Induction of Basic Helix-Loop-Helix Protein-Containing Complexes During Erythroid Differentiation

Abstract

The involvement of basic helix-loop-helix (bHLH) transcription factors in erythroid differentiation and development has been established by forced expression of the proteins TAL1 and Idl in cultured cell lines and by targeted disruption of the mouse TAL1 gene. To better understand the mechanism by which bHLH proteins regulate erythropoiesis, we have investigated HLH protein-DNA interactions in mouse erythroleukemia (MEL) cells before and during chemically induced differentiation. Three bHLH (E-box) binding activities were found to be induced in nuclei from differentiating MEL cells. Using specific antisera, we have demonstrated that these complexes are dimers of TALI and ubiquitous E proteins. Similar complexes were detected in nuclear extracts from a human erythroid cell line, K562, and from mouse fetal liver. All three bHLH complexes were disrupted in vitro by Id1, a dominant-negative HLH protein that we and others have previously shown to antagonize MEL cell differentiation. During differentiation of an Id1-overexpressing MEL cell line, induction of a complex containing TAL1 and E2A was not only blocked but reduced below the levels seen in undifferentiating cells. These observations are consistent with the idea that TAL1 and Id1 have opposing effects on erythroid differentiation and that the level of TAL1/E2A heterodimer and/or another E protein-containing complex may influence the decision of a cell to terminally differentiate.

COMMITMENT and differentiation of hemato-poietic stem cells along the erythroid lineage is controlled in part by tissue-restricted transcription factors [reviewed in (24,34,35)]. These include proteins of the basic helix-loop-helix (bHLH) family, which derives its name from a conserved structural motif consisting of a short basic region required for DNA binding, followed by two am-phipathic a-helices separated by an intervening loop of variable length (31). The HLH domain mediates dimerization, and dimers of bHLH proteins recognize the consensus element CANNTG, which is found in the promoters and enhancers of many cell-specific genes and is known as the E-box (31). bHLH proteins can be subdivided into several groups (31,32). Among these, E proteins (class A bHLH proteins) are ubiquitously expressed and can form homodimers and heterodimers. Class B proteins such as MyoD (10) are expressed in a tissue-specific manner and homodimerize poorly but form heterodimers with the class A proteins. The dominant-negative HLH proteins, exemplified by Idl (5), lack the basic region required for DNA binding. Therefore, complexes between Id and class A proteins or class B proteins cannot bind DNA and are inactive. Hence, by functioning combinatorially, this family of proteins comprises a system of regulators with the potential for complex and sensitive control of gene expression.

Mouse erythroleukemia (MEL) (14) cells are thought to be developmentally arrested at the proerythroblast stage and can be chemically induced to undergo a limited program of terminal differentiation (28,29). MEL cells express a number of HLH genes in regulated patterns during their differentiation (6,15,26,39,43,45,48). Therefore, it has been suggested that erythroid cell differentiation can be promoted or antagonized through changes in the relative levels of different HLH protein complexes (1,26,43). For example, transcription of the TAL1 gene, which encodes the major tissue-specific bHLH protein expressed in the erythroid lineage (4,15), increases during Me₂SO-induced differentiation of MEL cells (1,49). Overexpression of TAL1 enhances the spontaneous differentiation of MEL cells, whereas expression of antisense TAL1 RNA or a TAL1 mutant protein lacking the DNA binding basic region blocks chemically induced differentiation (1). Conversely, Idl messenger RNA is rapidly down-regulated in MEL cells during treatment with Me₂SO (6,26) and forced expression of Idl blocks their differentiation (26,43). Although TAL1 and Idl do not heterodimerize with each other (17,45), each may form heterodimers with E proteins such as those encoded by the E2A gene, E12 and E47/ E2-5 (5,17,50). By analogy with MyoD and Idl (21), TAL1 and Idl may compete for interactions with E proteins in erythroid cells such that the relative levels of active (TAL1/E) and inactive (Idl/E) complexes influence the expression of E-box-containing target genes involved in the switch to the differentiated state. No target genes for TAL1 have yet been identified in erythroid cells.

A requirement for TAL1 in embryonic blood development in vivo has been established by targeted disruption of the mouse TAL1 gene (38,40,42). Analysis of these mutants indicates that, in addition to its role during terminal erythroid differentiation, TAL1 also functions very early in hematopoiesis. TAL1 does not homodimerize but forms heterodimers with class A bHLH proteins to bind DNA (17,18). Surprisingly, genetic ablation of the mouse E2A gene, which encodes potential dimerization partners of TAL1 in erythroid cells, has no apparent effect on erythropoiesis (52,53). This result suggests either that E2A proteins do not function in erythroid development and differentiation or that they are redundant with other E proteins in erythroid cells. Consistent with this notion of overlapping functions is the absence of any apparent defects in the erythroid lineage in mice carrying homozygous-targeted mutations in either the HEB or E2-2 gene (54). The E proteins HEB (20) and E2-2 (16) could potentially substitute for E2A to form dimers with TAL1.

