Mi2β Is Required for γ-Globin Gene Silencing: Temporal Assembly of a GATA-1-FOG-1-Mi2 Repressor Complex in β-YAC Transgenic Mice

Activation of γ-globin gene expression in adults is known to be therapeutic for sickle cell disease. Thus, it follows that the converse, alleviation of repression, would be equally effective, since the net result would be the same: an increase in fetal hemoglobin. A GATA-1-FOG-1-Mi2 repressor complex was recently demonstrated to be recruited to the −566 GATA motif of the Aγ-globin gene. We show that Mi2β is essential for γ-globin gene silencing using Mi2β conditional knockout β-YAC transgenic mice. In addition, increased expression of Aγ-globin was detected in adult blood from β-YAC transgenic mice containing a T>G HPFH point mutation at the −566 GATA silencer site. ChIP experiments demonstrated that GATA-1 is recruited to this silencer at day E16, followed by recruitment of FOG-1 and Mi2 at day E17 in wild-type β-YAC transgenic mice. Recruitment of the GATA-1–mediated repressor complex was disrupted by the −566 HPFH mutation at developmental stages when it normally binds. Our data suggest that a temporal repression mechanism is operative in the silencing of γ-globin gene expression and that either a trans-acting Mi2β knockout deletion mutation or the cis-acting −566 Aγ-globin HPFH point mutation disrupts establishment of repression, resulting in continued γ-globin gene transcription during adult definitive erythropoiesis.


Introduction
The human b-globin locus is composed of five functional genes (e, G c, A c, d, and b) and a master regulatory region called the locus control region (LCR). These genes are arrayed in the order in which they are progressively expressed during development. Expression of the b-like globin genes undergoes two major switches. The first is an embryonic to fetal switch that occurs between 6 and 8 weeks of gestation and involves the silencing of the embryonic e-globin gene in the yolk sac and the activation of the fetal c-globin genes ( A cand G c-globin) in the liver. The second switch is from the fetal c-globins in the liver to the adult globins (mostly b-globin, with d-globin as a minor component) in the bone marrow. This switch is characterized by the progressive silencing of the c-globin genes, with the concomitant activation of b-globin gene expression, and is not completed until after birth. An understanding of the mechanisms that regulate the globin gene switching is of fundamental importance, since reactivation of the fetal hemoglobin expression during definitive erythropoiesis is well-established as therapeutic for hemoglobinopathies such as sickle cell disease (SCD) and b-thalassemias.
Hereditary persistence of fetal hemoglobin (HPFH) is a condition characterized by elevated synthesis of c-globin in adult definitive erythroid cells, which normally have only very low levels of fetal hemoglobin (HbF). HPFH mutations include both small and large deletions in the b-globin locus (deletional HPFH), as well as point mutations in the two c-globin gene promoters (nondeletional HPFH). When a HFPH mutation is co-inherited with a SCD mutation, the SCD patients present with a better clinical evaluation due to the high levels of HbF.
We identified a novel A c-globin gene silencer motif and an associated repressor complex that are linked to a new HPFH point mutation [1]. This silencer is located at 2566 relative to the mRNA CAP site in a GATA binding motif and repression is mediated by GATA-1 binding at this site, with Friend of GATA-1 (FOG-1) and Mi2 (NuRD) as protein partners in this repressor complex. Interestingly, a mutation in the analogous 2567 GATA site of the G c-globin gene in an Iranian-American family was recently associated with a HPFH phenotype and GATA-1 protein was shown to bind at this site when c-globin is not expressed [2]. Together, these studies demonstrate that the 2566 A cand 2567 G c-globin GATA sites are true silencers and that the GATA-1 protein is the DNA-binding component that mediates c-globin gene silencing.
GATA-1 is a zinc finger transcription factor that plays a role during development in the differentiation of several cell types including erythrocytes, megakaryocytes, eosinophils and mast cells [3]. GATA-1 recognizes the consensus sequence (A/T) GATA (A/ G) and, like many other transcription factors, binds to its cognate DNA sequence, facilitating target gene repression or activation through recruitment of co-activator or co-repressor proteins [4]. Previously published studies demonstrate that GATA-1 is capable of acting both as an activator and a repressor of transcription [1,5,6]. GATA-1 binds the co-regulator FOG-1, which assists in potentiating transcriptional activation or repression [7,8]. These two proteins were shown to associate with the NuRD complex and mediate the repression of certain genes, including c-globin [1,5]. A repressive GATA-1/FOG-1/MeCP1 complex binds to silenced hematopoietic genes in erythroid cells, with FOG-1 serving as the bridging factor between GATA-1 and the MeCP1 complex [6]. A recent study demonstrated that the GATA-1/FOG-1/NuRD complex is also associated with gene activation [9,10].
In this study, we demonstrate that Mi2b is required for c-globin gene silencing. c-globin was increased in definitive erythroid cells from Mi2b conditional knockout human b-globin locus yeast artificial chromosome (b-YAC) transgenic lines, corroborating the involvement of Mi2 (NuRD) in establishing the permanent silencing c-globin gene expression. In addition, we focused on the temporal events leading to GATA-1-FOG-1-Mi2-mediated cglobin gene silencing. We hypothesized that repression is established gradually over time in the developing mouse fetus. Chromatin immunoprecipitation (ChIP) experiments performed on post-conception day E12-E18 fetal liver samples from b-YAC transgenic mice showed that GATA-2 occupies the 2566 A cglobin GATA site early in fetal liver definitive erythropoiesis when c-globin is expressed (day E12). GATA-2 vacates this site and is replaced by GATA-1 at day E16, followed by recruitment of FOG-1 and Mi2 proteins at day E17. Finally, we demonstrate that c-globin is expressed during adult definitive erythropoiesis in b-YAC transgenic mice carrying the T.G HPFH point mutation at the 2566 GATA motif of the A c-globin gene. The presence of this mutation disrupted recruitment of the GATA-1-FOG-1-Mi2 repressor complex to this motif, resulting in reactivation of cglobin expression during adult definitive erythropoiesis.

