PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
 
Cancer Biol Ther. Author manuscript; available in PMC 2010 May 18.
Published in final edited form as:
Cancer Biol Ther. 2009 November; 8(22): 2126–2135.
Published online 2009 November 5.
PMCID: PMC2872176
NIHMSID: NIHMS198250

Molecular clustering of endometrial carcinoma based on estrogen-induced gene expression

Abstract

Identification of biomarkers potentially provides prognostic information that can help guide clinical decision-making. Given the relationship between estrogen exposure and endometrial cancer, especially low grade endometrioid carcinoma, we hypothesized that high expression of genes induced by estrogen would identify low risk endometrioid endometrial cancers. cDNA microarray and qRT-PCR verification were used to identify six genes that are highly induced by estrogen in the endometrium. These estrogen-induced biomarkers were quantified in 72 endometrial carcinomas by qRT-PCR. Unsupervised cluster analysis was performed, with expression data correlated to tumor characteristics. Time to recurrence by cluster was analyzed using the Kaplan-Meier method. A receiver operating characteristic (ROC) curve was generated to determine the potential clinical utility of the biomarker panel to predict prognosis. Expression of all genes was higher in endometrioid carcinomas compared to non-endometrioid carcinomas. Unsupervised cluster analysis revealed two distinct groups based on gene expression. The high expression cluster was characterized by lower age, higher BMI, and low grade endometrioid histology. The low expression cluster had a recurrence rate 4.35 times higher than the high expression cluster. ROC analysis allowed for the prediction of stage and grade with a false negative rate of 4.8% based on level of gene expression in endometrioid tumors. We have therefore identified a panel of estrogen-induced genes that have potential utility in predicting endometrial cancer stage and recurrence risk. This proof-of-concept study demonstrates that biomarker analysis may play a role in clinical decision making for the therapy of women with endometrial cancer.

Keywords: endometrial cancer, biomarkers, estrogen, clinical decision making, tumor recurrence

Introduction

While the overall incidence of cancer is falling, the incidence of endometrial cancer has remained relatively unchanged. It is the fourth most common cancer in women in the United States. The American Cancer Society estimated that there were 40,100 new diagnoses and 7,470 deaths from endometrial cancer in 2008.1

The role of estrogen in endometrial carcinogenesis has been well studied. Hyperestrogenic states, such as obesity, anovulation, early menarche, late menopause, nulliparity, and unopposed estrogen treatment, are established risk factors for the development of endometrial carcinoma, especially endometrioid type.24 Estrogen excess is associated with histologic changes in the endometrium, including atypical endometrial hyperplasia, which are associated with the development of endometrial carcinoma of the endometrioid type.5 Currently, it is thought that estrogen stimulates excessive cellular proliferation which leads to accumulation of genetic and epigenetic changes that disrupt and alter normal cellular functions.6 It is likely that an imbalance of growth stimulatory and inhibitory factors in the endometrium leads to the development of endometrial cancer.

Clinically, it is crucial to distinguish endometrioid carcinomas (related to estrogen excess) from non-endometrioid carcinomas (not related to estrogen excess). Non-endometrioid carcinomas are typically associated with advanced stage at diagnosis and poor prognosis compared to low grade endometrioid carcinoma. Therefore, an endometrial biopsy diagnosis of non-endometrioid endometrial carcinoma typically triggers a more aggressive treatment plan, including total abdominal hysterectomy, bilateral salpingo-oophorectomy, staging lymphadenectomy and post-surgical radiation therapy and/or chemotherapy. However, not all endometrioid type carcinomas are clinically indolent. It is well-known that a subset of women with endometrioid carcinoma will have lymph node metastases at the time of hysterectomy or experience recurrence following surgery. Women with these aggressive endometrioid carcinomas would likely benefit from more aggressive surgical staging and adjuvant treatment, similar to what is recommended for patients with non-endometrioid carcinomas. Unfortunately, at this time there is no reliable clinical or pathological parameter that allows for a priori identification of these high risk endometrioid carcinomas.

