Priscilla Lapalud, SysDiag, UMR3145 CNRS/BioRad, Cap Delta, 1682 rue de la Valsière, CS 61003, 34184 Montpellier, France. Tel.: +33 6 08 58 53 35; fax: +33 4 67 16 66 01. E-mail: email@example.com
Summary. Background: Acquired hemophilia A (AHA) is a severe life-threatening autoimmune disease due to the development of autoantibodies that neutralize the procoagulant activity of factor VIII (FVIII). In rare cases, AHA occurs in the postpartum period as a serious complication of an otherwise normal pregnancy and delivery. Due to its rarity, little is known about the features of the antibody response to FVIII in AHA.
Objectives: Our study wanted to (i) determine the epitope specificity and the immunoglobulin (Ig) subclasses of anti-FVIII autoantibodies in plasma samples from a large cohort of AHA patients, and (ii) compare the epitope specificity of anti-FVIII autoantibodies in plasma samples from postpartum AHA and other AHA patients.
Patients/Methods: Seventy-three plasma samples from patients with postpartum AHA (n = 10) or associated with malignancies (n = 16) or autoimmune diseases (n = 11) or without underlying disease (n = 36) were analyzed with three multiplexed assays.
Results and Conclusions: Our results showed a stronger response against the A1a1-A2a2-B fragments of FVIII and more specifically against the A1a1 domain in patients with postpartum AHA than in the other AHA groups (P <0.01). Moreover, although IgG4 was the predominant IgG subclass in all groups, anti-A1a1-A2a2-B and anti-A1a1 domain autoantibodies of the IgG1 and IgG3 subclasses were more frequently detected in postpartum AHA than in the other AHA groups. These findings support the involvement of the Th1-driven response in the generation of autoantibodies in women with postpartum AHA compared with the other groups of AHA patients in whom production of Th2-driven IgG4 was predominant.
Factor VIII (FVIII) plays a critical role in thrombin generation at the surface of activated platelets [1,2]. FVIII is a glycoprotein of 2332 amino acids and consists of a heavy chain (HC; A1a1-A2a2-B domains) and a light chain (LC; a3A3-C1-C2 domains) that are associated through ion metal interactions [3,4].
Acquired hemophilia A (AHA) is an autoimmune disease due to the development of IgG autoantibodies (autoAbs) that neutralize FVIII procoagulant activity or accelerate its clearance [1,2]. Although AHA incidence varies between 0.1 and 1.5 cases per million per year, the mortality rate is high (between 7.9% and 22%) [5,6]. Because of its rarity, the exact pathogenesis of inhibitors in AHA remains unclear and only a few data on the humoral response against FVIII are available.
AHA mostly occurs in elderly patients (two-thirds are older than 70 years [7,8]) without any identified underlying disease (idiopathic AHA), but can also be associated with autoimmune diseases (AID), or solid and hematologic malignancies in up to 50% of cases [5,9,10]. AHA has been reported also in young women (20–30 years old) during the postpartum period [7,10]. Postpartum AHA is a rare (7–11% of all AHA patients) and serious complication of an otherwise normal pregnancy and delivery [11,12]. AutoAbs develop after the first pregnancy and generally do not recur, although reappearance during subsequent pregnancies has been described [13,14]. Compared with AHA associated with other conditions, the outcome of postpartum AHA is often favorable, with a high percentage of spontaneous remissions [12,15], independently of treatment .
The inhibitor titer of plasma samples containing anti-FVIII neutralizing antibodies is usually assessed with the Bethesda assay [17,18], a functional test that measures the residual FVIII procoagulant activity. However, as acquired autoAbs often exhibit a non-linear FVIII inactivation pattern (type II kinetics), this method might underestimate the in vivo inhibitor potency, making more difficult the monitoring of AHA patients and the choice of adequate therapeutic strategies . Moreover, non-neutralizing antibodies (NNA) cannot be detected by the Bethesda assay. NNA may bind to non-functional FVIII epitopes that are mainly localized on the A1, C1 and B domains and increase FVIII clearance with possible clinical consequences, such as serious hemorrhages even with low inhibitor titre [20,21].
