Blood samples from 11 severe and moderately severe HA subjects were donated as part of the “INHIBIT” feasibility study (5U34HL 114674, MVR, PI). Three of these were known Int22Inv subjects, while the HA-causing mutation was unknown for six of these subjects. PBMCs from severe HA and normal control subjects that were banked from earlier studies were also utilized (USU IRB#1, protocols MED-83-3442 and MED-83-3426). All subjects provided informed consent consistent with the Principles of Helsinki. Additional de-identified normal control samples were obtained from the NIH Blood Bank (Exempt E4). Both freshly prepared and frozen (7% DMSO in FBS) PBMC samples were analyzed.

Inhibitor titers, expressed as Bethesda units (BU)/mL, were from clinical records for the subjects, which reported results of CLIA-certified clotting or chromogenic substrate assays (26). Both clotting and chromogenic inhibitor assays are functional assays that measure the inhibition of FVIII cofactor activity in the presence of anti-FVIII antibodies. The clotting assay measures plasma clotting time [using a 1-stage assay (27)], while the chromogenic assay measures cleavage of a chromogenic substrate for factor X; normal factor X kinetics require a normal FVIII level. In either assay, serial dilutions of the patient plasma are mixed with a normal pooled plasma sample, and the residual FVIII activity, compared to a parallel assay using the normal pooled plasma plus buffer, is measured and reported in BU/mL. One BU/mL is the reciprocal of the patient plasma dilution that results in 50% normal FVIII activity in the assay. The lower limit of quantification is generally considered to be 0.6 BU/mL, and titers > 5.0 BU/mL are considered high-titer inhibitors.

Recombinant (r)FVIII (Kogenate 27NZPO) was from Bayer (Leverkusen, Germany) or Baxter (Westlake Village, CA, USA; Recombinate TRH07504AA). Recombinant FVIII-C2 protein was purified in-house as described (28). Tetanus and Diptheria (TD) toxoid adsorbed was from Mass Biologics, (Boston, MA, USA) and Remel PHA (R30852801) (100μg/ml) were from Thermo Scientific (Rockford, IL, USA). Benzonase nuclease, 250 U/μl (021M0852) and 0.4% Trypan blue (T8154) were from Sigma Aldrich (St Louis, MO, USA). RPMI 1640 with 25 mM HEPES, DMEM/F-12(1(1:1), MEM nonessential amino acids (100X), 1X D-PBS (Ca/Mg-free) and 10X D-PBS were from Life Technologies (Grand Island, NY, USA). Ficoll-Paque PLUS was from GE Healthcare Life Sciences (Piscataway, NJ, USA). HCl was from JT Baker Tyco Products (Phillipsburg, NJ, USA). Human IFN-gamma ELISPOT pair (551849), Human IL-5 ELISPOT pair (551886), Human IL-10 ELISPOT pair (551883), AEC substrate set (551951) and TMB substrate were from BD Biosciences (San Jose, CA, USA). IL-7 (0.5ng/ml) was from Peprotech (Rocky Hill, NJ, USA). Fetal bovine serum was from Hyclone (Logan, UT, USA). ImmunO human serum type AB was from MP Biomedicals (Solon, OH, USA). Recombinant, extracellular domains of 10 HLA-DRB1 protein monomers, HLA Class II tetramers loaded with the relevant peptides and anti-HLA-DR antibody L243 were provided by the Tetramer Core Facility at the Benaroya Research Institute (Seattle, WA, USA). AIM-V medium and goat anti human IgG-HRP were from Invitrogen (Carlsbad, CA, USA). Sheep anti-human FVIII was from Cedarlane labs (Burlington, NC, USA). ELISA 12 well strips were from E & K Scientific (Santa Clara, CA, USA). Wallac enhancement solution, fluorescence assay buffer and europium-labeled straptavidin were from Wallac-Perkin Elmer (Turku, Finland). 96-well polypropylene plates and Corning Costar high-binding ELISA plates were from Sigma Aldrich (St. Louis, MO, USA), cat # CLS3799 and CLS2592, respectively). Recombinant human IL-2 (Chiron Il-2, Proleukin, 70 IU/ml) was from R & D systems, cat #202-IL-010)

Individual 20-mer peptides spanning the FVIII A2, C1 and C2 sequences, as well as equimolar pools containing 2-5 of these peptides, were from Anaspec, Inc. (Fremont, CA, USA, Table S1 ). Additional 15-mer peptides spanning the FVIII A1, A2, A3, C1 and C2 domain sequences, with 12-residue overlaps, were from Mimotopes, Inc. (Mulgrave, Victoria, Australia), which also provided large, medium and small equimolar pools of these peptides ( Tables S2–S5 ). An intron 22-23 ‘junction peptide pool’ was created by combining four 20-mer peptides (Mimotopes, Inc.) with overlapping sequences spanning the F8 exon 22-23 – encoded region (FVIII 2103-2146) in DMSO ( Table 1 ). All peptides were ordered at >70% advertised purity and their quality was verified by mass spectrometry.

