Molecular Analysis of the Rare In(Lu) Blood Type: Toward Decoding the Phenotypic Outcome of Haploinsufficiency for the Transcription Factor KLF1

ABSTRACT

KLF1 encodes an erythroid transcription factor, whose essential function in erythropoiesis has been demonstrated by extensive studies in mouse models. The first reported mutations in human KLF1 were found in individuals with a rare and asymptomatic blood type called In(Lu). Here, we show that KLF1 haploinsufficiency is responsible for the In(Lu) blood type, after redefining this peculiar blood type using flow cytometry to quantify the levels of BCAM and CD44 on red blood cells. We found 10 (seven novel) heterozygous KLF1 mutations responsible for the In(Lu) blood type. Although most were obligate loss-of-function mutations due to the truncation of the DNA-binding domain of KLF1, three were missense mutations that were located in its DNA-binding domain and impaired the transactivation capacity of KLF1 in vitro. We further showed that the levels of the hemoglobin variants HbF and HbA₂ were increased in the In(Lu) blood type, albeit differently. The levels of the membrane glycoproteins BCAM and CD44 were also differently reduced on In(Lu) red blood cells. This biochemical and genetic analysis of the In(Lu) blood type tackles the phenotypic outcome of haploinsufficiency for a transcription factor.

Introduction

Blood types result from genetic polymorphisms at different loci. Identifying these polymorphisms has allowed important improvements in transfusion safety and obstetrics, especially with the recent development of high-throughput platforms for blood group genotyping. Elucidating the genetic basis of a blood type has also sometimes provided unexpected insights on the physiological function of a gene in humans. For instance, the water channel gene AQP1 (MIM# 107776) had been postulated to be essential until the complete deficiency of AQP1 was found responsible for the rare Co(a-b-) blood type that has no clinical consequences [Preston et al., 1994]. More recently and as detailed below, this was also the case for the transcription factor gene KLF1 (MIM# 600599, formerly called EKLF), whose essential function in erythropoiesis has been established in mouse models (for a review, see [Siatecka and Bieker, 2011]).

The first investigations about the Lutheran blood type system (abbreviated as Lu) go back to the Second World War, with the identification in 1944 of an antibody against a “new” blood group antigen (Lu^a) in the serum of a polytransfused patient [Callender and Race, 1946]. It took more than a decade before the first example of an antibody against the antithetical and high-incidence antigen, Lu^b, was identified [Chanarin and Cutbush, 1956], but now the Lutheran system officially comprises 17 other antigens [Daniels, 2009]. In 1961, the immunohematologist Dr. M. N. Crawford fortuitously discovered in herself and in three generations of her family, a dominant Lu(a−b−) blood type that was subsequently referred to as the In(Lu) phenotype [Crawford et al., 1961]. In(Lu) was originally the name given for a dominant allele of a then unknown gene that inhibits the expression of the Lu^a and Lu^b antigens, whose genetic basis was still unknown. The molecular basis of the antithetical Lu^a and Lu^b antigens was only elucidated 36 years later: it consists in a nonsynonymous SNP in BCAM (MIM# 612773) that encodes a membrane glycoprotein with five immunoglobulin-like domains [El Nemer et al., 1997; Parsons et al., 1997]. It was another 11 years before Singleton and colleagues identified different heterozygous KLF1 mutations in most of the In(Lu) kindreds they analyzed [Singleton et al., 2008], and these authors concluded that the In(Lu) blood type (MIM# 111150) was most likely caused by inheritance of a loss-of-function mutation in one allele of KLF1. It is important to note that the In(Lu) blood type is not associated with any blood disorder.

Soon after this first report of KLF1 mutations in humans, Borg and colleagues identified an obligate loss-of-function allele of KLF1 (p.[Met39Leu; Lys288Ter]) that cosegregated with elevated levels of fetal hemoglobin (HbF, α₂γ₂) in the adults of one large family from Malta Island [Borg et al., 2010]. HbF is normally reduced to very low levels in adults (<1.0% of total hemoglobin, which is mainly constituted of HbA, α₂β₂) but may persist at higher levels with no adverse clinical consequences. Actually, elevated levels of HbF are known to lessen the symptoms of some hemoglobinopathies, such as β-thalassemia and sickle cell disease, and hence are of great clinical relevance. Borg and colleagues found the KLF1 allele p.[Met39Leu; Lys288Ter] only at the heterozygous state, which was fully consistent with the absence of obvious consanguinity in this Maltese family, and succeeded in providing a rational link between haploinsufficiency for KLF1 and elevated levels of HbF: KLF1 is an activator of BCL11A (MIM# 606557) that encodes a repressor of the γ-globin genes [Borg et al., 2010]. Since then, targeted sequencing of KLF1 in patients with elevated levels of HbF identified several other heterozygous KLF1 mutations [Gallienne et al., 2012]. This corroborates the hypothesis that haploinsufficiency for KLF1 may be a relatively frequent cause of hereditary persistence of fetal hemoglobin (HPFH), even though most identified mutations were missense and thus not obligate loss-of-function mutations.

