Erythroid gene suppression by NF-κB

NF-κB/Rel transcription factors play essential roles to mediate the immune response and apoptosis, and they have also been implicated in cellular differentiation such as erythropoiesis. To elucidate the possible role(s) of NF-κB in erythroid gene regulation and erythropoiesis, we have carried out transient transfection studies of the human embryonic/fetal erythroid cell line K562 and mouse adult erythroid MEL cells. It is shown that tumor necrosis factor-α represses the transcription activity directed by either α or ζ globin promoter in a dose-dependent manner. Furthermore, different NF-κB family members could effectively repress the transfected α-like globin promoters in K562 as well as in MEL cells. The involvement of NF-κB pathway is supported by the ability of a NF-κB-specific, dominant negative mutant to block the tumor necrosis factor-α or p65-mediated suppression of the α-like globin promoter activities. The suppression appears to be mediated through cis-linked HS-40 enhancer. Finally, stably transfected K562 cells overexpressing p65 contain reduced amounts of the p45/NF-E2 RNA and functional NF-E2 proteins. Our studies have identified a new set of targets of NF-κB. We suggest that the relatively high activity of the NF-κB pathway in early erythroid progenitors is involved in the suppression of erythroid-specific genes. Later in differentiation, together with other changes, the decline of the amounts of the NF-κB family of factors leads to derepression and consequent increase of NF-E2, which in turn would activate a subset of erythroid-specific genes.

The mammalian ␣-like (embryonic , adult ␣2, ␣1, and ) and ␤-like (embryonic ⑀, fetal ␥G and ␥A, adult ␦ and ␤) globin gene clusters together have provided a paradigm for the analysis of coordinated and differential gene expression. Understanding the molecular basis of human globin gene switch during development would also provide essential knowledge and alternative therapeutic strategies for the treatment of severe hemoglobinopathies including sickle cell anemia and thalassemias (1,2).
It has become clear that the developmental switch on and off of different globin gene transcription is controlled by the synergistic interactions between different globin promoters and their upstream regulatory elements (3)(4)(5)(6)(7). These elements, namely, the locus control region of the ␤-like globin gene cluster (4,6,7) and the HS-40 enhancer of the ␣-like globin gene cluster (5), and the promoters each consists of specific sequence motifs that are bound with nuclear factors in an erythroid celland developmental stage-specific manner (Refs. 8 and 9 and references therein). Among the DNA-binding proteins known to function in globin gene regulation are the erythroid-enriched transcription factors NF-E2 (10), GATA-1 (11), and EKLF (12). There are also cofactors that physically interact with the above DNA-binding transcription factors, for example, FOG (13) and chromatin remodeling complexes (14).
When compared with other gene systems, relatively little is known about the signal transduction pathways regulating the globin gene transcription. One study in human primary erythroid progenitors and the human embryonic/fetal erythroid cell line K562 showed that IL-6 1 down-regulated the ␥ globin mRNA level. However, the molecular basis of this repression remains unknown (15). Another transgenic mice study has shown that a NF-B-binding motif at the 3Ј side of the human globin gene is required for its complete silencing in the adult mice (16). Finally, it has been implicated that serine-threonine phosphorylation might regulate the activity of NF-E2 (17,18).
NF-B/Rel is a family of potent inducible regulatory factors consisting of p65/RelA, p50(p105), p52(p100), RelB, and c-Rel (Refs. 19 and 20 and references therein). The NF-B pathway could be activated by a variety of extracellular stimuli to regulate the expression of genes required for immune and inflammatory responses, for cell growth and differentiation, as well as for suppressing apoptosis (21)(22)(23). In general, upon stimuli such as exposure to the various cytokines, the IB inhibitory proteins are phosphorylated and degraded. The liberated NF-B is then translocated to the nucleus, binds to specific B elements (GGGRNNYYCC), and modulates by gene transcription. More recently, several lines of evidence have suggested that NF-B also plays a crucial role in erythropoiesis (24). RelB knockout mice develop normally but exhibit abnormalities in hematopoiesis and multiorgan inflammation (25). Rel/RelAdeficient mice displayed multiple hematopoietic cell defects and erythropoiesis impairment (26). Also, NF-B subunits p65, p50, and p52 are all expressed during early normal erythroid proliferation (day 7-10 erythroblasts), and their levels decline during differentiation (24). It was also postulated (24) that NF-B factors modulate the erythropoiesis via its downstream target genes, c-myb and c-myc, the expression of which are required for the erythroid development (27)(28)(29). We show below that activation of the NF-B pathway in two different types of erythroid cells represses the expression of the ␣-like globin genes. We have also carried out experiments to investigate the molecular basis of this NF-B-mediated repression.
