Haploinsufficiency for the erythroid transcription factor KLF1 causes hereditary persistence of fetal hemoglobin.

Hereditary persistence of fetal hemoglobin (HPFH) is characterized by persistent high levels of fetal hemoglobin (HbF) in adults. Several contributory factors, both genetic and environmental, have been identified but others remain elusive. HPFH was found in 10 of 27 members from a Maltese family. We used a genome-wide SNP scan followed by linkage analysis to identify a candidate region on chromosome 19p13.12-13. Sequencing revealed a nonsense mutation in the KLF1 gene, p.K288X, which ablated the DNA-binding domain of this key erythroid transcriptional regulator. Only family members with HPFH were heterozygous carriers of this mutation. Expression profiling on primary erythroid progenitors showed that KLF1 target genes were downregulated in samples from individuals with HPFH. Functional assays suggested that, in addition to its established role in regulating adult globin expression, KLF1 is a key activator of the BCL11A gene, which encodes a suppressor of HbF expression. These observations provide a rationale for the effects of KLF1 haploinsufficiency on HbF levels.

Hemoglobin (Hb) is composed of two α-like and two β-like globin chains, encoded by genes in the HBA and HBB clusters, respectively. Developmental regulation of globin genes results in expression of stage-specific Hb variants (Supplementary Fig. 1). HbF ameliorates the symptoms of β-thalassemia and sickle cell disease, and reactivation of the HBG1/HBG2 genes in adults is therefore of significant interest for the clinical management of β-type hemoglobinopathies. After birth, HbF is gradually replaced by adult hemoglobin (HbA) 4. Residual amounts of HbF continue to be synthesized throughout adult life. In the majority of adults, HbF contributes <2% to total Hb, but there is considerable variation 5. Genetic studies have revealed three loci that control HbF levels in adults: HBB (11p15.4) 6–7, HBS1L-MYB (6q23.3) 6,8–9 and BCL11A (2p16.1) 10–11. Together, these loci account for <50% of the variation in HbF, indicating that additional loci are involved 5. Genetic analysis of HPFH families is a particularly powerful approach to identify novel modifiers of HbF levels 8. Here, we describe a Maltese pedigree with HPFH. The proband (II-5, Fig. 1a) was referred to the clinic because of microcytosis. She presented with high HbF levels (19.5%). Additional family members were recruited and ten of twenty nine tested were identified with HPFH (Fig. 1a and Supplementary Table 1), suggesting an autosomal dominant inheritance of the trait. Linkage to the HBB locus (Supplementary methods) was excluded, indicating involvement of a trans-acting factor. We performed a genome-wide linkage analysis on twenty seven family members to identify candidate loci for the HPFH modifier. Whole genome multi-point parametric linkage analysis was conducted using the Merlin programme 12 with two software packages, easyLINKAGE 13 and dChip 14. The analyses resulted in one significant linkage peak with LOD scores of 2.7 and 4.2, respectively, on chromosome 19p13.12–13 (Fig. 1b and Supplementary Fig. 2). These analyses were performed using an autosomal dominant model, assuming a penetrance of 90% and 1% phenocopy rate. No evidence of significant linkage was observed to the previously reported trans-acting HPFH loci of chromosomes 2p16.1 10–11 and 6q23.3 6,8–9. This was further investigated by genotyping of the five individual SNPs linked to increased HbF levels. This ruled out involvement of the HBS1L-MYB locus and revealed that heterozygosity at SNP rs766432 in the BCL11A locus may have contributed to the increased HbF levels, but it was not the major determinant (Supplementary Table 1). HPFH individuals had a consistent haplotype at 19p13.12–13, and the inferred haplotypes revealed that all HPFH individuals shared one copy of an identical chromosome segment, presumably containing the putative HPFH locus (Supplementary Fig. 2). Recombination events delineating the linkage region are indicated with arrows. The distal boundary is determined by a recombination event in individuals IV-3 and IV-5 (Supplementary Fig. 2, white arrow). The proximal boundary is determined by individuals III-12, III-18, IV-6 and IV-7 (Supplementary Fig. 2, black arrow). These narrowed the region down to a 663 kb interval between rs7247513 and rs12462609. The KLF1 gene, encoding a key erythroid transcriptional regulator 2, resides in this area. Mutations in KLF1 have been reported to form the molecular basis of the rare blood group In(Lu) phenotype 15, but a connection with HPFH has not been made. DNA sequencing revealed two linked mutations in KLF1 present exclusively in all HPFH individuals (Fig. 1c). The first mutation, p.M39L, is most likely a neutral substitution since mouse Klf1 contains a leucine at this position 16. The second mutation, p.K288X, ablates the complete zinc finger domain and therefore abrogates DNA binding of the mutant protein 17. The KLF1 p.K288X variant was absent in a random sample from the general Maltese population (n=400). To identify differentially expressed genes, RNA was isolated from erythroid progenitors (HEPs) cultured