In this paper, we demonstrate that genes from a variety of different growth regulatory pathways, specifically components of retinoic acid synthesis, Wnt signaling and insulin-like growth factor (IGF) pathways, are induced by estrogen in the human endometrium and that expression of these genes is dysregulated in endometrial carcinoma. We hypothesize that these estrogen-induced genes will demonstrate higher expression in the low grade endometrioid carcinomas. In addition, as a proof-of-principle study, we determined the potential utility of using a panel of these estrogen-induced genes to distinguish low risk endometrioid carcinomas from high risk endometrioid carcinomas.

Results

Candidate estrogen-induced genes examined in endometrial cancer

Our previous studies had identified EIG121 and RALDH2 from a validated microarray analysis as being highly induced by estrogen in the endometrium of post-menopausal women.7,9 RALDH2 is the rate-limiting enzyme in retinoic acid synthesis. Retinoic acid is well-known to inhibit proliferation in the uterus.10,11 EIG121 is a highly conserved gene of unknown function. Here, we demonstrate that from this same microarray, two components of the Wnt signaling pathway, sFRP1 and sFRP4, and two components of IGF signaling, IGF-I and IGF-IR, are upregulated by estrogen (Fig. 1). sFRP's modulate Wnt signaling,12 and we have recently shown that sFRP4 inhibits Wnt signaling in endometrial epithelial cells.13 IGF-I and IGF-IR are known to stimulate proliferation in the uterus.14 Note that estrogen stimulation of normal post-menopausal endometrium results in the induction of both pro-proliferative (IGF-I, IGF-IR) and growth inhibitory factors (RALDH2, sFRP1, SFRP4). This pattern of gene induction by estrogen likely plays a major role in the controlled, balanced endometrial proliferation normally observed following stimulation by estrogen.

Figure 1
Expression of estrogen-induced genes, RALDH2 (A) sFRP1 (B), sFRP4 (C), IGF-I (D) and IGF-IR (E) in benign post-menopausal endometrium treated with placebo and estrone, compared to endometrial carcinoma. Data are expressed as medians with bars representing ...

We next quantified these transcripts in a series of endometrial carcinomas (n = 55 endometrioid; n = 14 non-endometrioid; Figs. 1 and and2).2). The clinical characteristics of these endometrial carcinoma patients are summarized in Table 1. Expression of these estrogen-induced genes was lower in the endometrial carcinomas as a whole compared to gene expression in estrogen-stimulated endometrium (Fig. 1). However, interesting patterns were observed when the tumors were sub-divided by grade and histotype (endometrioid vs. non-endometrioid). In general, the expression of the estrogen-induced genes tended to be highest in the well-to-moderately differentiated endometrioid carcinomas (grades 1 and 2), with lower expression observed in the poorly differentiated endometrioid carcinomas (grade 3) and the non-endometrioid carcinomas (Fig. 2). The exception to this trend was sFRP1, which had a similar expression pattern in the endometrioid and non-endometrioid carcinomas (Fig. 2).

Figure 2
Gene expression in the endometrial tumors by histologic grade. In general, gene expression was highest in the grade 1 and grade 2 endometrioid tumors (well- and moderately-differentiated). All values are medians of normalized expression data ×10 ...
Table 1
Characteristics of endometrial cancer patients

Unsupervised cluster analysis based on estrogen-induced gene expression

Unsupervised cluster analysis was performed for each estrogen-induced gene singly and in various combinations (data not shown). Only the unsupervised cluster analysis utilizing all six estrogen-induced genes resulted in a split into two well-separated clusters based on gene expression (agglomerative coefficient = 0.76; Fig. 3A). Cluster 1 had an overall lower gene expression of the six genes. Table 3 summarizes the clinical characteristics associated with each cluster. Cluster 1 demonstrated higher median age at diagnosis (p = 0.029) and lower mean body mass index (p = 0.033). Cluster 1 members were more likely to be of grade 3 endometrioid or non-endometrioid histology (p = 0.001). In addition, cluster 1 had a lower proportion of early stage (stage 1 and 2) patients than cluster 2, although this did not reach statistical significance (p = 0.072). At a median follow up of 33 months, 35% of patients (10/29) in Cluster 1 and 8% of patients (3/36) in Cluster 2 demonstrated recurrence. Thus, Cluster 1 had a recurrence rate 4.35 times higher than that of Cluster 2 (95% CI: 1.21–16.13; Fig. 3B). Interestingly, two of the patients with recurrence in cluster 1 had grade 1 endometrioid carcinomas, the type of tumor that, using standard clinical and pathological parameters, would be classified as low-risk for recurrence.