Previous studies have reported that anti-FVIII autoAbs are often IgG4, thus suggesting a Th2-driven immune response, and less frequently IgG1 (Th1-driven response) [22,23]. However, there are few data on the possible variations in epitope specificities and autoAb isotypes/classes in AHA patients relative to their specific clinical context, notably in postpartum AHA. Matsumoto et al.  reported a higher level of IgG1 in patients with AHA associated with a disease than in patients with idiopathic AHA and suggested a relationship between the presence of an underlying disease and the patient’s immunological characteristics. We therefore analyzed, using HC/LC multiplexed assays , the IgG profile and the specificity of the anti-FVIII autoAbs in seventy-three patients with AHA associated with different underlying conditions, including the postpartum. As the most common epitopes of anti-FVIII Abs are located on the A2 and C2 domains of FVIII and little is known about anti-A1 domain autoAbs, we investigated the immune response towards these domains using C2 and A2a2/A1a1 multiplexed assays.
Materials and methods
Plasma samples were obtained from 73 patients with AHA (median age, 71 years; range, 28–92; male:female ratio = 39:34) (Table 1) who belonged to the French SACHA cohort, a biological follow-up study of patients with acquired hemophilia . The study protocol was approved by French national ethics and data-protection committees and registered in the National Institutes of Health (NIH) clinical trial database (trial identifier: NCT00213473). The 63 informed patients reported mainly spontaneous bruises (49/63; 78%), soft tissue and mucous membrane hemorrhages (17/63; 27%) and muscle hemorrhages (11/63; 17%). Plasma samples were collected at diagnosis and kept at −80 °C until use. FVIII activity (in %) and inhibitor titer (in BU mL−1) were measured with the Nijmegen-modified Bethesda assay in the center where the specific patient was followed.
Table 1. Clinical and laboratory characteristics of the patients with AHA included in the study
Age, median in years (range)
Inhibitor titer, median in BU mL−1 (range)
FVIII activity, median in % (range)
Proteins and antibodies
Proteins The HC (A1a1-A2a2-B) and LC (a3A3-C1-C2) of recombinant FVIII (rFVIII) (from mammalian cells, Bayer Healthcare, Loos, France) were dissociated by incubation with 100 mm EDTA at room temperature for 30 min just before use. To obtain the A1a1 and A2a2 domains, EDTA-dissociated rFVIII was thrombin-cleaved as described . To produce recombinant C2, an N-ter KKKGG and C-ter poly-His tag were introduced into the cDNA sequence of the human C2 domain. The plasmid construct was transformed into the Origami E.coli competent strain. Bacteria were batch-cultured in 2 L of culture medium (Proteus Company, Nîmes, France) and the C2 domain was purified through a Cobalt column (Clontech Laboratories Inc., Mountain View, CA, USA). The recombinant C2 domain (1.8 mg mL−1 and 90% of purity) was recognized by the conformation-dependent monoclonal antibody (mAb) Bo2C11, thus indicating that it was correctly folded and that it could be used for the study.
Antibodies The mouse mAb ESH8 (American Diagnostica, Stamford, CT, USA) against the C2 domain was used as control or as capture mAb in the anti-LC and anti-C2 reactivity assays. The mouse mAb 8860 (Baxter/Healthcare, Stamford, CT, USA) against the A2a2 domain was used as capture mAb in the anti-HC and anti-A2a2 reactivity assays. The mouse mAb F7B4A8 against the acidic region a1  was used as capture mAb in the anti-A1a1 reactivity assay. The anti-C2 domain human mAb Bo2C11 (a gift from Dr Jacquemin, KU Leuven), the anti-B domain mAb 25H3 , the anti-A2 domain mouse mAb R8B12 and the anti-A1 domain mAb GMA8002 (Green Mountain Antibodies, Burlington, VT, USA) were biotinylated and used as controls in each reactivity assay.
Antibody reactivity assays
To detect anti-HC or LC antibodies, the HC/LC duplex assay, based on the x-MAP technology, was used, as previously described . Briefly, the mAbs 8860 and ESH8 were coupled to carboxyl beads (Bio-Rad Laboratories, Hercules, CA, USA) and then incubated with EDTA-dissociated FVIII for 1 h (Fig. 1A). After three washes (with PBS), control mAbs or diluted (1/100) plasma samples from AHA patients were added to the beads for 30 min. After washing, a specific phycoerythrin-labeled anti-human IgG isotype antibody (Beckman Coulter, Fullerton, CA, USA) was used to identify the IgG subclasses (IgG1 to IgG4) of the anti-FVIII autoAbs. A second multiplexed assay (A2a2/A1a1 duplex) was designed for the detection of anti-A2a2 and anti-A1a1 antibodies. First, the mAb 8860 or F7B4A8 was immobilized on carboxyl beads and incubated with EDTA-dissociated and thrombin-cleaved FVIII for 1 h; the subsequent steps were similar to those described for the HC/LC duplex assay. Finally, in the C2 simplex assay, recombinant C2 was directly coupled to beads for the detection of anti-C2 domain antibodies as before (Fig. 1A). All multiplexed assays were performed in duplicate. Data were expressed as the mean fluorescence intensity (MFI) read by the BioPlex™ (Bio-Red Laboratories, Hercules, CA, USA) instrument. To define the positivity threshold of a given AHA plasma sample, the reactivity of a FVIII-deficient plasma standard (used as negative control, Dade Behring, Deerfield, IL, USA) was compared with the reactivity of plasma samples from 60 non-hemophilic subjects with normal coagulation tests. This evaluation allowed us to validate and use the negative control in each assay (data not shown). All plasma samples with an MFI value that exceeded the mean MFI value of the negative control plus 3 standard deviations (SD) were considered positive. Results were then expressed in terms of relative antigenic reactivity (RAR), which corresponds to the ‘patient’s MFI value/negative control mean MFI value + 3 SD’ ratio. Any plasma sample with an RAR > 1 was considered as positive for the presence of autoAbs.
