Descriptive statistics of the PATH cohort
We conducted the PATH study on a cohort of 450 HA patients who self-reported either BA (45.8%) or WE (54.2%) racial/ethnic ancestry and were from the US or Canada (Table 1). The proportion of BA and WE subjects who had ever developed a FEI of any titer (AT), FEIAT+, prior to study entry (i.e., had a positive historical-FEI-status) was 35.1% and 26.6%, respectively (Table 1a), while the proportion of BA and WE subjects who had AT FEI (FEIAT+) at study entry (i.e., had a positive baseline-FEI-status) was 21.5% and 15.9%, respectively (Table 1b). Using Fisher’s method for combining p-values for the chi-square tests (p=0.06 and p=0.14, respectively), the joint tabular data are marginally not significant (p=0.06) but trend in the expected direction of greater prevalence of FEIs in HA patients with BA-racial/ethnic-ancestry.15 We identified causative F8 mutations in 406 subjects (90.2%): 188 of the 206 BA (91.3%) and 218 of the 244 WE (89.3%) subjects (Figure 1). The two most common mutations were I22-invs11,15,34,41,42,43—identified in 187 subjects (41.6%) including 70 (34.0%) and 117 (48.0%), respectively, with BA- and WE-racial/ethnic-ancestry—and the missense-SBSMs,5,10.11,12,14,44 although the later are occasionally nonsynonymous (ns)‑SNPs, especially when found in the FVIII B-domain.8,15,34,43,45 We found F8 missense-SBSMs in 97 subjects (21.6%) (Figure 1), including 60 (of 206) with BA (29.1%) and 37 (of 244) with WE (15.2%) racial/ethnic-ancestry. These 97 subjects carried 61 distinct F8 ns-SBSMs, including 15 that were previously unknown when the PATH study was conducted: 79E>Q; 155T>S; 414N>D; 669T>P; 964V>I; 979H>Q; 1193K>N; 1460G>D; 1499H>Y; 1776R>K; 1801E>K; 1846D>V; 1966R>Q; 1969E>K; and 2143P>Q.
Because the baseline-FEI-status of HA patients has never been investigated genetically previously, we estimated its heritability for the first time herein. (Though the genetics associated with FEI risk has been reported on extensively previously, the dependent variable in prior studies has been “lifetime-FEI-status”, i.e., what we refer to here as historical-FEI-status, which is: ‘YES’, if a patient has ever developed FEIs, of AT, at any time in life, irrespective of whether or not a FEI is present at ‘baseline’, on study entry; or ‘NO’, if a patient has never developed FEIs, of AT, at any time in life, including at ‘baseline’, on study entry.) We found that the residual additive genetic heritability for the presence of FEIs of AT (FEIAT+) on entry into the PATH study versus their absence (FEIAT-) was 0.47 (p<0.05) and the F8-mutation-specific heritability was 0.08 (p<0.05), the latter of which comprised ~15% of the total observable heritability (0.08/0.08+0.47). We leveraged the substantial significant additive genetic and F8-mutation-specific heritabilities for baseline-FEI-status to perform a GCAS in which the 101 distinct pleiotropic-IMD-genes on the IC—that met the criterion of having previously been implicated in the risk of developing either two or more distinct AADs (n=76) or FEIs and at least one AAD (n=21), or both (n=4)—were screened as candidate determinants for the presence or absence of FEIs (on study entry) in these previously-treated-patients (PTPs) with HA.
Novel and replicated IMD-genes/-gene-variants identified using the candidate gene/GCAS approach
Using the IC-genotypes from PATH subjects, we found that baseline-FEI-status—i.e., the presence versus absence of FEIs at enrollment (FEIAT+ vs FEIAT-)—was significantly associated with SNPs assigned to each of three pleiotropic-IMD-genes, NOS2A, B3GNT2, and CTLA4 (Figure 2), with NOS2A variations yielding the strongest signals (Table 2). Located on the proximal long-arm of chromosome (Chr)17, NOS2A comprises a centromerically-oriented, 27-exon-containing transcription unit that spans antisense-strand nucleotides 23,107,919-23,151,682 (NCBI36/hg18) (Figure 3A) and encodes the inducible isoform of nitric oxide (NO)-synthase (NOS) 2, which produces NO, a labile uncharged lipophilic gas molecule that freely passes phospholipid bilayers. Because NO signals by activating soluble guanylyl cyclase, which in turn increases synthesis of cGMP from cellular GTP, NOS2 is a vital regulator of several physiologic processes especially those in the immune system related to host defense as NO has potent microbicidal- and tumoricidal-actions.46,47 Though NOS2A had not previously been implicated in the development of FEIs, nor in their persistence and/or resolution, it likely influences the development of at least five AADs, including multiple sclerosis (MS), insulin-dependent diabetes mellitus (IDDM), rheumatoid-arthritis (RA), systemic-lupus-erythematosus (SLE) and myasthenia gravis (MG).46,47,48,49,50,51 We identified NOS2A’s involvement in the immunogenicity of tFVIIIs via three SNPs (Figure 2)—with the top being rs117382854 (p=3.2E-6), a C>G substitution at nucleotide 23,183,786, ~34 kbp upstream of its translation initiation-codon (Figure 3A)—whose minor alleles (MAs) comparably increased risk of having FEIs of AT on study entry (Table 2).
