|Home | About | Journals | Submit | Contact Us | Français|
L.S., K.H., I.T., M.F. and N.H. designed the experiments; L.S., K.H., A.M., I.P. and R.E. performed the experiments; and A.M. provided subject blood samples; S.U. provided the antibody to Kindlin-3; M.M. provided the EGFP-Kindlin-3 construct;. L.S., K.H., A.M., I.P., M.F., I.T. and N.H. were involved in data analyses. All authors contributed to the writing or editing of the manuscript. N.H. supervised the project and wrote the initial manuscript.
Integrins are the major adhesion receptors of leukocytes and platelets. β1 and β2 integrin function on leukocytes is crucial for a successful immune response and the platelet integrin αIIbβ3 initiates the process of blood clotting through binding fibrinogen1-3. Integrins on circulating cells bind poorly to their ligands but become active after ‘inside-out’ signaling through other membrane receptors4,5. Subjects with leukocyte adhesion deficiency-1 (LAD-I) do not express β2 integrins because of mutations in the gene specifying the β2 subunit, and they suffer recurrent bacterial infections6,7. Mutations affecting αIIbβ3 integrin cause the bleeding disorder termed Glanzmann’s thrombasthenia3. Subjects with LAD-III show symptoms of both LAD-I and Glanzmann’s thrombasthenia. Their hematopoietically-derived cells express β1, β2 and β3 integrins, but defective inside-out signaling causes immune deficiency and bleeding problems8. The LAD-III lesion has been attributed to a C→A mutation in the gene encoding calcium and diacylglycerol guanine nucleotide exchange factor (CALDAGGEF1; official symbol RASGRP2) specifying the CALDAG-GEF1 protein9, but we show that this change is not responsible for the LAD-III disorder. Instead, we identify mutations in the KINDLIN3 (official symbol FERMT3) gene specifying the KINDLIN-3 protein as the cause of LAD-III in Maltese and Turkish subjects. Two independent mutations result in decreased KINDLIN3 messenger RNA levels and loss of protein expression. Notably, transfection of the subjects’ lymphocytes with KINDLIN3 complementary DNA but not CALDAGGEF1 cDNA reverses the LAD-III defect, restoring integrin-mediated adhesion and migration.
Individuals with LAD-III (also called LAD-I variant)10 have Glanzmann’s thrombasthenia-like bleeding problems and LAD-I-like life-threatening infections. Their leukocytes fail to undergo β2 and β1 integrin-mediated adhesion and migration despite normal integrin expression, which is characteristic of the LAD-III disorder. We have investigated the nature of LAD-III lesions in a Maltese subject (family one)11 and two Turkish subjects, one previously described (family two; family five in ref. 12) and the other characterized here for the first time (family three; details in the Methods and Supplementary Fig. 1 online). Given the likelihood that LAD-III in the Turkish and Maltese families is a recessive condition caused by mutations inherited from a common ancestor, we used homozygosity mapping to show the most likely location of the gene responsible for the LAD-III disorder to be a region between 60.6 megabases (Mb) and 65.3 Mb on chromosome 11 (Supplementary Fig. 2 online).
The CALDAGGEF1 gene (chromosome 11q13.1: 64.25-64.27 Mb) lies on the distal boundary of this region. CALDAG-GEF1 is activated through binding of diacylglycerol and Ca2+ and is a guanine exchange factor for Rap1, a GTPase that has an essential role in the activation of integrins (refs. 13,14). The gene encodes two proteins as a result of alternative splicing, a 68-kDa cytosolic form and a 72-kDa form that is membrane localized owing to an additional amino-terminal myristoylated and palmitoylated domain15. Caldag-gef1-deficient mice mimic the LAD-III phenotype, showing Glanzmann’s thrombasthenia-like bleeding problems16 and defective function of neutrophil β1 and β2 integrins17. A C→A base change in the CALDAGGEF1 gene has been described recently in two Turkish subjects with LAD-III9. This mutation, in a putative splice acceptor site for exon 16, is reported to cause a splicing problem leading to a loss of CALDAGGEF1 mRNA and protein expression and is suggested to be responsible for LAD-III.
