Prognostic significance of erythropoietin expression in human endometrial carcinoma

Authors

  • Geza Acs M.D., Ph.D.,

    Corresponding author
    1. Department of Pathology and Laboratory Medicine, University of Pennsylvania Medical Center, Philadelphia, Pennsylvania
    • Department of Pathology and Laboratory Medicine, University of Pennsylvania Medical Center, 6 Founders Pavilion, 3400 Spruce Street, Philadelphia, PA 19104
    Search for more papers by this author
    • Fax: (215) 349-5910

  • Xiaowei Xu M.D., Ph.D.,

    1. Department of Pathology and Laboratory Medicine, University of Pennsylvania Medical Center, Philadelphia, Pennsylvania
    Search for more papers by this author
  • Christina Chu M.D.,

    1. Division of Gynecologic Oncology, Department of Obstetrics and Gynecology, University of Pennsylvania Medical Center, Philadelphia, Pennsylvania
    Search for more papers by this author
  • Peter Acs M.D., Ph.D.,

    1. Lombardi Comprehensive Cancer Center, Georgetown University Hospital, Washington, DC
    Search for more papers by this author
  • Ajay Verma M.D., Ph.D.

    1. Department of Neurology, Uniformed Services University of the Health Sciences, Bethesda, Maryland
    Search for more papers by this author

  • The opinions expressed herein do not necessarily reflect the views of the Uniformed Services University of the Health Sciences or the U.S. Government.

  • This article is a US Government work and, as such, is in the public domain in the United States of America.

Abstract

BACKGROUND

Erythropoietin (Epo), which is induced by hypoxia, controls erythropoiesis and protects neurons from hypoxic damage. Hypoxia in malignant disease is associated with invasion, metastasis, and resistance to therapy. The authors recently demonstrated hypoxia-stimulated expression of Epo and Epo receptor (EpoR) in human breast and cervical carcinomas, suggesting a role for autocrine Epo signaling in the hypoxic adaptations of carcinomas.

METHODS

The authors characterized the expression of Epo, EpoR, hypoxia-inducible factor (HIF)-1α, estrogen receptor (ER), and progesterone receptor (PR) by immunohistochemical methods using endometrial carcinoma samples from 107 women and benign endometrial samples from 59 women in various phases of the menstrual cycle. They then analyzed potential correlations of Epo and EpoR immunostaining and clinicopathologic tumor features with outcome.

RESULTS

In benign endometrial tissue samples, Epo and EpoR expression increased over the course of the cycle, with the highest levels observed in the late secretory phase. Epo expression in benign endometrial samples showed a negative correlation with ER and PR expression. The authors found Epo and EpoR expression in 95.3 % and 100% of endometrial carcinoma samples, respectively. Increased EpoR, but not Epo, expression in tumors was associated with advanced-stage disease, lymphovascular invasion, lymph node metastasis, and loss of ER expression. Increased Epo expression was observed in perinecrotic tumor regions in a pattern similar to the HIF-1α expression pattern. Increased Epo expression was significantly associated with adverse clinical outcome on both univariate and multivariate analysis.

CONCLUSIONS

Hypoxia-inducible autocrine Epo signaling in endometrial carcinoma may contribute to tumor progression and increased aggressiveness. Increased Epo expression in endometrial carcinomas may be an independent prognostic and/or predictive factor. Cancer 2004. Published 2004 American Cancer Society.

Erythropoietin (Epo) is a glycoprotein hormone responsible for regulating erythropoiesis.1 Epo normally is produced by the kidney,1 and its gene expression is modulated by tissue hypoxia via hypoxia-inducible factor-1 (HIF-1)-directed gene transcription.1, 2 Epo acts by binding to Epo-specific receptors (EpoR), which belong to the cytokine receptor type I superfamily.1, 3 EpoR stimulation in erythroblasts promotes proliferation and differentiation, leads to increased expression of the antiapoptotic proteins Bcl-2 and Bcl-XL, and inhibits apoptosis.3

