Almost every day new data about the composition of microbiomes are published. Many of these studies analyse the human microbiome, but also environmental samples.
Today we have the ability to sequence microbiomes in much more depth than a couple of years ago. Looking deeper sheds light on an important point: Contamination! In the very interesting publication of Salter et al. they could show that contaminating DNA is present in DNA extraction kits and other lab reagents.
The researchers sent dilutions of pure cultures of Salmonella bongori to three different institutes for DNA extraction and PCR, followed by sequencing on Illumina MiSeq. While S. bongori was the only organism identified in the undiluted samples, contaminating bacteria increased in relative abundance with higher degrees of dilution, and finally became dominant after the fifth dilution.
They did a similar analysis performing shotgun metagenomics of a pure S. bongori culture. This time, they used four different DNA extraction kits. Again, they saw that contamination increased with the degree of dilution, with contamination being the predominant feature after the fourth dilution. Also, they could show that each kit gave a different bacterial profile.
They also report on a study on the nasopharyngeal microbiota of children, analyzed over 2 years. They could show that using 4 different DNA extraction kits over time led to the false conclusion that differences in the microbial spectrum were associated with age. When DNA extraction was repeated on original samples using a different kit lot, the OTUs previously identified as contaminants were no longer detected.
In conclusion, contamination affected both 16S and metagenomic shotgun sequencing projects and was especially critical for samples with low biomass. Salter et al. present a list of potential contaminating organisms, as well as recommendations on how to cope with this problem. One recommendation is very obvious, and very effective: use negative controls!
Altogether, we should be very careful in planning our experiments in order to deliver results instead of artefacts. Especially, we need to be very careful when interpreting the data!