Erythropoietin (Epo) is the prime regulator of erythroid cell growth and differentiation. Additionally, a wide variety of normal and malignant nonhematopoietic cells have been reported to express the Epo receptor (EpoR) and in some cases, Epo has been shown to act on these cells, sometimes as a survival/antiapoptotic factor (reviewed in Ref.1).
Reports of erythropoietin receptor (EpoR) expression by a variety of cancer cell types and the use of recombinant human erythropoietin (epoetin) and darbepoetin to treat anemia in cancer patients have raised concern regarding the potential for antiapoptotic effects of these erythropoiesis stimulating agents on the cancer itself.2, 3, 4, 5, 6 Some clinical studies have proven especially worrisome. The Breast Cancer Erythropoietin Trial (BEST) was terminated early because of increased mortality in patients receiving Epo.6 There was an increase in early disease progression. A trial of Epo in head and neck cancer patients treated with radiotherapy resulted in an increase in locoregional recurrence and a decrease in survival in the Epo-treated group.5 Recently, a trial of darbepoetin to enhance radiosensitivity of head and neck cancer was stopped due to a significantly poorer outcome in the darbepoetin treatment group.7, 8
In an accompanying study, we show that each of 4 human ovarian cancer cell lines, including A2780, expresses both the EpoR and Epo at the mRNA and protein levels. Additionally, we show that these EpoR are functional in initiating Epo-dependent intracellular signaling and that an Epo/EpoR autocrine/paracrine mechanism may be operative in enhancing ovarian cancer cell growth.
In the present report, we have studied the effects of Epo on A2780 human ovarian cancer cells in vitro and report that long-term growth of the cells in the presence of Epo resulted in cells with enhanced Epo-mediated signaling. Importantly, they also exhibit an increased resistance to paclitaxel-induced apoptosis, a phenotypic property that is specific for paclitaxel.
Material and methods
Human ovarian cancer cell line A2780 was kindly provided by Dr. Thomas C. Hamilton (Fox Chase Cancer Center, Philadelphia, PA) and was maintained in RPMI-1640 medium (Mediatech, Herndon, VA) supplemented with 10% fetal bovine serum (FBS; Hyclone, Logan, UT), 100 U of penicillin/ml and 100 μg of streptomycin/ml, and with or without 0.25 U of insulin/ml (all from Invitrogen, Carlsbad, CA). In some experiments, cultures also were supplemented with specified concentrations of recombinant human erythropoietin (rhEpo) (Elanex Pharmaceuticals, Bothell, WA). For long-term rhEpo treatment, A2780 cells were grown in the presence of 5 U or 35 U rhEpo/ml for 2 months, resulting in cell lines designated A27802m5U and A27802m35U. Two additional cell lines were used as EpoR− and EpoR+ controls. BaF3 cells are a murine pre-B cell line and do not express the human EpoR while BaF3/EpoR cells are BaF3 cells stably transfected with the human EpoR.9 Both cell lines were maintained in RPMI-1640 containing 10% heat inactivated FBS and supplemented either with 5% WEHI-3B cell conditioned medium (for BaF3 cells) or with 1 U rhEpo/ml (for BaF3/EpoR cells). For Epo control cells, we used BHK (baby hamster kidney) and BHK/Epo cells. BHK cells do not express Epo. BHK/Epo cells are stably transfected with the human Epo gene.
RNA purification and semiquantitative RT-PCR
Total RNA was isolated from the specified cell lines at their subconfluent states using the RNAqueous-4PCR kit (Ambion, Austin, TX). Purified RNA was treated with DNase I to remove genomic DNA contamination and reverse transcribed using a RETROscript kit (Ambion) with random decamer as primer. The resulting cDNA products were used as templates for subsequent PCR analysis. Throughout this article, the quantities of cDNA are expressed corresponding to the quantities of RNA used to derive them (e.g., 10 ng of cDNA corresponds the amount of cDNA produced from 10 ng of RNA).
