Novel variants in Krueppel like factor 1 that cause persistence of fetal hemoglobin in In(Lu) individuals

Beta-hemoglobinopathies become prominent after birth due to a switch from γ-globin to the mutated β-globin. Haploinsufficiency for the erythroid specific indispensable transcription factor Krueppel-like factor 1 (KLF1) is associated with high persistence of fetal hemoglobin (HPFH). The In(Lu) phenotype, characterized by low to undetectable Lutheran blood group expression is caused by mutations within KLF1 gene. Here we screened a blood donor cohort of 55 Lutheran weak or negative donors for KLF1 variants and evaluated their effect on KLF1 target gene expression. To discriminate between weak and negative Lutheran expression, a flow cytometry (FCM) assay was developed to detect Lu antigen expression. The Lu(a−b−) (negative) donor group, showing a significant decreased CD44 (Indian blood group) expression, also showed increased HbF and HbA2 levels, with one individual expressing HbF as high as 5%. KLF1 exons and promoter sequencing revealed variants in 80% of the Lutheran negative donors. Thirteen different variants plus one high frequency SNP (c.304 T > C) were identified of which 6 were novel. In primary erythroblasts, knockdown of endogenous KLF1 resulted in decreased CD44, Lu and increased HbF expression, while KLF1 over-expressing cells were comparable to wild type (WT). In line with the pleiotropic effects of KLF1 during erythropoiesis, distinct KLF1 mutants expressed in erythroblasts display different abilities to rescue CD44 and Lu expression and/or to affect fetal (HbF) or adult (HbA) hemoglobin expression. With this study we identified novel KLF1 variants to be include into blood group typing analysis. In addition, we provide further insights into the regulation of genes by KLF1.

Blood group phenotyping of donors is crucial to provide matched blood to anemic patients to prevent transfusion reactions and alloimmunization. The genetic basis of most blood groups have been elucidated allowing for genetic prediction of blood group phenotype through detection of prevalent small nucleotide polymorphism (SNP), insertion or deletions. Importantly, whole genome sequencing, amplicon or PCR based genetic screenings may identify specific novel variants of which the phenotypic translation may be difficult to predict. Therefore, the identification of novel genetic variants that lead to blood group null and weak phenotypes is still important to supplement existing high throughput genotyping. During erythropoiesis the indispensable erythroid specific transcription factor Krueppel like factor 1 (KLF1, formerly erythroid KLF) transcriptionally regulates expression of specific blood group antigens. KLF1 variants can cause a multitude of phenotypes. These vary from benign loss of specific blood groups to inherently dominant diseases like congenital dyserythropoiesis type IV (CDA IV). KLF1 variants have been divided into different classes (class 1 to 4) according to their location within (and effect on) the gene 1 . One of the blood group systems affected by KLF1 variants is Lutheran (Lu), encoded by the BCAM gene 2 . BCAM is a glycoprotein containing 25 Lu antigens and putatively functions as a laminin binding receptor 3 . BCAM negativity is rare (1:3000; 0.005-0.032% 4 ). It is caused by KLF1 or GATA1 mutations, but also deletions/stop variants in the BCAM gene itself causes negativity for BCAM 2,5,6 . KLF1 variants specifically result in the dominant type of Lu(a−b−), termed In(Lu) or inhibitor of Lu. Recessive Lu(a−b−), also known as the [Lu null Donor blood sample collection and typing. Peripheral blood mononuclear cells (PBMCs) were obtained according to the Declaration of Helsinki (seventh revision, 2013). Written informed consent to extract and use human material (peripheral blood) as part of regular donations was obtained with approval of the local medical ethics committee (MEC; guidelines of NetCord FACT, by the Sanquin Blood bank, The Netherlands). Experimental protocols were approved by Sanquin (MEC). Lutheran blood typing was performed by serology using anti Lu(a) or Lu(b) antibodies (Pelicluster ® , Sanquin, the Netherlands) in standard agglutination assays. Table 1) and KLF1 exons 1-3 and promoter (Supplementary Table 2), were amplified by PCR cleaned by Exo-SAP ® as described by manufacturers (GE Healthcare, Exo Pro Star kit) and sequenced on ABI 3730xl DNA Analyzer (Thermo Fisher).

