Increased expression of the distal, but not of the proximal, Gata1 transcripts during differentiation of primary erythroid cells.

Gata1 is expressed from either one of two alternative promoters, the erythroid (proximal to the AUG) and the testis (distal to the AUG) promoter, both used by hemopoietic cells. To clarify the role of the distal and proximal Gata1 transcripts in erythroid differentiation, we determined by specific reverse transcriptase-polymerase chain reactions their relative levels of expression during the differentiation of erythroid precursors purified from the spleen of mice treated with phenylhydrazine (PHZ) or infected with the anemia-inducing strain of the Friend virus (FVA cells). PHZ cells are erythroid precursors that progress in vivo to erythroblasts in 3 days. Both PHZ and FVA cells synchronously proliferate and differentiate in vitro in the presence of erythropoietin (EPO). The levels of total and of distal, but not of proximal, Gata1 transcripts increased by five- to eightfold during in vivo and in vitro differentiation of FVA and PHZ cells. The increase in expression was temporally associated with an increase in the expression of Eklf, Scl, and Nfe2, three genes required for erythroid differentiation, and preceded by 24 h the repression of Gata2 and Myb expression. The day 1 PHZ cells that survived 18 h in the absence of EPO do not express globin genes and express detectable levels of distal but not of proximal Gata1 transcripts. These cells activate the expression of the globin genes within 2 h when exposed to EPO. Therefore, during erythroid differentiation of primary cells, increased expression of distal Gata1 transcripts underlies the increase in the expression of total Gata1 associated with the establishment of the erythroid differentiation program.

Function of caspases in regulating apoptosis caused by erythropoietin deprivation in erythroid progenitors.

Erythropoietin (EP) is required by late stage erythroid progenitor cells to prevent apoptosis. In a previous study (Gregoli and Bondurant, 1997, Blood 90:630-640), it was shown that rapid proteolytic conversion of procaspase 3 to the fully activated enzyme occurred when erythroblasts were deprived of EP for as little as 2 h. In the present study, protein and mRNA analyses of erythroblasts indicated the presence of the proenzyme precursors of caspases 1, 2, 3, 5, 6, 7, 8, and 9. The effects of various caspase inhibitors on caspase 3 processing and on apoptosis were examined. These inhibitors were benzyloxycarbonyl (z-) and fluoromethyl-ketone (FMK) derivatives of peptides that serve as substrates for selected caspases. z-VAD-FMK, t-butoxycarbonyl-aspartate-FMK (Boc-D-FMK), and z-IETD-FMK blocked the initial cleavage of procaspase 3, while z-DEVD-FMK, z-VEID-FMK, and z-VDVAD-FMK did not block the initial cleavage but had some effect on blocking apoptosis. The peptide inhibitor z-FA-FMK, which inhibits cathepsins B and L but is not known to inhibit caspases, altered caspase 3 processing to a final 19 kDa large subunit that appeared to retain enzymatic activity. The action of z-FA-FMK in preventing the usual conversion to a 1 7 kDa subunit suggests the possibility that a noncaspase protease may be involved in caspase 3 processing. Studies with the peptide inhibitors and EP were done to determine the short- and long-term effectiveness of the caspase inhibitors in protecting EP-deprived cells from apoptosis. Although several of the inhibitors were effective, z-IETD-FMK was studied most extensively because of its specificity for enzymes which cleave procaspase 3 at aspartate 175 (IETD175). Large percentages of EP-deprived erythroblasts treated with z-IETD-FMK appeared morphologically normal and negative by a DNA strand breakage (TUNEL) assay at 24 h (75%) compared to EP-deprived controls (10%) which were not treated with inhibitor. However, inhibitor-treated erythroid progenitors deprived of EP for 24 h and then resupplied with EP showed only a modest improvement in long-term survival compared to cells which did not receive the caspase inhibitor during the 24 h EP deprivation. Thus, while the manifestations of apoptosis were delayed in most cells by inhibiting caspase activity, the processes initiating the loss of cell viability due to EP deprivation were irreparablein the majority of the cells and eventually led to their deaths.

The roles of Bcl-X(L) and apopain in the control of erythropoiesis by erythropoietin.

