Re-examination of a published sample-set using shallow shotgun metagenomics
After establishing that 16S amplicon and shallow shotgun metagenomic sequencing yielded similar estimates of taxonomic patterns, we sought to determine whether a shallow shotgun sequencing approach could recapitulate the findings reported in a recently published 16S rRNA amplicon sequence dataset .
A primary finding of this paper was an apparent relationship between microbiome ASV-level beta-diversity and the spatial separation of sample collection locations. Spatial patterns in the microbiome were still present in an 83-sample subset of the previously published amplicon dataset (mantel test: r = 0.51, p < 0.01; Figure 5), even after controlling for sea sandwort availability (partial mantel test: r = 0.25, p < 0.01), a heterogeneously distributed dietary component shown to correlate with microbiome variation. A much weaker spatial effect was observed in the shallow shotgun metagenomic dataset (mantel test: r = 0.19,p < 0.01), and this effect was only marginally significant after controlling for sea sandwort availability(partial mantel test: r = 0.12, p = 0.04). Despite an absence of correlation in ASV/species level beta-diversity in a 13-sample set of re-sequenced DNA extracts (see above), we did observe a modest correlation in amplicon and shotgun metagenomic estimated Bray-Curtis dissimilarity in this expanded 83-sample dataset (mantel test: r = 0.44, p < 0.01). However, dissimilarity estimates in the amplicon dataset were consistently higher than those in the shallow shotgun dataset. This could be caused by a zero-inflated ASV count-table in the amplicon dataset. This pattern is common in 16S rRNA gene amplicon studies , and was evidenced by an apparent bi-modal distribution in ASV prevalence (Figure S4).
In addition to spatial effects, we previously demonstrated that horse access to sandwort appeared to affect the Sable Island horse bacterial microbiome . This pattern was still evident in a reduced 83-sample subset of these data (PERMANOVABray-Curtis: F = 5.300, R2 = 0.06, p < 0.01), and we observed an effect of comparable magnitude within the shallow shotgun metagenomic dataset (PERMANOVABray-Curtis:F = 6.146, R2 = 0.07, p< 0.01).
Analysis of 16S rRNA amplicon sequence data using ANCOM-BC tests indicated that Fibrobacteraceae , Spirochaetaceae , andPasteurellaceae were enriched in horses without access to sandwort, while horses with access to sandwort tended to possess greater abundances of Christensenellaceae , Anaeroplasmataceae ,Mycoplasmataceae , Peptococcaceae , Burkholderiaceae , and Puniceicoccaceae (Table S1A). Within the shallow shotgun dataset, Fibrobacteraceae and Nanosyncoccacae (candidate family) were significantly enriched among horses without access to sandwort, while Oxalobacteraceae and Muribaculaceae were of greater abundance in horses with access to sandwort (Table S1B). Despite non-identical patterns in the Families identified between datasets, there was a positive correlation in the test statistics (log-linear difference divided by the standard error) among families present in both datasets (R2 = 0.26, t = 3.717, p < 0.01; Figure 6).