Functional Stability and the Limits of 16S ‐Based Inference in Serpentine Biocrusts

ABSTRACT Biological soil crusts (biocrusts) shape surface ecosystem processes, but how serpentinite geochemistry and moisture structure their microbial communities remains unclear. We collected biocrust samples from serpentine and nonserpentine areas under frequent and erratic moisture regimes from the Barberton Greenstone Belt, South Africa. Bacterial communities were analyzed using 16S rRNA gene sequencing and shotgun metagenomics. Coarse bacterial diversity measures (alpha diversity and phylum‐level composition) did not differ significantly between soil types or moisture regimes. However, amplicon sequence variant (ASV)‐level analyses combining differential abundance testing, indicator‐species analysis, and Spearman's correlations ( ρ ≥ 0.75, p ≤ 0.05) identified taxa linked to soil chemistry and moisture. Three hundred sixty ASVs differed between soil types (199 enriched in serpentine) and 240 differed by moisture regime within serpentine soils (220 enriched under erratic moisture). Many serpentine‐associated microbes (e.g., Solirubrobacterales, Blastocatella , Mycobacterium , Chloroflexi clades) were linked to elevated nickel, cobalt, chromium, manganese, and iron, and sometimes to low‐calcium environments, reflecting tolerance to metal toxicity and characteristic serpentine geochemistry. In contrast, metabolic pathways were comparatively constrained across soils and moisture. Few pathways differed significantly, and no combined effect of soil type and moisture was detected on overall functional composition. In serpentine biocrusts, the moisture regime shaped the distribution of microbial functions without altering overall pathway composition. Erratic moisture was associated with enrichment of pathways linked to stress tolerance and detoxification. Frequent moisture was associated with greater representation of fermentation‐related pathways and cell‐envelope biosynthesis. Benchmarking showed that PICRUSt2 recovered broad sample‐level patterns but poorly captured pathway‐specific quantitative variation relative to HUMAnN3.