Like NOS2A, B3GNT2 likely influences the risk of at least five AADs—including IDDM, RA, Graves’-disease (GD), ankylosing spondylitis (AS), and Crohn’s disease (CD)52,53,54,55,56—and it has not previously been implicated in the development, persistence, and/or resolution of FEIs. Located on the short-arm of Chr2, B3GNT2 comprises a centromerically-oriented, two-exon-containing transcription unit spanning sense-strand nucleotides 62,196,114-62,224,730 (GRCh38/hg38) (Figure 3B) and encodes the widely expressed UDP‑GlcNAc:bGal b-1,3-N-acetyl-glucosaminyltransferase-2 enzyme (B3GNT2), the major golgi-resident glycosyltransferase which catalyzes the addition of repeating-polylactosamine-chains to N-glycan core structures of transmembrane- and secreted-glycoproteins.57 It is known that excessive immune responses are suppressed by poly-N-acetyl-lactosamine and that deletion of B3GNT2 yields hypersensitive/hyperresponsive immunocytes by markedly reducing its levels on cell-surfaces.57 B3GNT2 was identified here via three B3GNT2-assigned SNPs (Figure 2)—with the top being rs10176009 (p=5.1E-6), a C>A substitution at nucleotide 62,282,826, ~59 kb downstream of its stop-codon (Figure 3B)—whose MAs comparably increased risk of having FEIs at baseline (Table 2).
Under the GCAS-approach, we found that the presence of FEIs on study entry (FEIAT+) versus their absence (FEIAT-) was strongly associated with four CTLA4-assigned SNPs (Figure 2), with rs231780 yielding the strongest signal (p=2.3E-5) (Table 2), and suggestively associated with rs7528265, a SNP assigned to the interleukin (IL)-10 gene (IL10) (p=3.7E-3), which is closely linked to the three other Chr1 genes of the IL10-superfamily, i.e., IL19, IL20, and IL24 encoding IL-19, IL-20, and IL-24 (Table 2). Like NOS2A and B3GNT2, both CTLA4 and IL10 had previously been implicated in multiple AADs—including, e.g., RA, IDDM, and SLE58,59,60,61,62,63—but both had also been confirmed genetically to play a role in the development of FEIs.20,22,23 Located on the long-arm of Chr2, CTLA4 comprises a telomerically-oriented, four-exon-containing transcription unit which spans sense-strand nucleotides 203,867,771-203,873,965 (GRCh38/hg38) (Figure 3C) and encodes the T-cell receptor—cytotoxic T-lymphocyte-associated-protein-4 (CTLA4)—that transduces inhibitory signals to down-regulate immune responses.63 SNP rs231780 (203,871,974 A>G)—which is located at nucleotide position +475 of CTLA4 I3—revealed a statistically-significant, genotype-specific decrease in FEI risk (beta = -0.555) (Table 2). (Note that dbSNP lists “G” as the major allele of rs231780, but “A”—which was coded in the GCAS as the MA and yielded a positive beta, i.e., increased risk—is the major allele.) The other CTLA4-assigned SNPs associated with baseline-FEI-status also reside in functional gene regions including its proximal promoter, 3’-genomic-DNA, downstream of the cleavage-/polyadenylation-site encoding sequences, and I3 (Figure 3C). IL10 is a centromerically-oriented, five-exon-containing transcription-unit on the long-arm of Chr1 that spans antisense-strand nucleotides 206,767,602-206,772,494 (GRCh38/hg38) (not shown). This pleiotropic-IMD-gene is expressed by regulatory T-cells and encodes IL-10, an anti-inflammatory cytokine that helps balance immune responses, allowing clearance of infectious microbes but minimizing host damage.64 The IL10-superfamily was identified via rs7528265 (206,845,818 C>T)—an IL10-assigned SNP located ~73kbp upstream from its initiation-codon—which revealed a suggestive genotype-specific increase in FEI risk (p=3.7E-3) (not shown).
Multivariate logistic regression odds ratios
Having identified a novel role for NOS2A and B3GNT2¾and verified the previously implicated role for CTLA4 and IL10¾in FEI risk, we incorporated the top SNP assigned to each of these pleiotropic-IMD-genes altogether in a model, along with race, to evaluate their joint ability to predict baseline-FEI-status where all effects are additive. The OR and 95%-confidence interval (CI) lower- and upper-bound (LB and UB)¾reported as OR (95% CI LB, 95% CI UB)¾for rs117382854, rs10176009, rs231780, rs7528265, and race, respectively, were 2.99 (1.27, 7.02), 2.68 (1.21, 5.95), 1.41 (1.11, 1.79), 1.19 (1.01, 1.40), and 0.74 (0.57, 0.96) (Figure 4A, covariates A-E), which indicate that the MAs of three of these four IMD-gene-variations (i.e., NOS2, B3GNT2, and IL10) increased the risk of having FEIs of AT on study entry while the true MA of CTLA4 and WE-racial/ethnic-ancestry were protective.
Because the main goal of the PATH study was to confirm and further clarify the influence of race on FEI development, we sought to explicitly evaluate whether its role is mediated in part via genetic effects by Bayesian model selection of all possible interactions of race with these four pleiotropic-IMD-gene-variations.65 The best model identified involved a single interaction term between race and B3GNT2 (i.e., rs10176009), where rs117382854, rs10176009, and race, by itself, still exerted significant additive effects (Figure 4B, covariates F-J). The ORs and 95% CIs for rs117382854, rs231780, rs7528265, race×rs10176009, and race alone are respectively 3.00 (1.33, 6.77), 1.49 (1.07, 2.09), 1.20 (0.87, 1.640), 2.83 (1.47, 5.47), and 0.71 (0.54, 0.92).