Transcription Factor Erythroid Krüppel-like Factor (ELKF) Is Essential for the Erythropoietin-induced Hemoglobin Production but Not for Proliferation, Viability, or Morphological Maturation*

The erythroid Krüppel-like factor (EKLF) is essential for the transcription of βmaj globin in erythroid cells. We show here that RNA for this transcription factor did not alter during erythropoietin-induced differentiation of J2E cells; however, EKLF protein content decreased and was inversely related to globin production. This unexpected result was also observed during chemically induced maturation of two murine erythroleukemia cell lines. To explore the role of EKLF in erythroid terminal differentiation, an antisense EKLF construct was introduced into J2E cells. As a consequence EKLF RNA and protein levels fell by approximately 80%, and the cells were unable to manufacture hemoglobin in response to erythropoietin. The failure to produce hemoglobin was due to reduced transcription of not only globin genes but also key heme enzyme genes. However, numerous other genes, including several erythroid transcription factors, were unaffected by the decrease in EKLF. Although hemoglobin synthesis was severely impaired with depleted EKLF levels, morphological maturation in response to erythropoietin continued normally. Moreover, erythropoietin-induced proliferation and viability were unaffected by the decrease in EKLF levels. We conclude that EKLF affects a specific set of genes, which regulates hemoglobin production and has no obvious effect on morphological changes, cell division, or viability in response to erythropoietin.

The erythroid Krü ppel-like factor (EKLF) 1 gene produces a transcriptional regulator that was originally identified by its expression in immature red blood cells (1). EKLF contains three TFIIIA-like zinc fingers that share homology with the Krü ppel family of Drosophila transcription factors (1). It is expressed almost exclusively in erythroid precursors, although low level expression is detectable in mast cells. The anatomical sites of EKLF expression in vivo support the concept of it being an erythroid-specific transcription factor (2). Although EKLF binds monomerically, with high affinity to the DNA sequence CCACACCCT present in CACC boxes (1,3), a number of other transcription factors also bind to this motif, e.g. SP1 family members and TEF-2/BKLF (4,5).
CACC motifs are found in the promoters of a number of erythroid-specific genes including globins, GATA-1, erythropoietin (epo) receptor, porphobilinogen deaminase (PBG-D), carbonic anhydrase I, glycophorin B and the erythroid isoform of pyruvate kinase (6 -10). Significantly, mutations in the CACC box of human ␤-globin promoter have been identified in ␤-thalassemias (3,11). The CACC elements in erythroid-specific genes are generally situated in close proximity to motifs for GATA-1 (6,7,(12)(13)(14)(15). Recently it has been shown that EKLF contains domains that enable it to interact functionally with other proteins, including GATA-1 (16 -18). The EKLF promoter contains three GATA binding sites, one of which is crucial for expression, and it can be directly activated in nonerythroid cells overexpressing GATA-1 (19). Although the GATA-1 promoter contains a CACC element required for full promoter activity (9), it has been suggested that EKLF lies downstream or coincident with GATA-1 in the hierarchy of transcription factors (19). Interestingly, proerythroblasts lacking GATA-1 have EKLF transcripts reduced by 40 -80% (20).
To illustrate the functional role of EKLF in vivo, targeted disruption of the gene was achieved in embryonic stem cells (21,22). Homozygous null mice died from a severe anemia early in fetal life when the site of erythropoiesis switched from the yolk sac to the fetal liver. This anemia was attributed to a lack of ␤-globin transcription at the developmental stage when expression of embryonic ␥-globin is normally silenced and ␤-globin activated (21,22). Other studies have supported the concept that EKLF participates in ␥-globin to ␤-globin switching (23)(24)(25). The conclusion from the knockout studies was that EKLF was necessary for hemoglobin synthesis in definitive erythroid cells but was dispensable for primitive erythropoiesis and commitment to the erythroid lineage. The lack of EKLF appeared to be specific for ␤ maj -globin as other erythroid genes containing CACC boxes (GATA-1, epo receptor, PBG-D) appeared to be expressed normally (21,22). In vitro biochemical analyses have corroborated the crucial transactivation function of EKLF in ␤ maj -globin expression (26).