Erythropoietin (EP) is required by late-stage erythroid progenitor cells to prevent apoptosis. Several lines of evidence suggest that it is this action of EP that regulates erythrocyte production in vivo. To study the control of apoptosis in mouse and human erythroblasts, the expression of members of the Bcl-2 family of proteins and the expression and activation of the apoptosis-linked cysteine protease Yama/CPP32/apopain were examined. These proteins have been implicated as regulators of apoptosis in several cell models. The Bcl-2 family members analyzed were Bcl-2, Bcl-X, Bax, Bad, Bak, A1, and Mcl-1. Bcl-X expression in proerythroblasts was highly EP-dependent. Bcl-X was strongly increased during the terminal differentiation stages of human and mouse erythroblasts, reaching maximum transcript and protein levels at the time of maximum hemoglobin synthesis. This increase in Bcl-X expression led to an apparent level of approximately 50 times the level in proerythroblasts. In contrast, neither mouse nor human erythroblasts expressed Bcl-2 transcript or protein. Bax and Bad proteins remained relatively constant throughout differentiation, but diminished near the time of enucleation. Bak protein was present in early erythroblasts, but diminished progressively during differentiation. EP deprivation in both mouse and human erythroblasts led to activation of the cysteine protease, apopain, as was indicated by cleavage of the proenzyme into its proteolytically active fragments. Apopain activation was detectable within 2 hours of EP deprivation in mouse erythroblasts. These findings suggest an important role for Bcl-X in late erythroid differentiation and for apopain in apoptosis of erythroblasts caused by deprivation of EP.

C-myc expression affects proliferation but not terminal differentiation or survival of explanted erythroid progenitor cells.

The expression of c-myc was analyzed in murine and human erythroblasts throughout their differentiation in vitro into reticulocytes. The murine cells were splenic erythroblasts from animals infected with the anemia strain of Friend virus (FVA cells). In FVA cells cultured without EPO, the c-myc mRNA and protein levels decrease sharply within 3 to 4 h, showing that continual EPO stimulation is required to maintain c-myc expression. When cultured with EPO, the c-myc mRNA level of FVA cells is raised within 30 min of exposure. The c-myc mRNA and protein reach maxima at 1 to 3 h, then decline slowly to very low levels by 18 h. In contrast, c-fos and c-jun mRNA levels are not regulated by EPO in FVA cells. The human cells analyzed were colony-forming units-erythroid, CFU-E, derived in vitro by the culture of peripheral blood burst-forming units-erythroid (BFU-E). When grown in EPO and insulin-like growth factor 1 (IGF-1) these cells differentiate into reticulocytes over 6 days rather than the 2 days required for murine cells, but the c-myc mRNA kinetics and response to EPO parallel those of mouse cells at similar stages of differentiation. Both IGF-1 and c-kit ligand (SCF) cause an additive increase in c-myc mRNA in human CFU-E in conjunction with EPO. These additive effects suggest that EPO, IGF-1, and SCF affect c-myc mRNA accumulation by distinct mechanisms. Addition of an antisense oligonucleotide to c-myc in cultures of human CFU-E specifically inhibited cell proliferation but did not affect erythroid cell differentiation or apoptosis. When human cells were grown in high SCF concentrations, an environment which enhances proliferation and retards differentiation, antisense oligonucleotide to c-myc strongly inhibited proliferation, but such inhibition did not induce differentiation. This latter result indicates that differentiation requires signals other than depression of c-Myc and resultant depression of proliferation.

Erythropoietin stimulates phosphorylation of eIF-4E and identification of a 37-kD phosphoprotein that binds mRNA caps in erythroblasts.

To explore the mechanism of erythropoietin action on differentiation of erythroblasts, we have examined its effect on regulating phosphorylation of the 25-kD mRNA cap binding protein (eIF-4E). Erythroblasts from the spleens of mice infected with the anemia strain of Friend virus (FVA cells) were studied. Erythropoietin stimulated phosphorylation of eIF-4E in FVA cells within 30 minutes, and this effect was maximal at 60 minutes. Phosphoamino acid analysis and tryptic phosphopeptide map analysis of eIF-4E isolated from both control and erythropoietin-treated cells identified a predominant phosphopeptide containing phosphoserine. However, when cells were incubated with 1 muM okadaic acid, eIF-4E was phosphorylated on both serine and threonine residues and three additional tryptic phosphopeptides were detected. We also identified a 37-kD phosphoprotein (pp37) that bound specifically to the m7GTP cap structure and coimmunoprecipitated with eIF-kD protein was phosphorylated on both serine and threonine residues. These results indicate that phosphorylation of eIF-4E is a target in erythropoietin-initiated signal transduction events and that this phosphorylation precedes observable effects of erythropoietin on macromolecular biosynthesis. Although of pp37 remains to be studied, it may represent a developmentally regulated mRNA cap binding protein.

