Impact of polymorphisms of the major histocompatibility complex class II, interleukin-10, tumor necrosis factor-α and cytotoxic T-lymphocyte antigen-4 genes on inhibitor development in severe hemophilia A


Johannes Oldenburg, Institute of Experimental Haematology and Transfusion Medicine, University Clinic Bonn, Sigmund-Freud-Str. 25, Germany.
Tel.: +49 228 287 15175; fax: +49 228 287 14783.


Summary. Background: Approximately 25% of severe hemophilia A (HA) patients develop antibodies to factor VIII protein. Patients: In the present case-controlled cohort study, 260 severely affected, mutation-type-matched HA patients were studied for association of human leukocyte antigen (HLA) class II molecules and polymorphisms in the genes encoding interleukin-10 (IL-10), tumor necrosis factor-α (TNF-α) and cytotoxic T-lymphocyte antigen-4 (CTLA-4) and development of inhibitors. Results: Our results demonstrate a higher frequency of DRB1*15 and DQB1*0602 alleles as well as of the haplotype DRB1*15/DQB1*0602 in inhibitor patients [odds ratio (OR) 1.9; P < 0.05]. In TNF-α, the A allele of the −308G>A polymorphism was found with higher frequency in the inhibitor cohort (0.22 vs. 0.13, OR 1.80). This finding was more pronounced for the homozygous A/A genotype (OR 4.7). For IL-10, the −1082G allele was observed more frequently in patients with inhibitors (0.55 vs. 0.43; P = 0.008). The functional cytokine phenotype was determined for the first time, on the basis of the genetic background, and this showed that 12% of patients with inhibitors were high-TNF-α/high-IL-10 producers, as compared with 3% of non-inhibitor patients (OR 4.4). A trend for a lower frequency of the A allele of the CT60 polymorphism in CTLA-4 was found in inhibitor patients (0.42 vs. 0.50). Conclusions: In conclusion, the reported data clearly highlighted the participation of HLA molecules in inhibitor formation in a large cohort of patients. The higher frequencies of the −308G>A polymorphism in TNF-α and −1082A>G in IL-10 in inhibitor patients confirmed the earlier published data. The CT60 single-nucleotide polymorphism in CTLA-4 is of apparently less importance.


The development of antibodies against the factor VIII protein following replacement therapy currently represents the most serious complication of hemophilia treatment. FVIII antibodies develop in approximately 25% of patients with severe hemophilia A (HA); they complicate the management of hemorrhage, and result in higher therapy costs and increased morbidity and mortality [1,2].

Recently, significant progress has been made in our understanding of the mechanisms that lead to inhibitor formation. Both genetic and environmental factors have been shown to play decisive roles [3,4]. Environmental factors include intensity of treatment, age at start of prophylaxis, and type of administered substitution therapy product [5,6]. The first genetic factors with a major influence on inhibitor formation were the F8 gene defects, discovered in the 1990s [7,8]. More recently, polymorphisms of the immune response genes, including genes encoding the major histocompatibility complex (MHC) class II system, tumor necrosis factor-α (TNF-α), interleukin-10 (IL-10) and cytotoxic T-lymphocyte antigen-4 (CTLA-4), have been suggested to be contributing determinants of the inhibitor risk [9–13].

The human leukocyte antigen (HLA) class II molecules play an essential role presenting FVIII peptides to CD4-positive T-helper cells. In particular, the HLA class II alleles DRB1*15 and DQB1*0602 have been reported to have an influence on inhibitor development [9,10]. Additionally, functional polymorphisms of proinflammatory and anti-inflammatory cytokines are important modulators of the immune response. An imbalance in cytokine secretion may affect the clinical outcome following initial exposure to FVIII protein. Expression and production of cytokines is, in part, genetically determined. Polymorphisms involving the 5′-flanking regions (promoters) of TNF-α and IL-10 are known to affect transcription levels, and are therefore of functional relevance [14]. Current knowledge of the influence of polymorphisms on IL-10, TNF-α and CTLA-4 with respect to inhibitor formation is highlighted in the analysis of individuals from the Malmö Inhibitor Brother Pair Study (MIBS). The polymorphisms at the gene loci IL-10 microsatellite IL10.G, TNF-α−308 and CTLA-4−318 were identified as being either risk factors for or having protective roles against inhibitor formation [11–13].