Expression of c-globin in Mi2b conditional knockout b-YAC mice
The NuRD complex is composed of the ATPase Mi2, MTA-1, MTA-2, p66, RbAp46 (RBBP7), RbAp48 (RBBP4), MBD3 and the histone deacetylases HDAC1 and HDAC2 [11]. Given the association of NuRD with other transcriptional repressors and the presence of a histone deacetylase and an ATPase subunit in this remodeling complex, NuRD is frequently associated with transcriptional repression [11,12]. Earlier ChIP experiments demonstrated that Mi2 is recruited to the 2566 GATA site of the A c-globin gene when c-globin is no longer expressed [1]. To further examine the role of Mi2 in the silencing of c-globin expression, a conditional knockout of Mi2b was created by breeding floxed Mi2b mice [13] with our erythroid-specific Cre expression mice [14] and our wild-type b-YAC transgenic mice [15] as described in Materials and Methods. Six mice were obtained and correct genotypes were determined. Conditional knockout of the murine Mi2b gene in our mice was demonstrated at the transcript level by real-time qRT-PCR. Mi2b mRNA expression was reduced to 50% (average of 6 animals, P,0.01) in peripheral blood samples from these mice compared to wild-type b-YAC mice ( Figure 1A). Expression of the murine globins (e y , bh1, and b maj ) and human globins (c and b) was analyzed by qRT-PCR and compared to wild-type b-YAC transgenic mice. Human c-globin gene expression was increased 8-fold in peripheral blood from adult conditional Mi2b knock-out mice (P,0.05, Figure 1B). Human b-globin and murine adult b maj -globin gene expression were decreased, but not significantly ( Figure 1C-1D). The murine embryonic bh1and e y -globins were expressed at the same level as in wild-type b-YAC mice (data not shown).
Expression at the protein level confirmed the transcription results. Mi2 protein expression was decreased nearly 50% in adult blood from two of the Mi2b conditional knockout mice (3 and 4) mice ( Figure 2E and 2F) compared to wild-type mice ( Figure 2B) as measured by flow cytometry. The other two Mi2b conditional knockout mice (1 and 2) showed a modest 20% decrease or no decrease in Mi2 protein expression, respectively ( Figure 2C and 2D). Taken together these data indicate variability of Cre excision efficiency among the mice. c-globin (HbF)-expressing F cells were measured in parallel ( Figure 2G-2L). Although all Mi2b knockout mice showed substantial increases in F cells, levels were variable and not concordant with the decrease in Mi2b expression. Additionally, cystospins of adult peripheral blood from two Mi2b conditional knockout mice queried with anti-human HbF antibody displayed a pancellular distribution of F cells ( Figure 2O-2P), similar to the 2117 Greek HPFH b-YAC mice ( Figure 2N), although the Mi2b conditional knockout mice showed fewer strong HbF-positive cells.
Mature RBCs are enucleated, making it difficult to demonstrate that the nuclear-localized Mi2b protein is reduced in these cells. To further demonstrate decreased expression of Mi2b protein in our conditional knockout mice, we derived nucleated CIDdependent bone marrow cells (BMCs) from our Mi2b conditional knockout mice. BMCs obtained and immortalized in this manner reflect the globin gene expression pattern observed in the adult transgenic mice from which they are derived [16]. Western blotting was performed using an anti-Mi2b antibody; a 240 KDa fragment corresponding to Mi2b was detected in CID-dependent wild-type b-YAC BMCs, but not in the CID-dependent Mi2b conditional knockout b-YAC BMCs ( Figure 3A). Real time PCR corroborated this result at the transcript level (data not shown). Finally, a 7.5-fold induction of c-globin mRNA level was measured in the CID-dependent Mi2b conditional knockout b-

