Expression and function of erythropoietin receptors in tumors

Implications for the use of erythropoiesis-stimulating agents in cancer patients



Safety concerns surrounding the use of recombinant human erythropoietin (Epo) to treat anemia in cancer patients were raised after 2 recent clinical studies reported a worse survival outcome in patients who received epoetin α or epoetin β compared with patients who received placebo. Although those findings contrasted with previous clinical studies, which demonstrated no difference in survival for cancer patients who received erythropoiesis-stimulating agents (ESAs), some investigators have suggested a potential role for ESAs in promoting tumor growth through 1) stimulation of Epo receptors (EpoR) expressed in tumors, 2) stimulation and formation of tumor vessels, and/or 3) enhanced tumor oxygenation. The first and second hypotheses appeared to be supported by some EpoR expression and ESA in vitro studies. However, these conclusions have been challenged because of poor specificity of EpoR-detection methodologies, conflicting data from different groups, and the lack of correlation between in vitro data and in vivo findings in animal tumor models. For this report, the authors reviewed the biology of EpoR in erythropoiesis and compared and contrasted the reported findings on the role of ESAs and EpoR in tumors. Cancer 2007. © 2007 American Cancer Society.

Concerns that recombinant human erythropoietin (rHuEpo) may adversely affect the survival of cancer patients through promoting tumor growth recently were raised by 2 randomized clinical trials that used rHuEpo to prevent anemia in patients with breast or head and neck cancer.1, 2 In those trials, decreased survival and increased tumor progression were observed in patients who received epoetin α or epoetin β compared with patients who received placebo. These outcomes were in contrast to many clinical trials, which demonstrated that erythropoiesis-stimulating agents (ESAs) are safe and efficacious for the treatment of anemia in cancer patients.3, 4 These trials were criticized for poor study design and execution.4 However, a number of hypotheses were proposed to explain the potential role of ESAs in promoting tumor proliferation and reduced survival including, stimulation of tumor expression of the erythropoietin (Epo) receptor (EpoR), tumor neovascularization, enhanced tumor oxygenation, and, most recently, the role of coadministered iron,5 although that topic is not addressed herein.

This review reflects conclusions from a comprehensive survey of the literature that investigated the expression and function of EpoR in primary tumors and in transformed cell lines and endothelial cells. Multiple broad searches of the biologic and medical literature were performed on the Ovid system using multiple databases. Many of the citations described the effects of ESAs related to receptor expression and signaling, tumor cell proliferation, tumor growth, or tumor progression. The majority of the articles supported 1 of 2 themes: 1) a protumor-promoting effect of ESAs and 2) no tumor-promoting effect and/or an antitumor effect of ESAs.

Biology of EpoR

EpoR is a type-1, single-transmembrane receptor that is expressed as several forms in erythropoietic progenitor cells, including a full-length form (F-EpoR), a truncated form (T-EpoR), and a soluble form (S-EpoR). T-EpoR and S-EpoR contain the extracellular Epo-binding domain, but alternative splicing of transcripts truncates the cytoplasmic or transmembrane domains (Fig. 1A,B).6 T-EpoR knock-in and transgenic mice had sustained erythropoiesis; however, erythroid progenitor survival and proliferation at low concentrations of Epo and during stress-induced erythropoiesis were severely compromised.7, 8 It has been reported that S-EpoR acts as an antagonist in neuronal tissues by competing with F-EpoR for binding to erythropoietin.9 The physiological roles for these EpoR variants have not been established.

Figure 1.

The erythropoietin receptor (EPOR) gene and alternatively spliced forms of EpoR are illustrated. (A) The EPOR gene is illustrated with alternative transcripts identified: Exon numbers are shown. * indicate stop codons. (B) EpoR protein forms generated with proposed functions. EpoR peptides used to generate commercially available antibodies are indicated by the blue boxes. Tyrosines (Y) in the cytoplasmic domain essential for signaling are shown. Regions of nonhomology to full-length EpoR (F-EpoR) in truncated EpoR (T-EpoR) (lilac) and soluble EpoR (S-EpoR) (magenta) are shown. JAK2 indicates Janus kinase-2; aa, amino acid.

