Microbiomes and human healthEnvironmental microbiomes, diverse and ecologically complex microbial communities associated with inhabitable spaces, interact with and impact human-associated microbiota and human health outcomes1,2. Within these interactions, microbiomes have been found to influence many complex relationships related to long-term human health outcomes, including organ development, modulating organ function, and immune response1,2. Advancements in high-throughput metagenomic sequencing3 have improved our ability to explore the impacts of urban environmental microbiomes, which have characteristically low diversity and richness of many taxonomic groups4, with negative impacts to human health5: Mechanically ventilated indoor spaces, where urban residents spend over 90% of their time6, present low microbiome diversity and abundance4 as compared to more informal7 and non-urban settlements4. Indoor environmental microbiome exposures have subsequently been correlated with detrimental human health outcomes, where low abundance and diversity have been linked to increased instances of atopic skin conditions and allergy8,9, asthma8,10,11, obesity11,12,13, cancer14, depression15,16 as well as the transmission of potential pathogens17,18,19.Although environmental microbiome diversity is only one variable by which microbial communities might be compared, existing research indicates it is a useful metric that appears to predict human health outcomes. For example, emerging research indicates relationships between environmental microbiome diversity and exposure with human health outcomes are reversible: while common building maintenance protocols such as HVAC use and surface cleaning contribute to reducing microbial diversity while increasing antimicrobial resistance in indoor microbiomes20, environmental biodiversity interventions (such as vegetation21) can diversify indoor microbiomes with measurable benefits to human health22. Further study is required within this rapidly developing field to determine the mechanisms by which environmental microbiome diversity (as well as other metrics), human exposure, and ultimately interactions with human microbiomes might influence long-term health outcomes.Metagenomics and sampling protocolsIndoor spaces in which urban residents typically spend 90% of their time6 are a critical environment within which to understand microbiome exposure. These spaces are also challenging to study as they have characteristically low biological material23, making it difficult to meet the DNA yield requirements for reproducible metagenomic analyses23,24. This has resulted in challenges of replicability for studies characterizing the microbiomes of urban indoor environments as different sampling and analysis protocols influence the quantification of the microbiome, and methodological differences often overshadow biologically relevant variation25. In addition, poor analytical outcomes due to low DNA yields26 can translate to an unduly high signal for any present contaminants24.As the field continues to develop, strategies for avoiding, identifying, and removing contamination from analyses have emerged through a variety of methodological approaches27. Negative controls throughout both sampling and sequencing processes are necessary to identify contamination through DNA quantification, while sequencing analysis of negative control DNA is required to control for potential contamination within experimental samples analytically27. Computational approaches have also been proposed to re-analyze existing datasets and correct for analytic artifacts28, as have alternative extraction methods for future work (such as a liquid–liquid approach29). Although all such developments are important protocols to utilize in future work, they do not address the challenge of insufficient DNA input to metagenomic processes, which may require an alternative sample collection method to those commonly employed in indoor studies.Sample collection methods that have been employed thus far by indoor microbiome studies span a large range of protocols. Airborne bioaerosols have commonly been sampled through the use of active air flow sampling on filters30, however this sampling approach is typically only representative of short time periods. Another common strategy involves collecting material that has settled on surfaces31 which tends to be more representative of chronic exposures while capturing aspects of both airborne and dermal exposure, the latter of which is another major avenue of microbial transmission to humans32. Studies that sample microbiomes that have settled on surfaces often utilize swabs to collect existing biological material from horizontal surfaces33, however this method presents challenges as the deposition timeline of material collected by swabbing existing surfaces are often unclear, and DNA extraction efficiencies from commercially available swabs are commonly limited to 15–35%33.Harmonization of both sampling and analysis stages may help address challenges of reproducibility in sequencing-based experiments. Not only do swabs have reportedly low extraction efficiencies33, but sampling processes utilizing swabs as a sampling tool are likely inefficient: although swabs are often used to collect and transfer deposited biomass from surfaces of interest for