Quantitative peptide-MHC competition binding assays were carried out to determine binding avidities of four 20-mer peptides spanning the F8 exon 22-23 junction region to 10 HLA-DRB1 proteins using methodology described previously (29). Briefly, 1-µl aliquots of serially diluted (50nM – 50μM in DMSO) non-biotinylated FVIII peptides, were added to triplicate wells of 96-well polypropylene plates. Serially diluted non-FVIII reference peptides known to bind to specific HLA-DRB1 ( Table S6 ) were added to separate wells as positive controls. The HLA-DRB1 proteins were diluted to 50 nM in 150 mM citrate-phosphate, pH 5.4, 0.75% n-octyl-beta-D-glucopyranoside, 1 mM PMSF. 50 μl of the HLA-DRB1 solution was added per well, and the plates were sealed and incubated at 37°C for 30 min to allow peptide binding to the HLA-DRB1 protein. Next, 1 μl of the appropriate biotinylated reference peptide, at a concentration in the known binding range for its respective HLA-DRB1 ( Table S6 ), was added to each well, the solutions were mixed, and the plates were incubated at 37°C for 18-24 hrs. 100 μl aliquots of the anti-HLA-DR antibody L243 (10 μg/mL in 12.5 mM borate buffer, pH 8.2) were added to 96-well ELISA plates, and the plates were incubated at 4°C for 18-24 hrs, washed, and blocked with 1xDPBS containing 5% FBS for 3 hrs. The plates were washed again and 50 μl neutralization solution (50 mM Tris, pH 8.0, 0.75% n-octyl-beta-D-glucopyranoside) was added to each well. The contents of wells containing the peptide-HLA-DRB1 binding reactions were then transferred to wells of the ELISA plates containing the neutralization solution and the ELISA plates were incubated for 18-24 hrs at 4°C. After washing the plates 5X with 300 ul/well of 1XPBS, 0.05% Tween-20, 100 ul of Europium-labeled streptavidin (diluted 1:1,000 into Wallac assay buffer) was added to each well. The plates were covered with a black polystyrene lid, incubated at room temperature for 4 hrs and washed 5X with 300 ul/well of 1XPBS, 0.05% Tween-20. Wallac enhancement solution (100 μl/well) was then added, and the plates were incubated at room temperature for 30 min and then read on a Victor 2D time-resolved fluorometer. Sigmoidal binding curves were simulated and IC50 values calculated for the FVIII peptides, based on their competition with the reference peptides for each HLA-DRB1.

Predicted affinities of the four FVIII junction peptides for the 10 HLA-DRB1 alleles were obtained using the Immune Epitope Database (IEDB) server, using the NetMHCIIPan 4.0 prediction algorithm (30).

FVIII-specific CD4⁺ T cells were expanded by culturing PBMCs with or without rFVIII, rFVIII-C2 protein, or pooled or individual FVIII peptides, as follows. One vial of frozen PBMCs (~10 million cells) was thawed by slowly adding 9 ml of AIM-V medium containing 10% human serum + 1.8 μl of 25K benzonase nuclease. The cells were washed with AIM-V medium containing 10% human serum and re-suspended in AIM-V medium containing 15% human serum, and then plated at 1x10⁶ cells per well in 48-well cell culture plates (1 mL/well) with one of the following: rFVIII protein (1 μg/mL ~8nM), rFVIII-C2 protein (2 μg/mL = 50 nM), pooled FVIII peptides (50 nM and 70 nM), FVIII pooled junction peptides (200 nM), or individual FVIII peptides (50 or 100 nM). Positive control wells were stimulated with 0.02Lf (5 μl/mL) TD toxoid, PHA (10 μg/mL) and negative controls with 5 μl DMSO. The plates were incubated for five days at 37°C in a 5% CO₂ incubator. At day 6, cells were harvested and re-suspended in AIM-V medium (no serum) containing 0.5ng/mL IL-7 and then directly plated on 96-well flat-bottom ELISPOT plates. For some individual peptide ELISPOT assays, expansions were continued for an additional 3 days, adding fresh AIM-V medium and IL-2 (3.5 IU/mL final concentration).

