The TR2 and TR4 orphan nuclear receptors comprise the DNA‐binding core of direct repeat erythroid definitive, a protein complex that binds to direct repeat elements in the embryonic and fetal β‐type globin gene promoters. Silencing of both the embryonic and fetal β‐type globin genes is delayed in definitive erythroid cells of Tr2 and Tr4 null mutant mice, whereas in transgenic mice that express dominant‐negative TR4 (dnTR4), human embryonic ε‐globin is activated in primitive and definitive erythroid cells. In contrast, human fetal γ‐globin is activated by dnTR4 only in definitive, but not in primitive, erythroid cells, implicating TR2/TR4 as a stage‐selective repressor. Forced expression of wild‐type TR2 and TR4 leads to precocious repression of ε‐globin, but in contrast to induction of γ‐globin in definitive erythroid cells. These temporally specific, gene‐selective alterations in ε‐ and γ‐globin gene expression by gain and loss of TR2/TR4 function provide the first genetic evidence for a role for these nuclear receptors in sequential, gene‐autonomous silencing of the ε‐ and γ‐globin genes during development, and suggest that their differential utilization controls stage‐specific repression of the human ε‐ and γ‐globin genes.
The human β‐globin locus is composed of ε‐ (embryonic), Gγ‐ and Aγ‐ (fetal), and δ‐ and β‐globin (adult) genes, which are spatially arranged from 5′ to 3′ and developmentally expressed in the same order. The embryonic ε‐globin gene is transcribed during the first 8 weeks of human gestation in yolk sac (primitive) erythroid cells. The first switch in β‐globin transcription results in the silencing of ε‐globin and concomitant activation of fetal γ‐globin when definitive erythropoiesis ensues in the fetal liver. Gradually, at around the time of birth, a second switch from γ‐ to β‐globin transcription occurs as the site of hematopoiesis shifts again to the adult bone marrow (Stamatoyannopoulos and Grosveld, 2001).
From genetic analyses of transgenic mice harboring mutated human β‐globin loci, two nonexclusive mechanisms for globin gene ‘switching’ have been postulated: one is regulation by sequences located in the globin promoters (autonomous gene control) (Magram et al, 1985; Townes et al, 1985; Raich et al, 1990; Dillon and Grosveld, 1991), and the other is competition among the globin genes for activation by the locus control region (LCR), an element required for abundant expression of all the globin genes (Grosveld et al, 1987; Choi and Engel, 1988; Behringer et al, 1990; Enver et al, 1990). In a competitive model, the gene closer to LCR should have a higher probability of interaction with the LCR and hence be more abundantly transcribed, unless the gene is autonomously silenced (Hanscombe et al, 1991; Tanimoto et al, 1999). Autonomous control plays a major role in silencing the human embryonic ε‐ and fetal γ‐globin genes in definitive erythroid cells (Raich et al, 1990; Dillon and Grosveld, 1991), whereas competitive control plays a major role in silencing of the adult β‐globin gene during the embryonic and fetal stages (Tanimoto et al, 1999). However, the molecular basis for the seamless integration of these regulatory phenomena is incompletely understood.
In analyzing possible autonomous silencing mechanisms governing transcriptional regulation of the ε‐ and γ‐globin genes, direct repeat (DR) elements (AGGTCA repeats), consensus binding sites for nonsteroidal nuclear receptors, were identified in the proximal promoters of both genes (Figure 1A). Mutation of the DR sequences in the ε‐globin promoter led to ε de‐repression in definitive erythroid cells of transgenic mice (Filipe et al, 1999; Tanimoto et al, 2000). In the hematologic condition known as hereditary persistence of fetal hemoglobin (HPFH), the fetal γ‐globin gene is abundantly transcribed in adulthood, with elevated synthesis (up to 30%) of γ‐globin in adult erythrocytes (Stamatoyannopoulos and Grosveld, 2001). HPFH mutations include small and large deletions in the locus as well as point mutations in the γ‐globin promoters. Of 16 documented naturally occurring HPFH promoter mutations, six are located within DR elements (Huisman et al, 1997). Introduction of artificial or naturally occurring mutations into the DR element leads to derepression of γ‐globin transcription in transgenic mice (Berry et al, 1992; Omori et al, 2005). These observations initially suggested a central role for promoter DR elements in both ε‐ and γ‐globin silencing in definitive erythroid cells.
Two nuclear factors, DRED (direct repeat erythroid definitive) and COUP‐TFII, can bind to the DR elements in the ε‐ and γ‐globin promoters in vitro, and both have been implicated in their repression (Filipe et al, 1999; Tanimoto et al, 2000). When we purified DRED from nuclear extracts of murine erythroid (MEL) cells, mass spectrometric analysis revealed that DRED is a large macromolecular complex containing the TR2 and TR4 orphan nuclear receptors (Tanabe et al, 2002). These data indicated that DRED might play a key role in repressing definitive erythroid ε‐ and γ‐globin transcription.
TR2 and TR4 have diverse biological functions. They form homodimers and heterodimers to bind to AGGTCA DRs separated by a 0–6 nt spacer, and can either activate or repress cellular target genes (Lee et al, 2002). TR2 and TR4 share common functions in regulating cellular genes: the CNTF receptor, multiple thyroid hormone‐regulated genes as well as genes involved in retinoic acid signaling (Lee et al, 2002). Recent gene ablation studies have revealed that mice lacking TR2 are viable and have no overt phenotypes (Shyr et al, 2002), whereas TR4 germline mutants display reproductive and neurological deficiencies (Collins et al, 2004; Mu et al, 2004; Chen et al, 2005).
