Erythropoietin after a century of research: younger than ever


Wolfgang Jelkmann MD, Institute of Physiology, University of Luebeck, Ratzeburger Allee 160, D-23538 Luebeck, Germany Tel: +49 451 500 4150; Fax: +49 451 500 4151; e-mail:


In the light of the enthusiasm regarding the use of recombinant human erythropoietin (Epo) and its analogues for treatment of the anaemias of chronic renal failure and malignancies it is worth remembering that today's success has been based on a century of laborious research. The concept of the humoral regulation of haematopoiesis was first formulated in 1906. The term ‘erythropoietin’ for the erythropoiesis-stimulating hormone was introduced in 1948. Native human Epo was isolated in 1977 and its gene cloned in 1985. During the last 15 yr, major progress has been made in identifying the molecules controlling Epo gene expression, primarily the hypoxia-inducible transcription factors (HIF) that are regulated by specific O2 and oxoglutarate requiring Fe2+-containing dioxygenases. With respect to the action of Epo, its dimeric receptor (Epo-R) has been characterised and shown to signal through protein kinases, anti-apoptotic proteins and transcription factors. The demonstration of Epo-R in non-haematopoietic tissues indicates that Epo is a pleiotropic viability and growth factor. The neuroprotective and cardioprotective potentials of Epo are reviewed with a focus on clinical research. In addition, studies utilising the Epo derivatives with prolonged half-life, peptidic and non-peptidic Epo mimetics, orally active drugs stimulating endogenous Epo production and Epo gene transfer are reviewed.

The Biblical phrase ‘for the life of the flesh is in the blood’ (Moses 4th book; Leviticus 17:11) reveals that blood was considered a symbol of life in the early days. In fact, ancient medicine related almost all somatic and psychic diseases to disorders of the blood. Accordingly, the blood of a strong animal was believed to make a human recipient powerful and courageous, if ingested or used for a bath (1, 2). Interestingly, the ancient concept of blood as a medium transferring individual properties has recently received some verification with the curious finding that the chemosensory identity through odour is altered in rodents after bone marrow transplantation (3). The first blood transfusion in a human being was performed by Jean-Baptiste Denis, Physician in Ordinary to Louis XIV, and the surgeon Paul Emmerez in Paris in 1667. Reportedly, a febrile young patient, who suffered from symptoms of anaemia after he had been bled 20 times for therapy, improved greatly following the transfusion of lamb blood (1). Both for medical and ethical reasons, heterologous blood transfusion to humans was disapproved after a few more trials. In 1825 James Blundell, an obstetrician in London, performed the first successful homologous blood transfusion in a human being, a woman suffering from postpartum haemorrhage whose husband donated the blood (4). Immune reactions and blood clotting remained major plagues in transfusion therapy. Following the discovery of the human blood groups and through the use of anticoagulants, homologous blood transfusion has become an important life-saving procedure in the previous century. The remaining risks include febrile non-haemolytic transfusion reactions, graft-vs.-host-disease, acute or delayed haemolytic reactions and transmission of prions, viruses, protozoans and bacteria. Thus, homologous blood cell transfusion should be avoided whenever possible.

Recombinant DNA technologies enabling the manufacture of biopharmaceuticals in cultured bacterial, yeast and animal cells have gained medical and economic relevance during the past 25 yr. Over 20 different blood cell-modulating proteins of potential therapeutic value have been engineered. Some of these are lineage-specific haematopoietic growth factors (Table 1), while others act primarily on the immune system. The growth factors inhibit apoptosis of haematopoietic stem cells and progenitors of blood cells. Red blood cell production is regulated by the glycoprotein hormone erythropoietin (Epo) which is mainly of renal origin. Before recombinant human Epo (rHuEpo) became available for therapy, about 25% of patients with chronic kidney disease (CKD) needed regular transfusions of red cells. rHuEpo is also administered for treatment of non-renal anaemias, such as those associated with malignant or autoimmune diseases, and in surgical settings.

Table 1.   Therapeutically relevant haematopoietic growth factors
FactorSize (kDa)Primary cells of synthesisPrimary target cellsClinical review
  1. 1 Routinely used as recombinant drugs.

  2. CKD, chronic kidney disease; BFU, burst-forming unit; CFU, colony-forming unit; CSF, colony-stimulating factor; E, erythroid; G, granulocyte; M, monocyte; Eo, eosinophilic granulocyte; Meg, megakaryocyte.

Erythropoietin130Renal fibroblasts, hepatocytesCFU-ECKD (305, 306, 351)
Cancer (316, 352, 353)
Thrombopoietin60Hepatocytes, renal tubular cellsCFU-Meg, megakaryocytes, thrombocytes(354–356)
GM-CSF114–35T-lymphocytes, monocytes, endothelial cells, fibroblastsCFU-GM, CFU-G, CFU-M, CFUEo, CFU-Meg(357–364)
G-CSF120Monocytes, endothelial cells, fibroblastsCFU-G(357, 360, 361, 363, 364)

In the light of the therapeutic value of rHuEpo it should be remembered that today's success has been based on a century of laborious research into the basics of erythropoiesis. Apart from an anecdotal review, this article provides information on novel actions of Epo and pharmacological strategies for the treatment of anaemias.