downstream analysis, this process likely leaves biological material behind. Additionally, variability in individual techniques may result in differences in swabbing surface area and time, which could alter outcomes and further impact inter-study comparability. Although alternative collection materials to swabs have been explored, such as wipes34,35, such a substitution requires a specialized extraction protocol35, which does not allow for direct comparisons across studies24, and does not fundamentally address sample timeline challenges. A variety of deployable surfaces, such as petri dishes31 and common building materials36, have been utilized within environmental microbiome studies to provide a controlled deposition timeline, however these methods continue to rely on swabbing the deployed material which fails to address the challenge of low swab extraction efficiencies.A novel sampling approachPassive deployable material that allows biological material to accumulate and be transferred in its entirety into an extraction pipeline may represent an alternative approach for assessing the indoor microbiome. Here we present a promising alternative to swab-based microbiome sampling: passive, deployable polydimethylsiloxane (PDMS) films that can be inserted in their entirety (along with any material deposited on its surface) into commercially available DNA extraction kits. PDMS is a biocompatible, chemically inert material37 that has been used increasingly for work in direct contact with DNA such as microfluidic devices 38 for a variety of purposes, including solid phase extraction with direct amplification39. PDMS is an excellent material candidate for the proposed sampling technology because it is not only durable and flexible even when very thin40 (allowing for insertion into extraction tubes), but it is inexpensive, chemically inert37 and has been used in direct contact with nucleic acids without interference39,41.Passive, deployable PDMS samplers where the whole surface is inserted into commercially available extraction kits may address two challenges discussed thus far by (1) providing a clearly defined sampling timeline of surface deposition samples, and (2) increasing DNA yields as all captured material is utilized as input to processing and analytic protocols. Although promising, such an approach must be optimized in the field to allow researchers to collect DNA in quantities targeted for sequencing when collecting samples over extended periods: > 500 ng if standard sequencing approaches are utilized23, as little as 1 ng if “ultra low” sequencing approaches are utilized23,42. PDMS has already been utilized in a similar manner to passively collect environmental chemicals43 and respiratory viruses, including SARS-CoV-241, however DNA collection and extraction efficiencies for downstream sequencing have not yet been explored.Results & discussionWe present a preliminary benchmark study for an alternative to swab-based microbiome sampling that transforms DNA yield to facilitate sample replicability: deployable polydimethylsiloxane (PDMS) films, easily customized by the user (see Methods). The presented experiments compare DNA extraction yield, efficiency, yield per cm2, and sequencing outcomes using a microbial community standard (Zymo Cat# D6300) between PDMS films in a range of sizes with a representative swabbed petri dish method as a preliminary investigation into the potential benefits of the described novel sampling technology. A deposition model was used to replicate microbiome biological material settling onto a surface with a standardized mixture: 1 µl of a diluted microbiome community master mix was pipetted onto each cm2 of each sampling surface (Fig. 1A) and allowed to dry while covered. Three commercially available extraction kits were compared, referred to as “Miniprep” (MP), “PowerSoil” (PS), and “PowerWater” (PW). PDMS films were fabricated in the lab to optimally fill each extraction kit based on the height of the bead-beating tubes (Fig. 1B). A variety of lengths of the PDMS samplers were included to capture a range of sampler surface areas under each condition, where larger surface area resulted in both greater volume of the deposited standardized microbiome community, as well as a greater surface area of PDMS packed within the bead-beating tube (see Figures S 1–4). Swab samples were collected from commercially available 15 cm diameter petri dishes as a representation of a common swab-based sampling approach with which to compare the range of PDMS sampler protocols. Each Sampling/Extraction method was replicated three times.Fig. 1(A) Illustrating the relative sizes of the eight samplers. White dots represent 1 µL of the diluted microbial community master mix. Samplers are drawn according to scale. Rectangular diagrams represent PDMS samplers in a variety of sizes inserted in their entirety into extraction kit collection tubes. The circular diagram represents a petri dish where collection required swabbing and inserting the swab head into extraction kit collection tubes. (B) Comparing PDMS with swab inputs and relationships to extraction bead-beating tube size constraints. (C) Comparing the field replicability of three samplers from section “A” (boxed) illustrating how the size and shape of each sampler impacts the number of possible sample