Figure 3
Using expression data for 6 estrogen-induced genes in the endometrial cancers, unsupervised cluster analysis (A) was performed. Cluster 1 exhibited low expression while cluster 2 exhibited high expression. Normalized median gene expression is quantified ...
Table 3
Patient characteristics by cluster

ROC analysis

The cluster analysis allowed for the separation of our data into two biologically and clinically meaningful groups. We next wanted to examine a composite score of the six estrogen-induced genes within the endometrioid subtype alone to determine if it could be used to differentiate a low risk endometrioid group from a higher risk endometrioid group. An ROC analysis was performed to examine the ability of this score to predict risk level at diagnosis as defined in the methods. ROC analysis allowed for the prediction of risk level (low vs. intermediate/high) with a false negative rate of 4.8% (90% CI: 1.5–12.8%) based on level of gene expression. This relatively low false negative rate is important, as this figure represents the percentage of endometrial cancer patients who would be incorrectly predicted to have low risk disease. This low figure is impressive compared to the false negative rate of 24.5% reported in the literature for intra-operative pathological frozen section analysis, the current “gold standard” of predicting risk.15 Figure 4 demonstrates the estimated ROC curve together with a bivariate 90% confidence interval for true positive fraction (TPF, sensitivity) and false positive fraction FPF, 1-specificity using the cutoff with minimal empirical error rate. The step function shows the empirical ROC curve. The smooth curve shows a fitted ROC curve16 with an area under the curve of 0.95. The cutpoint used for predicting stage was a score of 1.74. In other words, if our linear combination of predictors was less than 1.74, we classified patients as more likely to have low risk disease than if it was 1.74 or more. It is important to note here that this model is proof-of-principle; validation of this model on a separate set of endometrial carcinomas is still necessary.

Figure 4
Estimated ROC curve with a bivariate 90% confidence interval for true positive fraction (TPF) and false positive fraction (FPF). The step function is the empirical ROC curve. The red shaded rectangle is a joint 90% confidence interval for (FPF, TPF) at ...

Discussion

In this study, we used cDNA microarray analysis to identify a panel of genes which were induced by estrogen treatment in the endometrium of postmenopausal women. Given the impact of estrogen exposure on endometrial cancer risk,4,17 we hypothesized that the estrogen-induced genes would be markers for estrogen associated endometrial cancers, typically the well differentiated, grade 1 endometrioid carcinomas. In accordance with our hypothesis, the estrogen-induced genes in our panel were highly expressed in grade 1 endometrioid endometrial adenocarcinoma compared to the expression in grade 3 endometrioid and nonendometrioid tumors.

In addition to our previous studies that have identified RALDH2 and EIG121 as being induced by estrogen in post-menopausal women,7,9 in the present study we demonstrated that estrogen also induced two inhibitors of the Wnt signaling pathway in the endometrium, sFPR1 and sFPR4. The Wnt signaling pathway is involved in cell proliferation, migration, differentiation and apoptosis. The deregulation of this pathway has been implicated in the carcinogenesis of multiple organ sites, including breast, lung and colon.18,19 Secreted-frizzled related proteins (sFRPs) act as molecular antagonists to the ligands in the Wnt pathway.20,21 The expression of sFRP4 in the endometrium varies with the menstrual cycle, indicating the possibility of hormonal regulation.22

While a difference in expression based on the grade of tumor was shown in a majority of the estrogen-induced genes in this panel, there was no single gene which achieved perfect correlation to low and high risk disease. Thus, an unsupervised cluster analysis based on level of gene expression was performed to assess the performance of all six genes. This analysis was successful in partitioning the endometrial carcinomas into two distinct groups. Overall, this analysis is validated by the fact that the clusters were discriminated by traditional characteristics. In general, the cluster with high estrogen-induced gene expression fit the “type I” phenotype. These patients were young, obese and had low grade and low stage endometrioid cancers. Conversely, patients in the low expression cluster had a higher median age, lower body mass index, and tended to have higher grade tumors, non-endometrioid histology and advanced stage.