ELISPOT plates were coated with capture antibody solutions (anti-IFN-γ, anti-IL-5 or anti-IL-10) diluted to 5 μg/ml in DPBS and incubated at 4C overnight. Wells were then washed once with 200 μl/well Blocking Solution (AIM-V medium containing 10% human serum). Another 200 μl Blocking Solution was then added to each well and the plates were incubated for 2 hrs at room temp and the Blocking solution discarded. The expanded PBMCs were then added in serial dilutions (carried out in duplicate when sufficient cells were available) with final concentrations of 2X10⁵, 1x10⁵ and 5x10⁴ cells/well, and incubated with the same antigenic stimulants as before. The plates were incubated for 24 hrs for IFN-γ ELISPOTS or 48 hrs for IL-5 and IL-10 ELISPOTS at 37C, 5% CO₂. The suspended cells were then discarded and the plates washed with 200 μl/well dH₂O (2X) and 3x with PBS containing 0.05% Tween 20 (wash buffer I). The biotinylated anti-human antibody provided in the kits (0.25 μg/ml, diluted in PBS + 10% human serum) was then added and the plates were incubated for 2 hrs at room temperature. The plates were washed 3X with 200 μl/well wash buffer I, washed 2X with 200 μl/well PBS, and developed with streptavidin-HRP (BD Biosciences, #557630) diluted 1:100 in PBS containing 10% human serum for 1 hr at room temperature. Plates were then washed 4X with 200 μl/well wash buffer I, and washed 2X with 200 μl/well PBS. Spots were then developed by adding 100 μl/well of BD ELISPOT AEC substrate and incubated for 5-30 minutes in the dark. The wells were then washed 2X with 200 μl/well dH₂0 and air-dried at room temp for 2 hrs. IFN-γ, IL-5 and IL-10-specific spot-forming cells (SFC) were then counted using software on a CTL Immunospot S6 Ultimate Image Analyzer. Stimulants were: rFVIII 5 μl/mL = 5 nM; rFVIII-C2 1μg/mL = 50 nM; FVIII peptides (pooled or individual): 50-100 nM; TT; 5 μl/mL; PHA: 5 μl/mL. The criteria for antigen-specific (positive) responses were: a minimum of 25 spots per million cells (based on spots per 200,000 cells/well) and wells with the stimulant had ≥3X the average number of spots counted for the unstimulated wells. In addition, antigen-specific responses to large FVIII peptide pools were “decoded”, when sufficient cells were available, by subsequent ELISPOT assays using smaller peptide pools and then with individual peptides, to map immunogenic FVIII regions and specific epitopes contributing to the cytokine response.

PBMCs were added to 48-well plates in 1ml of 10% human serum RPMI medium (1 million PBMCs per well) and then stimulated with individual or pooled 20-mer or 15-mer FVIII peptides (10 μM) and incubated for up to 19 days at 37C, 5% CO₂. At day 7, 50 μl of IL-2 was added (35 IU/mL final concentration) without removing any medium. At day 9, 400-500 μL supernatant was removed and replaced with fresh 15% human serum RPMI medium containing 50 μl IL-2 after which cells were fed with 50 μl IL-2 every 48-72 hours, splitting as needed. After 17-19 days, the cells were re-suspended, 100 μL aliquots were transferred to FACS tubes, and 1.5 μl of the appropriate peptide-loaded, PE-labeled HLA Class II tetramer (~10 μg/mL final concentration) was added. The tubes were incubated in the dark at 37C for 2 hrs and then put on ice. A mixture of anti-CD3-PerCP + anti-CD4 APC (3.75 μL each from stock solutions) was added to each tube, and tubes with no antibody added served as a negative staining control. The tubes were then incubated in the dark at 4C for 30 min. Additional tubes containing beads for compensation (ThermoFisher (ebiosciences), Rockville, MD, USA), were stained by adding 5 μL of anti-CD3-PerCP, 1.5 μl of anti-CD4-APC and 1.5 μl of anti-CD4-PE and tubes were incubated in the dark for 30 min. Sample and control cells were then washed and re-suspended in ice-cold FACS buffer, stored on ice in a covered container, and analyzed by flow cytometry and/or FACS acquisition. The gating strategy was: Singlets (FSC-H/FSC-A), Lymphocytes (SSC-A/FSC-A), followed by gating on anti-CD3-PerCP-high, followed by tetramer-PE/CD4-APC to detect antigen-specific CD4+ T cells.