Translocation of EpoR to the cell surface is an inefficient process, resulting in <1% of total cellular F-EpoR molecules reaching the cell surface of hematopoietic cells. This is a consequence of the short half-life of EpoR protein (1–2 hours), inefficient processing for surface expression, and protein degradation within the endoplasmic reticulum, proteosome, and lysosomes.10–14 Accessory factors that are required for EpoR trafficking to the surface also may be at limiting concentrations. For example, Epo-dependent and Epo-independent subclones of a hematopoietic cell line reportedly expressed EPOR transcripts and protein, but only Epo-dependent clones expressed surface receptor.15 In addition, enforced expression of EPOR messenger RNA in a human hematopoietic cell line did not increase total surface EpoR protein.16 One such accessory protein, Janus kinase-2 (Jak2), binds EpoR in the endoplasmic reticulum, induces correct protein folding, promotes surface expression in hematopoietic cells, and is essential for EpoR signaling.17

Epo or ESAs binding to EpoR homodimers activates Jak2, inducing a signaling cascade to promote erythroid progenitor survival, proliferation, and differentiation (for review, see Wojchowski et al.18) (Fig. 2). Src homology region 2 domain-containing phosphatase 1 (SHP-1) and suppressor of cytokine signaling (SOCS) proteins CIS, SOCS-1, and SOCS-319, 20 are negative regulators of EpoR signaling. Absence of negative regulation of EpoR signaling is associated with familial erythrocytosis because of cytoplasmic truncations of EpoR that remove suppressor binding sites.21, 22

Figure 2.

Erythropoietin receptor (EpoR) signal-transduction pathways are illustrated in erythroid progenitors and cell lines. The signal-transduction pathways used in erythroid cells are shown with physiologic effects indicated on the right. JAK2 indicates Janus kinase-2; STAT5, signal transducer and activator of transcription; MAPK, mitogen-activated protein kinase; AKT, protein kinase C; IRS-2, insulin receptor substrate 2; BAD, Bcl-2 antagonist of cell death; FKHR-LI, forkhead transcription factor; CDK, cyclin-dependent kinase.

EpoR Expression and Signaling in Tumors

EPOR transcription

Tumor cell lines and tumor tissues derived from neoplasias of the breast,23–26 kidney,25, 27 colon,25, 26 stomach,26 pancreas,25, 26 prostate,25, 26, 28 female reproductive organs,25, 26, 29–31 liver,25, 32 lung,26 brain,25, 32 melanocytes,25, 26, 33, 34 and head and neck35 and from hematopoietic neoplasias25, 26 reportedly were capable of transcribing the EPOR gene (as determined by nonquantitative reverse transcriptase-polymerase chain reaction [RT-PCR] analysis), and it has been suggested that this expression is indicative of a potential role in tumor progression. One study reported that 20 of 23 tumor cell lines expressed EPOR transcripts and correlated with immunostaining of EpoR protein in most cell lines.25 However, in a urinary bladder carcinoma cell line and in immortalized keratinocytes, no EPOR transcripts were detected, although the tumor cell lines reportedly were positive for EpoR protein. These discrepancies may have resulted from the lack of specificity of the anti-EpoR antibody used in that study. The methods used in these RT-PCR studies do not distinguish between the different splice variants of EpoR; thus, it is not clear whether functional EpoR is expressed in these cells. Indeed, high levels of alternatively spliced transcripts that encode attenuated or antagonistic EpoR forms have been reported in tumor cell lines.36

Using real-time, quantitative RT-PCR, EPOR transcripts have been detected in normal tissues, including kidney, heart, brain, endothelium, and smooth muscle,37 but are at significantly lower levels than the levels detected in bone marrow that contains the Epo-responsive erythroid progenitors. Similarly, EPOR transcript levels in breast, colon, lung, lymphoma, ovary, prostate, stomach, ileum, and kidney tumor tissues and tumor cell lines also were significantly lower than the transcript levels in bone marrow and were no higher than the levels observed in normal counterpart tissues.38 It also was reported that EPOR transcript levels were at equivalent levels in matched tumor and nontumor samples from patients with lung or colon cancer.38 These findings are in concordance with reports by other groups that have used semiquantitative means to evaluate EPOR transcription. For example, EPOR transcripts in neoplastic prostate were equivalent to levels in normal prostate28 and were equivalent in head and neck squamous cell carcinoma tumors and paired normal samples.39 Taken together, these data suggest that EPOR transcript levels in tumor cells are not elevated above the levels observed in normal tissues, thus suggesting that there is no selective advantage for tumors to overexpress the EPOR gene.

