Hypoxia-Inducible Erythropoietin Signaling in Squamous Dysplasia and Squamous Cell Carcinoma of the Uterine Cervix and Its Potential Role in Cervical Carcinogenesis and Tumor Progression

Cell Culture and Treatments

All culture media were purchased from Life Technologies, Inc. (Rockville, MD). The HeLa and SiHa human cervical carcinoma, Hep3B human hepatoma, and U251 human glioblastoma cell lines were obtained from the American Type Culture Collection (ATCC, Manassas, VA) and cultured according to ATCC directions. Forty-eight hours before hypoxic treatment cells were switched to serum-free medium. Hypoxic treatment of cells was performed in an enclosed chamber (Billups-Rothenberg Inc., Del Mar, CA) flushed with a premixed gas mixture (1% O₂, 5% CO₂, 94% N₂) for the times indicated. The O₂ concentration in the culture medium was confirmed to be 1% using an oxygen-sensitive electrode (World Precision Instruments, Sarasota, FL), and this oxygen concentration was maintained throughout the course of the experiments. Stimulation of tyrosine phosphorylation was performed by treating cells in phenol-free Dulbecco’s modified Eagle’s medium with recombinant human erythropoietin (rHu-Epo, epoetin-α, 10 U/ml; Amgen, Thousand Oaks, CA) or the Epo mimetic peptide Emp1 (AEHCSLNENITVPDTKV, 1 μg/ml) for 30 minutes.

Cytotoxicity Assay

For cytotoxicity studies cells were plated in 96-well plates at 3000 cells/well. The following morning cells were treated with rHuEpo (0, 25, 50, or 200 U/ml) with or without the general tyrosine kinase inhibitor genistein (10 μmol/L) for 60 minutes. Cells were then exposed to cisplatin (10 μg/ml) for 20 hours. Each condition was performed in quadruplicate. Viability of cells was assessed using the non-radioactive cell proliferation (MTT) assay (Promega, Madison, WI).

Annexin V Apoptosis Assay

Apoptosis was assayed by detection of membrane externalization of phosphatidylserine using the Annexin V-PE Apoptosis Detection Kit I (BD Biosciences, Palo Alto, CA). Cells were plated at 40 to 50% confluency in six-well plates and allowed to adhere overnight. Media was then removed and replaced with serum-free media with or without rHuEpo. Cells were pretreated with rHuEpo (0, 50, 100, or 200 U/ml) for 2 hours and then challenged with cisplatin (5 μg/ml) for 48 hours. Both adherent and floating cells were harvested for the apoptosis assay. Cells were washed once with phosphate-buffered saline, resuspended in binding buffer, and stained with phycoerythrin-conjugated Annexin V and 7-amino-actinomycin (7-AAD) according to the manufacturer’s protocol (5 μl for 1 × 10⁵ cells) for 15 minutes. Cells were analyzed using an EPICS XL-MCL (Beckman Coulter, Miami, FL) flow cytometer. Emission from Annexin V was collected through a 575-nm band-pass filter and emission from 7-AAD was collected through a 675-nm band-pass filter. A minimum of 10,000 events was collected for each sample and each treatment group was done in triplicates.

Reverse Transcriptase-Polymerase Chain Reaction (RT-PCR)

Measurement of Epo and EpoR mRNA were performed as described previously. 64,80 RNA was isolated using the RNeasy RNA isolation kit (Qiagen, Inc., Valencia, CA) and the Superscript One-Step RT-PCR System (Life Technologies, Inc., Gaithersburg, MD) was used for RT-PCR. Approximately 1 μg of RNA and 10 pmol of both the forward and reverse primers were used for each sample. RT was performed at 50°C for 30 minutes. The PCR cycle protocol for the Epo primers was 42 cycles at 94°C for 20 seconds, 61°C for 45 seconds, and 72°C for 1 minute, and for EpoR was 42 cycles at 94°C for 15 seconds, 55°C for 30 seconds, 72°C for 1 minute. Primers for Epo were: sense 5′-ACTCTGCTTCGGGCTCTGGGAGCCCAGAAG-3′ (corresponding to nucleic acids 581 to 610), and anti-sense 5′-AAGCAATGTTGGTGAGGGAGGTGGTGGAT-3′ (corresponding to nucleic acids 786 to 814). 64,80 Primers for EpoR were: sense 5′-GGCAGTGTGGACATAGTGGC-3′ (corresponding to nucleic acids 1304 to 1323), and anti-sense 5′-AGCAGGATGGATTGGGCAGA-3′ (corresponding to nucleic acids 1782 to 1801). These primer sets were used previously, with subsequent nucleotide sequencing of the 447-bp EpoR RT-PCR products obtained from MCF-7 human breast cancer and MG-U87 human glioblastoma cells demonstrating that these were identical to the normal human EpoR (BLAST NCBI search results). 64 The RT-PCR products were then run on a 1% agarose gel. The DNA standard used was the 100-bp DNA ladder from FMC Bioproducts (BioWhittaker Molecular Applications, Rockland, ME).