The initial goal of this study was to determine whether the individuals with the In(Lu) blood type exhibit elevated levels of HbF as heterozygous KLF1 mutations had been associated on one hand with the In(Lu) blood type [Singleton et al., 2008] and, on the other hand, with elevated levels of HbF [Borg et al., 2010]. This data would be extremely valuable for physicians interpreting hemoglobin analyses, but also for researchers looking for therapeutic strategies to increase the levels of HbF to alleviate the severity of sickle cell disease or β-thalassemia. We quickly realized that In(Lu) blood samples were not clearly identified in our and other blood collections, and actually were not distinguished from those with the Lu_null phenotype that results from null mutations in BCAM [Karamatic Crew et al., 2007]. The Lu_null blood type (MIM# 247420) is also referred to as the recessive Lu(a−b−) blood type, in contrast with the In(Lu) blood type, that is, the dominant Lu(a−b−) blood type. Hence, we decided to first redefine the In(Lu) blood type using flow cytometry of red blood cells (RBCs) in order not to analyze Lu_null samples instead of In(Lu) samples. We then identified and characterized several KLF1 mutations responsible for the In(Lu) blood type. Finally, we evaluated the levels of HbF associated with the In(Lu) blood type, as well as the levels of the adult hemoglobin variant 2 (HbA₂, α₂δ₂) that are normally <3.5% of total hemoglobin in adults. Thus, we ended up performing an analysis of the In(Lu) blood type at the biochemical and genetic levels, which allowed a better understanding of this unique phenotype resulting from haploinsufficiency for the KLF1 transcription factor.

Materials and Methods

Blood Samples

This study was conducted according to the ethical standards of the National Institute of Blood Transfusion (INTS, Paris, France). The retrospective study was conducted on historical samples that were cryopreserved in the rare blood collection of the National Reference Center for Blood Groups (CNRGS, Paris, France) to establish the In(Lu) blood type of other subjects and ultimately to identify the genetic basis of the In(Lu) blood type. Fresh blood samples were taken on EDTA after obtaining individual and signed informed consents.

Blood Group Serology

RBC typing was performed by indirect antiglobulin gel test (ID-Micro Typing System Coombs Anti-IgG, DiaMed, BIO-RAD, Hercules, CA, USA) with quality-controlled sera against Lu^a, Lu^b, P1, and AnWj from the serum collection of the National Reference Center for Blood Groups (CNRGS, Paris, France).

Flow Cytometry Analysis

RBCs thawed from frozen blood samples and stored in stabilization solution (ID-CellStab, DiaMed) were washed in Dulbecco's phosphate-buffered saline solution (DPBS, Gibco, Life Technologies headquartered in Carlsbad, CA, USA) and then resuspended at 0.2% in DPBS supplemented with 0.15% bovine serum albumin (BSA, Sigma-Aldrich, St. Louis, MO, USA). As a first step, RBCs were incubated with biotinylated anti-human BCAM (1:100 dilution; R&D Systems headquartered in Minneapolis, MN, USA; cat. BAF148). As a second step, RBCs were incubated with PE-conjugated Streptavidin (1:100 dilution; BD Pharmingen BD Biosciences, San Jose, CA, USA; cat. 554061) and APC-conjugated anti-human CD44 (1:20 dilution; BD Pharmingen, cat. 559942). BCAM and CD44 labelings were analyzed with a FACSCanto II flow cytometer (BD Biosciences) equipped with FACSDiva software (v. 6.1.2) (BD Biosciences). Ten thousand RBCs, gated on forward scatter (FSC) versus side scatter (SSC), were collected for each sample. Data were processed with FlowJo software (v. 7.2.5) (TreeStar, Ashland, OR, USA).