EpoR protein

Most studies that have reported high levels EpoR protein expression in tumors have used commercially available antibodies in immunoblotting and immunohistochemistry experiments.23–25, 27, 33, 34, 36, 40 Conclusions from these studies should be reexamined because of nonspecificity of the antibodies and confusion over the identity of the EpoR protein. The reported sizes of EpoR using commercial anti-EpoR antibodies ranged from 66 kilodaltons (kD) to 78 kD. However, the calculated size of the EpoR protein is 52.6 kD, and maturation with the addition of carbohydrate could increase the size to approximately 57 kD. However, when a tagged EpoR was overexpressed, antibodies against the protein tag and EpoR identified EpoR that was ≈59 kD.41, 42 This was confirmed further by protein microsequencing.41 All commercial anti-EpoR antibodies detected multiple proteins (up to 20 different bands) that differed in size compared with EpoR.41 The commonly used Santa Cruz polyclonal antibody C-20 detected a 66-kD band, which investigators believed was EpoR; however, when the 66-kD band detected by C-20 was protein sequenced, no EpoR peptides were identified.41 Heat-shock proteins (HSP) with homology to the EpoR peptide used to generate the C-20 antibody were identified in the 66-kD band, and HSP70 peptides were able specifically to block binding of C-20 to the 66-kD protein41 and reduced or abolished immunoreactivity of C-20 to nonsmall cell lung carcinoma tumor sections.43 This finding of cross-reactivity to non-EpoR proteins was not surprising, because the various antibody preparations contain polyclonal antipeptide antibodies. It is noteworthy that HSP70 is a potent survival factor for cells under stress; it is highly expressed in tumors; it is associated with metastasis, poor prognosis, and resistance to radiation therapy and chemotherapy; and it increases the tumorigenicity of cancer cells in animal models (for review, see Schmitt et al.44). Therefore, a recent C-20 EpoR expression study suggesting that EpoR expression in tissue sections was correlated with tumor progression and worse survival45 may have identified an association between the expression of HSP and prognosis.

A limited number of studies have used Epo-binding assays to investigate the presence of EpoR on the surface of tumor lines.27, 46, 47 In the tumor lines in which binding was reported, the binding affinity of EpoR was substantially lower than the binding affinity to primary erythroid progenitors and cell lines.27, 46–50 A rat pheochromocytoma cell line reportedly expressed EpoR with a binding affinity that was 160-fold lower than what was reported previously for EpoR on hematopoietic cells,46 the significance of which is uncertain. Other studies have reported no or extremely low levels of surface EpoR (≈17 Epo binding sites)38, 51, 52 on tumor cell lines. This was despite the finding that F-EpoR protein was detected by immunoblot analysis, in 1 instance at levels equivalent to those detected in hematopoietic cells.38, 51 Thus, EpoR protein may be synthesized but may not be transported to the cell surface at detectable levels or at low levels because of limiting accessory trafficking factors.

EpoR signaling

Some in vitro studies have reported a general increase in tyrosine phosphorylation in cell lines when exposed to suprapharmacologic doses of rHuEpo. It was suggested that this was a protumorogenic effect, but no specific pathways were determined.23, 29 It was reported that rHuEpo induced modest translocation of nuclear factor κB (NF-κB) to the nucleus and up-regulation of antiapoptotic gene transcription. Again, it was suggested that this was indicative of a survival response, although no survival studies were performed.32 These data were contradicted by the finding that rHuEpo inhibited NF-κB-induced antiapoptotic gene transcription in cell lines to enhance chemotherapy-induced cell death.53 None of those studies used appropriate vehicle controls; therefore, the results obtained may have been because of non-Epo proteins that were present in the diluent.