Here, we report biochemical and genetic analyses of TR2 and TR4 and their loss‐ and gain‐of‐function effects on the regulation of β‐type globin gene transcription. The analyses of Tr2 and Tr4 null mutant as well as dominant‐negative TR4 (dnTR4) mutant mice showed that TR2/TR4 are stage‐selective repressors of the human ε‐ and γ‐globin genes: an ε‐globin repressor in both primitive and definitive erythroid cells, but at the same time a definitive stage‐specific repressor of the fetal γ‐globin genes. Forced TR2/TR4 expression resulted in precocious repression of the human embryonic ε‐globin gene, but in contrast, to induction of the γ‐globin gene in definitive erythroid cells. This gene‐selective repression of the human embryonic and fetal β‐type globin genes by TR2 and TR4 provides critical new insight into the molecular basis for gene autonomous, sequential silencing of the ε‐ and γ‐globin genes during embryonic development, and in concert with the competition hypothesis, provides a compelling molecular rationale for how globin gene switching during human development ensues.
DRED differs in affinity for embryonic and fetal globin promoter DR sites
During embryonic development, primitive erythrocytes produced in the murine yolk sac predominantly express the embryonic εy‐ and βh1‐globins. At around 12.5 d.p.c., when definitive erythropoiesis ensues in the murine fetal liver, the εy and βh1 genes are gradually silenced with concomitant activation of the two adult β‐globin (βmajor and βminor) genes whose expression continues after birth as the site of erythropoiesis shifts to the bone marrow (Whitelaw et al, 1990). The promoters of the mouse embryonic globin genes contain either two (εy) or one (βh1) DR elements, and are similar to the equivalent regions of their human orthologues, the embryonic ε‐ and fetal γ‐globin genes, respectively (Figure 1A). Neither the mouse nor the human adult β‐globin genes contain recognizable DR sites. This suggests that both the mouse and human embryonic or fetal β‐type globin genes could be regulated by DRED (Tanabe et al, 2002).
To test the hypothesis that TR2/TR4 might negatively regulate embryonic and fetal globin gene transcription, we first asked which of these promoter DR elements could bind to DRED. A 32P‐labeled probe from the ε‐globin distal promoter DR element that conforms best to the consensus binding site for nuclear receptors was used in these experiments; an assortment of oligonucleotides corresponding to human and mouse β‐type promoter DR elements were used as competitors. From these studies, equilibrium dissociation constants (Ki) representing the affinities of competitor oligonucleotides for DRED were determined.
The Ki value of the human ε distal DR element displayed the highest affinity of all the competitors tested (Figure 1A). The affinity of the γ‐globin promoter DR element was equivalent to known functional binding sites for RXR (Medin et al, 1994) or HNF‐4 (Jiang et al, 1997), members of the same subfamily (Laudet, 1997). Not surprisingly, the affinity of the human β‐globin promoter was too low to generate functional association. The mouse globin DR‐binding sites were of generally lower affinity than their equivalent human promoter counterparts. Typical data for the competitive binding experiments are shown in Figure 1B. These data indicated that the human and mouse embryonic and fetal β‐type globin gene promoters have DR elements with differential affinities for DRED in the following order: human ε distal>ε proximal>γ; mouse εy distal>βh1>εy proximal.
We next compared the affinities of TR2 or TR4 alone and the TR2/TR4 heterodimer to the ε‐globin distal DR element by expressing TR2 and TR4 separately or together. After transfection into 293T (kidney) cells, nuclear extracts were prepared and examined for binding to the ε distal DR element by electrophoretic gel mobility shift assay (EMSA) (Figure 1C). Expression of TR2 alone led to a weak signal for a DNA–protein complex that co‐migrated with authentic DRED (from MEL cells), whereas expression of TR4 alone yielded a more robust EMSA complex, even though TR2 was slightly more abundant than TR4 (Figure 1C). Coexpression of TR2 and TR4 generated an EMSA complex whose signal intensity was comparable to that of TR4 alone. These data indicate that the affinity of the ε‐globin distal DR element for the TR2 homodimer is lower than that for either the TR2/TR4 heterodimer or TR4 homodimer, whose affinities are roughly equivalent.
Effects of Tr2 or Tr4 loss‐of‐function on embryonic/fetal globin gene transcription
To investigate the in vivo roles of TR2 and TR4 in β‐type globin gene regulation, we first analyzed the expression of the endogenous globin genes in Tr2 or Tr4 null mutant mice. The level of the embryonic εy‐, βh1‐, and adult β‐globin mRNAs in 10.5 d.p.c. yolk sac and 13.5 d.p.c. fetal liver was determined by semiquantitative RT–PCR and normalized to endogenous α‐globin mRNA. In the yolk sac, there was no significant difference in the expression of any of the globin genes in the homozygous Tr2 or Tr4 mutant embryos or their wild‐type littermates (data not shown). In fetal liver definitive erythroid cells, βh1 silencing was significantly delayed in both Tr2 and Tr4 mutant fetuses; at the same stage, expression of the mouse embryonic εy and adult β‐globin genes was unaffected (Figure 2A). The results indicate a role of TR2 and TR4 in the repression of βh1 in the fetal liver, consistent with the hypothesis that DRED represses both embryonic and fetal DR‐regulated β‐type globin genes. However, these results do not provide clear evidence regarding a role, if any, for TR2 and TR4 in the regulation of the murine εy gene. As TR2 or TR4 can bind to DR elements as homodimers (Figure 1), they may be functionally redundant in their ability to regulate the embryonic and fetal β‐type globin genes. To examine this possibility, we interbred the Tr2 and Tr4 mutants to generate compound mutants in which to analyze β‐globin gene expression.