Plasmid Constructions

An expressed sequence tag cDNA clone encoding full-length human KLF1 (IRAKp961E1684Q; IMAGE ID: 5243387) was obtained from imaGenes GmbH, sequence-verified (identical to NM_006563.3) and subcloned as an EaeI/XbaI fragment into the p3xFLAG-CMV-10 mammalian expression vector (Sigma-Aldrich) cut NotI/XbaI. The plasmids corresponding to KLF1 L326R, K288E, and H357Q were constructed by replacing the KpnI/XbaI fragment of p3xFLAG-CMV-10 KLF1 wildtype with the corresponding mutated and fully sequenced fragments, which were generated by site-directed mutagenesis of the IMAGE cDNA. To construct the human BCAM promoter–reporter plasmid, around 1 kb of BCAM gene upstream of the initiation codon was amplified from a gDNA clone obtained from imaGenes GmbH (IMGSB737E121026D; original name: CTD-2608C5), and cloned as a HindIII/PciI fragment (the HindIII site is located −1.085 bp upstream the initiation codon and the PciI site was introduced by PCR at the initiation codon) into the pGL4.12 luciferase reporter vector (Promega, Madison, WI, USA) cut HindIII/NcoI. Sequence files of all the DNA constructs can be obtained upon request.

Promoter–Reporter Assay

COS7 cells were cotransfected with a p3xFLAG-CMV-10-based KLF1 construct and the pGL4.12-based BCAM promoter-firefly luciferase construct, using TransIT-2020 Transfection Reagent (Mirus Bio LLC, Madison, WI, USA). After 24 hr, cells were washed in cold DPBS and lyzed in 1X Passive Lysis Buffer (Promega) with one freeze–thaw cycle. Luciferase activities were measured with a Luciferase Reporter Assay System (Promega) in a single-sample 1250 luminometer (Bio-Orbit Oy, Tuku, Finland). Expression of Flag-KLF1 constructs was analyzed by Western blot using equal amounts of the aforementioned cell lysates, anti-Flag (1:7,500 dilution; Sigma-Aldrich cat. F3165) and anti-GAPDH (1:7,500 dilution; Millipore, Billerica, MA, USA, cat. MAB374).

Hemoglobin Analysis

Hemoglobin analysis was performed by HPLC using a clinical Variant II analyzer and the built-in CDM 5.1 software (Bio-Rad) at the Hôpital Henri Mondor (Créteil, France).

Mutation Detection and Analysis

Genomic DNA was isolated from frozen blood samples using the automated MagNA Pure Compact System (Roche, Basel, Switzerland) and from fresh blood samples using the manual Wizard Genomic DNA Purification Kit (Promega). The different gDNA fragments of interest were amplified by PCR as mentioned below. PCR products were then treated with ExoSAP (GATC Biotech, Konstanz, Germany) and sequenced by ABI BigDye terminator chemistry (GATC Biotech) using the after-mentioned primers. All primers are described in Supp. Tables S1 and S2. Sequencing data were analyzed with DNA Workbench software (CLC bio, Aarhus, Denmark) and Mutation Surveyor software (SoftGenetics, State College, PA, USA). Mutation numbering relates to the reference sequence of the corresponding mRNA with the +1 position referring to the first nucleotide of the initiation codon. All KLF1 mutations reported in this paper were submitted to the KLF1-specific Leiden Open Variation Database (http://www.lovd.nl/KLF1).

KLF1 (NG_013087.1; NM_006563.3) was amplified by PCR as a single fragment using the KLF1–36 and KLF1–3 primers. Mutations were screened by sequencing the corresponding PCR product with the KLF1–37, KLF1–7, KLF1–34, and KLF1–13 primers, and heterozygous deletion/insertion mutations were confirmed by sequencing with the KLF1–4, KLF1–6, KLF1–11, and KLF1–12 primers.

BCAM (NG_007480.1; NM_005581.3) was amplified by PCR as two fragments using the BCAM-8 and BCAM-12 primers, and the BCAM-16 and BCAM-28 primers. The BCAM-8/12 PCR product was sequenced using the BCAM-19, BCAM-21, BCAM-24, BCAM-25, and BCAM-7 primers. The BCAM-16/28 PCR product was sequenced using the BCAM-22, BCAM-23, BCAM-26, and BCAM-29 primers.

rs7482144 (HBG2; NG_000007.3) was analyzed by XmnI RFLP of a 666-bp PCR product amplified with the 5′GG1 and 3′AG1 primers [Craig et al., 1993].

rs28384513 (HBS1L-MYB intergenic region; NC_000006.11) was analyzed by sequencing, with the rs28–5 primer, a 454-bp PCR product amplified with the rs28–3 and rs28–1 primers.

rs9399137 (HBS1L-MYB intergenic region; NC_000006.11) was analyzed by sequencing, with the rs93–4 primer, a 363-bp PCR product amplified with rs93–3 and rs93–2 primers.

rs4671393, rs11886868, and rs766432 (BCL11A; NG_011968.1) were analyzed by sequencing, with the rs46–1 and rs76–4 primers, a 1,480-bp PCR product amplified with the rs46–2 and rs76–3 primers.

