Laboratory contamination over time during low-biomass sample analysis

Abstract

Bacteria are not only ubiquitous on earth but can also be incredibly diverse within clean laboratories and reagents. The presence of both living and dead bacteria in laboratory environments and reagents is especially problematic when examining samples with low endogenous content (e.g., skin swabs, tissue biopsies, ice, water, degraded forensic samples or ancient material), where contaminants can outnumber endogenous microorganisms within samples. The contribution of contaminants within high-throughput studies remains poorly understood because of the relatively low number of contaminant surveys. Here, we examined 144 negative control samples (extraction blank and no-template amplification controls) collected in both typical molecular laboratories and an ultraclean ancient DNA laboratory over 5 years to characterize long-term contaminant diversity. We additionally compared the contaminant content within a home-made silica-based extraction method, commonly used to analyse low endogenous content samples, with a widely used commercial DNA extraction kit. The contaminant taxonomic profile of the ultraclean ancient DNA laboratory was unique compared to modern molecular biology laboratories, and changed over time according to researcher, month and season. The commercial kit also contained higher microbial diversity and several human-associated taxa in comparison to the home-made silica extraction protocol. We recommend a minimum of two strategies to reduce the impacts of laboratory contaminants within low-biomass metagenomic studies: (a) extraction blank controls should be included and sequenced with every batch of extractions and (b) the contributions of laboratory contamination should be assessed and reported in each high-throughput metagenomic study.

Keywords: ancient DNA; contaminant; contamination; metagenomics; microbiome; microbiota.

Figure 1

Lower diversity is observed in EBCs and NTCs compared to biological samples. (a) The number of sequenced reads from samples that were all pooled at equimolar concentrations is displayed on a box and whisker plot. (b) The alpha diversity of each type sample (i.e., the within sample diversity) was calculated using observed species metrics in qiime1 for rarefied 16S rRNA data. Each sample was rarefied up to 1,000 sequences in 100 sequence intervals; the standard error at each subsampling event is displayed using error bars. Calculus samples are shown in blue, while control samples (extraction blank controls [EBCs] and no‐template controls [NTCs]) from the ancient laboratory (AL) and the modern laboratory (ML) are shown in red and green, respectively [Colour figure can be viewed at http://www.wileyonlinelibrary.com]

Figure 2

Microbial phyla within controls are distinct from biological samples. The proportion of different microbial phyla are shown for a wide array of modern and ancient calculus samples and control samples (EBCs and NTCs) from both laboratory facilities (modern lab [ML] and ancient lab [AL]) and two different extraction methods: the method employed in ancient DNA research and a commercially available DNA extraction kit (kit). Rare phyla were collapsed if they represented <0.001% of the total phyla observed [Colour figure can be viewed at http://www.wileyonlinelibrary.com]

Figure 3

PCoA plots of control samples highlight differences in method and laboratory. PCoA plots of unweighted UniFrac values were plotted in qiime1 to compare beta diversity differences (between samples differences) in all samples (a) or in different laboratories (b). The different laboratory facilities are represented by ML (modern lab) and AL (ancient lab), and the two control types are represented by EBC (extraction blank control) or no‐template control (NTC) [Colour figure can be viewed at http://www.wileyonlinelibrary.com]

Figure 4

PCoA of the extraction method and seasonal variation in contaminant communities. The modern and ancient calculus samples were removed from the analysis presented in Figure 3, and a PCoA plot was constructed of only control samples to identify differences between the extraction method and laboratory in control samples (a). (b) UniFrac values from control samples (EBCs and NTCs) from the ancient laboratory over a 5‐year period (2012–2016) are coloured on a PCoA plot according to month [Colour figure can be viewed at http://www.wileyonlinelibrary.com]