In order to further investigate the previously reported findings, as well as additional single-nucleotide polymorphisms (SNPs) associated with variations of the cytokine levels, we genotyped HLA class II alleles and several polymorphic sites in TNF-α, IL-10 and CTLA-4 in a separate, mutation type-matched, unrelated case-controlled cohort of 260 severely affected HA patients with and without inhibitors. Thus, for the first time, the role of polymorphisms in the immune response genes are analyzed in a large, homogeneous and well-defined cohort of patients with HA.

Patients, materials and methods


Our study cohort consisted of 260 patients with severe HA from different hemophilia centers in Germany. One hundred and thirty patients had been diagnosed with inhibitors, and 130 patients were without inhibitors and matched with inhibitor patients for mutation type. Patients with a non-null mutation were excluded because of the possibility of the presence of endogenous FVIII protein. Patients with limited data on the presence of an inhibitor, as well as patients with transient inhibitors, were also excluded. All non-inhibitor patients had more than 150 exposure days. High-response inhibitors were defined as those with an inhibitor titre >5 BU mL−1 at any time and low-response inhibitors were those with an inhibitor titre persistently <5 BU mL−1 or less [15]. The study cohort did not contain sibling patients with hemophilia. Informed patient consent was obtained according to the Declaration of Helsinki.

Coagulation assays

Blood was collected by peripheral venous puncture as a 1 : 10 volume ratio in 3.8% trisodium citrate. FVIII activity for each sample was determined by an in-house-modified one-stage activated partial thromboplastin time-based assay, performed on a KC10A coagulation analyzer (Trinity Biotech, Lemgo, Germany) with FVIII-deficient plasma from Helena (Genzyme Virotech GmbH, Rüsselsheim, Germany). FVIII inhibitor activity was determined using the modified Njimegen Bethesda method [16].

F8 gene analyses

DNA isolation  High molecular weight genomic DNA was isolated from whole blood by a salting-out procedure [17].

Intron 22/intron 1 inversion polymerase chain reaction (PCR)  Detection of the intron 22 inversion was performed as described by Rossetti et al. [18]. Intron 1 inversion was analyzed by PCR as reported previously, with a modified annealing temperature of 59 °C [19].

Sequence analyses of the F8 gene  The F8 gene was examined by direct sequencing of all 26 exons and flanking intronic regions, for both strands, using the BigDye Terminator Cycle Sequencing V1.1 Ready Reaction kit and an automated ABI-3130 DNA sequencer (Applied Biosystems, Foster City, CA, USA). The Sequence Analysis software package (Applied Biosystems) was employed for final sequence reading and mutation documentation. Primers and sequencing conditions are available on request.

HLA class II allele genotyping

HLA class II (HLA-DRB1 and HLA-DQB1) typing was performed using the Dynal RELI PCR-SSOP test, following the manufacturer’s recommendations (Dynal Biotech, Wirral, UK). Briefly, the test is based on PCR target amplification (50 or 100 ng of genomic DNA), hybridization of the amplified products to an array of immobilized sequence-specific oligonucleotide probes, and detection of the probe-bound amplified product by colorimetric product formation. The sequence-specific oligonucleotide (SSO) probes are designed to hybridize with polymorphic target sequences of the corresponding HLA loci. Under appropriate washing conditions, the probes remain bound only to their corresponding complementary sequence in the amplified DNA. Evaluation of results was performed using pmp5.4.1 software (Dynal Biotech, Hamburg, Germany). All ambiguous typings, as well as HLA-DQB subtyping, were subjected to high-resolution analysis by the PCR sequence-specific priming technique, using the One Lambda’s High Resolution Trays according to the manufacturer’s specifications (One Lambda, Montpellier, France). PCR products were analyzed on 2.5% agarose gels and stained with ethidium bromide.

HLA allele assignments were made according to the World Health Organization Nomenclature Committee for Factors of the HLA System (13th International Histocompatibility Workshop, Victoria, Canada).