The P1/P2 blood group polymorphism was analyzed by NlaIII RFLP as previously described [Thuresson et al., 2011].

Statistical Analysis

The Mann–Whitney U test was used to assess the difference between data sets, and box-and-whisker diagrams were generated to interpret the distribution of data within a group using StatEl software (ad Science, Paris, France).

Results

Expression of BCAM and CD44 on In(Lu) RBCs

The RBCs of the individuals with the In(Lu) blood type usually seem to lack the Lu^a and Lu^b blood group antigens when tested by standard agglutination techniques, as those of the individuals with the Lu_null blood type. Thus, In(Lu) and Lu_null RBCs can be confused with each other because they exhibit the same Lu(a−b−) phenotype. Nevertheless, In(Lu) RBCs bind selected antibodies to the Lu^b antigen in adsorption–elution tests, suggesting that BCAM is reduced, but not completely absent, on In(Lu) RBCs. Of note, binding of antibodies to the Lu^a antigen is usually not investigated by this technique due to the low incidence of this antigen.

To evaluate the levels of BCAM on In(Lu) RBCs, we performed a flow cytometry analysis of these RBCs labeled with an antibody to BCAM that recognized all BCAM isoforms, independently of the single amino acid polymorphisms associated with the different Lutheran antigens. First of all, we analyzed RBCs from the princeps case with the In(Lu) blood type [Crawford et al., 1961], which were available in our collection of frozen blood samples. The vast majority of her RBCs actually showed a complete absence of BCAM reactivity by flow cytometry analysis, and only a very small fraction of her RBCs appeared to express BCAM (Fig. 1A, top left panel). Of note, all the In(Lu) RBCs analyzed in this study (thawed or fresh samples) showed a similar BCAM reactivity, with a tiny but reproducible “shoulder” population of RBCs expressing BCAM. Even though unexpected, this BCAM reactivity was fully consistent with the Lu(a−b−) phenotype of In(Lu) RBCs when determined by standard agglutination tests, as well as their Lu(b^weak) phenotype if eventually determined by adsorption–elution tests that are more sensitive. Furthermore, this BCAM reactivity is likely related to the well-known, but also unexplained, intra-individual variability of BCAM expression on normal RBCs that distinguishes BCAM between all blood group carrier proteins (Fig. 1A, bottom left panel).

Figure 1. Flow cytometry analysis of BCAM and CD44 on In(Lu) RBCs. A: Histograms of reference In(Lu) RBCs (top panels) and control RBCs (bottom panels) colabeled with anti-BCAM (red) and anti-CD44 (blue); the black overlays correspond to unlabeled RBCs. The arrow indicates the characteristic small population of BCAM-expressing RBCs observed in In(Lu) RBC samples. B: Plot of 28 potentially In(Lu) RBCs (triangles) and 18 control RBCs (circles) using the geometric mean of fluorescence intensity (GMFI) of BCAM labeling (axis) and CD44 labeling (ordinate); RBCs were colabeled with anti-BCAM and anti-CD44 as above. The quadrant gates three clearly distinct BCAM/CD44 profiles: BCAM^low/CD44^low (Q1), BCAM^low/CD44^high (Q2), and BCAM^high/CD44^high (Q3); only the RBCs with a BCAM^low/CD44^low profile (highlighted in purple) corresponded to In(Lu) individuals, as confirmed by KLF1 sequencing.