Stimulation of erythropoiesis by hypoxaemia

In an elegant book chapters on very early observations of respiration at high altitudes, Allan Erslev (5) cites the Jesuit priest Father Acosta who, in 1569, noted breathlessness, dizziness and vomiting on highland expeditions (the elements of air are in this place so thin and delicate…). Still, the important role of red blood cells as O2 carriers was not recognised until the second half of the 19th century (Table 2). Paul Bert, Professor of Physiology at the Sorbonne in Paris, who later switched over to politics and became governor-general of French Indo-China (6), and his mentor Denis Jourdanet discovered that the concentration of red blood cells depends on the O2 supply to the tissue (7). In following the order of Napoleon III, Jourdanet had moved to the highlands of Mexico where he made his important observations on the symptoms of chronic mountain disease. The patients were found to have thickened blood, while their symptoms were similar to those of anaemic persons (dyspnoea, tachycardia and syncope). Jourdanet coined the term ‘anoxyhémie’ to describe the lack of O2 in arterial blood. Paul Bert noted that animals living in La Paz at 4000 m altitude have a higher haemoglobin concentration and thus higher O2 capacity than lowlanders (7). Bert and Jourdanet believed that polyglobulia, the ‘thick blood’ of high-altitude residents was inherited. Similarly, Gustav von Hüfner (Hüfner's number: 1 g haemoglobin binds 1.34 mL of O2) stated in 1890 that the increased haemoglobin richness of the blood was acquired through generations (8). In the same year, however, this concept was disproved by the French anatomist Francois-Gilbert Viault who had travelled from Bordeaux to the highland of Peru. After he had stayed in Morococha (about 4500 m above sea level) for 23 d, the number of erythrocytes in his blood was found to have increased from 5 to 8 × 106/μL. His companions showed the same reaction (Dr. Mayorga, a few miners, a dog, a rooster and a llama). Viault concluded that erythropoiesis is acutely stimulated when the O2 content of the blood is reduced (9). Along these lines, Müntz reported that the generations of a herd of rabbits he had brought to the Pic du Midi (2877 m) in the Pyrenees exhibited a 75% increase in the iron content of blood (10). Friedrich Miescher in 1893 (11) reported the data of his colleague Egger from Basel, showing that the number of red blood cells was increased in anaemic patients with tuberculosis after they had spent a few weeks in an alpine health resort. Miescher explained the stimulation of erythropoiesis by a reduced O2 partial pressure in the bone marrow.

Table 2.   Seminal studies of the respiratory function of red blood cells
1840Blood gas measurementsH. Gustav Magnus
1860Analysis of function of haemoglobinFelix Hoppe-Seyler
1870Tissue respirationAlexander Schmidt, Eduard F.W. Pflüger
1870Role of bone marrow in haematopoiesisErnst Neumann, Guilio Bizzozero
1880Respiratory function at high altitudePaul Bert, Denis Jourdanet
1890Haemoglobin/O2 binding curveChristian Bohr

The idea of hormonal regulation of erythropoiesis was first formulated by Paul Carnot, Professor of Medicine at the Sorbonne, Paris, and his co-worker Deflandre in 1906 (12, 13). These investigators had subjected rabbits to a bloodletting (30 ml), taken another blood sample on the next day and injected the serum (5–9 mL) into normal rabbits. The concentration of red blood cells in the recipients increased by 20–40% within 1–2 d. Carnot and Deflandre concluded that the serum contained a haematopoietic factor (hémopoiétine). The more specific name ‘erythropoietin’ was introduced by Eva Bonsdorff and Eeva Jalavisto from Helsinki in 1948 (14).

The search for the elusive hormone

Extensive studies have shown that reticulocytosis occurs after a lag of 3–4 d, when the plasma Epo level increases (15). The concentration of red blood cells requires an even longer period to rise significantly. Thus, the results of Carnot and Deflandre (13) are almost impossible to reproduce in the described way. In addition, Carnot and Deflandre (13) reported that the ‘hémopoiétine’-containing serum was inactivated by heating to 56°C, whereas Epo is a heat-stable glycoprotein that loses little activity on short-term boiling. The observations made by Carnot and Deflandre (13) were probably based on good luck, i.e. haemoconcentration in the recipients. Yet this backward glance shows that a plausible concept may promote scientific progress even if derived from a misleading observation.

In the following 50 yr, some investigators reported that the injection of serum from anaemic or hypoxaemic animals increases the number of red blood cells in recipient animals (16–21); [for a review, see Ref. (22)]. Moreover, Oliva et al. published in 1949 that the transfusion of plasma (40–50 mL) from patients with pernicious anaemia to normal subjects led to reticulocytosis in the recipients within 12 h (23). Despite these reports, the existence of ‘hémopoiétine’ was doubted for many years because most investigations failed to reproduce Carnot's finding (24, 25). The interest in the alleged humoral erythropoietic factor was stimulated after Kurt Reissmann (26) and Gerhard Ruhenstroth-Bauer (27) demonstrated in 1950 that erythrocytic hyperplasia of the bone marrow and reticulocytosis occur in both partners of parabiotic animal pairs when anaemia or hypoxaemia is induced in one of them. Allan Erslev, who was born in Copenhagen in 1919 and later became Professor of Medicine at Thomas Jefferson University in Philadelphia, is generally credited with having provided definitive proof for the existence of Epo (28). Erslev transfused large volumes (50–200 mL) of plasma from anaemic rabbits (haematocrit <0.2) into normal rabbits. The recipients responded with significant reticulocytosis and, in the longer term, an increase in haematocrit values. Notably, Erslev predicted 50 yr ago the potential therapeutic value of the erythropoietic factor: ‘Conceivably isolation and purification of this factor would provide an agent useful in the treatment of conditions associated with erythropoietic depression, such as chronic infection and chronic renal disease’ (28).