As far as we are aware, the importance of EKLF to epoinduced terminal differentiation has not been examined directly. Therefore, we have used the v-raf/v-myc-transformed J2E erythroleukemic cell line (27) as a model to investigate the role of EKLF in epo-stimulated differentiation. In response to epo, J2E cells undergo typical erythroid maturation characterized by activation of erythroid-specific genes, accumulation of functional hemoglobin, enhanced proliferation, maintenance of viability and morphological alteration (27)(28)(29)(30)(31). Thus, the importance of EKLF to each facet of erythroid differentiation could be assessed in this system.
In this manuscript we show that as J2E cells differentiated, EKLF protein fell whereas globins were synthesized. However, reduction of EKLF levels before exposure to epo via an antisense EKLF construct blocked hemoglobin synthesis due to reduced transcription of globin and some heme enzyme genes. Lowering EKLF levels had no effect on transcription of numerous other genes and had no impact on epo-induced cell division, viability, or morphological maturation. We conclude that a threshold concentration of EKLF is essential for epo-stimulated transcription of certain genes associated with hemoglobin synthesis and that this transcription factor plays no significant role in other aspects of erythroid terminal differentiation.
Construction of Antisense EKLF Vector-An EcoRI fragment comprising nucleotides 270 -1345 of the murine EKLF coding region (1) was subcloned into the multiple cloning site of Bluescript KSϩ (Stratagene, La Jolla, CA). A 945-base pair BamHI-HindIII fragment was excised and ligated in the retroviral vector pRuf(TK) neo (37) in the antisense orientation. DNA for this construct was then electroporated into the PA317 amphotropic-packaging line (38). After Geneticin (Sigma) selection, the virus-producing fibroblasts were expanded to approximately 70% confluence. The media was changed, and the next day the viruscontaining supernatants were collected and passed through a 0.22-m filter (Millipore, Bedford, MA). J2E cells were then cultured with the virus-containing media before selection in Geneticin. Cells that emerged were then cloned in methylcellulose (Methocel, Fluka Biochemika, Switzerland), and single colonies were picked for expansion and further characterization. Southern blotting was used to identify individual clones with unique retroviral integration sites.
RNA Analyses-Total cellular RNA was isolated from cells by the method of Chomczynski and Sacchi (42). Poly(A) ϩ RNA was extracted and separated electrophoretically on 1% agarose gels with formaldehyde. After transfer to nylon membranes (Hybond Nϩ, Amersham), Electrophoretic Mobility Shift Assays-Nuclear extracts were prepared according to the method of Ramsay et al. (51), and DNA binding reactions were conducted as we have described previously (32). Briefly, 5 g of nuclear extract was incubated with a 32 P end-labeled oligonucleotide (5Ј-GCTAGCCACACCCTGAAGCT-3Ј) corresponding to the EKLF motif in the murine ␤ maj -globin promoter (1). The EKLF complex was identified with anti-EKLF antibodies as described by Crossley et al. (4).

EKLF Protein Levels Decrease during Erythroid Differentiation-
The expression of EKLF was examined in J2E cells induced to differentiate with epo. Nuclear run-on assays revealed that transcription of the EKLF gene did not change in maturing J2E cells (Fig. 1A). Therefore, transcription of the EKLF gene itself was not regulated by epo. In contrast, the transcription of ␤ maj -globin, a known target of EKLF (21,22,26), increased 6-fold during the 48-h period of observation. Similar data were obtained when steady state mRNA levels were determined by Northern blotting (Fig. 1B). Although EKLF transcript levels remained constant during J2E cell differentiation, a decline in EKLF protein was detected; in contrast, globin protein content rose significantly (Fig. 1C). It was concluded that EKLF protein levels did not correlate with globin production during the maturation of J2E cells and that translation or protein turnover may be important control points in EKLF expression.
EKLF expression was also examined in the F4N MEL cell line induced to differentiate with Me 2 SO. No changes in EKLF RNA were observed in differentiating MEL cells ( Fig. 2A) in agreement with previous reports (1); however, EKLF protein content fell dramatically, whereas globin proteins rose appreciably (Fig. 2B). Similar data were obtained with the 707 MEL cell line (data not shown). Thus, EKLF protein levels were inversely related to the appearance of globin proteins during both epo-stimulated and chemically induced erythroid terminal differentiation.