Donor sequencing. Blood collected from Lu negative and weak donors (Supplementary
Flow cytometry. Erythrocytes were diluted to 1 × 10 8 cells/ml and incubated with primary antibody (Supplementary Table 3) for 30' at RT and washed five times with PBS supplemented with 0.5% bovine serum albumin (BSA). Unconjugated antibodies were additionally stained with a secondary antibody goat anti-human APC and incubated for 30' on ice. Cells were washed once with ice-cold PBS/0.5% BSA. For intracellular staining of globin subunits, cells were fixed using 0.025% glutaraldehyde for 10 min, pelleted (1800 rpm, 5 min) and permeabilized using 0.05% NP40 for 10 min. After pelleting and washing using ice-cold PBS, cells were stained using HbA-PE (Santa Cruz) and HbF-APC (Invitrogen) for 30 min at 4 degrees Celsius. Samples were measured using a Flow Cytometer Canto II (BD Biosciences) and analysed using Flowjo software (Flowjo, LLC, USA).

HPLC.
Hb isoform expression was determined on a minimum of 1 × 10 7 cells by high-performance cationexchange liquid chromatography (HPLC) on Waters Alliance 2690 equipment as previously described 13 . Blood-MLPA assay. DNA samples from 55 Caucasian donors were extracted and genotyped by Multiplex ligation-dependent probe amplification (MLPA). These donors were genotyped with exon-specific probes for the Lutheran gene to evaluate the relative copy number of each DNA sequence. Each probe was composed of two half-probes (5′ and 3′ half-probes), consisting of a target-specific sequence and a universal primer sequence allowing the simultaneous multiplex PCR amplification of all targets 14 . Cloning of novel KLF1 mutants. The ORF from human KLF1 was amplified by PCR using primers with specific restriction enzymes within the oligo's and cloned into the lentiviral vector PAD5 (pRRLSIN.PPT. hCMV.2A-GFPpre 15 ), which was cut using XbaI (NEB, R0145S) and EcoRI (NEB, R0101S). The novel KLF1 variants identified were introduced into the wt KLF1 within PAD5 using Phusion Site-Directed Mutagenesis Kit according to manufacturer's guidelines (Thermo Fisher; primers in Supplementary www.nature.com/scientificreports/ CaCl 2 . Dropwise 500μL HEPES 2 × was added under shaking conditions. After 2 days medium was harvested, spun down to remove cell debris (1800 rpm, RT) and filtered using a 0.45 μm filter. Filtered medium was ultracentrifuged for 3 h at 20.000 g and resuspended in IMDM medium in 500 μL aliquots, snap frozen with liquid nitrogen and stored at − 80 C until used.
Primary erythroid cell culturing and transduction with KLF1 constructs. Erythroblasts were cultured from CD34 + cells purified from mobilized peripheral blood (MPB; consent was given in accordance with the declaration of Helsinki and the Sanquin internal ethics board). On day 1 isolated CD34 + we resuspended in Cellquin 16 , supplemented with stem cell factor (SCF, 239 T culture supernatant; 100 ng/ml), TPO (50 ng/nl), IL3 (10 ng/ml), IL6 (10 ng/ml), at a cell density of 1 × 10 6 cells/mL as described previously 16 . Cells were cultured for 4 days (37 °C, 5% CO2) and medium was refreshed every other day. On day 4, cells were washed and cultured as described previously 16  SDS PAGE and western blotting. Cells were lysed in ice-cold lysis buffer as previously described 17 .

Statistical analysis.
All measurements represent the average MFI of at least three single individual measurements (*** indicates a P value of < 0.001, Tukey's multiple comparisons test or unpaired student T-test as indicated in the figure legends). MFI is indicated as the Mean geometric Fluorescence Intensity of the signal. Statistical analysis was done using GraphPad Prism (version 6.01). First, One Way Anova test was performed, followed by a Tukey's multiple comparisons test.