Genotyping of IL-10, TNF-α and CTLA-4

We investigated three biallelic SNPs and a dinucleotide microsatellite polymorphism in IL-10, four biallelic polymorphisms in TNF-α, and three biallelic polymorphisms in CTLA-4. All polymorphisms were identified using the National Center for Biotechnology Information SNP database. The particular SNPs included in this study were as follows: IL-10– G>A −1082(rs1800896), C>T −819(rs1800871), and C>A −592(rs1800872); TNF-α– C>T for −827(rs1799724), G>A for −308(rs1800629), A>G for −238(rs361525), and A>G for 670(rs3093662); and CTLA-4– C>T for −318(rs5742909), A>G for 49(rs231775), and A>G for CT60(rs3087243). Genotype data were complete for the IL-10, TNF-α and CTLA-4 markers in all of the subjects. All biallelic polymorphisms were detected by PCR amplification and direct sequencing (primers and conditions are available on request). The multiallelic IL-10 microsatellite IL10.G was genotyped by fragment length analysis. Flanking primers were constructed: IL10.1, F-AGGATCCCCAGAGACTTTCC, and R-ATGGAGGCTGGATAGGAGGT. The forward primer was 5′-labeled with 6-FAM fluorescent dye (PE Applied Biosystems, Weiterstadt, Germany). Samples were run on an ABI Prism 3130 automatic sequencer (Applied Biosystems, Foster City, CA, USA). Each sample included an internal size standard (GS500 ROX; Applied Biosystems). The size of microsatellite-containing DNA fragments was measured by comparison with the DNA size standard using Gen Mapper software (version 4.0). Alleles were named according to Kube et al. [20], where the numbering directly reflects the number of CA repeats.

Statistical analysis

Gene frequencies were calculated by gene counting and expressed as a ratio of the number of each allele observed within a locus to the total number observed in that locus. A conventional chi-square test was applied to compare the frequencies in between the HA patients with and without inhibitors. Fisher’s exact test was used if sample sizes were < 5. All tests were two-sided, and a P-value below 0.05 was considered to indicate statistical significance. To correct the level of significance in the chi-square test and eliminate significance by chance, the P-value was corrected using the Bonferroni correction [21]. Odds ratios (ORs) were calculated, and were considered to be significant if the lower value of the 95% confidence interval (CI) was larger than 1.0. Haplotype frequencies were estimated using the program famhap. The P-values given for the single haplotypes are derived by the program using a likelihood ratio test. The ORs are calculated from the estimated haplotype frequencies.


In the present study, polymorphisms in the HLA class II system and IL-10, TNF-α and CTLA-4 were analyzed for their association with inhibitor development in 260 patients with severe HA. For each polymorphism, allele and genotype frequencies were determined. All polymorphisms were in Hardy–Weinberg equilibrium for both inhibitor and control patient groups.


The case-controlled cohort consisted of 130 patients with severe HA with inhibitors and 130 patients without inhibitors who were matched according to mutation type. The mutation spectrum for both groups included only F8 null mutations, as follows: 74 (57.1%) intron 22 inversions, 12 (9.2%) large deletions, 21 (16.1%) small deletions/insertions, 16 (12.3%) nonsense mutations and seven (5.3%) splice site mutations for each group. All patients enrolled in the study were German Caucasians.

HLA class II allele frequencies

The frequencies of HLA class II alleles in patients with HA with and without inhibitors are shown in Table 1. DRB1*1501 was found to be present more frequently in inhibitor than in non-inhibitor patients (0.20 vs. 0.11). The difference was statistically significant (P = 0.0054), with an OR of 1.99 (95% CI 1.21–3.25). All other alleles of the DRB1* loci exhibited no statistically relevant differences in both groups. On comparison of the results for the DQ locus, a statistically significant difference was detected only for DQB1*0602. Forty-two (16%) patients with inhibitors vs. 23 (9%) non-inhibitor patients exhibited DQB1*0602 (P = 0.0117), corresponding to an OR of 1.98 (95% CI 1.15–3.40).

Table 1.   Phenotypic frequencies of human leukocyte antigen class II DRB1 and DQB1 alleles in patients with hemophilia A with and without inhibitors
 Inhibitor patientsNon-inhibitor patientsP ORCI
  1. CI, confidence interval; OR, odds ratio. Statistically significant associations are shown in bold.

