The system for in vitro expansion and differentiation of human CD34+ cells described in this study has several advantages including the generation of sufficient numbers of untransformed cells for various biochemical studies, and the relatively synchronous nature of the differentiation that makes it possible to determine the sequence of events during very early to late stages of erythroid differentiation ( & and sFig 1). The serum free culture conditions described in our protocol provides good control over the growth and differentiation of CD34+ cells and eliminates unknown effects associated with serum and co-culture based cultures. Induction of erythropoiesis of the expanded CD34+ cells converts more than 85% of the cells into GYPA+ cells by day-5 of Epo treatment, and these cells progressively differentiate in a relatively synchronous manner to produce large number of cells at different stages of erythropoiesis ( and sFig 1). One caveat is that the conditions for differentiation are not fully physiologic, both with respect to the types of cytokines and cell to cell signaling, and environmental differences such as ambient rather than physiologic oxygen concentrations. Nevertheless, our data on the chromatin structure and transcription factor recruitment at the α–globin locus and the pattern of appearance of key erythroid transcription factors such as GATA1 and NF-E2 during erythropoiesis suggest that the system reflects many aspects of normal in vivo differentiation.
Although the expanded CD34+
cells express low levels of glycophorin-A until Day-11 in culture, they lack erythroid properties as evidenced by the silent α-globin locus and very low levels of GATA-1 and absence of other key erythroid specific genes such as β-globin and ALAS2 (). These are still uncommitted multipotent cells that can be differentiated into either erythroid ( & ) or myeloid lineages (data not shown). The order of peak appearance of the α and β like globin gene mRNAs during erythroid development of expanded CD34+
cells in our experiments follows the same temporal pattern as does their appearance during ontogeny (). Previously, expression of γ globin genes in erythroid precursor cells has been reported that is consistent with our observation in this study[18
]. However, we see early appearance of Mu globin, which peaks at day 2 of Epo treatments. This timing of Mu expression is different from the previous report wherein it was observed to peak at day-10 of Epo induction[43
]. In addition, we also detected unusually high amounts of θ globin message that persists at late stages of erythropoiesis ( and sFig 3). Earlier studies have reported that θ globin is expressed in fetal/adult stage and its expression is about 50 fold less than alpha message in the fetal erythroid stage[44
]. However, in bone marrow cells of patients with thalassemia and sickle cell disease, significantly higher amounts of θ-globin transcripts were detected[45
]. This increase was attributed to the accelerated erythropoiesis in the bone marrow of these patients. In this in vitro
cell culture system, cells are stimulated with non-physiological doses of cytokines at ambient oxygen levels, which may mimic accelerated erythropoiesis under stress, leading to the high levels of θ-globin transcript (). Further insights into possible functions of the θ-gene must await identification and characterization of its presumptive protein product.
Our observation of the absence of NF-E2 in very early stages of induced erythropoiesis and its subsequent steep induction until late stages () is consistent with the earlier studies[18
]. However, we detected significant levels of NF-E2 protein in uncommitted, multipotent, expanded CD34+
cells in culture. Interestingly, upon Epo treatment NF-E2 disappears initially to again reappear from day-3 onwards. Further studies are needed to understand the role of down regulation of NF-E2 in early erythropoiesis. ChIP experiments show recruitment of NF-E2 prior to erythropoiesis and at day-7 of epo treatment that suggest role of NF-E2 in α-globin gene regulation in erythroid as well as in non-erythroid cells (). Previously, presence of NF-E2 was demonstrated at the HS-26 sequence of the α–globin locus and at the HS2 sequence of the β-globin locus in murine FDCP-mix cells and NF-E2 expressing HeLa cells, respectively[31
]. Forced expression of NF-E2 in HeLa cells was observed to generate DNase hypersensitive regions at the β-globin HS2 sequences that suggest the role of NF-E2 in chromatin modification[49
]. Indeed, in erythroid cells NF-E2 was shown to mediate trimethylation of H3K4 by its association with MLL2 methyltransferase[50
]. The prior presence of NF-E2 at the α–globin locus in expanded CD34+
cells and its continued presence at the same sites, albeit at different levels, suggests that NF-E2 may keep the α-globin enhancer in a poised state prior to erythroid differentiation. Further studies are needed to understand the role of NF-E2 in chromatin modification and α-globin transcription before and after the onset of erythroid differentiation of CD34+
cells. Identification of all the NF-E2 interacting proteins in expanded CD34+
cells and their erythroid derivatives will be helpful in understanding the differential role of NF-E2 at the α-globin locus. Curiously, we have seen high levels of NF-E2 recruitment at the transcriptionally inactive Mu globin promoter and absence of detectable NF-E2 at ζ, α and θ promoter regions.
Basal levels of GATA-1 protein in expanded CD34+ cells and their exponential increase from day-4 of Epo treatment co-insides with the lack of GATA-1 binding at the α-globin locus prior to onset of erythropoiesis and its subsequent binding at day-7 of Epo treatment ( & ). Association of GATA-1 at HS-10 and HS-46 in erythroid cells is consistent with earlier observations in murine CFUe cells[31
]. However, in humanized mouse erythroid cells GATA-1 was shown to be present primarily at HS-40 [14
]. Interestingly, like NF-E2, significant level of GATA-1 was present at the transcriptionally inactive Mu globin promoter (). It remains to be seen if the presence of GATA-1 and NF-E2 at Mu globin promoter has any influence on the activity of surrounding α–like globin genes.