Antisense EKLF Suppresses Transcription of Several Erythroid Genes-The inverse relationship between EKLF and globins during erythroid terminal differentiation (Figs. 1, 2) appeared to contradict the need for EKLF in globin synthesis. To examine the role of EKLF during epo-initiated terminal differentiation, an antisense strategy was adopted. To this end, an EKLF cDNA between 270 and 1215 base pairs (1) was inserted in the reverse orientation into the retroviral vector pRuf-neo. Three of 12 independent clones (JpREK 1, 6, and 12) containing the antisense EKLF construct were chosen for analysis. Cells infected with the pRuf-neo vector alone (JRN) were used as controls. No differences in growth characteristics were observed between the JpREK and JRN clones.
Northern blotting of poly(A) ϩ RNA was used initially to determine the effect of the antisense EKLF construct on erythroid gene expression. Fig. 3 shows that the antisense vector in the JpREK cells had a marked effect on the endogenous EKLF gene expression as RNA levels were reduced by 80 -90%. Importantly, the amount of EKLF protein in the cells was approximately 20% of the parental cells (Fig. 4A). In addition, EKLF binding activity decreased substantially (Fig. 4C). These results demonstrate that the antisense construct effectively reduced EKLF mRNA and active protein in J2E cells. The effect of the antisense EKLF construct on a number of transcription factors known to be important for erythroid development was then examined. Little significant change was noted for SCL, BKLF, and p45 NF-E2 transcripts (Fig. 3) or protein (Fig. 4). However, GATA-1 levels were 50 -80% lower.
Analysis of the expression of several genes associated with hemoglobin synthesis revealed that the RNA levels for globin genes were reduced. Both ␣and ␤ maj -globin transcripts fell, but ␤ maj -globin was down-regulated to a greater extent (Fig. 5). In addition, some enzymes of the heme biosynthetic pathway had diminished mRNA levels. Transcripts for ALA-S(E) and FC were noticeably reduced; however, less significant decreases were seen with 5-aminolevulinic acid dehydratase and PBG-D. Subsequent nuclear run-on assays demonstrated that of 40 genes examined, only coproporphyrinogen oxidase (a heme enzyme gene) was appreciably down-regulated together with the globins, ALA-S(E), and FC (data not shown). Moreover, the transcriptional increase of globin and heme enzyme genes normally seen in response to epo ((35) Fig. 1) did not occur in the JpREK antisense clones (data not shown). It was concluded that below a threshold level of EKLF, transcription of a specific set of genes was severely impaired.
Reduced EKLF Expression Inhibits epo-induced Hemoglobin Protein was extracted from J2E cells and three antisense EKLF-containing clones (JpREK-1, 6, 12) and immunoblotted for the presence of EKLF protein. v-raf was used as the loading control in the two separate immunoblots (Panels A and B). GATA-1, BKLF, and p45 NF-E2 were other transcription factors analyzed in these cells. Panel C shows an electrophoretic mobility shift assay with reduced EKLF binding activity in the JpREK clones. Synthesis but Not Proliferation or Viability-Epo stimulates J2E cells to synthesize hemoglobin, undergo enhanced proliferation, mature morphologically, and remain viable in the absence of serum (27)(28)(29)(30)(31)52). The effect of the antisense EKLF construct on each of these responses to epo was examined. Initially, hemoglobin synthesis was monitored by benzidine staining, and Fig. 6 shows that each clone containing antisense EKLF failed to increase hemoglobin production during a 72-h incubation period with epo. In contrast, the vector-alone JRN clones manufactured hemoglobin just like parental J2E cells (27,28). Spectral scans confirmed that oxyhemoglobin did not increase with epo stimulation in the JpREK cells (data not shown). Fluorometric analyses and immunoblotting revealed that the decrease in hemoglobin in these cells was caused by a 4 -12-fold reduction in both heme and globin content. Significantly, the reduction in EKLF also restricted hemoglobin production initiated by sodium butyrate, a chemical inducer of J2E cell differentiation (data not shown).