To elucidate the molecular mechanism(s) by which bHLH proteins regulate erythroid differentiation, we have examined the expression of E-box binding activities during Me₂SO-induced differentiation of MEL cells. We present evidence for three activities in MEL cell nuclear extracts that increase in level by 12–24 h of induction. Using antisera against TAL1 and E-type bHLH proteins, we show that these inducible DNA binding activities contain E2A and TAL1, HEB, or an as yet unidentified protein. All three complexes are disrupted in vitro by addition of exogenous Idl and they are the only E-box binding activities so affected. Finally, in a cell line that constitutively overexpresses Id1, TAL1/E2A binding activity is reduced. Surprisingly, no heterodimers of TAL1 and HEB or E2-2 were detected in MEL cells either before or during differentiation. Together, these results suggest that the balance of heterodimers between E2A and TAL1 or other bHLH proteins and the negative regulator Idl control terminal differentiation in the erythroid lineage.

MATERIALS AND METHODS

Cell Culture, Nuclear Extract Preparation, and Electrophoretic Mobility Shift Assay

MEL cells (subline DS19/sc9) and K562 cells (subline RA6) were maintained and induced as described previously (26). Stably transformed MEL lines C6 and B5 (26) were maintained in medium containing 0.2 mg/ml G418. Nuclear extracts were prepared from cultures (0.25–1.0 1) using the method of Dignam et al. (11) with the addition of protease inhibitors (2 μg/ml pepstatin A, 1 μg/ml leupeptin, and 5 μg/ml aprotinin) to buffers A, B, and C. Nuclear extracts were prepared from the livers (approx. 15) of 14.5-day embryos after homogenization of tissue in phosphate-buffered saline containing proteinase inhibitors.

Electrophoretic mobility shift assays (EMSA) were performed as described in Trepicchio et al. (46), except that binding reaction mixtures contained 2 μg poly(dl-dC) and 3.5–12 μg nuclear extract, and the binding reactions were analyzed on 4% or 5% nondenaturing polyacrylamide gels. For supershift experiments, antiserum (0.5–1 μl) was incubated with nuclear extract (6 μl) on ice for 20 min before addition of the other components of the binding reaction. Labeled oligonucleotide probe was added last. Reaction mixtures were incubated for 30 min at room temperature and then analyzed by nondenaturing acrylamide gel electrophoresis.

The sequences of the oligonucleotides used as probes and as cold competitors are shown in Table 1. Double-stranded ollgonucleotides were radiolabeled at room temperature by end-filling reactions using Klenow polymerase for 30 min.

TABLE 1.

OLIGONUCLEOTIDES USED IN THIS STUDY

TAL1 preferred E-box oligonucleotide

5′ -TGACCTGAACAGATGGTCGGCT-3′

3′ -TGGACTTGTCTACCAGCCGAGT-5′

TAL1 mutant oligonucleotide

5′ -TGACCTGAACCGATTGTCGGCT-3′

3′ -TCCACTTGGCTAACAGCCGAGT-5′

μE5 oligonucleotide

5′ -TGCAAGAACACCTGCAAACACCTG-3′

3′ -GTTCTTGTGGACGTTTGTGGACTG-5′

MCK oligonucleotide

5′ -GATCCCCCCAACACCTGCTGCCTGA-3′

3′ -GGGGGTTGTGGACGACGGACTCTAG-5′

βmaj oligonucleotide

5′ -TG ATTGAGCAAATGCGTTCGC-3′

3′ -TAACTCGTTTACGCAAGCGGT-5′

E-boxes (bHLH binding sites) are underlined and shown in boldface letters (E-box consensus: 5′-CANNTG-3′).

COS-1 Cell Transfection

COS-1 cells were transfected essentially as described previously (47). Cells were transfected with 10 μg of a plasmid consisting of the vector pMT23 (Genetics Institute, Cambridge, MA) containing the entire coding sequence of human E47 (31). Forty-eight hours after transfection, cells were harvested and pellets were resuspended in 150 μl of lysis buffer (47). For EMSA, 1 μl of this lysate diluted 1:40 with lysate from mock-transfected cells was added to binding reactions.

