The CFU-E is an erythroid cell closely related to the proerythroblast.⁷ Under the influence of low concentrations of Epo, it gives rise (in 2 to 3 days in murine and in 5 to 8 days in human marrow) to colonies of 8 to 32 well-hemoglobinized cells^7,8,9 and ¹⁰ (Fig. 6.2). The clonal origin of these colonies has been demonstrated by glucose-6-phosphate dehydrogenase-isoenzyme analysis.¹¹ Morphologically, CFU-E purified from progenitor cell cultures appear as immature cells with fine nuclear chromatin; a well-defined, large nucleolus; a high nuclear-cytoplasmic ratio; a perinuclear clear zone; and basophilic cytoplasm with pseudopods.^12,13 On electron microscopy, this cell appears as a primitive blast with dispersed nuclear chromatin, a prominent nucleolus, and an agranular cytoplasm containing clumps of mitochondria and frequent pinocytotic vesicles.¹² The number of CFU-E in the human marrow ranges from 50 to 400/10⁵ light-density, nonadherent, nucleated cells and varies significantly with the methods used for cell separation and the culture conditions. The majority of CFU-E are in a phase of active DNA synthesis (S phase) as demonstrated by a 70% to 90% killing of cells after short exposure to ³H-thymidine in vitro (³H-thymidine suicide) or after administration of cycle-specific chemotherapeutic agents in vivo.^14,15 and ¹⁶ The size of the CFU-E compartment in intact animals depends on the levels of circulating Epo. Anemia associated with high Epo levels or the administration of Epo leads to expansion of the CFU-E compartment, whereas transfusion-induced polycythemia leads to low Epo levels and a significant reduction of the CFU-E compartment.¹⁶ From a number of in vitro studies, it has been established that the CFU-E is the most Epo-sensitive cell, carrying the highest density of Epo receptors (EpoR) on its surface, and it is also absolutely dependent on Epo for its survival. In the absence of Epo, CFU-E rapidly undergoes programmed cell death (apoptosis)^{12,17,18,19,20} Although the first phase of CFU-E differentiation is Epo-dependent, the later stages are not.²¹

Highly purified CFU-E has been isolated from murine fetal liver cells; this is defined in flow cytometry studies as expressing c-kit, the receptor for the cytokine stem cell factor (SCF), negative for a number of lineage markers (Ter119, B220, Mac-1, CD3, Gr-1, CD32/16, Sca-1, and CD41), and having high cell surface expression of the transferrin receptor (TfR) (CD71) and the heat stable antigen, CD24.¹³

The BFU-E is an erythroid progenitor that is much more immature than the CFU-E, and more closely related to the multipotent hematopoietic stem cell, as indicated by its cell size, buoyant density, and the relatively low percentage of these cells in active DNA synthesis (0% to 25%).^15,16,22,23 In contrast to CFU-E, BFU-E also has a (limited) capacity for self-renewal and is detectable in
the peripheral blood at a concentration of 0.02% to 0.05% of lightdensity mononuclear blood cells.^24,25 BFU-E can be separated from CFU-E by its slower velocity sedimentation at unit gravity.²⁶ Morphologically, the BFU-E appears as a very immature blast cell with slightly oval, moderately basophilic cytoplasm with occasional pseudopods, very fine nuclear chromatin, and large nucleoli.^13,19 On electron microscopy, the cytoplasm contains polyribosomes and is less abundant than in CFU-E, whereas the nucleus contains small amounts of clumped heterochromatin and prominent nucleoli.¹⁹

In the presence of Epo and under the influence of other cytokines that act on early hematopoietic cells, such as interleukin-3 (IL-3), IL-6, granulocyte-macrophage colony-stimulating factor (GM-CSF), thrombopoietin (Tpo), and SCF, the BFU-E progenitor gives rise in 5 to 7 days in mice and in 14 to 16 days in humans to clusters of many erythroid colonies (a large “burst”) containing a total of 500 to more than 30,000 to 40,000 well-hemoglobinized erythroblasts (Fig. 6.3). Cytokines such as transforming growth factor (TGF) β, tumor necrosis factor (TNF) α, and interferon (IFN) γ, on the other hand, suppress progenitor proliferation. The clonal origin of BFU-E-derived erythroblasts has been demonstrated by characterization of the type of hemoglobin produced by cells in single colonies in co-culture experiments of bone marrow cells from a patient with homozygous hemoglobin C and marrow cells from another patient with homozygous hemoglobin S disease.²⁷ The BFU-E can be considered as a progenitor of the CFU-E. Indeed, after 6 to 8 days in culture, cells generated from human BFU-E have all the functional characteristics of CFU-E.¹² The concentration of BFU-E in human bone marrow varies from 10 to 50/10⁵ nucleated cells; however, this number fluctuates widely depending on the cell separation methods and culture conditions. From both in vitro and in vivo experiments, it has been established that the early stages of BFU-E proliferation and differentiation are Epo independent.^16,19,23 BFU-E can survive in vitro for 48 to 72 hours in the absence of Epo, but it is absolutely dependent on IL-3 for survival.¹⁹ Only 20% of blood BFU-E expresses a very low density of EpoR detectable by autoradiography.¹⁹ The size of the BFU-E compartment in the marrow of animals remains unaffected by the acute changes in the levels of circulating Epo induced by anemia or transfusional polycythemia.¹⁶ Anemia can induce an increase in the cycling of BFU-E without affecting their numbers,²⁸ and in vitro Epo can induce BFU-E into DNA synthesis.²⁹ In humans, chronic administration of Epo is associated with an increase in the concentration and cycling status of marrow BFU-E; however, these changes are also seen in granulocytic -monocytic and megakaryocytic CFUs (CFU-GM, CFU-MK), and multilineage progenitors such as CFU-GEMM, indicating that, at the early progenitor cell level, the marrow responds to Epo as an organ in a non-lineage-specific manner.³⁰ All these evidences indicate that the early stages of erythropoiesis at the BFU-E level are Epo independent, and Epo dependence develops at a stage between BFU-E and CFU-E.¹⁹ The distinction between early (BFU-E) and late (CFU-E) erythroid progenitors, although valid, is by itself artificial. There is a variety of cells
between BFU-E and CFU-E that form a continuum of erythroid progenitors at different stages of differentiation with properties between those of BFU-E and CFU-E. As an example, a subclass of erythroid progenitors termed mature BFU-E has been described in human and murine marrow.^14,15 These cells share properties of both CFU-E and BFU-E. They have a proliferative potential lower than BFU-E but higher than CFU-E, their cycling status is also intermediate between CFU-E and BFU-E, and they do not exhibit IL-3 dependence, but show relative Epo dependence.^14,15,18 Thus, the evidence suggests that during erythroid development, early progenitors of high proliferative potential in a relatively low cycling status with absolute dependence on IL-3 and responsiveness to, but not dependence on, Epo differentiate progressively through various stages into later progenitors of low proliferative potential with a high cycling status that are IL-3 independent and totally Epo dependent.

Increasingly, in place of functional definition of erythroid progenitors, studies, especially in mice but also in humans, allow for prospective identification of progenitors using cell surface markers.^31,32 As noted above, recent work, for example, using murine fetal liver cells as a source of hematopoietic—at the fetal liver stage predominantly erythroid—cells allows isolation of BFU-E and CFU-E progenitors with 94% and 95% purity, respectively: although both these erythroid progenitors express c-kit, BFU-E is distinguished from CFU-E as it has much lower cell surface expression of CD24 and CD71.¹³

Distinct structural genetic loci exist for each of the polypeptide chains in hemoglobin (see also Chapter 34). Thus, there are α, β, γ, δ, and ε genes. In most human populations, the α genetic locus is duplicated, and there are four (two pairs of) identical α genes in normal subjects.^75,76 There are also at least two different pairs of γ genes, one (Gγ) coding for a γ-chain with glycine at position 136 and another (Aγ) coding for a γ-chain with alanine at the same position.⁷⁷ In contrast, only single (pairs of) genes code for the β-and δ-chains, respectively.