The In(Lu) blood type is characterized by weakened or suppressed expression not only of Lutheran antigens, but also of several high-incidence antigens that are not carried by BCAM [Crawford et al., 1974; Daniels et al., 1986], especially AnWj and P1 as exemplified in Figure 2 (members I.1 and II.1 are In[Lu]). Flow cytometry analysis of CD44, which carries the AnWj antigen, showed that In(Lu) RBCs had reduced, but significant, levels of CD44 (Fig. 1A, top right panel). In contrast, no reduced expression of CD44 was observed in Lu_null RBCs [Spring et al., 1988]. Therefore, the combined analysis of BCAM and CD44 by flow cytometry could better discriminate In(Lu) from Lu_null RBCs. We thus analyzed the cryopreserved RBCs of all the individuals who were registered with a Lu(a−b−) or Lu(a−b^weak) phenotype at the French National Reference Center for Blood Groups (CNRGS, Paris, France). These 28 potentially In(Lu) individuals represented 25 kindreds. Out of these 28 RBC samples, 23 had a BCAM^low/CD44^low profile (Q1 in Fig. 1B) and 2 had a BCAM^low/CD44^high profile (Q2 in Fig. 1B), and most likely corresponded to 23 In(Lu) and 2 Lu_null individuals, respectively. This result was consistent with the higher incidence of the In(Lu) than Lu_null blood type [Daniels, 2009]. We also found 3 Lu(a−b^weak) RBC samples that had a BCAM^high/CD44^high profile like control Lu(a−b+) RBC samples (Q3 in Fig. 1B). The Lu(a−b^weak) phenotype of these 3 RBC samples was double-checked and was indeed inconsistent with their high expression of BCAM. The genetic cause of this discrepancy is currently unknown; so far, no causative mutations were identified in BCAM, KLF1 and GATA1.

Figure 2. Pedigree analysis of an In(Lu) family. Fresh blood samples of each family member were serologically typed for the blood group antigens Lu^a, Lu^b, P1, and AnWj, as indicated with the color code, and analyzed for HbF and HbA₂ content. Of note, member I.1 was genotyped LUB/LUB, P2/P2; member I.2 was genotyped LUA/LUB, P1/P2; member II.1 was genotyped LUB/LUB, P1/P2; member II.2 was genotyped LUA/LUB, P2/P2; member II.3 was genotyped LUB/LUB, P1/P2. The arrow indicates the propositus while the black frames highlight the two family members with the In(Lu) blood type resulting from the heterozygous KLF1 mutation p.Tyr197Ter.

KLF1 Mutations Causing the In(Lu) Blood Type

We managed to extract genomic DNA from 13 cryopreserved RBC samples that showed a BCAM^low/CD44^low profile, and we first sequenced KLF1 in the 10 corresponding propositi. In each of these unrelated individuals with a BCAM^low/CD44^low profile, we found one heterozygous KLF1 mutation that was obligate or potential loss-of-function (Supp. Table S3A), and that was absent in the NCBI dbSNP database (build 135) as well as in 100 French blood donors (Supp. Table S3A). Surprisingly, none of these 10 propositi carried the same heterozygous KLF1 mutation, suggesting a wide variety of mutations causing the In(Lu) blood type while confirming that these 10 propositi were unrelated. We found six frameshift mutations (c.310_311insG, p.Ala104GlyfsTer249; c.519_520insC, p.Gly174ArgfsTer179; c.519_525dupCGGCGCC, p.Gly176ArgfsTer179; c.569delC, p.Pro190LeufsTer47; c.663delG, p.Leu222SerfsTer15; c.954_955insG, p.Arg319GlufsTer34), one nonsense mutation (c.591C>G, p.Tyr197Ter) and three missense mutations located in the zinc-finger domain of KLF1 (c.862A>G, p.Lys288Glu; c.977T>G, p.Leu326Arg; c.1071C>A, p.His357Gln) (Supp. Table S3A). To the best of our knowledge, only c.569delC, c.954_955insG, and c.977T>G were reported previously, as mentioned in the LOVD database for KLF1. We then sequenced BCAM in the 10 In(Lu) propositi to verify the absence of silencing mutations in BCAM (Supp. Table S4). We also confirmed the presence of these KLF1 mutations in the In(Lu) relatives of the few families available (n = 3). Altogether, these data indicated that the aforementioned KLF1 mutations were responsible for the In(Lu) blood type. On the contrary, the KLF1 mutations c.304T>C, p.Ser102Pro (rs2072597), c.311C>T, p.Ala104Val (rs182276666), and c.544T>C, p.Phe182Leu (rs2072596), as well as g.4813C>G (rs3817621) and g.4916G>A (rs79334031) in the proximal promoter of KLF1 are relatively frequent in the general population (Supp. Table S3A) and thus cannot be responsible for the rare In(Lu) phenotype. Of note, the p.Ser102Pro mutation was present in 7 out of 10 In(Lu) propositi, suggesting a potential bias. However, haplotype analysis revealed that this mutation was on the KLF1 allele carrying the In(Lu) blood type causing mutation only in 4/7 cases (Supp. Table S3B), which was fully consistent with the incidence of the p.Ser102Pro mutation in the general population (Supp. Table S3A). Hence, we concluded that there is no association of the p.Ser102Pro mutation with the In(Lu) blood type.