Although it was concluded in a xenograft study in vivo that Epo-EpoR interaction activated signal transducer and activator of transcription 5 (STAT5) in tumors, no signaling occurred in the parental cell lines in vitro.26 Because STAT5 is activated not only by Epo but also by numerous other cytokines (for review, see Seidel et al.54), the significance and relevance of these data to Epo-EpoR signaling are unclear. In an attempt to address the specificity of response, the investigators injected S-EpoR and anti-Epo antibodies directly into tumors in ex-vivo organ cultures. They reported that Jak2 and STAT5 signaling were reduced because of the inhibition of autocrine or paracrine Epo-EpoR signaling in tumors.30 Multiple sequential injections with large quantities of S-EpoR protein (0.3–1.6 μg/mg) and anti-Epo antibodies (32 μg/mg) into the tumors were required to achieve a reduction in phosphorylated STAT5 from 0 to approximately10-fold. The lack of appropriate negative controls, such as antibody isotypes or irrelevant protein controls, limits the ability to reach definitive conclusions from that study. In summary, experiments that analyze isolated signaling pathways ex vivo are difficult to interpret in terms of their physiologic or pathologic significance, because they are complicated by confounding influences, such as the presence of additional growth factors, hypoxia, the presence of endotoxins in recombinant proteins, and potential off-target effects of antibodies or soluble receptors, especially when they are used at high doses.

In Vitro Functional Studies

Tumor cell line proliferation in response to rHuEpo

Studies that assess the biologic response of tumor cell lines to rHuEpo may be more informative than those that measure only intracellular and cell-surface expression of EpoR or EpoR signaling pathways. Numerous investigators have assessed the proliferation of tumor cell lines in response to rHuEpo in vitro (Table 1). Pharmacologic levels of serum rHuEpo are approximately 1 U/mL after subcutaneous administration of a single dose of 600 U/kg (at the high end of recommended therapeutic dosing).55 Concentrations of rHuEpo up to 1000-fold greater than those used clinically did not lead to proliferation of 6 transformed cell lines, including those derived from breast, pancreatic, prostate, renal, and myelogenous tumors (in vitro treatment with rHuEpo up to 1000 U/mL), even though EpoR transcripts and protein reportedly were detected in each cell line.25 Likewise, in 8 other studies,56–63 there was no observed biologic response to rHuEpo, as measured by proliferation or clonogenic growth in 68 tumor cell lines. Although several of those studies did not investigate whether tumor cell lines expressed surface EpoR, 1 study reported that cell lines that expressed surface EpoR could not be induced to proliferate when they were exposed to rHuEpo.58 Selzer et al. reported that 50% of spontaneously transformed melanocyte lines expressed EpoR but that their growth characteristics in culture were no different than cell lines that did not express the receptor.33

Table 1. In Vitro Proliferation Studies Using Recombinant Human Erythropoietin
Cell linesTissue originEPO dose, U/mLResultStudy
  1. EPO indicates erythropoietin; AML, acute myeloblastic leukemia; ALL, acute lymphoblastic leukemia; NS, not statistically significant; GI, gastrointestinal.

81 AML and 5 ALL patient samplesBlood11.15-Fold increased proliferation (NS)Takeshita et al., 200064
MCF-7, BT-549Breast101.25-Fold increased proliferationAcs et al., 200123
FO-1, SK-MEL28, PLC, LXF-289, KTCTC-1M, KTCTL-26A, KTCTC-30, MCF-7, MDA-MB-231Skin, kidney, mammary, liver, colon, lung1–5000No proliferation; no colony-stimulating effectMundt et al., 199259
H69, N417, OCIM-1, GBLHU2, MCF-7, HL-50, KG1, HEL, PLB-985, K562Lung, stomach, brain, breast, blood0.5–5No proliferation; no colony-stimulating effectPedrazzoli et al., 199261
HTB-43, CCL-23, CCL-17, HTB-119, HTB-120, HTB-19, HTB-22, KATO, HTB-38, WiDr, CCL 187, HepG2, MIA PaCa, HTB-77, HTB-36, CRL-1427, U87MG, 87HG31, CCL-127, B, HTB-45, HTB-44Head/neck, lung, breast, stomach, GI, liver, pancreas, ovary, brain, prostate, kidney0.01–100No proliferation; no colony-stimulating effectBerdel et al., 199157
HTB-22, HTB-120, HTB-38Breast, lung, colorectal0–100No colony-stimulating effectBerdel et al., 199256
K562, HEL, HL-60, KG-1, PLB-985, H69, N417, MCF-7, OCUM-1, GBL-HU12Blood, lung, breast, stomach, brain0.5–10No colony-stimulating effectRosti et al., 199360
K562, HepG2, KTCTL-30, NMB, S117, RT112Blood, liver, kidney, neural, thyroid, bladder10–1000No proliferationWestphal et al., 200225
Caki-2, 786-0, RAG 3Kidney0.5–100From 1.25- to 4-fold increased proliferationWestenfelder and Baranowski, 200027
ACHN, Caki 1, CEM, HCT116, HL60, K567, U266Renal, blood, GI5–20No proliferationLiu et al., 200458
H838Lung12.5No proliferationDunlop et al., 200662
SHSY-5Y, LAN-5, Kelly, SK-N-LO, KS-N-MC, SH-SHEPBrain0.00125 to 10No proliferationRossler et al., 200463
PC3, LNCaP, 267B1, X/267B1, Ki/267B1Prostate0–100From 1.5-to 3.5-fold increased proliferation in some lines; no proliferation in othersFeldman et al., 200628