One potential utility for a biomarker panel in endometrial carcinoma is to guide clinical decision-making. Such “personalized cancer therapy” has recently been employed in estrogen receptor positive, axillary lymph node negative breast cancer. In the past, these women uniformly received adjuvant chemotherapy, despite the fact that only a small proportion developed recurrence. A molecular diagnostics assay, based on quantitative RT-PCR of a small panel of genes, has been identified and correlates to the risk of subsequent recurrence in this population of women. A score is calculated from the quantitative expression level of a panel of genes that can be used to stratify patient risk level and guide the need for adjuvant therapy.23

In endometrial cancer, we envision that this type of molecular assay could be applied to two clinical scenarios—determining the need for comprehensive surgical staging and determining the need for subsequent adjuvant radiotherapy and/or chemotherapy. Current Society of Gynecologic Oncologist guidelines promote the need for complete surgical staging, consisting of systematic pelvic and paraaortic lymphadenectomy, on all patients with early stage endometrial cancer. However, a prospective study of women with clinical stage I and occult stage II disease revealed the rate of lymph node metastasis in this group is only 11%.24

Despite the low rate of lymph node metastasis in early stage disease, retrospective data supported the performance of systematic lymphadenectomy to improve survival in these patients.2527 However, recent data from two large, randomized clinical trials provide further indication that systematic lymphadenectomy may not be necessary in all women with early stage endometrial carcinoma.28,29 The ASTEC trial, which evaluated the use of systematic lymphadenectomy in patients with preoperative stage I endometrial cancer patients, revealed that complete surgical staging did not provide a survival benefit in this population of women.28 Another large, prospective trial from Italy confirmed these results. This trial, which randomized 514 stage I endometrial carcinoma patients to surgical treatment with or without systematic lymphadenectomy, found that lymphadenectomy did not significantly improved recurrence-free or overall survival.29 Thus, a molecular tool for preoperative guidance regarding performance of lymphadenectomy may be beneficial.

Our panel of estrogen-induced genes performed quite well in the discrimination between low and intermediate/high risk endometrial tumors. The optimal cut off level of 1.74 yielded a sensitivity of 95.2% and a specificity of 88.9% for the designation of a tumor as low risk based on gene expression. The resultant false negative rate of 4.8% significantly out-performs the current practice of intra-operative frozen section analysis as reported by multiple institutions.15,30

In summary, we have identified a panel of genes that are upregulated by estrogen and are relevant to low grade, early stage endometrial cancer. This panel of estrogen-regulated genes may be a potential molecular tool to identify low risk patients or determine prognosis. Future studies are necessary to validate this panel in a larger series. In addition, the inclusion of other genes in the panel relevant to metastasis and invasion may be beneficial.

Materials and Methods

Candidate gene identification

To evaluate the effect of estrogen replacement therapy on postmenopausal endometrium, cDNA microarray analysis was performed to identify genes which were induced after estrogen treatment. These methods and results have been previously described.7 Briefly, for the purposes of microarray screening and real-time quantitative PCR verification, two different subsets of normal endometrial biopsies were randomly selected from a large group of healthy post-menopausal women (n = 210) participating in a clinical trial of estrogen replacement therapy. These 210 women were randomly divided into three groups receiving one of the following three treatments: (i) placebo; (ii) conjugated estrogens (2:1, w/w) of estrone sulfate and equilin sulfate (EES) (Wyeth Research, Philadelphia, PA) in the amount found in a 0.625 mg/day Premarin tablet; (iii) Premarin (Wyeth Research, Philadelphia, PA), 0.625 mg/day, for three months under conditions approved by the Human Ethics Committee of Escola Paulista de Medicina Universidade Federal de Sao Paulo, Brazil. Endometrial biopsies were obtained at baseline (pre-treatment) and at the end of the three-month treatment. For the initial microarray screen, RNA derived from 10 endometrial biopsies per treatment group (placebo, EES and Premarin) was pooled into one sample per group. Incyte spotted cDNA microarrays (Incyte Genomics, Palo Alto CA, ~7,000 probes per array) were hybridized using a two-color competitive approach by direct labeling of 20 ug of total RNA with Cy3 or Cy5 UTP. Two arrays were hybridized that compared placebo and Premarin and placebo and EES. Initial analysis was performed using GEMTools (Incyte Genomics). Subsequent validation of transcripts of interest was accomplished by real-time quantitative PCR, using a second subset of endometrial biopsies (n = 10 each from the placebo, synthetic estrogens or Premarin group) that were analyzed individually.

