancer detection, but also guide proper and effective treatment.
The tetraspan protein epithelial membrane protein-2 (EMP2) has recently been identified as a biomarker that can help predict the probability of EC progression as well as patient survival. Multivariate analysis revealed that higher EMP2 expression levels were a predictor for more aggressive tumor behavior, including higher stage and greater vascular invasion.16 EC with higher levels of EMP2 was also more likely to recur postsurgery and thus be more deadly.16
EMP2 is involved in regulating plasma membrane localization and signaling of several classes of membrane receptors involved in cell-cell interaction or recognition of the extracellular milieu.17-21 For example, EMP2 levels alter the localization and surface expression of specific integrin isoforms modulating such diverse processes as αvβ3 integrin-mediated blastocyst implantation or β1 integrins-focal adhesion kinase–mediated extracellular matrix adhesion, contraction, and invasiveness.17, 18, 20 Notably, the centrality of these integrin-mediated functions in the malignant traits of neoplastic cells suggests a potential biologic basis for the association of EMP2 expression with aggressive endometrial cancer.17, 22
The nexus of EMP2 expression with endometrial cancer raised the possibility that EMP2 might identify a subset of hyperplasic lesions destined for EC development. Accordingly, the present study was designed to 1) validate in an independent patient population our previous findings associating EMP2 expression within a subset of EC; and 2) determine whether EMP2 was an early predictive marker in hyperplastic lesions for EC development. To study this, we constructed a tissue microarray (TMA) from metachronous endometrial samples of 207 female patients (aged 30-85 years at initial diagnosis) who were seen at the University of California at Los Angeles (UCLA) Medical Center. Significantly, this study confirms the association of EMP2 expression with aggressive EC. Moreover, it reveals that increased expression of EMP2 in hyperplastic endometrium is a strong prognostic indicator of future EC development.
MATERIALS AND METHODS
Endometrial Tissue Microarray
The TMA was constructed as previously described.16 Archival formalin-fixed, paraffin-embedded endometrial tissue samples were obtained from the Department of Pathology and Laboratory Medicine at the David Geffen School of Medicine at UCLA with institutional review board approval. The TMA was designed to have a wide representation of histopathologies of the endometrium as well as to represent tissue samples from biopsies or hysterectomies from individuals over time (metachronous) who either developed or did not develop EC. The metachronous progression TMA consisted of 535 surgical cases from 207 patients (Fig. 1). The following describes the breakdown of patient cases on the TMA. Note that for each case, there were approximately 3 TMA spots from each histopathology (discussed below).
Cases were selected from patients who underwent endometrial sampling through biopsy, curettage, and/or resection procedures from 1982 to 2002. The case histologies were: 1) benign endometrium (n = 231) (consisting of a composite of atrophic, weakly proliferative, menstrual, proliferative, secretory, progestational effect because of hormone therapy, disordered proliferative, and polypoid endometrium); 2) simple hyperplasia (n = 105); 3) complex hyperplasia (n = 36); 4) simple atypical hyperplasia (n = 10); 5) complex atypical hyperplasia (n = 44); and 6) primary endometrial adenocarcinoma (n = 109). Hyperplasias with simple, complex, simple atypical, or complex atypical morphology are collectively referred to in this paper as potential preneoplastic lesions. The adenocarcinomas were limited to endometrioid types, as this represents the majority of type I endometrial cancers,23 as well as related variants, including ciliated, secretory, papillary (villoglandular), adenoacanthoma, and adenosquamous. Tumor typing follows the WHO classification.24 In total, the TMA contained 535 surgical cases from 207 patients. Of these, 46 patients had metachronous samples and had disease progression with >1 year follow-up (Fig. 1). Table 1 summarizes clinical variables and patient groups as well as EMP2-positive and-negative samples in each subgroup. Table 1 also summarizes whether there is a statistically significant difference in EMP2 expression for each of the clinical variables.
Table 1. Clinical Variables and Patient Groups
All Patients (%)
EMP2 Negative (% of Total)
EMP2 Positive (% of Total)
BMI indicates body mass index; EC, endometrial cancer.