from peripheral blood 18 of four HPFH and four non-HPFH family members and used for genome-wide expression analysis. Comparison to the reported gene expression profiles of mouse Klf1 null erythroid progenitors 19 identified a set of common differentially regulated genes (Supplementary Table 2). Cluster analysis with this set of genes separated the non-HPFH samples from the HPFH samples (Fig. 2a), consistent with the notion that KLF1 activity is compromised in the HPFH individuals. Deregulation of these KLF1 target genes could explain the mild hypochromic microcytic indices displayed by the HPFH individuals (Supplementary Table 1). Of note, the embryonic Hbb-y and HBE1 genes were highly upregulated (Supplementary Table 2) while expression of the fetal globin repressor BCL11A 3 was downregulated (Supplementary Table 2 and Supplementary Fig. 3). Expression of fetal/adult globins can not be measured quantitatively on the microarrays owing to saturation effects. Quantitative RT-PCR (qPCR) confirmed downregulation of BCL11A, and showed increased expression of HBG1/HBG2 genes in the HPFH samples (Fig. 2b). Next, we investigated the effects of KLF1 knockdown in HEPs derived from healthy donors. Efficient knockdown of KLF1 was observed with two out of five lentiviral shRNA constructs 20 tested (Fig. 3a). Quantitative S1 nuclease protection assays 21 demonstrated significantly increased HBG1/HBG2 expression upon KLF1 knockdown (Fig. 3b–d), which was confirmed by qPCR (Fig. 3e). In addition, we found that BCL11A expression was diminished after KLF1 knockdown, both at the protein- (Fig. 3a) and at the mRNA level (Fig. 3e). Thus, the effects of KLF1 insufficiency on HBG1/HBG2 and BCL11A expression in HEPs from healthy donors were similar to those observed in KLF1 p.K288X heterozygotes, supporting the causative role of this mutation in the HPFH phenotype. To further investigate this, we transduced HEPs with lentiviral vectors expressing the KLF1 p.K288X truncation mutant or full length KLF1. The transgenic proteins were expressed at physiological levels in control HEPs (Supplementary Fig. 4a). This did not affect HBG1/HBG2 expression levels (Supplementary Fig. 4b, c), indicating that the truncated form of KLF1 does not act as a dominant-negative factor. In HPFH HEPs, lentivirus-mediated expression of full-length KLF1 resulted in considerable downregulation of HBG1/HBG2 mRNA levels, while expression of truncated KLF1 had no effect (Fig. 4). BCL11A protein levels were increased after transduction with full-length KLF1 lentivirus, while no such changes were observed upon transduction with either GFP- or truncated KLF1 lentiviral vectors (Fig. 4a). We noted that the endogenous truncated KLF1 protein was not or at best barely detectable in HPFH HEPs. This suggested that RNA transcribed from the KLF1 p.K288X allele was subject to nonsense-mediated decay 22, further emphasizing that it was dysfunctional. Consistent with this notion, we found that KLF1 mRNA expression was reduced in HPFH HEPs (Supplementary Fig. 3). It is well established that KLF1 preferentially activates the HBB gene, at the expense of HBG1/HBG2 gene expression, through direct interactions with regulatory elements in the HBB promoter 23–25. The molecular analysis of the Maltese HPFH family is consistent with this function of KLF1. In addition, it revealed a novel mechanism by which KLF1 tips the balance from HBG1/HBG2 to HBB expression: through activation of the gene encoding the HBG1/HBG2 repressor BCL11A 3. The promoter area of the BCL11A gene contains several putative KLF1 binding sites (CACC boxes; Fig. 5a). We performed chromatin immunoprecipitation (ChIP) assays to investigate whether KLF1 was bound to the BCL11A promoter in vivo. We used human fetal liver erythroid progenitors, which express high levels of HBG1/HBG2, and HEPs from adult peripheral blood in which the HBG1/HBG2 genes are suppressed. In adult HEPs, we observed strong binding of KLF1 to the BCL11A promoter (Fig. 5b). This was similar to the binding of KLF1 to the HBB promoter which served as a positive control 26. Neither promoter appeared to be bound by KLF1 in fetal liver-derived erythroid progenitors. ChIP reactions with the unrelated CD71 antibody were negative in all cases. We conclude that in adult HEPs KLF1 is bound to the BCL11A promoter in vivo. Diminished KLF1 activity, either through mutation of one KLF1 allele as occurs in the Maltese HPFH individuals or experimentally through shRNA-mediated knockdown in HEPs from normal donors, decreased BCL11A expression. Conversely, BCL11A levels increased upon restoration of KLF1 activity in Maltese HPFH HEPs. This identifies KLF1 as a double-barreled regulator of fetal-to-adult globin switching in humans (Supplementary Fig. 5). Firstly, it acts on the HBB locus as a preferential activator of the HBB gene 27. Secondly, it activates expression of BCL11A, which in turn represses the HBG1/HBG2 genes. This “double whammy” ensures that in most adults HbF levels are <2% of total Hb. In conclusion, we have identified haploinsufficiency for KLF1 as a cause of HPFH. We suggest that attenuation of KLF1 activity may be a fruitful approach to raise HbF levels in patients with β-type hemoglobinopathies.