To isolate FVIII-specific T-cell clones, CD4^hiTetramer^hi cells were single-cell sorted into 96-well round-bottom plates containing 100 μl RPMI medium per well. To each well, (200,000) irradiated HLA-mismatched PBMC were added as feeder cells in a volume of 100 μL RPMI Medium containing 15% human serum AB plus PHA (5 μg/mL). The cells were incubated for 24 hrs at 37C, 5% CO₂, and then IL-2 (10 IU/mL) in 25 μl RPMI medium was added to each well. The cells were then incubated at 37C in 5% CO₂ for another 14-21 days. For wells containing expanding clones, tetramer staining was performed to confirm their specificity. If clones were not expanding, the cells were re-stimulated with irradiated, HLA-mismatched PBMCs and PHA as before, adding fresh medium and IL-2 (3.5 IU/mL final concentration) as needed to continue the expansion or else frozen in 10% DMSO in FBS.

The experimental FVIII peptide-HLA-DRB1 affinities are in Table 2A and their predicted affinities are in Table 2B . Only HLA-DR0404, DR0701, DR0901 and DR1501 showed moderate or strong binding affinity to one or more peptides spanning the F8 exon 22-23-encoded region of FVIII, while weak binding was observed to HLA-DR0101 and DR0401. The MHCIIpan algorithm (30) predicted moderate-to-strong binding affinity to additional HLA-DR proteins. Based on the experimental results, ELISPOT assays were then carried out using PBMCs from nine HA subjects with an intron 22 inversion mutation who were also known to have HLA-DR1, DR4, DR7, DR9 or DR15 alleles.

Several strategies were used to map T-cell epitopes, in order to test specific hypotheses and/or to rationally choose which peptides to utilize when there were not sufficient PBMCs to systematically test responses to larger and then smaller peptide pools followed by individual peptides (due to competing studies and resulting blood volume limitations in subjects with hemophilia A). The number of PBMCs per subject was highly variable, as samples were obtained from both adults and children, and the amounts of blood drawn also reflected the stated preferences of the subjects. Therefore, the number of experiments and replicate wells per experiment varied somewhat according to the number of cells available for the assays.

rFVIII and/or rFVIII-C2 domain protein ELISPOTs were included as a specific control in almost all ELISPOT experiments. For positive responders, IFN-γ secretion was generally more robust when cells were stimulated with 50 nM rFVIII-C2 than with 8 nM rFVIII. Unfortunately, FVIII protein becomes toxic to CD4⁺ T cells at higher concentrations. The stronger responses to rFVIII-C2, and to pooled or individual FVIII peptides in some assays, may simply be due to stimulation of lower-avidity CD4⁺ T cells by the higher concentrations of rFVIII-C2 and FVIII peptides. Of note, the rFVIII-C2 preps were routinely tested to confirm low endotoxin levels (31).

Potential CD4⁺ T-cell responses to exon 22-23 junction sequences were tested using samples from nine HA subjects with an intron-22 inversion mutation who also had one of the HLA-DRB1 alleles shown experimentally to bind peptides with these sequences: HLA-DR1, DR4, DR7, DR9 or DR15 ( Table 2A ). Eight of these subjects failed to show a response to the pooled FVIII exon 22-23 junction peptides when tested using IFN-γ ELISPOT assays ( Table 3 ). Five of them had a current inhibitor and four had no inhibitor history. The one subject who showed a response to the junction peptides, G4, had a low-titer inhibitor. He had an intron 22 inversion mutation as well as a missense mutation, H1499Y, and his HLA-DRB1 alleles were HLA-DRB1*0301, 1501. Therefore, additional MHCII binding predictions were made using the IEDB analysis resource Consensus tool (32, 33) with the input sequence FVIII 1485-1513, lpktsgkvellpkvhiyqkdlfptetsng, which contains all possible 15-mer sequences containing residue H1499. Interestingly, the IEDB prediction algorithm identified two overlapping 9-mer core sequences containing H1499 as potential HLA-DRB1*1501-restrictred CD4⁺ T-cell epitopes, ranking them in the top 15-20% based on predicted binding affinity ( Table S7 ).

