KLF1 drives the expression of fetal hemoglobin in British HPFH

Fetal hemoglobin (HbF) levels are typically downregulated postnatally to ∼1% of total hemoglobin. However, in some individuals, higher levels of HbF persist throughout adulthood. This benign condition, hereditary persistence of fetal hemoglobin (HPFH), has been associated with mutations in regulatory regions of fetal γ-globin. Several HPFH variants have been identified, often in families that presented to the clinic as a result of other disease-causing hemoglobin mutations. Family members who were HPFH carriers tended to experience milder symptoms of disease and, thus, high levels of HbF are considered beneficial.¹ 

A −198T>C mutation in the promoter of the γ-globin gene was first described in 1974 in a large British family^2,3  and is known as British-type HPFH (Figure 1A; supplemental Figure 1, available on the Blood Web site). Carriers have HbF levels of ∼5% to 20%. Previously, it was shown that the mutation creates a sequence resembling a specificity protein 1 (SP1)–binding site and that the activator SP1 could bind this site in vitro.⁴  Accordingly, it was proposed that the ubiquitous SP1 transcription factor is responsible for driving high HbF levels in British HPFH individuals.^4,5  It was therefore unexpected that subsequent transactivation assays showed that elevated activity of the mutant promoter is erythroid-specific,^4,5  suggesting that an erythroid-specific transcription factor is involved in the overexpression of HbF. Given these results, we reinvestigated and found that the erythroid transcription factor Krüppel-like factor 1 (KLF1)^6-8  binds to and drives the expression of the −198T>C promoter.

In vitro binding assays were performed as previously described.⁹ 

HUDEP-2 cells¹⁰  were used as a model to study the effects of −198T>C HPFH. Cells were transfected with clustered regularly interspaced short palindromic repeat (CRISPR)/CRISPR-associated protein 9 (Cas9) plasmid px458¹¹  encoding a guide RNA targeting the γ-globin promoter and a single-stranded oligodeoxynucleotide containing the −198T>C mutation. Transfected cells were enriched by fluorescence-activated cell sorting and clonal populations were established.

γ-globin expression was determined by quantitative reverse transcription polymerase chain reaction (qRT-PCR) using primers to amplify the region of interest, and expression was normalized to 18S ribosomal RNA (rRNA). Flow cytometry was performed on permeabilized cells stained with an allophycocyanin (APC)-conjugated HbF antibody.

Chromatin immunoprecipitation (ChIP) was performed as previously described¹²  using an antibody against KLF1 and control immunoglobulin G.

Hemoglobin high-performance liquid chromatography (HPLC) was performed as previously described.¹³ 

Chromatin-conformation-capture (3C) assays were performed as previously described.¹² 

See supplemental Methods for more details.

SP/KLF transcription factor family members have DNA-binding domains composed of 3 Cys₂His₂-type zinc fingers that recognize CNCCC motifs where N can be any base but mostly is G or A.¹⁴  The DNA-binding domains of all members of the SP/KLF transcription factor family are virtually identical in the 3 residues in each zinc finger that make primary contacts to DNA (Figure 1B). Hence, it is not surprising that the binding sites of different SP/KLF family members are highly similar as determined by ChIP followed by sequencing (ChIP-Seq) experiments^15-18  (Figure 1C). Accordingly, we reasoned that, as SP1 has been shown to bind in vitro,^4,5  any of the members of the SP/KLF family could potentially interact with the −198T>C γ-globin promoter.

To test this prediction in vitro, we performed EMSAs with erythroid KLF (KLF1/erythroid KLF; Figure 1D) and 12 other KLF proteins (supplemental Figure 2), and assessed binding to the WT and −198T>C γ-globin promoters. All tested KLFs preferentially bound the mutant γ-globin promoter. Next, we investigated the abundance of KLF and SP proteins in erythroid tissues and cell lines because the −198T>C mutation activates γ-globin expression erythroid-specifically.^4,5  We found that KLF1 is the most abundantly expressed in differentiated human CD34⁺ bone marrow cells¹⁹  and erythroblast HUDEP-2 cells¹⁰  making it a likely candidate to be accountable for elevated HbF levels in British HPFH (supplemental Figure 3).

We then moved to engineer HUDEP-2 cells to stably carry the mutation. HUDEP-2 cells are an attractive recently described model for primary adult erythroblasts, given their expression patterns of globins¹⁰  and of the appropriate KLFs/SP proteins (supplemental Figure 3). We introduced targeted substitution mutations into the γ-globin locus using CRISPR/Cas9 genome editing and homology-directed repair to establish clonal cell populations carrying the −198T>C mutation (supplemental Figure 4).

First, we examined messenger RNA (mRNA) expression levels of β-like globins by qRT-PCR (Figure 2A). Clonal WT HUDEP-2 cells express ∼0.5% to 1% γ-globin mRNA [γ/γ+β]. Introducing the −198T>C mutation into a single fetal globin promoter elevated the percentage of γ-globin mRNA to 4% to 6%. Accordingly, we also observed that the percentage of HbF-immunostaining cells (“F cells”) increased in cells carrying the −198T>C mutation (Figure 2B).