Figure 1

Chromosome 19 locus linked to HPFH in a Maltese family

a) The Maltese HPFH pedigree. HbF levels are indicated as percentage of total Hb (%HbF). HPFH individuals are shown as half-filled symbols.

b) LOD scores derived from genome-wide linkage analysis. The putative HPFH locus on chr. 19 is indicated by an arrow. pLOD = parametric LOD score; MPT = multi point test; cM = centiMorgan.

c) Sequence analysis of KLF1. HPFH individuals were heterozygous for two mutations (arrows; Supplementary Table 1). The predicted effects of the mutations on KLF1 protein are shown below.

Figure 2

KLF1 target genes are downregulated in wt/KLF1 p.K288X HEPs

a) RNA isolated from HEPs derived from normal (wt/wt) and HPFH (wt/KLF1 p.K288X) family members was used for genome-wide expression analysis. Deregulated genes common between wt/wt and wt/KLF1 p.K288X and mouse wt/wt versus Klf1 null mutant erythroid progenitors 19 (Supplementary Table 2) were used for cluster analysis.

b) Validation of key target genes by qPCR. Expression levels of BCL11A were normalized using GAPDH as a reference. Expression levels of HBG1/HBG2 (HBG) were calculated as ratio to HBA1/HBA2 (HBA) expression. Medians are indicated by red lines in the box plots. Asterisk: p=0.0209.

Figure 3

Increased HBG1/HBG2 expression after knockdown of KLF1 in normal HEPs

a) HEPs derived from normal donors were transduced with shRNA-expressing lentiviruses. Cells were harvested five days after transduction, and nuclear extracts prepared. Top panels: KLF1 protein expression assessed by Western blot analysis. Bottom panels: BCL11A protein levels were reduced upon KLF1 knockdown. NPM1 served as a loading control. None = mock transduction; TRC = control non-specific shRNA; sh1 and sh2 = two independent shRNAs targeting KLF1. A non-specific band is indicated (<).