The α-gene cluster (approximately 30 kb) is located on the short arm of chromosome 16 and also contains the locus encoding for the ζ-chain,⁷⁸ while the β-gene cluster (approximately 50 kb) is located on chromosome 11 and includes the genes for the Gγ-, Aγ-, δ-, and ε-globins.^78,79 A schematic representation is shown in Figure 6.7.

The differentiation of erythroid progenitors to erythroblasts is accompanied by the activation of the genes involved in erythroid differentiation, including the globin genes.^46,80 The active genetic regions of DNA (5% to 10% of the genetic material in erythroblasts⁷⁸) make up the open portion, or euchromatin, of nuclear material, whereas unexpressed genes (the majority of genes in the nucleus) are included in the condensed, or heterochromatin, fraction. Chromatin is described as a nucleoprotein that contains, packages, and provides an instructive scaffold for nuclear DNA.⁸¹ Thus modulation of the expression of genes, including the globin genes may be imposed by the chromatin structure, which includes not only strands of DNA, but also histone and nonhistone proteins.

Globin mRNA, like most eukaryotic mRNAs, is synthesized in a precursor form that is two to three times as long as the molecule that ultimately serves as the template for protein synthesis.^75,78,80 These precursor molecules, heterogeneous nuclear RNA, undergo “processing” to be converted into the final mRNA.⁸² Posttranscriptional processing includes “capping” at the 5′ end of the molecule, polyadenylation at the 3′ end, and “splicing,” which results in removal of so-called intervening sequences or introns. Abnormal splicing of, for example, β-globin mRNA transcripts is a common cause of thalassemia.

The primary structure of mRNA can be divided into four regions: the 5′ untranslated region (which includes the cap), the translated or coding region, the 3′ untranslated region, and the polyadenosine region. The “cap” is characterized by an atypical 5′-triphosphate-5′ linkage with guanosine-5′-triphosphate (GTP) and methylation of adjacent nucleotides. This structure appears to be essential for maximal translational activity and enhances mRNA stability. The cap is followed by an untranslated region (“5-UTR”) of 36 nucleotide bases in α-globin mRNA and 53 bases in β-globin mRNA. The difference may explain the observation that β-mRNA is translated more efficiently than α-mRNA.^78,80 Normally, this relatively inefficient translation of the α-chain is compensated for by an increased amount of α-mRNA. The translated sequence begins with an initiator sequence of three bases (AUG) followed by a sequence of triplet codons, each of which corresponds to an amino acid in globin, according to the genetic code. The translated sequence ends with a terminator codon (UAA), which is followed by a noncoding area and a terminal polyadenosine region, of variable length, that affects the stability or half-life of the molecule. The number of adenosine residues appears to decrease as mRNA ages⁸³; in older reticulocytes, mRNA may contain little or no polyadenosine. Globin mRNA is quite stable, with a half-life of up to 48 hours, allowing reticulocytes to continue to produce hemoglobin after enucleation.

Heme is of particular importance in controlling the rate of globin synthesis.^84,85 It stimulates globin synthesis in intact reticulocytes and cell-free systems, and, in its absence, polyribosomes disaggregate.^86,87 and ⁸⁸ The major effect of heme is exerted on the chain-initiation step in translation. In the absence of heme, an inhibitor of globin synthesis accumulates.^89,90 This inhibitor, heme-regulated eIF2α kinase (HRI), acts by phosphorylating the α-subunit of an initiation factor, eIF-2 that promotes binding of tRNA^met_F to ribosomes, to shut down protein synthesis.^91,92 HRI, which has two heme binding sites thus serves as a sensitive intracellular heme “sensor” that closely coordinates heme availability with globin chain production to prevent the accumulation of excessive unfolded globin proteins.⁹³ Studies of the hri knockout mouse verify the importance of this protective mechanism during high-level hemoglobin synthesis. HRI function is especially important in iron (resulting in heme) deficiency: In these circumstances, the continued globin production in hri^−/− mice results in cytotoxic globin protein precipitates, causing oxidative stress, and apoptosis of late erythroid precursors.⁹⁴ Interestingly, the red cells of iron deficient hri^−/− mice are normocytic/hyperchromic rather than microcytic/hypochromic and have globin chain inclusions. HRI also serves to ameliorate the phenotype of β-thalassemia in mice models, minimizing the production and accumulation of α-globin aggregates, erythroid precursor apoptosis, and ineffective erythropoiesis. Recent work indicates that HRI is activated by oxidative stress (e.g., during erythropoiesis in β-thalassemia).⁹⁵

In addition to its effects on globin translation, heme also exerts positive effects on globin transcription, through nuclear binding to the globin transcriptional repressor Bach1 during erythroid differentiation.⁹⁶

The enzyme responsible for catalyzing the condensation of succinyl CoA and glycine is ALA synthase (ALAS; EC 2.3.1.37). The products of the condensation reaction are ALA, CO₂, and free CoA. ALAS requires pyridoxal 5′-phosphate as a cofactor in the reaction as first suggested by nutritional studies in pigs.¹¹² There are two forms of the enzyme, one specific for erythroid cells (ALAS2) and the other (ALAS1) present in all other tissues, especially liver. ALAS2 uniquely appears to associate with succinyl CoA synthetase in the mitochondria to promote heme synthesis.¹¹³ The genes encoding for human ALAS have been identified and cloned.¹¹⁴ The erythroid-specific gene is present on chromosome Xp11, whereas ALAS1 is encoded by a gene on chromosome 3p21.^115,116

ALAS1 and ALAS2 differ predominantly at their amino terminal ends. The ALAS2 gene has 11 exons and extends over 22 kb. Notably, the mRNA contains a 5′ iron-response element (IRE) in exon 1. IRE are short hairpin structures that allow binding by the iron regulatory proteins 1 and 2 (IRP1 and IRP2) according to the cellular iron status, regulating the expression or stability of mRNAs encoding proteins involved in iron uptake, storage, utilization, or export^117,118 (see Chapter 23). ALAS2 translation begins in exon 2, which encodes the mitochondrial targeting sequence for this nuclear-encoded gene. As also seen with ALAS1 the catalytic domain is encoded by exons 5 to 11. Mutations of ALAS2, especially of exons 5 and 9, are the most common cause of sideroblastic anemias (see Chapter 24). The promoter contains binding sites for GATA1, KLF1, and NF-E2 (although assays suggest the latter may not be functional¹¹⁹). However, regions further
upstream may also be important for erythroid-specific ALAS2 induction.¹²⁰ In addition, analysis of intron 8 indicated the presence of DNAse I hypersensitive sites; the intron appears to contain functional binding sites for GATA1 and Sp1.¹²¹ Of relevance, recent studies identifying SLC25A38 as another frequent genetic cause of congenital sideroblastic anemias indicate it also affects either ALA synthesis or transport. Studies of yeast deficient in the yeast ortholog suggest the gene encodes a mitochondrial transporter mediating either ALA export into the cytosol or glycine import into the mitochondrial matrix, as ALA levels were low and the heme synthesis impairment was rescued by glycine or ALA supplementation.¹²²

The monopyrrole porphobilinogen functions as a precursor of the hemes, chlorophylls, and the cobalamins.¹²³ Because virtually all known forms of life require at least one of these classes
of tetrapyrroles, it follows that porphobilinogen is biologically ubiquitous. Porphobilinogen is formed by the condensation of two molecules of ALA and the loss of two water molecules. The enzyme that catalyzes this reaction is porphobilinogen synthase or ALA dehydratase (EC 4.2.1.24). ALA dehydratase is a soluble enzyme found in the cytosol⁹⁸ and abundant in tissues such as bone marrow and liver where heme biosynthesis is active. It is also active in mature circulating erythrocytes, even though these cells are not actively synthesizing heme. Its persistence results from the enzyme’s inherent stability.

The mammalian enzyme is an octamer of 31 kDa subunits containing zinc atoms required for stability and activity.¹²⁴ The enzyme is inhibited by heavy metals, particularly lead, which displaces the catalytically active zinc ions and thus enzyme activity can serve as an index of environmental pollution by lead.^125,126 Free sulfhydryl groups (-SH) are also essential for activity,^127,128 and these -SH groups seem to be protected by the zinc. Three isoforms of ALA dehydratase have been reported and the gene identified (on chromosome 9q34) and cloned.^129,130 The gene consists of 12 exons, with two promoters generating different transcripts (but encoding the same protein) in erythroid and nonerythroid tissues. The erythroid promoter contains GATA1-binding and CACCC sites.