Endometrial cancer specimens

The characteristics of the endometrial carcinoma patients (n = 72) are summarized in Table 1. RNA was extracted from frozen endometrial cancer specimens obtained at the time of hysterectomy at our institution. Diagnoses were confirmed following light microscopic examination of H&E stained slides by a gynecologic pathologist. Poor quality RNA was obtained from 3 of the specimens. Therefore, for transcript quantification by qRT-PCR, n = 69 endometrial carcinomas are reported.

RNA preparation

At the time of collection, tissues were frozen in liquid nitrogen and stored at −80°C. Tissues were homogenized in TriReagent (Molecular Research Center, Cincinnati, OH) and precipitated with isopropanol. The RNA was applied to RNeasy spin columns (QIAGEN, Valencia, CA), eluted and treated with RNase-fee DNase for 30 minutes at 37°C. DNase I was heat-inactivated at 75°C for 10 minutes and samples were stored at −80°C.

Reverse transcription and real-time quantitative PCR analysis

Six specific genes which were greatly induced by estrogen (EIG121, RALDH2, sFRP1, sFRP4, IGF-I and IGF-IR) were selected for analysis. To quantify the expression of these genes, qRT-PCR was performed using the frozen endometrial carcinoma specimens. Forty ng of RNA (10 ng/μL) were reverse transcribed in triplicate on a 7700 format 96-well plate (ISC Bioexpress, Kaysville, UT) in 6 μL of reaction master mix containing 400 nM assay-specific reverse primer, 500 μM deoxynucleotides, Stratascript buffer, and 10 U Stratascript reverse transcriptase (Stratagene, Cedar Creek, TX). The assay of each sample also included a non-amplification control well, which contained all reagents and RNA but was lacking the reverse transcriptase enzyme. The plates were incubated for 30 minutes at 50°C followed by 20 minutes at 72°C in a thermocycler. Afterwards, 40 μL of PCR master mix containing 400 nM assay specific primers, 100 nM assay specific probe made up of a 5' 6-FAM (5-carboxyfluorescein) and a 3' Black Hole Quencher 1 (BHQ1), 5 mM MgCl2, 200 μM deoxynucleotides, PCR buffer and 1.25 U Taq polymerase was added to each well and amplified in an ABI Prism 7700 sequence detection system (Applied Biosystems, Foster City, CA) under the following cycling conditions: 95°C for 1 minute, 40 cycles of 95°C for 12 seconds, and 60°C for 30 seconds. The results were analyzed using SDS 1.9.1 software (Applied Biosystems) with SuperROX (BioSearch, Novato, CA) as a reference dye. Final analyses were normalized to the geometric mean of the transcript levels of β-actin, 36B4 and 18S. Table 2 summarizes the primer and probe sequences and accession numbers for each assay.

Table 2
Probes and primers for real-time PCR assays

Cluster analysis

We performed a cluster analysis to cluster tumor samples based on the expression for EIG121, RALDH2, sFRP1, sFPR4, IGF-I and IGF-IR. We used a log transformation of the gene expression measured by quantitative PCR to obtain better approximation to normal distributions. The data were then standardized to a mean of 0 and a standard deviation of 1 to ensure that no one particular subset of gene was more influential in determining the clusters solely because of the scale. The data were clustered using agglomerative hierarchical techniques. Average linkage was used to determine the difference between clusters, with distance calculated using the Manhattan method. Finally, an agglomerative coefficient was calculated to assess how well the data clustered.