Stem cell factor retards differentiation of normal human erythroid progenitor cells while stimulating proliferation.

Stem cell factor (SCF), the ligand for the c-kit tyrosine kinase receptor, markedly stimulates the accumulation of erythroid progenitor cells in vitro. We now report that SCF delays erythroid differentiation among the progeny of individual erythroid progenitors while greatly increasing the proliferation of these progeny. These effects appear to be independent of an effect on maintenance of cell viability. Highly purified day-6 erythroid colony-forming cells (ECFC), consisting mainly of colony-forming units-erythroid (CFU-E), were generated from human peripheral blood burst-forming units-erythroid (BFU-E). Addition of SCF to the ECFC in serum-free liquid culture, together with erythropoietin (EP) and insulin-like growth factor 1 (IGF-1), resulted in a marked increase in DNA synthesis, associated with a delayed peak in cellular benzidine positivity and a delayed incorporation of 59Fe into hemoglobin compared with cultures without SCF. In the presence of SCF, the number of ECFC was greatly expanded during this culture period, and total production of benzidine-positive cells plus hemoglobin synthesis were ultimately increased. To determine the effect of SCF on individual ECFC, single-cell cultures were performed in both semisolid and liquid media. These cultures demonstrated that SCF, in the presence of EP and IGF-1, acted on single cells and their descendants to delay erythroid differentiation while substantially stimulating cellular proliferation, without an enhancement of viability of the initial cells. This was also evident when the effect of SCF was determined using clones of ECFC derived from single BFU-E. Our experiments demonstrate that SCF acts on individual day-6 ECFC to retard erythroid differentiation while simultaneously providing enhanced proliferation by a process apparently independent of an effect on cell viability or programmed cell death.

Erythropoietin stimulates transcription of the TAL1/SCL gene and phosphorylation of its protein products.

Activation of the TAL1 (or SCL) gene, originally identified through its involvement by a recurrent chromosomal translocation, is the most frequent molecular lesion recognized in T-cell acute lymphoblastic leukemia. The protein products of this gene contain the basic-helix-loop-helix motif characteristic of a large family of transcription factors that bind to the canonical DNA sequence CANNTG as protein heterodimers. TAL1 expression by erythroid cells in vivo and in chemical-induced erythroleukemia cell lines in vivo suggested the gene might regulate aspects of erythroid differentiation. Since the terminal events of erythropoiesis are controlled by the glycoprotein hormone erythropoietin (Epo), we investigated whether the expression or activity of the TAL1 gene and its protein products were affected by Epo in splenic erythroblasts from mice infected with an anemia-inducing strain of Friend virus (FVA cells). Epo elicited a rapid, dose-related increase in TAL1 mRNA by increasing transcription of the gene and stabilizing one of its mRNAs. An Epo-inducible TAL1 DNA binding activity was identified in FVA cell nuclear extracts that subsequently decayed despite accumulating mRNA and protein. Induction of DNA binding activity was associated temporally with Epo-induced phosphorylation of nuclear TAL1 protein. These results indicate that Epo acts at both transcriptional and posttranscriptional levels on the TAL1 locus in Friend virus-induced erythroblasts and establish a link between Epo signaling mechanisms and a member of a family of transcription factors involved in the differentiation of diverse cell lineages.

Distinct roles of erythropoietin, insulin-like growth factor I, and stem cell factor in the development of erythroid progenitor cells.