Sequencing of all 18 exons and intron-exon boundaries of the CALDAGGEF1 gene for the three subjects with LAD-III and their relatives revealed that the Turkish subjects with LAD-III were homozygous for the C→A base change (Fig. 1a and Supplementary Fig. 3 online). The parents of the Turkish subjects were heterozygous, whereas the sister in family two was homozygous for the normal C allele. In contrast, the C→A change was not present in the Maltese family, and no further nucleotide changes were detected in the CALDAGGEF1 gene (data not shown). The observation that a C→A change was present in the two Turkish subjects with LAD-III in this study and also in the previously described two Turkish subjects9, but not in the Maltese subject, suggests that this mutation is exclusive to LAD-III subjects of Turkish origin and implies a Turkish founder effect maintained through consanguineous marriage.
The C→A mutation has been reported to prevent correct splicing of the CALDAGGEF1 transcript, resulting in unstable CALDAGGEF1 mRNA9. Quantitative RT-PCR with probes that recognize either total CALDAGGEF1 mRNA (Supplementary Fig. 4 online), or the larger alternatively spliced form of CALDAGGEF1 mRNA (data not shown), revealed that the subjects with LAD-III all expressed CALDAGGEF1 mRNA, although at somewhat variable levels. Furthermore, all subjects with LAD-III expressed the 72-kDa and 68-kDa CALDAG-GEF1 proteins (Fig. 1b). In summary, we find that the C→A change present in the Turkish subjects had no impact on either CALDAG-GEF1 mRNA or protein levels, in contrast to results previously described9. Finally, the C→A base change was not present in the Maltese family.
To test the possibility that the C→A base change might affect the function of CALDAG-GEF1, we generated Epstein-Barr virus (EBV)-transformed B cells from the subjects with LAD-III and their parents. The parents’ B cells, but not the subjects’ B cells, were able to attach and migrate without added stimulant on the intercellular adhesion molecule-1 (ICAM-1) that is the ligand of the β2 integrin lymphocyte function associated antigen-1 (LFA-1; Fig. 1c). Migration on fibronectin, which is mediated by α4β1 and α5β1, showed the same difference (data not shown). Expression of a transfected human CALDAGGEF1 cDNA construct that includes the membrane localization domain15 had no effect on the migration of the B cells from subjects with LAD-III (Fig. 1c) or on their individual cell tracking patterns (data not shown). A similar negative result was obtained with cDNAs encoding either human or mouse cytosolic CALDAG-GEF1 (data not shown).
We next used interference reflection microscopy (IRM) to investigate more directly the effect of CALDAG-GEF1 on cell adhesion to ICAM-1. Cells from family 3 subject’s mother showed a pattern of close contact with ICAM-1, whereas cells from the family 3 subject were only lightly attached, whether or not they expressed wild-type CALDAG-GEF1 (Fig. 1d). We reached the same conclusion after quantification of the areas of attachment (Fig. 1d). These experiments show that expression of several forms of wild-type CALDAG-GEF1 fails to repair the adhesion and migration defect caused by the LAD-III disorder.
Taken together, these data show that mutation in the CALDAGGEF1 gene is not the cause of the LAD-III disorder in the subjects in our study. Moreover, the properties of CALDAG-GEF1 do not totally conform to the characteristics of LAD-III disorder, in which the functions of β1, β2 and β3 integrins are all substantially compromised. For example, knockdown of CALDAG-GEF1 expression in human T cells affects LFA-1-mediated adhesion to ICAM-1, but not α4β1 adhesion to vascular cell adhesion molecule-1, which is protein kinase C dependent18; in some circumstances the CALDAG-GEF1 and PKC pathways operate separately to activate mouse platelet αIIbβ3 (ref. 19).