The identification of the Epo-producing organs

Observations of patients with a patent ductus arteriosus led to the finding that hypoxia of the lower part of the body causes polyglobulia (29, 30). The important role of the kidney in the production of Epo became clear when Leon Jacobson and his colleagues published in 1957 that nephrectomised rats fail to respond with the normal increase in the plasma Epo level on hypoxic stress (31). Subsequently, initial reports of low plasma Epo activity in anaemic patients with CKD appeared (32, 33). In 1961 Kuratowska et al. (34) and Fisher and Birdwell (35) detected erythropoietic activity in the blood perfusate of isolated rabbit kidneys and in situ perfused dog kidneys, respectively. Nevertheless, for many years it was doubted that renal cells synthesise Epo directly because attempts failed to extract Epo from the kidney. The alternative concept was put forward that hypoxic kidneys would release an enzyme (erythrogenin) capable of splitting Epo from a plasma protein (36). The erythrogenin concept was disapproved in 1974 when Erslev demonstrated Epo activity in isolated serum-free perfused rabbit kidneys (37). Subsequently, Epo (38, 39) and Epo mRNA (40–42) were extracted from the kidneys of hypoxic rodents. The hormone was shown to be mainly present in the renal cortex and not in the medulla (38). In situ hybridisation studies identified a subgroup of peritubular fibroblasts as the site of Epo gene expression in the kidney (43). In addition to the kidneys, liver, spleen, lung, bone marrow and brain were shown to express Epo mRNA (44–46). As in other mammals, the liver is the main site of Epo synthesis in human foetuses (47). Brain-derived Epo, which is unlikely to enter the general circulation in significant amounts because of the blood–brain barrier, is thought to act as a paracrine neuroprotective factor (48).

Isolation and chemical characterisation of Epo

The initially slow progress in Epo research is understandable in view of the low concentration of the hormone in fluids and tissues, which rendered its detection difficult (49). Normally 1 L of plasma contains about 50 ng Epo. Plzak et al. (50) first developed a reliable bioassay for Epo in which radioactive iron was injected into rats followed by measurements of the incorporation of the iron into haemoglobin. An increased sensitivity was achieved when the assay was carried out in polycythaemic mice, whereby polycythaemia was induced by blood transfusion (51) or previous hypoxia exposure (52) to suppress endogenous Epo production. One Epo unit (U) was defined as the dose producing the same erythropoiesis-stimulating response as 5 μmol cobaltous chloride in experimental animals. In 1966, Epo standard A (ESA; prepared from sheep plasma) was replaced by standard B, which was of human urinary origin and nominated as the first International Standard [IU; (53)]. In 1972, the second International Reference Preparation [2nd IRP; (54)] of impure human urinary Epo (specific activity 2 IU/mg protein) and in 1992 a purified recombinant DNA-derived human Epo [IS 87/684; 130 000 IU/mg fully glycosylated protein; (55)] were established. Garcia, Sherwood and Goldwasser first developed a reliable radioimmunoassay for Epo (56–58). Today, commercial enzyme-linked immunoassay kits are most commonly used for the measurement of Epo (59). The concentration of immunoreactive Epo in serum or plasma of non-anaemic humans amounts to about 15 IU/L.

The purification of human Epo was a difficult task. Based on their experience with the extraction of Epo from sheep plasma, Goldwasser and Kung (60) calculated in 1968 that a volume of 3250 L of urine from anaemic patients was required to purify 10 mg pure human Epo, which ‘would represent about 3 yr's daily collection from a single patient, or one month's collection from 36 patients, which does not seem to be an impossible goal.’ The authors succeeded after 9 yr (61). The pure human urinary Epo enabled the partial identification of its amino acid sequence and the subsequent isolation of the human Epo gene (62, 63). With a view to litigations about patent rights between the pharmaceutical companies involved, Eugene Goldwasser has published a very informal and personal review of the rivalry for being the first to clone the Epo gene and express rHuEpo for therapeutic purposes (49).

Human Epo is an acidic glycoprotein with a molecular mass of 30.4 kDa. The peptide core of mature Epo is composed of 165 amino acids which form two bisulphide bridges (Cys7–Cys161, Cys29–Cys33). The carbohydrate portion (40% of the molecule) comprises three tetra-antennary N-linked (Asn24, Asn38 and Asn83) and one small O-linked (Ser126) glycans (64, 65). Endogenous circulating Epo (66) as well as rHuEpo (67) have several glycosylation isoforms. The N-glycans are essential for the in vivo biological activity of Epo (68). Of major importance are the terminal sialic acid residues of these glycans (69). Like other asialo-glycoproteins, asialo-Epo is rapidly removed via galactose receptors of hepatocytes (70), because galactose is the preterminal sugar of the glycans. On the contrary, the introduction of additional N-glycans into recombinant Epo by site-directed mutagenesis results in a prolonged in vivo survival of the molecules (71, 72).

The human Epo gene is located on the long arm of chromosome 7 (q11–q22) (73–75). It contains five exons, which encode a 193-amino acid prohormone, and four introns (63). The amino acid leader sequence of 27 residues is cleaved prior to secretion (76). In addition, circulating human Epo lacks the carboxy-terminal arginine that is expected from the nucleotide sequence. Tissue-specific expression of the Epo gene depends on distinct upstream (5′) DNA sequences (77, 78). Epo expression is controlled by several transcription factors. The 5′-promoter possesses GATA-binding sites (79). GATA-4 is thought to recruit chromatin-modifying activity, promoting the expression of the Epo gene. The low GATA-4 level in adult vs. foetal hepatocytes may explain the declining role of the liver in Epo production after birth (80). In contrast to GATA-4, GATA-2 appears to inhibit Epo gene expression (81, 82). Furthermore, the Epo promoter and the 5′-flanking region contain binding sites for nuclear factor κB (NF-κB) (83). GATA-2 and NF-κB are assumed to be responsible for the inhibition of Epo gene expression in inflammatory diseases (84–86).

With respect to the hypoxic induction of Epo, of primary importance are hypoxia response elements (HRE) in the 3′-Epo enhancer to which specific heterodimeric hypoxia-inducible transcription factors (HIF-α/β) can bind. The first member of the HIF family, HIF-1, was discovered by the group of Semenza 15 yr ago (87, 88). HRE and HIF are not only relevant for Epo-producing cells but are also expressed by almost all nucleated cells (89). More than 100 genes have been identified that are regulated by HIF-binding to HRE, including those encoding the vascular endothelial growth factor (VEGF), the glucose transporters 1 and 3 and several glycolytic enzymes (90–92).