Each relevant histology in a given case was typically represented by 3 cores (or spots) on the TMA with a diameter of 1.0 mm. Therefore, the total number of spots/cores in the TMA was 2025. The spot histologies of the endometrial array were assigned based on a review of the hematoxylin and eosin slides and pathology reports. The per spot breakdown by core histology includes: benign (n = 927), simple hyperplasia (n = 348), complex hyperplasia (n = 126), simple atypical hyperplasia (n = 70), complex atypical hyperplasia (n = 216), and primary endometrial adenocarcinoma (n = 338).
Disease progression cases
The TMA was also designed to include tissue samples for disease progression. In this regard, 75% percent of these patients (n = 150) had multiple, metachronous samples taken over time, giving a total of 457 cases. Figure 1 summarizes the criteria for inclusion or exclusion for the studies of disease progression. Patients were excluded for the following reasons: 1) if they ultimately developed a nonendometrioid carcinoma in follow-up, 2) if they had no follow-up tissue after the initial sample collection, and 3) if their only follow-up time point was <1 year after the initial sample acquisition. Cases were also excluded if during array construction the appropriate histopathology was not accurately covered (n = 50). In addition, spots were excluded if they became physically dislodged, if they were completely depleted from the TMA block, or if after sectioning they no longer contained relevant cell types (ie, epithelium or tumor; 5 cases). In Figure 1, these categories are referred to as being excluded because of “lack of EMP2 informative information” from the TMA. Disease progression criteria has previously been defined.6, 25 In particular, Lacey et al. described the high risk of carcinoma development in women who have atypical hyperplasia (AH), in contrast to nonatypical hyperplasias.6, 25 Thus, disease progression was considered to be development of endometrial carcinoma or AH.
The TMA was stained with human EMP2 antisera or a preimmune control as previously described.16 Briefly, for antigen retrieval, sections were incubated at 95°C for 20 minutes in 0.1 M citrate, pH 6.0. EMP2 was detected using a primary polyclonal rabbit antihuman polyclonal EMP2 antiserum at a dilution of 1:40016 followed by visualization using the Vector ABC kit (Vector Labs, Burlingame, Calif) according to the manufacturer's instructions. Negative controls included the corresponding amount of preimmune sera. Specificity of this reagent is well established as previously described and has been shown to be titratable, inhibitable by EMP2 protein or peptide, and reactive against tissue or cells shown to express EMP2 by Western blot or reverse transcriptase polymerase chain reaction (RT-PCR), and not to be reactive against tissues or cells shown not to express EMP2 by Western blot or RT-PCR.16-19, 26
We conducted a semiquantitative analysis of the EMP2-stained endometrial progression TMA by a pathologist (R.A.S.) who was blinded to clinical information. A second pathologist (O.H.) spot-checked the data. The percentage of cells with staining intensities of 0 to 3 (0 = below the level of detection, 1 = weak, 2 = moderate, 3 = strong) was scored.
Western Blot Analysis
Frozen tissue samples of surgically removed specimens were obtained from the UCLA Department of Pathology and Laboratory Medicine under institutional review board approval. Endometrial samples had histologies of either normal (proliferative endometrium), hyperplasia (1 sample simple hyperplasia, 1 sample complex hyperplasia), or carcinoma. Tissue samples were homogenized and resuspended in Laemmli buffer. Samples were treated for 2 hours at 37°C with PNGase (New England Biolabs, Beverly, Mass) to remove N-linked glycans.17 Proteins were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis and transferred to a nitrocellulose membrane (Amersham Pharmacia, Piscataway, NJ). Membranes were blocked with 10% low-fat milk in phosphate-buffered saline + 0.1% Tween-20 and probed with antihuman EMP2 antisera. Protein bands were observed using an antirabbit horseradish peroxidase-labeled secondary antibody (Southern Biotechnology Associates, Birmingham, Ala) followed by enhanced chemiluminescence detection reagents (Amersham Pharmacia).
Statistical analyses were performed as previously described using R (http://www.R-project.org) and StatView software (Adept Scientific, Acton, Mass)27-29 (S.H.). To examine the differences in EMP2 expression between different histologies, the Kruskal-Wallis test and Spearman correlation were used. The Mann-Whitney U test was used to examine differences in EMP2 expression related to tumor development for each histology, as well as to compare continuous, clinical variables to levels of EMP2 expression. To arrive at a single expression value across multiple, informative spots within each histology of each case, the mean pooled value for EMP2 was calculated. Pooling criteria for case-based analyses was previously described.27-29 The dependence between categorical variables was tested using the Fisher exact test. Curves estimating the probability of remaining cancer-free were generated using the Kaplan-Meier method, and comparisons were made by using the log-rank test. Hazard ratios and prognostic significance were estimated using the Cox proportional hazards model. Positive EMP2 expression was defined as being >0. For all results, P ≤ .05 was considered significant.
