PMA treatment successfully depletes signal from relic DNA in simple synthetic communities
Our evaluation of PMA’s ability to deplete signal from relic DNA started with the validation in simple synthetic communities, where we built cultures with known mixtures of living and heat-killed Escherichia coli and Streptococcus sanguinis (Fig. 1, Methods). The two species were chosen as representatives of Gram-negative and Gram-positive strains, respectively, known to be a major determinant of PMA susceptibility [13, 17, 18]. In communities composed of mostly living cells (Fig. 1a-b, Groups 1, 3 and 5), PMA treatment resulted in lowered DNA yields, suggesting the existence of non-viable or damaged cells in these cultures or the non-specific effects of PMA treatment to viable cells (though in these communities 16S gene copies did not vary significantly; paired t test, p = 0.36, 0.27, and 0.46 for Group 1, 3, and 5, respectively). DNA yields and 16S gene copy numbers both decreased after PMA treatment in the heat-killed synthetic communities (Groups 2, 4, and 6), as expected, though the killing process itself likely induced some DNA degradation.
Sequencing results (Fig. 1c) suggested that signals from relic DNA were removed in cultures with only one viable microorganism (Groups 9 and 10). In monocultures containing only one non-viable microorganism (Groups 2 and 4), which would ideally result in "no" sequenced DNA, the small number of sequences still obtained were (as expected) from either the non-viable bacteria added, or from the one that was not added, likely due to low-level contamination or bleedthrough from other samples (as would occur in any near-empty amplicon library, regardless of viability). Most importantly in such a setting, the overall depletion of relic DNA in these groups (2 and 4) were supported by much lower DNA yields and qPCR signals. PMA treatment of mixed communities comprising two microorganisms (Groups 7 and 8) was qualitatively successful but different from the target mix proportions. Particularly, we expected a ~30% reduction in DNA yield and 16S gene copies after PMA treatment, while we did observe less DNA after PMA treatment (Fig. 1b, top), the resulting 16S copies (Fig. 1b, bottom) and ratio of the two species (Fig. 1c) was not consistent. It is worth noting that this inconsistency may not lay in PMA-treatment alone, as any amplicon-based taxonomic profiling tends to be affected by factors such as PCR efficiency during target amplification and library preparation, 16S rRNA gene copies in different taxa, and the overall composition of microbial community . Taken together, though, these results indicated that PMA-seq was able to successfully reconstruct the communities of viable/heat-killed E. coli and S. sanguinis in the ten synthetic communities, regardless of Gram stain, with some degree of quantitative accuracy (Fig. 1).
PMA-seq does not accurately quantify microbial viability in spiked complex communities
To further test the accuracy of PMA-seq, we evaluated its performance by spiking control microbes into complex microbiomes from a variety of environments (Fig. 2). Two were high-biomass, high-complexity communities (i.e. soils and human saliva), and two were representatives of low-biomass communities sourced from expected low-viability environments (i.e. computer screens and computer mice) (Methods, Additional file 1: Figure S2). Additionally, we spiked each of these samples with different concentrations of a 1 ml mixture of live / dead E. coli cells at the ratio of 1:1 (Methods) to assess whether PMA treatment was able to effectively remove known relic DNA within a complex community.
Within these community types, PMA-seq was unable to denote microbial viability quantitatively, with its performance largely dictated by each environmental source’s characteristics, which added to existing distortions from amplicon-based sequencing (Fig. 2). The low biomass samples indicated larger compositional dissimilarities between PMA-free vs. PMA-treated samples. Conversely, only minor changes were observed in the spiked and resident microbes of the high biomass samples (Fig. 2b). Sample type, as expected, accounted for most of the observed differences in these communities, explaining 46.5% of the variation (PERMANOVA FDR adjusted p = 0.0015) (Additional file 4). The effects of PMA-treatment differed significantly by sample types (PERMANOVA FDR adjusted p = 0.039), arguably more effective in low biomass samples, in which somewhat fewer competing microbial, chemical, and matrix-driven effects were present to prevent intercalation and nucleotide depletion.
To explore the potential of PMA-seq as a semi-quantitative tool, we compared the abundance of known spike-in cultures with and without PMA treatment. The addition of E. coli controls to spiked-in (“Spike+”) groups boosted the abundance of Enterobacteriaceae, as expected. Indoor samples (computer screens and computer mice) were further enriched for Clostridium species in the PMA-free groups, possibly resulting from contamination in these samples (Fig. 2a). Further increases of Enterobacteriaceae were observed upon PMA treatment in the Spike+ groups, likely resulting from the elimination of a subset of dead microorganisms from the community, as expected.
It should be noted that the abundance change of a taxon in PMA-seq only represents a “relative viability,” where an increased abundance could be due to the taxon being “more viable” or from the changes of other microbes. We thus calculated normalized PMA efficacies independently in the four sample types based on the relative abundance of Enterobacteriaceae, combined with the 16S rRNA gene copies obtained from qPCR in the four aliquots of samples (Methods, Additional file 3). Briefly, to determine the amount of correct PMA tagging (efficacy) in each sample, we divided the amount of dead Enterobacteriaceae that was successfully removed after PMA treatment by the total amount of spiked Enterobacteriaceae. The efficacy should equal to 0.5 under ideal conditions, given that each spike-in culture was a mixture of live / dead E. coli cells at the ratio of 1:1; an efficacy over 0.5 indicated the unintentional removal of viable microbes, while below 0.5 suggests incomplete depletion of non-viable cells. The high efficacy in computer screens (1.01), computer mice (0.96) and soil (0.87) samples suggests a partial toxicity of PMA to viable cells, which somewhat explained the drastic compositional changes in the communities from screens and mice. The efficacy in saliva samples is relatively low (0.35), indicating the incomplete elimination of relic DNA. These results reiterated that PMA-seq did not accurately quantify microbial viability in complex communities, with efficacy varying in different sample types.