KINDLIN3 (the encoded protein is also known as FERMT3, URP2 or MIG-2B) has recently become a prime candidate as a causative gene for LAD-III because Kindlin-3-deficient mice show a Glanzmann’s thrombasthenia-like phenotype20 and have leukocytes with integrin function defects (see Moser et al.21 in this issue of Nature Medicine). The KINDLIN3 gene (chromosome 11: 63.73-63.75 Mb) is located 0.5 Mb from CALDAGGEF1 at chromosome 11q13.1 in the region of shared haplotypes in the LAD-III-affected families in this study (Supplementary Fig. 2). Furthermore, Kindlin-3 is exclusively expressed by hematopoietically-derived cells22. It is homologous to the adhesion plaque protein Kindlin-1, which is expressed in epithelial cells, and also to Kindlin-2, which is widely expressed22,23. Loss-of-function mutations in KINDLIN1 lead to Kindler syndrome, a human disease characterized by skin blistering due to integrin dysfunction24,25. The kindlins have a FERM (protein4.1-ezrin-radixin-moesin) domain highly homologous to the FERM domain of talin26-28. Kindlin-2 and Kindlin-3 both bind to the integrin β subunit through their FERM domains; of note, the integrin activation-inducing activity of the talin FERM domain is enhanced by integrin-bound kindlins20,29,30.
We screened the human KINDLIN3 gene for germline mutations in the three subjects with LAD-III and their immediate families (Fig. 2). The two Turkish subjects with LAD-III (families two and three) were homozygous for a C→T nonsense mutation at nucleotide 1525 in exon 12 that changes a CGA codon to a TGA termination codon (R509X; Fig. 2a). The parents were heterozygous for this change, and the healthy sister in family two was homozygous for the wild-type allele (Fig. 2a). This mutation lies within the amino acid sequence coding for the carboxy-terminal half of the FERM F2 subdomain and could have an impact on the integrin binding F3 subdomain (Fig. 2c).
In contrast to the Turkish families, all members of the Maltese family were homozygous for the wild-type C allele at nucleotide 1525. However, further screening of this family revealed an A→G mutation at the splice acceptor site of exon 14 in the KINDLIN3 gene (Fig. 2b). The Maltese subject with LAD-III was homozygous for the G allele and the three other family members were heterozygous (Fig. 2b). This mutation, predicted to affect the correct splicing of exon 13 to exon 14, would also prevent the generation of an intact FERM F3 subdomain, thus having implications for the binding of KINDLIN-3 to integrin (Fig. 2c). The Turkish subjects’ families were all homozygous for the wild-type A allele.
We used quantitative RT-PCR to investigate the effect of these two mutations on KINDLIN3 mRNA expression in the transformed B cell lines. We used a probe specific for exons 6-7 (distant from the mutation sites) and one for exons 13-14 (designed to measure the integrity of the mRNA distal to the termination codon mutation in the Turkish subjects or in the neighborhood of the exon 14 splice acceptor site mutation in the Maltese subject). The parents of all three LAD-III-affected families had similar levels of KINDLIN3 mRNA as detected with both sets of probes compared to normal controls (Fig. 3a). The two Turkish subjects with LAD-III expressed substantially lower levels of KINDLIN3 mRNA, as detected by both probe sets (Fig. 3a), indicating that the C→T mutation that specifies a termination codon at nucleotide 1525 in exon 12 affects the stability of the KINDLIN3 mRNA.
For the Maltese subject, KINDLIN3 mRNA was undetectable by the exon 13-14 probes that covered the region where the A→G mutation was located (Fig. 3a). However, an RT-PCR assay covering exons 12-15 did generate a low level of product (Fig. 3b). Analysis of this product revealed a mixture of cDNAs, all of which contained a normal sequence up to the exon 13-14 boundary, followed by a variety of aberrant splice products (Fig. 3b). These results are consistent with the A→G mutation in the exon 14 splice acceptor site causing abnormal splicing of the exon 13 to exon 14 junction in the Maltese subject. However the cells of the Maltese subject also expressed lower levels of mRNA as detected using the exon 6 and 7 probes (Fig. 3a), suggesting that the A→G mutation, like the Turkish C→T mutation, destabilizes the KINDLIN3 mRNA upstream of the mutation.