Tr2/Tr4 mutant loss‐of‐function effects on human β‐type globin transcription were assessed by breeding to a wild‐type human β‐globin YAC transgenic line (TgβYAC, line 264; Tanimoto et al, 2000). The total amount of transgene‐derived human β‐type globin mRNAs in animals bearing this YAC was only about 10% of mouse endogenous α‐globin transcript. ε‐ and γ‐globin silencing was significantly delayed (9‐ or 3.6‐fold increased ε‐ and γ‐globin gene expression in 14.5 d.p.c. fetal livers, respectively) in compound Tr2/Tr4 homozygous null mutant fetuses as compared to wild type, whereas expression of the human adult β‐globin gene was unaffected (Figure 2B). In the fetal livers of the Tr2–/–:Tr4–/+ or the Tr2–/+:Tr4–/– embryos, expression of the ε‐ and γ‐globin genes was induced to levels lying between those of wild‐type and compound homozygous null mutant fetuses. The data indicate that TR2 and TR4 play key roles in repression of the ε‐ and γ‐globin genes, and that TR2 and TR4 are genetically partially redundant. Furthermore, the data suggest that TR4 plays a more prominent role than does TR2 in vivo, as the Tr2–/+:Tr4–/– mutants displayed a more severe phenotype than did the Tr2–/–:Tr4–/+ mutants.
To determine whether these changes in mRNA accumulation were due to altered transcriptional activity, we quantified primary RNA transcripts recovered from fetal liver samples of human β‐type globin genes by RT–PCR using primer sets spanning exon–intron junctions (one primer for an exon sequence, and the other for an adjacent intron sequence). The abundance of β‐, γ‐, or ε‐globin primary transcripts in the 14.5/15.5 d.p.c. fetal livers of wild‐type mice was about 0.5, 0.2, or 0.2% of the corresponding mRNA abundance, respectively. In the fetal livers of the Tr2–/–:Tr4–/− mutant fetuses, ε and γ primary transcript levels were elevated 2.9‐ or 2.4‐fold, respectively, compared to wild type (Figure 2C). These data show that TR2 and TR4 exert repressive effects on ε‐ and γ‐gene transcription in definitive erythroid cells of the fetal liver, and thus that the effects of TR2/TR4 are transcriptional and are not due to (e.g.) altered longevity of primitive erythroid cells or erythroid cell‐specific mRNAs.
Transgenic mice that express wild‐type or dominant‐negative TR2/TR4
We next wished to address the possible roles of TR2 and TR4 in β‐type globin gene regulation after forced transgenic expression of wild‐type or mutant receptors. The crystal structure of the DNA‐binding domain of RXRα, a member in the same subfamily as TR2 and TR4 (Laudet, 1997), has been resolved, and the amino‐acid residues that determine the base and phosphate backbone contacts in complex with the DR element have been identified (Zhao et al, 2000) (Figure 3A). Based on that structure, we predicted which residues in TR4 would make contacts with the DR elements, and introduced three amino‐acid substitutions for those residues (Lys or Arg to Glu) to generate a potential dnTR4 mutant protein that should be defective in DNA binding, but should retain other (dimerization or co‐regulator interaction) activities.
After transfection of dnTR4 into 293T cells, nuclear extracts were examined for binding of the force‐expressed receptor to the human ε‐distal DR element (Figure 1C). The dnTR4 mutant produced no DNA–protein complex on EMSA, even though the dnTR4 protein is abundantly expressed in the transfected cells (from Western blotting) but is devoid of DNA‐binding activity. The dnTR4 mutant was next coexpressed in cells at the same time as the wild‐type TR2 or TR4 proteins to ask whether the mutant could, in fact, serve as a dominant‐negative receptor (Figure 3B). The DNA‐binding activities of TR2 and TR4 were significantly diminished by coexpression of dnTR4 protein. These data showed that dnTR4 could serve as a mutant receptor isoform that could block the intrinsic DNA‐binding activity of both TR2 and TR4.
In order to restrict their expression exclusively to hematopoietic cells, the dominant‐negative or wild‐type TR2 or TR4 cDNAs were cloned into G1‐HRD, a Gata1 construct that is sufficient to drive expression of any cDNA exclusively in (primitive and definitive) erythroid cells (Onodera et al, 1997). The eGFP gene was also placed under G1‐HRD control, and by microinjecting the two constructs together into fertilized oocytes, transgenic lines carrying the dnTR4 mutant (TgdnTR4) as well as wild‐type TR2 (TgTR2) or TR4 (TgTR4), or both (TgTR2/TR4) were generated (co‐integration of eGFP and nuclear receptor transgenes was verified by following multiple generations of breeding in each line).