Hypoxia-inducible transcription factors cooperate with the hepatocyte nuclear factor 4 (HNF-4) (93) by direct protein–protein interaction and through the recruitment of the transcriptional co-activator complex CBP/p300 (94). A DR2 element (TGACCTCTTGACCC) near the HIF-binding site augments the hypoxic induction of Epo expression (79). DR2 elements are activated by binding the retinoic acid receptor which seems to be of major importance for Epo production by the foetal liver (95). In addition, retinoic acid has been shown to stimulate Epo production in rat kidneys and human hepatoma cell cultures (96, 97).

Hypoxic induction of Epo

Teleologically, the primary function of Epo is to maintain the blood haemoglobin concentration in the normal range during steady-state conditions and to hasten red cell mass recovery after haemorrhage. The concentration of circulating Epo increases exponentially with decreasing haemoglobin levels in uncomplicated anaemia (absence of renal disease or inflammation). However, from the early days of Epo research it has been clear that the controlled variable is not the concentration of erythrocytes or of haemoglobin in blood. Instead, feedback regulation is based on the tissue O2 pressure (pO2), which depends on the haemoglobin concentration, the arterial pO2, the O2 affinity of the haemoglobin and the rate of blood flow. The kidney is very appropriate for controlling Epo production because the pO2 in the renal cortex is little affected by the rate of blood flow as the renal O2 consumption changes in proportion with the glomerular filtration rate (98). Whether extrarenal sites, primarily the hypothalamus (99), modulate O2-dependent Epo production in the kidney is still a matter of debate (100).

Because the tissue pO2 is the controlled variable, residence at high altitudes leads to a stimulation of erythropoiesis, and in the long term to erythrocytosis. Faura et al. (101) first demonstrated increased Epo activity in the urine of lowlanders taken to Morococha, whereby Epo activities reached peak values after 24 h and then fell to a new plateau at about twice the sea-level values. Abbrecht and Littell (102) confirmed the dynamic response by measurements of serum Epo in a high-altitude expedition in Colorado (4360 m). Note that South American high-altitude natives often suffer from erythrocytosis and chronic mountain sickness (103), while Tibetans living at about 4000 m altitude have low haemoglobin concentrations and do not develop chronic mountain sickness (104, 105). These findings are significant because they show that there has been natural selection and genetic adaptation in the evolution of tolerance of humans to hypoxia (106, 107).

The molecular mechanism of O2 sensing

Several concepts were earlier developed to explain the pO2-dependent expression of the Epo gene [for a review, see Ref. (108)]. Based on the pioneering in vitro observation that CO suppresses Epo gene expression, while transition metal ions such as cobalt or nickel mimic the response to hypoxia, it was suggested that a haem protein might be involved in O2 sensing (109). Haemoproteins of the microsomal mixed functional oxidases (P450) were implicated in Epo production (110), which has been reconsidered recently (111). Other investigators proposed that mitochondrial reactive O2 species (ROS) might trigger hypoxia-induced transcription of the Epo gene (112). In contrast, our own group found that the production of H2O2 is reduced under hypoxic conditions (113). Studies utilising mutant cells lacking mitochondrial DNA have shown that O2 sensing does not require the mitochondrial respiratory chain apparatus for hypoxia-induced gene expression (114–116). Note, however, that ROS appear to mediate the ligand-induced activation of HIF-dependent gene expression in oxygenated cells (117).

The discovery of HIF-1 by Semenza et al. has provided a clearer insight into the molecular mechanisms of O2 sensing (87, 88). HIF-1 is a heterodimeric protein composed of the subunits α (HIF-1α, 120 kDa) and β (HIF-1β, 91–94 kDa) (118, 119), which belong to the family of basic helix loop helix Per-ARNT-Sim proteins (bHLH-PAS). The N-terminal bHLH-PAS domains are required for dimerisation and DNA-binding (120–122). Although both HIF-1α and HIF-1β are continuously translated, HIF-1α is usually not detectable in normoxic cells (123–125). The C-terminus of HIF-1α comprises a transactivation domain (TAD) that can be subdivided into an N-terminal (N-TAD) and a C-terminal (C-TAD) part. The N-TAD overlaps with an O2-dependent degradation domain (O-DDD). The critical residue in this domain is a proline in position 564 (126). Catalysed by specific prolyl-4-hydroxylases (HIF-PHD) this Pro564 is hydroxylated in the presence of O2, Fe2+ and 2-oxoglutarate (127–131). The prolyl hydroxylated HIF-1α combines with the von Hippel-Lindau tumour suppressor protein (pVHL) to form a complex that is polyubiquitinated by an E3 ligase (125, 132, 133) and undergoes immediate proteasomal degradation (134). The HIF-PHDs also catalyse the hydroxylation of a second prolyl residue (Pro402) of HIF-1α in normoxia (126).

More recently, two bHLH-PAS HIF-α isoforms have been identified, namely HIF-2α (also known as endothelial PAS domain protein 1, EPAS 1) and HIF-3α (135–138). These are also O2-labile and can dimerise with HIF-1β in hypoxia but are different with respect to their tissue-specific mRNA expression pattern (139). HIF-2α is detectable in the Epo-producing renal fibroblasts of hypoxic rats while tubular cells express HIF-1α (140, 141). Indeed, recent studies indicate that HIF-2 is the primary transcription factor inducing Epo gene expression (142). HIF-2α is degraded on hydroxylation of Pro531 (143). In contrast to HIF-1α and HIF-2α, HIF-3α lacks a transcriptional activation domain and can suppress the expression of hypoxia-responsive genes (144).