Subjects P1 and P11 (both with an intron-22 inversion) generously donated sufficient blood to carry out fairly comprehensive epitope mapping by ELISPOT assays. P1 had a past inhibitor that had resolved (peak titer 15.2 BU/mL approximately 10 years prior to donating this sample). IFN-γ ELISPOT assays showed robust responses to both rFVIII and rFVIII-C2 protein ( Figure 2A ). His responses to A2 and C2 domain peptides were tested next, in order to test the hypotheses that tolerance to A2 epitopes was conferred by translation from the inverted F8 mRNA (exons 1-22, containing the A2 domain sequence) and/or from F8B mRNA (exons 23-26, containing the C2 domain sequence). No responses to pooled A2 domain peptides were revealed under these assay conditions (not shown). However, his PBMCs responded to one of the pools of 20-mer peptides spanning the C2 domain sequence ( Figure 2B ). Decoding of this response by testing the individual peptides comprising this pool identified robust responses to peptides corresponding to C2-2210-2229, C2-2226-2245 and C2-2242-2261 ( Figure 2C ).

Subject P11 also had a previous inhibitor that resolved four years before donating this blood sample. His PBMCs showed robust IFN-γ secretion when stimulated with rFVIII, rFVIII-C2 protein, peptide pools spanning the FVIII A2 domain sequence, and peptide pools spanning the FVIII C2 domain sequence ( Figures 3A–C ). His PBMCs also responded to rFVIII stimulation with secretion of IL-5 and IL-10 ( Figure 3D ).

IFN-γ ELISPOT assays were carried out to determine if PBMCs from six severe HA subjects with an HLA-DR11 allele would respond to a known HLA-DRB1*1101-restricted T-cell epitope in the FVIII A2 domain, FVIII-589-608, which was identified in a HA subject with missense substitution R593C (20). The overlapping 20-mer peptides FVIII-A2-28 and/or A2-29 both contained the epitope of interest. Two of the subjects had an Int22Inv mutation, two had a frameshift mutation, and the HA-causing mutation was unknown for the other two. One of the Int22Inv subjects responded to peptide A2-29, and both of the Int22Inv subjects also showed a robust response to the FVIII-C2 protein ( Figure S1 ). Together, these results indicated the Int22Inv mutation did not confer tolerance to either the FVIII A2 or C2 domains.

ELISPOT assays were first carried out for subject G21, who had an Int22Inv mutation and was monogenic for HLA-DRB1*0301. His PBMCs showed positive IFN-γ responses to rFVIII and to large peptide pools LP1, LP2 and LP3, which together span FVIII residues 1-737 ( Figure 4A ). The FVIII heavy chain consists of the A1 and A2 domains, residues 1-740. There were not enough PBMCs to allow systematic decoding via ELISPOT assays using smaller peptide pools and then individual peptides covering 740 residues. Therefore, the Immune Epitope Database (IEDB) server (33) was used to predict peptides within the FVIII A1 and A2 domains that would bind with high affinity to HLA-DR0301. Twenty-five 15-mer FVIII peptides from our library contained a total of 18 unique 9-mer motifs that were predicted to be in the top 2% for binding to HLA-DR0301 ( Table S8 ). In several cases where overlapping 15-mer peptides contained the same motif the 2-3 peptides were pooled, generating a total of 18 individual or pooled FVIII peptides for further testing ( Table S9 ). Stimulation of his PBMCs with four of these individual peptides, A1-41, A1-58, A2-47 and A2-59, produced IFN-γ secretion slightly above background levels ( Figure 4B ).

Next, PBMCs from three additional Int22Inv HA subjects who all had a current inhibitor and an HLA-DR3 allele, G16, G17 and G18, were tested for IFN-γ responses to these 4 peptides by ELISPOT assays ( Figure 4C ). The strongest response was from G18, responding to peptide A2-59. Therefore, HLA-DR0301 tetramers loaded with the 4 pooled peptides, and with individual peptide A2-59, were ordered.