Author Summary
Sickle cell disease (SCD) is one of the most common genetic diseases, affecting millions of people worldwide. SCD affects red blood cells' shape and renders them ineffective, resulting in anemia along with attendant complications. The disease is caused by a single point mutation in the coding sequence of the adult b-globin gene that changes normal adult hemoglobin (HbA) to sickle hemoglobin (HbS). Scientific evidence has demonstrated that continued expression of the fetal c-globin genes (fetal hemoglobin, HbF), which are normally silenced after birth, is the best treatment for SCD, since the pathophysiology is largely ameliorated. Our therapeutic goal is to reactivate the c-globin genes to substitute for the defective adult b-globin gene. We identified a novel cglobin gene silencer sequence and demonstrated that a GATA-1-FOG-1-Mi2 repressor complex binds to this sequence and silences c-globin synthesis. However, data regarding the requirement of Mi2 for silencing is controversial. We demonstrate that c-globin synthesis increases as Mi2 expression decreases. We also show that repressor complex components assemble sequentially during development; completion of assembly coincides with c-globin gene silencing. Disruption of either the repressor complex or mutation of its binding site induces c-globin. Understanding this mechanism will reveal potential new targets for treating SCD.
YAC BMCs relative to CID-dependent wild-type b-YAC BMCs ( Figure 3B). Together, these data confirm the role of Mi2 as an essential component of the c-globin silencing complex.
Temporal repression of c-globin by sequential recruitment of GATA-1, FOG-1, and Mi2 to the A c-globin 2566 GATA silencer We previously demonstrated that GATA-1 was recruited to the A c-globin 2566 GATA silencer by day E18 in fetal liver from wild-type b-YAC transgenic mice, a developmental time point at which c-globin is no longer expressed [1]. GATA-1 was not present at this site at day E12, when c-globin is at its highest expression level in the fetal liver. However, we did not examine recruitment during the intervening days. Thus, the assembly of the GATA-1 repressor complex at the 2566 silencer region might occur in a sequential manner, with each component recruited in a temporal fashion between days E12 and E18 with GATA-1 recruitment coinciding with the onset of c-globin gene silencing. To test this hypothesis, chromatin immunoprecipitation (ChIP) analyses were performed using wild-type b-YAC transgenic mouse staged fetal liver samples from days E12 to E18. Our data demonstrated that GATA-1, FOG-1 and Mi2 proteins do not occupy the 2566 GATA silencer until day E16 and E17 ( Figure 4A-4C). Although no recruitment was demonstrated until day E16 when silencing begins, we observed a temporal recruitment of the previously identified repressor components. GATA-1 alone occupied the 2566 GATA silencer at day E16 ( Figure 4A), but FOG-1 or Mi2 occupancy was not observed until day E17 ( Figure 4B and 4C, respectively). The complete GATA-1/FOG-1/Mi2 protein complex was observed at day E18 as previously demonstrated [1].
GATA-2 occupies the 2566 A c-globin silencer prior to repression by GATA-1/FOG-1/Mi2 GATA-1 and GATA-2 are reciprocally expressed during erythropoiesis, with GATA-1 levels rising when GATA-2 levels decline [17,18]. GATA-1 and GATA-2 share a common WGATAR DNA motif, present at cis-regulatory elements that activate transcription in an erythroid cell-specific manner [18]. These data prompted us to investigate whether GATA-2 was bound to the 2566 GATA silencer prior to GATA-1-mediated repression, even though GATA-2 is thought to not play a role in globin gene switching once the erythroid lineage has been established [3]. ChIP experiments were performed using day E12 and E18 fetal liver samples from wild-type b-YAC mice, where we previously demonstrated the absence (day E12) and presence (day E18) of GATA-1 recruitment. GATA-2 occupancy was observed in day E12 samples from the wild-type b-YAC transgenic mice ( Figure 5A). Occupancy at the Gata-2-2.8 Kb region by GATA-2, a positive control, was observed in day E12 samples ( Figure 5B), but not in day E18 samples from these mice (data not shown). This control is consistent with previous data where GATA-2 was demonstrated to bind the 22.8 Kb region of the Gata-2 locus when the locus is transcriptionally active, but is replaced by GATA-1 to initiate repression [19,20].
