Insufficient sampling constrains our characterization of plant microbiomes

Plants host diverse microbial communities, but there is little consensus on how we sample these communities, and this has unknown consequences. Using root and leaf tissue from showy milkweed (Asclepias speciosa), we compared two common sampling strategies: (1) homogenizing after subsampling (30 mg), and (2) homogenizing bulk tissue before subsampling (30 mg). We targeted bacteria, arbuscular mycorrhizal (AM) fungi and non-AM fungi in roots, and foliar fungal endophytes (FFE) in leaves. We further extracted DNA from all of the leaf tissue collected to determine the extent of undersampling of FFE, and sampled FFE twice across the season using strategy one to assess temporal dynamics. All microbial groups except AM fungi differed in composition between the two sampling strategies. Community overlap increased when rare taxa were removed, but FFE and bacterial communities still differed between strategies, with largely non-overlapping communities within individual plants. Increasing the extraction mass 10 × increased FFE richness ~ 10 ×, confirming the severe undersampling indicated in the sampling comparisons. Still, seasonal patterns in FFEs were apparent, suggesting that strong drivers are identified despite severe undersampling. Our findings highlight that current sampling practices poorly characterize many microbial groups, and increased sampling intensity is necessary for increase reproducibility and to identify subtler patterns in microbial distributions.