GST Fusion Proteins

The plasmid pGEX-2t-Id_PAH2ΔJ, encoding an Id1 helix-2 mutant (37) fused to glutathione 5-transferase, was generated by replacing a AlwNl-Eag I fragment of pGEX-2t-IdΔJ (a gift from R. Benezra) spanning nucleotides 32–399 of the Idl coding region (5) with the corresponding AlwNl-Eag I fragment from IdPAH2er (26). The presence of the mutation was confirmed by DNA sequencing. Glutathione 5-transferase fusion proteins were prepared according to a standard protocol (44). Protein concentration was determined by Bradford protem assay (Biorad) and integrity of the protein was verified by SDS-PAGE followed by Coomassie staining.

For EMSA experiments, nuclear extract (3.5 μg) was incubated with GST protein (0.5 μg) for 20 min at 37°C before addition of other binding reaction components.

RESULTS

MEL Cell Nuclear Extracts Contain Several E-Box Binding Activities That Differ in Their Affinities for Distinct E-Box Sequences

MEL cells are known to transcribe several HLH genes (5,20,39,45,49), but translation of HLH proteins and their interactions in this system have been less well characterized. We and others have previously shown that overexpression of the dominant-negative HLH protein Id1 blocks MEL cell differentiation in response to Me₂SO (26,43), presumably by disrupting or preventing the formation of one or more DNA binding bHLH species. Therefore, to understand how HLH proteins are involved in promoting this differentiation switch, we asked which E-box DNA binding activities are present in MEL cells.

Nuclear extracts were prepared from undiffer-entiated MEL cells and used in EMSA experiments with a variety of different E-box DNA probes. Here we discuss results obtained only with three of these probes, the ones (TAL1, μE5, and MCK) for which some changes in protein binding complexes were detected during differentiation. The TAL1 preferred binding site (19) (hereafter referred to as the TAL1 E-box), μE5 (31), and MCK (2) E-boxes each contain the canonical CANNTG motif but differ in the central and flanking residues (see Table 1). The μE5 E-box sequence is found in the immunoglobulin heavy-chain enhancer, and the MCK E-box probe corresponds to the MCK-R site found in the murk muscle creatine kinase enhancer. As shown in Fig. 1A, each E-box probe is recognized by a distinct pattern of DNA binding activities in the nuclei of undifferentiated MEL cells. Based on their mobilities, we have denoted a subset of these activities A through E (Fig. 1). Complexes A and B (which appear as a single wide band in some of the figures but on lighter exposures and longer electrophoretic runs can be clearly seen as two distinct complexes; Fig. 1C) were present at similar levels for all three probes, whereas the slower migrating complex C was observed only with the TAL1 E-box probe (lane 1) and complexes D and E were observed only using the μE5 and MCK E-box probes (lanes 2 and 3). Several other complexes were detected for one or both of the μE5 and MCK probes (e.g., those indicated with arrows in Fig. 1A). This article will focus on complexes A–E.

FIG. 1.

E-box binding activities in the nuclei of undifferentiated MEL cells. (A) Three E-boxes each give distinct patterns of complexes. Uninduced MEL cell nuclear extracts (5 μg) were incubated with 0.3 ng of each of the three E-box probes and analyzed by EMSA on a 4% nondenaturing acrylamide gel. Complexes A and B were observed with each of the three probes. Complex C was most prominent for the TAL1 probe, whereas complexes D and E were most abundant for the μE5 and MCK probes. Arrows indicate other complexes. F indicates free probe. (B) Specificity of complexes binding the TAL1 preferred E-box probe. The TAL1 E-box oligonucleotide probe (3 ng) was incubated with 10 μg of uninduced MEL cell nuclear extracts in the presence or absence of unlabeled competitor oligonucleotides. Lane 1, no competitior; even lanes 2–10, 6 ng (20-fold molar excess) competitor; odd lanes 3–11, 60 ng (200-fold molar excess) competitor. Competitor oligonucleotides were as follows: lanes 2–3, TAL1 E-box; lanes 4–5, TAL1 mutant E-box; lanes 6–7, μE5 E-box; lanes 8–9, MCK E-box; lanes 10–11, β-major globin promoter putative E-box. F indicates free probe. Binding reactions were analyzed by nondenaturing electrophoresis on a 5% acrylamide gel. Proteins in complexes A, B, and C specifically recognized the E-box motif, with a preference for the TAL1 E-box sequence. (C) Specificity of complexes containing the μE5 and MCK E-box probes. The indicated labeled oligonucleotide probes (0.12 ng) were incubated with 6 μg nuclear extracts from MEL cells treated for 4 days with Me₂SO and analyzed by EMSA on a 5% nondentauring acrylamide gel. A 250-fold molar excess (3 ng) of the indicated competitor oligonucleotides was included in reactions in lanes 2–4 and 6–8. Complexes A, B, D, and E are all specific for the E-box motif. The complex indicated by the arrow is not specific for the E-box sequence (figure and also data not shown).