DRB1*0940.01100.155642Fisher exact 
DQB1*0606  1    

Further comparison highlights the distribution of DRB1*1501 and DQB1*0602 with respect to mutation type. The group including only patients with intron 22 inversions was compared with those with patients carrying other types of null mutation. The frequencies of DRB1*15 for inhibitor-positive patients showed no difference with respect to either the intron 22 inversion or the group of the other mutation types (0.20 vs. 0.19); this was also the case for the group of non-inhibitor patients (0.11 vs. 0.10).

Multivariate analysis of the results, performed to investigate the relative predisposing effect of DRB1*1501 and DQB1*0602, showed that 32 (24.6%) inhibitor patients were positive for both DRB1*1501 and DQB1*0602, 13 (10%) were positive for DRB1*1501 and negative for DQB1*0602, and six (4.6%) were positive for DQB1*0602 and negative for DRB1*1501. Nineteen (14.6%) of the non-inhibitor patients possessed the DRB1*1501/DQB1*0602 haplotype. The comparative results showed a positive association of the DRB1*1501/DQB1*0602 haplotype with inhibitor formation (0.25 vs. 0.15; P = 0.0423; OR 1.9; 95% CI 1.01–3.57). The complete source data are provided in supplementary Table S1.

TNF-α genotype and allele frequencies

Genotype and allele frequencies for TNF-α are presented in Table 2. The only polymorphism that showed statistically significant differences between the groups of inhibitor and non-inhibitor patients was −308G>A. The data showed that 63% and 75% of patients with and without inhibitors, respectively, were GG homozygous, whereas 30% of inhibitor patients and 24% of non-inhibitor patients were AG heterozygous at −308. The incidence of AA, however, was significantly different between the two groups analyzed (P = 0.031). The inhibitor patient group showed a 7% incidence, whereas the non-inhibitor patients showed a rate of 2%. The OR was 4.76 (95% CI 1.00–22.47). The allele frequency data confirmed the higher representation of the −308A allele in inhibitor patients (0.22) than in non-inhibitor patients (0.13) (P = 0.0114; OR 1.80; 95% CI 1.13–2.86).

Table 2.   Frequencies of tumor necrosis factor-α gene (TNF-α) polymorphisms and haplotypes in patients with severe hemophilia A with and without inhibitors
TNF-αInhibitor patientsNon-inhibitor patientsPORCI
  1. CI, confidence interval; OR, odds ratio. Statistically significant associations are shown in bold.

G>A −308
Allele frequency
G>A −238
 A/A0 0    
Allele frequency
C>T −827
Allele frequency
A>G 670
Allele frequency
 308 −238 670
 A G A550.21250.090.0002 2.481.51–4.19
 G G A1880.722050.780.08270.710.46–1.04
 G G G70.03140.050.11890.470.19–1.25
 G A A50.0260.020.76050.830.25–2.75
 G A G30.0100 0.83 
 A G G20.0170.030.09270.30.05–1.36
 A A A 030.01 0 

A single TNF-α haplotype (AGA) showed a strong association with inhibitor formation (P = 0.0002; OR 2.48; 95% CI 1.51–4.19). This association remained significant after Bonferroni correction for multiple testing.

IL-10 genotype and allele frequencies

In the IL-10 promoter region, three SNPs at positions −1082, −819 and −592, associated with changes in the expression of the IL-10 gene, were analyzed. Additionally, the CA repeat unit polymorphisms, also located in this region, have been included in the study. The allele and genotype frequencies for these patients are summarized in Table 3. As reported earlier, the −819 and −592 polymorphisms were found to be in complete linkage disequilibrium with each other [22]. The only polymorphism that exhibited a statistically significant difference between the groups of patients with and without inhibitors was −1082G>A. The frequency of the G allele was greater in inhibitor than in non-inhibitor patients (0.55 vs. 0.43; P = 0.008; OR 1.5). No significant difference in the distribution of the multiallelic IL-10 microsatellite IL10.G between the inhibitor and non-inhibitor patients was observed.

Table 3.   Frequencies of interleukin-10 gene (IL-10) polymorphisms and haplotypes in patients with severe hemophilia A with and without inhibitors
IL-10Inhibitor patientsNon-inhibitor patientsPORCI
  1. CI, confidence interval; OR, odds ratio. Statistically significant associations are shown in bold.