Epo was long considered to be a specific stimulator of erythropoiesis. However, recently, other sites of Epo production have been reported, including brain astrocytes,4 breast epithelial cells,5–7 and human female reproductive organs, including the uterus.8, 9 EpoR is expressed by a variety of cell types as well, including endothelial cells,10 neurons,11 mammary epithelial cells,5–7 and human endometrial cells,9 suggesting a wider biologic role for Epo signaling unrelated to erythropoiesis. Epo also stimulates proliferation and migration of human endothelial cells10 and promotes angiogenesis.8 Because uterine Epo production is stimulated by estrogen, Epo signaling may play an important role in cyclic uterine angiogenesis.8

Previous studies have shown that cultured human breast and cervical carcinoma cells, as well as other carcinoma cells, express high levels of EPO/Epo and EPOR/EpoR mRNA and protein.5, 12 Exposure of the tumor cells to recombinant human Epo (rHuEpo) stimulated tyrosine phosphorylation and DNA synthesis and proliferation, suggesting that Epo signaling is biologically active in malignant cells.5, 12 EpoR expression also has been found in tumors of the female genital tract, including ovarian and endometrial carcinomas.13

Endometrial adenocarcinoma is the most common malignancy of the female genital tract and the fourth most common malignancy among women in the United States.14 The most common histologic type is endometrioid adenocarcinoma. Most of these tumors are estrogen sensitive and develop in perimenopausal or postmenopausal women.15 In contrast, serous and clear cell adenocarcinomas typically occur in older, postmenopausal women, are less strongly associated with hormonal stimulation, and exhibit more aggressive clinical behavior.15 The importance of traditional prognostic factors for endometrial carcinomas, such as International Federation of Gynecology and Obstetrics (FIGO) stage, tumor grade, and type, has been well established.15 However, during the last decade, efforts have focused on attempting to identify molecular events that are correlated with the malignant potential of endometrial carcinomas.

In the current study, we examined the immunohistochemical expression of Epo and EpoR in benign endometrial samples obtained at various points during the menstrual cycle and in a series of endometrial carcinoma samples. We then analyzed potential correlations of Epo and EpoR expression levels with estrogen receptor (ER), progesterone receptor (PR), and hypoxia-inducible factor (HIF)-1α expression levels, as well as with various clinicopathologic tumor features and clinical outcome.

MATERIALS AND METHODS

Clinical Tissue Samples

Benign endometrial tissue samples from 59 premenopausal women undergoing hysterectomy for leiomyoma were selected from the files of the University of Pennsylvania Medical Center (Philadelphia, PA). None of the patients received hormonal treatments or used intrauterine contraception for at least 6 months before surgery, and all patients had regular menstrual cycles. Histologic dating of the endometrial samples was independently performed by two pathologists according to the standard histologic dating method,16 and the respective phases of the menstrual cycle were verified by establishing correlations between the date of the last menstrual period and histologic findings. Data obtained from individual uterine specimens were grouped according to the phase of the menstrual cycle: proliferative (Days 1–14, n = 22), early secretory (Days 15–19, n = 11), midsecretory (Days 20–22, n = 13), and late secretory (Days 23–28, n = 13). Eleven gestational endometrial samples were obtained from elective abortion specimens (Weeks 8–12 of gestation).

One hundred seven hysterectomy specimens containing endometrial carcinoma were selected from the files of the University of Pennsylvania Medical Center. All hematoxylin and eosin–stained slides were reviewed to confirm the diagnoses; this review included assessment of histologic type, tumor grade, and FIGO stage. Staging was defined according to the FIGO surgical staging system.17 Pelvic and/or paraaortic lymph node sampling was performed for 79 patients. Of the 28 patients with unknown lymph node status, Stage III or IV was assigned to 15 patients based on the presence of adnexal involvement, omental metastasis, or positive peritoneal cytology. The remaining patients with unknown lymph node status included 2 with clear cell carcinoma and no myometrial invasion and 13 with Grade 1 endometrioid carcinoma and no or only superficial myometrial invasion. For these patients, disease stage was assigned based on the combination of pathologic and surgical findings. Histologic classification was performed according to the World Health Organization system.18 Grading was based on the degree of glandular differentiation, in accordance with the FIGO guidelines.17 The clinicopathologic features of the endometrial carcinomas are summarized in Table 1. Follow-up of patients was performed on the basis of information reported in the clinical histories. All patients were included in the analysis of data. Only the records of patients who died of disease were considered to be uncensored; the records of all patients who were alive at follow-up or who did not die of disease (or a related cause) were considered to be censored. During the follow-up interval, tumor recurrence was observed in 37 patients (34.6%), and 19 patients (17.8%) died of disease. Study protocols were approved by the University of Pennsylvania Institutional Review Board.