Cell Lines and DNA Transfection-Human erythroid K562 (33) and mouse erythroid MEL (34) cells were cultured under 5% CO 2 at 37°C in RPMI 1640 medium (Invitrogen) supplemented with 10% fetal bovine serum, 2 mM L-glutamine, 50 units/ml of penicillin, and 50 g/ml of streptomycin (Invitrogen). For transient DNA transfection, the seeding density of K562 cells was 2.5-3.0 ϫ 10 5 /ml at 5 h prior to the transfection. For MEL cells, it was 5 ϫ 10 5 /ml at 1 day before transfection. A total of 2-3 ϫ 10 6 (K562) or 5 ϫ 10 6 (MEL) cells were collected by centrifugation and then resuspension in 0.4 ml of RPMI 1640 medium containing the indicated number of micrograms of the expression constructs plus 1 or 2 g of the luciferase or GH reporter plasmids. The total amount of transfected DNA was kept constant with the addition of the cloning vectors. Electroporation was carried out with a Bio-Rad Gene Pulser at a capacity of 975 microfarads and 250 mV (K562) or 300 -350 mV (MEL). The transfected cells were kept in 6-well culture plates for 24 h for the luciferase assay and 48 h for the growth hormone assay. Unless indicated otherwise in the figure legends, the transfection data represent the averages of at least three independent experiments. For the cytokine induction, 50 ng/ml of recombinant human TNF-␣ and IL-6 (R&D Systems) were added to the culture medium at 5-6 h post-transfection. The cells were then incubated at 37°C for 24 h before the assays. For hemin induction, 25 M of hemin (Sigma) were added to the culture medium at a cell density of 2-4 ϫ 10 5 /ml, and the cells were incubated at 37°C for 2 days before the assays.
For the establishment of stable K562 pools overexpressing p65, K562 cells were transfected with pEGFP-HAp65. The clones were selected in the presence of 800 g/ml of active G418 for 2 weeks. GFP-positive cells were then sorted out by flow cytometry as follows. The cells were collected by centrifugation at 1,000 rpm for 5 min, washed by resuspension in 1 ml of phosphate-buffered saline, and recentrifugated. The washed cells were resuspended in phosphate-buffered saline and 5% fetal bovine serum and then sorted with a FACStar-plus instrument (Becton Dickinson). Propidium iodide (0.1 g/ml) was added to restrict the sorting to living cells. The control cell pool, K(V), was generated in the same manner as above, except that the vector pEGFP-C1 was transfected and stably integrated. The sorted cells were maintained in the presence of 400 g/ml of active G418 for use in further assays. Two pools, K(p65) and K(p65)Ј, independently selected and sorted in this way were analyzed for the expression levels of their ␣, , p45/NF-E2, and control genes. Of the two pools with p65 overexpression, K(p65) was further characterized in more detail by EMSA and DNA transfection.
GH Assay-Forty-eight hours after DNA transfection, the levels of human GH secreted into the media by the transfected cells were quantified with the Allegro human GH radioimmunoassay kit as described previously. In cotransfection assays of Figs. 2-4, the GH values were normalized with the protein contents because coexpressed p65 or IB would affect the CMV promoter activity of the commonly used internal plasmid pCMV-CAT or pRSV-CAT.
Luciferase Assay-The cells were harvested at 24 h post-transfection.