Western Blotting

For detection of EpoR, whole cell lysates and clinical biopsy sample protein extracts were normalized for protein. One hundred μg of proteins from each sample were subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis, transferred to nitrocellulose, and stained with anti-EpoR antibody (rabbit polyclonal, C20, 1:1500 dilution; Santa Cruz Biotechnology, Inc., Santa Cruz, CA) in the absence or presence of EpoR blocking peptide (10:1 peptide:antibody ratio, Santa Cruz Biotechnology). Secondary antibody was goat anti-rabbit antibody conjugated with horseradish peroxidase (Amersham Pharmacia Biotech, Piscataway, NJ). Immunoreactive bands were visualized using chemiluminescence (SuperSignal WestPico Chemiluminescence kit; Pierce, Rockford, IL). A horseradish peroxidase-conjugated monoclonal antibody (PY20, SC-508, 1:1000 dilution; Santa Cruz Biotechnology) was used to detect phosphotyrosine. Nuclear extracts and Western blot analysis for HIF-1α were performed as previously described 81 using a mouse monoclonal anti-HIF-1α antibody (clone 54, 1:500 dilution; BD Transduction Laboratories, San Jose, CA). Secondary antibody was rabbit anti-mouse conjugated with horseradish peroxidase (Amersham Pharmacia Biotech).

Clinical Samples and Clinical Data

Study protocols involving human material were approved by the University of Pennsylvania Institutional Review Board. Fresh tissue from seven cervical squamous cell carcinomas and adjacent benign cervical squamous mucosa (in one case) was obtained at the time of surgery. A frozen section was cut from each sample and examined to confirm the presence or absence of tumor tissue: all tumor samples contained at least 80% neoplastic cells, whereas no tumor cells were present in the sample of benign mucosa. Tissue samples were stored at −80°C until used for protein extraction.

Ninety-seven cases of cervical biopsies, cone biopsies, or hysterectomies performed for benign disease (leiomyomas), cervical dysplasia, or invasive squamous cell carcinoma (ISCC) were selected from the surgical pathology files of the University of Pennsylvania Medical Center. Hematoxylin and eosin (H&E)-stained slides of all cases were reviewed and the diagnoses confirmed. Grading of squamous dysplasias (CIN I to CIN III) was performed on the H&E-stained slides according to established criteria. 9 Invasive carcinomas were also evaluated to determine the presence or absence of tumor cell keratinization and necrosis. Information regarding clinicopathological tumor features were retrieved from the pathology reports and are summarized in Table 1 ▶ .

Thirty-nine SurePath (formerly known as AutoCyte PREP; Tripath Imaging, Burlington, NC) liquid-based preparations of cervicovaginal cytological samples were retrieved from the cytopathology files of the University of Pennsylvania Medical Center. All Papanicolaou-stained cytological slides were reviewed and the diagnoses confirmed. The cytological diagnoses were as follows: within normal limits, 17 cases; low-grade squamous intraepithelial lesion (HPV infection and CIN I), 5 cases; and high-grade squamous intraepithelial lesion (CIN II and CIN III), 16 cases. Because benign squamous cells were present in all cases, a total of 39 samples of benign squamous cells were analyzed.