PBMCs from 11 severe and moderate HA subjects, three of whom had a confirmed Int22Inv mutation, were tested for responses to rFVIII and rFVIII-C2 protein ( Figure S2 ). Eight of them showed robust cytokine secretion in response to rFVIII and 9 responded to rFVIII-C2. In general, comparing results of ELISPOT experiments testing reactivity to rFVIII and rFVIII-C2, among the positive responders IFN-γ secretion was usually more robust when cells were stimulated with 50 nM rFVIII-C2 than with 8 nM rFVIII. Unfortunately, FVIII protein becomes toxic to CD4 T cells at higher concentrations. The stronger responses to rFVIII-C2, and to pooled or individual FVIII peptides in some assays, may simply be due to stimulation of lower-avidity CD4 T cells by the higher concentrations of rFVIII-C2 and FVIII peptides. Of note, the rFVIII-C2 preps were routinely tested to confirm low endotoxin levels.

Control ELISPOT experiments indicated that IFN-γ secretion from non-HA PBMCs in response rFVIII stimulation under these assay conditions ( Figure S3 ) was rare, while responses of non-HA PBMCs to rFVIII-C2 protein (included as a control in several experiments) were seen more frequently but still were not common.

Two severe HA and two non-HA control subjects showed high background IFN-γ secretion (seen in ELISPOT wells with DMSO alone added); they were therefore excluded from this study and are not included in any figures or tables.

HLA-Class II tetramer staining was carried out using CD4⁺ T cells isolated from four subjects with an Int22Inv mutation. Subject G16 was an adult inhibitor subject who had an HLA-DRB1*0301, 1201g haplotype. Subject G19 was an adult inhibitor subject with HLA-DRB1*0302, 0901 haplotype. Subject G20 (age range 4-17 years) had the HLA-DRB1*0301, 1101 haplotype and a current inhibitor. Subject G21 (age range 4-17 years) had the HLA-DRB1*0301, 0301 haplotype and no inhibitor history. Tetramer staining was also carried out for subject G18, an adult with an intron 1 inversion mutation, a current inhibitor, and HLA-DRB1*0301, 1401g haplotype, in order to further test for HLA-DRB1*0301-restricted epitopes in FVIII. Tetramer staining of CD4⁺ T cells isolated from these subjects’ PBMCs was followed by sorting of tetramer-hi cells and expansion of clones in culture.

PBMCs from Int22Inv subject G20 were queried using HLA-DR1101 tetramers loaded with peptide pool A2-6, which consisted of four 20-mer peptides spanning FVIII residues 565-616. This region contains a known HLA-DRB1*1101-restricted T-cell epitope that was recognized by HA subjects with an R593C missense mutation (20). Figure 5A shows staining of representative T-cell clones with HLA-DR1101 tetramers loaded with peptide pool A2-6, obtained following expansion of CD4⁺ T cells using this same peptide pool. Results clearly indicate recognition of one or more epitopes in the region FVIII 565-616.

CD4⁺ T-cell clones were isolated from Int22Inv subject G19 following expansion of CD4⁺ T cells using a 20-mer peptide pool spanning the FVIII C2 domain region 2289-2332. Figures 5B–D shows tetramer staining of representative HLA-DRB1*0901-restricted T-cell clones recognizing these pooled peptides. A second staining of expanded clones using tetramers loaded with the individual FVIII-C2 peptides comprising this pool identified FVIII-2297-2316 as an HLA-DRB1*0901-restricted T-cell epitope recognized by CD4⁺ T cells from this subject.

PBMCs from three Int22Inv subjects and one intron 1 subject were stimulated with a pool of four immunogenic peptides identified by ELISPOT assays: A1-41, A1-58, A2-47 and A2-59 ( Figure 4B ). The expanded cells were then stained and single-cell sorted using HLA-DR0301 tetramers loaded with these pooled peptides (not shown). Clones were obtained from three of these subjects. Specificity of the staining was then confirmed using an HLA-DR0301 tetramer loaded with peptide A2-59. This tetramer produced strong staining for two clones expanded from Int22Inv subject G16, two clones from Int22Inv subject G21, and three clones from intron 1 inversion subject G18, confirming that this peptide contained a FVIII epitope. Representative tetramer staining results are in Figure 6 . The gating strategy and staining of the remaining clones are in Figure S4 .

Positive controls stimulating PBMCs from subjects with tetanus-diptheria toxoid peptides demonstrated the validity of tetramer-based epitope mapping via isolation of multiple TT-specific T-cell clones ( Figure S5 ). Negative controls for all tetramer experiments consisted of tetramers loaded with an irrelevant peptide to rule out nonspecific binding of the tetramer to CD4⁺ T cells.

Results of all assays, grouped by assay type and presented as per-subject results, as well as clinical and demographic data, are summarized in Table 4 .