Median values and ranges were used to describe continuous variables and frequencies for categorical variables. AHA groups with different underlying conditions were compared with the Kruskal–Wallis test. The Wilcoxon and Fisher’s tests were used to detect differences in, respectively, continuous and categorized variables of patients with postpartum AHA and patients with AHA associated with other conditions (malignancy, autoimmune disease and idiopathic groups). A multiple testing correction was applied to adjust the P-values. Spearman’s correlation coefficients were computed to examine the association between inhibitor titers (BU mL−1) and the RAR of the different IgG subclasses towards each FVIII fragment. The significance level was set at 5% for all analyses.
Patients’ clinical and laboratory characteristics
Idiopathic AHA was diagnosed in 49% of patients (n = 36/73). The remaining AHA cases were associated with a solid (i.e. breast, prostate, pancreas or gall-bladder) or hematologic (i.e., lymphoma or leukemia) cancer (n = 16/73; 22%) or with other A.I.D. (Sjögren’s syndrome, hypothyroidism or thrombocytopenic purpura) (n = 11/73; 15%) or it occurred during the postpartum period (median elapsed time after delivery = 2 months) (n = 10/73; 14%). The autoAb titer at diagnosis ranged from 1 to 2320 BU mL−1and the median FVIII activity was 2% (Table 1). Patients with AHA associated with malignancies had the highest autoAb titers (median = 29 BU mL−1) and those with postpartum AHA the lowest titers (median = 4.5 BU mL−1), but this difference was not significant.
Prevalence of anti-FVIII antibodies: higher occurrence of anti-A1a1 domain antibodies in postpartum AHA
Three multiplexed assays (Fig. 1A) were developed to measure the binding of four IgG subclasses (IgG1 to IgG4) to different FVIII fragments (HC and LC, A2a2 and A1a1, and C2) in plasma samples. These assays were validated by assessing the reactivity of the control mAbs 25H3 and Bo2C11 (HC/LC duplex assay), R8B12 and GMA8002 (A2a2/A1a1 duplex assay) and ESH8 (C2 simplex assay) towards their target FVIII fragments (Fig. 1B). The fluorescence intensity values in the three multiplexed assays showed that each control mAb recognized specifically its target FVIII fragment without any reactivity towards the other FVIII regions (Fig. 1B) and confirmed the correct EDTA-dissociation and thrombin-cleavage of FVIII.
The analysis of the 73 plasma samples showed that anti-LC (a3A3-C1-C2) and anti-HC (A1a1-A2a2-B) autoAbs were predominantly of the IgG4 subclass in at least 75% of patients (Table 2). An IgG1 anti-LC and anti-HC immune response was also detected, whereas the IgG3 and IgG2 responses were less frequent (e.g. anti- A1a1 domain IgG1, 37%; IgG3, 31%; IgG2, 9%; Table 2).
Table 2. Frequency of the different IgG subclasses of autoantibodies against the LC and HC fragments of FVIII
Then, patients were divided into two groups (women with postpartum AHA and patients with AHA associated with other conditions) and the specific IgG subclass distribution of the autoAbs against the different FVIII fragments (LC, HC) and domains (C2, A2a2, A1a1) in their plasma samples was analyzed. While the distribution of the IgG subclasses that targeted the LC (a3A3-C1-C2 or C2 alone) was similar, as indicated by the two largely overlapping areas in the radar chart (Fig. 2A), binding of the IgG1 and IgG3 subclasses to the HC (A1a1-A2a2-B) was more frequently observed in the postpartum AHA group (80% for both IgG subclasses) than in the other AHA group (32% and 19% for IgG1 and IgG3, respectively). Similarly, IgG1 and IgG3 reactivity towards the A1a1 domain was detected in 89% of patients with postpartum AHA (for both subclasses) compared with 29% and 22% in the other AHA group (Fig. 2B, P <0.01).