Immunohistochemistry and Immunocytochemistry

Immunohistochemical assays were performed on formalin-fixed paraffin-embedded sections as described previously. 65 Five-μm-thick sections were cut and deparaffinized in xylene and rehydrated in graded alcohols. SurePath cytological slides were decoverslipped in xylene, rehydrated in graded alcohols, and postfixed in 10% phosphate-buffered formaldehyde for 30 minutes. All slides were steamed in 0.01 mol/L of sodium citrate buffer (pH 6.0) for 20 minutes. Endogenous peroxidase activity was blocked by 3% hydrogen peroxide in methanol for 20 minutes. Slides were incubated with the antibodies against Epo (rabbit polyclonal, H-162, 1:200 dilution; Santa Cruz Biotechnologies), EpoR (rabbit polyclonal, C-20, 1:400 dilution; Santa Cruz Biotechnologies), bcl-2 (clone 124, mouse monoclonal, 1:50 dilution; DAKO Corp., Carpinteria, CA), p16^INK4a (clone 16P07, mouse monoclonal, 1:100 dilution; Neomarkers, Fremont, CA), and HIF-1α (clone H1alpha67, mouse monoclonal, 1:10,000; Neomarkers) overnight at 4°C. In the cases of Epo, EpoR, bcl-2, and p16^INK4a assays, the slides were then washed five times with Tris-buffered saline containing Tween 20 (TBST, pH 7.6; DAKO) and incubated for 30 minutes at room temperature with horseradish peroxidase-labeled dextran polymer coupled to anti-rabbit or anti-mouse antibody (DAKO EnVision + System HRP, DAKO). Slides were then washed three times with TBST, developed with diaminobenzidine for 10 minutes, and counterstained with hematoxylin. For HIF-1α the DAKO CSA signal amplification kit was used according to the manufacturer’s recommendations. For Epo and EpoR, slides of fetal liver 82 and placenta 63,71 were used as positive controls. For bcl-2 slides of human tonsil, and for p16^INK4a slides of human cervical squamous cell carcinoma known to show strong p16^INK4a immunostaining were used as positive controls. For HIF-1α slides of oral squamous cell carcinoma with large areas of necrosis known to show HIF-1α overexpression were used as positive control. A negative control was done in each case by omission of the primary antibody. The specificity of the Epo and EpoR antibodies were confirmed previously. 64,66,70,75,83 In addition, the specificity of the EpoR and Epo immunoreactivity was also evaluated by antibody absorption test: the primary antibody was preincubated with blocking peptide for EpoR (Santa Cruz Biotechnologies Inc.) or human recombinant Epo (rHuEpo; R&D Systems, Minneapolis, MN) (10:1 peptide:antibody ratio), which resulted in complete abolishment of immunohistochemical staining. The specificity of the immunostaining reaction is further supported by other experiments using a mouse monoclonal anti-Epo (clone 9C21D11, R&D Systems) and a rabbit polyclonal anti-EpoR antibody (Upstate Biotechnology, Lake Placid, NY), 84 which resulted in an immunostaining pattern similar to that obtained with the antibodies used in the current study (Acs et al, unpublished observation).

Double Immunohistochemistry for HIF-1α and Epo

For double-immunohistochemical staining slides were immunostained for HIF-1α using the DAKO CSA kit as described above. After incubation with diaminobenzidine, the slides were washed three times in deionized water and three times in TBST. Slides were then incubated with the primary Epo antibody as described above, washed five times with TBST, and incubated for 30 minutes at room temperature with alkaline phosphatase-labeled dextran polymer coupled to anti-rabbit antibody (DAKO EnVision + System AP, DAKO). Slides were then washed three times with TBST, developed with fast red chromogen for 10 minutes, and counterstained with hematoxylin.