EPOR expression levels were determined by semiquantitative PCR (sqPCR) using the sense primer (5′-gcgatatcaccgtgtcatccacat, 529–552 bp of EPOR mRNA) and the antisense primer (5′-attctggtacagccacagctggaa, 1,046–1,069 bp of EPOR mRNA). For EPO expression levels, the sense primer (5′- ctgtgctgaacactgcagcttgaa, 346–369 bp of EPO mRNA) and the antisense primer (5′- tcagcagtgattgttcggagtgga, 646–669 bp of EPO mRNA) were used. The levels of GAPDH in the same templates were determined by PCR using the sense primer (5′-aaggctgagaacgggaagctt, 241–261 bp of GAPDH mRNA) and the antisense primer (5′-tccaccaccctgttgctgta, 1,018–1,037 bp of GAPDH mRNA) and were used as normalization controls. All 3 primer sets were designed to include one or more intron regions, so that the PCR product from genomic DNA contamination, if any, could be distinguished by virtue of its different size. No significant products were amplified from control cDNA samples lacking reverse transcriptase.
A modified touch-down PCR cycle was used as follows: 94°C/3 min, 20 cycles of 94°C/30 sec–70°C (−1°C per cycle)/30 sec-72°C/x-sec, y-cycles of 94°C/30 sec-50°C/30 sec–72°C/x-sec, 72°C/5 min, where x = 30, y = 10 for EPOR, x = 20, y = 18 for EPO, and x = 60, y = 17 for GAPDH amplification. The amounts of cDNA and the number of cycles used for the sqPCR were predetermined so that the band intensity of amplified DNA would be in a linear range of amplification. The effective number of cycles was 20 for EPOR, 28 for EPO and 25 for GAPDH, based on the theoretical melting temperature of the primers. PCR products were analyzed by agarose gel electrophoresis and the images were processed using ImageJ software (National Institutes of Health, Bethesda, MD) for quantification.
Cells were seeded into 6-well plates (3–5 × 105 cells/well) and grown for 24 hrs in their usual growth medium. Cells were washed twice with phosphate-buffered saline (PBS) and serum-starved overnight by incubation in phenol red-free RPMI1640 medium (Mediatech) containing 0.5% FBS. Cells were treated with 50 rhEpo U/ml, or with vehicle alone (0.1% bovine serum albumin in PBS), for 0–60 min. The incubation was concluded by washing the cells with ice-cold PBS and placing the plates on ice. Cell lysates were prepared by scraping the cells into 2 × SDS-PAGE sample buffer containing protease inhibitors and protein phosphatase inhibitors (Calbiochem, San Diego, CA), followed by boiling for 5 min. Equal amounts of cell lysate protein were subjected to immunoblotting analysis using rabbit polyclonal anti-phospho-Erk1/2 (Cell Signaling Technology, Danvers, MA) and rabbit polyclonal anti-Erk1/2 (Cell Signaling Technology) as described below.
Cell growth/survival assay
Cells were seeded into 96-well cell culture plates at a density of 3 × 104 cells/ml, 100 μl/well, in culture medium and allowed to adhere for 24 hr. An additional 100 μl of culture medium, without or with specified concentrations of rhEpo, paclitaxel, cisplatin or carboplatin were then dispensed into the appropriate wells. After incubation for specified times, MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide) (Sigma, St. Louis, MO)10, 11, 12, 13, 14, 15, 16, 17 was added to a final concentration of 0.2 mg/ml. After 4 hr of incubation at 37°C, 100 μl of 20% SDS in 0.01 N HCl were added to dissolve the MTT-formazan product, and the incubation was continued overnight. Absorbance at 570 nm was measured using an EL×800 Universal Microplate Reader (Bio-Tek Instruments, Winooski, VT).
A photometric enzyme immunoassay for the quantitative in vitro determination of cytoplasmic histone-associated DNA fragments (mono- and oligonucleosomes) was used (Roche Diagnostics, Indianapolis, IN). Briefly, 20 μl of cell lysate was pipetted into a streptavidin-coated microplate. A mixture of anti-histone-biotin and anti-DNA-peroxidase (POD) (80 μl) was added and incubated for 2 hr at room temperature. ABTS (2, 2′-azino-di(3-ethylbenzthiazoline-6-sulfonate) solution (100 μl) was added as a substrate for POD and, after color development, the absorbance at 405 nm was measured against ABTS solution as a blank (reference wavelength 490 nm).