Table 1. Summary of Clinicopathologic Features of Endometrial Carcinomas
CharacteristicAll (n = 107)Endometrioid (n = 74)Serous (n = 19)Clear cell (n = 14)
  1. FIGO: International Federation of Gynecology and Obstetrics; MI: depth of myometrial invasion; LVI: lymphovascular invasion; LN: lymph node.

Age (yrs)    
 Median65626867
 Mean (range)63.6 (25–91)62.0 (25–91)68.1 (45–82)66.1 (50–80)
Tumor size (cm)    
 Median3.543.752.9
 Mean (range) 3.9 (0.5–13) 3.8 (0.5–10) 4.1 (1–12) 3.6 (1–13)
FIGO grade    
 1363600
 2201622
 351221712
MI    
 None9603
 <50%564097
 ≥50%4228104
LVI    
 Absent574377
 Present5031127
LN status    
 Negative524075
 Positive271665
 Unknown281864
Necrosis    
 None392955
 <5%362286
 ≥5%322363
FIGO stage    
 IA8602
 IB383341
 IC11713
 IIA1001
 IIB7520
 III231652
 IV19775

Immunohistochemistry

Immunohistochemical assays were performed on formalin-fixed, paraffin-embedded sections as described previously.6, 12 Slides were incubated with antibodies against Epo (rabbit polyclonal, H-162 [1:200 dilution]; Santa Cruz Biotechnologies, Santa Cruz, CA), EpoR (rabbit polyclonal, C-20 [1:400 dilution]; Santa Cruz Biotechnologies), and HIF-1α (mouse monoclonal, clone H1alpha67 [1:10,000 dilution]; Neomarkers, Fremont, CA) overnight at 4 °C. For ER (mouse monoclonal, clone 1D5 [1:100 dilution]; Dako, Carpinteria, CA) and PR (mouse monoclonal, PgR636 [1:300 dilution], Dako) immunostaining, slides were incubated for 30 minutes at room temperature. For Epo, EpoR, ER, and PR, the Dako EnVision+ System HRP was used. For HIF-1α, the Catalyzed Signal Amplification kit (Dako) was used according to manufacturer's recommendations. For Epo and EpoR, slides of fetal liver specimens were used as positive controls. A negative control was obtained in each case by omitting the primary antibody. The specificity of the Epo and EpoR antibodies has been confirmed in previous reports.5, 7, 11 In addition, the specificity of EpoR and Epo immunoreactivity also was evaluated using the antibody absorption test. For example, the primary antibody was preincubated with blocking peptide for EpoR (Santa Cruz Biotechnologies) or rHuEpo (R & D Systems, Minneapolis, MN) (10:1 peptide-to-antibody ratio), which resulted in the complete elimination of immunohistochemical staining.

Immunohistochemical stains for Epo, EpoR, ER, PR, and HIF-1α were interpreted semiquantitatively by assessing the intensity and extent of staining using a four-tiered scale.6, 12 For Epo in cytoplasm, EpoR in cytoplasm and/or the cell membrane, and ER, PR, and HIF-1α, nuclear immunoreactivity was considered to indicate positivity. The percentage of weakly, moderately, and strongly staining cells was determined, and a staining score was calculated as follows: score (maximum, 300) = 1 × the percentage of weakly staining cells + 2 × the percentage of moderately staining cells + 3 × the percentage of strongly staining cells. For the purposes of statistical analysis, for ER and PR, tumors were considered to exhibit positive expression if nuclear staining was observed in ≥ 10% of tumor cells. Tumor samples showing any nuclear staining were considered to be positive for HIF-1α expression. For Epo and EpoR, expression in tumor samples was considered to be elevated if the immunohistochemical staining score of a tumor sample was higher than the mean plus 2 standard deviations for all tumor samples (> 100 for Epo and > 250 for EpoR).