The abundance and stability of transgene expression was first estimated by eGFP fluorescence, as measured by flow cytometric analysis of adult peripheral blood. Lines that expressed eGFP in >80% of the erythrocytes were selected for further analysis. Expression of the transgenes in the 14.5 d.p.c. fetal liver was determined by real‐time quantitative PCR; the level of dnTR4 mRNA was 11‐ or 9‐fold higher than endogenous TR2 or TR4 mRNAs, respectively. In the gain‐of‐function lines, the level of Tg‐derived TR2 mRNA was 1.7‐ to 7‐fold higher than endogenous TR2, whereas Tg‐derived TR4 mRNA levels were from 8‐ to 13‐fold greater than those of endogenous TR4 (Figure 3C).
The human ε‐ and γ‐globin genes are activated by dnTR4
We analyzed the effects of transgenic dnTR4 expression on human β‐type globin transcription by breeding to a wild‐type human β‐globin YAC transgenic line (Tanimoto et al, 2000) to generate TgdnTR4:TgβYAC compound transgenic animals. The abundance of human β‐type globin mRNAs in the 8.5, 10.5, and 12.5 d.p.c. yolk sacs or 14.5 d.p.c. fetal liver was determined by semiquantitative RT–PCR, normalized to mouse α‐globin mRNA (Figure 4A). The data show that the peak of ε‐globin mRNA accumulation in the yolk sac was induced 1.5‐fold in comparison to TgβYAC littermates upon concomitant expression of the dnTR4 mutant, and that ε‐globin gene silencing in the fetal liver (Figure 4B) was significantly delayed (4.2‐fold increased expression) compared to littermates without TgdnTR4. In contrast, γ‐globin mRNA accumulation was only slightly reduced in comparison to YAC‐containing littermates in yolk sac erythroid cells, whereas γ‐globin transcription, like ε, was induced 1.9‐fold in the fetal liver RNA samples. These data verify the Tr2 and Tr4 null mutant mouse analysis, and underscore the conclusion that TR2/TR4 is a repressor of the mouse βh1 gene, as well as the human ε‐ and γ‐globin genes. The data also suggest that TR2/TR4 may act in a stage‐selective manner: as an ε‐globin repressor in primitive (Figure 4A) and definitive (Figure 4B) erythroid cells, but as a fetal γ‐globin repressor only in definitive erythroid cells.
The embryonic εy gene is repressed by forced TR2 and TR4 expression
We next analyzed the consequences of forced transgenic expression of TR2 or TR4 on β‐type globin gene transcription. The abundance of mouse endogenous globin mRNAs in the 9.5 d.p.c. yolk sac and 14.5 d.p.c. fetal liver was determined by semiquantitative RT–PCR (Figure 5). In the embryonic yolk sac of the two TgTR4 lines, εy expression was reduced to 70 or 25% in comparison to their wild‐type littermates, whereas βh1 mRNA accumulation was unaffected. Unexpectedly, in both TgTR4 lines, expression of the adult β‐globin gene was induced by 2.0‐ or 2.5‐fold. In contrast, TR2 forced expression did not cause a change in expression of any of the β‐type globin genes (data not shown). The activation of adult β‐globin transcription upon forced TR4 transgenic expression was unpredicted, but may be due to a secondary consequence of promoter competition by the LCR due to coordinate εy repression.
In the fetal livers of the TgTR2 lines, embryonic εy globin gene expression was repressed to 45% of wild type (line 2), whereas βh1 and adult β transcription was unaffected. TR4 forced expression resulted not only in εy repression to 45% of wild type (line 2), but also in mild induction of βh1 (1.9‐ or 2.5‐fold higher than in wild‐type fetal livers). In the TgTR2/TR4 lines, εy gene expression was severely repressed (to 35 or 15% of wild type), but βh1 was even more robustly activated (by 8.8‐ or 4.3‐fold over wild type), whereas again adult β mRNA accumulation was unaffected. The data indicate that both TR2 and TR4 homo‐ and heterodimers can repress εy transcription. However, the data also show that forcibly expressed wild‐type TR2 and TR4 can induce βh1 expression, thereby causing a significant delay in ‘fetal’ gene silencing in the liver, thus superficially contradicting the loss‐of‐function data (showing that TR2/TR4 acts as a repressor of the βh1 gene in the fetal liver; see Discussion). The TgTR2/TR4 lines displayed essentially the same, but more robust, phenotypes in the fetal liver than fetuses bearing only TR2 or TR4 transgenes, underscoring the possibility that TgTR2 and TgTR4 function additively or synergistically. Forced expression of TR2 plus TR4 also resulted in induction of βh1 (2.4‐ and 6.7‐fold in the TgTR2/TR4 lines 1 and 2, respectively) in the adult spleen, but did not cause a significant change in εy or adult β‐globin mRNA (data not shown).
Finally, we performed a complete time‐course analysis of mouse β‐type globin mRNA abundance in one of the TgTR2/TR4 lines (line 1; Figure 6). The data show that TgTR2/TR4 forced expression reduced the peak level of εy transcription in the yolk sac, and accelerated εy silencing in the liver. In contrast, βh1 transcription was unchanged in the yolk sac, but induced in the fetal liver, causing a significant delay in βh1 silencing. Adult β‐globin transcription was unaffected except in the yolk sac, suggesting that its transient induction was secondary to repression of εy transcription.