The HIF-PHDs are likely to play a major role in the control of Epo production, because they prevent HIF-α from entering the nucleus under normoxic conditions. As PHD-2 and PHD-3 are themselves HIF-target genes, their expression increases while HIF-α levels decline on exposure to hypoxia in the long term (145–147). This finding may explain the reduced Epo production during prolonged stay at high altitudes. Percy et al. (148) have recently described members of a family in whom erythrocytosis is associated with an inherited mutation in PHD-2. The binding of pVHL is a prerequisite for the degradation of HIF-α. Thus the mutation of pVHL, demonstrated in patients suffering from congenital Chuvash polycythaemia, is associated with increased transcription of the Epo gene (149).

The transcriptional activity of the HIFs is further suppressed by a third O2-dependent hydroxylation, namely at Asn803 in HIF-1α and Asn851 in HIF-2α. This reaction is catalysed by a HIF-α-specific asparaginyl hydroxylase that is also termed ‘factor-inhibiting HIF-1’ (FIH-1) (150, 151). As a result of the asparaginyl hydroxylation, the binding of the transcriptional co-activator CBP/p300 to the C-TAD of HIF-α is prevented (150, 152, 153). Like the PHDs, FIH-1 is an Fe2+-containing and α-oxoglutarate-requiring dioxygenase. On binding to Fe2+ (a non-haem iron), O2 is split with one O-atom being transferred to the amino acid residue of HIF-α and the other to α-oxoglutarate, forming CO2 and succinate. The Km values of the three PHDs for O2 are above the arterial pO2 (approximately 170 mmHg), whereas the asparaginyl hydroxylase operates at a lower pO2 (approximately 60 mmHg) (154). α-Ketoglutarate antagonists, which inhibit the hydroxylation of HIF-α (HIF stabilisers), stimulate Epo production in the kidney and the liver (155).

Fandrey (156) has recently described the molecular mechanisms controlling Epo gene expression in more detail. Reviews are also available on the role of HIF-1 and -2 activities in diseases of the cardiovascular (157) and central nervous systems (158) and in cancer (157, 159).

Erythrocytic progenitors and the Epo receptor

Evidence was provided about 50 yr ago that blood cell production is sustained by a relatively small pool of self-renewing stem cells in haematopoietic organs. Till and McCulloch (160) showed that mouse bone marrow cells can re-populate the spleen of heavily irradiated mice to form erythrocytic, granulocytic, megakaryocytic and mixed colonies. The murine pluripotent stem cell was termed ‘colony-forming unit spleen’ (CFU-S). Subsequent studies identified subtypes of stem cells restricted to generate either myeloid or T-cell progenies (161). In the erythropoietic lineage the ‘burst-forming unit erythroid’ (BFU-E) was shown to produce large multiclustered bursts of haemoglobin-synthesising cells when cultured with high doses of Epo (162). Two different types of BFU-E were detected in mouse bone marrow: a mature type giving rise to 50–200 erythroblasts within 3 d in culture and a primitive type giving rise to several hundred erythroblasts within 8–10 d (163). The human analogues produce bursts after 10–12 and 17–20 d, respectively (164). Importantly, the primitive type of BFU-E was also detected in human blood (165, 166). The ‘colony-forming unit erythroid’ (CFU-E) was identified as a more differentiated and strictly Epo-dependent erythrocytic progenitor (163, 167). CFU-Es produce small clusters of eight to 60 erythroblasts, which takes 2–4 d for mouse (168) and 7–8 d for human cells (164). CFU-Es are not stem cells, as they have no potential for self-replication (169). The number of colonies that grow from CFU-Es correlates closely with the Epo concentration (170, 171), which can be utilised for in vitro bioassay of Epo. Koury and Bondurant (172, 173) first showed that the primary mechanism by which Epo maintains erythropoiesis is prevention of apoptosis. CFU-Es express abundant GATA-1 (174), which is an important transcription factor in erythrocytic development (175). The balance between GATA-1 and caspases determines the balance between apoptosis, proliferation and differentiation of erythrocytic progenitors (176). For example, GATA-1 was shown to induce the anti-apoptotic protein bcl-xL (177).

Chang et al. (178) and Krantz and Goldwasser (179) first provided evidence that Epo binds to a membranous receptor (Epo-R) of its target cells. The murine Epo-R was cloned and characterised by D'Andrea et al. in 1989 (180). Epo-R was found to belong to the cytokine class I receptor superfamily (181, 182) whose members are characterised by an extracellular N-terminal domain with conserved cysteines and a WSXWS-motif, a single hydrophobic transmembrane segment and a cytosolic domain that lacks enzymatic activity. The mature human Epo-R is a 484-amino acid glycoprotein with 1 N-glycan. The calculated mass of human Epo-R of 52.6 kDa increases to about 60 kDa because of glycosylation and phosphorylation (183). Two of the membrane-spanning Epo-R molecules form a homodimer that binds one Epo molecule. The Epo dissociation constants for the two Epo-R-binding sites differ by three orders of magnitude (184). Most of the ligand/Epo-R interaction occurs in a hydrophobic flat region of the Epo-R (Phe93, Met150, Phe205) (185, 186). Crystal structure analysis data of the Epo-R are available (187). With respect to novel pharmacological compounds it is of interest that the affinity of Epo analogues for Epo-R decreases with glycosylation (71). Darling et al. (188) have demonstrated that the carbohydrate portion of the glycoprotein prevents receptor binding through electrostatic forces.