Statistical Analysis

The Wilcoxon signed rank test was used to compare median Epo and EpoR immunohistochemical expression levels in tumor samples and in adjacent benign endometrial tissue samples. Median Epo, EpoR, HIF-1α, ER, and PR immunohistochemical expression levels were compared using the Kruskal–Wallis one-way analysis of variance by ranks followed by the Dunn multiple comparison test, when appropriate. Increased/positive versus not increased/negative immunostaining in tumor samples was compared using the chi-square test. Correlations of ER and PR expression levels with Epo immunostaining were assessed using the Spearman rank correlation test. Survival curves were plotted using the Kaplan–Meier method and compared using the log-rank test. A Cox proportional hazards model was used to assess the effect of tumor variables on survival. The endpoint of the analysis was overall survival as measured from the day of surgery. For each test, statistical significance was indicated by a two-sided P value of less than 0.05. Computations were performed using Graphpad Prizm (Version 3; GraphPad Software, San Diego, CA) and SYSTAT (Version 10.2; SYSTAT Software, Richmond, CA) software.

RESULTS

Benign Endometrial Tissue Samples

The results of the immunohistochemical assays of benign endometrial samples are illustrated and summarized in Figures 1 and 2. ER and PR immunoreactivity decreased significantly in the endometrial glands over the course of the normal cycle. Changes in ER and PR expression in stromal cells were less prominent, and PR expression remained relatively stable in stromal cells. In contrast to ER and PR, both Epo and EpoR expression in endometrial glands increased significantly over the course of the cycle. However, the change in Epo expression was more prominent. The highest levels of Epo and EpoR expression were observed in late secretory and gestational endometrial samples. Epo and EpoR expression in stromal cells also increased over the course of the cycle, with the highest levels observed in decidualized stromal cells in late secretory and gestational endometrial samples. Except for rare decidualized stromal cell staining in gestational endometria, no evidence of significant HIF-1α expression was observed in either glandular or stromal cells. Highly significant negative correlations were found between glandular Epo and ER expression levels (Spearman test: r = −0.6750; P < 0.0001) and between glandular Epo and PR expression levels (Spearman test: r = −0.6867; P < 0.0001) (Fig. 3A,B).

Figure 1.

Immunohistochemical expression of estrogen receptor (ER), progesterone receptor (PR), erythropoietin (Epo), and erythropoietin receptor (EpoR) in benign endometrial tissue samples at various stages of the menstrual cycle. H & E: hematoxylin and eosin. Immunohistochemical stains with hematoxylin counterstain, original magnification × 200.

Figure 2.

Expression of estrogen receptor (open triangles), progesterone receptor (filled triangles), erythropoietin (filled squares), and erythropoietin receptor (open squares) in benign endometrial tissue samples at various stages of the menstrual cycle. Data points represent mean immunostaining scores, and error bars indicate the standard error of the mean. P: proliferative phase; ES: early secretory phase; MS: midsecretory phase; LS: late secretory phase; G: gestational endometrium.

Figure 3.

Correlation of immunohistochemical expression of erythropoietin (Epo) with estrogen receptor (ER) and progesterone receptor (PR) expression in (A,B) benign endometrial tissue samples and (C,D) endometrial carcinoma samples. Calculated regression lines also are shown on the correlation plots.

Endometrial Carcinoma Samples

The results of the immunohistochemical assays of endometrial carcinoma samples are summarized in Table 2. Heterogeneous, weak-to-moderate Epo immunostaining was found in 102 of 107 endometrial carcinoma samples (95.3%). However, strong, prominent staining was present in viable tumor cells adjacent to necrotic areas (Fig. 4A,4B). Necrotic tissue areas exhibited no staining for Epo. Compared with the adjacent benign endometrial cells present in the assayed samples, Epo immunostaining was significantly increased in carcinomas (Wilcoxon signed rank test: P < 0.001) (Fig. 4C,D). No significant difference in Epo expression was found among the various subtypes of endometrial carcinoma (Table 2). Epo expression in endometrial carcinoma samples did not show any significant correlation with any of the clinicopathologic tumor features examined (Table 3). In addition, no significant correlation was found between Epo and ER expression or between Epo and PR expression in carcinoma samples (Spearman test: P > 0.05) (Fig. 3C,D).