Receiver operating characteristic (ROC) analysis

We performed an ROC analysis to investigate the potential use of the estrogen-induced gene biomarker panel to predict low risk for recurrence within those tumors of endometrioid histology. Low risk was defined as endometrioid histology, grade 1 or 2, and stage less than Ic. Grade 3 endometrioid tumors at stage Ic or higher were considered high risk. All other endometrial tumors were classified as intermediate risk. The test score was computed as a linear combination of the six markers and an additional interaction indicator for an interaction of IGF-IR, IGF-I and RALDH2. The interaction term is motivated by the fit of a classification and regression tree (CART) for stage as a function of the 6 genes.8 The proposed score is a linear combination of the 6 estrogen-induced genes and the interaction indicator.

Statistical analysis

Statistical differences between groups was analyzed using Fisher's Exact, Mann-Whitney, Jonckheere-Terpstra, and Student's t-tests. Time to recurrence by cluster was analyzed using the Kaplan-Meier method. Time to recurrence was defined as the time from date of treatment completion until date of confirmed recurrence. Statistical significance was set at a p value <0.05.

Acknowledgements

NIH T32 Training Grant (S.W., R.L., M.M.). NIH SPORE for Uterine Cancer (S.W., R.B., L.D., A.M., K.L., D.U., M.M., P.M., D.L.). This original research was presented in part at the 97th United States and Canadian Academy of Pathology Annual Meeting, March 1–7, 2008, Denver, CO and at the 40th Society of Gynecologic Oncologists Annual Meeting on Women's Cancer, February 5–8, 2009, San Antonio, TX.