Binding of Epo induces a conformational change and a tight connection of the two monomeric Epo-R molecules (189–191). Thereby, two Janus kinases 2 (JAK2) which are in contact with the cytoplasmic region of Epo-R are activated (191, 192). As a result, several tyrosine residues of the Epo-R are phosphorylated to provide docking sites for signalling proteins containing SRC homology 2 (SH2) domains (193, 194). The pathways through which Epo signals include phosphatidyl-inositol 3-kinase (PI-3K)/Akt, STAT5, MAP kinase and protein kinase C (195–197). The effect of Epo is terminated by the action of the haematopoietic cell phosphatase (HCP) which catalyses JAK2 de-phosphorylation (198, 199). Mutations of the cytoplasmic C-terminal regions of Epo-R and functional deficiencies of HCP may lead to erythrocytosis (200). In vitro studies have shown that the acute effects of Epo last for about 30–60 min (201). The Epo/Epo-R complex is internalised following de-phosphorylation of Epo-R. The proteasome controls the duration of Epo signalling by inhibiting the renewal of cell surface receptor molecules (201, 202). In using Ba/F3 and UT-7/Epo cells as in vitro model systems, Gross and Lodish (203) have recently shown that 60% of internalised Epo is re-secreted, while 40% is intracellularly degraded. It remains to be investigated, however, whether brief binding to Epo-R without internalisation suffices for Epo signalling and whether re-secreted Epo is in vivo biologically active.

Non-haematopoietic actions of Epo

Epo-R mRNA is expressed in various non-haematopoietic tissues such as endothelium, neuronal cells and placenta (181). Probably due to anaemia and tissue hypoxia the complete knock-out of the Epo-R gene results in a phenotype of severe cardiac malformations and foetal death at day 13.5 in mice (204). Non-anaemic mice with a tissue-specific Epo-R knock-out outside the haemopoietic tissue exhibit no abnormalities (205). Functional Epo-R in non-erythrocytic cells were first demonstrated in endothelial cell cultures (206, 207). In vitro, Epo promotes the proliferation and migration of endothelial cells (206, 208), stimulates the production of modulators of vascular tone (209–211), favours a pro-angiogenic phenotype and induces neovascularisation (212). In vivo, Epo increases the number of circulating endothelial progenitor cells (213, 214). Vascular smooth muscle cells also express Epo-R (215, 216). Here, Epo causes Ca2+ mobilisation (217, 218), phospholipase C activation (219) and smooth muscle contraction (218). Other effects include the activation of the MAPK (216, 220) and PI-3K/Akt pathways (215), which mediate the inhibition of apoptosis (221, 222). Furthermore, Epo-R is expressed by human cardiomyocyte cell lines (223) and in human adult cardiac tissue (224). Preclinical studies have shown that Epo exerts cardioprotective effects (225–227).

With respect to the central nervous system both Epo-R and Epo become detectable 5 wk after conception in the brain of human embryos (228–230). Epo-R is expressed by neurones and astrocytes (231, 232) and by brain capillary endothelial cells (233, 234). Epo exerts neuroprotective and neurotrophic effects (231, 235, 236). Erbayraktar et al. (237) have reported that asialo-rHuEpo, which is devoid of erythropoiesis-stimulating activity, crosses the blood–brain barrier after i.v. administration, binds to neurones within the hippocampus and cortex, and is neuroprotective in cerebral ischaemia, spinal compression and sciatic nerve crush in experimental animals. Likewise, carbamoylated Epo (often incorrectly referred to as carbamylated Epo), which does not bind to the haematopoietic Epo-R and does not stimulate erythropoiesis (238, 239), has been reported to confer neuroprotection in animal models (240). The hypothesis has been put forward that non-haematopoietic effects of Epo are mediated by heteromers of Epo-R and β-subunit receptor molecules (240, 241). In view of the neuroprotective potential of Epo in experimental animals, Ehrenreich et al. (242) performed a clinical trial in patients with acute stroke. In a double-blind randomised proof-of-concept study of 40 patients, rHuEpo treatment showed a strong trend for reduction in infarct size as judged from magnetic resonance imaging. The reduction in infarct size correlated with a markedly improved neurologic recovery (242). rHuEpo may also prove useful as an additive neuroprotective therapeutic in schizophrenia and other human diseases characterised by a progressive decline in cognitive performance (243). Detailed reviews of the cardioprotective and neuroprotective effects of Epo are available (244–248).

A matter of current debate is whether Epo promotes tumour growth (249, 250). Epo-R mRNA and/or Epo-R protein have been detected in breast carcinoma (251–253), lung carcinoma (254), renal carcinoma (255), tumours of the cervix and of other organs of the female reproductive tract (256, 257), and various paediatric tumours (258). However, a recent study has shown that the antibodies used for detection of Epo-R protein are non-specific (259). In addition, in vitro effects of Epo in tumour cells were generally seen only when extremely high Epo concentrations were applied that exceeded the concentrations occurring after administration of Epo to oncologic patients by several orders of magnitude (251, 255, 258). Many investigators failed to demonstrate any relationship between Epo-R expression and Epo signalling or tumour growth (260–268). Liu et al. (269) have reported that Epo may not influence the basal viability of tumour cells but may prevent the cytotoxic effect of chemotherapeutics such as cisplatin. Likewise, Epo has been reported to cause resistance to cisplatin in HeLa cells (257) and to dacarbazine in melanoma cells (270). However, the understanding of the combined effects of Epo and chemotherapeutics on cancer cells is still insufficient, as one group has reported that renal carcinoma cells undergo a higher degree of apoptosis with a combination of daunorubicin and Epo or vinblastine and Epo than with either of these agents alone (271).