By using an independent TMA patient population, our goals for this present study were to confirm and extend our previously reported association of elevated EMP2 expression with EC16 and to extend these finding to determine whether EMP2 was a early biomarker for EC development and/or progression. Accordingly, we constructed a disease progression (metachronous) TMA that contained surgical cases from 207 patients, of whom 199 patients fulfilled study criteria (see Materials and Methods and Fig. 1).
The cellular and histological expression pattern for EMP2 was identical to those previously observed,16 with representative images shown in Figure 2. Consistent with earlier studies, EMP2 is typically expressed in the cytoplasm and/or at the plasma membrane of the epithelium. In addition, for benign tissue, we observed that EMP2 expression was slightly yet significantly higher in the secretory phase (0.31 ± 0.04) versus the proliferative phase (0.20 ± 0.03) (P = .010). We next semiquantified EMP2 expression for each spot on the TMA and compared these values for each histology. As shown in Figure 3A, on a per-spot basis there was a stepwise increase in relative EMP2 expression from benign to hyperplasia to atypical hyperplasia to EC. This trend was maintained when the spot data were pooled to establish a case-level expression index (Fig. 3B). As validation, similar results were observed when EMP2 protein was detected by Western blot analysis from freshly obtained surgically resected tissue samples obtained from the UCLA Department of Pathology and Laboratory Medicine. Representative samples are shown in Figure 3C, and display an increased expression of EMP2 in EC compared with morphologically normal or hyperplastic tissue. Importantly, these results are consistent with our previous findings using an independent patient population at the Memorial Sloan-Kettering Cancer Center.16
However, despite this general trend, there was some degree of heterogeneity of EMP2 expression in each histological group; some individuals displayed higher EMP2 levels, whereas there was no detectable EMP2 in others. We therefore examined whether EMP2 expression in benign or potentially preneoplastic lesions provided information regarding future tumor development. To start addressing this issue, we considered EMP2 expression in TMA samples from metachronous individuals (ie, those who had had multiple surgical visits to the UCLA Medical Center). Included were samples of benign endometrium (including secretory, proliferative, and menstrual), potential preneoplastic lesions (simple, complex, and atypical hyperplasia), and malignant endometrium. We considered patients who had an initial diagnosis of benign or potentially preneoplastic lesions, who then subsequently did or did not develop EC. As shown in Figure 3D, there was a striking although not statistically significant increase in EMP2 expression in potentially preneoplastic lesions in patients who ultimately developed EC after ≥1 year compared with those who did not develop malignancy. In contrast, the EMP2 level in benign tissue did not correlate with tumor development (Fig. 3D).
On the basis of these results, we considered in more detail the EMP2 expression in potentially preneoplastic lesions in patients from whom there was available endometrial tissue with at least 1 year follow-up. We dichotomized the population into those who had EMP2-positive lesions versus those that had no detectable EMP2 (EMP2 negative). Significantly, if EMP2 was positive in the potentially preneoplastic lesion, there was a much greater probability of endometrial cancer development than if EMP2 was negative (Fig. 4; P = .05). This trend also held when EMP2 was analyzed as a continuous variable (hazard ratio, 3.64; 95% confidence interval, 1.16-11.4, P = .03). Therefore, based on these results, EMP2 expression appeared to be an early predictor of EC development. Interestingly, EMP2 positivity was significantly, albeit slightly, associated with the age of the patient at the time of first biopsy, and its expression associated with an early onset of menopause (Table 1). Additional studies will be needed to examine whether assessment of EMP2 expression levels will be most valuable in relatively younger individuals. No other clinical variables were considered to be associated with EMP2 expression levels (Table 1).