They were isolated by centrifugation at 1000 rpm for 5 min at 4°C and washed twice by resuspension in 1 ml of ice-cold phosphate-buffered saline and recentrifugation. The extracts were prepared by first lysis of the cells in appropriate amounts of a reporter lysis buffer (100 mM potassium phosphate, pH 7.8, 1 mM EDTA, 10% glycerol, 1% Triton X-100, 7 mM ␤-mercaptoethanol) for 5 min at room temperature. The supernatants were collected by centrifugation at 14,000 rpm for 5 min.
The luciferase measurements were done on a luminometer (LKB1251). The values of the luciferase activities were also normalized to the protein concentrations, which were determined by the Bradford assay using bovine serum albumin as the standard (Bio-Rad). Relative luciferase activities were indicated as the average values with the standard deviation.
Western and Northern Blot Analyses-K562, K(V), K(p65), and K(p65)Ј cells at the density of 1-2 ϫ 10 6 cells/ml were first arrested in 0.5% serum RPMI 1640 for 24 -36 h. After 12-24 h of serum stimulation, the cells were harvested for Northern blot analysis of the levels of globin RNAs. For Northern and Western blot analysis of the p45/NF-EL expression, serum stimulation of the cells was for 2-3 and 6 -12 h, respectively. The blot hybridization procedures essentially followed Sambrook et al. (35). For Western blot analysis, the extracts were prepared by cell lysis in 25 mM HEPES, pH 7.5, 0.3 M NaCl, 1.5 mM MgCl 2 , 0.2 mM EDTA, 0.5% Triton X-100, 3 mM dithiothreitol, 30 mM ␤-glycerophosphate, 50 mM NaF, 1 mM Na 3 VO 4 , 10 g/ml each of aprotinin and leupeptin, and 0.5 mM phenylmethylsulfonyl fluoride. 50 g of the total protein were separated with SDS-PAGE, transferred to nitrocellulose (Millipore), and hybridized with antisera against HA (Roche Applied Science), p45, p65, and ␣-tubulin (all from Santa Cruz Biotechnology), respectively. The immunocomplexes were detected by the reaction with anti-rabbit or anti-mouse IgG conjugated to horseradish peroxidase (Amersham Biosciences), followed by ECL detection (Amersham Biosciences) according to the manufacturer's instructions. For Northern blot analysis, total RNAs were extracted by means of a commercial Trizole reagent (Invitrogen). 10 g of each RNA sample were separated on a 1.0% agarose, 6% formaldehyde gel, transferred to a nylon membrane, and hybridized with DNA fragments amplified by PCR and containing ␣ globin, globin, and ␤-actin cDNA, respectively. A 1.0-kb HindIII fragment containing human p45 cDNA was also used as the probe. The probes were labeled with [ 32 P]dCTP by random priming. After overnight hybridization at 42°C with standard buffer containing 50% formamide, the blot was washed successfully in 2ϫ SSC, 0.1% SDS once and 0.2ϫ SSC, 0.1% SDS at 55°C. The blots were subjected to autoradiography or direct quantitation with a PhosphorImager.
EMSA-The nuclear extracts were prepared essentially as described by Dignam et al. (36). All of the DNA binding reactions were performed using 10 g of nuclear extracts in a volume of 20 l of reaction mixture. The binding reaction for the IgB site was carried out in a buffer containing 20 mM HEPES, pH 7.9, 60 mM KCl, 1 g of poly(dI-dC), 10% glycerol, 0.2 mM EDTA, 0.5 mM phenylmethylsulfonyl fluoride, and 0.5 mM dithiothreitol. 3.2 mM MgCl 2 was included in the reaction mixtures for NF-E2 or NF-Y binding. For Sp1 binding, the reaction mixture contained 10 mM HEPES, pH 7.9, 50 mM KCl, 0.1 mM ZnCl 2 , 10% glycerol, 0.5 mM phenylmethylsulfonyl fluoride, and 0.5 mM dithiothreitol.