Table 2. Summary of Immunohistochemical Analysis of Endometrial Carcinomas
CharacteristicAllEndometrioidSerousClear cell
  1. Epo: erythropoietin; EpoR: erythropoietin receptor; HIF: hypoxia-inducible factor; ER: estrogen receptor; PR: progesterone receptor; SEM: standard error of the mean.

Epo    
 Median score62.5705080
 Mean ± SEM64.4 ± 4.166.2 ± 5.149.1 ± 8.376.1 ± 11.5
 No. increased171313
 No. not increased90611811
 Kruskal–Wallis P0.174   
 Chi-square P0.373   
EpoR    
 Median score210200220235
 Mean ± SEM212.2 ± 5.3204.6 ± 6.4221.8 ± 10.7236.4 ± 14.9
 No. increased55321310
 No. not increased524264
 Kruskal–Wallis P0.072   
 Chi-square P0.032   
HIF    
 Median score9.59.5138
 Mean ± SEM27.0 ± 3.827.3 ± 4.621.7 ± 7.132.7 ± 12.4
 No. positive79541510
 No. negative282044
 Kruskal–Wallis P0.968   
 Chi-square P0.647   
ER    
 Median score104000
 Mean ± SEM44.1 ± 6.360.4 ± 8.110.7 ± 5.41.0 ± 0.7
 No. positive565150
 No. negative51231414
 Kruskal–Wallis P<0.0001   
 Chi-square P<0.0001   
PR    
 Median score42.5122.500
 Mean ± SEM84.6 ± 10.1118.2 ± 12.029.8 ± 16.90 ± 0
 No. positive665970
 No. negative41151214
 Kruskal–Wallis P<0.0001   
 Chi-square P<0.0001   
Figure 4.

(A–C) Immunohistochemical staining for erythropoietin (Epo) in malignant endometrial tissue. (D) Comparison of Epo immunostaining scores in malignant endometrial tissue and adjacent benign endometrial tissue. Bars indicate median values (Wilcoxon signed rank test: P < 0.001). (E,F) Immunohistochemical staining for hypoxia-inducible factor (HIF)-1α in malignant endometrial tissue. Similar spatial distributions of expression were found for Epo and HIF-1α, with prominent staining in tumor cells adjacent to necrotic areas (*). Epo expression was increased in endometrial carcinoma cells compared with adjacent benign endometrial glandular cells (arrow). Immunohistochemical stains with hematoxylin counterstain, original magnification × 50 (A,E); × 100 (B,F); × 200 (C).

Table 3. Statistical Analysis for Epo, EpoR, HIF-1α, ER, and PR Expression in Endometrial Carcinoma According to Various Clinicopathologic Features
VariableP value
All tumorsEndometrioid tumors
EpoEpoRHIFERPREpoEpoRHIFERPR
  1. Epo: erythropoietin; EpoR: erythropoietin receptor; MI: depth of myometrial invasion; LVI: lymphovascular invasion; LN: lymph node; ER: estrogen receptor; PR: progesterone receptor; HIF: hypoxia-inducible factor; FIGO: International Federation of Gynecology and Obstetrics; N/A: not applicable.

FIGO grade          
 Chi-square0.8220.0150.6160.0000.0000.9250.1180.5710.0090.011
 Kruskal–Wallis0.7170.0690.694N/AN/A0.6530.2340.303N/AN/A
FIGO stage          
 Chi-square0.9220.0010.9870.0110.0110.7440.0070.6650.080.021
 Kruskal–Wallis0.2990.0020.851N/AN/A0.6290.0380.746N/AN/A
MI          
 Chi-square0.6100.3280.6220.8090.0780.5820.1150.1490.2790.018
 Kruskal–Wallis0.2010.2260.386N/AN/A0.1680.1490.503N/AN/A
LVI          
 Chi-square0.4000.0050.6800.0150.0280.4840.0020.4160.0060.008
 Kruskal–Wallis0.2260.0030.706N/AN/A0.1070.0080.368N/AN/A
LN status          
 Chi-square0.5610.0070.4630.0490.0140.7020.0280.3220.3980.261
 Kruskal–Wallis0.4890.0060.518N/AN/A0.8360.0720.581N/AN/A
ER status          
 Chi-square0.3500.0100.143N/A0.0000.4050.0830.086N/A0.000
 Kruskal–Wallis0.2870.0140.735N/AN/A0.1430.1420.607N/AN/A
PR status          
 Chi-square0.4470.0940.5930.000N/A0.4420.1570.9170.000N/A
 Kruskal–Wallis0.1590.0680.293N/AN/A0.2700.2350.451N/AN/A
Necrosis          
 Chi-square0.1660.2390.0610.1280.1910.1010.0270.0210.0260.006
 Kruskal–Wallis0.4440.2190.000N/AN/A0.8800.1440.001N/AN/A