Among the 10 KLF1 mutations causing the In(Lu) blood type that we found, three resulted in a single amino acid change (p.Lys288Glu (K288E), p.Leu326Arg (L326R), and p.His357Gln (H357Q)) in the DNA-binding domain of KLF1 that consists in three zinc fingers (Fig. 3A). Each of these missense mutations affected a different zinc finger and was predicted to be “probably damaging” by the PolyPhen-2 software (v.2.1.0). To confirm that these three missense mutations impaired the transactivation capacity of KLF1, we performed a promoter–reporter assay by cotransfecting cells with a plasmid encoding firefly luciferase under the control of BCAM promoter and a plasmid encoding Flag-tagged KLF1 wildtype, K288E, L326R, or H357Q. Of note, the 1 kb proximal promoter of BCAM, which was used in this promoter-reporter assay, contains several potential KLF1 binding sites (14 CCNCNCCCN sites [Feng et al., 1994] and 6 CCMCRCCCN sites [Tallack et al., 2010]), strongly suggesting that BCAM was a target gene of KLF1. All KLF1 constructs were expressed at the same level, which indicated that the three missense mutations did not alter the stability of KLF1 (Fig. 3B). However, the three point mutants K288E, L326R, or H357Q showed a reduced transcriptional activity when compared with KLF1 wildtype (Fig. 3B). Together, these results confirmed that these three missense mutations induced a loss of KLF1 function, such as the frameshift or nonsense mutations that truncated its DNA-binding domain (Fig. 3A).

Figure 3. KLF1 mutations causing the In(Lu) blood type. A: Diagram of the KLF1 transcription factor that shows the location and the predicted amino acid change of the heterozygous mutations causing the In(Lu) blood type in 10 unrelated individuals (the corresponding nucleotide changes can be found in Supp. Table S3A); the three zinc fingers (ZF) composing the DNA-binding domain of KLF1 are highlighted in black. B: BCAM promoter–reporter assay in COS7 cells that shows the loss of KLF1 function associated with the missense mutations p.Leu 326Arg (L326R), p.Lys288Glu (K288E), and p.His357Gln (H357Q) causing the In(Lu) blood type; the mutation p.Glu325Lys (E325K) has been also studied. COS7 cells were cotransfected with a BCAM promoter-luciferase construct and the indicated Flag-tagged KLF1 constructs; the results are shown as means ± SD (n = 4) of relative light units (RLU). The bottom panel shows the expression of exogenous Flag-tagged KLF1 constructs, as well as endogenous GAPDH, at the time of the luciferase assay (top panel).

We also studied a previously described missense mutation in the second zinc finger of KLF1, c.973G>A, p.Glu325Lys (E325K), which is not responsible for the asymptomatic In(Lu) blood type contrary to the three previous missense mutations, but that is responsible for congenital dyserythropoiteic anemia type IV (CDAN4; MIM# 613673) [Arnaud et al., 2010; Singleton et al., 2009]. Flag-tagged KLF1 E325K was able to transactivate BCAM promoter as well as, and even better than the wildtype transcription factor (Fig. 3B), in line with the fact that CDAN4 is not associated with a reduced levels of BCAM on RBCs [Arnaud et al., 2010; Parsons et al., 1994]. This result demonstrated that a missense mutation in the zinc-finger domain of KLF1 did not necessarily induce a loss of its transactivation capacity, and hence substantiated the relevance of the results obtained for K288E, L326R, and H357Q.

Expression of HbF and HbA₂ in In(Lu) RBCs

To evaluate the levels of HbF associated with the In(Lu) blood type, we performed a standardized Hb fraction analysis by high-performance liquid chromatography (HPLC) in the cryopreserved RBCs of the 10 aforementioned In(Lu) propositi. The In(Lu) blood type of all these individuals has been ascertained by the presence of an heterozygous loss-of-function KLF1 mutation (see above). As shown in Figure 4(A), the In(Lu) individuals showed increased, albeit variable, levels of HbF (mean 2.14 ± 0.97% [n = 10]) when compared with control blood donors (mean 0.63 ± 0.12% [n = 10]), as well as with the normal range (<1.0%). A similar result was obtained with fresh RBCs taken from the only informative In(Lu) family available for this study (Fig. 2). Toward confirming the effect KLF1 haploinsufficiency on the levels of HbF in adults, we analyzed six SNPs commonly associated with HPFH: rs7482144 located in the promoter of HBG2 [Gilman and Huisman, 1985], rs28384513 and rs9399137 located in the intergenic region HBS1L-MYB [Thein et al., 2007], as well as rs4671393, rs11886868, and rs766432 located in the second intron of BCL11A [Menzel et al., 2007; Uda et al., 2008]. We found no significant correlation between the status of these SNPs and the increased levels of HbF observed in the In(Lu) individuals (Supp. Table S5). We concluded from all these data that the In(Lu) blood type resulting from KLF1 haploinsufficiency was associated with increased levels of HbF.