Activation of the NF-B Pathway Represses Human ␣ and
Globin Promoters-Transient transfection was used to test whether the NF-B pathway might be involved in the regulation of the ␣-like globin gene expression. The human erythroleukemia cell line K562 was transfected with the reporter plasmids pHS40-␣590Luc or pHS40-597Luc. At 5-6 h posttransfection, the cells were treated with TNF-␣ or IL-6, respectively, for 24 h prior to the reporter assay. As shown in Fig. 1, treatment with TNF-␣ but not IL-6 reduced the luciferase activity of either (Fig. 1A) or ␣ (Fig. 1B) globin promoter. To further examine whether TNF-␣ exerted this effect through the NF-B pathway, the cells were cotransfected with pCMV-IB␣(S32A/S36A). This plasmid, termed pCMV-IB␣A 32 A 36 in the following, expresses IB␣(S32A/S36A), also known as IB␣AA or IB, which inhibits TNF-␣-or IL-1-induced NF-B activation (37). Indeed, TNF-␣-mediated repression of the human and ␣ globin promoters was relieved by cotransfection of pCMV-IB␣A 32 A 36 (Fig. 1). These data together suggested that activation of the NF-B pathway by TNF-␣ could suppress the transcriptional activities of transfected human ␣-like globin promoters in K562 cells.
The functional roles of NF-B subunits in suppressing the human ␣-like globin promoters were further studied by the cotransfection assay. Consistent with Fig. 1, a dose-dependent suppressive effect by p65 in K562 cells was observed when either growth hormone ( Fig. 2A) or luciferase (Fig. 2B) was used as the reporter. In particular, cotransfection with 10 g of pCMV-p65 decreased the and ␣ promoter activities by 90 and 80%, respectively ( Fig. 2A, middle bars). Although IB␣AA by itself had no effect, coexpression of this subunit reversed the p65-mediated suppression of either globin promoter ( Fig. 2A,  top two bars). These data indicate that p65 could down-regulate the transcriptional activities of the human and ␣ globin promoters through the NF-B pathway. Cotransfection of MEL cells with pHS40-␣590 GH and pCMV-65 gave similar results (data not shown).
Members of the NF-B Family Other Than p65 Are Also Capable of Repressing the ␣-Like Promoters-Similar to p65, several other NF-B members including p50, p65-p50, and c-Rel significantly repressed the globin promoter in transfected K562 cells (Fig. 2B, left panel). Coexpression of p65-p50 also repressed transfected human ␣ globin promoter in MEL cells (Fig. 2B, right panel). In addition, the repression by these factors correlated well with the expression of NF-B activity in the transfected cells, as measured by a NF-B-dependent reporter assay (data not shown). Furthermore, coexpression of a dominant negative mutant of p50, p50⌬Sp, had no suppression effect on the transfected globin promoters (Fig. 2B, top bars of both panels). The data of Fig. 2B indicated that the HS-40mediated activities of the human ␣-like globin promoters could be repressed by activation of the NF-B pathway through the functioning of different members of the NF-B family.

Target Sites of Suppression of the ␣-Like Globin Promoters by NF-B-To determine the sites of action by the NF-B factors
to suppress the ␣-like globin promoters, we compared the promoter activities in transfected K562 cells in the presence or absence of the cis-linked HS-40 enhancer. Interestingly, the suppression phenomena were not seen for the enhancerless plasmid p␣590GH (Fig. 3) or p597GH (data not shown). This result suggests that suppression by p65 or other NF-B factors is mediated mainly through the HS-40 enhancer region instead of the promoter(s) per se.