The three complexes formed on the TAL1 E-box probe in Fig. 1A were specific for the E-box motif, as shown in Fig. 1B. Addition of a 20- to 200-fold molar excess of unlabeled E-box oligonucleotide abolished complex formation with the labeled probe (lanes 2, 3, 6-9), whereas a mutated TAL1 E-box oligonucleotide, in which CCGATT was substituted for the CAGATG E-box core (see Table 1), had no effect (lanes 4, 5). Wild-type μE5 or MCK oligonucleotides competed for protein binding by the TAL1 probe less effectively than the TAL1 oligonucleotide itself (compare lanes 6 and 8 with lane 2, and lanes 7 and 9 with lane 3), indicating a preference of these proteins for the TAL1 E-box. An oligonucleotide spanning the region of the mouse 0-major globin promoter from −268 to −250 (Table 1) (36) and containing a potential E-box (AGCAAATGCG) did not compete for binding (lanes 10, 11). We have been unable to detect specific protein binding to this sequence (data not shown).

Similar competition experiments indicated that complexes D and E are also specific for the E-box (Fig. 1C). Complexes A, B, D, and E, observed with the μE5 (lanes 1–4) and MCK (lanes 5–8) E-box probes, were competed by a 200-fold molar excess of unlabeled oligonucleotides corresponding to the μE5 (lane 2), MCK (lane 6), or TAL1 (lanes 3, 7) E-box, but not by the mutated TAL1 E-box (lanes 4, 8).

Induction of E-Box Binding Activities During Erythroid Differentiation

We next examined how the pattern of each DNA binding activity changes during MEL cell differentiation. Nuclear extracts prepared from uninduced MEL cells and from MEL cells induced to differentiate for 2 or 4 days were examined for E-box binding activity by EMSA. For each of the three probes tested, the pattern of protein-DNA complex formation changed with time of induction (Fig. 2A). The level of complex C, containing the TAL1 probe, increased by ∼5-fold after 2 days (compare lanes 2 and 3) then decreased to the uninduced level by 4 days (lane 4). A similar pattern was observed for complex C when the MCK E-box was used as a probe (lanes 10–12). Complex D, formed by interaction of MEL cell proteins with the μE5 and MCK E-box probes, increased strongly by 4 days of induction (Fig. 2A, lanes 6–8, 10–12), as did complex E with the μE5 probe. Levels of complex A, containing the TAL1 or μE5 E-box probe, increased by 4 days (Fig. 2A, lanes 4 and 8). No significant changes were observed for complex B.

FIG. 2.

Changes in E-box binding activities during Me₂SO-induced MEL cell differentiation. (A) Four-day time course. Labeled oligonucleotide probes (lanes 1–4, 0.12 2n TAL1; lanes 5–8, 0.06 ng μE5; lanes 9–12, 0.06 6n gCK) were incubated without protein (lanes 1, 5, 9) or with 5 μg nuclear extracts from unin-duced MEL cells (lanes 2, 6, 10) or MEL cells treated with Me₂SO for 2 days (lanes 3, 7, 11) or 4 days (lanes 4, 8, 12), and analyzed by nondenaturing electrophoresis on a 5% acrylamide gel. Complexes A and B were observed for all three probes. Complex C was most prominent for the TAL1 probe. Complexes indicated by the arrows are not specific for the E-box sequence. (B) Two-day time course. Labeled oligonucleotide probes (lanes 1–5, 0.24 ng TAL1; lanes 6–10, 0.24 ng μE5; lanes 11–14, 0.0 2ng MCK) were incubated without protein (lanes 1, 6) or with 6μ nuclear extract from uninduced MEL cells (lanes 2, 7, 11) or MEL cells treated with Me₂SO for 12 h (lanes 3, 8, 12), 24 h (lanes 4, 9, 13), or 48 h (lanes 5, 10, 14). The level of complex C (TAL1 probe) increased during the period from 12 to 48 h after addition of Me₂SO but dropped by 4 days (A). Complex D showed a biphasic expression pattern with a low-level increase within 12 h (μE5 and MCK probes) and more sharply by 4 days of Me₂SO treatment.

