ABO antigen and secretor statuses are not associated with gut microbiota composition in 1,500 twins

Abstract

Background: Host genetics is one of several factors known to shape human gut microbiome composition, however, the physiological processes underlying the heritability are largely unknown. Inter-individual differences in host factors secreted into the gut lumen may lead to variation in microbiome composition. One such factor is the ABO antigen. This molecule is not only expressed on the surface of red blood cells, but is also secreted from mucosal surfaces in individuals containing an intact FUT2 gene (secretors). Previous studies report differences in microbiome composition across ABO and secretor genotypes. However, due to methodological limitations, the specific bacterial taxa involved remain unknown.

Results: Here, we sought to determine the relationship of the microbiota to ABO blood group and secretor status in a large panel of 1503 individuals from a cohort of twins from the United Kingdom. Contrary to previous reports, robust associations between either ABO or secretor phenotypes and gut microbiome composition were not detected. Overall community structure, diversity, and the relative abundances of individual taxa were not significantly associated with ABO or secretor status. Additionally, joint-modeling approaches were unsuccessful in identifying combinations of taxa that were predictive of ABO or secretor status.

Conclusions: Despite previous reports, the taxonomic composition of the microbiota does not appear to be strongly associated with ABO or secretor status in 1503 individuals from the United Kingdom. These results highlight the importance of replicating microbiome-associated traits in large, well-powered cohorts to ensure results are robust.

Keywords: ABO; Blood group antigens; FUT2; Heritability; Microbiome; Secretor status.

Fig. 1

Neither ABO or secretor status associated with broad compositional differences of the gut microbiota in the TwinsUK. None of the top 100 principal coordinates (PCs) from principal coordinate analysis of unweighted UniFrac distance are significantly associated with either ABO or secretor status. The first two PCs are shown, colored by ABO status (a) and secretor status (b). c Discriminant analysis of PCA (DAPC) is largely unsuccessful at predicting ABO or secretor status from microbiome data. The mean accuracy from 5-fold cross validation is plotted for ABO status, secretor status, and ABO status only in secreting individuals (yellow). Significance was determined by comparing the accuracy of each test to the accuracies of permuted data, which took into account twin relationships (gray). Significance codes: P ≤ 0.05 = *, not significant = NS

Fig. 2

Microbiome diversity does not significantly differ by ABO or secretor status. Within sample diversity (Faith’s phylogenic diversity) is not significantly different (P > 0.05) across the ABO groups in all individuals (a), secretors versus non-secretors (b), or across ABO groups in only secreting individuals (c). d Microbiomes are more similar for siblings versus pairs of unrelated individuals, as measured by unweighted UniFrac distance. Additionally, pairs of monozygotic twins have significantly more similar microbiomes than dizygotic twins. However, microbiomes of pairs of individuals concordant for either ABO or secretor status are not more similar than for pairs of individuals who are discordant. This holds true when all individuals in the dataset are considered (“all individuals”) or when only one individual from each twin pair is examined (“one twin per family”). The total number of pairs of individual within each boxplot is indicated with “n = “. Significance codes: P ≤ 0.05 = *, P ≤ 0.01 = **, P ≤ 0.001 = ***, P ≤ 0.0001 = ****, not significant = NS

Fig. 3

The relative abundances of several Firmicutes OTUs differ in secreting individuals. a QQ-plot displaying the expected –log₁₀(P-value) compared to the –log₁₀(P-value) for all taxa tested in linear mixed models 6 (light gray points) and 8 (dark gray points). Model 6 identified taxa differentially abundant across the four ABO classes in secreting individuals only while model 8 incorporated interaction terms between ABO groups and secretor status. All differentially abundant taxa passing a significance threshold of q ≤ 0.1 are indicated with larger point sizes and the distributions of covariate-corrected, transformed relative abundances of those taxa are displayed in (b-e)