By western blot analysis, KINDLIN-3 was detectable in the parents’ cells for all three families, but, as predicted from the low levels of mRNA, was absent from the cells of subjects with LAD-III (Fig. 3c). Thus, two different mutations in the KINDLIN3 gene had the effect of preventing expression of the KINDLIN-3 protein.
Kindlin-3 is crucial for activating integrins on mouse hematopoietic cells and for their adhesion-related functions20,21. However there has been no comparable study of human KINDLIN-3, and it was essential to test whether the defects in adhesion and migration shown by the cells from subjects with LAD-III could be repaired by the expression of the wild-type protein. Therefore, we transfected B cells from the three subjects with LAD-III with EGFP-KINDLIN3 cDNA that is able to reverse the integrin activation defect in Kindlin-3-deficient mouse cells21,22. As assessed by IRM, control EGFP-transfected LAD-III cells made poor contacts compared with EGFP-transfected parents’ cells (Fig. 4a). In contrast, expression of the EGFP-KINDLIN3 cDNA in LAD-III B cells increased their adhesion to ICAM-1. Both IRM images and quantification of the areas of attachment provided evidence that Kindlin-3-transfected cells from the Maltese and Turkish subjects with LAD-III were able to make adhesions equivalent in area to the parents’ cells (Fig. 4a).
To show further that expression of wild-type Kindlin-3 was able to overcome the LAD-III phenotype, we tested the same cells for their ability to undergo LFA-1-mediated migration on ICAM-1-coated surfaces. LAD-III cells from all three families expressing EGFP-Kindlin-3 migrated on ICAM-1 similarly to control EGFP-transfected parents’ cells, whereas this was not the case for control EGFP-transfected LAD-III cells (Fig. 4b and Supplementary Videos 1-4 online). B cells from the Maltese and Turkish subjects with LAD-III expressing wild-type Kindlin-3 also had an identical pattern of random motility compared to the parents’ cells (Fig. 4c and data not shown).
In summary, we describe two independent disabling mutations in the KINDLIN3 gene in subjects with LAD-III. The Maltese subject with LAD-III has a homozygous inactivating mutation within the splice acceptor site of exon 14, whereas both Turkish subjects contain an inactivating mutation in exon 12 resulting in a translational stop codon. This C→T nonsense mutation was recently described in three other Turkish subjects in a correspondence published after submission of this manuscript31. Both mutations lead to an overall decrease in KINDLIN3 mRNA levels and loss of protein expression. Furthermore, the failure of leukocytes from subjects with LAD-III to adhere and migrate, as is typical of the LAD-III phenotype, is restored by expression of wild-type Kindlin-3.
Kindlin-3 binds the cytoplasmic tails of β1, β2 and β3 integrins at the membrane distal NPXY motif 20,21,29. The evidence suggests that the binding of Kindlin-3 enhances talin binding at the membrane-proximal NPXY site, which causes an increase in integrin affinity32,33 Further details of the relationship between Kindlin-3 and integrin activation remain to be investigated. However, the implication of our results is that KINDLIN-3 is essential for the generation, maintenance or both of integrin activity on human leukocytes and platelets and that this crucial step is faulty in individuals with LAD-III.
A distinguishing feature between individuals with LAD-III and individuals with other integrin deficiency syndromes is the normal expression, but lack of function, of the β1, β2 and β3 integrins expressed by their leukocytes and platelets. This failure of integrin function leads to immune deficiency and bleeding problems8. To date, 14 individuals with LAD-III have been reported, of which the majority are of Turkish origin10-12,34,35. The nature of the disabling mutation(s) has been a focus of interest for more than a decade since the first reports of these individuals. In this study we identify mutations in the KINDLIN3 gene in the subjects with LAD-III and show that expression of wild-type Kindlin-3 can overcome the LAD-III defects by generating integrin adhesive contacts and integrin-mediated migration of LAD-III lymphocytes. This provides strong biological evidence for the presence of inactivating mutations in the KINDLIN3 gene as a cause of the LAD-III phenotype.
