Our data show that KLF3 binds the adult mouse α-globin promoter in erythroid tissue in vivo. However, KLF3 does not appear to functionally repress the endogenous promoter in red blood cells. Similarly, we have previously observed KLF3 occupancy at the adult β-globin (Hbb-b1) promoter in erythroid cells and no associated perturbation of Hbb-b1 transcription upon ablation of KLF3 . It is notable that KLF3 binding is highest at the late stages of erythroid maturation (compare Figure 4A with 4B, and 4D with 4E) when the adult globin genes are expressed at very high levels and their promoters are presumably highly accessible. This is also when KLF3 levels peak  and it is possible that KLF3 gains access to these regions but is not sufficiently potent to limit KLF1 driven activation of the genes. This observation highlights the view that transcription factor binding sites discovered by ChIP may not always have functional relevance in the context in which they are identified, but may instead reflect the dynamic nature of transcription factor binding at permissive loci. Indeed, a number of recent ChIP-Seq experiments, performed in association with transcriptome analysis of gene knockout models have revealed that transcription factor binding is not always associated with changes in gene activity [34, 35].
In contrast to the endogenous mouse α-globin promoter, we have shown that KLF3 does appear to regulate the expression of a human transgenic promoter in erythroid cells. The transgene is driven by a minimal human α-globin promoter and HS-40 and perhaps this subset of elements is more reliant on repression by KLF3 than the entire set of globin regulatory elements. In the case of the endogenous α-globin locus, chromatin conformation capture experiments suggest that gene expression is dependent upon chromosomal looping of distal enhancers to the proximal promoter, in a process that is dependent upon many regulatory factors . The removal of such complexity in the transgene most likely offers a far greater opportunity for observing the contribution that single factors make to expression levels. Alternatively, it should be noted that the experiments presented here primarily analyzed KLF3 function in murine cells, and thus it remains possible that KLF3 may play a role in α-globin regulation in human erythroid cells. Indeed, the related factor KLF4 has been shown to positively regulate the human α-globin promoter in reporter assays and to drive the endogenous HBA gene in K562 cells .
The up-regulation of GFP expression in Line3::Klf3−/− mice shows that KLF3 can functionally repress the transgenic α-globin regulatory sequences in vivo, and may function as an epigenetic modifier of transgene expression. KLF3 mediates repression of its target genes by binding the co-repressor CtBP , which in turn recruits a repressive complex that includes several epigenetic modifiers, such as LSD1, G9A, EUHMT, PC2, HDAC1, and HDAC2 [37, 38]. These factors facilitate histone methylation, demethylation and deacetylation, and are responsible for the addition of repressive epigenetic marks and gene silencing. It is possible that the absence of KLF3 in Line3::Klf3−/− erythrocytes prevents CtBP from being recruited to the transgene, and it is this that allows the rewriting of epigenetic marks permissive for transcription, resulting in the up-regulation of GFP expression. Indeed, the Line3 mice have frequently been used in ENU mutagenesis screens for modulators of variegated expression, and these screens have predominantly culminated in the identification of epigenetic modifiers, including HDAC1 [39–43].
Another possible explanation for the lack of de-repression of the endogenous α-globin gene in red blood cells is that the locus is already fully open and maximally expressed, so significant further de-repression cannot occur. In contrast, the transgene contains only a limited subset of regulatory sequences, and may therefore be expressed at lower levels allowing its up-regulation in the absence of KLF3. To circumvent this, we examined regulation in murine embryonic fibroblasts, as α-globin mRNA expression is limited to low but detectable levels in this cell type. In these non-erythroid cells, we identified a modest but significant increase in α-globin gene expression in the absence of KLF3. Moreover, in support of a role for CtBP in the regulation of the α-globin locus, we note that another group have observed a similar de-repression (4-fold) of α-globin gene expression from microarray analysis of Ctbp−/− murine embryonic fibroblasts .
Both the human and mouse α-globin loci lie in an open chromosomal region, surrounded by a number of actively expressed genes and in non-erythroid cells these loci retain the hallmarks of constitutively accessible chromatin . This contrasts significantly with the more isolated β-globin gene cluster, where in non-erythroid cells a silent heterochromatic state is established and maintained. It therefore appears that the α-globin locus employs different silencing mechanisms to prevent expression in non-red blood cells. In the case of the human locus, this is achieved by targeted recruitment of the repressive polycomb complex, PRC2, to CpG islands in the promoter regions . However, these CpG islands have been significantly eroded in the murine α-globin locus (Figure 2) and recruitment of PRC2 has not been detected, most likely due to loss of polycomb recruitment sites . The mechanism of α-globin gene silencing in non-erythroid tissue in the mouse therefore remains unclear. Here we suggest that KLF3 participates in this silencing and may do so not through direct interaction with the α-globin proximal promoter but via distal regulatory regions such as HS-26. In erythroid cells, HS-26 is an enhancer element that loops to the α-globin promoter and is required for appropriate regulation of expression . In non-erythroid cells, such looping is disrupted and occurs at a much lesser frequency . Whilst these observations allude to the functional importance of the HS-26 element, it should be noted that loss of HS-26 only modestly deregulates α-globin expression in erythroid cells and has not been reported to perturb non-erythroid silencing [48, 49]. Thus it is likely that correct tissue-specific control of the locus is achieved by a complex interplay between multiple cis-acting regulatory regions and positively- and negatively-acting trans factors such as KLF3 and KLF1.