Anaemia of renal failure

Richard Bright at Guy's Hospital in London is commonly credited with first having described urea retention as a cardinal sign of renal insufficiency (272). Although Bright noted ‘… after a time, the healthy colour of the countenance fades’ (272) inquirers have stated that the anaemia in association with renal failure was first recognised by Sir Robert Christison from Scotland, who stated in 1839 that ‘… by far the most remarkable character of the blood in the advanced stage of the Bright's disease is a gradual and rapid reduction of its colouring matter…’ (273). The anaemia in CKD is usually normochromic and normocytic, and its severity increases with the impairment of renal function (274, 275). Significant anaemia develops in general when the glomerular filtration rate falls below 40 to 20 mL/min and 1.73 m2 body surface area (274, 276–278). Low Epo levels relative to the degree of anaemia were first shown by bioassay (276, 279–281) and later by radioimmunoassay (274, 275, 281–284). However, the observation that the plasma Epo level can increase on acute hypoxic stress in CKD patients indicates that the capacity to produce Epo is not necessarily abolished (274). The O2 sensor in control of the synthesis of Epo appears to operate at a lower set-point in some of the patients. Epo production by the remaining renal fibroblasts may be suppressed by proinflammatory cytokines such as interleukin (IL)-1 and tumour necrosis factor (TNF)-α (285). Other factors contributing to the anaemia in CKD include shortening of red blood cell survival, bleeding, malnutrition (286), and inhibition of the growth of erythrocytic progenitors by cytokines (287, 288) and uraemia toxins (261, 282, 289, 290). A recent re-investigation into the role of cytokines has shown that T cells from poor responders to Epo therapy are in an activated state [increased expression of interferon (IFN)-γ, TNF-α, IL-10 and IL-13] and that pentoxyfilline treatment can increase haemoglobin levels in such patients (291).

Prior to the availability of rHuEpo, cobaltous salts (292, 293) and androgens (294) were used for treatment of anaemia in CKD. The molecular mechanism of the action of cobalt is not fully clear but it probably is related to HIF-α stabilisation. Exposure to industrial cobalt is thought to contribute to the erythrocytosis in high-altitude dwellers suffering from chronic mountain sickness (295). Despite the toxicity of cobalt, there is suspicion that cobalt supplementation may be misused by athletes to increase erythropoiesis (296). In contrast to cobalt, which enhances Epo production, androgens can directly stimulate the development of erythrocytic precursors (297). At least under physiological conditions, androgens do not appear to increase Epo production significantly. Treatment of men with the luteinising hormone-releasing factor agonist nafarelin results in decreased haemoglobin levels, while the concentration of Epo in plasma remains constant (298). Along these lines, the concentration of circulating Epo does not differ when female and male human beings are compared at given haemoglobin levels (299).

The erythropoietic activity of exogenous Epo in humans was first demonstrated by Oliva et al. in 1949 (23), who transfused plasma from patients with pernicious anaemia into normal subjects, and by Essers et al. in 1974 (300), who treated patients with renal failure with plasma from anaemic persons with normal kidney function. Clinical nephrology has made major progress since then, with dialysis and renal transplantation having become standard therapeutic procedures. rHuEpo was introduced as an anti-anaemic drug for treatment of CKD patients 20 yr ago (301, 302). Eschbach (303) has published a typical case report of that time: within 8 wk of rHuEpo therapy haematocrit was raised from 0.19 to 0.38 in a 39-yr-old patient with nephrotic syndrome, who had experienced androgen treatment, unsuccessful renal transplantation, transfusion of 313 units of red blood cells, and human immunodeficiency virus (HIV) infection. The patient's clinical improvement was dramatic, he was able to perform sports and to meet all the physical demands of running his country store (303). Indeed, the availability of rHuEpo has transformed the lives of hundreds of thousands of CKD patients. Epo raises haematocrit and blood haemoglobin concentration in a dose-dependent and predictable way. It abolishes the need for red cell transfusions, prevents the hyperdynamic cardiac state and improves brain function. It may also slow down the progression of chronic kidney disease (304). The target haematocrit in patients under Epo therapy is usually set at 0.33–0.36 (haemoglobin 110–120 g/L) according to the European Best Practice Guidelines (EBPG) (305) and the US-based National Kidney Foundation's Kidney Disease Outcome Quality Initiative (K/DOQI) (306). The question whether haematocrit values should be raised into the normal range has been a matter of debate (307, 308) as a randomised prospective long-term multicentre study of 1233 patients with cardiac disease showed the 1- and 2-yr mortality rates to be 7% higher in the normal haematocrit (0.42) than in the low haematocrit (0.30) group (309). A recent evidence-based systematic literature review of the relationship between haematocrit and/or haemoglobin values and all-cause mortality in dialysis patients has revealed that the published literature is insufficient for generalisation of benefits or risks of a haemoglobin level >110 to 120 g/L (310). Moreover, the just published results of the re-investigating trials Cardiovascular Risk Reduction in Early Anemia Treatment with Epoetin beta (CREATE) and Correction of Hemoglobin and Outcomes in Renal Insufficiency (CHOIR) with Epoetin alpha support the practice to keep the patients’ haemoglobin level above the recommended level of 110 g/L and to prevent an elevation above 130 g/L (311, 312).

Causes of the rare cases of hyporesponsiveness to Epo include iron deficiency, infection and inflammation, bleeding and haemolysis, coexisting medical conditions such as malignancy, secondary hyperparathyroidism, aluminium toxicity, vitamin B12 or folate deficiency, protein malnutrition and inadequate Epo dose. Reduced iron availability is indicated by a serum ferritin concentration <100 μg/L, a transferrin saturation <20% and a proportion of hypochromic red cells >10% (313). Resistance to Epo in inflammation and infection is mainly caused by the action of IFN-γ and TNF-α (314). Measurements of C-reactive protein and baseline fibrinogen concentrations in serum may enable early recognition of the probability of response to Epo (315).