All affected children in the three families under study showed characteristics typical of the LAD-III disorder; cerebral hemorrhages at birth (presumed to result from a failure of platelet integrin function), an increased leukocyte blood count (presumed to result from the difficulty of leukocytes exiting the circulation as a result of a lack of leukocyte β1 and β2 integrin function) and recurrent tissue infections. In family one, the female Maltese subject was previously characterized as having LAD-III11 (see Supplementary Fig. 1). This child was successfully bone marrow transplanted in November 2001 (ref. 11). In family two, the subject, a 4-year-old Turkish boy, was also previously identified as having LAD-III (family five in ref. 12). A female Turkish subject with LAD-III (family three) who shows clinical features in keeping with a LAD-III diagnosis has not been previously reported on. This subject (<1 year old) first presented with anemia, thrombocytopenia and leukocytosis (35-100 × 109 cells per liter; neutrophilia and lymphocytosis in equal proportions). This subject had a history of delayed cord separation, hematuria, melena, petechias and severe infection. These symptoms all persisted, necessitating repeated erythrocyte transfusions. The subject has a clinical phenotype suggestive of the severe form of LAD-I: septicemia, axillary ulceration without pus collection, diffuse cellulitis on the right arm and delayed cord separation (surgical separation at day 20). Additionally, platelet aggregation tests showed abnormalities as seen in Glanzmann’s thrombasthenia.
We obtained approval for studies with the newly reported human primary cells from the Local Ethical Committee of the SB Ankara Diskapi Children’s Hospital, Ankara, Turkey. Additionally, informed consent was obtained from the donor families and control individuals.
Monoclonal antibodies 38 (specific for αL integrin, CD11a), TS1/18 (specific for β2 integrin, CD18), HP2/1 (specific for α4 integrin, CD49d) and P5D2 (specific for β1 integrin, CD29) have been previously reported11,36. We also used the following antibodies: rabbit polyclonal antibody to CALDAG-GEF1 (serum 3752) and DM 1A, an α-tubulin-specific monoclonal antibody (Sigma-Aldrich). We coupled a human-specific KINDLIN-3 peptide (EPEEELYDLSKVVLA; amino acids 156-170) to Imject maleimide activated keyhole limpet hemocyanin (Pierce) and used it to immunize rabbits. We prepared the affinity-purified antibodies as previously described22.
We used cDNAs encoding human cytosolic Myc-CALDAG-GEF1 and membrane-localized CALDAG-GEF1 (RASGRP2-Flag)15 and mouse cytosolic CALDAG-GEF1-Flag in this study. The mouse EGFP-KINDLIN3 DNA construct has been described previously22. We prepared full-length ICAM-1-Fc protein as previously described11.
We prepared T lymphocytes and expanded them in culture as previously described11. EBV-transformed B lymphoblastoid cells were derived from peripheral blood mononuclear cells of subjects and their relatives by Research Cell Services, Cancer Research UK, who derived these cells by standard procedures. We obtained standard control EBV-transformed cell lines from the same source. We maintained all cells in RPMI-1640 medium with 10% FCS.
We washed EBV-transformed B cells in OptiMEM + GlutaMAX (Invitrogen) and electroporated 2 × 107 cells with 10 μg per reaction of CALDAGGEF1 or KINDLIN3 cDNA constructs or 5 μg of EGFP-N1 (BD Biosciences) with a Gene Pulser with Capacitance Extender (Bio-Rad UK) set at 960 μF and 260 mV. We maintained the DNA-transfected cells in RPMI 1640 with 10% FCS for up to 24 h. We evaluated the efficiency of the transfection by flow cytometry, and we sorted the EGFP-positive cells with a MoFlo Cell Sorter (Beckman Coulter) before use in migration assays.