Diffuse, moderate-to-strong cytoplasmic and membrane EpoR immunostaining was observed in all 107 endometrial carcinoma samples (Fig. 5). EpoR staining was uniform throughout the tumor samples, although further increases in staining intensity were observed in areas surrounding necrotic regions. In addition, strong EpoR expression was found in the tumor vasculature (Fig. 5B). Compared with adjacent benign endometrial tissue samples, endometrial carcinoma samples exhibited significantly increased EpoR immunostaining (Wilcoxon signed rank test: P < 0.0001) (Fig. 5C,D). EpoR expression tended to be higher in serous and clear cell carcinomas compared with endometrioid tumors; however, this tendency did not reach statistical significance (Table 2). Significantly increased EpoR immunostaining was observed in carcinomas in association with advanced stage, lymphovascular invasion, lymph node metastasis, and lack of ER expression (Table 3).

Figure 5.

(A,B) Strong immunohistochemical expression of erythropoietin receptor (EpoR) in endometrial carcinoma tissue is uniformly distributed. Note the EpoR expression in vascular endothelial and smooth muscle cells (*). (C,D) EpoR expression is markedly increased in endometrial carcinoma cells compared with benign endometrial glandular cells. Bars indicate median EpoR immunostaining scores (Wilcoxon signed rank test: P < 0.0001). Immunohistochemical stains with hematoxylin counterstain, original magnification × 50 (A); × 200 (B); × 100 (C).

Nuclear HIF-1α immunostaining was observed in 79 of 107 endometrial carcinoma samples (73.8%) (Table 2). HIF-1α expression was focal in all samples with positive staining and was concentrated in areas surrounding necrotic regions (Fig. 4E,F). The spatial distributions of HIF-1α expression and prominent Epo expression were quite similar in tumor samples. There was no difference in HIF-1α expression among tumor samples of various types. Significantly increased HIF-1α expression was observed in tumor samples containing areas of necrosis, but we found no correlation between HIF-1α expression and any of the other clinicopathologic tumor features examined (Table 3).

Survival Analysis

The median time to death for the uncensored subgroup was 9.3 months (range, 4.1–72 months), whereas the median follow-up of censored patients was 39.4 months (range, 1–190.3 months). The results of the univariate survival analysis are summarized in Table 4. FIGO stage, tumor grade, tumor type, depth of myometrial invasion, presence of lymphovascular invasion, presence of lymph node metastases, and lack of PR expression were associated with poor prognosis. Increased Epo expression in endometrial carcinoma samples also was associated with poor disease-related survival (Fig. 6), but not with recurrence-free survival. Patients with tumors that exhibited increased EpoR expression tended to have poorer disease-related and recurrence-free survival results; however, this tendency did not reach statistical significance.

Table 4. Univariate Analysis of the Association of Outcome with Clinicopathologic Features and with the Expression of Epo, EpoR, HIF-1α, ER, and PR in Endometrial Carcinoma
VariableP value
Overall survivalRecurrence-free survival
All tumorsEndometrioid tumorsAll tumorsEndometrioid tumors
  1. Epo: erythropoietin; EpoR: erythropoietin receptor; ER: estrogen receptor; PR: progesterone receptor; HIF: hypoxia-inducible factor; FIGO: International Federation of Gynecology and Obstetrics; N/A: not applicable.

