Although the human β-globin LCR may function as a holocomplex within an active chromatin hub, we provide evidence that within the aggregate hypersensitive site activation domain of the holocomplex, the individual HSs mediate preferential activation of specific globin genes during development. As a model to test this hypothesis we chose to mutate LCR 5′HS3 in conjunction with a globin gene order alteration (replacement) within the context of a human β-globin locus YAC. Several previous papers clearly identified the phenotypes associated with 5′HS3 deletion mutant β-YAC transgenic mice (Supplementary Table S1
). A 2.3-kb complete deletion of 5′HS3 (Δ5′HS3) in β-YAC transgenics resulted in small decreases in ε- and γ-globin gene expression, with essentially normal β-globin gene expression (28
), a relatively mild expression phenotype. A 225-or 234-bp deletion of the 5′HS3 core (Δ5′HS3c) in β-YAC transgenics abrogated ε-globin gene expression during primitive erythropoiesis, without affecting γ-globin gene expression at this developmental stage (23
). During fetal liver definitive erythropoiesis, no to very low γ-globin expression was observed, but β-globin expression was unaffected. However, β-globin synthesis suffered from position effect variegation in the adult stage of definitive erythropoiesis. In spite of the similarity of the phenotypes between the large and small deletions, the severity was markedly amplified in the smaller core deletion mice. That is, the smaller 5′HS3c deletion was more catastrophic on gene expression than the larger 5′HS3 deletion. Addition of the −117 A
γ-globin Greek HPFH mutation to the 234 bp Δ5′HS3c β-YAC resulted in a phenotype indistinguishable from the 234 bp Δ5′HS3c transgenics (49
). Singly mutant Greek HPFH β-YAC mice were previously shown to exhibit a strong hereditary persistence of fetal hemoglobin (HPFH) phenotype in adult mice (50
). However, this HPFH point mutation had no effect on moderating the negative effect of the Δ5′HS3c mutation on γ-globin expression in transgenic mice. There was no γ-globin observed in adult mice even with the HPFH mutation, highlighting the importance of the 5′HS3 core sequences in regulating γ-globin expression. Finally, mutation of one of the seven GT motifs in the 225-bp 5′HS3 core, the GT6 motif, reduced the expression of the ε- and γ-globin genes during embryonic erythropoiesis (36
). γ-globin gene expression was significantly reduced during fetal definitive erythropoiesis, but β-globin gene expression was not affected. Thus, the 5′HS3 GT6 motif is required for normal ε- and γ-globin transcription in the yolk sac and for γ-globin transcription in the fetal liver. We concluded that mutation of a single transcriptional motif in the LCR can have profound effects on gene expression. Importantly, we observed a general conservation of phenotype associated with a Δ5′HS3, deletion of the core element only or mutation of a single transcriptional motif within the core.
The lack of ε-globin expression in the Δ5′HS3c β-YAC transgenic mice suggested that 5′HS3 sequences of the LCR are involved directly in ε-globin gene activation. This reduction of ε-globin gene transcription in Δ5′HS3 or Δ5′HS3c β-YAC transgenics can be explained by two hypotheses. The first hypothesis posits that within the LCR holocomplex or ACH, the major determinant of LCR-globin gene interaction is LCR HS site specificity; i.e. for each globin gene, a specific HS or subset of the HSs is required for gene activation and the others are dispensable. During embryonic erythropoiesis, the interaction between the LCR and the ε-globin gene promoter involves specific sequences of 5′HS3 and specific sequences of the ε-globin gene promoter. When 5′HS3 or its core is deleted, these interactions do not take place and ε-globin gene transcription is diminished (28
). 5′HS3 is not required for γ-globin transcription in the primitive yolk sac, but is necessary in the fetal definitive liver (30
). Thus, LCR5′HS3 shows specificity for activation of ε-globin during primitive erythropoiesis and for γ-globin during fetal definitive erythropoiesis.
The second hypothesis states that the conformation of the LCR in the ACH is the most important determinant of LCR–globin gene interaction. If this hypothesis is true, than in the embryonic stage, the LCR would be expected to adopt a three-dimensional conformation that favors interaction with the first gene in the complex, the ε-globin gene. Consistent with this hypothesis, following the first switch from ε-globin gene expression in the primitive yolk sac to γ-globin gene expression in the fetal liver, the LCR would be predicted to assume an alternate conformation favorable to γ-globin activation. When 5′HS3 is deleted, an alternate conformation is assumed that decreases the chance that there will be an interaction between the LCR and the ε-globin gene. However, in 5′ΔHS3c mice, the next genes in locus, the γ-globin genes, are expressed (30
). Thus, we assume that the LCR must still interact with the γ-globin genes during primitive erythropoiesis. Although γ-globin gene expression is normal during primitive erythropoiesis in these mutant mice, expression is extinguished during the fetal stage of definitive erythropoiesis, in contrast to mice carrying a normal β-YAC construct. These data suggest that a conformational change occurs in the Δ5′HS3c LCR during the switch from embryonic to definitive erythropoiesis, from one that supports γ-globin gene expression to one that does not (). Alternately, the embryonic trans
-acting environment may allow the mutant LCR to interact with and activate the γ-globin genes, but the fetal trans
-acting environment may not support this interaction in the absence of the 5′HS3 core.
Figure 9. Model for LCR 5′HS3 gene activation specificity. These illustrations emphasize the interaction of 5′HS3 with a specific globin gene at each developmental stage. Panels A–C represent the interaction of the intact wt LCR with the (more ...)