Interpretation of Immunohistochemical Stains

Immunohistochemical stains for Epo, EpoR, p16^INK4a, and HIF-1α were interpreted semiquantitatively by assessing the intensity and extent of staining on the entire tissue sections present on the slides according to a four-tiered (0 to 3) scale. For Epo cytoplasmic, and for EpoR cytoplasmic and/or membrane immunoreactivity was considered positive. For HIF-1α nuclear staining, and for p16^INK4a nuclear and/or cytoplasmic staining was considered positive. In the case of dysplasias or in situ carcinomas, first, the percentage of total epithelial thickness showing positive staining was determined (eg, 50% if the basal half or 75% if the basal three-fourths of the squamous epithelium showed positive immunostaining, and so forth). In the case of invasive tumors, first the total percentage of positively staining tumor cells was determined. Then the percentage of weakly (1), moderately (2), and strongly (3) staining cells was determined, so that the sum of these categories equated with the overall percentage of positivity. A staining score was then calculated as follows: score (out of maximum of 300) = sum of 1 × percentage of weak, 2 × percentage of moderate, and 3 × percentage of strong staining. To assess the immunostaining for Epo and HIF-1α adjacent to areas of tumor cell necrosis and keratinization, immunostaining was scored as described above in at least 10 high-power fields of viable tumor cells within 1 mm of necrotic or keratinized areas, and in tumor regions away from such areas. Bcl-2-stained slides were interpreted by determining the percentage of squamous epithelial cells showing cytoplasmic immunoreactivity of any intensity. Immunohistochemical stains were evaluated independently by two pathologists (GA and PJZ). Slight differences in interpretation were resolved by simultaneous viewing.

Immunofluorescence

For immunofluorescence staining slides were pretreated and incubated with the primary antibody against p16^INK4a or EpoR as described above. Slides were then washed five times with TBST and incubated with a fluorescein isothiocyanate-conjugated rabbit anti-mouse secondary antibody (1:20 dilution, DAKO) or a tetramethyl-rhodamine isothiocyanate-conjugated swine anti-rabbit secondary antibody (1:40 dilution, DAKO) for 30 minutes on 37°C in the case of p16^INK4a and EpoR, respectively. Slides were then coverslipped with 4,6-diamidino-2-phenylindole and examined using a Leica DMRA Research Microscope equipped with a Megapixel charge-coupled device camera. For simultaneous immunostaining of tissue sections and cytological preparations for p16^INK4a and EpoR, slides were first stained for EpoR as described above. Slides were then washed five times with TBST and immunostained for p16^INK4a as described above.

Statistical Analysis

The Wilcoxon signed rank test was used for the comparison of median p16^INK4a, EpoR, Epo, HIF-1α, and bcl-2 immunohistochemical expression levels in dysplastic epithelia and the adjacent benign squamous epithelium, and for the comparison of Epo and HIF-1α immunostaining in tumor areas adjacent to or away from necrosis and keratinization. Median p16^INK4a, EpoR, Epo, HIF-1α, and bcl-2 immunohistochemical expression levels in benign epithelia, CIN I, CIN II, CIN III, and ISCC were compared using the Kruskal-Wallis one-way analysis of variance by ranks followed by Dunn’s multiple comparison test, when appropriate. The correlation between the levels of Epo and EpoR staining and p16, bcl-2, and HIF-1α staining was estimated using the Spearman rank correlation test. The normal versus abnormal distribution of HIF-1α expression in benign and dysplastic epithelia was compared using the Fisher’s exact test. Statistical significance was determined if the two-sided P value of a test was less than 0.05.

Hypoxia-Inducible Expression of Functional EpoR in Human Cervical Carcinoma Cell Lines and Human Cervical Carcinomas

Using RT-PCR, HeLa and SiHa human cervical carcinoma cells cultured under normoxia (21% O₂) or hypoxia (1% O₂) were analyzed for Epo and EpoR mRNA expression. Abundant expression of EpoR mRNA was detected in HeLa cells under both conditions, whereas SiHa cells showed a prominent increase in EpoR mRNA after hypoxia treatment (Figure 1A) ▶ . No Epo mRNA expression could be detected in HeLa and SiHa cervical carcinoma cells (Figure 1A) ▶ . In contrast, human hepatoma Hep3B and glioblastoma U251 cells expressed abundant Epo mRNA (Figure 1B) ▶ , as reported previously. 64

Western blot analysis using EpoR antibodies demonstrated that HeLa cervical carcinoma cells express a prominent and specific immunoreactive band at M_r ∼66,000 corresponding to the EpoR protein (Figure 2A) ▶ . The immunostaining was specifically abolished when the antibody was preincubated with the EpoR peptide against which the antiserum was raised (not shown). Hypoxia treatment induced a marked increase in the expression of both HIF-1α and EpoR in the cells (Figure 2A) ▶ . Functional EpoR signaling in HeLa cells was evidenced by induction of tyrosine phosphorylation after treatment of the cells with rHuEpo or Epo mimetic peptide (Figure 2B) ▶ .