The HS-40 enhancer consists of at least five functional motifs that are bound with factors in vivo in an erythroid lineage-and developmental stage-specific manner (8,33,38,39). These motifs include a GT motif, two GATA-1-binding sites, and two NF-E2/AP1-binding sites, 5Ј-NA and 3Ј-NA (8). In vitro, the NA motif(s) could be recognized by a number of different factors including the erythroid-enriched NF-E2 and more ubiquitously expressed AP1, small maf (avian musculoaponeurotic fibrosarcoma virus) homodimers, and Bah1, Bah2, Bah3, etc. (Refs. 38, 40, and 41 and references therein). To examine whether the NA motifs are involved in the suppression of the HS-40 enhancer function by p65, transient transfection experiments were carried out with a reporter plasmid, pGL3-(NA) 2 ␣(Ϫ87/ϩ41)Luc, containing two tandem copies of the NF-E2/AP1-binding sites, or NA, in front of an ␣ globin minimal promoter. The tandem NA sites confer red activation of the reporter by coexpression of p45 and p18, i.e. NF-E2 (Fig. 4, compare top and bottom bars of  left panel). Coexpression of p65 in K562 cells, on the other hand, resulted in a marked decrease of the reporter activity (Fig. 4, left panel, second bar from bottom). Similar to the experiments with pHS40-␣590GH or pHS40-597GH, cotransfection with IBAA reversed the p65-mediated suppression. In contrast, the expression of p65 and/or IBAA had little effect on the control reporter construct, pGL3-␣(Ϫ87/ϩ41)Luc, without the two NA sites (Fig. 4, right panel). The data of Fig. 4 together with Fig. 3 suggest that presence of cis-linked NA sites is both necessary and sufficient for p65-mediated repression of the ␣-like globin promoters.
Repression of Endogenous ␣-Like Globin Messages by Exogenous p65-To study the effects of p65 on the expression of endogenous ␣-like globins, stably transfected K562 cells over- expressing p65 were generated. This was accomplished by the transfection of K562 cells with pEGFP-HAp65 encoding GFP as well as a HA-p65 fusion protein. As a control, K562 cells were also transfected with the vector pEGFP-C1 encoding only the GFP. After selection with G418, two pools with integrated pEGFP-HAp65, K(p65) and K(p65)Ј, and one pool integrated with the vector, K(V), were isolated by cell sorting. Expression of recombinant p65 in these stable clones was then determined by Western blot analysis. As exemplified by K(p65), both the HA and p65 epitopes could be detected in K562 overexpressing p65 (Fig. 5A, left two panels, lanes 3) but not in the parental K562 or K(V) cells (Fig. 5A, left two panels, lanes 1 and 2). Consistent with previous reports that p65 positively regulates I〉-␣ (42,43), the K(p65) cells (Fig. 5A, right top panel, lane 3) and K(p65)Ј cells (see Fig. 7B) also have elevated levels of I〉.
Next, we performed electrophoretic mobility shift and transient transfection assays to test the DNA binding abilities and transactivation potential of the recombinant p65 in the K(p65)  (Fig. 5B, top panel). Two specific protein-DNA complexes could be observed when K(p65) nuclear extract was incubated with a radiolabeled oligonucleotide, Ig〉, the B site from the immunoglobulin light chain enhancer (Fig. 5B, lane 2). In contrast, there was no obvious formation of these complexes in the control K(V) extract (Fig. 5B, lane 1). The cold wild-type competitor oligonucleotide, but not a mutant one, efficiently inhibited the complex formation (Fig. 5B, compare lanes 3 and  4). In addition, anti-p65 or anti-p50 (Fig. 5B, lanes 5 and 6), but not unrelated anti-IB and anti-c-Jun antibodies (data not shown), supershifted these complexes. Thus, these complexes most likely resulted from binding of p65 homodimer (Fig. 5B, upper band) or p65/p50 heterodimer (Fig. 5B, lower band) to the IgB probe. There was no difference in the formation of the NF-Y complex in K(V) and K(p65) extracts (Fig. 5B, bottom  panel).
To further explore whether the recombinant HA-p65 in K(p65) cells could play a role in transcriptional regulation, a functional assay was performed. A reporter plasmid pB-Luc was transfected into proliferating K(p65) and K(V) cells. As shown in Fig. 5C, the luciferase activity in transfected K(p65) cells was 35-fold higher than that of the K(V) cells, suggesting that HA-p65 indeed could induce gene expression through binding to the 〉 enhancer in the plasmid. Taken together, the above immunoblot analysis, DNA binding study, and p〉-Luc reporter assay suggested that recombinant HA-p65 is functional in K562 cells.