C>T −819
Allele frequency
C>A −592
Allele frequency
A>G −1082
Allele frequency
 CA repeats       
 180 20.01   
 2730.0140.020.2692Fisher Exact 
 2820.0110.000.3757Fisher Exact 
 −1082 −819 −592
 G C C1440.551140.430.00651.61491.14–2.28
 A C C530.20840.320.00200.53650.36–20.79
 A T A630.24620.24   

As specific IL-10 haplotypes have been linked to different phenotypes of IL-10 production [23], we performed an analysis of data comparing IL-10 haplotypes between the inhibitor and non-inhibitor patients. Upon analysis of each individual allele at the three IL-10 promoter sites, we found only six genotypes, which correspond to the three haplotypes previously described in the Caucasian populations (GCC, ACC and ATA at positions −1082, −819, and −592, respectively) (Table 3). A notably increased incidence of the GCC haplotype was detected in patients with inhibitors (0.55 vs. 0.43, P = 0.0065), corresponding to an OR of 1.6.

Analysis of interaction between the TNF-α and IL-10 alleles

The G to A polymorphism at position −308 of TNF-α is associated with high levels of cytokine production [14]. The polymorphism located in the IL-10 promoter region at position −1082 (G to A) results in lower expression levels of IL-10 [24]. We assessed the incidence of inhibitor development among patients in relation to different TNF-α/IL-10 genotype combinations (Table 4). There was a strong association of the high-TNF-α/high-IL-10 haplotype [i.e. TNF-α−308(AA/GA)/IL-10−1082(GG)] with inhibitor formation (0.12 vs. 0.03; P = 0.0052; OR 4.4). Conversely, the low-TNF-α/low-IL-10 haplotype [i.e. TNF-α−308(GG)/IL-10−1082(AA/AG)] was predominantly expressed in non-inhibitor patients (0.46 vs. 0.62; P = 0.0089; OR 0.5).

Table 4.   Frequencies of tumor necrosis factor-α (TNF-α) and interleukin-10 (IL-10) low-producer and high-producer predicted phenotypes in patients with severe hemophilia A with and without inhibitors
 Inhibitor patientsNon-inhibitor patientsPORCI
  1. CI, confidence interval; OR, odds ratio. Statistically significant associations are shown in bold. *Potential phenotypes were assigned after typing of TNF-α and IL-10 gene promoter polymorphisms as follows. Homozygous TNF-α: −308G-positive subjects were assumed to be low-TNF-α producers. Homozygous and heterozygous TNF-α: −308A-positive subjects were assumed to be high-TNF-α producers. Homozygous IL-10:−1082G-positive subjects were assumed to be high-IL-10 producers. Homozygous and heterozygous IL-10: −1082A-positive subjects were assumed to be low-IL-10 producers [18].

High TNF-α/low IL-10*330.25290.220.56041.18490.66–2.09
High TNF-α/high IL-10160.1240.030.00524.42111.43–13.61
Low TNF-α/low IL-10600.46810.620.00890.51850.31–0.85
Low TNF-α/high IL-10210.16160.120.37471.37270.68–2.76

CTLA-4 genotype and allele frequencies

No statistically significant difference with regard to the genotype analysis for the three tested polymorphisms of CTLA-4 was found between the groups of inhibitor and non-inhibitor patients (Table 5). However, a trend for lower frequency of the A allele of the CT60A>G polymorphism was observed in patients with inhibitors (0.42 vs. 0.50; P = 0.06; OR 0.72).

Table 5.   Phenotypic frequencies of cytotoxic T-lymphocyte antigen-4 gene (CTLA-4) polymorphisms in patients with severe hemophilia A with and without inhibitors
CTLA-4Inhibitor patientsNon-inhibitor patientsPORCI
  1. CI, confidence interval; OR, odds ratio.

G>A 49
Allele frequency
C>T −318
Allele frequency
A>G CT60
Allele frequency


In the present study we examined, for the first time, the association of HLA and polymorphisms of the IL-10, TNF-α and CTLA-4 genes with respect to development of inhibitors in a case controlled cohort of 260 severely affected hemophilia A patients. Our results demonstrate a higher frequency of DRB1*15 and DQB1*0602, as well as the haplotype DRB1*15/DQB1*0602, among patients with inhibitors than among those without inhibitors. The −308A allele in TNF-α and the IL-10−1082G allele were found with higher frequencies in the inhibitor cohort. The results of our study failed to demonstrate a substantial correlation between the presence of CTLA-4 polymorphisms and inhibitor formation. However, a trend for a lower frequency of the A allele in the CT60 polymorphism was found in patients with severe HA and inhibitors.