SDS-PAGE and western blot analyses
Cells were washed twice with ice-cold PBS, incubated on ice for 15 min in RIPA buffer (PBS, 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% SDS) freshly supplemented with Complete Protease Inhibitor Cocktail (Roche Diagnostics) for 15 min and removed from the dishes by scraping. Cell lysates were transferred to individual microcentrifuge tubes and passed through 21 gauge needles to shear the DNA. Sheared lysates were incubated on ice for 30–60 min, centrifuged at 10,000g for 10 min, 4°C and the supernatants were transferred to fresh microfuge tubes. Protein samples were separated on a 12% SDS-polyacrylamide gel and transferred electrophoretically to an Immobilon-P PVDF membrane (Millipore Co., Billerica, MA). The following primary antibodies were used to probe the western blots: mouse monoclonal antibodies anti-Bcl-2 (C-2), anti-Bax (B-9), anti-Bcl-xL (H-5) (all from Santa Cruz Biotechnology, Santa Cruz, CA); and rabbit polyclonal antibody anti-Bcl-10 (Lab Vision Co., Fremont, CA). HRP-conjugated secondary antibodies were obtained from Santa Cruz Biotechnology. Blots were developed using SuperSignal West Dura substrate (Pierce, Rockford, IL). Protein bands were quantified using ImageJ software or Gel-Pro Analyzer™ software (version 3.1, Media Cybernetics, Silver Spring, MD).
Human ovarian cancer cell lines A2780, A27802m35U and A27802m5U express both EpoR and Epo
We examined ovarian cancer cell lines A2780, A27802m35U and A27802m5U for EPOR and EPO expression by semiquantitative RT-PCR. As seen in Figure 1a, we generated a concentration curve using a plasmid containing the entire EpoR coding sequence as control. A specific 541 bp EPOR fragment was amplified. Similarly, a specific 541 bp EPOR fragment was amplified from cDNA generated from the positive control BaF3/EpoR cells, which express the human EpoR, but not from parental BaF3 cells, which do not express human EpoR. Importantly, an identical 541 bp EPOR fragment was amplified from each of the ovarian cancer cell lines, A2780, A27802m35U and A27802m5U. Expression levels were virtually identical in all 3 lines. Amplification of a 797 bp fragment of the GAPDH gene from each sample was included as a control. As seen in Figure 1b, we generated a concentration curve using a plasmid containing the entire Epo coding sequence as control. A specific 324 bp EPO fragment was amplified. Similarly, a specific 324 bp EPO fragment was amplified from cDNA generated from the positive control BHK/Epo cells, which express human Epo, but not from parental BHK cells, which do not express human Epo. An identical 324 bp EPO fragment was amplified from each of the ovarian cancer cell lines A2780, A27802m35U and A27802m5U. Again, expression levels were virtually identical in all 3 lines.
Signaling from the EpoR is enhanced in A27802m35U and A27802m5U cells
Like the EpoR on A2780 (see accompanying publication, Jeong et al.), the EpoR on A27802m35U and A27802m5U are functionally active, an activity that is enhanced over A2780. We incubated A2780, A27802m35U and A27802m5U cells for specified times in the absence or presence of rhEpo, lysed the cells and used western blotting to probe the lysates for phospho-Erk1/2, an important intermediate in the EpoR signaling cascade (Fig. 2). In the absence of rhEpo, relatively low levels of phospho-Erk1/2 were detected in all three cell lines, with the level in A27802m35U being about 3-fold higher than A27802m5U and A2780. Addition of rhEpo triggered a time-dependent increase in phospho-Erk1/2 that reached a maximum in 10 minutes and then returned toward the baseline level in all three cell lines. However, the maximum amplitude differed rather markedly, with A2780 being the lowest (5), A27802m5U significantly higher (17) and A27802m35U the highest (20). Addition of vehicle alone was without effect (not shown).