Whether the endogenous ␣-like globin gene transcription in K562 cells was affected by expression of exogenous p65 was examined by Northern blot analysis (Fig. 5D). Remarkably, the steady-state levels of ␣ and globin transcripts were greatly reduced, by at least 70 and 50%, respectively, in K(p65) cells (Fig. 5D, compare lane 1 with lane 2). Similar to previous studies of K562 cells, hemin also induced the levels of the globin transcripts in both K(V) and K(p65) cells (Fig. 5D, compare lanes 1 and 3 and lanes 2-4). However, the repression of the ␣-like globin gene expression in K(p65) cells appeared to persist under the conditions of hemin treatment (Fig. 5D, compare lanes 4 and 3). The ␣-like globin gene transcription in the other pool, K(p65)Ј, is also greatly reduced ( Fig. 7A and data not shown). Thus, NF-B not only suppresses transfected ␣-like globin promoters, but it also suppresses transcription of the endogenous globin genes, further supporting the physiological roles of NF-B pathway in the regulation of the ␣-like globin genes during erythroid differentiation.
Formation of NF-E2/DNA Complex Is Reduced in K(p65) Cells-As already shown in Figs. 3 and 4, the suppression of the ␣-like globin promoters by p65, and probably by other NF-B components as well, is most likely mediated through protein-DNA complex(es) formed at the NA motifs of the HS-40 enhancer. It is thus logical to examine whether p65 exerts this suppression effect by interference with NF-E2-DNA complex formation at the NA motifs. To accomplish this, we used EMSA to assay nuclear extracts prepared from K(V) and K(p65) cells (Fig. 6). Similar to the previous studies (for example, see Ref. 44), several prominent protein-DNA complexes, including the major AP-1 and a minor NF-E2, were observed with the control K(V) extract (Fig. 6A, lane 2). These complexes could be com- peted out with the cold NA oligonucleotide (Fig. 6A, lane 3) but not with the IgB oligonucleotide (Fig. 6A, lane 4). The addition of anti-p45 and anti-c-Jun antibodies supershifted the NF-E2 and AP1 complexes, respectively (Fig. 6A, lanes 5 and 6). Use of preimmune serum had no obvious effect (Fig. 6A, lane 7). Interestingly, however, the NF-E2 complex could barely be seen with the nuclear extract from K(p65) cells (Fig. 6A, compare lanes 1 and 2). These data indicated that the presence of p65 in K(p65) reduced or inhibited the formation of NF-E2/ DNA complex. No significant variations in the intensities of Sp1 and NF-Y complexes in the two nuclear extracts were detected (Fig. 6B), which argued against a general inhibition of protein-DNA complex formation by p65.
Expression of p45 Is Repressed in p65 Overexpressing Cells-Among the possible reasons for the above reduction of NF-E2/ DNA complex formation is the reduced level of NF-E2 as caused by the presence of p65. To address this possibility, we investigated the expression level of the erythroid-enriched sub-unit p45 of NF-E2 in both K(p65) and K(p65)Ј in comparison with K(V). Either Northern (Fig. 7A) or Western (Fig. 7B) blot analysis showed that indeed the level of p45 expression in K(p65) or K(p65)Ј cells is significantly lower than in the control K(V) cells (Fig. 7, top panels). Quantitation of the globin RNA and the IB protein served as the positive controls (Fig. 7, middle panels). DISCUSSION We have investigated the regulation of the ␣-like globin promoter activities as affected by the activation of the NF-B pathway. It appears, from DNA transfection analysis, that different NF-B factors could suppress the ␣-like globin promoters in erythroid cells derived from either embryonic/fetal or adult lineage. The suppression required the presence in cis of the HS-40 enhancer and was mediated at least in part through the NF-E2/AP-1-binding sites, or NA motif, in the enhancer. Indeed, activation of the NF-B pathway repressed the expression of the erythroid-specific subunit, p45, of NF-E2, thus leading to the reduction of NF-E2/DNA complex formation with the NA motif(s).