High level of autoimmune reactivity towards the A1a1 domain in patients with postpartum AHA
The level of antibody binding (expressed as RAR) to each FVIII fragment (LC, HC) and domain (C2, A2a2, A1a1) was then compared in the different AHA groups. Again, no significant difference in the IgG profiles towards the LC (a3A3-C1-C2 and C2 alone) and the A2a2 domain was found (Table 3). However, the anti-HC autoAb reactivity was significantly different in the four groups (e.g. for anti-A1a1-A2a2-B IgG1 and IgG3, P =0.018 and P =0.005, respectively; Kruskal-Wallis test, Table 3). These differences were mainly due to a stronger reactivity of IgG1 and IgG3 towards the A1a1-A2a2-B FVIII moiety in samples from patients with postpartum AHA compared with the other AHA groups (IgG1, P =0.0042; IgG3, P =0.0028; Wilcoxon test, Table 3) even after adjustment for gender (P <0.01). This difference seemed to be due to the significantly higher anti-A1a1 domain IgG response in the postpartum AHA group than in the other AHA groups (IgG1, IgG3 and IgG4, P =0.0028; IgG2, P =0.0017; Wilcoxon test, Table 3), even after adjustment for gender (P <0.01). Analysis of the IgG subclass distribution and epitope profiles of the autoAbs in the postpartum AHA group and in the group that included all the other AHA patients (Fig. 3A,B) confirmed the high involvement of IgG4 in AHA and showed that, in patients with postpartum AHA, anti-A1a1 domain IgG1 and IgG3 antibodies were also important components of the immune response. Indeed, their median RAR level was 6.8 and 3.0, respectively, in the postpartum AHA group (Fig. 3A), whereas it was < 1 in the other AHA group (Fig. 3B).
Table 3. Relative antigenic reactivity (RAR) of anti-FVIII autoantibodies in the different groups of patients with AHA
Correlation between autoantibody titers and IgG reactivity
The inhibitor titer (BU mL−1) of the 73 plasma samples and the reactivity (as RAR) of each subclass of IgG autoAbs were not correlated. However, when each AHA group was analyzed separately, strong correlations between the IgG4 reactivity towards the LC and HC fragments and BU (r = 0.78, P <0.001 and r = 0.86, P <0.0001, respectively; Fig. 4A,B) were observed in patients with AHA associated with malignancies. In this group, a correlation between anti-C2 IgG4 autoAbs and BU was also observed (r = 0.82, P < 0.001, Fig. S1 and Table S1). A significant correlation between antibody titer and anti-FVIII-LC IgG4 was also found in idiopathic AHA (r = 0.63, P < 0.001, Fig. 4G).
Here, we employed recently developed immunoassays that allow mapping conformational epitopes in the A2, C2 and A1 domains of FVIII that are recognized by both neutralizing and non-neutralizing antibodies. We could screen 73 patients with AHA, with variable underlying conditions. This is the largest cohort examined so far, as previous studies that investigated anti-FVIII auto-Abs in AHA patients included much smaller populations (one to 22 individuals) [30–34]. Moreover, some of the previously used approaches have some weaknesses. For instance, the presence of NNA could not be investigated when activity-based immunological assays were employed, whereas antibodies that recognize discontinuous epitopes were overlooked by peptide-based methods. Our results confirm the antigenicity of the C2 and A2a2 domains as antibodies against these domains were found in 82% and 70%, respectively, of patients. However, the reactivity of anti-C2 antibodies was relatively lower compared with that of anti-LC antibodies, suggesting that, in AHA patients, autoAbs target mainly epitopes outside the C2 domain, such as the A3 domain. We also found a high prevalence (78%) of antibody reactivity against the A1a1 domain of FVIII.