In contrast, proliferation of tumor cells in response rHuEpo has been reported in other studies.23, 64 However, the biologic significance of these data is uncertain for a number of reasons. First, the increase in proliferation was within the range of the background noise observed in typical cell proliferation assays. Second, no dose/response relations were established. Furthermore, the proliferative response observed in a breast tumor cell line23 was not observed in 4 other studies that used the same cell line.56, 57, 59, 60 To date, there appear to be only 2 studies that have investigated the relation between rHuEpo dose and tumor cell line proliferation.27, 28 In those studies, the proliferation of a murine cell line and 7 human renal and prostate cell lines increased 1.25 to approximately 4-fold after treatment with 0.5 U/mL to 100 U/mL rHuEpo in both serum-free and serum-containing media.27, 28 The absence of vehicle controls and the general use of suprapharmacologic doses of rHuEpo in those studies limits their usefulness. These data contrast with the effect of rHuEpo on an erythroid cell line, in which pharmacologically relevant doses between 0.01 U/mL and 0.4 U/mL rHuEpo stimulated a 650% increase in proliferation.65

Tumor cell line survival in response to rHuEpo

Another proposed mechanism of Epo action on tumor cells is the inhibition of apoptosis. Studies29 reported that rHuEpo (25–200 U/mL) in combination with cisplatin improved HeLa cell survival by approximately 5% to 60% compared with cisplatin alone. However, only 10% of the cell death induced by cisplatin was attributed to apoptosis, and 200 U/mL rHuEpo (a concentration never achieved in humans) appeared to have a maximal effect that only reduced this apoptotic fraction to 5%. Treatment of hematopoietic EpoR-expressing 32D cells with rHuEpo (5 U/mL) reportedly suppressed cisplatin-induced growth arrest and apoptosis at 24 hours but, subsequently, resulted in increased necrotic cell death when cell cultures were monitored for longer periods.66 Suprapharmacologic levels of rHuEpo (>1 U/mL, the maximum concentration observed with rHuEPO in clinical studies) were required to observe these effects. In similar studies, the addition of rHuEpo (30 U/mL) to U87 glioblastoma and HT100 cervical carcinoma tumor cell lines reportedly made the cells more resistant to ionizing radiation and cisplatin, although no EpoR-mediated signaling was investigated.67 In contrast to those reports, other investigators have observed that the magnitude of cell killing by cisplatin in a variety of cell lines that were pretreated with rHuEpo (10 U/mL) was not reduced compared with treatment using cisplatin alone.58 This was not because the cells were unresponsive to rHuEpo, because activation of mitogen-activated protein kinase was reported.58 Taken together, the results from studies that investigated the protective effects of rHuEpo on tumor cells in vitro are conflicting.

ESAs in Animal Models of Tumor Growth

Erythropoietins in animal tumor models

Animal models are more indicative of a physiological role of ESAs on tumor growth. The in vivo tumor model studies with ESA treatment generally can be grouped into 1 of 3 categories: 1) tumor regression, 2) enhancement of tumor-ablative therapies, and 3)no enhancement of tumor-ablative therapies (Table 2). In all 23 in vivo studies, ESA administration alone had no tumor-promoting effect.

Table 2. Effect of Erythropoiesis-stimulating Agents in Xenograft or Syngenic Tumor Models
Tumor type and originEPO/DA doseTumor and survival outcomesStudy
  1. EPO indicates erythropoietin; DA, darbepoetin alpha; QD, once daily; TIW, 3 times per week; RT, radiotherapy; CT, chemotherapy; PT, photodynamic therapy; QW, once weekly; Q2W, every 2 weeks; IL-12, interleukin-12; Q3D, every 3 days.