replicates within a hypothetical deployment area of 400cm2.Sample replicabilityAs sampler size increased, sample replicability, as indicated by the number of samplers that can be deployed per area, decreased as a function of sampler surface area (Fig. 1C). Although field-deployable replicability was reduced for larger samplers, larger sampler sizes also resulted in greater DNA yields per sample (Fig. 2A). This relationship between sampler size, replicability per unit field area, and DNA yield per sampler indicates a clear cost–benefit analysis that must be considered where the available space within an identified field-deployment location, required sample number, and expected DNA yield must be weighed to optimize sequencing and analytic outcomes.Fig. 2(A) Line range plots of total DNA yield. (B) Line range plots of calculated extraction efficiencies in reference to positive controls (see Methods). Median values of the largest PDMS samplers included within each extraction kit are circled. The lowest quantified extraction efficiencies within each group are denoted. (C) Line range plots of DNA yield per cm2 of sampler area. Average DNA yield per cm2 is denoted for discussed samplers.Extraction results: DNA yieldThe largest average DNA yield corresponded with the 117 cm2 PDMS/PowerWater protocol (31.6 ng, Fig. 2), which was 43% higher than the Swab/PowerWater mean yield (20.4 ng, Fig. 2) despite being 60 cm2 smaller and receiving 60 μL less biological material than the swabbed petri dish (177 cm2). The largest yield overall was collected from one of the180 cm2 PDMS/PowerWater sampler/extraction protocols, however this group resulted in a lower average yield (28.8 ng) and was accompanied by the largest range in yield (+ /− 13.0 ng). The smallest films sized for the Miniprep and PowerSoil kits all resulted in DNA yields 6.25 ng or below (Fig. 2), which are only suitable for “ultra-low” sequencing approaches that require as little as 1 ng of DNA23,42, however the largest average yield for this group was measured in the 12cm2/PowerSoil protocol (5.2 ng). The presented yield results in combination with the replicability results (above) indicate that the 117 cm2 PDMS/PowerWater group represents a promising sampling and extraction protocol that maximizes both DNA yields suitable for sequencing and sample replicability in the field.Extraction results: extraction efficiencyExtraction efficiencies were calculated and analyzed in comparison to positive control yields within each extraction kit (see Methods) in an attempt to control for differences between extraction kits and explore the reported discrepancies in DNA yields between samplers despite differences in biological input. All PDMS films returned higher average extraction efficiencies than the swab samples extracted with the same kit (p < 0.05) with the exception of the 180 cm2 PDMS/PowerWater method which was not statistically significantly higher (p = 0.47) (Fig. 2B). This result indicates higher extraction efficiencies of DNA from PDMS samplers (irrespective of extraction kit) likely contributed to the 43% greater DNA yield captured by the 117 cm2 PDMS sampler than the corresponding swab samples, despite a smaller surface area and volume of biological material.The extraction efficiency results within the PDMS samplers also provides preliminary evidence of an interaction between PDMS sampler size and extraction efficiency outcomes. Four of the smallest PDMS samplers (8–12 cm2) extracted with the Miniprep and PowerSoil kits, returned extraction efficiencies over 100% in comparison to the positive control yields (see Table S1). This indicates PDMS sampler sizes that are small relative to the size of extraction kit bead-beating tubes may increase extraction efficiency, however due to the proximity of these samples to the limit of detection of the DNA quantification method (and thus a greater possibility of quantification error, see Methods) this result requires further exploration. In considering the impact of PDMS sampler sizes that are large relative to the extraction kit bead-beating tubes, the largest PDMS films resulted in the lowest calculated extraction efficiencies within each extraction kit (see Fig. 2B, circled). In addition, it was observed that these largest sized PDMS films filled the extraction kit bead-beating tubes nearly completely (see Figure S3 and 4). These outcomes indicate that extraction efficiency likely declines as a function of sampler size: small sampler sizes may somehow enhance extraction processes, perhaps by physically enhancing the bead-beating process, while larger volumes of sampler material may somehow inhibit the same processes or impede collection of the supernatant following the bead-beating step. Further testing is required to confirm the replicability of such patterns, as well as identify potential mechanisms.Extraction results: DNA yield per cm2