Figure 4. HPLC analysis of HbF and HbA₂ in In(Lu) RBCs. Box-and-whisker diagram of the levels of HbF (A) or HbA₂ (B) in 10 In(Lu) and 10 control RBC samples; lower and upper boundaries of boxes represent the 25th and 75th percentiles, horizontal lines the medians, whiskers the minimum, and maximum nonatypical values; P by Mann–Whitney U test.

Hemoglobin fraction analysis by HPLC has the advantage to quantify HbF and HbA₂ at the same time. Thus, we observed that the levels of HbA₂ were slightly, but statistically significantly, increased in our cohort of In(Lu) individuals (mean 3.12 ± 0.28% [n = 10]) when compared with control blood donors (mean 2.59 ± 0.21% [n = 10]) (Fig. 4B). The levels of HbA₂ that we measured in In(Lu) individuals are considered as borderline levels because the upper limit of the normal range is 3.5%. Furthermore, the distribution of the increased levels of HbA₂ in In(Lu) individuals was clearly different from that of the increased levels of HbF, indicating that KLF1 haploinsufficiency has not the same influence on the expression of these two Hb variants.

Expression of P1 Antigen in In(Lu) RBCs

Finally, we decided to investigate the suspected influence of the In(Lu) blood type on the expression of the P1 blood group antigen. Suppressed or weakened expression of the P1 antigen on In(Lu) RBCs had been noted early [Crawford et al., 1974] but precise studies were almost impossible until the recent discovery of the genetic basis of the P1/P2 blood group polymorphism [Thuresson et al., 2011]. Indeed, an In(Lu) individual can exhibit the P1 phenotype because he is simply homozygous for the P2 allele as approximately 20% of the general population (this was actually the case for the propositus of the In(Lu) family depicted in Figure 2). We determined the P1 phenotype and P1/P2 genotype of the ten aforementioned In(Lu) propositi, as well as of the members of the In(Lu) family depicted in Figure 2. We found that nine In(Lu) individuals did not express P1 while being heterozygous for P1, and that one expressed only weakly P1 while being homozygous for P1. This result corroborated that the expression of P1 is suppressed in the In(Lu) blood type.

Discussion

Predicting the phenotypic outcome of haploinsufficiency for a transcription factor is almost impossible. When Singleton and colleagues proposed that the In(Lu) blood type was most likely caused by inheritance of a loss-of-function mutation in one allele of KLF1 [Singleton et al., 2008], In(Lu) RBCs appeared as a paradigm. Indeed, blood is one of the rare tissues that can be collected in vivo with a minimally invasive technique, and RBCs are a relatively simple cellular model system.

Although the In(Lu) blood type has been studied quite intensively by blood group serology (for a review, see [Daniels, 2009]), little was known at the molecular level. In fact, even the identification of this rare blood type remained challenging and its confirmation often required knowing whether it was inherited as a dominant trait. Here, we show that In(Lu) RBCs can be easily and surely identified by double-flow cytometry analysis of BCAM and CD44: In(Lu) RBCs exhibit a unique BCAM^low/CD44^low profile. Of note, this analysis relies on commercially available antibodies and thus circumvents the usual limitations of sera to blood group antigens in terms of availability and standardization. Out of 28 RBC samples that we analyzed by flow cytometry after they were suspected to be In(Lu) by serology, we found five RBC samples that did not exhibit a BCAM^low/CD44^low profile. These five RBC samples were not In(Lu) as confirmed by the absence of KLF1 mutations. Similarly, Singleton and colleagues found no KLF1 mutations in 3 out of 24 RBC samples that were expected to be In(Lu) even though the authors did not mention their criteria of selection [Singleton et al., 2008]. Analyzing the BCAM/CD44 profile of these three RBC samples by flow cytometry would likely indicate that they are actually not In(Lu). It is worth mentioning that our flow cytometry analysis of BCAM and CD44 discriminates the In(Lu) phenotype not only from the Lu_null phenotype but also from another, yet uncharacterized, Lutheran phenotype that can be confused with the In(Lu) phenotype due to the weakened expression of the Lutheran antigens.