The degree of enhanced signaling of long-term Epo-treated A2780 cells depends on the length of treatment and is stable to withdrawal of long-term Epo treatment. As seen in Figure 3 (Panel a), the signaling was increased slightly after 4 days of Epo treatment (5 U/ml), and modestly after 15 days. Two months of treatment (i.e., A27802m5U cells) yielded a strikingly robust maximum phospho-Erk1/2 response. As seen in Panel b, the A27802m5U cells still exhibited a robust phospho-Erk1/2 response after withdrawal of long-term Epo treatment for 4 or 15 days.
Long-term Epo treatment leads to increased paclitaxel resistance
Because of the known antiapoptotic action of Epo, we studied the effect of Epo treatment on the response of ovarian cancer cells to paclitaxel, a drug that causes apoptosis and that is widely used in first-line treatment of ovarian cancer.18, 19, 20, 21, 22 A2780 cells and their derivatives have been used frequently to model drug resistance in ovarian cancer.23, 24, 25, 26, 27 They were cultured in the absence or presence of Epo for 24 hr (designated A278024hr cells) or for 2 months with refeeding and passaging as appropriate (designated A27802m35U and A27802m5U cells). For assay, the Epo was removed by washing the cells, and replicate cultures were grown in the absence of Epo and in the presence of specified concentrations of paclitaxel. After 72 hr, cell growth/survival was measured using MTT. As seen in Figure 4a, increasing concentrations of paclitaxel resulted in a corresponding reduction in viability that was virtually identical for A2780 and A278024hr cells. The slight apparent increase in viability exhibited by A278024hr cells at the lowest paclitaxel concentrations was not statistically significant. In contrast, A27802m35U cells exhibited a marked increase in viability at almost all paclitaxel concentrations. The viability curve for A27802m35U cells is right-shifted ∼10-fold in paclitaxel concentration, indicating a ten-fold increase in paclitaxel resistance in these cells. A27802m5U cells were also more resistant to paclitaxel, but less so than A27802m35U cells (Fig. 4b).
We also tested cell growth/survival after 24 hr of paclitaxel exposure followed by 72 hr in its absence (Fig. 5). At each of the relatively higher paclitaxel concentrations tested, A27802m35U cells exhibited paclitaxel resistance compared to A2780 cells. All further studies focused on A27802m35U cells.
Long-term Epo treatment does not result in cisplatin or carboplatin resistance
The paclitaxel resistance exhibited by ovarian cancer cells after long-term Epo treatment does not extend to cisplatin or carboplatin, two other agents widely used in first-line treatment of ovarian cancer but which have a mechanism of action distinct from that of the taxanes. As seen in Figure 6, both A2780 and A27802m35U cells exhibited identical dose-response curves when treated with cisplatin or with carboplatin. This result strongly suggests that the paclitaxel resistance observed after long-term Epo exposure is drug specific.
Epo-induced paclitaxel resistance is due toan inhibition of apoptosis
Epo is well-known to have an antiapoptotic effect on both hematopoietic and nonhematopoietic cells.1 Since paclitaxel treatment results in apoptosis of ovarian cancer cells, we examined the apoptotic response of A2780 and A27802m35U cells to the drug. Cells were plated in replicate cultures in the absence of rhEpo and grown in the absence or presence of specified concentrations of paclitaxel for 24 hr. Apoptosis was quantified by measurement of mono- and oligonucleosome formation. As seen in Figure 7, treatment of A2780 cells with paclitaxel resulted in a marked concentration-dependent increase in mono- and oligonucleosome formation, consistent with markedly increasing apoptosis. In contrast, treatment of A27802m35U cells yielded a much less pronounced dose-response curve, indicating substantially less paclitaxel-induced apoptosis in these long-term Epo-treated cells.