Results
Defining negativity for the mixed field Lu blood group by flow cytometry. A cohort of 55 Lu weak (n = 25) or negative (n = 30) donors were identified using standard agglutination techniques (Supplementary Table 1). Detection of the Lu blood groups by antibody-mediated agglutination results in different degrees of agglutination, even in Lu positive samples. This is termed a mixed field result, which can lead to difficulties in discriminating between Lu negativity or weak expression 18 Table 1). In 11 out of 55 cases, serology typed negative for Lu(a) or Lu(b) while FCM could still detect traces of Lu(a) or Lu(b), which was in agreement with genotyping. In 7 out of 55 cases serology could pick up a weak Lu(a) or Lu(b) signal while FCM was negative with the threshold set ( Fig. 1B,C). Lutheran antigens are notoriously very variable in strength partly due to the mixed field results and complicating serology. Indeed, adsorption and elution tests are sometimes required to detect weak Lua or Lub.  www.nature.com/scientificreports/ are located in the second linker, outside the second zinc finger (Fig. 2). Both variants alter residues that are highly conserved throughout evolution 21 . The c.1003G > A variant leads to a p.Gly335Arg substitution altering the glycine residue. Two 2nd linker variants at the p.Gly335 site have been described before: Thr334Lys and p.Thr334Arg (rs483352841) and correlated with HPFH 22 . The novel c.1001C > T variant leads to p.Thr334Met substitution and possibly causes an alternate start site in the linker 2 region of KLF1. Of note, one Sickle cell disease carrier was found in the Lu weak group (Supplementary Table 1, indicated by **).

In(Lu) individuals display lowered CD44 expression compared to control and other Lutheran negative individuals.
Variants in the transcription factor KLF1 may lead to additional deregulation of KLF1 target genes. CD44 (harboring the Indian blood group system) has been described as a KLF1 target gene 2 .
In addition, variants in KLF1 have been shown to result in upregulation of fetal hemoglobin 9,23 and correlate with increased adult hemoglobin A2 24 . We further characterized the Lu weak/negative cohort (including individuals without KLF1 variants), by measuring CD44, (harboring the Diego blood group system) and), GPA (harboring MNS antigens) and band 3 expression levels by FCM (Fig. 3A, Supplemental Fig. 3A,B). GPA and Band3 were used as controls to indicate that the changes in specific markers affected in In(Lu) are not caused by differences in the populations analyzed. Wild type, control erythrocytes display a variable scattered pattern of CD44 GM (Fig. 3A). Significantly lower expression of CD44 was found in donors displaying weak or negative Lu expression (Supplementary Table 5). In(Lu) donors showed the lowest average CD44 expression compared to control individuals. One of the 13 p.Leu354 variants, displayed a Lu(b) weak expression by FCM (but was Lu(b) negative by serology). In contrast, the same individual also showed a normal CD44 expression. Band3 and GPA membrane expression were not significantly different (Supplemental Fig. 3A,B, Supplementary Table 5).