Taken together, our results support a model of temporal repression, in which GATA-2 first occupies the 2566 A c-globin silencer at day E12, followed by GATA-1 occupancy at day E16 and FOG-1 and Mi2 at day E17. The c-globin silencing might be initiated by the change in the GATA factor occupancy at the 2566 GATA motif, suggesting that GATA switches may play a role as a determinant of the onset of temporal repression by GATA-1 at the 2566 silencer region. To definitively prove that a HPFH mutation identified by us and another group [1,2] had the expected phenotype, we introduced the T.G mutation at position 2566 relative to the A c-globin mRNA start site into the normally located copy of the A c-globin gene in the b-YAC and produced transgenic mice. The GATA to GAGA alteration (and the absence of others) was confirmed by DNA sequence analysis of a PCR product amplified from the promoter region of the resultant YAC. Three 2566 T.G A c-globin HPFH b-YAC transgenic lines were obtained (lines 18, 20 and 25). Structural analysis was performed using radioactively-labeled DNA probes spanning the locus from 59HS3 through the HPFH6 breakpoint on Southern blots of pulsedfield gels to confirm integrity of the b-globin transgene loci and copy numbers were determined as described in Materials and Methods (data not shown). Only line 20 was suitable for further analysis. The 2566 T.G A c-globin HPFH mutation maintains cglobin expression in adult definitive erythropoiesis To test whether the 2566 T.G point mutation reproduced a human HPFH phenotype and maintained c-globin expression in the adult YAC transgenic mice, human b-like globin gene expression was measured by qRT-PCR in blood from F 2 or F 3 generation adult mice. Mouse a-globin and Gapdh served as internal controls to quantitate human b-like globin transgene expression levels. All values were normalized to these internal controls and corrected for transgene and endogenous gene copy number. Overall, the average of line 20 animals showed a 20-fold increase of c-globin expression (P,0.05; Figure 6A) and a 1.5-fold increase of b-globin expression, but this increase was not statistically significant ( Figure 6B). The variance of both c-globin and b-globin gene expression observed among different animals from the same lines and between lines suggests that position effect variegation is operative in 2566 A c-globin HPFH. However, these results clearly demonstrate that c-globin gene expression is increased during adult definitive erythropoiesis when the 2566 HPFH mutation is present. The increase is small compared to the 2117 G.A A c-globin Greek HPFH, in which c-globin transcription is induced 300-fold (unpublished data) [21,22].
We also determined the ratio of human c-globin protein chains to total human b-like globin protein chains (c-globin/(c-globin+bglobin) by reversed-phase high-performance liquid chromatography (RP-HPLC) in adult blood hemolysates from 2566 A c-globin HPFH b-YAC line 20 mice compared to wild type b-YAC and the 2117 Greek HPFH b-YAC transgenic mice ( Table 1). The 2566 A c-globin HPFH mice showed a small, but significant increase in c-globin chain expression (7.5%) compared to wild-type b-YAC mice (5.1%), but less than that measured in 2117 Greek HPFH mice (9.5%). These data corroborate the qRT-PCR data.
Increased levels of c-globin expression (F cells) were also demonstrated by flow cytometry analysis ( Figure 6C-6F). The 2566 A c-globin HPFH b-YAC mice showed a 23.8% and 20.5% increase of F cells ( Figure 6E and 6F) compared to a wild-type b-YAC transgenic control (3.4% F cells; Figure 6C) and the positive control, the previously characterized 2117 Greek HPFH b-YAC mice (26.2% F cells; Figure 6D). Immunostaining of 2566 A cglobin HPFH b-YAC line 20 peripheral blood cytospins demonstrated a heterocellular distribution of F cells in this line ( Figure 6I), compared to a pancellular distribution in 2117 Greek HPFH b-YAC mice ( Figure 6H); [21,22]. Although only one representative microscope field is shown in each panel of Figure 6G-I, the number of positively stained cells was approximately 6-fold higher compared to wild-type b-YAC transgenic mice ( Figure 6G). The modest increase of c-globin expression associated with the 2566 HPFH mutation should be therapeutic for sickle cell patients [23,24].

