Antibody signatures against viruses and microbiome reflect past and chronic exposures and associate with aging and inflammation

Summary Encounters with pathogens and other molecules can imprint long-lasting effects on our immune system, influencing future physiological outcomes. Given the wide range of microbes to which humans are exposed, their collective impact on health is not fully understood. To explore relations between exposures and biological aging and inflammation, we profiled an antibody-binding repertoire against 2,815 microbial, viral, and environmental peptides in a population cohort of 1,443 participants. Utilizing antibody-binding as a proxy for past exposures, we investigated their impact on biological aging, cell composition, and inflammation. Immune response against cytomegalovirus (CMV), rhinovirus, and gut bacteria relates with telomere length. Single-cell expression measurements identified an effect of CMV infection on the transcriptional landscape of subpopulations of CD8 and CD4 T-cells. This examination of the relationship between microbial exposures and biological aging and inflammation highlights a role for chronic infections (CMV and Epstein-Barr virus) and common pathogens (rhinoviruses and adenovirus C).


Antibody responses against rhinoviruses are correlated with longer telomeres and cellular composition changes
We explored the observed association of antibodies against rhinovirus and longer telomeres.
Due to the lack of a clear selection criterion for a single representative rhinoviral peptide, we employed a regularized linear model (lasso penalty) regression on different bootstraps of the complete dataset to identify rhinoviral peptides that were consistently associated with different telomere lengths (TLs).Our analysis revealed that one individual peptide, twist_35344, targeting 'polyprotein & VP1 capsid protein' and a score representing the breadth of antibody-bound peptides of rhinoviral origin, total number of anti-rhinovirus antibodies, were frequently selected as features in the regularized model (median among cell types of 93% and 75.8%, respectively).We then compared TLs from all individuals with the breadth of antibody-bound rhinoviral proteins and twist_35344 and found that TLs were more strongly associated with the breadth of antibody-bound rhinoviral proteins than with twist_35344 (linear-mixed model of the effect on all TL, Pbreadth_antibody=1.8x10 -11 , Ptwist_35344=1.38x10 - ).Consequently, we utilized the count score representing the breadth of rhinoviral antibodies to investigate the associations with TL.
It is worth noting that rhinoviral antibody-bound peptides are often observed in younger individuals (1,2), while cytomegalovirus (CMV) infections are more prevalent at older age.To address the potential confounding effect of age, we adjusted for CMV status and compared the strength of the association with and without adjustment for age (which was treated as a categorical variable to account for possible non-linear effects).Remarkably, the association between the breadth of rhinoviral antibodies and TL remained significant after adjusting for age (linear-mixed model of the effect on all TL, average effect breadth rhinovirus antibodies in all TLs without accounting for age=2.688x10 - , P=5.82x10 -10 ; average effect breadth rhinovirus antibodies in all TLs accounting for age effect=1.677x10 - , P=2.70x10 -5 ), indicating that the association of the presence of rhinoviral antibodies to longer telomeres is partially independent of participant age.Similarly, after matching 393 participants with antibody responses against twist_35344 (rhinoviral 'polyprotein & VP1 capsid protein') with those with no antibody response, based on the nearest age-sex match (see Methods), we still identified significant positive effects of rhinovirus on TLs (multivariable model with age, CMV and twist_35344, effectmatched=0.21,Pmatched=2.5x10 -4 , effectnotmatched=0.173,Pnotmatched=4.27x10 -4 ).Furthermore, we investigated the independence of the rhinoviral association from smoking, as smoking is often associated with increased rhinoviral infections and typically considered a factor negatively associated with TL.Our analysis revealed that the association between the breadth of rhinoviral antibodies and TL remained significant after adjusting for smoking (effect on all TL, effect=1.841x10 - , P=4.61x10 -6 ).
Overall, after accounting for CMV, smoking habits, age and sex, the breadth of rhinoviral antibodies was associated with all TLs (P<1.5x10 -0 ), but this effect was significantly different between cell types (likelihood ratio test (LRT) model with interaction term of cell type and rhinovirus vs model without, P=8.576x10 -7 ).With respect to TLs, rhinoviruses were more strongly associated with the TLs of memory T-cells (estimate=0.02,P=4.05x10 -6 ), lymphocytes (estimate=0.019,P=3x10 -5 ) and naïve T-cells (estimate=0.019,P=4.61 x10 -5 )     [Table S4].We did not find statistical significance supporting differences between age groups (LRT model with age group interaction with rhinovirus vs model without, P=0.46).
cells (effect=9.1x10 - , P=1.3x10 -3 ), among others [Fig S1A].To explore the mediating role of cell composition and TL in these associations, we conducted a mediation analysis that included both CMV infection and the breadth of rhinovirus antibodies.The results suggest that CMV and rhinoviral effects are independent and that cell composition partially mediated the changes in TL [Fig S1B].Specifically, the effect of rhinovirus on TL in naïve T-cells was found to be partially mediated by the predicted cell counts of CD8+ naïve cells, accounting for 17.1% (95% CI, 0.07-0.35) of its effect on TL.Cell composition association to CMV serostatus in single-cell data In our previous results using measured cell counts and cell counts predicted from bulk RNAseq, we found a CMV-associated expansion of CD8+ T-cells, particularly CD8+ EM, and a decrease of proliferative and naïve CD4+ T-cells.Using scRNA-seq data, we used both the low (l1)-and high (l2)-resolution cell-type-annotations predicted by Azimuth (3) to classify cells in order to closely reflect the resolution of the measured and deconvoluted blood cell counts (see Methods).At l1 level, we replicated the previously observed association [Fig 3B] between CD8+ T-cells and CMV serostatus (effect=0.74,P=4x10 -7 , FDR=3.2x10 -6 ) [Fig S2A].At l2 level, we identified four significant associations (FDR<0.05)[Fig 4A] and replicated three previously observed cell proportion-CMV associations [Fig 3B]: the negative association of