Human ε‐globin transcription is repressed in TR2/TR4 transgenic mice
We next analyzed the effects of transgenic TR2/TR4 expression on human β‐type globin transcription by analysis of TgTR2/TR4:TgβYAC compound transgenic animals (Figure 7A). The abundance of human β‐type globin mRNA in the 10.5 d.p.c. yolk sac, 15.5 d.p.c. fetal liver, or adult spleen was determined by semiquantitative RT–PCR and once again normalized to endogenous mouse α‐globin mRNA abundance.
In the embryonic yolk sac, forced expression of TR2/TR4 repressed human ε‐globin to 28% (line 1) or 3% (line 2) of wild‐type (TgβYAC alone) levels, but did not change fetal γ‐ or adult β‐globin (data not shown) transcription at the primitive stage. In the fetal liver, TgTR2/TR4 expression repressed embryonic ε‐globin transcription to 54% (line 1) or 10% (line 2) of wild type, and activated γ‐globin by 3.9‐fold (line 1) or 3.6‐fold (line 2), precisely as observed at the same stage in the mouse orthologue, βh1. In the adult spleen, γ‐globin expression was also induced 3.9‐fold (line 1) or 5.1‐fold (line 2), but neither ε‐globin (data not shown) nor β‐globin accumulation was affected.
We performed a complete time‐course analysis of human β‐type globin transcription in the TgTR2/TR4:TgβYAC compound transgenic mice using the TgTR2/TR4 line 1 (Figure 8). The data show that elevated TR2/TR4 expression reduced the peak level of ε‐globin transcription in the yolk sac and significantly accelerated ε transcriptional silencing. In contrast, fetal γ‐globin expression in the yolk sac was not significantly affected, but was induced 2.2‐ to 7.3‐fold in the fetal liver, causing a delay in silencing of the γ gene in definitive erythroid cells. Adult β‐globin mRNA accumulation was unchanged in the yolk sac, but was repressed by 20% in the fetal liver. The consistent repression of human adult β‐globin synthesis in the fetal liver was unpredicted, but may be simply a secondary consequence of activation of the γ‐globin gene via promoter competition for LCR activity (Choi and Engel, 1988; Carter et al, 2002; Tolhuis et al, 2002). These data indicate that the transcriptional effects of forced TR2 and TR4 expression are conserved between the mouse and human embryonic and fetal orthologues: between εy and ε, as well as between βh1 and γ.
We quantified the primary RNA transcripts of the human β‐type globin genes to analyze the actual transcriptional activity in erythroid cells of the fetal liver and adult spleen. The levels of γ primary transcript of TgTR2/TR4 line 2 animals increased 2.2‐ and 1.5‐fold in the fetal liver or adult spleen, respectively, compared to the wild‐type TgβYAC transgene alone, whereas the ε primary transcript level was not significantly altered in the fetal liver (Figure 7B). These data indicated that forcibly expressed TR2 and TR4 induced γ‐globin transcription in definitive erythroid cells of the fetal liver and adult spleen. In contrast, the reduction of ε‐globin mRNA abundance in the fetal livers of TgTR2/TR4:TgβYAC compound transgenic mice can be largely ascribed to the TR2/TR4 effects on ε‐gene transcription in the few residual primitive erythroid cells that remain in the fetal liver.
We next analyzed the effects of TgTR2/TR4 on mutant human β‐globin YAC transgenes that bear mutant DR elements in either the ε‐ or Aγ‐globin promoters (Tanimoto et al, 2000; Omori et al, 2005) in order to determine whether or not the effects were mediated directly through globin promoter DR elements. One transgenic line bears a mutant YAC with 9 nt substitutions in the ε‐globin gene promoter that ablates both of its DR elements (Bepsi) (Tanimoto et al, 2000). This YAC mutant (TgBEPSI) was bred to TgTR2/TR4 line 2, and expression of the mutant ε‐globin gene in the 10.5 d.p.c. yolk sac was determined. The data show that repression of the ε‐globin gene observed in the wild‐type YAC is abrogated by the DR site mutations in the TgBEPSI YAC (Figure 7C). We also examined a different transgenic line: mutDR has a 4 nt substitution in the Aγ‐globin gene promoter that specifically abolishes the DR1 element (Omori et al, 2005); this mutation induced Aγ‐globin gene expression by six‐fold compared to the wild‐type YAC transgene (line 264). TgmutDR was bred to TgTR2/TR4 line 2, and expression of the (unmodified) Gγ‐globin gene (as the internal control) and the mutant Aγ‐globin gene was individually quantified. In contrast to the behavior of the wild‐type TgβYAC (Figure 7D), only the Gγ gene (bearing an intact DR element) in the TgmutDR YAC was activated. The YAC mutant data show that the effects of TR2/TR4 on transcription are direct, and are mediated by the binding of TR2 and TR4 to the DR elements in both the ε‐ and γ‐globin promoters.