Disruption of GATA-1 mediated silencing by the 2566 A c-globin HPFH mutation
To validate our hypothesis that the 2566 A c-globin HPFH mutation reactivates c-globin gene expression during adult erythropoiesis by preventing the recruitment of the GATA-1/ FOG-1/Mi2 repressor complex, ChIP experiments were carried out on day E18 fetal liver samples from our 2566 A c-globin HPFH b-YAC transgenic line 20. Matched samples from wildtype b-YAC mice were employed as a control, where we previously demonstrated recruitment of the GATA-1/FOG-1/ Mi2 repressor complex at this developmental stage [1]. These proteins were not recruited to the 2566 GATA silencer region in 2566 A c-globin HPFH b-YAC transgenic mice in contrast to wild-type b-YAC transgenic mice ( Figure 7A). A 6-fold average increase of c-globin transcription was observed in the E18 blood samples from two 2566 A c-globin HPFH b-YAC transgenic animals ( Figure 7B). However, no significant increase was detected in E16 blood samples from three 566 A c-globin HPFH b-YAC animals. Thus, the 2566 HPFH mutation prevents recruitment of the GATA-1-mediated repressor complex and reactivates c-globin gene expression.

Discussion
Our studies provide evidence that a temporal mechanism of cglobin gene silencing is operative at the 2566 A c-globin GATA motif. GATA-1 is recruited first, at day E16, followed by the recruitment of FOG-1 and Mi2 at day E17, indicating that assembly of the GATA-1-FOG-1-Mi2 repressor complex occurs sequentially over a 24 hour period. The binding of the GATA-1 repressor complex might change the ''transcription-ready'' state to a more permanently silenced state by altering the chromatin into a heterochromatic state, preventing c-globin gene transcription (temporal repression model; Figure 8).
Our data also demonstrate that the 2566 GATA motif is occupied by GATA-2 early in fetal definitive erythropoiesis (day E12), followed by a change to GATA-1 occupancy at day E16, suggesting that GATA factor occupancy switching may play a role in the silencing of c-globin expression. GATA-2 is crucial for the maintenance and proliferation of immature hematopoietic progenitors, whereas GATA-1 is essential for the survival of erythroid progenitors and for the terminal differentiation of erythroid cells [3]. Changes in global gene expression patterns during hemoglobin switching are accompanied by changes in the expression of GATA-2 and GATA-1 (GATA switching), which in part coordinates cellular maturation [3,18]. These changes in GATA factor occupancy, combined with changes in the transcriptional factor milieu as maturation proceeds, may contribute to transcriptional repression and negative chromatin remodeling.
As human erythroid development proceeds, the proper b-like globin genes are activated or repressed, giving rise to the different hemoglobin chains expressed throughout development. Fetal hemoglobin (c-globin) is silenced shortly after birth, and the adult hemoglobins (b-and d-globin) are activated reciprocally. However, the c-globin genes remain in a ''transcription-ready'' state, since they can be reactivated following inducing treatments such as hydroxyurea or 5-azacytidine, or by naturally occurring HPFH mutations. It is possible that the loss of GATA-2 occupancy after day E12 at the 2566 A c-globin GATA site ( Figure 8A) results in the simultaneous loss of transcriptional co-activators associated with GATA-2, dictating the initial event in the onset of c-globin silencing ( Figure 8A-8B). Thus, the change in GATA occupancy, from GATA-2 during early fetal definitive erythropoiesis to GATA-1 at late fetal definitive erythropoiesis observed at this site may be orchestrated by an alteration in the nearby chromatin, post-translational modification of proteins and/or changes in the transcription co-factors available in the neighborhood ( Figure 8A-8C).
The demonstration of co-localization of GATA-1, FOG-1, and Mi2 by ChIP does not prove interaction between those proteins. Since we are analyzing a small region in the more distal promoter region of the A c-globin gene, it is possible that these proteins are associated with other complexes in the neighborhood, but still detected by ChIP due to the cross-linking step and size of the fragments after sonication. Hence, we do not exclude the hypothesis that other transcription factors and cofactors are recruited to nearby sites and contribute additively to silencing.