Anti-FVIII Abs are usually of the IgG4 and, less commonly, of the IgG1 subclass. IgG4 appear to be predominant in inhibitor neutralization tests and immunoblotting studies [35–37] and our results are in agreement with these findings. The IgG1 antibody response was less frequent and both IgG1 and IgG4 were directed against the LC and HC of FVIII in most patients. In congenital hemophilia, the levels of anti-FVIII IgG1 and IgG4 correlate well with the inhibitor titers as measured by the Bethesda assay . Similarly, in a previous study , the level of IgG4 autoAbs, assessed by ELISA, was correlated with the inhibitor titer also in patients with AHA. Conversely, such a correlation was not found in the 73 AHA patients with anti-FVIII autoAbs that we studied. As both ELISA and x-MAP procedures measure inhibitory and non-inhibitory anti-FVIII antibodies, the different number of patients in the two studies (11 vs. 73) could explain this discrepancy. However, when only the 16 patients with malignancy-associated AHA were considered, a strong correlation was found between inhibitor titer and IgG4 reactivity towards the LC, HC and C2 domains (r = 0.78, 0.86 and 0.82, respectively). These patients also had the highest inhibitor titers compared with the other groups.
The development of autoAbs in patients with AHA is associated in 50% of cases with a variety of underlying diseases . Matsumoto et al.  suggested differences in the immunological characteristics of patients with AHA associated (n = 7) or not with a disease (n = 9). Our analysis of plasma samples from 73 AHA patients confirms that a differential immunological profile exists; however, it is different from the one proposed by Matsumoto et al. Our most important observation is that autoAbs in postpartum AHA are directed particularly towards the A1a1-A2a2-B fragment and target specifically the A1a1 domain. The higher reactivity of anti-A1a1 autoAbs in comparison to anti-HC autoAbs could be due to variations in the molecule presentation. Indeed, the A1 domain of thrombin-activated FVIII is probably more accessible in the A2a2/A1a1 than in the HC/LC duplex assay. In postpartum patients, anti-FVIII-HC IgG1 and IgG3 are more often present (80%) and more reactive than in the other AHA groups (frequency and reactivity, P <0.01). Conversely, in the study by Matsumoto et al., the plasma samples of only three patients contained anti-FVIII-HC antibodies. Two of these patients had postpartum AHA with anti-FVIII-HC autoAbs, in accordance with our findings. Postpartum AHA is a rare event and the etiology of pregnancy-related FVIII autoAbs remains unclear [39–42]. As these autoAbs usually develop in the postpartum period, a plausible hypothesis is that the mother is exposed to fetal FVIII during delivery . In our study, production of Th2-driven IgG4 autoAbs in AHA was observed in most patients; on the other hand, in postpartum AHA, Th1-driven IgG1 and IgG3 were also important components of the immune response. During pregnancy the maternal immune system is modified to achieve immune tolerance towards paternal antigens that are expressed by fetal cells. Hormonal changes during pregnancy are associated with increased production of Th2 cytokines and reduced expression of Th1 cytokines, resulting in a Th2 polarization of the immune response . A Th1-Th2 shift might explain why Th2-mediated autoimmune diseases, such as systemic lupus erythematosus, tend to develop or worsen during pregnancy , while Th1-mediated diseases, such as rheumatoid arthritis, tend to improve . In the postpartum period, there is a switch from a Th2- to a Th1-mediated immune response that favors, for example, the occurrence of postpartum thyroiditis . The higher occurrence and reactivity of anti-FVIII-HC autoAbs of the IgG1 and IgG3 subclasses in postpartum AHA in our study might be explained by the greater involvement of the Th1-driven response in this group of patients compared with other AHA groups. Moreover, Reding et al.  suggested a larger contribution of the Th1-driven response in successfully treated HA and AHA. This feature could play a role in the good outcome of postpartum AHA . Finally, the higher anti-A1a1 domain reactivity and the lack of correlation between inhibitor titer and IgG4 autoAbs suggest the presence of NNA.
Due to the limited size of our patients’ subgroups our conclusions should be considered cautiously and larger cohorts of patients are needed to strengthen these findings. Further studies are also required to compare the Th1/Th2 cytokines profiles in different groups of AHA patients  to confirm that pregnancy/delivery may lead to postpartum exacerbation of the Th1-driven immune response. Indeed, the Th1 and Th2 cellular immune responses provide a useful model for explaining the different pathogenic mechanisms involved in the development of AHA associated with different clinical conditions.
PL performed most of the laboratory work. TA performed the C2 simplex laboratory work. CC and CP performed methodology works. EMD and JB were in charge of the statistical analysis. HL and JYB coordinated and promoted the SACHA study. YG revised the article. JFS, CG and GLL designed the study and coordinated the research. PL, CG and GLL wrote the paper.
We wish to thank C. Combescure (Geneva, Switzerland) for performing the first statistical analysis of the preliminary data, and B. Guillet (Rennes, France) and S. Lacroix-Desmazes (Paris, France) for revising our preliminary study. We are very grateful to Novonordisk® and LFB, France, for their support.
Disclosure of Conflict of Interests
The authors state that they have no conflict of interest.