Porphobilinogen is a rather unstable, chemically reactive molecule. Within a few hours, a solution of porphobilinogen exposed to air and light develops a deep orange-red color. The color results from the formation of porphobilin, a poorly defined mixture of mono-, di-, and tripyrrolic oxidation products.⁹⁸ This phenomenon can be observed in the urine of patients with acute intermittent porphyria, who excrete large quantities of porphobilinogen (see Chapter 30).

If instead porphobilinogen is incubated in solution at an acid pH then nonenzymic condensation or cyclization occurs, forming the tetrapyrrole macrocycle uroporphyrinogen.⁹⁸ All four possible isomers of uroporphyrinogen are formed under these conditions. The reaction is often referred to as a “head-to-tail” condensation because of the apparent orientation of the precursor molecules. The penultimate reaction product apparently is an open-chain (linear) tetrapyrrole called hydroxymethylbiliane (HMB). On loss of the amino group attached to ring A, the -CH₂⁺ can attack the free α-position on the D-ring pyrrole, thus forming the macrocycle.¹²⁴

In vivo at neutral pH, these reactions are catalyzed by the cytosolic enzyme HMB synthase, formerly known as uroporphyrinogen I synthase or porphobilinogen deaminase (EC 2.5.1.61). In the cytosol HMB synthase works in concert with a second enzyme, uroporphyrinogen III synthase¹³¹ (EC 4.2.1.75), to form uroporphyrinogen III, an asymmetric cyclic tetrapyrrole that serves as the common precursor of all known functional tetrapyrroles. Uroporphyrinogen III synthase is a very fast enzyme, outcompeting the uncatalyzed reaction of HMB that results in uroporphyrinogen I (above). HMB synthase is present in humans in two tissue-specific isoenzymes, an erythroid tissue-specific and a shorter nonerythroid form, both of which are products of a single gene located on chromosome 11q23.^{98,124,132,133} and ¹³⁴

The molecular mechanism by which uroporphyrinogen III synthase effects the “turning around” of the D ring has been studied intensively.^98,135,136 The molecular mechanism involved in this reaction has been clarified by nuclear magnetic resonance spectroscopy.^123,135,136 and ¹³⁷ HMB synthase first catalyzes the head-totail condensation of four porphobilinogen molecules, yielding the aminomethyl tetrapyrrole. The tetrapyrrole is then deaminated, yielding a macrocycle that has been termed preuroporphyrinogen,¹³⁷ which is released from the surface of HMB synthase and serves as the substrate for uroporphyrinogen III synthase.¹³⁷ This enzyme opens the bond linking the methylene bridge carbon to the D ring, allowing the D ring to rotate 180° about the carbon-nitrogen bond linking rings A and D. A new carbon-carbon bond is formed linking rings C and D (1,3-sigmatropic shift). The carbon-nitrogen bond then opens, and a new carbon-carbon bond is formed linking rings A and D (1,5-sigmatropic shift) and yielding uroporphyrinogen III.¹³⁷

The gene for uroporphyrinogen III synthase, located on 10q25.3, has alternative promoters in nonerythroid tissues (with transcripts starting in exon 1) and erythroid tissues (transcripts begin in exon 2). As there is no translation initiation site in exon 1, however, identical transcripts are generated. The intron between exons 1 and 2 contains multiple putative GATA1 binding sites. Mutations of the uroporphyrinogen III synthase gene are associated with congenital erythropoietic porphyria (Chapter 26).

The formation of coproporphyrinogen III is accomplished by the enzymic decarboxylation of the four acetic acid side chains of uroporphyrinogen III,⁹⁹ a reaction catalyzed by the cytosolic enzyme, uroporphyrinogen decarboxylase (EC 4.1.1.37). The decarboxylation proceeds in a clockwise fashion, starting with the acetic acid on the D ring of uroporphyrinogen III and continuing with the successive decarboxylations of the acetic acid residues on rings A, B, and C.¹³⁸ The enzyme exists as a dimer and it was proposed that the two catalytic centers interact functionally, perhaps shuttling reaction intermediates between subunits.¹³⁹ Recent mutational studies, however, indicate that shuttling is unlikely: Dimerization mediates a more active enzyme than the monomer but without any change in the reaction products.¹⁴⁰

The gene, located on chromosome 1p34, consists of 10 exons spread over 3 kb. Mutations of the gene encoding uroporphyrinogen decarboxylase are associated with porphyria cutanea tarda (PCT), the most common human porphyria. The structural consequences of some of these mutations have been described.¹⁴¹ The genetics of the disease are complex, however, with some variants without evidence of gene mutations (sporadic PCT type I; see Chapter 30 for details). A cytosolic competitive inhibitor of the enzyme, porphomethene, has been described in a murine model of PCT as a potential explanation for PCT patients with normal levels of the protein but reduced hepatic enzyme activity.¹⁴²

The formation of protoporphyrinogen IX from coproporphyrinogen III is catalyzed by the enzyme coproporphyrinogen oxidase (EC 1.3.3.3). The exact location of coproporphyrinogen oxidase in the mitochondria is unknown, with studies suggesting it is present in the intermembranous space and/or loosely associated with the inner surface of the outer membrane.^143,144 It may instead form a macrocomplex with protoporphyrinogen IX oxidase and Fech to allow funneling of the substrates from the cytosol into the matrix¹¹¹ (reaction products after uroporphyrinogen III being poorly soluble in aqueous solutions). The enzyme, which has a long mitochondrial targeting sequence, functions as a dimer and has an absolute requirement for molecular oxygen.^143,145,146 and ¹⁴⁷ It sequentially and oxidatively decarboxylates the propionic acid side chains in rings A and B (but not C or D) of coproporphyrinogen III to form vinyl groups.¹⁴⁸ The transporter mediating influx of cytosolic coproporphyrinogen III into the intermembranous space is unknown.

The gene encoding coproporphyrinogen oxidase is located at 3q12, contains seven exons and spans 14 kb. Studies of the murine promoter indicate that SP1, GATA1 binding sites, and a novel transcription factor-binding sequence termed CPRE are important for induction of the enzyme during erythroid
differentiation of MEL cells.¹⁴⁹ Mutations of the gene are associated with hereditary coproporphyria and, more rarely, with erythropoietic harderoporphyria (see Chapter 30).

The product of the coproporphyrinogen III oxidase reaction is protoporphyrinogen IX. To serve as a substrate for the final enzyme in the pathway (ferrochelatase), protoporphyrinogen IX must first be oxidized to protoporphyrin IX. Although protoporphyrinogen IX is easily oxidized nonenzymatically to protoporphyrin in vitro (or in the cytosol), an enzyme is required to catalyze this reaction in vivo. A membrane-associated, mitochondrial, oxidizing dimeric enzyme—protoporphyrinogen IX oxidase (EC 1.3.3.4)— has been demonstrated in mammalian cells, including rat liver, human fibroblasts, reticulocytes, and leukocytes.¹⁵⁰ The crystal structure of the tobacco plant mitochondrial enzyme has been described.¹¹¹ The protein lacks a typical mitochondrial targeting leader sequence and is targeted by just 17 residues in the amino terminus.¹⁴⁵ Flavin adenine dinucleotide (FAD) serves as an essential cofactor and molecular oxygen is utilized to terminally accept the electrons.¹⁴⁴

The gene is present on chromosome 1q23.3 and consists of 13 exons extending over 4.2 kb. The promoters of the murine and human genes have been characterized^151,152; both contain SP1 and GATA1 binding sites that may be of importance during erythroid differentiation. Mutations of the gene are associated with variegate porphyria (see Chapter 30).