Tumor regression alone
 Murine myelomas MOPC-315, 5T33 MM30 U EPO QDTumor regression and prolonged survivalMittleman et al., 200170
 Murine BCL-1 leukemia/lymphoma30 U EPO QDTumor regression and prolonged survivalMittleman, 2003102
Enhanced tumor ablation
 Neurogenic sarcoma ENE2750 U/kg EPO TIWImproved RT therapy in anemic miceStuben et al., 2003103
 Lewis lung carcinoma60 U/kg EPO (2 doses)No effect alone; enhanced CTSigounas et al., 2004104
 Rat DS-sarcoma1000 U/kg EPO TIWNo effect alone; improved ablative RTThews et al., 1998105
 Ovary adenocarcinoma20 U EPO TIWNo effect alone; improved CTSilver and Piver, 199968
 Glioblastoma HTZ II1000 U/kg EPO TIWNo effect alone; improved RT in anemic miceStuben et al., 2001106
 Rat DS-sarcoma1000 U/kg EPO TIWNo effect alone; improved CT in anemic ratsThews et al., 2001107
 Colon adenocarcinoma1000 U/kg EPO QDRestored PT in anemic miceGolab et al., 2002108
 Human glioblastomas GBM, Nan1 and U87300 U/kg QDNo effect on tumor alone; enhanced RTPinel et al., 200472
 Murine SCC VII squamous cell carcinoma and RIF-1 fibrosarcoma30 μg/kg DA QW or Q2WNo effect alone; improved RT in anemic miceNing et al., 2005100
 Lewis lung carcinoma10 μg/kg DA QWNo effect alone; improved CTShannon et al., 2005109
No enhanced tumor regression
 Murine MmB16 melanoma20 U EPO twice dailyNo effect alone; no enhanced IL-12 therapyGolab et al., 1998110
 Rat DS-sarcoma1000 U/kg EPO TIWNo effectKelleher et al., 199699
 Rat R3230 mammary carcinoma3 μg/kg DA TIWNo effect to enhance RTKirkpatrick et al., 200674
 Murine C26-B adenocarcinoma25 U EPO QD to 25 U TIW per mouseNo effect on tumor; decreased body weight lossvan Halteren et al., 200473
 Rat R3230 mammary carcinoma, murine CT26 colon carcinoma, human HCT-116 colon carcinoma, human FaDu head and neck carcinoma2000 U/kg EPO TIWNo effect aloneHardee et al., 200588
 RatR3230 mammary carcinoma2000 U/kg EPO TIWNo effect alone; no enhanced CTHardee et al., 200669
 Human breast carcinomas MDA-MB-231 and MCF-72.5 μg/kg epoetin α, 7.5 μg/kg DA, and 2.5 μg/kg epoetin βNo effect alone; no enhanced CTLaMontagne et al., 200651
 Head and neck squamous cell carcinoma LU-HNxSCX-7400 U/kg epoetin α Q3DNo effect alone; slight increased tumor growth with surgical transactionKjellen et al., 200675
 Human squamous cell A431, colorectal carcinoma HT25150 U/kg EPO TIWNo effect alone; enhanced CTTovari et al., 200589

In some (but not all) studies, ESAs reportedly restored the effectiveness of radiation therapy, photodynamic therapy, and chemotherapy in anemic animals that had tumors (Table 2). For example, severe combined immunodeficiency (SCID) mice bearing small, subcutaneous ovarian tumors showed a significant decrease in tumor progression after treatment with rHuEpo plus cisplatin compared with control animals.68 In contrast, rats bearing tumors that reportedly expressed and signaled through Epo-EpoR did not demonstrate an enhanced tumor-ablative effect with chemotherapy.69 There was no enhanced tumor promotion in the presence or absence of chemotherapy. In another setting, rHuEpo reportedly induced tumor regression in a murine model of myeloma by provoking an antitumor immune response.70 The same investigators recently reported prolonged survival in 6 patients with advanced multiple myeloma when they received rHuEpo with or without chemotherapy.71

Others explored rHuEpo in combination with fractionated radiation in a murine cancer cachexia model: They concluded that rHuEpo had a beneficial effect by contributing to radiosensitization and/or reducing weight loss.72, 73 In contrast, no radiosensitizing effect was observed in a rat tumor model with darbepoetin alpha (DA) at a high dose, even though there was an increase in tumor oxygenation.74 Therefore, the data surrounding tumor effects of ESAs were consistent in all models: There was no evidence of tumor enhancement. However, the ability of ESAs to enhance the effect of chemotherapy or radiotherapy was observed inconsistently.