FIGO stage0.00010.00010.00010.0001
FIGO grade0.00010.0010.00010.001
Tumor type0.017N/A0.001N/A
Myometrial invasion0.0030.0050.00010.0001
Lymphovascular invasion0.0010.0040.00010.0001
Lymph node status0.0010.00010.00010.0001
ER status0.0840.7510.0240.416
PR status0.0040.0310.0030.010
Increased Epo expression0.0370.0080.6050.231
Increased EpoR expression0.3980.5890.1720.178
HIF-1α expression0.2430.4910.1740.455
Figure 6.

Kaplan–Meier disease-related survival curves stratified according to increased erythropoietin (Epo) expression. (A) All endometrial carcinomas combined (P = 0.037). (B) Tumors of endometrioid subtype (P = 0.008).

For the stepwise logistic regression models, we included the following variables: FIGO stage, tumor grade, tumor type, lymph node status, ER and PR status, and Epo, EpoR, and HIF-1α expression. Backward elimination according to Cox regression results led to a model containing four independent terms that were predictive of disease-related survival: Epo expression (P = 0.002), FIGO stage (P = 0.008), lymph node status (P = 0.013), and PR expression (P = 0.015).

DISCUSSION

We performed immunohistochemical analysis to characterize the expression of Epo, EpoR, HIF-1α, ER, and PR in a series of 107 endometrial carcinoma samples and 59 benign endometrial tissue samples obtained at various points throughout the menstrual cycle. Expression levels for both Epo and EpoR were low in benign endometrial glands in the proliferative phase but increased significantly in the secretory phase, with the highest expression levels noted in late secretory and gestational endometria. Our results confirm recent reports of increased Epo expression in secretory endometrial tissue compared with proliferative endometrial tissue19 and reveal the first documented correlations of Epo and EpoR expression with the distributions of ER and PR expression and with cyclic changes in the expression of these markers in human endometrial tissue.

We observed expression of Epo, EpoR, and HIF-1α in 95.3%, 100%, and 73.8% of the endometrial carcinoma samples, respectively. Epo and EpoR immunostaining in endometrial carcinoma samples was significantly increased compared with the adjacent benign (usually proliferative or inactive) endometrial tissue samples. Epo immunostaining was heterogeneous, with the most prominent expression observed in tumor cells adjacent to necrotic areas. The spatial distribution of increased Epo immunoreactivity was quite similar to that of HIF-1α, the expression of which was also confined to tumor regions adjacent to necrotic areas, which are believed to be among the most hypoxic sites.20 The patterns of Epo and HIF-1α expression were similar to those observed in cervical squamous cell carcinomas12 and were consistent with the known regulation of Epo expression by HIF-1.2 In contrast, EpoR expression was uniform and homogenous, although increased expression was present in tumor cells adjacent to necrotic areas. These observations support our recent finding of hypoxic up-regulation of both Epo and EpoR in breast,5 cervical,12 and endometrial carcinoma cell lines (unpublished data).

Epo immunostaining in endometrial carcinoma samples was not correlated with any of the clinicohistopathologic features examined, including FIGO stage, tumor type and grade, lymphovascular invasion, lymph node status, and ER and PR status. These results are similar to those of earlier breast carcinoma studies, which found no correlation between Epo expression6 or intratumoral hypoxia21 and clinicopathologic tumor features. In contrast, high levels of EpoR expression in tumors were associated with advanced FIGO stage, lymphovascular invasion, lymph node metastases, and loss of ER expression.

The human endometrium is characterized by constant cell proliferation, differentiation, and breakdown during the menstrual cycle, and it is highly responsive to ovarian steroid hormones. It is well established that ER expression and PR expression vary with menstrual cycle phase. This variability is related to the known effects of estrogen and progesterone on steroid hormone receptor expression.22 Although steroids control endometrial function, paracrine and autocrine growth factor/cytokine signals are now viewed as key mediators of reproductive function. It has been suggested that neovascularization in normal endometrial tissue is indirectly modulated by ovarian steroids via the production of locally active angiogenic factors, such as vascular endothelial growth factor.23 The presence of EPO/Epo and EPOR/EpoR mRNA and protein in endometrial cells has been demonstrated previously,9 and it has been suggested that Epo, acting in an autocrine/paracrine manner, may be involved in the cyclic proliferation of endometrial epithelial cells and in vascular proliferation.8, 9 Our results also support the role of Epo signaling in the endometrial changes that occur during the menstrual cycle.