Complex microbial communities from the BE do not respond consistently to PMA
Based on this assessment, we applied PMA-seq to real-world microbial community samples from the Boston subway system as a representative (transit) built environment (Fig. 3). The relative abundances of several taxa changed significantly after PMA treatment (Additional file 6), but these PMA-reactive taxa varied among different sample types. As one example, the family Porphyromonadaceae was enriched consistently after PMA-treatment in samples from seats and walls, but remained stable in those from grips and touchscreens (Additional file 1: Figure S4).
Large compositional changes were observed in samples from seats and walls after PMA treatment, while grip and touchscreen communities exhibit less of a composition shift. This is arguably concordant with what we observed from the spiked samples and from previous studies , in which PMA treatment performed more effectively in less-complex samples with lower biomass (i.e. here, seats and walls). Likewise, Bray-Curtis dissimilarities differed more in walls and seats between PMA-free and PMA-treated samples (Fig. 3b). Independent of PMA treatment, sample type was again the main force driving taxonomic differences, explaining 37.8% of the variation (PERMANOVA FDR adjusted p = 0.002). Overall, and critically for future use of PMA-seq, our results indicate that microbial communities do not respond consistently to PMA treatment, with unique “PMA-reactive” taxa enriched after treatment in a context-dependent manner.
“PMA-resilient” taxa are shared among studies while “PMA-responsive” taxa vary
A variety of factors are known to account for PMA responsiveness in complex communities . To characterize common patterns of community structural changes upon PMA treatment, we compared our results with previous PMA-seq studies spanning three diverse environments: dust from a clean room environment, soil habitats, and stool [2, 5, 10]. We re-profiled each study’s raw data to ensure consistency (Methods) and identified the taxa with the greatest changes in relative abundance before and after PMA treatment in each sample type. Changes were calculated by dividing the relative abundance difference between PMA-free and PMA-treated samples by the initial relative abundance in PMA-free samples, allowing us to identify the taxa with the largest fold changes in response to PMA (i.e. PMA-responsive taxa).
Due to underlying differences in initial viability and community background, change in the relative abundance of a microorganism was observed to be variable upon PMA treatment between sample types. Thus, PMA-responsive microbes were mostly unique in each independent source (Additional file 1: Figure S6). PMA treatment was able to remove several taxa (to undetectable) while also revealing others that were undetectable in the initial communities. On low-touch surfaces (cleanroom floors, computer screens, subway seats and walls), the largest fold changes of relative abundances were mainly caused by the reduction of typically environmental taxa after PMA application, indicating that many non-viable microbes pertaining to those surfaces were successfully removed by PMA treatment. By contrast, the most responsive microorganisms on high-touch surfaces (computer mouse and subway grips), and in human stool and saliva samples, changed in the opposite direction after PMA treatment. These taxa, underrepresented or undetected in the PMA-free communities, were sometimes orders of magnitude more abundant after PMA treatment, which would indicate that they may be viable but rare organisms in these environments.
To better compare PMA treatment across different sample types, we ranked the most PMA-responsive taxa in each sample type and selected 30 taxa with the largest average relative abundance fold changes in at least two sample types (Fig. 4a). As expected, abundance changes of these taxa varied greatly in different samples. The abundance of the human commensal genus Peptoniphilus decreased from 10-2 to undetectable after PMA treatment on computer screens and subway walls, while remaining stable on clean room floors, computer mice, and subway grips. While in saliva and soil samples, it increased repeatedly from undetectable to a low fraction of ~10-5 after PMA treatment. Several other human commensals changed in similar directions across different sample types, such as the genera Finegoldia and Gardnerella, the Veillonellaceae family, and typically environmental microbes from the genus Brevibacterium. Biologically, these results suggest that functionally high-touch BE surfaces, intuitively, retain more viable human commensals compared to low-touch ones.
Using similar approaches, we also identified “PMA-resilient” taxa with the smallest average relative abundance fold changes within each sample type (Fig. 4b). In contrast to PMA-reactivity, PMA-resilience was more likely to be determined taxonomically, with less dependence on community background, biomass, or microbes’ abundance. Taxa of widely ranging abundances could be PMA-resilient (Fig. 4b, Additional file 1: Figure S7). Several highly abundant, human-derived genera, including Corynebacterium, Staphylococcus, and Lactobacillus, showed repeated stability upon PMA treatment in stool, saliva, soil, and on different BE surfaces. This would suggest that these abundant human commensals exist in viable forms in the BE; alternatively, they may simply be more resistant to PMA treatment. In summary, the above results suggest that PMA-based viability quantitation within communities may be both context-specific and taxon-specific, with “PMA-resilient” taxa often shared among similar ecosystem types while “PMA-responsive” taxa vary.