How to Design a Skin Microbiome Study, Part I: Sampling

In this post, the first of a 2-part series on skin microbiome research, we will discuss technical issues surrounding sampling of human skin microbes.

Let’s say a researcher sets out to study bacteria on the human skin—the body’s largest organ, which is teeming with microbes from each domain of life, and viruses. The scientific question has been identified and the funding to conduct a pilot study has been secured. Perhaps her group has some experience studying the gut microbiome. For the most part, she has discovered, getting bacteria out of stool is not too difficult; a small amount of material contains enough microbial DNA to sequence the most prominent members of the microbiome. But unlike the intestine, the skin does not support a high-biomass microbiome. If microbes on the skin are present in low abundance, how does our researcher decide on reasonable sampling and sequencing strategies that together capture a representative picture of bacterial diversity?

Before collecting samples, the researcher must consider key advantages and limitations of available sampling protocols. Commonly used methods involve variations of swabbing – the repeated rubbing of a defined area of the skin with a sterile, pre-moistened swab. When the objective is to obtain sufficient microbial DNA from skin sites with variable and/or low microbial biomass, swabbing can be complemented with scalpel scrapping (Oh et al., 2014). If access to deeper layers, including the dermis, is required, punch biopsies are a viable alternative, but require specialized expertise and are more invasive, reducing the number of sites that can be sampled from the same subject.

Because each method samples a slightly different environment, we would expect different microbial profiles to arise from variations in sampling method. This prediction has not been thoroughly tested (but see Chng et al., 2016). As part of ongoing efforts to improve sampling methods, the Microbiome Insights team is currently evaluating whether D-Squame and Sebutape tape stripping, used for peeling off epidermal layers and sebum, respectively, provide a reliable means of sampling skin microbes. Compared with swabbing, tape stripping recovers ~2- to 3-fold less bacterial DNA. We have yet to evaluate, via amplicon sequencing, whether lower bacterial yield results in different microbial profiles. We are also exploring if coating pre-wetted swabs with aluminum oxide particles maximizes bacterial DNA recovery. The results of this experiment will be made available in a forthcoming technical note.

Handling samples with low microbial biomass is challenging. Even if the sampling method affects how much microbial biomass is collected, the amount of DNA recovered from skin is always low. (Of course, the DNA extraction method affects DNA yield and microbial composition from study to study. But because most studies are comparative in nature, methodological consistency is vitally important.) As most of the DNA is human, obtaining enough genetic material for microbial profiling can be difficult. Microbial load can be increased by instructing participants not to wash with soap or bathe at least 24 hrs prior to sampling, although the effect of cleansing is probably minor compared with the combined influence of sampling and extraction.

Another challenge of low microbial biomass samples is dealing with environmental contamination. Contamination can be introduced during sample collection, DNA extraction, and sequencing library preparation. For instance, bacterial DNA is often found in DNA extraction kits and in other reagents used for preparing samples. And while it is tempting to create lists of “usual suspect” contaminants, this may be futile when studying skin microbes or other human-associated bacteria because, for example, Staphylococcus, a common skin inhabitant, and Escherichia have been identified as potential kit reagent contaminants (Salter et al., 2014).

Processing negative controls alongside low-biomass specimens is critical, because the proportion of microbial DNA attributable to contamination is higher in low-biomass samples compared with high-biomass samples. Usually, we include at least four replicates for each of two types of negative controls on each 16S sequencing run: (1) DNA extraction controls, to assess if kit reagents carry a detectable signal, and (2) template-free PCR blanks, to pinpoint contamination that may arise during downstream processing. For skin microbiome analysis, sterile swabs opened at the site of sample collection are co-processed with the swabbed samples. In general, the number of sequencing reads in our negative controls is about 3- to 4-fold lower than the average in samples derived from skin sites. This is what we would expect for samples containing little to no DNA.

Stay tuned for the second post in this series: amplicon sequencing in skin microbiome studies.