Long-term Epo treated paclitaxel-resistant A27802m35U cells exhibit changes in expression of apoptosis-related proteins
Long-term Epo treated paclitaxel-resistant A27802m35U cells express an antiapoptotic (apoptosis-resistant) phenotype with respect to paclitaxel treatment. This prompted us to measure expression at the protein level of several apoptosis-related proteins (Bcl-2, Bcl-10, Bax and Bcl-xL), hypothesizing that changes in expression levels of one or more of these proteins could explain the A27802m35U phenotype. Erk1/2 was measured as a control. Replicate cultures of A2780 and A27802m35U cells were grown in the absence of rhEpo and in the absence or presence of a low (0.75 nM) or high (50 nM) concentration of paclitaxel for 24 hr. The cells were lysed, and the lysates were subjected to SDS-PAGE and western blotting (Fig. 8). Among the 4 apoptosis-related proteins surveyed, the expression of Bax and Bcl-xL (known to mediate Epo's direct antiapoptotic action28, 29, 30, 31, 32, 33, 34, 35) were virtually identical in the paclitaxel-sensitive A2780 and the relatively paclitaxel resistant A27802m35U cells. However, expression of both Bcl-2 and Bcl-10 was substantially lower in the A27802m35U cells, and paclitaxel treatment may have reduced these levels further. Similar results were obtained in a repeat experiment. There are mechanistic bases for the involvement of both Bcl-2 and Bcl-10 in the paclitaxel resistance exhibited by A27802m35U cells (see Discussion). After paclitaxel treatment at either low or high concentration, the expression levels of all 4 proteins remained relatively stable.
Inhibition of Erk1/2 phosphorylation does not restoreBcl-2 levels nor does it change paclitaxel resistance
To determine if the enhanced Erk1/2 phosphorylation seen in long-term Epo treated cells was the primary cause of Bcl-2 down-regulation and/or of paclitaxel resistance, we used the MEK inhibitor PD98059, which we have shown previously to block MEK-dependent phosphorylation of ERK1/2 in the Epo signaling cascade.36, 37 As seen in Figure 9, addition of either 4 μM or 50 μM PD98059, which are the IC50 concentrations that inhibit Erk1 and Erk2 phosphorylation, respectively,38, 39, 40 did not increase Bcl-2 protein levels in any of the 3 cell lines, despite confirmation that endogenous Erk1/2 phosphorylation was inhibited significantly. The same results were obtained in the absence or presence of 50 nM paclitaxel. The moderate down-regulation of Bcl-2 by PD98059 seen in these experiments has been reported previously in human pancreatic cancer cells.41 Similarly, treatment of the cells with PD98059 did not cause any changes in paclitaxel resistance (not shown).
In the present study, we have shown that long-term Epo treatment of ovarian cancer cells results in a stable alteration in phenotype characterized by enhancement of Epo signaling and increased resistance to paclitaxel but not to cisplatin or carboplatin. This alteration in phenotype remained even after cessation of long-term Epo treatment. The mechanism of the Epo effect leading to enhanced signaling and paclitaxel resistance remains under investigation, but at least two possibilities deserve prominent consideration. The first is the possibility that Epo treatment may lead to the selection of paclitaxel-resistant cells due to preferential survival/growth in the presence of Epo. Although our short-term growth studies did not demonstrate an Epo effect (see accompanying publication, Jeong et al.), growth over several days–weeks may be needed to detect an effect. A second possibility rests in the likelihood of Epo-induced epigenetic effects leading to changes in gene expression that stabilize. Obviously, a combination of these and other mechanisms must be considered. Further work will be needed to clarify this, as yet, unclear mechanism. Although the concentrations of rhEpo used for long term treatment of A2780 cells were supraphysiological, pharmacokinetic studies of rhEpo in humans have shown that serum levels of 3–8 U/ml are reached in humans subjects after intravenous administration of clinically relevant doses of the hormone.42 These concentrations, achieved clinically, are in the range of those concentrations shown to be active in long term Epo treatment of A2780 cells in our study.