Next, we examined the expression of EpoR protein using Western blotting in clinical samples of cervical squamous cell carcinoma. All seven samples of carcinoma examined showed expression of a specific immunoreactive band at M_r ∼66,000 corresponding to the EpoR protein (Figure 2C) ▶ . Pretreatment of the antibody with the EpoR-blocking peptide completely abolished the immunostaining. No EpoR expression was seen in the sample of benign squamous mucosa.

The majority of chemotherapy regimens used to treat patients with cervical carcinoma includes cisplatin. 85-87 Given the potent anti-apoptotic effect of Epo 46,52,53,56,59 , we examined whether EpoR signaling may inhibit cisplatin-induced cytotoxicity and apoptosis in cervical carcinoma cells. We treated HeLa cells with cisplatin in the absence and presence of rHuEpo. Pretreatment of HeLa cells with rHuEpo showed a dose-dependent protective effect on HeLa cells treated with the chemotherapeutic agent cisplatin, as determined by the MTT assay (Figure 3A) ▶ . Using the annexin V assay, ∼10% of the cell killing caused by cisplatin was purely apoptotic, and rHuEpo dose dependently inhibited this apoptosis (Figure 3B) ▶ .

Immunohistochemistry

To further investigate the potential role of Epo signaling in cervical squamous dysplasia and carcinoma we examined the expression of Epo and EpoR by immunohistochemistry in a series of clinical samples of cervical dysplasia and carcinoma and compared their expression with those of p16^INK4a, HIF-1α, and bcl-2. The results of the immunohistochemical assays are summarized in Table 2 ▶ and illustrated in Figures 4 and 5 ▶ .

None of the 85 benign squamous epithelia showed expression of p16^INK4a, whereas strong nuclear and cytoplasmic immunostaining was seen in 11 of 17 CIN I, all CIN II and CIN III, and 14 of 15 ISCCs (Table 1 ▶ and Figures 4A and 5 ▶ ). The difference in p16^INK4a expression between benign squamous epithelia, CIN I, CIN II, CIN III, and ISCCs was highly significant (see Table 2 ▶ ). Immunofluorescence staining for p16^INK4a showed a similar staining pattern as seen for immunoperoxidase staining.

Cytoplasmic and/or membrane immunostaining for EpoR was present in all samples (Table 2 ▶ , Figures 4B and 5 ▶ ). In benign squamous epithelia EpoR expression was confined to the basal and parabasal cell layers. In contrast, in squamous dysplasia EpoR expression increasingly appeared in the more superficial cell layers and was significantly increased compared to the adjacent benign epithelia (P = 0.0002 for CIN I, P = 0.0003 for CIN II, and P < 0.0001 for CIN III; Wilcoxon signed rank test). In CIN I and CIN II cases EpoR expression was seen in the basal 50% and 70% of the epithelial thickness, whereas in CIN III the full thickness of the dysplastic epithelia showed EpoR expression (Figure 6) ▶ . The mean EpoR staining score increased with increasing grade of dysplasia (slope 51.72 ± 4.61, r 2 = 0.9844, P = 0.0078, linear regression). Prominent EpoR expression was also present in the endothelial and smooth muscle cells of blood vessels (not shown). Immunofluorescence staining for EpoR resulted in a similar immunostaining pattern, including prominent labeling of endothelial and smooth muscle cells of blood vessels (Figure 6C) ▶ .

Cytoplasmic granular Epo immunostaining was seen in 20 of 85 benign squamous epithelia and was restricted to the basal cell layer (Table 2 ▶ , Figures 4C and 5 ▶ ). Epo immunostaining was significantly increased in squamous dysplasias compared to adjacent benign squamous epithelia (P = 0.0313 for CIN I, P = 0.0469 for CIN II, and P = 0.0105 for CIN III; Wilcoxon signed rank test). Epo immunostaining was present in 14 of 15 ISCCs, and was usually localized to small areas adjacent to tumor cell necrosis and keratinization, and to the infiltrating edge of tumors (Figure 7, A and B) ▶ . Epo immunostaining was significantly increased adjacent to necrotic regions (P = 0.0002, Wilcoxon signed rank test) and areas of tumor cell keratinization (P = 0.0039, Wilcoxon signed rank test) (Table 3) ▶ .