The insertion of ferrous iron into protoporphyrin IX to form heme is catalyzed by the enzyme ferrochelatase within the mitochondrial matrix. Ferrochelatase (EC 4.99.1.1) is the best characterized of the heme biosynthesis enzymes.¹⁵³ The enzyme, which functions as a dimer appears to be tightly bound to or is an integral part of the inner mitochondrial membrane,¹⁵⁴ likely complexed with protoporphyrinogen IX oxidase dimers on the opposite side of the membrane.¹⁴⁴ Insertion of the iron appears to involve distortion of the planar porphyrin by ferrochelatase.¹⁵³ No cofactors are required for activity. Although the in vivo substrates are ferrous iron and protoporphyrin, in vitro, the enzyme can also catalyze incorporation of several metals (iron, cobalt, and zinc) into several dicarboxylic porphyrins (protoporphyrin, mesoporphyrin, and deuteroporphyrin).⁹⁸

Studies in cell lines demonstrate that ferrochelatase interacts with Mfrn1, a mitochondrial inner membrane importer of iron up-regulated during erythroid differentiation.¹⁵⁵ In elegant studies it has been shown that Mfrn1 is stabilized by an inner membrane ATP-binding cassette (ABC) transporter known as ABC-me or ABCB10,¹⁵⁶ allowing high-efficiency iron uptake into mitochondria for optimal heme synthesis.¹⁵⁷ Further studies show that ferrochelatase, Mfrn1, and ABCB10 are co-induced during MEL cell erythroid differentiation and that Mfrn1 and ABCB10 interact with ferrochelatase.¹⁵⁵ The gene encoding ABCB10 was originally identified in a screen of mouse genes up-regulated by GATA1 and expression is regulated by heme.¹⁵⁶ The potential substrate of the ABCB10 transporter has not been identified, but heme, which needs to be transferred from the matrix across two mitochondrial membranes into the cytosol, is a candidate.

Apart from its use for heme synthesis, mitochondrial iron is also required for mitochondrial Fe-S cluster biogenesis. Fe-S clusters are modular protein co-factors consisting of iron and sulfur, usually linked by bonds joining the cysteine sulfur atoms of a polypeptide (“scaffold”) protein to iron atoms of the cluster. They function, for example, as part of enzyme catalytic centers (e.g., aconitase, succinate dehydrogenase).¹⁵⁸ Ferrochelatase contains a NO-sensitive Fe-S cluster that is attached at the C-terminus.¹⁵³ Although the cluster is not required for catalytic function or as a supply of ferrous iron, the enzyme is sensitive to the availability of Fe-S clusters. For example, when Fe-S cluster synthesis is impaired in MEL cells, due to deficiency of the scaffold proteins or iron required for Fe-S cluster biosynthesis, then apo-Fech is rapidly degraded in mitochondria, indicating a direct link between biosynthesis of Fe-S clusters and heme.¹⁵⁹

The gene encoding ferrochelatase is comprised of 11 exons spread out across 45 kb on chromosome 18q21.3.¹⁶⁰ The promoter region has been examined in detail using in vitro and in vivo studies.^161,162 and ¹⁶³ Sp1, NF-E2, and GATA1 elements have been identified in the promoter region and a fragment of the promoter containing these binding sites allows expression in hematopoietic cells derived from transgenic embryonic mouse cells where a single copy of the reporter construct was inserted. In vivo erythroid specificity is mediated by NF-E2 elements ˜300 bp upstream of the transcriptional start site (-275 bp) along with additional erythroid-specific elements that lie between -375 bp and -1,100 bp upstream from the start site.¹⁶¹ In in vitro assays in K562 cells, the Kruppel-like transcription factor KLF-13 activates the promoters for porphobilinogen deaminase, δ-aminolevulinate synthase, and ferrochelatase genes.¹⁶⁴ Mutations of the ferrochelatase gene are associated with erythropoietic porphyria (see Chapter 26).

About 15% of the daily production of heme is generated in the liver by ALAS1 for cytochromes and enzymes. The amount of ALAS1 is regulated by induction and repression of enzyme synthesis¹²⁷ and may increase by a factor as great as 300-fold.¹⁶⁶ The enzyme has a short half-life, allowing a rapid response to changes in the demand for heme and, thus, ALA.¹⁶⁷ The enzyme may be induced by a number of chemicals, drugs, and nonglucocorticoid steroids.¹²⁷ Heme plays a critical central role in ALAS1 regulation, repressing transcription,¹⁶⁸ decreasing the half-life of the mRNA, and, through binding to heme regulatory motifs (HRMs) in the 5′ end of the protein, translocation of the enzyme into the mitochondrial matrix.¹⁶⁹ When the amount of intracellular heme is high, ALAS1 synthesis is repressed; when the amount of heme is low, synthesis is induced. Thus, agents that interfere with heme synthesis can induce ALAS1, and agents that induce the synthesis of hemoproteins (e.g., induction of cytochrome p450 enzymes by barbiturates), potentially depleting a putative pool of “free” or “uncommitted” heme, can produce a similar effect.^170,171 Agents that exert these effects on ALAS1 synthesis induction are clinically important, as they may precipitate acute attacks in patients with acute intermittent porphyria and related disorders of porphyrin metabolism (Chapter 26).

Early studies indicated that ALAS1 synthesis is induced during fasting, which can precipitate porphyria attacks, and that nutritional supplements (e.g., glucose loading) may ameliorate
these acute episodes. Handschin et al. recently demonstrated that this occurs because ALAS1 is induced by the concerted actions of peroxisome proliferator-activated receptor γ coactivator 1α (PGC-1α) and the transcription factors FOXO1 and nuclear respiratory factor-1 (NRF-1) on the ALAS1 promoter.¹⁷² PGC-1α, which co-activates nuclear receptors or transcription factors to regulate mitochondrial biogenesis and oxidative phosphorylation, is induced by low glucose levels or glucagon and repressed by high glucose or insulin.¹⁷³ Presumably ALAS1-mediated heme synthesis is required for de novo respiratory cytochrome synthesis with mitochondrial biogenesis as a response to decreased cellular ATP levels. In contrast, regulation of ALAS2 in erythroid cells is coordinated with heme synthesis, iron assimilation, and globin synthesis.

With erythroid differentiation there is coordination of cellular iron assimilation, heme, and globin synthesis to safely allow maximal hemoglobin synthesis within a short timespan without the buildup of individual, potentially cytotoxic components (see below). Regulation of erythroid heme involves the induction of the enzymes of the heme biosynthetic pathway and their regulation once induced,^105,107 regulation of iron uptake and its delivery to ferrochelatase in the mitochondria, and the regulated export of the newly formed heme from the mitochondria to the cytosol to bind to globin chains.

Initial studies suggested that the heme biosynthesis pathway enzymes were sequentially induced with erythroid differentiation (from ALAS2 to ferrochelatase¹⁷⁴); however, subsequent studies in MEL cells¹⁵² and maturing populations of human erythroblasts¹⁷⁵ indicate early up-regulation of ferrochelatase with ALAS2, with synthesis of mRNAs for the terminal three pathway enzymes up-regulated within 12 hours of erythroid induction of MEL cells with dimethylsulfoxide (DMSO). In contrast to the repressive effects of heme on ALAS1 in hepatocytes, in erythroid cells intracellular heme appears to be necessary for induction of the biosynthetic pathway enzymes, perhaps through up-regulation of the erythroid transcription factor NF-E2,¹⁷⁶ and/or inhibition of the transcriptional repressor Bach1.⁹⁶ For example, studies of ALAS2-deficient MEL cells following their “erythroid” induction with DMSO demonstrate a lack of erythroid differentiation (as assessed by a lack of up-regulation of mRNAs for ALA dehydratase, porphobilinogen deaminase, ferrochelatase, and β-globin) that is at least partially reversed by addition of heme to DMSO, in keeping with studies by others of ALAS2-deficient cell lines.¹⁷⁷