The generation of FVIII inhibitors is a CD4+ T-cell-dependent process involving a cascade of events, comprising antigen presentation, activation of T lymphocytes, cytokine secretion, and differentiation of B lymphocytes into FVIII-specific antibody-secreting plasmocytes [25]. In this process, HLA class II molecules play an important role in presenting FVIII-derived peptides to the T cells. It is known that the degree of tolerance towards exogenous therapeutic FVIII of the immune system of a given patient is dependent on the type of the F8 gene mutation responsible for HA, and on the associated production or absence of endogenous FVIII. Accordingly, the underlying genetic alteration has been defined as one of the major risk factors for inhibitor development [26]. To exclude the possibility that the type of mutation may mask the association between inhibitor risk and polymorphisms in immune response genes, our cohort included only patients with high inhibitor risk null mutations, fully matched between the groups of inhibitor and non-inhibitor patients.

Numerous studies have demonstrated associations between HLA alleles and disease susceptibility. The high polymorphicity of the HLA region made it the first candidate to be studied for association with inhibitor development in HA. In the previous work of Oldenburg et al. [10], the data, exclusively derived from patients with intron 22 inversion, showed higher frequencies among inhibitor-positive patients for DRB1*15 DQB1*0602 (DR1501, 0.36 vs. 0.19; DQB0602, 0.31 vs. 0.14) than those reported in the present article (DR1501, 0.20 vs. 0.11; DQB0602, 0.16 vs. 0.09). A possible explanation for the observed lower frequencies in this study might be that, since the first study was conducted, new HLA alleles have been discovered and more sensitive test systems have been introduced, reducing the frequencies of homozygous DR15 and DQB0602 cases observed in the earlier study [27]. In the present study, no diversity in the frequencies of DRB1*15 and DQB1*0602 was detected when the patients were grouped with respect to mutation type (intron 22 inversion vs. all other null mutations).

In contrast to these findings, Astermark et al. [12] found no correlation of any HLA allele with inhibitor formation in the MIBS cohort. It may be hypothesized that, because MIBS contains brother pairs that share 50% of alleles, the influence of genetic factors of limited impact such as HLA may have escaped detection. Additionally, MIBS investigated an ethnically and geographically diverse group of patients; it is well documented that HLA alleles vary considerably with geography and the race of patient cohorts.

Specific polymorphisms in the promoter regions of genes encoding the TNF-α and IL-10 cytokines and the CTLA-4 costimulatory molecules have previously been associated with inhibitor development in patients with HA [11–13]. TNF-α is located in the region of the MHC genes; a genetic association has been proposed between particular TNF-α alleles, levels of TNF-α production, extent of immune responses, and susceptibility to diseases [28–30]. The polymorphism G>A at position −308 of TNF-α results in an increase in transcription by six-fold to seven-fold and of TNF-α production [14,30–32]. In accordance with a previous study of Astermark et al. [15], the −308G>A TNF-α polymorphism was significantly associated with inhibitor formation in our cohort. The −308A allele (also referred to as the TNF2 allele) and the TNF-α high-producer genotype (G/A or A/A) were present at higher frequencies among inhibitor patients than among non-inhibitor patients. Notably, the TNF-α haplotype AGA (P = 0.0002; OR 2.48) showed an association with inhibitor development, which was most likely due to the presence of the −308A allele.

IL-10 is a pleiotropic cytokine that may act as a potent suppressor of systemic inflammatory immune responses, while promoting activated B lymphocytes to secrete immunoglobulins. Twins and family studies have suggested that nearly 75% of the variation in IL-10 production is genetically determined, and appears to be controlled at the transcriptional level [33,34]. Polymorphisms located at three sites in the IL-10 promoter may alter the expression of the IL-10 gene: at positions −1082 (G>A), −819 (T>C), and −592 (C>T). In vitro studies indicate that the −1082G allele is associated with higher IL-10 production and the A allele with lower IL-10 production [35]. The data for the genotype frequencies for IL-10 at positions −1082 in our study showed a higher frequency of the IL-10 G allele among inhibitor patients than among non-inhibitor patients (0.55 vs. 0.43). We also found that the high-producer genotype (−1082G/G) was present in 33% of inhibitor-positive patients and in 22% of non-inhibitor patients. A clear predominance of the high-producer GCC haplotype (0.55 vs. 0.32; P = 0.0065) and a lower frequency of the low-producer ACC haplotype (0.20 vs. 0.32; P = 0.002) was observed in patients with inhibitors.