Erythropoietin (EP), insulin-like growth factor I (IGF-I) and stem cell factor (SCF) each reduce apoptosis of human erythroid progenitor cells. To determine if these growth factors have additional roles in stimulating erythropoiesis, the proliferation, maturation, and survival of highly purified human erythroid colony-forming cells (ECFCs) were studied during the application of different combinations of these growth factors in a serum-free liquid culture. EP maintained cell viability and supported heme synthesis during erythroid maturation, with little increase in viable cell number or stimulation of DNA synthesis. The addition of SCF with EP resulted in a substantial increase in DNA synthesis, which was greater than that seen with the addition of EP and was associated with a large expansion in the number of ECFCs. Thus EP, by itself, produces little increase in cell proliferation, and expansion of the number of erythroid cells depends upon the presence of SCF with EP. The addition of IGF-I with EP led to enhanced heme synthesis and moderate cellular proliferation, but also greatly enhanced nuclear condensation and enucleation in the late erythroblasts. Thus EP, by itself, is not sufficient for complete end-terminal nuclear condensation/enucleation and the presence of IGF-I is necessary for this complete process. While EP greatly reduced apoptosis during 16 h of incubation at 37 degrees C, the addition of SCF and IGF-I with EP had little additional effect, but these additions enhanced DNA synthesis > 3.4-fold. Thus SCF may have an additional role in directly stimulating proliferation through a process that is distinct from apoptosis. Our observations indicate that EP prevents apoptosis and maintains erythroid cell viability and development. IGF-I enhances erythroid maturation and proliferation, but the proliferation of erythroid progenitors is mainly controlled by the addition of SCF with EP, independent of an effect on apoptosis.

Apoptosis in erythroid progenitors deprived of erythropoietin occurs during the G1 and S phases of the cell cycle without growth arrest or stabilization of wild-type p53.

Erythropoietin (Epo) inhibits apoptosis in murine proerythroblasts infected with the anemia-inducing strain of Friend virus (FVA cells). We have shown that the apoptotic process in FVA cell populations deprived of Epo is asynchronous as a result of a heterogeneity in Epo dependence among individual cells. Here we investigated whether apoptosis in FVA cells correlated with cell cycle phase or stabilization of p53 tumor suppressor protein. DNA analysis in nonapoptotic FVA cell subpopulations cultured without Epo demonstrated little change in the percentages of cells in G1,S, and G2/M phases over time. Analysis of the apoptotic subpopulation revealed high percentages of cells in G1 and S, with few cells in G2/M at any time. When cells were sorted from G1 and S phases prior to culture without Epo, apoptotic cells appeared at the same rate in both populations, indicating that no prior commitment step had occurred in either G1 or S phase. Steady-state wild-type p53 protein levels were very low in FVA cells compared with control cell lines and did not accumulate in Epo-deprived cultures; however, p53 protein did accumulate when FVA cells were treated with the DNA-damaging agent actinomycin D. These data indicate that erythroblast apoptosis caused by Epo deprivation (i) occurs throughout G1 and S phases and does not require cell cycle arrest, (ii) does not have a commitment event related to cell cycle phase, and (iii) is not associated with conformational changes or stabilization of wild-type p53 protein.

Survival or death of individual proerythroblasts results from differing erythropoietin sensitivities: a mechanism for controlled rates of erythrocyte production.

Murine erythroid progenitors infected with the anemia-inducing strain of Friend virus (FVA cells) undergo apoptosis when deprived of erythropoietin (EPO). When cultured with EPO, they survive and complete terminal differentiation. Although cell volume is decreased and nuclear chromatin is condensed during both apoptosis and terminal differentiation, morphologic and biochemical distinctions between these two processes were observed. In apoptosis, homogeneous nuclear condensation with nuclear envelope loss occurred in cells that had not reached the stage of hemoglobin synthesis. In terminal erythroid differentiation, nuclear condensation with heterochromatin, euchromatin, and nuclear envelope preservation occurred simultaneously with hemoglobin synthesis. Cells with apoptotic morphology appeared asynchronously in EPO-deprived cultures, indicating that only a portion of the cells were undergoing apoptosis at any given time. The percentages of apoptotic cells and cleaved DNA increased with time in EPO-deprived cultures. Inhibition of DNA cleavage was directly proportional to EPO concentration over a wide physiologic range, demonstrating a heterogeneity in susceptibility to apoptosis based on variability in the EPO sensitivity of individual cells. A subpopulation of FVA cells with increased EPO sensitivity (decreased EPO requirement) was isolated from EPO-deprived cultures. This increased EPO sensitivity did not result from differences in EPO receptor number, affinity, or structure, suggesting that the differences are in the signal transduction pathway. These results indicate that control of red blood cell production involves both prevention of apoptosis by EPO and heterogeneity in the EPO requirement of individual progenitor cells.