HIF-1α immunostaining was predominantly nuclear, although in a few cases weak cytoplasmic reactivity was also observed. In benign squamous epithelia, HIF-1α immunostaining was seen in the parabasal cells and in the middle third of the epithelium in 72 of 85 cases (Table 2 ▶ and Figures 4 and 5 ▶ ). Although overall HIF-1α immunostaining was not significantly different in benign squamous epithelia, dysplasias, and ISCCs (Table 2) ▶ , when compared to the respective benign squamous epithelia, HIF-1α expression was significantly increased in all grades of dysplasia (P = 0.0156 for CIN I, P = 0.0215 for CIN II, and P = 0.0139 for CIN III; Wilcoxon signed rank test). In addition, in dysplastic epithelia HIF-1α immunostaining was not restricted to the middle third of the epithelium as in benign cases (normal staining), but it was also present in the basal and/or the superficial portions (abnormal staining) (Figure 8) ▶ . The difference in the distribution of immunostaining (normal versus abnormal) in benign and dysplastic epithelia was highly significant (P < 0.0001, Fisher’s exact test). In invasive carcinomas, HIF-1α expression was seen in 12 of 15 cases: expression was focal (less than 50% of tumor cells) in 10 cases, whereas in 2 cases it was expressed diffusely in the majority of the tumor cells. When focal, HIF-1α immunostaining was concentrated to areas surrounding necrotic regions, tumor cell keratinization and the infiltrating edge of tumors (Figure 7, C and D) ▶ . HIF-1α immunostaining was significantly increased adjacent to necrotic regions (P = 0.001, Wilcoxon signed rank test) and areas of tumor cell keratinization (P = 0.0156, Wilcoxon signed rank test) (Table 3) ▶ .

In benign squamous epithelia, cytoplasmic bcl-2 staining was found in 81 of 85 cases and was sharply restricted to the basal cell layer (Table 2 ▶ , Figure 5 ▶ ). In CIN I and CIN II, cytoplasmic bcl-2 staining appeared in the more superficial cell layers, however, the percentage of positively staining cells did not differ significantly compared to benign epithelia (P = 0.125 for CIN I and P = 0.1094 for CIN II, Wilcoxon signed rank test). In contrast, in cases of CIN III bcl-2 immunostaining not only extended to the superficial cell layers, but the percentage of positively staining cells was significantly increased as well (P = 0.0007, Wilcoxon signed rank test). Bcl-2 expression was present in 7 of 15 ISCCs, whereas no bcl-2 staining was seen in the rest of the cases.

EpoR expression showed a highly significant correlation with the expression of p16^INK4a (r = 0.8644, P < 0.0001, Spearman’s test) (Figure 9A) ▶ , Epo (r = 0.4801, P < 0.0001, Spearman’s test, not shown), and bcl-2 (r = 0.4632, P < 0.0001, Spearman’s test) (Figure 9B) ▶ . Double-immunofluorescence staining showed co-expression of EpoR and p16^INK4a in squamous dysplasias and invasive carcinomas (Figure 10) ▶ . Double-immunohistochemical staining also showed co-expression of Epo and HIF-1α in invasive carcinomas (Figure 7, E and F) ▶ .

CytologicalSamples of Cervical Squamous Dysplasia

Cytoplasmic EpoR immunostaining was seen in benign squamous cells in 2 of 39 cases. In contrast, dysplastic squamous cells showed strong cytoplasmic EpoR expression in all 16 high-grade squamous intraepithelial lesion and 4 of 5 low-grade squamous intraepithelial lesion cases (Figure 11) ▶ . The difference in EpoR expression between benign and dysplastic squamous cells was highly significant (P < 0.0001, Fisher’s exact test). Double-immunofluorescence staining showed co-expression of EpoR and p16^INK4a in dysplastic squamous cells, whereas benign squamous cells showed no staining with either antibody (Figure 11) ▶ .