Hierarchical Euclidean complete clustering of FCM data using CD44, Lu(a) and Lu(b) z-scores segregated the Lu negative individuals with variants from the rest of the samples strengthening the link between KLF1 variants and low CD44 expression (Fig. 3B). The K-means clustering defines Lu(b) + cells is a separate cluster from Lu(a) or Lu(a,b), which may suggest correlations based on the Lutheran phenotype itself. Note that some heterozygous  Mutations in KLF1 lead to changes in hemoglobin subunit distribution. Next we examined if the Lu weak and negative individuals display altered expression of specific hemoglobins by HPLC (Fig. 4, Supplementary Fig. 4A,B). Increased HbF and HbA2 expression in In(Lu) individuals was observed compared to normal controls and the remaining cohort (Fig. 4B,C, Supplementary Table 5). This was accompanied with a concomitant drop in HbA1 (α2β2) and resembles the HPFH phenotype (Fig. 4A). Some KLF1 mutants, such as the p.Trp271Ter and p.Lys288Glu show high HbF levels of 5% and 1.8% respectively. KLF1 variants that occur outside the DNA binding region also display phenotypic characteristics associated with In(Lu) and HPFH (Supplementary Table 1). Most of the In(Lu) individuals contained the class 2 c.1060C > G variant, which correlated with low CD44 expression, Lu negativity and elevated HbA2 levels but not with increased HbF levels. Euclidian hierarchical complete clustering of the HPLC and flow cytometry data of the Lutheran negative and Lutheran weak erythrocytes confirms the correlation between HbF and HbA2 as well as the correlation between individuals with KLF1 variants versus individuals with no KLF1 variants (Fig. 4D). Of note, within the Lu weak/negative donor population, the c.304 T > C polymorphism (rs2072597) shows a distribution of 45.5% for the homozygous TT allele, 47.3% for the heterozygous TC allele and 7.3% for the homozygous CC in concordance with frequencies observed in the European population ( Supplementary Fig. 5). This polymorphism did not show a correlation with red blood cell membrane markers CD44, band 3, Lu(a), Lu(b), HbF or HbA2 expression levels. This is in agreement with the international society of blood transfusions workshop on red cell immunogenetics and blood group terminology which termed this particular KLF1 variant obsolete concerning In(Lu) 25 .  Fig. 6). In addition, Western blotting showed expression of c.704dupT and c.813G > A, however, c.1042 T > C was barely detectable (Supplementary Fig. 6B). However, as the KLF1 mutants were expressed using a lentiviral   (Fig. 5A,D,G and Supplementary Fig. 7A,D). KLF1 knockdown results in decreased expression of CD44 and BCAM, confirming transcriptional control by KLF1 (Fig. 5B,C,E,F, Supplementary Fig. 7B,C). Re-expression of KLF1 WT rescues CD44 and Lu expression but not to untreated or scrambled shRNA control sample levels. Expression of CD44 and Lu upon over-expression of KFL1 variants c.704dupT and c.813G > A comparable to knockdown of KLF1 and thus did not rescue. The c.1042 T > C variant led to an intermediate rescue of CD44 and Lu expression. Of note, the percentage of GFP positive cells did not change over the course of differentiation suggesting that KLF1 overexpression nor knockdown did not lead to a competitive disadvantage in the first 72 h of differentiation ( Supplementary Fig. 7E). Upon further differentiation (120 h), the erythroid cells in which the endogenous KLF1 was knocked down showed a low percentage of viable cells and a further loss of CD44 expression com-   Fig. 8A-D), which confirmed the crucial role of KLF1 during erythropoiesis.

Expression of mutant KLF1 in primary human erythroblasts recapitulates the HPFH phenotype. Next we evaluated if expression of KLF1 variants interfered with hemoglobin subunit expression and/
or resulted in elevated HbF expression. HBG1/2 and HBB1 protein levels were measured on different time points of EBL differentiation using FCM (Fig. 6A). Expression of a scrambled shRNA did not lead to increased HBG1/2 expression in erythroblasts. However, erythroblasts transduced with shRNA against KLF1 show a significant increase of HBG1/2 expression during differentiation indicating the prominent role of KLF1 in repressing gamma globin expression (Fig. 6B). Re-expression of wild type KLF1 led to a reduction HBG1/2. Erythroblasts expressing the c.1042 T > C, c.813G > C and c.704dupT variants have higher averages of HBG1/2 when compared to EBL treated with scrambled RNA or wild KLF1. Only c.704dupT variant demonstrated significantly higher HBG1/2 levels. These data indicate that specific expression of KLF1 mutants in primary erythroblasts can cause HPFH.