The observation of highly significant negative correlations between Epo and ER expression and between Epo and PR expression, together with data from the literature,8, 9 suggests that Epo expression in normal endometrial tissue is regulated by estrogen and progesterone. This hypothesis also is supported by the finding that the 5′-flanking region of the human EPO gene contains sequences that are highly homologous to the sequence responsible for ER binding.8 However, the situation appears to be different for endometrial carcinomas. We found no correlation between Epo and ER or PR expression in endometrial carcinomas, suggesting that in malignant endometrial epithelial cells, Epo expression may be independent of steroid hormone action. The enhanced expression of Epo in endometrial carcinoma cells may instead be associated with tumor hypoxia. In the current study and in previous reports, we have shown that Epo immunoreactivity in tumor samples is most prominently associated with viable cells in the hypoxic perinecrotic zones.12 It is believed that the principal determinant of HIF-1α accumulation in tumor cells is tissue hypoxia.24 The similarity in the expression patterns of Epo and HIF-1α in tumor samples further supports the hypothesis that hypoxia plays a role in elevating Epo expression and is consistent with the well known regulation of Epo expression by HIF-1.2 In contrast, EpoR immunostaining was uniform throughout the tumor tissue samples, with further accentuation near necrotic areas. These findings are similar to those that we previously described in human breast and cervical carcinomas.6, 12 Because the EPOR gene is not known to have HIF recognition elements, these findings suggest that although hypoxia induces EpoR expression, other hypoxia-activated or oncogenic mechanisms could promote increased EpoR expression in neoplastic cells.

In addition to making solid tumors resistant to radiation and chemotherapy, tumor hypoxia may also increase aggressiveness and advance tumor progression.25 To our knowledge, the current report is the first to show that increased Epo expression in human endometrial carcinoma is associated with significantly poorer disease-related survival; this finding suggests that Epo expression may represent an independent prognostic and/or predictive factor. One of the major functions of Epo in erythroblasts and neurons is inhibition of apoptosis.1, 3 We have shown that Epo also inhibits the cytotoxic and apoptotic actions of cisplatin in cervical carcinoma cells.12 The action of Epo on EpoR-expressing malignant cells also may be mediated by mechanisms similar to those described for erythroblasts and neurons and may contribute to the treatment resistance, propensity for progression, and increased aggressiveness associated with hypoxic tumors.

rHuEpo is used frequently to manage treatment-related anemia in patients receiving curative radiotherapy and chemotherapy.26 Epo is a potent growth factor and may stimulate proliferation and inhibit apoptosis of EpoR-bearing tumor cells.5 Epo also stimulates proliferation and migration of vascular endothelial cells and augments angiogenesis.8, 10 We have shown previously5, 12 and in the current study that the vasculature of solid tumors expresses EpoR. Epo, therefore, may promote tumor angiogenesis as well. Local inhibition of Epo signaling has been shown to reduce vascularization and promote destruction of ovarian and uterine carcinoma xenografts.27 Moreover, recently, two prospective multicenter studies involving patients with breast carcinoma and patients with head and neck carcinoma have suggested that despite elevating hemoglobin levels, rHuEpo, compared with placebo, may have an adverse effect on overall prognosis.28, 29 Cellular responses to Epo may collectively promote the growth of EpoR-bearing tumors, and these actions may be enhanced further by either high levels of endogenous Epo production or by exogenous Epo administration. Therefore, until it has been clearly demonstrated that pharmacologic doses of Epo lack such trophic effects in vivo, we suggest that the administration of rHuEpo to treat patients with malignant disease should be performed with some degree of caution.

In summary, we have shown that human endometrial carcinomas express Epo and EpoR. Although Epo expression in benign endometrial cells appears to be mediated by estrogen and progesterone, in endometrial carcinomas, Epo expression is associated with tumor hypoxia. To our knowledge, the current report is the first to suggest that increased Epo expression in endometrial carcinomas may be an independent prognostic and/or predictive factor.

Ancillary