Five years ago, Susannah Salter and her colleagues swabbed the throats of 20 infants monthly until the children reached 2 years old—and documented a remarkable change. By sequencing the swabs’ DNA, Salter, a microbiologist at the Wellcome Trust Sanger Institute in Hinxton, U.K., learned that infants and toddlers have quite different microbial communities in their throats. Or so it seemed, until the team discovered that change was an artifact of contamination in two different batches of the DNA isolation kit used in the study. "The pattern wasn’t real," says Salter’s former Sanger colleague Alan Walker, now a microbiologist at the University of Aberdeen in the United Kingdom.

Worried that many microbiome researchers are falling into the same pitfall, Salter, Walker, and their colleagues have now done a systematic study of how contamination can affect DNA survey results. They find that most testing kits and other reagents for analyzing DNA contain stray microbial DNA that can overpower the signal of the native bacteria.

Their study, reported this week in BMC Biology, "is of huge significance, because it either implicitly or explicitly debunks quite a large number of microbiome papers," says Mick Watson, a bioinformaticist at the Roslin Institute of the University of Edinburgh in the United Kingdom. "I think everyone kind of knew that there might be contamination in reagents and kits, but no one really knew the extent of the problem.”

To perform their test, Salter, Walker, and their colleagues cultured a single Salmonella strain and then dissolved the bacteria in an otherwise sterile broth. Then they repeatedly diluted the broth, testing it each time using several DNA isolation kits and other laboratory reagents and methods. Early in the process, when the sample bacteria were abundant, the small proportion of stray DNA in the test kits was not enough to skew the outcome, but when the sample was diluted to just 1000 bacterial cells, the contaminant DNA swamped the real signal, the team reports.

The team thinks such contamination will not be a problem for microbiome studies of fecal samples, as they are teeming with gut bacterial DNA. But it looms large for samples typically containing relatively few bacteria, such as dust, spinal fluid, or blood, Walker says. Manufacturers of DNA isolation kits don't guarantee their materials are sterile, he notes, but many people might not realize that. "This study will have a big impact on how we design and carry out our lab protocols and sequencing studies in the future," says microbial ecologist Holly Bik of the University of Birmingham in the United Kingdom.

Veterans of microbiology have for decades warned about the need to take contamination precautions seriously and exchanged horror stories. In 1998, for example, scientists were finding the same microbes in all sorts of extreme environments; it turned out the species in common came from water pipes. This “issue is as old as the field itself and it’s being continuously rediscovered,” says Ruth Ley, a microbial ecologist at Cornell University. And it can be dealt with, says Gary Huffnagle, who studies lung microbes at the University of Michigan, Ann Arbor. "This study should not be used to dismiss analysis of low biomass samples," he says. "Running controls is always key."