Discussion
Here we described novel variants in the KLF1 gene that result in HPHF and in all cases causes dominant In(Lu). Interestingly, FCM revealed that Lu positive donors always display a negative population, which explains the mixed field observed in Lu(a + b +) donors upon agglutination based serology. The ratio is different between individuals but maintained upon erythrocyte aging. Stringent gate settings aid separating Lutheran negative individuals from extremely low expressing individuals thereby facilitating selection of better matched blood upon transfusion when Lutheran negative blood is necessary. Of note, since In(Lu) individuals are characterized by carrying not only reduced expression of antigens in the Lutheran system but also significantly reduced P1, In b , and AnWj antigens, alloimmunization by these antigens may also not be expected 2 . However, the thresholds set by FCM here, may still not discriminate extremely low Lutheran expressing RBC from total negative erythrocytes. Thus, although In(Lu) blood donors may score Lu(a-b-) on serology and FCM tests, they may still contain traces of Lu(a +)/Lu(b +). Importantly however, the thresholds used here segregated the Lu negative individuals with KLF1 variants from low Lu expressing individuals without KLF1 variants. It should be noted however, that transfusion dependent recessive Lu(a− b−) patients might have a risk of developing alloimmunization when receiving blood from In(Lu) donors, however the chance of this actually occurring is extremely low. In 24 out of 30 Lu negative individuals (80%) variants in KLF1 were found, these individuals can thus be classified as In(Lu), and, the frequency found here is comparable to that of other studies 28 . To explain the Lu negative phenotype in the remaining 6 out of 30 individuals, more research needs to be done. A possible explanation may come from homozygous autosomal-recessive mutations in the BCAM gene, or hemizygosity for a recessive X-linked GATA1 mutations that have previously been shown cause Lu negative phenotypes 6 . From all individuals harboring KLF1 variants, 9 (c.704dupT, c.813G > A, c.862A > G, c.1001C > T, c.1057C > T, c.1003G > A, and in two out of 13 c.1060C > G cases) show elevated levels of HbF. Of these, all class 3 variants and both individuals carrying the c.1057C > T variant in different donors, showed elevated HbF, indicating that this is a mutant specific effect. 20 individuals display elevated HbA2 (> 2.8%). Indeed, class 2 and 3 KLF1 variants have been shown to correlate with elevated HbA2 expression 24,] 29 . According to the recently proposed KLF1 mutant classes 1 , we have identified 4 novel class 2 and 2 novel class 3 variants. Over-expression of a selection of mutants in a KLF1 knockdown setting recapitulated the observed phenotypes in KLF1 mutant donors, providing a model to study www.nature.com/scientificreports/ specific KLF1 mutants in the same genetic background and underscoring the role that KLF1 plays in regulation of erythroid surface proteins and specific globin expression. KLF1 functions in a large range of processes during erythropoiesis and specific variants within KLF1 result in a range of different phenotypes and diseases 1 . This underscores the pleiotropic effect of KLF1 in wt and mutant settings. In addition, KLF1 is also involved in regulating globin expression either directly through transcriptional activation of the beta globin gene or indirectly through repression of gamma globins, e.g. through BCL11A or LRF 10,11,30 . Haploinsufficiency or expression of KLF1 variants may induce persistence of fetal hemoglobin. Interestingly, families with the same KLF1 mutant display significant differences HbF levels (range between 2 and 20%), indicating additional contributing factors independent or dependent on KLF1 31 . Here we also observed elevated HbF levels in some but not for all In(Lu) class 2 variants underscoring either the pleiotropic effect of these different mutants or the confounding consequence of other genetic variation between the In(Lu) individuals 1 . Due to genetic variation it is relatively difficult to segregate the mild effect of other specific KLF1 variants on HbF regulation from other confounders. As shown here, the re-expression of KLF1 variants in the same genetic background is useful to compare and understand the contribution of different KLF1 mutants to target gene regulation without completely understanding underlying confounding genetic parameters. For instance, expression of the KLF1 c.1060C > G Zn-finger variant (p.Leu354Val, rs1397962733; found 13 times) presented with elevated HbF in only 2 out of 13 donors and decreased CD44 12 out of 13 donors. Despite its high frequency, it has not been reported before putatively reflecting a founder effect within this Dutch cohort. This missense variant was never bi-allelic and homozygosity may be incompatible with life.