The effect of antisense EKLF on epo-stimulated cell division and viability was investigated next. Cell numbers were determined daily after the addition of epo to the cultures, and unlike hemoglobin production, no difference was observed between cells containing the vector alone and the antisense EKLF clones (Table I). [ 3 H]Thymidine assays revealed that the cells expressing antisense EKLF were still able to undergo enhanced DNA synthesis in the presence of epo. Moreover, reducing EKLF levels had no impact on the ability of epo to support cell survival. These data demonstrate that the antisense EKLF construct inhibited only one aspect of epo-induced maturation in J2E cells i.e. hemoglobin production; it did not influence enhanced cell division or maintenance of viability.
Antisense EKLF Does Not Affect epo-induced Morphological Maturation-Having observed a significant effect of antisense EKLF on hemoglobin production (Figs. 6, Table I), epo-induced changes to cell morphology were then analyzed. Unexpectedly, no differences in morphological maturation were observed with the clones containing the antisense EKLF (data not shown). The proportion of cells at each morphologically identifiable stage of erythroid terminal differentiation was comparable with our previous reports (29).
In an attempt to separate hemoglobin synthesis from morphological changes, cells were stained with DAF to detect hemoglobin, whereas morphological alterations were identified microscopically. DAF was used instead of benzidine as it does not precipitate with serum (54) and is a more sensitive stain for detecting the presence of hemoglobin in J2E cells (34). The data presented in Fig. 7 demonstrate that after epo stimulation, JRN control cells produced hemoglobin and simultaneously underwent morphological changes associated with normal erythroid maturation. In contrast, the JpREK clones did not produce sufficient hemoglobin to react with DAF but did undergo characteristic alterations in shape and size. Thus, morphological maturation of J2E erythroid cells did not require high levels of EKLF and could be separated from the hemoglobin synthesis.
Antisense EKLF Does Not Affect epo Signaling-One possible explanation for the inability of JpREK clones to produce hemoglobin was due to defective epo signaling, induced somehow by the antisense EKLF construct. To address this issue, immunoprecipitation and immunoblotting experiments were conducted on molecules known to be associated with the epo signaling cascade. Fig. 8 shows that the epo receptor was phosphorylated with a normal time course in JpREK cells. Although the level of receptor phosphorylation appeared less than in parental J2E cells, we have demonstrated previously that activation of 10% of the epo receptors is sufficient to generate a complete differentiation response (55). Significantly, stimulation of STAT5, MAP kinase, and general protein phosphorylation was comparable with JRN control cells. It was concluded from these observations that antisense EKLF did not interfere with epoinitiated signaling events. DISCUSSION In this manuscript we have shown that below a threshold concentration of EKLF, epo is unable to stimulate hemoglobin synthesis. Although EKLF protein decreased during erythroid terminal differentiation, down-regulation of EKLF by the antisense construct prior to induction by epo severely inhibited transcription of globin and heme enzyme genes. Strikingly, EKLF was not needed for other aspects of epo-stimulated differentiation. The reduction in EKLF levels had no effect on morphological maturation, elevated cell division, or maintenance of viability. Thus, this erythroid-restricted transcription plays a unique and highly specialized role during erythroid terminal differentiation, viz. production of the oxygen carrier.
We have demonstrated here that EKLF is needed for hemoglobin production and not for morphological maturation. This observation clearly shows that the processes of hemoglobin synthesis and cytological changes are under separate control. Epo transmits a signal that initiates both processes, but they must bifurcate at some point. EKLF is involved in hemoglobin production alone, and other factors must regulate the change in cellular morphology. Recently, different aspects of the differentiated macrophage phenotype could be linked with specific regions of the common ␤ chain of the granulocyte macrophagecolony stimulating factor/interleukin 3/interleukin 5 receptors (56). The observations in vitro using the J2E cell line were comparable with the data obtained from EKLF Ϫ/Ϫ mice (21,22). Although the number of erythroid precursors was normal in the fetal livers of these animals, the enucleate erythrocytes formed lacked hemoglobin due to a deficiency in ␤-globin pro-tein (21,22). In addition, EKLF Ϫ/Ϫ erythrocytes are shortlived due to the imbalance between globin chains, leading to precipitation of excess ␣-globin and Heinz body formation (57).