The use of in situ hybridization to study erythropoietin gene expression in murine kidney and liver.

In situ hybridization has been used to localize erythropoietin (EPO)-producing cells in murine kidney and liver. Peritubular interstitial cells were the only cell type that produced EPO in the kidney. The EPO-producing cells were primarily concentrated in the inner cortex but were also seen in the outer medulla and outer cortex. EPO-producing cells represented less than 10% of the total interstitial cell population. The number of EPO-producing cells per square centimeter of cortex directly correlated with the amount of renal EPO mRNA and varied in an inverse exponential manner with hematocrit. These results suggest that EPO is expressed in an all-or-none fashion in peritubular interstitial cells and that the oxygen carrying capacity of blood is the major regulator of renal EPO production. Peritubular interstitial cells were also identified as the renal source of human EPO in transgenic mice that expressed human EPO mRNA is a regulated fashion in the kidney. Transgenic mice exhibiting inducible supranormal liver expression of human EPO were used to identify EPO-producing cells in the liver. Hepatocytes surrounding central veins produced human EPO in these mice. Individual hepatocytes were able to modulate their production of human EPO depending upon the severity of anemia to which they were subjected. Two types of widely scattered cells produced EPO in severely anemic nontransgenic mice. Eighty percent of EPO-producing cells were hepatocytes and 20% were classified as being nonepithelial based on their nuclear morphology and location in venous sinusoids.

Differentiation and erythropoietin receptor gene expression in human erythroid progenitor cells.

Partially purified human burst-forming unit-erythroid (BFU-E) cells from peripheral blood were cultured for 6 to 8 days to obtain colony-forming unit-erythroid (CFU-E) cells. When these BFU-E-derived CFU-E were further purified and recultured in liquid suspension cultures with erythropoietin (EPO), they matured and differentiated into reticulocytes in vitro. A maximum rate of hemoglobin synthesis was observed at day 10 of cumulative culture time by measuring 59Fe incorporation into heme. Withdrawal of EPO from erythroblast cultures at various times during development showed that between day 10 and day 11 (when the majority of the cells are in the polychromatic erythroblast stage), these cells became independent of EPO. The timing of the disappearance of the EPO requirement in these cells coincided with the marked decline in proliferation. Measurement of EPO receptor messenger RNA (mRNA) levels by Northern analysis showed that there is a slight decline during the day 8 to day 10 time period, followed by a rapid decline between days 10 and 14. Binding of 125I-EPO to erythroblasts also showed a steady decline of the cell surface binding during maturation and terminal differentiation. The half-life of the human EPO receptor was 90 minutes in the presence of the transcriptional inhibitor actinomycin D and the half-life measured at two different times during the 8- to 14-day culture period remained constant. These results indicate that human EPO receptor mRNA must be transcribed continuously to maintain the levels seen by Northern analysis. The human cell system described here is well suited for the study of a wide variety of biochemical events during late erythroid differentiation.

Regulation of programmed death in erythroid progenitor cells by erythropoietin: effects of calcium and of protein and RNA syntheses.

Erythropoietin (EPO) retards DNA breakdown characteristic of programmed cell death (apoptosis) and promotes survival in erythroid progenitor cells. The mechanism by which EPO inhibits programmed death is unknown. In the well-characterized model of glucocorticoid-treated thymocytes, activation of a Ca2+/Mg(2+)-sensitive endonuclease and new protein and RNA syntheses have been found necessary for apoptosis. We examined the effects of EPO on the free intracellular calcium ion concentration ([Ca2+]i), and the roles of Ca2+ and RNA and protein syntheses on DNA cleavage in erythroid progenitor cells. The murine model of erythroid differentiation using Friend leukemia virus-infected proerythroblasts (FVA cells) was used. EPO did not affect the [Ca2+]i in FVA cells. Decreasing [Ca2+]i by extracellular Ca2+ chelation with EGTA facilitated DNA breakdown. Increasing [Ca2+]i with the calcium ionophore 4-bromo-A23187 increased DNA cleavage; however, DNA fragments generated by high [Ca2+]i were much larger than those seen in the absence of EPO or presence of EGTA. Increased [Ca2+]i also inhibited DNA breakdown to small oligonucleosomal fragments characteristic of cells cultured without EPO. However, no concentration of ionophore protected the high molecular weight DNA as did EPO. Cycloheximide inhibited DNA breakdown in a dose dependent manner in cultures lacking EPO, but two other protein synthesis inhibitors, pactamycin and puromycin, did not prevent DNA breakdown. Inhibition of RNA synthesis with actinomycin D did not prevent DNA breakdown. Cells with morphological characteristics similar to those reported in other cells undergoing programmed death accumulated in EPO-derived cultures. These studies demonstrate that although DNA cleavage and morphological changes are common to apoptotic cells, the roles for Ca2+ and protein and RNA syntheses are not universal and suggest that apoptosis can be regulated by different biochemical mechanisms in different cell types.