One out of the two class 3 variants (p.Leu236Profs*117) expressed the lowest CD44 levels found in this cohort. This finding indicates that an extremely low CD44 expression may be a variant specific effect for class 3 variants. The class 3 frameshift variant p.Leu236Profs*117, and the premature stop variant p.Trp271X, both result in the loss of the three Zinc-finger domains. The p.Trp271X variant displays the highest increase in HbF. They phenotypically match previously reported class 3 variants, p.Ser270X, p.Pro190LeufsX47 and p.Arg319GlufsX34 2,32 , which described nonsense mediated decay of KLF1 RNA resulting in haploinsufficiency. The effects of missense variants are much harder to predict, for example the heterozygous class 1 p.Ser102Pro (c.304 T > C) variant was found in 26 of 55 times in our cohort and does not seem to have a significant effect on CD44 expression as suggested previously 25   Similarly the c.977 T > G variant showed increased HbA2 but no elevated HbF 34 . As expected no changes in expression of surface proteins Band3 and GPA in the In(Lu) cohort were observed. Even though the class 1 p.Met39Leu (c.115A > C) variant presented with lowered Lu and CD44 expression in this study, it was previously reported as possibly neutral as mouse Klf1 contains a leucine at this position 10 . The compound heterozygous KLF1 promoter variant (c.-148G > A, rs79334031) has been described before to be associated with increased HbF levels 19 . It has previously been shown that CD44, HBB and BCAM are down regulated after in vitro culture of In(Lu) erythroblast 2 , which was a result of decreased binding to the promoters of these genes 12 , however, this was never shown on protein level. By in vitro knock down of endogenous wild type KLF1 and re-expression of mutated KLF1 co-expressed in healthy human erythroblasts we studied the effect of the novel frameshift variant p.(Leu236Profs*117) (c.704dupT), nonsense variant p.Trp271X (c.813G > A), and missense variant p.Phe348Leu (c.1042 T > C). Our in vitro experiments confirmed the in vivo found phenotypes. In contrast to haploinsufficiency, knockdown of KLF1 always resulted in similar phenotypes irrespective of the donor, indicating that genetic variation is of minor influence upon significant KLF1 knockdown.
In conclusion, findings described here suggest that all found variants lead to BCAM negativity and decreased CD44. Class 3 variants showed an intensified decrease of CD44. Two out of the eight type 2 variants and all class 3 variants resulted in HPFH, indicating the variant specific effects. The c.1060C > G variant showed variability between individuals concerning hemoglobin expression and blood group distribution, indicating that undetermined donor differences play a role. Our findings were confirmed by ectopic expression of specific KLF1 mutants that did not rescue BCAM and CD44 expression upon knocking down endogenous KLF1, indicating that these mutants were responsible for the In(Lu) phenotype. We found that, ectopic mutant expression gave a higher HbF level compared KLF1 knock down conditions. This effect can be further explained through very high HbF levels (> 70%) found in a child where both parents were carriers of class 3 KLF1 variants 35 . Furthermore, in previously reported mouse studies, where the combination of KLF1 haploinsufficiency and a complete absence of BCL11A resulted in γ-globins are not expressed to the full extent 36 .
Ectopic expression of KLF1 mutants in 3 different donors showed no confounding effects. The fact that only 1 patient has been described to date that is null for KLF1 displaying severe anemia and complete transfusion dependency indicates the importance of KLF1 in erythropoiesis 35 . It is intriguing to observe such differential effects of KLF1 haplo-insufficiency on a selection of target genes like CD44 and BCAM, while clearly not detrimental to the donor or affecting expression of other known KLF1 target genes 37 . Whether this can all be explained by lowered binding affinity of KLF1 for promoter regions or that competition at the level of cofactor recruitment plays a role remains to be investigated. Using over-expression of specific mutants in human pro-erythroblasts as shown here may help to uncover the underlying mechanisms.
In conclusion, we have found 6 novel KLF1 variants that lead to In(lu) phenotype. These mutants have subtle but clear effects on the expression of globin subunits and CD44 in erythrocytes, which were confirmed during in vitro erythropoiesis for KLF1p.(Leu236Profs*117) (c.704dupT), p.Trp271X (c.813G > A), and p.Phe348Leu (c.1042 T > C). Finding novel KLF1 mutations that lead to in(lu) is important in blood group profiling of individuals.