As discussed above, unlike ALAS1, there is an IRE in the 5′UTR of the ALAS2 gene. Therefore depletion of cytosolic iron (believed to exist in a putative labile iron pool) in erythroid precursors should result in binding by IRP1 and IRP2 to the IRE of ALAS2 transcripts, preventing translation. Thus, it is the supply of iron to the erythroid precursor that ultimately controls heme synthesis. To allow maximal heme synthesis—which requires ˜20 mg of iron for the 20 g of erythrocytes generated in adult humans every day—iron is delivered to the bone marrow in the form of ferric-transferrin (Fe-Tf) that is rapidly bound by TfR1 present in large numbers on the cell surface of erythroid precursors (up to 10⁶ receptors per cell¹⁷⁸). In addition, as noted above, the mitochondrial iron importer MFN1 becomes stabilized in a macrocomplex with ferrochelatase and ABCB10 following up-regulation of ABCB10 with erythroid differentiation, thus facilitating the transfer of iron across the mitochondrial inner membrane to ferrochelatase. By necessity, there must also be up-regulation of an unidentified heme transporter to export heme from the mitochondria into the cytosol. As a feedback mechanism “uncommitted” or “free” heme appears to inhibit either Fe-Tf-TfR endocytosis or iron release from Tf to prevent unnecessary iron uptake.^105,179 Like ALAS1, ALAS2 also has HRMs located in the 5′ end of the preALAS2 protein and, in vitro, micromolar quantities of heme inhibit translocation of ALAS2 into isolated mitochondria; however, whether excess uncommitted heme also impedes ALAS2 translocation in erythroid precursors is unknown.¹⁶⁹

ALAS2 and ferritin transcripts both contain an IRE in their 5′-UTR and their translation is thus susceptible to low cytosolic iron levels; on the other hand, TfR1 contains five IRE modules in its long 3′-UTR and the mRNA should be stabilized (predominantly by IRP2¹⁸⁰) under the same conditions. However, the standard posttranscriptional regulatory model involving cytosolic iron levels controlling iron transport, utilization, and storage via the IRE/IRP system (see Chapter 23) appears to become uncoupled in differentiating erythroid cells: Here ALAS2 is translated yet ferritin translation blocked, whereas high expression of TfR1 persists despite high Fe-Tf delivery. Recent analyses of mass cultures of erythroid progenitors indicate that, with differentiation, these cells behave as if a “low cytosolic iron level” condition exists.¹⁸¹ This is in keeping with the “kiss and run” hypothesis of iron delivery proposed by Ponka and co-workers,¹⁸² whereby iron released from Fe-Tf-TfR complexes in the endosome bypasses the cytosol to be delivered to the mitochondria by direct contact between these two organelles. An alternative theory is that MFN1 functions as a highly efficient mitochondrial iron importer driving cytosolic iron transfer across inner mitochondrial membranes for heme or Fe/S synthesis. In addition, it is suggested that the presence of large numbers of ALAS2 transcripts may overwhelm the IRE/IRP system.¹⁸¹

In closing this section, note should be made of a number of recently described mechanisms to protect erythroid precursors from the results of any imbalance among iron assimilation, heme, and globin synthesis. Although each component is essential for hemoglobin synthesis, individually they are all potentially cytotoxic and it is therefore crucial that the α– and β-globin chains and heme are produced in the 2:2:4 ratio necessary to form the stable complex of α₂β₂ and 4 heme molecules that comprise hemoglobin A. The toxicity of free iron is well known, related to its intrinsic ability to generate highly reactive hydroxyl radicals from hydrogen peroxide in the Fenton reaction, whereas “free” or uncommitted heme is lipophilic and toxic to cells, promoting lipid peroxidation and ROS production, resulting in membrane injury and ultimately cell apoptosis.^106,183,184 The cytotoxicity derived from an imbalance in the production of α-globin and β-globin chains is best illustrated by the pathophysiology of β-thalassemia (Chapter 34), where the relative excess in α-globin production and the resultant precipitation of these globin chains triggers oxidative stress and cytotoxicity (first described in the 1960s^185,186).

These protective systems include the following.

(i) The heme-regulated inhibitor of translation, HRI

As mentioned, HRI, a heme-regulated protein kinase that phosphorylates and inhibits eIF2α and thus general protein translation, serves as a sensor of intracellular heme and is important for coordinating heme and globin production. Studies of the hri knockout mouse verify the importance of this protective mechanism during high-level hemoglobin synthesis. HRI function appears to be especially important in iron (resulting in heme) deficiency: In these circumstances, the cessation of globin production seen in control mice presumably does not occur in hri^−/− mice, resulting in the observed cytotoxic globin protein precipitates, ROS production and oxidative stress, and apoptosis of late erythroid precursors.⁹⁴ Interestingly, the red cells of irondeficient hri^−/− mice are normocytic/hyperchromic (rather than
microcytic/hypochromic, as observed in the iron-deficient control mice), have globin chain inclusions, and a more severe anemia. Similarly, mice with combined HRI and severe ferrochelatase deficiency (Fech^m1Pas/m1Pas mice, another model of heme deficiency) have more severe anemia and globin chain inclusions than control Fech^m1Pas/m1Pas mice. Notably, these mice have a 30-fold increase in red cell protoporphyrin IX compared with Fe^{chm1Pas/m1Pas} mice, emphasizing that HRI-mediated regulation of erythroid precursor protein synthesis also affects heme enzyme biosynthesis and the severity of the porphyria.¹⁷²

HRI also ameliorates the phenotype of β-thalassemia in mice models, minimizing the imbalance in production between α– and β-globins, the accumulation of α-globin aggregates, apoptosis of erythroid precursors, and the resultant ineffective erythropoiesis. Recent work indicates that HRI is also activated by oxidative stress (e.g., during stress erythropoiesis or erythropoiesis in β-thalassemia).⁹⁵ HRI, by activating the transcription factor ATf4, induces an antioxidant stress response that seems to be important for erythroid differentiation.

(ii) The feline leukemia virus subgroup C receptor, FLVCR

FLVCR1 is the human ortholog of the feline cell surface receptor for feline leukemia virus subgroup C (FeLV-C). The virus infects all feline hematopoietic cells, impairing feline FLVCR1 function due to binding of the receptor by viral envelope that is continuously synthesized within infected cells. Cats infected with FeLV-C develop a red cell aplasia characterized by a block in erythroid differentiation at the CFU-E/proerythroblast stage. The impairment in differentiation is also observed upon conditional deletion of flvcr1 in neonatal mice,¹⁸⁷ who develop a severe anemia within 5 weeks of deletion¹⁸⁷ that may be due to erythroid cell apoptosis.¹⁸⁸ FLVCR1 functions as a mammalian cell-surface heme exporter that thus appears to protect differentiating erythroid progenitors from potential heme excesses and subsequent cytotoxicity resulting from any imbalances between heme and globin synthesis.^104,188 As noted, heme synthesis by erythroid ALAS2 is not subject to transcriptional repression by heme and catabolism by heme oxygenases does not normally occur during differentiation of human erythroid progenitors¹⁸⁹ or murine erythroid cell lines.¹⁹⁰ It is hypothesized that heme oxygenase is not induced in order to prevent futile cycles of simultaneous erythroid heme synthesis and catabolism.^188,189 The severity of the anemia in feline and murine models of FLVCR knockdown suggests that excess heme synthesis occurs frequently at the CFU-E/proerythroblast stage (likely prior to initiation of high-level globin synthesis) or that FLVCR has yet another cell function. A recent study suggests another FLVCR isoform functions to export heme from its site of synthesis in the mitochondria into the cytosol^190a. Of interest, knockdown of murine flvcr1 is embryonic lethal with the embryos displaying a phenotype similar to that of patients with Diamond-Blackfan anemia (DBA), a congenital red cell aplasia.¹⁹¹ The most common genetic cause of DBA is haplo-insufficiency of large or small ribosomal protein subunits, impairing erythroid protein (i.e., predominantly globin) synthesis, suggesting DBA is caused by an imbalance of heme and globin synthesis.¹⁹²

(iii) The α-hemoglobin stabilizing protein, AHSP

The gene encoding this small protein (102 aa) is strongly induced by GATA-1 during erythroid differentiation. AHSP primarily binds αHb (i.e., the holoprotein, α-globin-heme), stabilizing it and inhibiting its pro-oxidant properties.⁵³ It functions as a chaperone, helping newly synthesized apo α-globin chains to fold and promoting the refolding of denatured chains, which may be particularly important in heme deficiency.¹⁹³ Studies have shown that AHSP forms a heterodimer with αHb and when βHb is added to these complexes, the AHSP is displaced and tetrameric HbA (α₂β₂) forms, suggesting that AHSP stabilizes αHb and then passes it to βHb to help form HbA in vivo. Deletion of the gene in mice¹⁹⁴ results in a mild hemolytic anemia with the red cells containing Heinz bodies (eosinophilic inclusions derived from denatured hemoglobin), indicating perhaps that AHSP function is critical only when there is a large imbalance in synthesis between αHb and βHb, such as occurs in β-thalassemia. In subsequent studies, interbreeding of AHSP^−/− mice with β-thalassemic mice (like the Hri knockout above) was indeed shown to worsen the β-thalassemic phenotype.¹⁹⁵