References

1. Jemal A, Siegel R, Ward E, Hao Y, Xu J, Murray T, et al. Cancer statistics, 2008. CA Cancer J Clin. 2008;58:71–96. [PubMed]
2. Amant F, Moerman P, Neven P, Timmerman D, Van Limbergen E, Vergote I. Endometrial cancer. Lancet. 2005;366:491–505. [PubMed]
3. Purdie DM, Green AC. Epidemiology of endometrial cancer. Best Pract Res Clin Obstet Gynaecol. 2001;15:341–54. [PubMed]
4. Rose PG. Endometrial carcinoma. N Engl J Med. 1996;335:640–9. [PubMed]
5. Kurman RJ, Kaminski PF, Norris HJ. The behavior of endometrial hyperplasia. A long-term study of “untreated” hyperplasia in 170 patients. Cancer. 1985;56:403–12. [PubMed]
6. Shang Y. Molecular mechanisms of oestrogen and SERMs in endometrial carcinogenesis. Nat Rev Cancer. 2006;6:360–8. [PubMed]
7. Deng L, Shipley GL, Loose-Mitchell DS, Stancel GM, Broaddus R, Pickar JH, et al. Coordinate regulation of the production and signaling of retinoic acid by estrogen in the human endometrium. J Clin Endocrinol Metab. 2003;88:2157–63. [PubMed]
8. R Development Core Team . A language and environment for statistical computing. R Foundation for Statistical Computing; 2004.
9. Deng L, Broaddus RR, McCampbell A, Shipley GL, Loose DS, Stancel GM, et al. Identification of a novel estrogen-regulated gene, EIG121, induced by hormone replacement therapy and differentially expressed in type I and type II endometrial cancer. Clin Cancer Res. 2005;11:8258–64. [PubMed]
10. Bo WJ, Smith MS. The effect of retinol and retinoic acid on the morphology of the rat uterus. Anat Rec. 1966;156:5–9. [PubMed]
11. Boettger-Tong HL, Stancel GM. Retinoic acid inhibits estrogen-induced uterine stromal and myometrial cell proliferation. Endocrinology. 1995;136:2975–83. [PubMed]
12. Kawano Y, Kypta R. Secreted antagonists of the Wnt signalling pathway. J Cell Sci. 2003;116:2627–34. [PubMed]
13. Carmon KS, Loose DS. Secreted frizzled-related protein 4 regulates two Wnt7a signaling pathways and inhibits proliferation in endometrial cancer cells. Mol Cancer Res. 2008;6:1017–28. [PubMed]
14. Rutanen EM. Insulin-like growth factors in endometrial function. Gynecol Endocrinol. 1998;12:399–406. [PubMed]
15. Frumovitz M, Slomovitz BM, Singh DK, Broaddus RR, Abrams J, Sun CC, et al. Frozen section analyses as predictors of lymphatic spread in patients with early-stage uterine cancer. J Am Coll Surg. 2004;199:388–93. [PubMed]
16. Lloyd CJ. Using smoothed receiver operating characteristic curve to summarize and compare diagnostic systems. J Amer Stat Assoc. 1998;93:1356–64.
17. Hecht JL, Mutter GL. Molecular and pathologic aspects of endometrial carcinogenesis. J Clin Oncol. 2006;24:4783–91. [PubMed]
18. Polakis P. Wnt signaling and cancer. Genes Dev. 2000;14:1837–51. [PubMed]
19. Taipale J, Beachy PA. The Hedgehog and Wnt signalling pathways in cancer. Nature. 2001;411:349–54. [PubMed]
20. Finch PW, He X, Kelley MJ, Uren A, Schaudies RP, Popescu NC, et al. Purification and molecular cloning of a secreted, Frizzled-related antagonist of Wnt action. Proc Natl Acad Sci USA. 1997;94:6770–5. [PubMed]
21. Eaton S. Planar polarization of Drosophila and vertebrate epithelia. Curr Opin Cell Biol. 1997;9:860–6. [PubMed]
22. Abu-Jawdeh G, Comella N, Tomita Y, Brown LF, Tognazzi K, Sokol SY, et al. Differential expression of frpHE: a novel human stromal protein of the secreted frizzled gene family, during the endometrial cycle and malignancy. Lab Invest. 1999;79:439–47. [PubMed]
23. Paik S, Shak S, Tang G, Kim C, Baker J, Cronin M, et al. A multigene assay to predict recurrence of tamoxifen-treated, node-negative breast cancer. N Engl J Med. 2004;351:2817–26. [PubMed]
24. Creasman WT, Morrow CP, Bundy BN, Homesley HD, Graham JE, Heller PB. Surgical pathologic spread patterns of endometrial cancer. A Gynecologic Oncology Group Study. Cancer. 1987;60:2035–41. [PubMed]
25. Mohan DS, Samuels MA, Selim MA, Shalodi AD, Ellis RJ, Samuels JR, et al. Long-term outcomes of therapeutic pelvic lymphadenectomy for stage I endometrial adenocarcinoma. Gynecol Oncol. 1998;70:165–71. [PubMed]
26. Cragun JM, Havrilesky LJ, Calingaert B, Synan I, Secord AA, Soper JT, et al. Retrospective analysis of selective lymphadenectomy in apparent early-stage endometrial cancer. J Clin Oncol. 2005;23:3668–75. [PubMed]
27. Larson DM, Broste SK, Krawisz BR. Surgery without radiotherapy for primary treatment of endometrial cancer. Obstet Gynecol. 1998;91:355–9. [PubMed]
28. ASTECStudy Group. Kitchener H, Swart AM, Qian Q, Amos C, Parmar MK. Efficacy of systematic pelvic lymphadenectomy in endometrial cancer (MRC ASTEC trial): a randomised study. Lancet. 2009;373:125–36. [PMC free article] [PubMed]
29. Benedetti Panici P, Basile S, Maneschi F, Alberto Lissoni A, Signorelli M, Scambia G, et al. Systematic pelvic lymphadenectomy vs. no lymphadenectomy in early-stage endometrial carcinoma: randomized clinical trial. J Natl Cancer Inst. 2008;100:1707–16. [PubMed]
30. Case AS, Rocconi RP, Straughn JM, Jr, Conner M, Novak L, Wang W, et al. A prospective blinded evaluation of the accuracy of frozen section for the surgical management of endometrial cancer. Obstet Gynecol. 2006;108:1375–9. [PubMed]