The initiating form of Pol II (recognized by monoclonal antibody 8WG16) is associated with transcriptionally active as well as poised promoters[42
]. The transcriptionally silent α-globin locus prior to erythropoietin seems to be devoid of Pol II binding (). At the onset of erythropoiesis, as early as day-2, when GATA-1 and NF-E2 are not at detectable levels and CTCF is stripped off the locus, Pol II appears on the HS-40, then spreads to other regions at the later stages of erythropoiesis (). Thus HS-40 seems to be the earliest recruitment site for Pol II. At Day-2 of Epo treatment we find a mixture of CD71+
cells (sFig 1). It will be interesting to find out which of these cell populations show the appearance of Pol II at HS-40. The presence of the initiating form of Pol II at the enhancer region () and of phospho-S5 Pol II at the α-globin promoter (sFig 2) suggests that the α-globin enhancer may be a nucleus for formation of the initiation complex. In addition, presence of Pol II-phospho S2 at the HS-46 region (sFig 2) of the enhancer suggests the participation of the α-globin enhancer in the elongation of transcription also. Similarly, a transcription elongation function was suggested for the β-globin LCR[51
]. However, De Gobbi et al and Anguita et al found Pol II prominently at the α-globin promoters in the peripheral blood mononuclear cell derived erythroid cells and murine TER119+ cells using antibodies against the N-terminal region of the large subunit of the Pol II[14
]. The antibodies raised against the N-terminal region of the Pol II would not distinguish between the initiating and elongating forms. Thus, the discrepancy in the Pol II binding could be related to the cell type, or choice of antibody used. Interestingly, we detected high levels of Pol II at the θ globin promoter in erythroid cells, which may reflect a difference in the regulation of α and θ promoters.
CTCF is a multifunctional protein that associates with insulators, causes looping of the chromatin, regulates noncoding RNA transcription and establishes local chromatin structure at the repetitive elements in the mammalian genome[52
]. CTCF is also reported to activate transcription by directly interacting with large subunit of the Pol II complex[40
]. Many binding sites for CTCF are constitutive, and the sites found in the α–globin cluster of undifferentiated CD34+
cells closely resemble those seen, for example, in non-erythroid HeLa cells (data not shown). The pattern of CTCF recruitment at the 5′ and 3′ ends of the α-globin locus and its subsequent rearrangement during erythropoiesis suggests dual roles of CTCF as an insulator prior to erythropoiesis and as a facilitator of transcription in erythroid cells (). Prior to erythropoiesis, CTCF may insulate the α-globin locus from the activity of surrounding genes. Upon erythropoiesis, the insulator CTCF is stripped off of the α-globin locus and is rearranged on the erythroid specific HS-33 site, perhaps acting as a positive transcription factor or changing the looping structure of the local chromatin. Further detailed studies are needed to understand the roles of CTCF as an insulator and facilitator of α-globin gene transcription during erythroid development.
Dimethylation of H3K4 (H3K4Me2) is broadly associated with transcriptionally active genes as wells as their enhancers[53
]. The H3K4Me2 modification pattern described in this study () is consistent with the earlier observations[14
]. However, our result for the H3K9Ac modification pattern at the α-globin locus differs from the earlier reports that show H3 histone H3 acetylation predominantly at the α-globin genes[14
]. This discrepancy could be due to the choice of antibodies. The antibody used in the pervious studies recognizes acetylation of histone H3 at K9 as well as K14 positions, where as in this study we have used the antibody that is specific to the H3K9Ac modification (). Although doubly acetylated histone H3 at K9 and K14 [53
] as well as the acetylation at the single K9 site[55
] are generally associated with the transcriptionally active promoters, not all transcriptionally active genes have H3K9Ac modification[56
]. This is in conformity with lack of this modification at transcriptionally active α-globin promoters in the erythroid derivatives of CD34+
cells (). Instead, we detected H3K9 specific acetylation at HS-33, HS-46 and HS-48 enhancer sites and elevated levels of H3K9Me3 at the α–like gene promoters (). These site-specific H3K9 acetylations and methylations may be a way to modulate the correct recruitment of transcription factors ( & ) and promote enhancer-promoter interactions. We also observe high levels of H3K9Me3 modification at the transcriptionally inactive Mu globin promoter, consistent with the general association of this modification with inactive genes[55
]. Overall, our data shows that when CD34+
cells are differentiated into erythroid lineage, the α-like globin promoters are trimethylated at H3K9 and the H3K9Ac modification is associated with upstream enhancer sequences. The extent of H3K9Me3 modification may vary between transcriptionally active and inactive genes as in case of α–globin and Mu globin gene promoters.
Taken together, our results indicate that there is a prominent change in the status of the upstream activator elements and less at the promoters of the α–globin genes themselves during the stages of erythroid differentiation that we have studied. Remodeling of the upstream elements may be the primary event in activation of α–globin gene expression at endogenous loci. Activation of α-globin genes upon Epo treatment involves initial binding of Pol II and removal of pre-existing factors like NF-E2 and CTCF, then rearrangements of CTCF and concurrent or subsequent binding of transcription factors like GATA-1 and NF-E2. The ζ-Globin gene was conspicuously devoid of chromatin modifications and transcription factor recruitment. A unique set of chromatin remodeling and transcription factor binding may be playing the role of inhibition of ζ-globin transcription in adult erythroid cells.
Perhaps the most unexpected result in the present studies was the finding that CTCF disappears from the α–globin locus after stimulation by erythropoietin, without a significant change in total cellular CTCF levels or in binding at a non-erythroid site (). Subsequently, with the progression of erythropoiesis, CTCF is rebound to these regions with a different distribution than seen in non-erythroid cells. This would be consistent with a rearrangement of higher order chromatin structure in this region early in erythropoiesis.