DNA yield per cm2 was analyzed to more directly compare yield outcomes by controlling for input of the mock community biological material, which must be accounted for when considering both total DNA yield results and extraction efficiency calculations. The results support the finding reported in the extraction efficiency section whereby sampler size appears to impact extraction outcomes resulting in greater yields per cm2 for smaller sampler sizes, and lower yields per cm2 for larger sizes. Within the PDMS samplers, the greatest average DNA yield per cm2 was calculated for the smallest sizes (Miniprep: 0.24 ng/cm2, PowerSoil: 0.45 ng/cm2, PowerWater: 0.35 ng/cm2), while the lowest average DNA yield per cm2 was calculated for the largest sampler sizes (Miniprep: 0.15 ng/cm2, PowerSoil: 0.23 ng/cm2, PowerWater: 0.16 ng/cm2) and intermediate PDMS sizes resulted in intermediate yields (Fig. 2C, Table S1).The 117 cm2 PDMS sampler mean DNA yield per cm2 (0.27 ng) was 76% greater (p < 0.05) than the mean DNA yield per cm2 PowerWater swab samples (0.12 ng). Although this was not the highest calculated yield per cm2, this outcome supports previous findings that the 117 cm2 PDMS sampler/PowerWater extraction protocol, with greater DNA yield per cm2 than a traditional swab-based method, represents a promising protocol that could be utilized to maximize DNA yield per cm2 of deployed sampler area for sequencing-based experiments.Sequencing results: negative controls & non-reference speciesNegative controls, including sterile samplers and kit-blanks, were included for each sampler size and extraction kit (see Methods). All negative controls, including sterile samplers, resulted in DNA concentrations that were below the detection threshold (see Methods). Only one negative kit-blank sample, extracted with the Miniprep kit, was successfully sequenced (due to low DNA concentrations), therefore comparisons between extraction kits were not possible. In addition, this sample resulted in significantly more non-reference species than any experimental sample, likely due to analytic artifacts stemming from insufficient DNA26. Within the Miniprep negative sample sequencing results, 54 species were identified. 18 of these species were also identified in experimental samples, 36 were unique to the negative sample (see Table S2).Within the 13 sampler/extraction kit protocols, all non-reference contaminant species were at an acceptably low level where even low-abundance reference species would not be excluded analytically. Within the 39 experimental samples the aggregated relative abundances for all non-reference species fell below 1% for 37 samples, with the two highest aggregated abundances falling at 3.1% and 1.2% for single samples (see Fig. 3). Analyzed on a species-specific basis, these are acceptably low contaminant abundances: Published literature indicates any species represented by less than 20% of the most abundant contaminant (in this case, non-reference) species should not be considered a true finding44. The lowest expected abundance of all the reference species was C. neoformans (0.37%, see Fig. 4), meaning the threshold relative abundance to disqualify this expected value was 1.85%. The most abundant contaminant (non-reference) species identified in this study was C. acnes. The protocol with the greatest C. acnes abundance was the 8 cm2 PDMS/Miniprep protocol with a maximum of 3%, but an average of 1.05% (see Fig. 3, Table S3), the highest of which falls above the exclusion threshold, but the average of which does not. This pattern was confirmed within the data where C. neoformans relative abundances were often lower than the expected 0.37% (see Table SIII). Given these results, all reference species would be considered true findings for all protocols as long as exclusions were determined based on average outcomes for all identical sampler replicates.Fig. 3Relative abundance (%) of the non-reference species identified by sampler/extraction kit protocols. A full list and average relative abundance by sampler can be found in Table SIV.Fig. 4A metagenomic analysis of the community structure for three representative samplers (Figure 1A&C). Outcomes were compared to the provided mock community reference numbers “R” (Bray–Curtis Dissimilarity), and species-level outcomes were compared within PDMS/Swab sampler groups (Kruskal–Wallis).Within the experimental samples, 27 non-reference species were identified, 9 of which were not identified within the Miniprep negative sequencing data (see Fig. 3, Table S4). In comparing the extraction kit outcomes, samplers extracted with the Miniprep kit resulted in the greatest number (p < 0.05) of non-reference species identified (21), followed by the PowerSoil samples (7), and the PowerWater samples (6). The Miniprep kit also resulted in the greatest average relative abundance (p < 0.05) of non-reference species (0.09%) in comparison to the PowerSoil (0.08%) and PowerWater (0.06%) kits, although all non-reference identification outcomes were within an acceptable range 44.Sequencing results: reference species comparison & accuracyOverall, the sequencing results did not differ significantly between the sampling strategies, however differences were identified between extraction kits. The PowerSoil extraction kit was the most accurate (5.1–7.8% dissimilarity from the reference), while the PowerWater kit was the least accurate (22.1–22.2%) (Fig. 4). In addition to community-level analyses, many of the species-level relative abundances differed between experimental groups. These results align with previous work that report differences in species abundance between extraction kits28.
A deployable film method to enable replicable sampling of low-abundance environmental microbiomes
