3.1 Sequencing results
10,000–25,000 sequences of 250 bp were obtained, samples with under 13,000 sequences were discarded and others are rarefied at 13,000 sequences, resulting in 24 samples. 3296 OTUs were identified after denoising.
3.2 System performance
Effects of O3 on performance data during the period was reported in detail in a previous publication (Saingam et al. 2017), both biofilters had approximately 65% – 70% removal rate treating inlet toluene ranging from 500 to 1500 mg/m3, with no significant difference under t-test (p = 0.62).
3.3 Biodiversity
Biodiversity index comparing the two biofilters are shown in Fig. 1. Intra-group diversity spanned over 400 in both groups of observed diversity, but median and average diversity of ozone biofilter is consistently higher than those of the control samples, indicating an increase in phylogenetic diversity range and average evenness.
3.4 Community differences and propensity of variables
Clear separation of microbiome composition is seen between the two biofilters, whereas intra-group differences also fluctuated considerably with time, relative abundance of major phyla Actinobacteria, Proteobacteria, Firmicutes, Bacteroidetes and TM7 were added as constraining variable in Fig. 2b to show overall compositional difference of OTUs level and abundance differences of major phyla. In addition to prediction of microbiome functions, microbiome phenotypic metabolizing activity from Biolog ECO plates were adopted as constraining variable to show inclination of communities for different carbon sources in Fig. 2b. Proteobacteria remained dominant in relative abundance, ozone stimulated its presence despite it consisting mostly of gram-negative species, along with Firmicutes, consisting mostly of gram-positive species. Actinobacteria is another known gram-positive bacteria seen in biofilters, decreased with ozonation. Metabolization ability of groups of amino acids and carboxylic acids increased in microbiome under ozonation, each containing a consortium of numerous compounds from the groups, while metabolization rates for complex carbohydrates decreased, metabolization data of individual compounds are provided in supplementary material 1.
3.5 Functional characteristics and traits caused by ozonation
Phenotypic traits of microbiome from both biofilters are shown in Fig. 3. In contrast of the
greater intra-group differences seen in biodiversity and OTU level community differences, results of functional traits are relatively clustered, with expected increases in microbes capable of mobility, biofilm formation and stress tolerance with induction of ozonation, yet median proportion of gram-positive strains dropped from 42–27% after ozonation, in contrast with common understandings.
Metacyc pathways predicted by PICRUSt2 that changed substantially with an ALDEx effect size of over 2.5 is shown in Fig. 4. Among 15 pathways highly specific to ozonation, mycolate biosynthesis, taxadiene biosynthesis (engineered), superpathway of heme biosynthesis from glycine, (5Z)-dodec-5-enoate biosynthesis are crucial pathways for assembling of proteins for cell membrane synthesis, although gram-negative bacteria increased with ozonation, repair and synthesis of cell membrane to mitigate oxidative distruption of ozone is seen in microbiome, along with production of anti-oxidants such as heme production and UDP-glucose-derived O-antigen biosynthesis. Of all enriched pathways after ozonation, phynylacetate degradation I (aerobic) pathway is the most significant with effect size of over 2.5, phenylacetate and degradation of other compounds containing acetate groups are widely seen in essential steps of toluene and xylene degradation, both recalcitrant compounds commonly seen as VOCs pollutants (Vardar et al. 2004, Leutwein et al. 2001). Purine ribonucleosides degradation feeding the urea cycle and aspartate degradation and pyrimidine deoxyribonucleosides degradation all contributes to to degradation of amino acids, though not directly fed to the microbiome, could be enriched. with the continuous death and accelerated cycle of cells under ozonation, contributing to the increased ability for utilization of amino acids shown in phenotypic metabolization results.
Decrease in TCA cycle participation is seen in ozonation filter, and pathways for degradation of other metabolites in VOCs degrading pathways such as 3-phenylpropanoate and inositol degradation is seen, indicating a motley ununiform changes in pathways after ozonation, no decisive changes affecting overall functional ability are seen between groups, but a mixture of degradation inclination changes favoring different parts of the VOCs degradation pathways.
3.6 Correlation of system performance and dominant taxa
Due to high redundancy known to bacterial populations (Escalas, Hale et al. 2019), majority of function is commonly assumed to be performed by the most abundant taxa in certain microbiome. Dominant taxa of both OTU and Genus level with top 20 relative abundance is selected and their abundance in different samples and operation time were plotted with removal efficiency and mineralization rates of corresponding samples, to investigate difference and similarity of reaction of dominant taxa and their relationship with performance measures, as shown in Fig. 5. By comparing both removal efficiency and mineralization rate, insights can be drawn from removal of toluene and complete degradation to CO2 and hence intermediate metabolites.
Both correlation of OTU and Genus level are chosen to investigate possible intra-genus differentiation and radicality of differences. 9 out of 20 genera are monotonously correlated with removal efficiency in both ozone and control biofilter, whereas only 3 out of 20 OTUs are monotonously correlated with removal efficiency, indicating a high specificity and different reaction of OTUs under same genus towards ozonation. genus Rhodococcus was reported in numerous studies to be dominant in degrdation of xenobiotics in biofilter systems (Portune, Perez et al. 2015, Allievi, Silveira et al. 2018), all four OTUs from the genus Rhodococcus present in this system are strongly proportional with removal efficiency in control biofilter but all negatively correlates in ozonated biofilter, suggesting a shift of degrading contribution by Rhodococcus and more towards OTUs of genera Devosia, Aquamicrobium and Rhizobiales with high positive correlation with performance, similar trends of these three genus are seen in combined genus level.
3.7 Topological analysis and co-occurrence network construction of microbiome
91 and 107 nodes, 146 and 256 links, 5.1 and 3.6 average path distance were found for control and ozone biofilters respectively, with the increase of average degree from 3.2 to 4.7 under identical specifications for network construction. A more connected and even network is seen in ozone microbiome, smaller path distance coupled with smaller centralization of betweenness is commonly seen as higher stability as major hubs are more diverged and less nodes are likely to be affected under shocks. Major hubs in control biofilter such as OTU68 and OTU309 are substantially irrelevant in ozone bioifilter, indicating a radical change in microbial network and hub distribution, yet OTU26 corresponding to the genus Pandoreae remain highly relevant in betweenness and degree count in both system, genus Pandoreae was reported to be highly enriched in species capable of effective degradation of xenobiotics (Peeters, De Canck et al. 2019), and is positively correlated with r > 0.7 in both biofilters in as shown in Fig. 6. Exact OTU taxonomy can be seen in supplementary material 2.