Standard filtration practices significantly distort planktonic microbial diversity estimates

Filtration is the standard method for isolating planktonic microbial biomass for analysis. It is unclear how the taxonomic composition of biomass on a filter changes as a function of filtered water volume, potentially due to filter clogging. Using seawater from a marine oxygen minimum zone, we conducted experiments to quantify the 16S rRNA gene composition of biomass on a prefilter (GF/A, 1.6 um pore size) and a downstream collection filter (Sterivex, 0.2 um) over a range of typical collection volumes, from 50 to 5000 ml. Significant community shifts occurred in both filter fractions, and were most dramatic in the prefilter community. Sequences affiliated with Vibrionales decreased from ~40-60% of the prefilter datasets at low volumes (50-500 ml) to less than 5% at higher volumes, while groups such at the Chromatiales and Thiohalorhabdales followed opposite trends, increasing from minor representation to become the dominant taxa at higher volumes. Taxa shown previously to be associated with marine particles, including diverse members of the Deltaproteobacteria, Planctomycetes and Bacteroidetes, were among those showing the greatest increase with filter volume (4 to 27-fold). Metrics of taxon richness (97% sequence clusters) also varied significantly with volume, and in opposing directions depending on filter fraction, highlighting potential biases in community complexity estimates. These data raise serious concerns for studies using filter fractionation to separate biomass for quantitative comparisons of aquatic microbial diversity, for example between free-living and particle-associated communities.