Because downregulation of Id1 mRNA occurs within hours after addition of Me₂SO to the medium (26), we evaluated binding to the three E-box probes using nuclear extracts prepared at 0, 12, 24, or 48 h (Fig. 2B). For the TAL1 probe, the levels of complex C increased gradually from 0 to 48 h (lanes 2–5), whereas complex D (clearly visible in this particular experiment with all three probes) showed a biphasic expression pattern, increasing markedly by 12 h (lanes 3, 8, 12) and then decreasing by 24 h (lanes 4, 9, 13). Complex E, not visible in this particular experiment, displayed kinetics similar to those of complex D at 12 and 24 h (data not shown).

Protein Composition of Inducible E-Box Binding Activities

The TAL1 protein exists in two forms (331 and 156 amino acids) resulting from translational initiation at distinct AUG codons (8). The smaller protein pp22^TAL1 corresponds to residues 176–331 of the larger protein pp42^TAL1. Both protein species have been found in MEL cells (33) and heterodim-ers of each protein with E2A have distinct mobilities in EMSA experiments (17). To examine the protein composition of TAL1 complexes, we used an antiserum raised against the carboxy-terminal 94 amino acids common to both polypeptides (19) and another that is specific for residues 1–121 found only in the longer translation product. Nuclear extracts from MEL cells grown for 48 h in the presence of Me₂SO were preincubated with immune or control serum prior to addition of the TAL1 E-box probe and DNA binding was assayed by EMSA. Figure 3A shows that the inducible complex C was specifically ablated by preincubation of nuclear extract with either TAL1 antiserum (lanes 2, 4), whereas control sera had no effect (lanes 3, 5). No changes were observed for the faster-migrating complexes A and B. We conclude that most or all of the TAL1 protein in DNA binding complexes in MEL cells is the longer translation product, pp42^TAL1.

FIG. 3.

Supershift and/or disruption of TAL1 E-box binding complexes with antisera to HLH proteins. (A) Inducible complex C contains full-length TAL1 protein. Nuclear extracts (4 μg) from MEL cells induced for 2 days with Me₂SO were incubated with antisera against the C-terminus (lane 2) or N-terminus (lane 4) of the TAL1 protein, or control sera (lanes 3, 5) prior to the addition of 0.6 ng TAL1 E-box probe and other binding reaction components. Samples were analyzed by electrophoresis on a 5% nondenaturing acrylamide gel. Complex C was ablated by the C-terminal antiserum, which recognizes both TAL1 polypeptides, and the N-terminal antiserum, which recognizes only the full-length protein. No other complexes were affected. Pre-immune serum (lane 3) and normal rabbit serum (lane 5) were used as controls. (B) Complex C is a heterodimer of pp42^TAL1 and E2A. Nuclear extracts (5 μg) from MEL cells treated for 2 days with Me₂SO were incubated with 1 ml of the indicated immune (i) or control (preimmune, c) antisera on ice for 20 min before addition of 0.12 ng TAL1 E-box oligonucleotide probe and other binding reaction components. After 30-min incubation at room temperature, reactions were analyzed by nondenaturing electrophoresis on a 5% acrylamide gel. The sample analyzed in lane 1 contained TAL1 E-box probe but no protein and the sample analyzed in lane 2 contained TAL1 E-box probe and 5 μg protein but no antiserum. *Nonspecific binding activity contributed by serum (data not shown). F, free probe. The inducible complex C was disrupted only by E2A (lane 3) and TAL1 (lane 9) antisera. Complexes A and B were not affected by any of the antisera.

We performed a similar analysis using antisera against the E proteins E2A (31), HEB (20), and E2-2 (16). The E2A antiserum (17) (Fig. 3B, lane 3), but not the anti-HEB (41) or anti-E2-2 (2) sera (lanes 5, 7), blocked formation of the inducible complex C but had no effect on complexes A or B. Extracts from a mouse B cell line were used as positive controls for the HEB and E2-2 antisera (not shown). A complex containing pp42^TAL1 and E2A was also detected in nuclear extracts from the human erythroid cell line K562 (data not shown). These experiments did not distinguish between the two proteins encoded by the E2A gene (El2 and E47/E2-5) (16,31).

We next assayed for the presence of TAL1 and E proteins in inducible complexes D and E. Nuclear extracts from MEL cells treated for 4 days with Me₂SO were preincubated with antisera against the E proteins and TAL1 before addition of the E-box probe. In experiments with the , μE5 and MCK E-box probes, antiserum against E2A supershifts and/or ablates both complexes D and E (Fig. 4, lane 2). Complex D is not affected by any other antiserum; thus, it appears to contain E2A but not TALI or either of the other two E proteins, HEB or E2-2. Complex E was super-shifted not only by anti-E2A serum but also by antiserum against HEB (lane 4), suggesting that this complex is a heterodimer of E2A and HEB proteins. As with the TAL1/E2A complex, we detected both “E2A/X” and E2A/HEB in K562 nuclear extracts (see below and data not shown).