Factors such as BCL11A, the orphan nuclear receptors TR2 and TR4, NF-3/COUP-TFII and Ikaros have been associated with cglobin silencing [25][26][27][28]. More recently, Ikaros was shown to interact with GATA-1, since a lack of Ikaros reduced GATA-1 binding at the c-globin promoter and delayed c-globin gene silencing [29]. This study demonstrated that Ikaros functioned in Overall, the data presented in this study provide clear evidence of the involvement of GATA-1 and Mi2 in silencing c-globin gene expression. In a recent study, Miccio and Blobel [10] used mutant mice expressing an altered FOG-1 that abrogated NuRD binding. The authors demonstrated that the FOG-1/NuRD interaction is dispensable for silencing c-globin expression, but is required for FOG-1-dependent activation of human adult globin expression [10]. These data do not discriminate whether these proteins directly interact to form a mega-complex, with repressive and activator protein partners, or if a sub-population of the proteins interacts to form a distinct repressor complex and another subpopulation interacts to form a distinct activator complex. A deficiency of Ikaros reduced GATA-1 binding at the A c-globin promoter, enhanced chromosomal proximity between the LCR and the A c-globin promoter and delayed c-globin silencing. An Ikaros-related consensus binding sequence is found at the 2566 position of the A c-globin gene [29], thus it is provocative to suggest that Mi2 associates with Ikaros and GATA-1 to form a fetal cglobin repressor complex that also contains FOG-1 ( Figure 8D). However, GATA-1-FOG-1 may interact with a different NuRD component, such as MTA1, and perhaps other NuRD subunits, to form an adult b-globin activator complex [10]. A significant reduction of adult-type human and murine b-like globin gene expression was observed in the bone marrow of adult b-YAC transgenic mice when the FOG-1/NuRD interaction was disrupted, suggesting that NuRD is required for FOG-1-dependent activation of adult globin gene expression [9,10]. Bowen et al. suggested that the Mi2/NuRD complex is, in fact, a set of distinct complexes with similar biochemical properties [11]. The existence of different NuRD complex sub-types could explain the distinct roles and functions of the NuRD complex in globin regulation. One sub-type complex might be associated with activation of the adult b-globin gene and another sub-type, with shared, but also unique subunits, might be associated with repression of c-globin  [30], which supports our data showing that Mi2 is required for c-globin silencing.
Finally, our studies also show that maintenance of c-globin expression observed with the 2566 A c-globin HPFH point mutation resulted from the disruption of GATA-1-FOG-1-Mi2mediated repression ( Figure 8E). This finding was corroborated by the increased expression of c-globin in the Mi2b conditional knockout lines ( Figure 8E). Although the HPFH phenotype produced by the 2566 A c-globin HPFH point mutation was weak, it was still at a level therapeutic for the treatment of hemoglobinopathies [23]. Heterocellular HPFH represents approximately 10% of the F cell trait population, with HbF levels between 0.8 and 5% [23]. The modest levels of c-globin produced by the 2566 A cglobin HPFH might be characteristic of a heterocellular HPFH, as demonstrated by cytospin preparations of RBCs ( Figure 6G-6I). In contrast, the Mi2b conditional knockout resulted in a pancellular HPFH ( Figure 2O-2P). The Mi2b knockout has a broader effect within RBCs than the cis-linked 2566 A c-globin HPFH mutation; the loss of Mi2b may generally affect a number of c-globin repressive mechanisms, leading to a pancellular F cell distribution, whereas the 2566 mutation variably affects binding of a single cglobin repressor complex, producing a heterocellular distribution. Data from HFPH patients bearing a mutation at the 2567 G cglobin GATA motif also suggested variance in the levels of HbF caused by the point mutation. Chen et al. [2] demonstrated that the father and his 9-year-old son had moderately elevated Hb F at 10.2% and 5.9%, respectively [2]. The variance in the levels of cglobin observed between different 2566 A c-globin HPFH b-YAC transgenic animals from individual lines suggests position effect variegation (PEV) is operative. Bottardi et al. [29] demonstrated that interaction between the LCR and the A c-globin gene is reduced by binding of Ikaros to the A c-globin promoter at the time of the cto b-globin switch. Thus, the chromatin organization of the c-globin promoter might be essential to maintain the long-range interaction with the LCR. The presence of the 2566 point mutation may prevent the promoter from fully interacting with the LCR, blocking full engagement with the LCR necessary for complete transcriptional activation, resulting in PEV.
In conclusion, our study is the first to demonstrate the temporal assembly of a GATA-1 repressor complex in vivo. We also demonstrated that the temporal repression mechanism is disrupted by a Mi2b mutation or a HPFH mutation, alleviating the stagespecific silencing of the A c-globin gene by the GATA-1-FOG-1-Mi2 repressor complex. This mechanism potentially provides a new target for treatment of sickle cell disease and other hemoglobinopathies.