One group reported that epoetin β had no effect on tumor growth alone but slightly increased tumor growth in a nonvalidated transection model of mock surgery.75 The relevance of that study is unclear based on the nonvalidated model that was used.

EpoR agonists or antagonists in tumor growth models

Some investigators have utilized anti-Epo antibodies or soluble EpoR to treat tumors in nude mice or rats to investigate the role of the Epo-EpoR axis. Injection of an anti-Epo antibody or soluble EpoR directly into female reproductive organ-derived tumors in nude mice reportedly reduced tumor size.31 It has been suggested that tumor regression is related to apoptotic death of tumor and capillary endothelial cells. The relevance of these results is unclear, because essential controls were lacking: No isotype or irrelevant protein controls were examined, and no inhibition of EpoR signaling was investigated or documented. In subsequent experiments, several tumor cell lines were transplanted into nude mice and treated with the Epo antagonistic (EMP-9) and agonistic (EMP-1) peptides.26 Administration of EMP-9 to mice with stomach choriocarcinoma or melanoma xenografts reportedly inhibited angiogenesis and tumor cell survival; opposite effects were reported with EMP-1. The results have been challenged on theoretical grounds because of the low in vivo potency of EMP-1 (compared with rHuEpo, its erythropoietic activity is reduced >65,000-fold76) and lack of hematocrit changes in animals that were administered EMP-1 or EMP-9 compared with control animals that received saline.26 Thus, it is not clear whether tumor effects unrelated to EpoR agonism or antagonism account for these results.

Local injection of anti-Epo antibodies, soluble EpoR, and the Jak2 antagonist AG490 reportedly increased the death of syngenic tumors in rats.24 Those data are difficult to interpret, because essential controls were not included: Matched antibody isotype and unrelated soluble receptor controls were not performed. Furthermore, AG490 is not specific to Jak2 and is a more potent antagonist of EGFR (approximately 100-fold),77, 78 and it also is active on Bcr-Abl,79 Her2,77 and Jak3.80

Overexpression of Epo in Mice and Humans

Biologic disorders that alter the levels of Epo in humans and animals provide additional insights into the potential role of ESAs in tumor induction or progression. Both primary congenital disorders associated with increased Epo production and mutations in EPOR that lead to hypersensitivity to Epo are associated with erythrocytosis in humans.21, 81 However, a greater cancer incidence has not been observed in patients with familial and congenital polycythemias. Transgenic mice have been developed that express high levels of Epo.82 There was no increase in tumorigenesis, including erythroleukemia, in those animals after 2 years of observation.

Epo in Tumor Angiogenesis and Vasculogenesis

EpoR expression has been reported in neovessels in explanted human myocardium,83 chick chorioallantoic membrane,40 human umbilical cord,84 human placenta,84 tumor xenograft models, and primary human tumors.26, 30, 85 However, most of those studies should be viewed with caution, because they relied on immunoblot or immunohistochemistry analyses that used nonspecific anti-EpoR antibodies (described above). However, using Epo-binding studies, cultured endothelial cells reportedly expressed between 10,000 and 27,000 low- affinity (Kd 860 to 3800 pM) receptors per cell.48, 86

In vitro, rHuEpo reportedly stimulated the proliferation of primary endothelial cells48, 86 and endothelial cell lines.40 It is noteworthy that high doses of rHuEpo (10U/mL and 30 U/mL) induced the release of vascular endothelial growth factor (VEGF) in some cultured tumor cell lines,32 and it has been suggested that VEGF acts in a comitogenic fashion with rHuEpo.87 Endothelial migration in response to rHuEpo also has been reported48 in addition to endothelial cord formation in primary and immortalized endothelial cell lines40 and capillary sprouting from explanted human myocardium and aortic rings.83

In vivo, using both rat mammary adenocarcinoma and mouse colon carcinoma models, no difference was reported in angiogenesis between rHuEpo-treated and placebo-treated tumors.88 The effect of rHuEpo on tumor vascular density also was examined in 2 different glioma xenograft models, and no increased density was noted.72 In addition, there were no morphologic alterations in blood vessels in the treated tumors. However, recent xenograft studies have suggested that rHuEpo, used at therapeutically relevant doses, acts on tumor endothelial cells to enhance tumor shrinkage in the presence of chemotherapy, potentially by increasing the bioavailability of anticancer drugs.89 Therefore, rHuEPO may provide a beneficial therapeutic effect when combined with chemotherapy.