FIG. 4.

Supershift/disruption of μE5 and MCK E-box complexes: inducible complexes D and E contain E2A. Nuclear extract (top panel, 4.5 μg; bottom panel, 5 μg) from MEL cells treated for 4 days with Me₂SO was incubated with 0.5 μ1 of the indicated immune (i) or control (c) sera prior to the addition of 0.6 ng μE5 (top) or 0.24 ng MCK (bottom) probe and EMSA analysis on a 4% (top) or 5% (bottom) nondenaturing acrylamide gel. Complex D was affected only by the anti-E2A serum, whereas complex E was affected by anti-E2A and anti-HEB sera. Normal rabbit serum was used as a control for the sample in lane 11; all other control samples contained preimmune serum.

The observation that complex D reacted with antisera against E2A proteins but not with any of the other antisera and its preferential binding to the μE5 E-box suggest that this activity may be a homodimer of E2A proteins. To address this question, we prepared lysates from COS-1 cells transfected with a human E47 construct and compared the mobility of human E47 homodimers produced in these cells with that of complex D from K562 cells. As shown in Fig. 5, two closely migrating complexes containing the μE5 E-box probe were observed in EMSA with COS-1 lysates from transfected (lane 3) but not mock-transfected cells (lane 2). The slower migrating of the two bands migrated with the same mobility as complex D from K562 cells (lane 1). It is therefore most likely that complex D is a homodimer of E2A proteins.

FIG. 5.

Complex D comigrates with E47 homodimer. Radiolabeled μE5 probe (6 ng) was incubated with 6 μg of nuclear extract from K562 cells (lane 1) or μ1 pi of lysate from mock-transfected COS-1 cells (lane 2) or COS-1 cells transfected with an expression construct for human E47 (lane 3). Samples were analyzed by EMSA on a 4% nondenaturing acrylamide gel. E47-transfected COS-1 cells contained two binding activities the slower of which comigrated with K562 complex D.

TAL1/E2A and E2A/X Are Present in Mouse Fetal Liver

To examine the potential relevance of the E-box binding activities observed in transformed erythroid cell lines to erythroid development in vivo, nuclear extracts were prepared from mouse fetal livers (14.5 days postcoitum) and analyzed by EMSA. At this stage of development, the liver is the primary site of hematopoiesis [for review, see (12)], and TAL1 protein is highly expressed in this tissue (23). In nuclear extract from mouse fetal liver, we detected a profile of TAL1 E-box binding activities similar to that found in MEL cells (Fig. 6A, lanes 1 and 2). Antibody supershift experiments established that in normal erythroid cells of the mouse, as in MEL cells, the majority of TAL1 protein in DNA binding complexes is found as heterodimers between the longer TAL1 translation product and E2A proteins (complex C, Fig. 6B). Similar analysis using μE5 or MCK E-box probes identified a complex that comigrated with MEL complex D and was disrupted only by the E2A antiserum (data not shown). Therefore, two complexes observed in induced MEL cells at high abundance are both found in nuclear extracts from normal erythroid cells.

FIG. 6.

Complex C, the TAL1/E2A heterodimer, is present in nuclear extract from mouse fetal liver. Nuclear extracts (4 μg) from MEL cells treated for 2 days with Me₂SO (lane 1) or prepared from mouse fetal livers (lanes 2–12) were incubated with the indicated immune (i) or control (c) sera prior to addition of 0.6 ng TAL1 E-box probe and other binding reaction components, followed by electrophoresis on a 5% nondenaturing acrylamide gel. Formation of complex C was blocked by antisera against E2A and TAL1 but not by control serum. None of the complexes were affected by HEB or E2-2 antiserum. *Nonspecific binding activity contributed by the serum. F, free probe.

The TAL1 /E2A and E2A /X Complexes Are Disrupted by Idl

To extend our analysis to bHLH proteins for which antisera are not available, we used a protein complex disruption assay. Glutathione S-transferase fusion proteins containing wild-type or mutant Idl were incubated with MEL cell nuclear extract before addition of other binding reaction components. GST-Id l contains the wild-type Idl protein, whereas GST-Idl_PAH2 contains a proline-to-valine substitution in amphipathic a-helix 2 (37). This mutant is unable to dimerize with E proteins in vitro (26,37) or to inhibit MyoD activity (37) or MEL cell differentiation (26) in vivo.