b) RNA was isolated from HEPs five days after transduction with the indicated lentiviruses, and used in quantitative S1 nuclease protection assays to measure globin expression. Arrows indicate protected fragments diagnostic for HBA1/HBA2 (HBA), HBG1/HBG2 (HBG) and HBB mRNAs.

c) Quantitation of data shown in b) by Phosphorimager analysis.

d) Box plots of HBG/HBA ratios after sh1-mediated KLF1 knockdown in HEPs derived from three independent donors. Medians are indicated by red lines. Asterisk: p=0.0463.

e) Box plots of qPCR analysis of AHSP, BCL11A and HBG expression after sh1/sh2-mediated KLF1 knockdown. AHSP is a known KLF1 target gene 28 and serves as a positive control. Expression levels of AHSP and BCL11A were normalized using GAPDH as a reference. Expression levels of HBG were calculated as ratio to total β-like globin expression (HBG + HBB) expression. Medians are indicated by red lines. Circles: points outside the range of the error bars. Asterisk: p=0.020; double asterisks: p<0.003.

Figure 4

Expression of exogenous KLF1 in HPFH HEPs

a) HEPs derived from individual II-5 were transduced with lentiviral constructs expressing GFP, KLF1 truncated at amino acid 288 (TR) or full-length KLF1 (FL). Seven days after transduction, nuclear extracts were prepared and expression of BCL11A and KLF1 was assessed by Western blot. NPM1 served as a loading control.

b) RNA was isolated from II-5 HEPs seven days after transduction with the indicated lentiviruses, and was used for quantitative S1 nuclease protection assays to measure globin mRNA expression. Arrows indicate protected fragments diagnostic for HBA1/HBA2 (HBA), HBG1/HBG2 (HBG) and HBB mRNAs.

c) Quantitation of data shown in b) by Phosphorimager analysis.

Figure 5

KLF1 binds to the promoter of the BCL11A gene in vivo

a) Schematic drawings of the promoter areas of the BCL11A, HBB and RASSF1A genes. Positions of potential KLF1 binding sites (CACC boxes) and PCR primers used are indicated. Arrows indicate transcription start sites.

b) ChIP analysis of KLF1 binding to the BCL11A promoter in human fetal liver cells and adult HEPs. The HBB promoter served as a positive control 26. RASSF1A was used as a negative control, and the unrelated CD71 antibody served as a control for the specificity of the KLF1 antibody. Asterisk: p<0.05; double asterisks: p<0.01. Error bars: standard deviation.

Online Methods

Molecular genetic analysis

The proband (II-5; Fig. 1a) was referred to the clinic because of microcytosis. She presented with 19.5% HbF, and therefore additional family members were approached to participate. Blood samples were obtained with informed consent and standard hematological indices were determined (Supplementary Table 1). Genomic DNA was extracted from ~1 × 106 cells from whole blood using a modified salting out procedure 29. Control DNA samples isolated from 400 random Maltese individuals were available from the Laboratory of Molecular Genetics, Biomedical Sciences Building, University of Malta. The family members were genotyped in the HBB, HBD genes and the HBG1/HBG2 gene promoters to detect point mutations and small insertions/deletions leading to β-, δ-thalassemia or HPFH, respectively, following routine procedures 30. Gap PCR was carried out to detect possible genomic rearrangements leading to deletional HPFH or δ β-thalassemia 31. This excluded linkage of the HPFH phenotype to the HBB locus. Occurrence of common α-thalassemic mutations (SEA, 3.7 and 4.2 deletions) was also excluded. The NspI mapping 250K set (Affymetrix, Santa Clara, CA, USA) was used to analyze twenty seven DNA samples from the HPFH family, starting with 250 ng of genomic DNA per array. Individual SNPs in the HBS1L-MYB (rs28384513, rs9399137, rs4895441) and BCL11A (rs766432, rs11886868) loci 32 were genotyped manually.