Although in most cases NF-B acts as an activator for gene expression, inhibition by NF-B is not without precedents. For example, studies on Dorsal, a Drosophila homologue of the members of the mammalian NF-B/Rel family, suggested that NF-B negatively regulated the expression of maternal effect genes controlling the dorsal-ventral pattern formation (45). Also, TNF-␣ reduced extracellular matrix deposition through inhibition of the expression of the structural matrix components such as the type I collagen (46). Finally, as already mentioned, two studies have linked globin gene silencing with the NF-B signaling (Refs. 15 and 16; see further discussion below). In contrast to the positive regulation of gene expression by NF-B in which the factor(s) binds to B motif(s) and interacts with the basic transcription complex or coactivators such as CBP/p300 (47), the molecular mechanisms through which the NF-B factors exert their negative regulatory effects are not well defined (46,48).
In the case of the human ␣ and globin genes, we showed that their expression in erythroid cells could be negatively regulated by activation of the NF-B pathway. This could be accomplished either by treatment of the cells with TNF-␣ ( Fig.  1) or by ectopic expression of the NF-B family of factors such as p65, p50, and c-Rel (Fig. 2). Further analysis of the amounts of endogenous ␣-like globin messages and p45 in K562 cells either stably expressing exogenous p65 (Figs. 5D, 6, and 7) or The expression of p45/NF-E2, globin, IB, ␣-actin, and ␣-tubulin in K(V), K(p65), and K(p65)Ј cells were assayed by Northern (A) and Western (B) analyses. Note that the level of IB is higher in K(p65) than in K(p65)Ј cells, and this is due to the higher level of p65 in K(p65) cells (data not shown). treated with TNF-␣ (data not shown) suggested that negative regulation of the ␣-like globin promoters by NF-B may result in part from repression of the p45 gene. As shown previously by different groups, p45 is a subunit of NF-E2, a DNA-binding transcription activator (10) stably bound in vivo in the HS-40 enhancer and the locus control region of the ␤-like globin gene cluster (8,49) and is required for transcriptional activation of the globin genes in erythroid cells (Refs. 8, 50, and 51 and references therein). The negative regulation of expression of either the ␣-like globins or p45 gene by NF-B could be regulated at the transcriptional level either directly by NF-B factor(s) or indirectly by its downstream targets, such as c-myb (28,29) and the CCAAT/enhancer-binding protein factor (46) through binding to negative regulatory DNA motif(s). Alternatively, the suppression of p45 expression by NF-B could occur post-transcriptionally. In fact, comparison of the Northern and Western blot data of Fig. 7 strongly suggested that suppression of the p45/NF-E2 expression by NF-B also occurred at the translational or post-translational level. Finally, it should be noted that the suppression of the ␣-like globin gene expression by NF-B might not entirely be mediated through p45 suppression.
In combination with several previous studies, we propose a model for the functional relationship between NF-B and NF-E2 in erythroid gene regulation and erythropoiesis. It is known that NF-E2 not only transactivates erythroid-specific genes such as the globins (50), but it is also involved in erythropoiesis (52,53). On the other hand, the NF-B factors are highly expressed in early burst-forming units-erythroblast-derived precursors, but their levels decline during erythropoiesis (24). This is in contrast to the increase of NF-E2 during maturation of the erythroid cells (10). Our demonstration of the suppression of p45/NF-E2 by NF-B thus provides a partial explanation for the reciprocal changes of the amounts of NF-B and NF-E2 during erythropoiesis. To expand further, through repression of p45, the NF-B factor could be involved in the silencing of not only the globins but also other NF-E2-activated genes in hematopoietic progenitors and early stage erythroid cells. Although the model needs to be refined, it strongly suggests that inactivation of the NF-B pathway is one requirement for the activation of a subset of erythroid-specific genes including the globins, as well as for the normal progression of erythropoiesis.