Lastly, “protein quality control mechanisms” has been proposed as a name for a number of cellular posttranslational mechanisms that serve to stabilize and aid folding of newly forming proteins (e.g., chaperones such as AHSP), or recognize misfolded proteins or protein aggregates and target them for degradation by the ubiquitin proteosome system or autophagy.¹⁹⁶ These protein quality control mechanisms appear to be particularly important in erythroid precursors during high-level hemoglobin synthesis, and when these systems are overwhelmed—for example, in severe thalassemias—accumulation of unstable insoluble proteins and cytotoxicity occurs.¹⁹⁷

Epo is a glycoprotein hormone produced by the kidney, which functions as the major humoral regulator of red cell production. The hormone was originally purified from the urine of patients with aplastic anemia.¹⁹⁸ It has an MW of 34 kDa as determined electrophoretically and contains 30% carbohydrate (11% sialic acid, 11% total hexose, and 8% N-acetylglucosamine).¹⁹⁹ The potency of Epo is expressed in units, with one unit defined as the amount of Epo present in one tenth of the International Reference Preparation.²⁰⁰ This unit was originally defined in starved rats as the amount of Epo that produced the same erythropoietic response (increase in serum Epo level) as treatment with 5 µmol of cobalt.² The potency of purified human urinary Epo has been determined to be 70,400 U/mg of protein.¹⁹⁸

The Epo gene has been isolated and cloned²⁰¹; the genomic locus extends over 5.4 kb on chromosome 7q22²⁰² and contains four introns and five exons, encoding a 193-amino acid polypeptide. The protein includes a 27-amino acid signal peptide, which is cleared during Epo secretion, and a 166-amino acid peptide with an MW of 18.4 kDa.⁴⁶ The C-terminal arginine is absent from the recombinant and native proteins, presumably because of posttranslational modification by a carboxypeptidase.²⁰³ Human Epo contains four cysteine residues linked by disulfide bonds, which, when reduced or alkylated, lead to significant loss of activity.^199,204

Recombinant Epo (rEpo), as synthesized by mammalian cells such as a Chinese hamster ovary cell line, is highly glycosylated, and the carbohydrate structure of the recombinant hormone is similar to but distinct from the glycosylation pattern of native Epo in kidney cells.²⁰³ Glycosylation is absolutely necessary for in vivo activity of Epo, with the bulk of glycosylation occurring at a single site of N-linked carbohydrate. Asialated Epo and nonglycosylated rEpo produced in bacteria have no activity in vivo, which is at least partially attributed to rapid clearance of the hormone by the liver via hepatocyte galactose receptors.^205,206

Recognition of the importance of glycosylation of Epo to its in vivo activity and half-life led to modifications of the Epo gene/protein to make a more effective pharmaceutical. For example, the gene has been modified by adding a second site of N-linked glycosylation, such that when the gene is expressed in Chinese hamster ovary cells, the amount of carbohydrate attached to the modified protein is almost doubled. This new product, called darbepoetin, has a longer in vivo half-life than rEpo, thus fewer injections per week are required for therapeutic efficacy.²⁰⁷

Studies on the amino acid sequences of human and murine Epo have shown a very high degree of conservation of the molecule structure in these two species.^208,209 Analyses of the NMR structure of Epo and the crystal structure of Epo complexed to the extracellular component of the EpoR have been performed.²¹⁰ Similar to the structure of hGH and GCSF (Chapter 5), Epo consists topologically of a bundle of α-helices (four in Epo²¹⁰). Two Epo-EpoR binding sites on Epo have been identified, a high-affinity site 1 (K_d ˜ 1 nM) and low-affinity site 2 (K_d ˜ 1 µM).²¹¹ The studies are in keeping with the interaction of a single molecule of Epo with EpoR dimers.

More than 50 years ago Jacobson et al. established that the kidney is the major site of Epo production in adult rats.²¹² Humans with end-stage renal failure were also found to have low serum Epo concentrations, which were restored to normal following renal transplantation.²¹³ The cloning of the murine Epo gene has allowed studies on the production of Epo-specific mRNA in anemic mice. Induction of anemia leads, within an hour, to the appearance of Epo-encoding mRNA in the kidney and liver of anemic mice and rats.^213,214 After bleeding, the Epo mRNA in the kidney increases 500 to 1,000 times when compared with Epo mRNA levels in normal kidney, whereas the liver produces only 7% of the total Epo mRNA.²¹⁵ These changes in Epo mRNA synthesis are followed by parallel changes in serum Epo concentration, as determined by radioimmunoassay, indicating that Epo production in response to anemia represents de novo synthesis rather than the release of preformed hormone.²¹⁴ Murine Epo mRNA is detectable by ribonuclease protection assay at 14 days of gestation in the fetal liver and a week later in the kidneys, which assume a major role in Epo production after birth.²¹⁶ In cases of paraneoplastic erythrocytosis, Epo mRNA is detected in the neoplastic cells.^217,218

Specialized cells producing Epo have been identified in renal and hepatic parenchyma by in situ hybridization techniques, using radioactive probes specific for Epo mRNA.^216,219,220 These rare Epo-producing cells are found in the renal parenchymal interstitium (outside the tubular basement membrane), predominantly in the inner cortex and outer medulla. The bulk of experimental evidence indicates that these are fibroblast-like type I interstitial cells.^221,222 In the liver, Epo mRNA is detected in hepatocytes. The number of interstitial renal Epo-producing cells increases (approximately exponentially) in response to anemia, indicating that increased Epo production is met by an increase in the number of Epo-producing cells; presumably with worsening anemia, increased numbers of these cells become sufficiently hypoxic to trigger Epo synthesis.⁶² Notably, there is no detectable storage of the hormone, and increased levels of circulating Epo do not repress further Epo production. Epo and EpoR are also expressed at low levels in other tissues including the spleen, bone marrow, lung, testis, eye, and brain.²²³ A possible paracrine function, potentially supporting low-level erythropoiesis, has been ascribed to Epo production within bone marrow hematopoietic progenitors,^224,225 and production of Epo within the CNS appears to protect EpoR-bearing neurons from ischemic damage and apoptosis.^226,227

The mechanism by which hypoxia leads to Epo synthesis has been determined. A sequence located in a region flanking the 3′ end of the Epo gene is oxygen-sensitive and involved in regulation of expression.²²⁷ It has been shown, for example, that this oxygen-sensitive enhancer sequence when fused to a plasmid construct comprising the Epo promoter and a reporter gene confers to transfected cells the ability to respond to hypoxia as detected by an increase of the protein encoded by the reporter gene and studies of transgenic mice expressing defined 5′ or 3′ sequences of the human Epo gene support these findings.²²⁸ The ligand for this oxygen-sensitive enhancer was identified as a 120 kDa protein termed hypoxia-inducible factor 1 (HIF-1).^228,229 This DNAbinding protein is tightly regulated by intracellular oxygen tension and serves as the physiologic regulator of Epo transcription.²³⁰