DNA Linkage Analysis

Multipoint parametric linkage analyses was performed using the Merlin v1.0.1 12 program with two software packages; EasyLinkage v5.05 Beta 13 and dChip 14, in order to calculate parametric LOD scores. Parametric analysis was carried out using an autosomal dominant mode of inheritance. Penetrances used for the dominant model were 0.01, 0.90 and 0.90 for the wild-type homozygote, mutant heterozygote and mutant homozygote, respectively. We assumed a disease allele prevalence frequency of 0.0001, and a phenocopy rate of 1%. A co-dominant allele frequency algorithm was used for the analysis. These analyses were carried out using the sex-averaged 500K Marshfield genetic map provided with the easyLINKAGE software package 13. Mendelian inheritance check was performed for all family members, using the program PedCheck 33 and incompatibilities were omitted from the analysis. This increased the power and accuracy. The analysis was performed by taking HbF as a quantitative hematological value and classifying family members as ‘affected’ with HbF >2%, and ‘non-affected’ with HbF <2%. Replicates of the linkage analysis and inferred haplotypes were constructed and visualized using dChip 14. DNA from 400 random Maltese individuals was used to check for the presence of the KLF1 p.K288X mutation in the population.

Cell culture

Human erythroid progenitor cells (HEPs) were cultured as described 18 in the presence of recombinant human Epo (1 unit/ml, kind gift of Ortho-Biotech, Tilburg, The Netherlands), recombinant human SCF (50 ng/ml, kind gift of Amgen, Breda, The Netherlands) and dexamethasone (5 × 10−7 M; Sigma, St. Louis, MO, USA). Cells were counted with an electronic cell counter (CASY-1, Schärfe System, Reutlingen, Germany).

Transcription profiling

A minimum of 1.5 × 106 HEPs were harvested at day12 of culture and RNA was extracted with Trizol reagent (Sigma) and purified using the RNeasy Mini Kit (Qiagen, Crawley, UK), including an on-column DNaseI digestion, according to the manufacturer’s instructions. RNA yield was determined using the 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA). 8–10μg of total RNA was analysed by microarrays using cells from day 12 of culture. Quality of the total RNA samples and the resulting cRNA was assessed on the Bioanalyzer. Fragmented biotinylated cRNA was prepared and 15μg hybridized to HG-U133 plus 2 GeneChips, according to the manufacturer’s protocols (Affymetrix). The data files have been deposited in MIAME-compliant format in the NCBI GEO database (GSE22109). Single Array Expression Analysis was performed using the Affymetrix GeneChip Operating Software (GCOS). A global scaling strategy was used to give an average target intensity of 500 for each array. Data from all eight arrays were filtered to exclude probe sets called either absent or marginal in all arrays. Control probe sets with the prefix AFFX were also removed prior to subsequent data analysis. Filtered data were transformed to a log2 scale and analysed to determine differentially expressed genes. A 1.5-fold change threshold and test statistic of p<0.05 were used as cut-off. A list of genes differentially expressed in mouse Klf1 null erythroid progenitors (p<0.05) 19 was downloaded from http://data.genome.duke.edu/EKLFDef.

Quantitative S1 nuclease protection assays

To measure globin mRNA levels directly, we used quantitative S1 nuclease protection assays 21. The probe fragment for detection of HBG1/HBG2 mRNAs was amplified by PCR using the primers S1-HBG-S and S1-HBG-A (Supplementary Table 3). Sizes of probes/protected fragments are: HBA1/HBA2: 700 nt/218 nt; HBG1/HBG2: 350 nt/165 nt; HBB: 525 nt/155 nt 21. Quantitation was performed using a Typhoon Trio Phosphorimager (GE Healthcare, Chalfont St Giles, UK) and corrected for specific activity of the probes.

qPCR analysis

Total RNA (1 μg) isolated from HEPs was converted to cDNA using SuperScript II reverse transcriptase according to the manufacturer’s instructions (Invitrogen, Carlsbad, CA, USA). Expression levels of mRNAs were analysed by quantitative real-time PCR (qPCR). Amplification reactions were performed with primers designed with Primer Express software v2.0 (Applied Biosystems, Foster City, CA, USA). All amplifications used SYBR Green PCR Master Mix (Applied Biosystems). qPCR was performed with an Optical IQ Thermal Cycler (Bio-Rad Laboratories, Hercules, CA, USA) with the following conditions: 50°C for 2 minutes and 95°C for 10 minutes, followed by 45 cycles of 95°C for 15 seconds and 62°C for 45 seconds. All reactions were performed in triplicate. Gene expression levels were calculated with the 2 (−DeltaDeltaC(T)) method 34. Target gene expression was normalized to GAPDH expression, unless indicated otherwise. Primers used are listed in Supplementary Table 3.