The previous study of Astermark et al. [11] showed significant expression of the 134 allele in CA repeat microsatellites of the IL-10 promoter. Notably, both the 134 allele and the G allele of the −1082 IL-10 polymorphisms are associated with an increase in the secretion of IL-10. It is noteworthy that the additional IL-10 SNPs described in this study were not studied in MIBS. The reason why the association was not found between inhibitor formation and IL-10.G microsatellites might be the low frequency of the 134 allele, different ethnicity, diversity of the F8 gene defects, and the family composition of MIBS (sibling pairs).

We have evaluated for the first time the combined cytokine phenotype on the basis of the genetic interaction between IL-10 and TNF-α in promoting the development of inhibitors. Our results show that 12% of inhibitor-positive patients were high-TNF-α/high-IL-10 producers, whereas only 3% of non-inhibitor patients had this functional phenotype. The association between inhibitor development and the high-TNF-α/high-IL-10 genotype yielded an OR of 4.4. Thus, the combination of the IL-10 and TNF-α genotypes appears to be an aggravating risk factor for inhibitor development. However, it should be taken into consideration that a correlation of the genotypic profile of the polymorphisms found in the present work and in previous studies and the relative cytokine production in patients with hemophilia still remains to be demonstrated.

CTLA-4 is a second costimulatory molecule and is a homolog of CD28. It is expressed only on activated T cells, binds to accessory molecule B7, and downregulates T-cell-dependent immune responses [36]. It is of note that CTLA-4–Ig constructs, which suppress the interaction between B7 on antigen-presenting cells and CD28 on T lymphocytes, have been shown to prevent the onset of the anti-FVIII immune response in FVIII-deficient mice [37]. From the three studied polymorphisms, only for the A>G CT60 polymorphism was a trend for a lower frequency of the A allele observed in patients with inhibitors (0.42 vs. 0.50; P = 0.06; OR 0.72). This tendency towards a protective function of the A allele was slightly stronger in the homozygous A/A genotype (0.18 vs. 0.28; P = 0.05; OR 0.56).

The CT60 polymorphism was not evaluated in MIBS, but instead the T allele at −318 was associated with a protective role against inhibitors [13]. In addition, in a previous study, we have shown that the polymorphic site at +49 seems to be associated with acquired hemophilia [38]. Altogether, the data suggest that polymorphisms in CTLA-4 have the potential to exert a minor effect on the immune response to FVIII by modulating the level of this costimulatory molecule at the cell surface.

Several patients included in our study failed to develop FVIII inhibitors despite the fact that they had HLA haplotypes and/or polymorphisms in TNF-α and IL-10 associated with an increased risk for alloimmunization towards FVIII. This strongly suggests that multiple factors, either genetic or environmental, play a role in predisposing a patient with HA to develop FVIII inhibitors.

In conclusion, the data reported in our study clearly highlight the participation of HLA molecules in inhibitor formation in a large cohort of patients with hemophilia. Although different polymorphisms in TNF-α and IL-10 were investigated and found to contribute to inhibitor formation as compared with earlier studies, the association of polymorphisms in this and other cytokine genes with inhibitor development opens up a new avenue of risk factors that add to our understanding of the complex processes of FVIII-specific tolerance and immunity in patients with HA.


A. Pavlova collected data, analyzed and interpreted data, and wrote the manuscript; D. Delev and R. Schwaab performed experiments; S. Lacroix-Desmazes and J. Astermark interpreted data and wrote the manuscript; M. Mende and R. Fimmers performed statistical analysis; and J. Oldenburg designed the research and concept of the study, interpreted data, and wrote the manuscript.


We are very thankful to J. Junen and A. Schmitt for technical assistance.

Disclosure of Conflict of Interests

This study was supported by a grant from Novo Nordisk (N-045.0099) to J. Oldenburg. The other authors state that they have no conflict of interest.