An interesting yet potentially complicating variable in the study was the prevalent use of hormones (progestin alone and/or estrogen + progesterone) in approximately 84% of the individuals studied. However, the presence or absence of hormone supplements or therapy showed no apparent correlation with EMP2 levels or the predictive strength of EMP2 positivity for EC development. Specifically, 15 patients had received progesterone only, 23 had received progesterone + estrogen, 2 had had estrogen alone, and 7 individuals had received no therapy or tamoxifen. Of the 38 women on progesterone alone or estrogen + progesterone, 21 women had positive expression of EMP2 (ie, increased risk of tumor formation), and 17 were EMP2 negative (ie, no increased risk of tumor formation). There was no statistical difference between these groups. Furthermore, the widespread use of hormone therapy within this patient population confounded the predictive ability of clinical variables (ie, estrogen receptor, progesterone receptor) commonly associated with endometrial cancer (Table 1).
In this study, we have identified EMP2 as an early marker for EC development. The increased EMP2 expression in EC observed in an earlier TMA was validated using an independent patient cohort.16 Moreover, our results showed that EMP2 positivity within lesions that are potentially preneoplastic (ie, simple, complex and atypical hyperplasias) strongly predicts the future development of EC.
The importance of identifying potential early detection markers for EC cannot be overemphasized. EC is the most common gynecologic malignancy in the United States and other developed countries.1-4 Although survival for this disease has been increasing, current treatment or surgical options can be severe, particularly for women of childbearing age. Moreover, the WHO classification of endometrial hyperplasia, complicated by the lack of interobserver reproducibility, provides only a limited ability to discern which hyperplastic lesions will undergo malignant progression.7-9 Attempts have been made to devise algorithms based on histology and other clinical parameters to define which lesions will progress, but such strategies are only effective with atypical hyperplasia.6 Furthermore, although several studies have suggested that the PTEN/AKT pathway may play a role in endometrial cancer progression,25, 30 to date there are no molecular biomarkers that can consistently predict the tumor potential in hyperplastic endometrium.25, 31 Therefore, EMP2 holds promise as the first predictor of tumor development in preneoplastic endometrium. Presently, we are expanding these studies to encompass a larger-scale population trial to further assess the clinical utility of EMP2 expression.
Whether EMP2 plays a causal role in the transition of hyperplasia to malignancy, or in the pathogenesis of EC, remains to be determined. Nevertheless, what is known about the cell biology of EMP2 provides a conceptual framework in which to formulate a mechanistic role. EMP2 is thought to regulate the transport of specific proteins.17, 18, 21 The cohort of proteins with which EMP2 interacts or regulates in this fashion appears to be tissue-specific, but frequently involves 1 or more integrin family members.17, 18, 21 In the endometrium, EMP2 associates with both integrin αvβ3 and focal adhesion kinase (FAK), and can regulate αvβ3 integrin expression and localization. Given the importance of integrins and FAK in adhesion, proliferation, and migration,32, 33 dysregulation of EMP2 is predicted to have profound cellular consequences. For example, there is extensive evidence that integrin clustering and FAK activation can activate ERK/MAP kinase and AKT pathways, while down-regulating p53 mediated apoptosis.32-34 In epithelial cells, β3 integrin activation directly leads to Src and FAK phosphorylation, resulting in stable focal adhesions.35, 36 Analysis of EMP2-β3 integrin-FAK association suggests that in the endometrium, EMP2 expression promotes FAK and Src phosphorylation (data not shown). We therefore hypothesize that EMP2 may help to stabilize or promote integrin-FAK-Src complexes. We are currently examining the consequences of αvβ3 integrin, FAK, and EMP2 association in both cell culture and TMA platform systems. It is interesting to note that both FAK and Src phosphorylation have been implicated in endometrial cancer invasion.37, 38
In conclusion, the results from this and prior studies point to the potential of EMP2 as not only an indicator of EC progression, but also as an early detection marker in preneoplastic lesions. Therefore, these results strongly suggest that EMP2 has a potential utility in the early diagnosis of EC. Whether EMP2 or associated genes such as FAK or Src might provide beneficial therapeutic targets remains an interesting possibility.
CONFLICT OF INTEREST DISCLOSURES
This work was supported in part by National Institutes of Health grants T32 CA 009056-32 (O.H.), HD48540 (J.B.), and CA131756 (M.W.), the Early Detection Research Network NCI CA-86,366 (L.G.), the Iris Cantor Seed Grant (M.W.), and the UCLA Jonsson Comprehensive Cancer Center (J.B.).