In addition to the anaemia of CKD, the anaemias associated with cancer, acquired immunodeficiency syndrome (AIDS), hepatitis C infection, bone marrow transplantation, myelodysplastic syndromes, autoimmune diseases and heart failure can be prevented by treatment with rHuEpo or its analogues (316). Miller et al. (317) first showed that the concentration of circulating Epo is relatively low in many adult cancer patients, when related to the blood haemoglobin concentration. The primary goals of Epo therapy in tumour patients are to maintain the haemoglobin values above the transfusion trigger, increase the exercise tolerance, prevent fatigue and improve quality-of-life parameters (318–320). Anti-anaemic therapy is considered beneficial for the outcome of radiotherapy and chemotherapy (321). The efficacy of the treatment of the anaemia associated with malignancy and chemotherapy is lower than in CKD patients, and higher dosing of erythropoietic drugs is required. The overall response rates have ranged from 40% to 85% (318, 322, 323). Erythropoietic stimulation in anaemic patients with chronic heart failure improves cardiac function and exercise capacity, and it reduces the need for hospitalisation (324, 325). In the surgical setting, Epo may be administered pre-operatively in order to stimulate erythropoiesis in phlebotomy programmes for autologous re-donation or for correction of a pre-existing anaemia, and post-operatively for recovery of red blood cell mass (326, 327).

Novel strategies for stimulation of erythropoiesis

Table 3 gives an overview of novel pharmacological approaches to stimulate erythropoiesis. HIF stabilisers and GATA antagonists are investigated for their potential to stimulate endogenous Epo gene expression. These compounds are orally active. The HIF stabiliser FG-2216 has already been administered in a dose-escalation study to anaemic patients with CKD (328).

Table 3.   Novel compounds and strategies for stimulation of erythropoiesis
Compound/strategy (status)Reference
HIF stabilisers for endogenous Epo gene activation (clinical studies)(328, 365)
GATA antagonists for endogenous Epo gene activation (animal studies)(86, 366)
‘Biosimilars’ (clinical studies)(329, 330)
Gene-activated Epo (Epoetin delta; clinical studies)(337, 338)
Pegylated rHuEpo (‘CERA’; clinical studies)(367)
Epo fusion proteins (animal studies)(344, 345, 345, 368–371)
Synthetic erythropoiesis proteins (SEP; animal studies)(372, 373)
Peptidic Epo mimetics (clinical studies)(186, 374, 375)
Non-peptidic Epo mimetics(376, 377)
Inhibitors of a haemopoietic cell phosphatase (animal studies)(378)
Human Epo gene therapy (clinical studies)(350)

The original rHuEpo preparations (Epoetin-α and -β) engineered in Chinese hamster ovary (CHO) cells transfected with the human Epo gene have been used routinely in clinical medicine for 20 yr. As the patents for these products have expired recently in the EU and elsewhere, other manufacturers will bring biosimilar Epoetins (‘Biosimilars’, ‘Follow-on Biologics’) on the market. The Biosimilars will have to prove that they are safe and effective (329, 330). Epoetin-omega, an rHuEpo produced in baby hamster kidney (BHK) cells has been applied in clinical trials in some countries outside the EU and the US (331–333). Note that the structure of the N-glycans of rHuEpo produced in CHO and BHK cells differs (334–336). Another new recombinant product is Epoetin-delta, which is produced by gene activation in human fibrosarcoma cells (line HT-1080) into which a DNA fragment was transfected that activates the Epo promoter (337, 338).

The rHuEpo mutein Darbepoetin-alpha was approved as an anti-anaemic drug 5 yr ago. Darbepoetin-alpha has an increased molecular mass (37.1 kDa) and an increased portion of carbohydrate (51%) compared with the Epoetins (40%). The amino acid sequence of Darbepoetin-alpha is modified by site-directed mutagenesis compared with that of the Epoetins at five positions resulting in two additional N-glycans at Asn30 and Asn88 (71). The terminal half-life of i.v. administered Darbepoetin-alpha is three- to fourfold longer than that of Epoetin-alpha and -beta [25 vs. 6–9 h; (339)]. The prolonged survival of Darbepoetin-alpha offers the opportunity for longer dosing intervals with weekly up to monthly administration. Pegylated Epoetin beta [continuous erythropoiesis receptor activator (CERA)], which is in clinical trials, has an even longer half-life (130–140 h) than Darbepoetin-alpha in circulation (340, 341). CERA contains a single methoxy-polyethylene polymer of approximately 30 kDa integrated via succidinimidyl butanoic acid. Phase I and phase II studies in healthy humans (342) respectively patients with multiple myeloma (343) have shown that CERA stimulates erythropoiesis in a dose-dependent way. Epo dimers with a polypeptide linker and Epo derivates with C-terminally added polypeptides without linker (344) and Fc-fusion proteins (345, 345) are also characterised by a prolonged half-life compared with the Epo monomers. Other pharmacological approaches such as the use of peptidic and non-peptidic organic molecules that mimic the action of Epo have been considered in detail recently (346, 347).

Epo gene therapy is an attractive alternative to the administration of erythropoietic drugs in anaemic patients (348). Here, the main problem is to adapt Epo production to the blood haemoglobin concentration. Rinsch et al. (349) first engineered cell lines to release human Epo as a function of the pO2 by using a vector containing the human Epo cDNA driven by the hypoxia-responsive phosphoglycerate kinase promoter. Lippin et al. (350) have recently performed a trial of ex vivo human Epo gene therapy in 13 CKD patients. The authors used a tissue protein factory based on dermal cores (Biopump) harvested and implanted autologously. The human Epo was expressed under the control of the cytomegalovirus (CMV) promoter. The serum Epo level increased to therapeutic levels from day 1 after implantation reaching a peak during the first week of follow-up. The rise in the Epo concentration resulted in a transient increase in reticulocyte counts. According to the authors (350) implantation of dermal cores ex vivo transduced with human genes could provide a useful tool to express therapeutic serum proteins in the clinical setting.


Thanks are due to Ms Lisa Zieske for her expert secretarial work on the manuscript.