A c-globin HPFH b-YAC construct
A 213 Kb yeast artificial chromosome carrying the human bglobin locus with the T.G A c-globin HPFH point mutation was synthesized as follows, using previously described methods [31]. Briefly, a marked A c-globin gene ( A c m ) contained as a 5.4 Kb SspI fragment (GenBank file U01317 coordinates 38,683-44,077) in the yeast-integrating plasmid (YIP) pRS406 [29] was mutagenized using the Quick Change Site-Specific Mutagenesis Kit (Stratagene, La Jolla, CA). The presence of the 2566 point mutation was confirmed by DNA sequencing and the mutation was introduced into the b-YAC by ''pop-in'', ''pop-out'' homologous recombination in yeast [1]. The mark in the A c m -globin gene is a six-base pair deletion at +21 to +26 relative to the A c-globin translation start site allowing preliminary discrimination of the modified b-YAC from the wild-type b-YAC by restriction enzyme digestion following homologous recombination. The presence of the mutation in clones passing this test was confirmed by DNA sequence analysis of a PCRamplified fragment encompassing the mutated region. YAC transformation, screening of positive clones, purification, and mouse transgenesis were performed as described previously [1].

Structural analysis
Transgene and copy number structural analyses of F 2 generation animals were performed by standard PCR, Southern blot analyses [1] and quantitative real-time PCR (qPCR) [32,33]. Initially, structural analysis was performed by a PCR-based approach to confirm the presence of the LCR 59HS3, e-, cand b-globin genes in the 2566 HPFH b-YAC transgenics (data not shown). Further structural studies were performed by Southern blot hybridization of pulsed-field gels [1]. The primer and probe sequences used were as described previously [21,34]. The transgene copy number was established by qPCR, using the standard curve method [32,35], comparing dilutions from the 2566 HPFH b-YAC mice to samples from our wild-type b-YAC mice line 26223, which has a well characterized copy number [31,36]. Values were normalized to the murine a-globin and Gapdh genes.

Mi2b conditional knockout b-YAC mice
Generation of the floxed Mi2b mice and the erythroid-specific m'LCR-b promoter (pr)-Cre recombinase transgenic mice was described previously [13,14,37]. These mice were crossed to obtain m'LCR-b pr-Cre, floxed -Mi2b/Mi2b + heterozygotes, which in turn were crossed with homozygous floxed Mi2b b-YAC transgenic mice to produce mice bearing an erythroid-specific Mi2b knockout and a

ChIP assay
ChIP assays were performed as described with some modifications [1]. Fetal livers from wild-type b-YAC transgenic mice at post-conception days E12-E18 were utilized. Fetal livers from 2566 HPFH b-YAC transgenic mice at post-conception days E12 and E18 were employed as controls. Cross-linking was performed using a two-step dual cross-linking method [38]. Cells were incubated for 30 minutes with 1.5 mM ethylene glycol bis[succinimidylsuccinate] (EGS), followed by 1% formaldehyde (fresh paraformaldehyde) for 10 minutes at room temperature. Chromatin was sonicated to a size range between 200 and 1,000 bp. The samples were pre-cleared with species-matched normal serum. Immunoprecipitations (IPs) were carried out with anti-GATA-1, anti-GATA-2, anti-FOG-1 or anti-Mi2 specific antibodies or isotype-matched IgG (rabbit, mouse or goat) and protein G conjugated to magnetic beads (Invitrogen Dynal, AS, Oslo, Norway). The immunoprecipitate was washed, the crosslinks were reversed and the genomic DNA was purified. Recruitment of GATA-1, GATA-2, FOG-1 and Mi2 proteins was measured by real-time qPCR, using gene specific primers as described previously [1]. The antibodies used were rat anti-GATA-1

Real-time quantitative PCR and RT-PCR (qPCR and qRT-PCR)
ChIP samples were analyzed in duplicate by real-time qPCR with SYBR Green dye using a MiniOpticon or CFX96 systems (Bio-Rad, Hercules, CA). To allow comparison among primer sets, input samples from each condition were diluted serially from 1:10 to 1:10,000 and used as standards for all PCR samples. Enrichment of protein binding to a specific DNA sequence was calculated using the standard curve method [32]. PCR primer sequences were as previously described [1] and additional primer sequences are listed in Table 2. ChIP experiments were performed using duplicate samples and each qPCR experiment was performed two to four times for each sample set. Murine GAPDH and a-globin genes were used as internal controls for the expression data. Data is shown as the mean 6 the standard deviation of the mean. The Student's t-test was used to determine statistical significance at P,0.05 and P,0.01.
Globin and Mi2b (Chd4) gene expression was measured by realtime qRT-PCR using the relative quantification, as previously described; primers sequences are listed in Table 2 [1,33,39].