Statistical analysis

Statistical analysis of gene expression data obtained from quantitative S1 nuclease protection assays and qPCRs was performed with Mann Whitney tests using STATA data analysis and statistical software (StataCorp LP, College Station, TX, USA).

KLF1 expression constructs

Human KLF1 cDNA clone (BC040000, Imagenes, Berlin, Germany) was amplified by PCR with an att-specific set of primers (Invitrogen) in order to fuse the cDNA with a V5 tag at the C-terminus of the protein. Primers used were KLF1-F and KLF1-R1 (Supplementary Table 3). In parallel, part of the clone was amplified, truncating the protein at amino acid 288, with att-specific primers using a different reverse primer KLF1-R2. The PCR products were introduced into the lentiviral expression vector pRRLsin.sPPT.CMV.Wpre 35 modified for Gateway cloning (Invitrogen). The final clones were verified by sequencing.

Lentiviral transduction of human erythroid progenitors

Lentivirus was produced by transient transfection of 293T cells according to standard protocols 36. Two days after transfection, the supernatant was collected, filtered and concentrated by centrifugation at 20 krpm for 2h at 4°C. HEPs cultured for one week were transduced in 24 well plates. We used 0.5×106 cells per well and sufficient amounts of virus to transduce ~80% of the cells. When appropriate, puromycin (1μg/ml final concentration) was added to the cells after 2 days, and selection was performed for 2–3 days. At day 5–7 after transduction cells were harvested and nuclear extracts were prepared 37. RNA was extracted with the Trizol reagent. For knockdown experiments, clones from The RNAi Consortium (TRC 20; Sigma) were used. The non-target SHC002 vector was used as a control. (SHC002: 5’-CAACAAGATGAAGAGCACCAA-3’). Five shRNA clones targeting KLF1 were tested: TRCN0000016273, TRCN0000016274, TRCN0000016275, TRCN0000016276 and TRCN0000016277. Efficient knockdown of KLF1 expression was observed with TRCN0000016276 (sh1) and TRCN0000016277 (sh2). Sequences are listed in Supplementary Table 3.

Western blotting

Nuclear extracts were separated on denaturing polyacrylamide gels followed by semi-dry blotting to PVDF or nitrocellulose membranes. The membranes were probed with the following primary antibodies: BCL11A (sc-56013, Santa Cruz Biotechnology, Santa Cruz, CA, USA), NPM1 (ab10530, Abcam, Cambridge, UK), KLF1 26, and anti-V5-HRP (R961-25, Invitrogen). For detection, the appropriate secondary antibodies were used. The enhanced chemoluminescence kit (GE Healthcare) or the Odyssey Infrared Imaging System (Li-Cor Biosciences, Lincoln, NE, USA) was used to develop the membranes.

Chromatin immunoprecipitations

Fetal liver and adult HEPs were cultured 18 and used for ChIP reactions were performed as described 38 with the KLF1 antibody, and a CD71 antibody (347510, BD Biosciences, San Jose, CA, USA) as a negative control. qPCR was performed on the input and immuno-precipitated samples using primers for the RASSF1A, HBB and BCL11A genes. The relative fold enrichment was calculated as 2− [(CT x ChIP y – CT input y)-(CT KLF1-ChIP HEP RASSF1A- CT input HEP RASSF1A] (where ‘x’ is the antibody and ‘y’ the sample), i.e. setting the relative fold-enrichment of the RASSF1A amplicon by the KLF1 antibody in HEPs to 1. Primers used are listed in Supplementary Table 3.