We showed that the TR2/TR4 heterodimer can bind in vitro to the DR elements of both the human and mouse embryonic and fetal β‐type globin gene promoters with affinities that are comparable to the known association constants for TR2, TR4 and closely related nuclear receptors. The genetic analysis provides compelling in vivo evidence for repression of the human embryonic ε‐ and fetal γ‐globin genes by these receptors. In the transgenic mice bearing the human β‐globin YAC as well as a dnTR4 mutant, the human ε‐globin gene was activated in both primitive and definitive erythroid cells. In contrast, the human fetal γ‐globin gene was only slightly perturbed during the yolk sac stage of erythropoiesis, but was induced at the fetal liver (definitive erythroid) stage. These data suggest that TR2/TR4 comprise a stage‐specific, gene‐selective repressor of the human embryonic ε‐ and fetal γ‐globin genes, and may provide at least part of the molecular basis for explaining the gene‐autonomous, sequential silencing of those genes during development: that is β‐globin gene ‘switching’.
Forced erythroid‐specific TR2/TR4 expression resulted in precocious repression of human ε‐globin transcription, but in contrast, to induction of the γ‐globin gene in definitive erythroid cells by a currently unexplained mechanism. The analysis of DR mutant YAC transgenic mice demonstrates that the effects of TR2 and TR4 on the ε‐ and γ‐globin genes are direct. The expression of the DR‐mutated Aγ‐globin gene was elevated six‐fold in the fetal liver compared to the wild‐type Aγ gene, whereas γ transcription was elevated 3.6‐fold in the compound Tr2/Tr4 homozygous null mutation. These data indicate that the majority of repressor activity acting through the γ promoter DR element can be attributed to TR2 and TR4, although the involvement of other currently undefined, redundant factors in repression of the γ‐globin genes through their DR elements cannot be excluded.
Control of β‐type globin gene switching mediated by TR2/TR4
Based on the results presented here, we propose a hypothetical model for how TR2 and TR4 regulate human β‐type globin gene transcription (Figure 9). In primitive erythroid cells of the yolk sac, TR2/TR4 binds to the DR elements of the ε‐globin gene as a component of the DRED complex, thereby repressing ε‐globin transcription. In contrast, TR2/TR4 exerts little or no effect on the γ‐globin gene DR sites at this developmental stage; this is consistent with the results of the dnTR4 mutant analysis, and with the lack of any effect on TgmutDR transcription in primitive erythroid cells (Omori et al, 2005). Such gene‐selective repression could result from a greater number of ε promoter‐binding sites or to the higher affinity of DRED for the two ε‐globin promoter DR sites compared to the single γ‐globin promoter DR site.
In definitive erythroid cells of the fetal liver, the DRED complex represses both the ε‐ and γ‐globin genes in a gene‐autonomous manner, thus enabling the subsequent γ‐ to β‐globin switch, possibly because the activity or abundance of DRED increases in definitive erythroid cells (we have no evidence to either support or refute this hypothesis). Although ligands for TR2 or TR4 have not been identified, it is also possible that the activity of TR2/TR4 could be mediated by the differential availability of small molecule ligands (if such exist), which would provide a non‐cell‐autonomous basis for the developmental stage specificity and synchronicity of globin gene switching.
Although autonomous control has been shown to play a major role in the silencing of the human embryonic ε‐ and fetal γ‐globin genes in definitive erythroid cells (Raich et al, 1990; Dillon and Grosveld, 1991), roles for the LCR hypersensitive sites 2, 3, and 4 in gene‐selective or stage‐specific transcriptional activation of the β‐type globin genes have been demonstrated (Bungert et al, 1995, 1999; Navas et al, 1998, 2001). In this context, we would speculate that DRED may repress the ε‐ and γ‐globin genes not only by gene‐autonomous mechanisms, but also possibly by modulating the interaction of the globin genes with specific LCR hypersensitive sites.
The precocious repression of the embryonic mouse εy‐ and human ε‐globin genes after forced transgenic expression of TR2/TR4, when considered along with the induction of the ε‐globin gene in the compound Tr2/Tr4 homozygous null fetuses and after breeding to the dnTR4 mutant, provides compelling genetic evidence that TR2/TR4 is a bona fide repressor of the mouse and human embryonic globin genes. In contrast, the absence of repression of the murine embryonic βh1‐ and human fetal γ‐globin genes in primitive erythroid cells, and the unexpected induction of those genes in definitive erythroid cells upon TR2 or TR4 forced expression, superficially contradicts the proposed repressor function of TR2 and TR4 for the βh1‐ and γ‐globin genes deduced from the loss‐of‐function and dominant‐negative analyses. The absence of βh1 or γ repression in primitive erythroid cells could be a secondary consequence of promoter competition for LCR activity due to profound repression of the εy‐ or ε‐globin gene, but the induction of the βh1‐ or γ‐globin gene in definitive erythroid cells cannot be explained by promoter competition, as the βh1‐ and γ‐globin genes are induced even in the adult spleen where the repression of εy‐ or ε‐globin genes is no longer observed, and as the induction of the γ‐globin gene in the fetal liver is disproportionately greater than the degree of repression of ε.