To distinguish between these two hypotheses, β-YAC lines were produced in which the ε-globin gene was replaced with a second marked β-globin gene (βm
), coupled to an intact LCR, a 2.9-kb 5′ΔHS3 or a 234-bp 5′ΔHS3c (). Δ5′HS3c Δε::βm
β-YAC mice expressed βm
-globin throughout development beginning at Day 10 in the yolk sac (A). γ-globin was expressed in the embryonic yolk sac, but at a reduced level in the fetal liver compared with transgenic lines containing an unmodified LCR (A and A; and ). Some wt β-globin was expressed in addition to βm
-globin in adult mice, but at much reduced level compared with βm
-globin (), which is probably due to the proximity of the βm
-globin to the LCR, as demonstrated previously (46
). The γ-globin phenotype is consistent with published data on Δ5′HS3c β-YAC mice (30
Although ε-globin was not expressed in Δ5′HS3c β-YAC mice, βm
-globin inserted in its place was expressed in Δ5′HS3c Δε::βm
β-YAC embryos, demonstrating that the 5′HS3 core was necessary for ε-globin expression during embryonic erythropoiesis, but not for βm
-globin expression, nor for γ-globin. In the 12-day fetal liver, expression of γ-globin was reduced in both LCR mutants compared with constructs with unmodified LCRs, but that result could be attributed to gene competition between the β- and γ-globin promoters for interaction with the LCR (51
). Interestingly, β-globin expression was higher in the Δ5′HS3c Δε::βm
β-YAC lines throughout development than with a full deletion of the 5′HS3 in the Δ5′HS3 Δε::βm
β-YAC construct, suggesting that the HS3 core deletion made the LCR even more permissive for β-globin expression (B). LCR HS structure has been implicated in directing specificity of globin gene expression by alteration of DNA conformation at the HSs (52
), which may explain the striking difference between the ε- and β-globin expression levels when the two genes are at the exact same location. The core deletion has an effect only on ε-globin expression in the yolk sac, whereas β-globin expression is largely unaffected by the core and entire 5′HS3 deletions.
If the LCR holocomplex conformation was the major determinant of LCR-mediated β-like globin gene activation, then the βm
-globin gene should not have been expressed in transgenic mice with a deletion of 5′HS3 or the 5′HS3 core, similar to previous data with 5′HS3 mutations (28
). We present evidence for a site specificity model of direct LCR HS–globin gene interaction with the confines of the holocomplex, where the holocomplex defines the three-dimensional structure of the open locus, but the specific LCR HS–gene interaction is the final determinant of temporal and spatial globin gene expression.
Our results, in combination with previous studies of human β-globin locus transgenic mice, contrast with similar analyses of the endogenous mouse β-globin locus, in which LCR 5′HSs were deleted (24
). For example, and relevant to our study, a 5′HS3 deletion reduced the overall expression of the locus, although it preferentially decreased βmin
-globin expression over βmaj
-globin expression (53
). HS site specificity for globin gene activation has not been observed in the murine locus; deletion of any of the individual HSs results in a phenotype similar to that for 5′HS3. This fundamental difference between the human and mouse loci is not readily explained, given that the HS core sequences of the human and mouse β-globin LCRs are highly conserved and the loci are similarly organized. Arguments against the human β-YAC transgenic data supporting the model presented above have been based on ectopic genome location of the transgene, coupled with possible resultant position effect variegation (PEV) of the transgene. When PEV was observed, a 150-kb β-YAC transgene was employed (55
), whereas we have used a 248-kb (now more accurately known to be 213 kb) β-YAC containing more extensive locus-flanking sequences. Our use of the larger β-YAC in this and many other studies, as well as by others, over the last two decades has obviated PEV as an explanation for our results (56
). We further minimize any chance for PEV by analyzing mice with complete locus copies that usually include extensive locus-flanking 5′- and 3′-sequences (Supplementary Figures S1–S3
) and using multiple lines per construct. In our experience, controlling these two variables, coupled with the larger size of the YAC transgene, we have avoided or overcome PEV associated with the smaller YAC transgene. Regardless of the LCR mutation, our mice have consistently shown site-independent, copy-number-dependent expression whether the LCR is intact, carries a HS deletion or bears a simple mutation in the HS (28,30,36,49, this study). Our data occasionally show an outlier, such as the Δε::βm
Δ5′HS3c line 61 on Day 10 yolk sac sample (). However, this line showed expression consistent with the other two lines at all other developmental time points and for all globin genes assessed. Thus, for this mutant β-YAC, PEV was not observed for any of the genes within the locus, nor at any developmental stage. In this report, an additional control was built in by virtue of the β- for ε-globin gene swap in the constructs. For example, Δ5′HS3 mutant LCR function is normal regarding γ-globin gene expression, but not ε-globin gene expression, during primitive erythropoiesis (28
) and the negative phenotype associated with ε-globin was reversed when β-globin was inserted in its place (this study). PEV does not account for the sum of all of these experimental outcomes; a more plausible explanation for our data is HS site specificity for globin gene activation, as our work posits. Therefore, our studies, when added to the data of previously published work, make a strong case for HS site specificity for globin gene activation. As such, these results may reflect inherent differences as to how the LCR functions within the human and mouse β-globin loci. These findings wait testing in genetically modified human progenitor cell systems or other human model systems.
One interpretation of these contrasting results in human versus mouse may be explained by taking into account the existence of two LCR functions, chromatin-opening activity and expression-enhancing activity. Although the mouse locus does not appear to require chromatin-opening activity, integrated human locus transgenes may require this activity in order to perform expression-enhancing activity. The 5′ΔHS3c might suppress the chromatin-opening activity, but the full-size deletion might have less influence on this activity. Integration of a wt, a core deletion or a full deletion transgene at a single genomic location to measure activity at the same single site might verify this possibility and allow understanding the mechanism.