Integrating Structural and Functional Studies Leads to a New Model of β-Globin Activation That Suggests Distinct Initiation and Maintenance States

Abstract 349

The human and mouse β-globin loci share a conserved structure in which the locus control region (LCR) and genes are flanked by three CTCF bound DNaseI hypersensitive sites (HSs); 3‘HS1 downstream, and 5‘HS5 and human and mouse orthologues HS-111 or HS-62 upstream. In mice HS-62 and 3‘HS1 delineate a DNase sensitive domain. During erythroid differentiation, high-level expression of the β-globin locus is associated with LCR-dependent re-localization of the locus from the nuclear periphery to the nucleoplasm, where it associates with foci of serine-phosphorylated PolII deemed transcription factories (TFs).

To investigate the relationships among chromatin structure, nuclear localization and β-globin expression during human erythropoiesis, CD34 progenitor cells were differentiated and analyzed by ChIP-array, primary transcript FISH, immuno-FISH, and chromatin conformation capture, carbon copy (5C). Localization of the β-locus away from the nuclear periphery and to TFs, and detection of β-nascent transcripts are rare events at day 4 (proerythroblasts), whereas by day 15 (polychromatic erythroblasts), nearly all loci are centrally located, associated with TFs and actively expressing. Three megabase profiles of complementary active and repressive histone marks (H3 lysine 4 di-methylation (DiK4) and H3 lysine 27 tri-methylation) reveal that DiK4 is enriched in the LCR and adult genes in undifferentiated CD34 cells and nears maximal enrichment by day 4. Thus the chromatin landscape is set up prior to erythroid commitment and is increased at day 4, but shows little change with activation. These profiles also reveal a previously un-described 257kb domain spanning from HS-111 to +146 relative to the ε-gene cap, with CTCF bound at its boundaries.

5C analysis reveals a high linkage frequency between the LCR and β-gene at day 4, prior to β-gene activation. Thus proximity may be necessary, but is not sufficient for high-level expression. In addition, the LCR and adult genes have frequent contact with surrounding regions, but interactions are sharply demarcated by HS-111 and +146, linking the above histone modification domain and 5C structure. The flanking regions HS-111, 3‘HS1 and +146 associate with the LCR and genes in an active chromatin hub (ACH)-like structure. By combining 5C with the Integrated Modeling Platform, a high-resolution three-dimensional (3D) model of chromatin structure was generated and revealed that the CTCF containing flanking regions, HS-111, 5'HS5, 3‘HS1 and +146 are in proximity and anchor loops of the intervening regions. The LCR and β-gene lie in close proximity (<100nm) within a tight chromatin globule. At day 15, 5C-interactions become more restricted. Throughout the 1Mb assayed there is a global decrease in linkages and, unlike the ACH model, associations of the flanking HSs with each other diminish. In contrast, the LCR and β-gene are more highly linked. While the distance between the LCR and β-gene remains under 100nm at day 15, most of the remaining 3D structure is less compact. One notable exception is with differentiation the LCR is in closer proximity to the β-gene 3‘enhancer. In addition, β-gene activation is associated with an increase in the contour length of the region, possibly correlating with DNaseI accessibility.

In summary, our results reveal that acquisition of DiK4 precedes erythroid commitment. Enrichment of this active histone modification may occur in the nuclear periphery and is associated with a compact structure in which the flanking HSs, LCR and β-gene are in close proximity. Notably, close association of the LCR and β-gene precedes association with a TF and is not sufficient for expression. This suggests that if an ACH structure is important for function, its role may be limited to the initiation, but not the maintenance of gene expression. Alternatively, this compact structure may reflect the mafK-mediated recruitment of co-repressors to the LCR, as we demonstrated previously. The exchange of these repressors for activators could lead to the observed relocation from the periphery to TFs where high-level expression occurs and provides an explanation for the large change in expression that occurs despite a subtle change in the proximity of the LCR and β-gene. Preliminary studies on cell lines harboring intact, and LCR deficient chromosomes suggest the LCR may effect boundary element formation, domains of histone modifications and structure of the region.