As shown in Fig. 7, preincubation of MEL nuclear extract with GST-Id l, but not with the mutant protein GST-Id_lPAH2, eliminated nearly all of the TAL1/E2A DNA binding activity (compare lanes 2 and 3), and completely eliminated the E2A/X (lanes 5, 8) and E2A/HEB (lane 5) complexes. The two faster migrating complexes A and B were unaffected by the addition of either protein to the binding reaction mixture, suggesting that they are formed by bHLH proteins that are resistant to Idl, such as those of the bHLH-leucine zipper class [e.g., USF or Myc-like proteins (7,27)].

FIG. 7.

Disruption of the inducible MEL cell complexes by Id1. Nuclear extract (3.5 μg) from MEL cells treated for 2 days (lanes 1–3) or 4 days (lanes 4–9) with Me₂SO was incubated with PBS/10 mM EDTA (lanes 1, 4, 7) or with 0.5 μg GST-Idl (lanes 2, 5, 8) or GST-Idl_PAH2 (nondimerizing mutant) (lanes 3, 6, 9) for 20 min at 37°C before addition of the indicated E-box probe (0.6 ng) and other binding reaction components. Formation of the TAL1/E2A, E2A/X, and E2A/HEB complexes was specifically disrupted by GST-Idl but not GST-Idl_PAH2 protein. Complexes A and B were unaffected by addition of either fusion protein.

Induced But Not Basal TAL1/E2A DNA Binding Levels Are Reduced in an Id l-Overexpressing Cell Line

We (26) and others (43) had previously shown that MEL cell lines constitutively overexpressing Id1 are blocked in their ability to terminally differentiate. To determine whether the levels of any of the inducible complexes are affected in these cell lines, nuclear extracts were prepared and examined by EMSA. The line B5 constitutively overexpresses Id1 at high levels. As a control, the C6 line, which was transfected with but does not express the Id1 construct (26), was used. As shown in Fig. 8 (lanes 2 and 6), after 48 h of Me₂SO treatment, an increase in TAL1/E2A binding was observed for the control line C6 (consistent with the results shown in Fig. 2A for parental MEL cells), whereas the Id1-overexpressing line B5 showed a decrease in TAL1/E2A binding below the levels seen in uninduced cells (lanes 4, 8). The decrease in TAL1/E2A binding activity observed for the Id1-overexpressing line B5 correlates with its block in differentiation; however, we have not detected consistent differences in E-box binding for other Id1-overexpressing cell lines (data not shown). No significant change in expression level was observed for complexes A, B, D, or E (Fig. 8; data not shown). Therefore, only complex C (TAL1/E2A) showed any change in these cells.

FIG. 8.

Reduction of TAL1/E2A levels in an Id1-overexpressing cell line treated with Me₂SO. Labeled TALI E-box probe (0.12 ng) was incubated with 12 μg of nuclear extract from uninduced (odd-numbered lanes) or 2-day Me₂SO-induced (even-numbered lanes) cells. Samples in lanes 3, 4, 7, and 8 contained extract from the Id-overexpressing line B5. Samples in lanes 1, 2, 5, and 6 contained extract from the control line C6. EMSA analysis was carried out on a 5% nondenaturing acrylamide gel. In the two experiments shown here, reduced levels of TAL1/E2A complex were observed for the Id1-overexpressing cell line B5.

The observation that overexpression of Id1 reduces the induced but not the basal level of TAL1/ E2A binding suggests that there is some difference in complex C detected in undifferentiated versus induced cells. To examine this possibility, we examined these complexes in supershift experiments using anti-TAL and other anti-bHLH sera, at a number of time points following induction. However, no differences were observed, suggesting that if complex C changes upon induction, the change is not one that can be detected by available antisera.

DISCUSSION

Our previous work established a functional relationship between downregulation of Idl and terminal differentiation of erythroid cells. In those studies, we showed that overexpression of Idl blocked Me₂SO-induced differentiation of mouse erythroleukemia cells, and that this inhibition required a functional Id1 helix-loop-helix dimerization domain (26). By analogy with the effect of Id1 on differentiation of myoblasts in culture (21), we reasoned that Idl in proliferating erythroid cells may prevent the formation of particular dimeric bHLH complexes involved in activating the differentiation program. In this report, we have begun to identify the components of such complexes by determining biochemically which E-box binding activities are present in actively dividing (undifferentiated) and terminally differentiating erythroid cells.