Abundance and stability of erythropoietin receptor mRNA in mouse erythroid progenitor cells.

The abundance and stability of erythropoietin (EPO) receptor mRNA was measured in Friend virus-infected mouse splenic erythroblast (FVA) cells. FVA cells correspond to the colony-forming unit-erythroid (CFU-E) and cluster-forming unit-erythroid stages of differentiation and they mature in vitro into reticulocytes during 45 to 60 hours of culture. After 20 hours, there was a rapid decline in the EPO receptor mRNA level through the 45-hour culture period. Levels of actin and beta-globin mRNAs were monitored as two representative proteins that are actively synthesized in early versus late stages of terminal erythroid differentiation. A general decrease of actin mRNA level was apparent, whereas globin mRNA levels increased throughout the culture period. The greatest decrease in both EPO receptor and actin mRNA was observed when the cell population was at the stage of basophilic and polychromatophilic erythroblasts. The half-life for EPO receptor mRNA was approximately 75 minutes, as determined using the transcriptional inhibitor actinomycin D. The half-life measured at several times during the 48-hour culture period remained constant. These results indicate that EPO receptor mRNA must be transcribed continuously until late in the maturation process of FVA cells in order to maintain the levels seen by Northern analysis. The number of copies of EPO receptor mRNA at 0 hours was determined to be 25 copies per cell. This low number and the decline of EPO receptor mRNA correlate with the low receptor numbers on FVA cells, as well as the decline of binding sites with cell maturation.

The mechanism of erythropoietin action.

Erythropoietin (Epo) is a glycoprotein hormone produced in the kidney that acts on erythroid progenitor cells in the bone marrow. A negative feedback system, in which tissue oxygenation controls Epo production and Epo controls red blood cell (RBC) production, provides homeostasis in oxygen delivery to body tissues. The target cells for the action of Epo are committed erythroid progenitor cells, which have specific receptors for the hormone. The Epo receptor is a member of a larger family of hematopoietic growth factor receptors. No known second messenger system has been implicated in signal transduction from the Epo receptor. Although Epo may have some effect on mitosis in early erythroid progenitor cells, its control of RBC production appears to occur in later stages of erythroid cell development, where it prevents programmed cell death.

Localization of cells producing erythropoietin in murine liver by in situ hybridization.

In situ hybridization using antisense RNA probes was used to localize cells that produce erythropoietin (EPO) in the livers of anemic transgenic mice expressing the human EPO gene and in livers of anemic nontransgenic mice. In transgenic mice bled from a hematocrit of 55% to one of 10%, hepatocytes surrounding central veins synthesized large amounts of human EPO mRNA. EPO-producing cells were very rare in the area of portal triads. In transgenic mice bled to a hematocrit of 20%, a similar number and distribution of cells contained human EPO mRNA as was found with a 10% hematocrit, but the cells were less heavily labeled, indicating increased EPO production per cell at 10% hematocrit as compared with 20% hematocrit. No human EPO mRNA was detected in the kidneys of anemic transgenic mice, although endogenous murine EPO mRNA was strongly expressed in cortical interstitial cells. In sections of livers from nontransgenic mice bled from a hematocrit of 45% to one of 10%, only isolated cells produced EPO. When the types of cells could clearly be identified, approximately 80% of these cells were hepatocytes, while 20% had a nonepithelial morphology and were located in or adjacent to the sinusoidal spaces. When the sense strand was used as the RNA probe for in situ hybridization, no labeled cells were seen in normal or anemic livers. These results demonstrate that hepatocytes are responsible for production of EPO in both transgenic and nontransgenic mice and that a second cell type that is similar in morphology to EPO-producing interstitial cells in the kidney also produces EPO in the livers of nontransgenic mice.