HbF detection by flow cytometry
Detection of HbF (F cells) and Mi2 was performed by flow cytometric analysis [40]. Briefly, mouse blood was collected from the tail vein in heparinized capillary tubes. Ten ml of whole blood was washed in PBS and fixed in 1 ml 4% fresh paraformaldehyde (Sigma Aldrich, Saint Louis, MO). The cells were centrifuged, the supernatant discarded and the pellets were resuspended in 1 ml icecold acetone: methanol (4:1) for 1 minute. Cells were washed twice in ice-cold PBS/0.1% BSA and resuspended in 800 ml of PBS/0.1% BSA/0.01% Tween 20 (PBT). One mg sheep anti-human hemoglobin F FITC-conjugated antibody (A80-136F Bethyl laboratories, Montgomery, TX) or anti-Mi2 (sc-11378, Santa Cruz Biotechnology, Santa Cruz, CA) was added to 100 ml of the cell suspension and incubated for 40 minutes at room temperature. Cells were washed with 1 ml ice-cold PBS/0.1% BSA and the pellets were resuspended in 100 ml of PBT. 100 ml Alexa 488 (Invitrogen, Molecular Probes)conjugated secondary goat anti-rabbit antibody diluted 1:200 in PBT was added to the cell suspension as secondary antibody to the anti-Mi2 antibody, and incubated at room temperature for 20 minutes, in the dark. Cells were washed with 1 ml ice-cold PBS/0.1% BSA and the pellets were resuspended in 200 ml of PBS [41,42]. Cells were analyzed using a BD LSRII (BD Biosciences, San Jose, CA) with an emission filter 530/30 nm (FITC/GFP). Data from 30,000 events was acquired for analysis using BD FACSDiva software (BD Biosciences, San Jose, CA).

HbF detection by cytospin preparation
Ten ml of anti-human hemoglobin F FITC-conjugated antibody-stained cells were added to 190 ml of PBS/0.1% BSA, the liquid was placed on slides, which were spun down in a cytocentrifuge at 700 rpm for 3 minutes. Cytospin images were acquired with a Leica DM5000 B microscope outfitted with a Leica DC500 digital camera. The Leica DC500 software runs through the Adobe Photoshop platform.

Western blot analysis
Chemical inducer of dimerization (CID)-dependent wild-type b-YAC bone marrow cell [16] and CID-dependent floxed Mi2b Cre b-YAC bone marrow cell lysates were prepared as described [1]. Protein concentrations were measured spectrophotometrically using the Bradford assay. Forty mg cellular lysate was mixed with loading dye (50 mM Tris, pH 6.8, 100 mM DTT, 2% SDS, 0.1% bromophenol blue, 10% glycerol) and heated at 95uC for 5 minutes, followed by separation in a 10% SDS-polyacrylamide gel using Tris-glycine buffer. Western blotting was performed as previously described [1].
Reversed-phase high-performance liquid chromatography (RP-HPLC) protocol b-like globin protein chains were separated by RP-HPLC. Hemolysates were prepared from packed red cells by freezethawing in water. Briefly, half capillary tubes of blood were collected (30-40 ml) and mixed with 2 ml 50 mM EDTA. The samples were washed three times with 0.9% NaCl. The RBCs were finally resuspended in 200 ml of water, vortexed for 10 seconds, centrifuged at for 20 min at 4uC to pellet debris, and the supernatant was transferred to a fresh tube. Hemoglobin concentration was determined by adding 5 ml lysate to 995 ml Drabkin's reagent, measuring the OD 540 and multiplying by 285.7. The sample was then diluted to 2 mg/ml in buffer A (20% acetonitrile, 0.1% TFA) and filtered through a 0.2 mm PES syringe filter. 400 mg samples were run through a Vydac large-pore C4 column (214TP54) on a Waters 600S Controller and 996 Photodiode Array Detector. Buffers used consisted of buffer A and buffer B (60% acetonitrile, 0.1% TFA). The gradient was 44 to 60% buffer B over an hour [43]. Quantitation of the human globins was performed using Empower 2 software. Seven to 12 individual samples were run for each transgenic mouse line.

Statement of ethical approval
The animal studies were performed in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The protocol was approved by the Institutional Animal Care and Use Committee (IACUC) of the University of Kansas Medical Center (Protocol ID Number: 2012-2060; approved 06/20/ 12).