HIFs are heterodimeric helix-loop-helix transcription factors consisting of two subunits, an oxygen-labile protein, HIF-α, and a constitutively expressed β subunit, HIF-β.²³¹ Note that three genes, HIF1A, HIF2A or EPAS1, and HIF3A, encoding different isoforms of HIF-α are present in the human genome; here HIF-α refers to HIF-1α or HIF-2α. The concentration and transcriptional activity of HIF-α increase in a geometric fashion upon exposure to hypoxia. HIF-α mRNA is constitutively expressed under normoxic conditions, but the protein is rapidly degraded via the ubiquitin proteosome complex following binding by von Hippel-Lindau protein (pVHL). The recognition of HIF-α by pVHL requires prior hydroxylation of specific HIF-α proline residues by prolyl-hydroxylase domain (PHD)-containing proteins.^232,233,234 and ²³⁵ These PHDs are oxygen- and iron-dependent enzymes. Under hypoxic conditions, little or no proline hydroxylation takes place; thus, pVHL does not bind to HIF-α, which accumulates in the nucleus, heterodimerizes with HIF-β, and recruits the transcriptional coactivators p300/CREB-binding
protein, with the whole complex then binding to the Epo enhancer to positively influence Epo promoter activity and gene transcription. The recruitment of p300 to the complex can itself be inhibited by hydroxylation of asparagine-803 in HIF-α, which is catalyzed by asparaginyl-hydroxylase, another oxygen-sensitive enzyme.²³⁶ It seems that these two amino acid hydroxylases, by their dependence on normal intracellular oxygen for their function, act as the oxygen sensor in the Epo-producing interstitial cells in the kidney, and, by regulating the function of HIF-α at two distinct points,²³⁵ ultimately control Epo synthesis and production. Not surprisingly, mutations of this pathway may be associated with an increased red cell mass: Chuvash polycythemia is due to a homozygous mutation of Vhl that impairs HIF-1α degradation, resulting in a mild increase in Epo levels²³⁷ whereas mutations of the genes encoding HIF-2α or PHD2 are rarer causes of familial erythrocytosis.²³⁸

In addition to Epo, a large number of other HIF target genes (e.g., glucose transporters, glycolytic enzymes, and vascular endothelial growth factors) are up-regulated during hypoxia to aid cells adapt to hypoxic conditions.²³⁹ Apart from the indirect effects of hypoxia on erythropoiesis through HIF-mediated renal Epo production, erythroid progenitor cells are also subject to direct cellular effects of hypoxia and HIF production. A recent analysis of early erythroid progenitor genes up-regulated by glucocorticoids (such as cortisol, which is released from the adrenal glands during acute anemic or hypoxic “stress erythropoiesis”^38,240) found that the promoter regions of many of these erythroid cell genes also contain HIF binding sites. Furthermore, in in vitro studies, HIF synergizes with glucocorticoids to expand erythroid progenitors dramatically, perhaps by increasing BFU-E progenitor self-renewal,³⁸ a HIF effect previously observed by others during stress erythropoiesis.²⁴¹

Epo binds to specific molecules on the cell surface, the EpoR. The expression of both Epo and EpoR is necessary for adult life. Deletion of either of the genes encoding for Epo or EpoR in mice results in the identical phenotype of fetal death at embryonic days E11.5 to E13.5 because of a lack of definitive erythropoiesis in the fetal liver and severe anemia.²⁴² Arguably, the most important control point of erythropoiesis is the interaction of Epo with the receptor for Epo.^38,243,244 and ²⁴⁵ The activation of EpoR generates an intracellular signal in immature erythroid cells that promotes survival of cells that would otherwise undergo apoptosis. Epo also appears to promote erythroblast proliferation and differentiation.

EpoR is expressed on hematopoietic cells that respond to Epo and has been identified on human¹² and murine erythroid cells,²¹⁶ on erythroleukemia cell lines,²¹⁶ in murine fetal liver tissue rich in erythroid elements, in mouse and rat placenta,^217,246 and on megakaryocytes.^246,247 EpoR expression on erythroid cells is relatively low (approximately 1,000 molecules per cell) and correlates with the cell’s responsiveness to and dependence on Epo.²⁴⁷ EpoR is detectable by autoradiography on human BFU-E; its density increases as BFU-E matures to CFU-E.¹⁹ Erythroid cells at a stage between CFU-E and proerythroblast seem to have the highest density of EpoR, which decreases as the proerythroblast matures and eventually disappears at the stage of orthochromatic erythroblast.^19,246 The receptor is not expressed on reticulocytes or red cells. The presence of EpoR on megakaryocytes^246,247 and ²⁴⁸ explains why Epo at physiologic concentrations promotes megakaryocyte differentiation and can thus affect platelet levels. Receptors for Epo are also observed on nonhematopoietic tissues including neurons and cardiac myocytes, endothelial cells, the kidneys, and embryonic muscle. The expression of EpoR in nonerythroid tissues was not believed to be required for normal embryonic development as erythroid tissue-specific EpoR expression (under control of the GATA-1 promoter) in EpoR^−/− embryonic stem cells gives rise to apparently normal mice.²⁴⁹ However, recent studies indicate that EpoR^−/− murine embryos develop neural defects,²⁵⁰ defects in angiogenesis,²⁵¹ and cardiac ventricular hypoplasia, which is not a result of generalized hypoxia.²⁵² In addition, some of the adverse effects observed in patients with certain solid tumors receiving Epo have been ascribed to activation of EpoR expressed on tumor cells.²⁵³

The MEL EpoR gene encodes a 507-amino acid peptide of 62 kDa with extracellular, single-membrane-spanning and intracellular domains.²⁵⁴ The EpoR undergoes glycosylation and phosphorylation before being incorporated into the cell membrane as a 70 to 78 kDa protein.²⁵⁴ Cells transfected with this gene express both high- and low-affinity EpoR.²⁵⁴ The human gene is located on chromosome 19p13.2, encodes a protein of 508 aa, and is 68 to 72 kDa depending on the degree of glycosylation.²⁵⁵

It is now recognized that structurally EpoR is part of a large family of type I cytokine receptors, which includes receptors for IL-2 to IL-7, GM-CSF, and Tpo. Type I cytokine receptors share basic structural features and are characterized by four conserved cysteine residues and a tryptophan-serine-x-serine-tryptophan (WSXSW) motif in the extracellular domain and by conserved box1/box2 regions in the intracytoplasmic domain adjacent to the membrane. Some of these type I cytokine receptors share common subunits and thus are heterodimeric; however, the receptors for Epo, Tpo, and G-CSF consist of homodimers.^256,257 Crystallographic studies confirm that, as is seen with other cytokine receptors, one molecule of Epo simultaneously binds to two EpoR.²⁵⁸ After binding, both Epo and EpoR are rapidly endocytosed and degraded.^{12,216,247,255} Although it is clear that EpoR activation by Epo can lead to formation of EpoR homodimers,²⁵⁹ evidence also indicates these dimers exist prior to Epo binding, and that binding shifts and stabilizes an active receptor conformation bringing the two EpoR into closer contact.^260,261

EpoR signaling pathways are outlined in Figure 6.10. Tyrosine phosphorylation of the EpoR²⁶² is the first observable event after Epo binding. Because EpoR lacks a kinase domain, a tyrosine protein kinase must therefore associate with the receptor. JAK2, a member of the Janus family of cytoplasmic tyrosine kinases, is the primary EpoR-associated kinase and binds to a conserved sequence of amino acids found in cytoplasmic domains of the EpoR^263,264 (and other receptors related in sequence to EpoR). Interestingly, recent studies indicate that the activation of EpoR through binding of Epo initiates a scissorlike rotation of the EpoR dimers, separating the intracellular domains of the two receptor molecules to allow room for the associated Jak2 molecules; in addition there appears to be a self-rotation of each monomer to allow them to orient properly for transphosphorylation of Jak2.²⁶⁵ Deletion of the JAK2 gene in mice results in fetal death on days 12 to 13 that is associated with a severe anemia mirroring the phenotype after deletion of Epo or EpoR,²⁶⁶ indicating the importance of Jak2 for Epo signaling.

To summarize current research,^38,243,244 and ²⁴⁵ one Epo molecule binds two EpoR molecules, activating JAK2 kinases associated with the juxtamembrane regions (box 1/box 2) of each receptor by physically bringing the inactive (or low-activity) kinases into close proximity during the induced rotational shift in EpoR conformation,²⁶⁷ such that these kinases cross-phosphorylate each other, gaining full activity. The activated kinases phosphorylate all eight conserved tyrosine residues of the EpoR cytoplasmic tail. The phosphorylated tyrosine (PY) residues then serve as docking sites for up to 20 different signaling molecules or adaptor proteins that may be phosphorylated by JAK2 to become active, and leading to various mitogenic, differentiative, and antiapoptotic responses. The signaling molecules and adaptor proteins contain either Src homology 2 (SH2) or other phosphotyrosine binding domains that mediate recognition of a PY residue in the context of specific
adjacent (EpoR) amino acids. The major pathways activated by Epo binding to EpoR include the signal transduction activator of transcription 5 (STAT5) pathway, and the phosphatidylinositol-3-kinase (PI3-kinase)/Akt and ras/raf/MAPK (mitogen-activated protein kinase) signaling cascades (Fig. 6.10).