After confirming that the −198T>C mutation reproduces an HPFH phenotype in HUDEP-2 cells, we investigated the binding of transcription factors to the −198T>C γ-globin promoter. We first sought to generate homozygous HUDEP-2 −198T>C cells. As it proved difficult to reproducibly target all 4 copies of the γ-globin genes, we introduced an intermediate step and first engineered a parental cell line carrying only 1 γ-globin gene and promoter on each allele (HUDEP-2(Δ^Gγ)WT; supplemental Figure 5). We then introduced the −198T>C mutation into these cells homozygously, applying the same strategy as outlined previously (HUDEP-2(Δ^Gγ)−198T>C). We found that γ-globin mRNA levels were strongly elevated compared with the parental cells and that the amount of F cells increased substantially in clonal HUDEP-2(Δ^Gγ)−198T>C cells (Figure 2C-D). We also performed HPLC for hemoglobin. In HUDEP-2(Δ^Gγ)WT cells, only adult hemoglobin was detected whereas HbF made up 52% to 77% in cells with −198T>C (Figure 2E). Western blots confirmed this finding and HbF repressor levels (BCL11A/ZBTB7A) were unchanged (supplemental Figure 6).

We then performed KLF1 ChIP and saw consistently higher enrichment of the γ-globin promoter in HUDEP-2(Δ^Gγ)−198T>C cells compared with the WT (Figure 2F). This effect was augmented when we performed the KLF1 ChIP in 5-day differentiated HUDEP-2(Δ^Gγ)WT and −198T>C cells (Figure 2G). To ensure that the antibody is specific for KLF1, we also carried out ChIP experiments across a range of loci using KLF1 knockout and WT HUDEP-2 clones. In KLF1 knockout cells, we observed only background binding, equivalent to that obtained with control immunoglobulin G, thereby validating the specificity of the antibody (supplemental Figure 7). Interestingly, we also observed that binding of KLF1 to the β-globin promoter was significantly reduced in HUDEP-2(Δ^Gγ)−198T>C cells, suggesting that KLF1 may switch from activating β-globin in WT^7,20  to driving γ-globin expression in −198T>C cells. Hence, we then looked at interaction frequencies of the LCR with the globin promoters by 3C. We observed that the LCR indeed interacts more frequently with the γ-globin promoter and less often with the β-globin promoter in HUDEP-2(Δ^Gγ)−198T>C compared with WT cells (Figure 2H).

These findings support our hypothesis that the −198T>C mutation creates a functional binding site for the erythroid activator KLF1 and that de novo binding of KLF1 contributes to the upregulation of γ-globin expression in British HPFH by shifting the LCR toward the fetal globin promoters (Figure 2I). This is consistent with the commonly accepted model of the LCR interacting competitively with the globin promoters. Although present in appropriate levels, BCL11A and ZBTB7A clearly do not silence the HbF expression at the −198T>C promoter, suggesting that de novo binding of KLF1 overrides the developmentally determined gene expression program of globins. This is analogous to what we¹²  and others²¹  have seen when other transcription factors that induce chromatin looping are targeted to the γ-globin promoter.

Several approaches have been proposed to reactivate the expression of HbF in adult life, including targeting the HbF repressor BCL11A²²  and its regulatory elements^23,24  and forced chromatin looping mediated by an artificial transcription factor.²¹  If the efficiency of targeting is sufficient,¹⁸  then introducing naturally occurring mutations associated with HPFH could represent an alternative approach.

Beyond the potential therapeutic aspects, our work is of interest as it shows that the −198T>C mutation generates a de novo binding site. An important insight gained from genome-wide association studies is that cis-regulatory mutations are commonly associated with genetic disease, and are as abundant as coding variants.²⁵  Mutations will typically either disrupt a transcription factor binding site or create a new site. Because a range of deletions, insertions, or substitutions can disrupt elements, but only precise changes can create new elements, one might expect that disruptions are more common than de novo sites. However, the first 2 HPFH mutations we studied, −198T>C and the previously reported −175T>C,¹²  create de novo sites for erythroid-specific transactivators. Similarly, a de novo GATA1 site can profoundly affect the expression of genes within the α-globin locus.²⁶  It may be that disruptions are less prevalent than expected because of redundancy observed between multiple cis-acting elements²⁷  but de novo sites are in fact more common than initially anticipated.

The CRISPR/Cas9 genome-editing plasmid px458 was a gift from Feng Zhang, Broad Institute, Cambridge, MA (Addgene plasmid # 48138). The authors thank Laura Norton, Alister Funnell, and Richard Pearson of UNSW Sydney (Sydney, NSW, Australia) for valuable discussions and reagents. The authors also acknowledge Helene Lebhar and the UNSW Recombinant Products Facility for assistance in the analysis of hemoglobin variants.

This work was supported by funding from the Australian National Health and Medical Research Council (APP1098391) (M.C.). B.W. was supported by a University International Postgraduate Award. G.E.M. was supported by an Australian Postgraduate Award.

Contribution: B.W. performed most of the experiments; G.E.M. performed EMSA with KLF1; R.K. and Y.N. provided HUDEP-2 cells; K.G.R.Q. and M.C. supervised the study; B.W., K.G.R.Q., and M.C. designed experiments and wrote the manuscript; and all authors have read and approved the contents of the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Merlin Crossley, School of Biotechnology and Biomolecular Sciences, University of New South Wales, Sydney, NSW 2052, Australia; e-mail: m.crossley@unsw.edu.au.