We analyzed genomic DNA from all subjects, relatives and nonrelated controls for sequence alterations in all exons and intron-exon boundaries of the CALDAGGEF1 (reference genomic data from http://genome.ucsc.edu/ (chr11: 64,250,959-64,269,504)) and KINDLIN3 (chr11: 63,730,782-63,747,930) genes by direct DNA sequencing in both orientations. Details and PCR conditions are available from the authors.
We quantified human CALDAGGEF1 and KINDLIN3 mRNA levels with TaqMan technology. Briefly, we extracted RNA from leukocytes with the GenElute Mammalian Total RNA Miniprep Kit (Sigma-Aldrich). We reverse-transcribed the RNA with the AffinityScript QPCR cDNA Synthesis Kit (Agilent Technologies) and purified it with the QIAquick PCR Purification Kit (Qiagen). We amplified a total of 40 ng of cDNA per reaction by TaqMan Gene Expression Assays from Applied Biosystems: for CALDAGGEF1, Hs00183378_m1 (exons 8-9) and Hs01057123_m1 (exons 2-3); for KINDLIN3, Hs00258828_m1 (exons 13-14) and Hs01075695_m1 (exons 6-7); and for control genes: Hs99999905_m1 (GAPDH) and Hs00202782_m1 (SF3B1).
We analyzed the samples on the ABI 7900HT Sequence Detection System instrument. We ran each sample in quadruplicate and expressed the data as a function of threshold cycle (CT). We corrected the CT values for reactions amplifying CALDAGGEF1 or KINDLIN3 by the CT value for two control housekeeping genes, SF3B1 and GAPDH to give ΔCT. The difference in ΔCT values between test and control cDNA samples allowed us to quantify the relative expression of the gene as follows: 2^- [(ΔCT test - ΔCT control) - (ΔCT control - ΔCT control)]
We coated either 35-mm glass-bottom microwell dishes (MatTek) or ibiTreat μ-Slides VI (Thistle Scientific) overnight with 3 μg ml-1 ICAM-1-Fc and then blocked them with BSA. We allowed the lymphocytes (4 × 105 cells per ml in HBSS with 20 mM HEPES buffer) to settle for 10 min at 37 °C and then removed the nonattached cells with gentle washing. We took images with a Nikon Diaphot 300 microscope, using a 20× or 63× lens and AQM2001 Kinetic Acquisition Manager software (Kinetic Imaging). We tracked the cells at 15-s intervals with Motion Analysis software (Kinetic Imaging) and analyzed the data with a Mathematica notebook (Wolfram Research) developed by D. Zicha (Cancer Research UK).
We plated EBV-transformed B cells on ibiTreat μ-Slides coated with 3 μg ml-1 ICAM-1. We acquired images of close substrate contact of the migrating cells between 10 min and 30 min of attachment to ICAM-1 with a Zeiss Axiovert 100 M inverted confocal microscope with a 63 × NA1.4 Plan-Apochromat oil-immersion objective lens. For evaluation of adhesion status, we measured the area of contact with MetaMorph Offline 7.1 (n = 35 cells per sample type).
The adhesion and migration assays are presented as the means ± s.e.m. We used the unpaired Student’s t test on the data with GraphPad Prism software version 4 for Macintosh computers. We analyzed the quantitative RT-PCR data by two-way analysis of variance. We separately assessed the data for each family and compared each subject with his or her relatives. Significant differences are indicated as follows: *P < 0.05, **P < 0.01 and ***P < 0.001.
Detailed methodology is described in Supplementary Methods online.
We are indebted to R. Fässler and D. Wagner for helpful discussion; J. Crittenden and A. Graybiel (Massachusetts Institute of Technology) for antibodies to CALDAG-GEF1, cDNA constructs and helpful discussion; and J. Hancock (University of Queensland Medical School) for CALDAGGEF1 constructs. We are also grateful to our Cancer Research UK London Research Institute colleagues D. Harvey for generation of the EBV-transformed cell lines and G. Kelly for help with the statistical analyses. L. S. was supported by a Marie Curie Individual Fellowship.