We have identified three inducible E-box binding activities in MEL cell nuclear extracts. One (complex C) preferentially recognizes a TAL1 preferred binding site out of a sampling of E-boxes and is a heterodimer between E2A and pp42^TAL1 proteins. The other two inducible E-box binding activities also contain E2A and interact preferentially with a different E-box core sequence. Complex E is a heterodimer between HEB and E2A; the binding partner of E2A in complex D is unknown. The levels of all three complexes change during differentiation. Induction of complexes D and E is biphasic, peaking first at 12 h and again at 96 h (4 days). TAL1/E2A levels increase steadily for 48 h before declining by 4 days. It has been reported that the DNA binding activity of TAL1 and E2A synthesized in vitro (19,50) or expressed in cultured hematopoietic progenitor cells (9) can be antagonized by addition of Id proteins. We have shown that all three of the inducible complexes in MEL cells are disrupted by Idl. Me₂SO triggers a rapid downregulation of Idl mRNA in MEL cells (5,26); the subsequent decrease in the level of Idl protein may therefore account for the observed increases in TAL1/E2A, E2A/HEB, and E2A/X during MEL cell differentiation. Although induction of TAL1/E2A is blocked in one Id1-overexpressing MEL cell line, we have not been able to detect consistent differences in the levels of E2A-containing complexes in all of our Id1-overexpressing lines. Therefore, other bHLH complexes important for erythroid differentiation may not have been detected by the E-box probes used in this study. Others have reported that over-expression of Id1 inhibits the increase in an MCK E-box binding activity at 4 days of Me₂SO induction (43). This activity may correspond to the E2A/X complex, which we also find is most abundant after 4 days of Me₂SO treatment.

The full-length TAL1 protein, but not its amino terminal-truncated form, was detected in E-box binding complexes. Both proteins are present in MEL cells (33) and have DNA binding activity as heterodimers with E2A proteins in vitro and in vivo (17). The amino-terminus of pp42^TAL1 has been reported to contain a weak transcriptional activation domain (51). This region is absent in the truncated form of TAL1. Whether or not the two proteins have distinct functions in erythroid cells is not known (40). Although no gene has yet been demonstrated to be a bona fide target for TAL1, good matches for the TAL1 preferred E-box are found in several promoters (18). Recent work from two laboratories has implicated a conserved E-box in hypersensitive site 2 (HS2) of the (β-globin locus control region in regulating LCR enhancer activity (13,25). In one study (13), mutation of the E-box at position 8701 diminished overall enhancer activity. In vitro, TAL1 but not E2A was detected in complexes formed by MEL cell nuclear proteins with this E-box sequence. El-nitski and coworkers (13) also reported inducible binding of the bHLH/ZIP protein USF to the HS2 8701 E-box. Complex A (this work) displays properties consistent with those expected of USF, including specificity for the E-box, mobility relative to the TAL1/E2A complex, and insensitivity to Id1.

We speculate that complex D (E2A/X) may be an E2A homodimer based on its preference for binding to the μE5 E-box and its mobility relative to that of human E47 expressed in COS-1 cells, but we cannot unequivocally rule out the possibility that it contains another, unidentified protein. The TAL1-related protein LYL1 is also expressed in MEL cells [our unpublished results; (20)], and can form heterodimers with E2A proteins. However, because LYL1 displays an almost identical E-box preference as TAL1 (30) and complex D shows a clear preference for the μE5 and MCK E-boxes over the TAL1 E-box, we think it is unlikely that the partner of E2A in complex D is LYL1. The presence of complexes containing LYL1 or the truncated form of TAL1 might have been obscured by other, more abundant complexes in MEL cell nuclear extracts.

E2A proteins were present in all three inducible complexes in MEL cells. Curiously, although the TAL1 gene is required for erythropoiesis (40,42), the products of the E2A gene are dispensable (3,53). These eindings suggest that trythroid dells contain other proteins that can compensate for E2A activity. However, although TAL1/E2A complexes were easily detected in MEL cell nuclei, we did not detect TAL1/HEB or TAL1/E2-2 complexes in nuclei from any erythroid cells, including nontransformed mouse fetal liver erythrocytes. E2A homodimers are known to regulate the expression of immunoglobulins (3,22,53) and are necessary for development of the B cell lineage (3,53), whereas simers so f2A and HEB have been previously shown to bind an E-box in the CD4 enhancer in T cells (41). Whether or not either of these complexes regulate transcription of genes in erythroid cells, either directly or indirectly (e.g., by influencing the levels of other HLH dimers), is unknown.

In conclusion, we have shown that the DNA binding activity of members of the helix-loop-helix protein family is regulated during MEL cell differentiation, consistent with the suggested role of these proteins in erythropoiesis. Complexes containing an E2A protein dimerized with TAL1, HEB, or another as yet unidentified partner may be of particular importance for terminal erythroid differentiation.