75 cohorts. Heterogeneity coefficients (I2 and Cochran’s Q) were estimated per association. Meta- analysis was conducted by pooling summary statistics for both cohorts and under random and fixed-effects assumptions using the meta R package (v4.19-0).61 FDR was estimated35 from the resulting associations. Results Antibody-bound peptide repertoires are personalized, linked to shared environments (co- housing) and time-dependent We interrogated a total of 344,000 peptides in 1,778 samples from 1,437 individuals (for 341 of whomwe have data at two time points 4-years apart) from a northern Dutch population cohort 22 (Figure 1A). After immunoprecipitation with protein A/G (binding primarily IgG antibodies21) and sequencing, we detected an enrichment of sequenced reads (compared to a null distribution without immunoprecipitation) of 175,242 (antibody-bound) peptides in at least one participant (average number of peptides bound per person = 1,168, range = 3–3,161) (see Methods). Peptide seropositivity was defined as a presence/absence binary score (enriched/not enriched) that was used for all subsequent analyses. Most antibody-bound peptides showed low seroprevalence, indicating the individual-specificity of the antibody epitope repertoire (Figure 1B). Based on peptide sequence identity and prevalence (see Methods for details), we chose 2,815 peptides for further analyses (Supplementary Table 1.1). The large variability in the antibody-bound peptide enrichment profile can be seen through a principal component analysis (PCA), where the amount of variability recovered by the first 10 components was just 15.5% and 709 components were needed to retrieve 90% of the total antibody-bound peptide variability (Figure 1C). Despite the relatively low variability accounted for by the first two PCs (6.3%), we observed two clusters in PC2 that were driven by cytomegalovirus (CMV)-related antibody-bound peptides (K-means, k = 2) (Figure 1A) (removal of these peptides resolves PC2 clustering, Supplementary Figure 1A). This is consistent with a previous observation that nearly 50% of the Dutch adult population are seropositive for this herpesvirus.62 In contrast, PC1 was highly related to the number of seropositive peptides (affine linear model R2 = 72%). In a permutational multivariate analysis of variance (PERMANOVA), the person-to-person antibodybound peptide repertoire dissimilarity showed effects of age (R2=1.4%), lifestyle phenotypes (smoking R2=0.18%), blood measurements (cholesterol R2=0.12%) and blood cell counts (lymphocytes relative abundance, R2=0.16%), among many others (while correcting for age, sex and sequencing plate) (Supplementary Table 2.1). In agreement with previous reports, we observed temporal consistency in the antibodybound peptide repertoire20,21 for the 322 LLD participants who were followed-up after 4 years. We observed that the distance between samples taken from the same individuals 4-years apart was on average lower than the distance of unrelated individuals (p < 5x10-4; 2,000 label permutations) (Figure 1D). Overall, the distance between baseline and follow-up was not associated with baseline age or sex. The temporal consistency of antibody-bound peptides showed a binomial distribution, with most peptides consistent between timepoints while a subset tended to change Determinants of the human antibody epitope repertoire
RkJQdWJsaXNoZXIy MjY0ODMw