The data presented here showed that in addition to ␤ majglobin, the heme enzyme genes ALA-S(E) and FC appear to be important targets for EKLF. There were substantial decreases in transcription and mRNA levels for these heme enzyme genes in the JpREK clones. It is noteworthy that ALA-S(E) and FC are the first and last enzymes in the heme biosynthetic pathway, respectively, and important sites of regulation (58). Significantly, both genes contain CACC motifs in their promoter regions (47,59), and Surinya et al. (60) have recently shown that EKLF can transactivate the ALA-S(E) promoter.
As physical and functional interactions are being defined for EKLF, GATA-1, Rbtn2, SCL, and others as well as NF-E2 and maf (17,18,(61)(62)(63)(64)(65)(66), it is possible that these molecules form a matrix that is vital for transcription of genes in response to epo. Perhaps transcription is not initiated by epo if the concentration of any component is suboptimal. Significantly, Kulessa et al. (67) have demonstrated a concentration-dependent effect of GATA-1 on lineage determination and McDevitt et al. (68) have recently shown that erythroid differentiation is dose-dependent with respect to GATA-1. Other modifications such as phosphorylation of DNA-binding proteins after epo stimulation may also affect their activity.
Transcription of EKLF itself was unaffected by epo stimulation, and RNA levels remained constant during J2E cell differentiation. Similarly, it has been demonstrated that EKLF transcripts did not change during MEL cell differentiation ((1) Fig.  2). In contrast with the constant amount of RNA, EKLF protein content declined in both J2E and MEL cells. These changes to EKLF were similar to the decrease in SCL protein late in erythroid differentiation, despite an accumulation of mRNA (69). It appears that reduced translation and/or increased protein degradation significantly influenced the production of both DNA binding proteins during the final stages of erythropoiesis. Interestingly, GATA-1 levels also fall late in the erythroid maturation process (32,70,71). It is conceivable, therefore, that these erythroid transcription factors are needed for the activation of transcription by epo, but once established, RNA synthesis of target genes can be maintained with lower levels of these proteins. Significantly, Ford et al. (72) observed that temporal changes of C/EBP isoforms at the myeloperoxidase enhancer mediated the transition from a primed state in multipotential cells to a transcriptionally active configuration in promyelocytes. They suggested that "differentiation-associated changes in factor occupancy at critical regulatory elements may prove to be a common theme in the activation of lineage specific loci." The effect of suppressing EKLF levels in vitro has parallels with the inhibition of other transcription factors in erythroid cell lines. Antisense SCL constructs have been shown to inhibit a Determined by absorbance at 414 nm (35). b Measured by eosin exclusion (27). c Ascertained by [ 3 H]thymidine incorporation (28). d Evaluated by eosin exclusion of cells deprived of serum (30,31,52).

FIG. 7. Morphological maturation of JpREK.
Vector alone JRN control cells and antisense EKLF JpREK-6 cells were induced to differentiate with epo. The cells were stained with DAF to detect hemoglobin up to 72 h post-induction and viewed microscopically. Note the dark staining in JRN cells, indicating hemoglobin production as they matured morphologically (Panels A-F). In contrast, cellular changes were noted in JpREK cells without hemoglobin production. Me 2 SO-induced MEL cell differentiation and influence the maturation of K562 cells (73,74). Similarly, the loss of NF-E2 from MEL cells prevents hemoglobin production in response to chemical stimulation, whereas expression of exogenous NF-E2 restores the ability to manufacture hemoglobin (75,76). However, it is not known whether other features of erythroid terminal differentiation, such as viability, proliferation, and morphological changes are also affected by SCL and NF-E2.
Epo has a number of effects on immature erythroid cells, viz. proliferation, viability, hemoglobin production, and morphological changes. Recently, we have been able to show that the tyrosine kinase Lyn is essential for epo-induced differentiation but is not required for supporting cell viability (52). By suppressing EKLF, we have separated hemoglobin production from morphological maturation and other aspects of erythroid terminal differentiation. It is hoped that epo-activated intracellular signaling and regulatory pathways can be dissected further using this model system.