Systematic Bias Introduced by Genomic DNA Template Dilution in 16S rRNA Gene-Targeted Microbiota Profiling in Human Stool Homogenates.

Variability in representation of microbial communities can be caused by differences in microbial composition or artifacts introduced at sample collection or processing. Alterations in community representation introduced by variations in starting DNA concentrations have not been systematically investigated in stool samples. The goal of this study was to evaluate the effect of the genomic DNA (gDNA) concentration in the resulting 16S rRNA gene library composition and compare its effect to other sample processing variables in homogenized human fecal material. Compared to a gDNA input of 1 ng/μl, inputs of ≤1.6 × 10 ng/μl resulted in a marked decrease in the concentration of the 16S rRNA gene amplicon ( < 0.001). Low gDNA concentrations (≤1.6 × 10 ng/μl) were also associated with a decrease ( < 0.001) in the number of operational taxonomic units and significant divergence in β-diversity profiles (unweighted UniFrac distance, < 0.001), as characterized by an overestimation of and underestimation of . Even a gDNA concentration of 4 × 10 ng/μl showed a significant impact on the β-diversity profile (unweighted UniFrac distance, = 0.03). Overall, the gDNA concentration explained 22.4% to 38.1% of the microbiota variation based on various β-diversity measures ( < 0.001). By comparison, the DNA extraction methods and PCR volumes tested did not significantly affect the microbial composition profile, and the PCR cycling method explained less than 3.7% of the microbiota variation (weighted UniFrac distance, = 0.03). The 16S rRNA gene yield and the microbial community representation of human homogenized stool samples are significantly altered by gDNA template concentrations of ≤1.6 × 10 ng/μl. In addition, data from studies with a gDNA input of ≤4 × 10 ng/μl should be interpreted with caution. The genomic DNA input for stool samples utilized for microbiome composition has not been determined. In this study, we determined the reliable threshold level under which conclusions drawn from the data may be compromised. We also determined the type of microbial bias introduced by less-than-ideal genomic input.