We previously showed that DRED is a >500 kDa complex, consisting of TR2 and TR4 as well as other currently unidentified proteins (Tanabe et al, 2002). One likely explanation for the induction of the γ‐ (and βh1‐) globin gene may be a dominant‐negative effect of forcibly expressed TR2 and TR4, which are DNA‐binding subunits of the functional DRED holo‐repressor complex. The forcibly expressed orphan receptors could sequester and dilute limiting γ‐specific co‐repressor(s) in the DRED complex, thereby precluding formation of a functional DRED complex on the γ promoter DR element (model 1 in Figure 9). The lack of γ‐ (and βh1‐) globin gene repression in primitive erythroid cells by the forcibly expressed TR2/TR4 may be also explained by hypothetical γ‐specific co‐repressor(s), which would also be limited in abundance in the context of forced expression of TR2/TR4. Alternatively, it is possible that TR2/TR4 may bear an inherent context‐dependent transcriptional function that is specific to the γ‐ (βh1‐) globin gene, and is dependent on interaction with specific coactivators that are available only in definitive erythroid cells, in keeping with the dual regulatory capacity of most nuclear receptors (model 2 in Figure 9) (Glass and Rosenfeld, 2000). Distinguishing between these two possibilities is an immediate goal.
We finally speculate that the very modest activation of the mouse adult β‐globin gene in the yolk sac, and the repression of the human adult β‐gene in the fetal liver, upon forced expression of TR2 and TR4 may be ascribed to promoter competition for the LCR due to mouse εy‐gene repression, or human γ‐globin activation, respectively. However, the possibility that forcibly expressed TR2/TR4 are directly modulating mouse and human adult β‐globin gene transcription cannot be formally disproven at this time.
Genetic analyses examining transgenic mice harboring mutated human β‐type globin loci have revealed that silencing of the embryonic and fetal β‐type globin genes is initiated by gene‐autonomous mechanisms (Magram et al, 1985; Townes et al, 1985; Raich et al, 1990; Dillon and Grosveld, 1991), although the relative contributions of autonomous control versus competitive silencing (Tanimoto et al, 1999) are unresolved. In analysis of possible cis‐regulatory elements governing transcription of the ε‐ and γ‐globin genes, multiple silencing elements have been proposed to lie in the proximal or distal regions of those promoters (Raich et al, 1992, 1995; Peters et al, 1993; Li et al, 1998a, 1998b; Tanimoto et al, 2000), including those identified through the analysis of the HPFH mutations (Berry et al, 1992; Ronchi et al, 1995; Li et al, 2001; Omori et al, 2005). However, identifying the repressors that mediate gene silencing through any of these cis‐regulatory elements has proven to be remarkably difficult. Nonetheless, some well‐known activators (e.g. GATA‐1) have been suggested to harbor context‐ or stage‐specific erythroid repressor functions (Li et al, 1997).
Taken together, the data presented here represent the first genetic evidence showing that a trans‐acting factor controls gene‐autonomous silencing of the human embryonic and fetal β‐type globin genes directly through well‐defined cis‐regulatory elements in their promoters. Considering the presence of many other hypothetical as well as verified negative cis‐regulatory elements in the proximal promoters of the β‐type globin genes, we speculate that multiple trans‐acting factors, including TR2/TR4, collectively control autonomous embryonic and fetal globin gene silencing. Identifying, and then clarifying the specific roles of each of these transcriptional effectors as well as the co‐regulators recruited to these genes via those sequence‐specific transcription factors, will be essential for finally elucidating the molecular mechanisms that control β‐globin gene switching.
TR2/TR4 as a target for therapeutic intervention in sickle‐cell disease
Sickle‐cell disease is caused by a missense mutation in the adult β‐globin gene and affects millions of people worldwide (Stuart and Nagel, 2004). Based on biochemical and epidemiological evidence, therapeutic agents that increase γ‐globin production are widely expected to benefit sickle‐cell patients. While the fetal hemoglobin (HbF) level in erythrocytes of normal human adults is <1% of total hemoglobin, HbF concentrations among sickle‐cell patients can vary from 0.1 to 30%, and any increment in HbF level was found to increase survival (Steinberg, 2005). A previous clinical study showed that sickle‐cell patients with higher (>8.6%) HbF levels experience significantly reduced mortality (Platt et al, 1994).
In this study, we found that the abundance of human fetal γ‐globin mRNA in fetal liver definitive erythroid cells of compound Tr2/Tr4 null mutants was 9.5% of human adult β‐mRNA abundance, and was nearly four‐fold higher than that of genetically matched wild‐type (2.5%) embryos. These data, therefore, provide direct evidence that TR2/TR4 may comprise a useful molecular target for possible therapeutic intervention in treating sickle‐cell disease: pharmacological inhibition of the repressor activity of TR2/TR4 in definitive erythroid cells is predicted to induce γ‐globin synthesis and thereby ameliorate the disease. Identification of natural ligand(s) of TR2 and TR4, if they exist, would facilitate the design of antagonists that would block the repressor function of TR2/TR4. Similarly, identification of co‐repressor proteins in the larger DRED complex (Tanabe et al, 2002) may also be crucial for developing new therapeutics, as the interfaces between TR2/TR4 and those co‐factors would constitute additional targets for drug design that may selectively block effective interactions and therefore diminish DRED repressor function.
Materials and methods
For transient expression in cell culture, the mouse TR2 and TR4 cDNAs were appended to Flag‐tags at their amino termini (Tanabe et al, 2002), and then cloned into a CMV promoter‐driven expression vector pEGFP‐N3 (Clontech), replacing the eGFP gene. The cDNA encoding the dnTR4 mutant was generated by PCR‐directed mutagenesis, and cloned into the same vector. For transgenic expression, the Flag‐tagged mouse TR2 or TR4 cDNAs, the dnTR4 mutant, and eGFP from the pEGFP‐N3 plasmid were ligated to the KpnI–NotI fragment from IE3.9int‐LacZ (GATA1‐HRD; Onodera et al, 1997).