The pathways through which paclitaxel and cisplatin induce apoptosis in ovarian cancer cells may differ, thus providing a means for the development of relative resistance to one drug versus the other. Paclitaxel reportedly operates through the caspase-3 pathway.43 In addition, activation of caspase-3 may results from both a JNK/SAPK dependent and a JNK/SAPK independent pathway.44 In contrast, cisplatin reportedly operates through both caspase-3 dependent and independent pathways.45 Furthermore, the drug results in reduced expression of X-linked inhibitor of apoptosis (XIAP) and increased Akt cleavage.46 Akt appears to mediate cisplatin resistance by way of its interaction with p53.47 Epo has been reported to affect cisplatin sensitivity in other tumor cells, however. In one study, Epo was used to increase the hematocrits of transplanted SCID mice, and an increase in cisplatin sensitivity of a papillary serous carcinoma was observed.48 In contrast, Belenkov et al. studied the effect of Epo on the human malignant glioma cell line U87 and the primary cervical cancer cell line HT100. The authors reported that the addition of Epo induced both cell lines to become more resistant to ionizing radiation and to cisplatin.49
Separate from the mechanism by which Epo treatment resulted in paclitaxel resistance is(are) the mechanism(s) of paclitaxel resistance itself. Our results suggest that the observed paclitaxel resistance may be due, at least in part, to altered expression of Bcl-2 and Bcl-10, two proteins involved in apoptosis. The mechanisms by which a reduction in each of these two proteins leads to apoptosis resistance are probably different. Bcl-10, which has not been studied before in ovarian cancer cells, is involved in antigen receptor induced lymphocyte proliferation and has been assigned a pro-apoptotic action.50, 51, 52 Thus, it would be expected that a reduction in its expression should be antiapoptotic. The potential mechanism by which a reduction in Bcl-2, generally considered to be an antiapoptotic protein,53, 54, 55, 56 might lead to an increase in paclitaxel resistance is, presumably, more complex. However, in this regard, Ferlini et al. have reviewed the evidence for a potential dual role for Bcl-2 in paclitaxel-induced apoptosis and have provided experimental data supporting the concept that paclitaxel acts not only on microtubule turnover but also on the mitochondrion, and that Bcl-2 is its target.57 A direct action of paclitaxel on isolated mitochondria had already been reported.58 Ferlini et al. showed that A2780 cells resistant to increasing concentrations of paclitaxel exhibited correspondingly reduced Bcl-2 expression, an effect that was reversible upon artificially increasing Bcl-2 expression. Our observation that the increased paclitaxel resistance of long-term Epo treated A2780 cells is correlated with decreased Bcl-2 expression (see Results) is supported by the data of Ferlini et al. and their cited earlier observations. Interestingly, Epo has been reported to up-regulate Bcl-2 in retinal ganglion cells,59 HCD-57 erythroid cells60 and rat cardiac myocytes.61
Recombinant Epo (epoetin) and its somewhat longer acting form (darbepoetin) are being used increasingly at the clinical level in ovarian cancer patients and other cancer patients as well. A randomized, controlled multicenter trial of Epo to prevent anemia and reduce red blood cell transfusions in 122 ovarian cancer patients receiving platinum-based therapy reported good Epo efficacy that was well-tolerated,62 although more recently a “definitive” phase III trial has been called for.63 A correlation between anemia and decreased survival in ovarian cancer has been described.64
The potential clinical applications and consequences of our observation relating to the use of paclitaxel (and taxanes in general) in ovarian and other cancers will depend upon whether the in vitro studies presented here are borne out in vivo. The preliminary observation that Bcl-2 expression may be decreased in clinical samples from paclitaxel-resistant ovarian cancer patients57 serves to heighten concern regarding the long-term use of epoetin and darbepoetin in this and other clinical settings and should prompt the design of appropriate clinicopathological studies.
We thank Dr. Steven Cannistra for instructive discussions during the inception of this work. This work was supported in part by NIH Grant R01 CA 89204, DOD Grant DAMD17-03-1-0233, grants from the Elsa U. Pardee Foundation and the Gustave and Louise Pfeiffer Research Foundation, and by a sponsored research agreement from DNAPrint Pharmaceuticals, Inc. to A.J.S. and by the Slovak Science and Technology Assistance Agency contract No. APVT-20-012104 and by grant VEGA No. 1/3253/06 to P.S.