Several studies have suggested that rHuEpo can modestly mobilize endothelial progenitor cells (EPCs) from the bone marrow to peripheral blood to promote neovessel development.90, 91 However, the true identification of EPCs92 and the correlation between circulating EPC levels and tumor growth remain quite controversial. In some studies, bone marrow-derived EPCs differentiated into mature endothelial cells and were incorporated into the tumor vasculature of animals.93 In other studies, bone marrow EPCs did not contribute to tumor vasculature in mice94 and also reportedly were infrequent in humans, contributing to <5% of the tumor vascular endothelium.95 Therefore, if rHuEpo mobilized EPCs, then the extent would be modest: The significance of this mobilization is uncertain, and additional factors may play a role in disease settings.

Tumor Oxygenation

Cellular response to hypoxia includes significant changes in gene expression that are believed to contribute to cellular resistance to radiotherapy and chemotherapy and may increase malignancy potential through gene amplification and mutation. Indeed, it was observed that patients who have tumors with lower oxygen pressure had significantly worse disease-free and overall survival, largely because of locoregional failures.96 It has been proposed that these tumor changes occur through the induction of the hypoxia-inducible factor 1α (HIF-1α) system (for review, see Leo et al.97). However, these findings conflict with the observation that patients with Chuvash polycythemia had high levels of HIF-1α expression (through acquisition of VHL protein mutations) but did not have an apparent increase in tumor incidence.81 These findings suggest that factors other than HIF-1α contribute to tumor initiation, but hypoxia may play a role in tumor progression.

It has been hypothesized that anemia contributes to the problem of tumor hypoxia and increases the relative risk of death.98 Studies investigating the use of ESAs to alleviate the problems of anemia and hypoxia in preclinical animal models have reported that rHuEPO improved tumor oxygenation in xenograft models72, 99, 100 but also in a mechanism that was independent of its hemoglobin-raising effects.101 DA corrected anemia in tumor-bearing mice and sensitized tumor cells to radiation therapy.100 However, a sensitizing effect also was observed when DA was given 2 hours before radiation therapy, before an effect on hemoglobin could occur.100 Conversely, in other studies, the tumor-sensitization effect was not observed in nonanemic rat tumor models even though tumor oxygenation was improved.74 Therefore, there was inconsistent evidence of the ability of ESAs to enhance the efficacy of radiotherapy in preclinical tumor models.


Numerous investigators have reported EpoR expression in tumors and have assumed a negative impact on tumor progression and survival. However, many of the findings and conclusions of these studies are questionable because of problems with the methods that were used to detect EpoR protein, lack of appropriate controls, and lack of detection of physiologically relevant surface EpoR on tumor cells. The evidence that ESAs confer a proliferative or survival advantage for tumor cells is not compelling. For tumor cell lines that responded to rHuEpo in vitro, either the response was marginal and/or suprapharmacologic levels of rHuEpo were required to evoke a biologic response. In contrast, many tumor cell lines do not respond to rHuEpo, even when EpoR is expressed. These findings may reflect a possible absence of intracellular signaling after ligand-receptor interaction, low EpoR density, or nonfunctional EpoR at the cell surface in tumor cell lines. Of most physiological relevance are in vivo tumor studies, some of which have demonstrated that treatment with ESAs reduced tumor growth through radiosensitization, reduced hypoxia, or enhanced tumor immunity. Most noteworthy, at as of the time that this article was written, no animal in vivo tumor study has demonstrated that ESAs treatment alone enhances tumor progression or decreases survival. The on-going, well-controlled clinical trials will clarify whether the use of ESAs in anemic cancer patients is safe.


We thank Jon Oliner, Bill Fanslow, and Ildiko Sarosi for valuable scientific discussions; Kathryn Boorer of Amgen Inc. for editorial assistance; and Barbra Sasu, Luis Borges, and Paul Hughes for reviewing the article.