Upon binding to EpoR at PY343 (EpoR numbering here refers to the murine protein), STAT5 is phosphorylated by Jak2,²⁶⁸ dimerizes, and translocates to the nucleus to mediate gene transcription (e.g., of the mitogenic transcription factor c-myc, Id1,¹⁰ TfR1, and IRP-2²⁶⁹). Targeted deletions of the eight potential PY residues in EpoR, to dissect the Epo/EpoR/Jak2 signaling axis, have also revealed the importance of pathways downstream of Jak2.²⁷⁰ Although deletion of all eight residues results (surprisingly) in only mild anemia (the hematocrit in adult mice is reduced by 25% to ˜0.37),²⁷¹ restoration solely of Epo/EpoR/Jak2/STAT5 signaling (i.e., by deletion of all tyrosines except the Y343 residue) results in near-wild-type EpoR activities,²⁷² and in particular, STAT5 signaling restores the erythron capacity for proliferative, stress erythropoiesis responses. The precise role of STAT5 in EpoR signaling in steady-state erythropoiesis, however, is still unclear. STAT5 is expressed from two very similar genes, STAT5a and STAT5b. In animal studies in which both genes were deleted or “knocked out” (STAT5a/b^−/− mice), a marked fetal anemia was reported with lesser but significant anemia in newborn and adult animals.²⁷³ More recent studies indicate these knockout STAT5a/b^−/− animals contain hypomorphic rather than null STAT5a/b^−/− alleles, expressing an N-terminally truncated phosphorylated Stat5 in erythroblasts that appears to retain some biologic STAT5 functions.²⁷⁴ In addition, Epo weakly activates other STATs in primary erythroid cells, such that STAT3²⁶⁹ and/or STAT1 activation²⁷⁴ may also partially compensate for STAT5, resulting in a relatively mild inhibition of murine erythropoiesis.²⁷⁵

A potential target of STAT5 is the Bcl-x gene, which encodes an antiapoptotic protein, Bcl-x_L,, believed to be essential for Epo-dependent survival. Two recent studies of Bcl-x_L-deficient animals confirmed Bcl-x_L expression is required for normal erythropoiesis, but also demonstrate that it promotes the survival of mature erythroid cells that no longer depend on Epo for survival.^276,277 Thus, Bcl-x_L must also be regulated by factors other than Epoactivated STAT5,²⁷⁴ likely the interplay of GATA-1 and Gf-1B at the Bcl-x promoter.²⁷⁸ Epo/EpoR signaling may promote survival earlier, at the proerythroblast stage, through up-regulation of a
number of other antiapoptotic genes including Pim1, Pim3, Trib3, and Serpina 3g.²⁵⁵

The distal end of the EpoR cytoplasmic domain and phosphorylation of Y479 are required for activation of the PI3-kinase/Akt and MAP kinase signaling cascades (Fig. 6.10). Binding of the p85 regulatory subunit of PI3-kinase to EpoR mediates translocation of the kinase from the cytoplasm to phosphorylate lipids at the cell membrane,²⁷⁹ resulting in modulation of survival, metabolism, and translation through the subsequent effects of Akt (a kinase directly downstream of PI3-kinase) on mTOR, GSK3, FOXO, and GATA-1 transcription factors.^280,281 The central role of PI3 kinase in signaling downstream of the receptor for Epo is suggested by the ability of a constitutively active Akt partially to rescue erythroid development when expressed in JAK2^−/− fetal liver cells.²⁸⁰ With regard to the ras/raf/MAPK cascade, Jak2-initiated phosphorylation of tyrosines at the distal end of EpoR allows a Shc adaptor protein to associate with the receptor, which then recruits GRB2 and SOS, resulting in activation of cell membrane-bound Ras and then Raf activation. This signaling cascade gives rise (via MEK) to phosphorylation of ERK1/2 which subsequently phosphorylates up to 60 substrates, promoting erythroid cell-cycle progression and proliferation.²⁸² In contrast to the two pathways described above, however, knockdown of Ras in erythroid cells has only subtle effects on terminal erythropoiesis.²⁸³ Although Epo may act on progenitor cells to promote survival, drive proliferation, and direct erythroid maturation, it is not clear if some intracellular signaling pathways can distinctly activate only one of these events. Epo-dependent activation of the PI3-kinase pathway appears important for both cell survival and proliferation whereas the MAP kinase pathway appears more important in directing proliferation.²⁷⁹

Apart from these three major pathways there are a number of other well-described mediators of responses downstream of Epo/EpoR that likely affect erythropoiesis. These include the inositide-specific phospholipases C (e.g., knockdown of the PLC-γ isoform impedes erythroid development²⁸⁴) and the protein kinase C pathways (e.g., inhibition of the PKCα isoform impairs Epo-induced differentiation, whereas PKCε up-regulation protects erythroid cells from TRAIL-induced apoptosis^285,286). In addition, knockdown of the Src family tyrosine kinase Lyn results in attenuated EpoR signaling and decreased erythroid precursor survival.^270,287,288 The adaptor SH2-containing proteins CrkL (which is phosphorylated by Lyn and indirectly activates Erk1/2²⁸⁹), Lnk (which attenuates JAK2 signaling²⁹⁰), and Spry1, which appears to down-regulate Erk1/2 and Jak2 activation²⁹¹ also appear to be important regulators of EpoR signals. Of interest, recent studies indicate that polymeric IgA (pIgA1-oligomers of IgA joined by their J-chains), produced in small amounts by plasma cells, binds to TfR1 present on the erythroblast cell surface. Binding of pIgA1 or Fe-Tf to TfR1 appears to transmit an intracellular signal that results in activation of erythroblast Akt and ERK1/2, stimulating erythroblast proliferation and differentiation.²⁹² This pathway may be important to boost erythroid output during stress erythropoiesis as hypoxia increases pIgA1 levels, or during iron deficiency when Fe-Tf levels are low. Recent studies identify over 160 genes that are significantly affected by Epo/EpoR signaling in murine primary bone marrow derived CFU-E-like progenitors,²⁵⁵ suggesting much remains to be learned.

As is found with other cell signaling cascades there is a need for checks and balances in the form of inhibitory or regulatory factors, to prevent overstimulation of erythroid cells by Epo/EpoR-mediated growth and survival signals. The distal end of EpoR, for example, acts as a negative regulatory domain to which SH2-containing tyrosine protein phosphatases dock (to PY401, 429, 431) to dephosphorylate substrates such as Jak2 and STAT5 and attenuate Epo signaling. For example, a transgenic animal expressing a truncated human receptor Epo developed severe erythrocytosis, mimicking primary familial and congenital polycythemia (PFCP, see Chapter 44) where patients have elevated red cell mass due to mutations in the EpoR gene.²⁹³ A number of kindreds with PFCP due to mutations in this distal regulatory region of EpoR have since been reported.²⁹⁴ Identified regulatory phosphatases²⁹⁵ include the SH2-containing tyrosine phosphatases SHP1, SHP2 (PTPN11³²), and PTP-1B. Other regulatory factors include the control of EpoR trafficking from the ER by Jak2,²⁹⁶ EpoR internalization upon interaction with Epo,²⁹⁷ proteasomal degradation of EpoR²⁹⁸ and signaling adaptor molecules, and the inducible expression of specific inhibitors such as the suppressor of cytokine signaling protein family members SOCS-1, SOCS-3, and CIS-1 (cytokine inducible SH2-containing protein) by STATs. The SOCS family of proteins down-regulates receptor signaling in a negative feedback manner by (a) competing with STAT5 for binding to EpoR phosphotyrosines and (b) binding and inhibiting the Jak2 kinase activation loop, or (c) ubiquitination and proteosomal targeting of Jak2.²⁹⁹