The 10 unrelated RBC samples with the In(Lu) phenotype (i.e., with a BCAM^low/CD44^low profile) that we managed to sequence had a heterozygous loss-of-function mutation in KLF1, actually different in each RBC sample, arguing that the In(Lu) blood type results from haploinsufficiency for KLF1. Although a loss of function can be easily predicted for a nonsense or frameshift mutation that truncates the DNA-binding domain of KLF1, this is not the case for a missense mutation even though located in a zinc finger of the DNA-binding domain. Therefore, we used a BCAM promoter–reporter assay to demonstrate that the three “In(Lu)” mutations p.Lys288Glu, p.Leu326Arg, and p.His357Gln caused a loss of function of KLF1. However, we noted that the three corresponding point mutants of KLF1 retained a residual activity in this promoter–reporter assay, suggesting that the loss of function created by these three missense mutations was not complete. Alternatively, this result reflected the limitation of this in vitro assay. It should be noted that COS7 cells were used for this BCAM promoter–reporter assay because BCAM promoter was not at all transactivated by exogenously expressed KLF1 in K-562 cells despite their erythroid character.

Whether total or partial, the loss of function created by an “In(Lu)” mutation affects only one allele of KLF1 because the second allele never appeared to be mutated, likely in line with the embryonic lethality of Klf1^−/− mice [Nuez et al., 1995; Perkins et al., 1995]. Thus, at least 50% of KLF1 function is expected to be present in an In(Lu) individual. However, BCAM is completely absent on the vast majority of the RBCs of an In(Lu) individual, suggesting that threshold levels of KLF1 function are required for BCAM gene expression. In contrast, the levels of CD44 are only reduced on In(Lu) RBCs, suggesting that CD44 gene expression does not require threshold levels of KLF1 function or that this threshold is below 50%.

As for BCAM versus CD44, we observed a different effect of KLF1 haploinsufficiency on the levels of HbF versus HbA₂, even though these two Hb variants are increased in In(Lu) individuals: the levels of HbA₂ are similar and below the upper reference limit (mean 3.12 ± 0.28% [n = 10]) whereas the levels of HbF are variable and above the upper reference limit (mean 2.14 ± 0.97% [n = 10]). Our results are in total agreement with the recent study of Perseu and colleagues [Perseu et al., 2011]. In particular, these authors reported that the KLF1 mutation p.Ser270Ter, which they found to be highly prevalent in the Sardinia Island, is associated with borderline levels of HbA₂ (mean 3.6 ± 0.2% [n = 42]) and slightly elevated levels of HbF (mean 2.1 ± 1.2% [n = 42]) [Perseu et al., 2011]. In contrast, we did not observe HbF > 5% in In(Lu) individuals as reported by Borg and colleagues in their princeps study describing a Maltese family carrying the KLF1 mutation p.Lys288Ter [Borg et al., 2010]. This discrepancy may be explained by the presence of other HbF-modifying gene alleles in this particular Maltese family. Indeed, the same laboratory has recently reported that such elevated levels of HbF were not found in a second Maltese family carrying the same KLF1 mutation [Felice et al., 2011], confirming that KLF1 haploinsufficiency was not directly responsible for the very high levels of HbF observed in the first Maltese family. In summary, although our data corroborate the statement of Borg and colleagues that KLF1 haploinsufficiency causes HPFH, they also indicate that KLF1 haploinsufficiency can not alone account for levels of HbF > 5% in adults.

In conclusion, this study shows that KLF1 haploinsufficiency has not the same effect on the expression of different erythroid proteins, likely reflecting the variable dependence of their respective genes for this transcription factor. We characterized the effect of KLF1 haploinsufficiency on two RBC membrane proteins, BCAM and CD44, and two Hb variants, HbF and HbA₂; but future studies should yield information regarding the spectrum of erythroid proteins whose expression is affected by KLF1 haploinsufficiency. In fact, this first analysis of the In(Lu) blood type at the biochemical level paves the way for further investigation of this rare blood group, which turns out to be a unique opportunity to decipher the phenotypic outcome of the haploinsufficiency for the transcription factor KLF1, and maybe to understand its physiological role(s) in humans.

ACKNOWLEDGMENTS

We thank R. Galanello (Ospedale Regionale per le Microcitemie, Cagliari, Italy) for sharing unpublished results, and M. L. Olsson (Lund University, Lund, Sweden) and M. Goossens (Hôpital Henri Mondor, Créteil, France) for their kind support. We also thank M. Le Gall and G. Nicolas for helpful discussions and critical reading of the manuscript.

Disclosure statement: The authors declare no financial conflict of interest.