The Tr2 and Tr4 null mutant mice were described previously (Shyr et al, 2002; Collins et al, 2004). The phenotypes of these mutant mice were confirmed in other independent Tr2 and Tr4 null mutant mouse lines (O Tanabe, unpublished). For generation of transgenic mice expressing wild‐type TR2 or TR4, or the dnTR4 mutant, the expression DNAs were separated from the plasmid backbones and purified by gel electrophoresis and electroelution. The constructs were injected into fertilized mouse oocytes (CD1; Harlan) that were then transferred to foster dams (Nagy et al, 2003). Founder offspring were screened by PCR for the presence of transgenes, and then bred to wild‐type CD1 mice; F1 offspring were analyzed for the presence of the transgenes by Southern blots of tail DNA. Transgenic mouse lines bearing the wild‐type human β‐globin YAC (line 264; Tanimoto et al, 1999) and the mutant YAC transgenes Bepsi (line 588; Tanimoto et al, 2000) and mutDR (line 74; Omori et al, 2005) were described previously.
Total RNA from transgenic lines was extracted from the yolk sac, fetal liver, or adult spleen (6‐week‐old, treated with 1‐acetyl‐2‐phenylhydrazine to induce anemia), and then used as a template for first‐strand cDNA synthesis. Details of the RT–PCR assay and primer sequences are described in Supplementary data. To determine the ratio of Gγ to Aγ mRNAs, both cDNAs were coamplified with the common γ primers, and then digested with PstI (Omori et al, 2005). The relative abundance (molar ratio) of β‐type globin RNAs normalized to mouse α‐globin mRNA was calculated according to the following equation:
Rβ where Rα is the radioactivity of PCR product for β‐type or mouse α globin measured by PhosphoImager; Cβ, Cα number of C nucleotides incorporated by PCR in each amplicon for β‐type or α globin; Eβ, Eα amplification efficiency of a primer set for β‐type or α globin; Nβ, Nα number of PCR cycles for β‐type or α globin.
The amplification efficiency for each primer set was experimentally determined for each tissue by plotting radioactivity of PCR products against cycle numbers over a 6–8 cycle range.
Real‐time PCR analysis for quantifying TR2, TR4, and GATA‐1 cDNA
Real‐time PCR analysis was performed with 0.1 μl of 14.5 d.p.c. fetal liver cDNA prepared as described above in a 25 μl reaction using an ABI Prism 7000 and SYBR Green PCR Master Mix (Applied Biosystems). Sequences of the primers are described in Supplementary data. All the primer sets were designed to span introns. The abundance of each cDNA was determined based on its Ct value and an experimentally determined amplification efficiency for each primer set, and then normalized to the abundance of GATA‐1 cDNA (as the internal control).
Cell culture and transient transfection
The mouse erythroleukemia cell line MEL and human embryonic kidney cell line 293T were cultured in DME medium (Gibco) with 10% fetal calf serum. For transfection, 2 × 106 293T cells were plated in a 10 cm dish the day before transfection. A total of 20 μg of expression plasmid was transfected into each dish using lipofectamine 2000 (Invitrogen). At 48 h after transfection, cells were harvested and nuclear extracts were prepared.
Nuclear extracts preparation, binding reactions, and electrophoresis were performed as described previously (Tanimoto et al, 2000). The oligonucleotides used are described in Supplementary data. To determine the affinity of each competitor by EMSA, the dissociation constant (Kd=0.56 nM) for the 32P‐labeled ε distal DR probe was initially determined by saturation binding experiments using nuclear extracts from MEL cells. The competitive binding experiments were then performed using 1.1 nM 32P‐labeled ε distal DR probe and 70 pM to 2 μM of each competitor to determine the 50% inhibitory concentration (IC50). The equilibrium dissociation constant (Ki) for each competitor was then determined as follows (Cheng and Prusoff, 1973):
Rabbit antisera against TR2 and TR4 were generated by fusing cDNA fragments (for the amino‐terminal regions of mouse TR2 (Leu35–Leu100) and TR4 (Ala43–Tyr116)) into the pET‐42a plasmid (Novagen), and then expressed as GST‐fusion proteins in Escherichia coli BL21‐CodonPlus (DE3)‐RIL (Stratagene). The fusion proteins were affinity‐purified with glutathione Sepharose 4B (Amersham), and then used as antigen for the preparation of rabbit antisera (Cocalico Biologicals Inc.). For Western blotting with the anti‐Flag mouse monoclonal antibody (Sigma), or the anti‐TR2 and ‐TR4 antisera, horseradish peroxidase‐conjugated secondary IgG was used for detection using the ECL system (Amersham).
Supplementary data are available at The EMBO Journal Online (http://www.embojournal.org).
We thank N Leff, W Song, and M Thompson for outstanding technical assistance, and L‐N Wei, Q Liu, F Katsuoka, and K‐C Lim for insightful discussions. Salary support was provided by the Cooley's Anemia Foundation and a research grant from AmGen Inc. (OT). This work was supported by research grants from the NIH (HL73465 and HL24415; JDE) and in part by the NIH through the University of Michigan Cancer Center Support Grant (CA46592).
- Copyright © 2007 European Molecular Biology Organization