Chemical composition of material extractives influences microbial growth and dynamics on wetted wood materials

The impact of material chemical composition on microbial growth on building materials remains relatively poorly understood. We investigate the influence of the chemical composition of material extractives on microbial growth and community dynamics on 30 different wood species that were naturally inoculated, wetted, and held at high humidity for several weeks. Microbial growth was assessed by visual assessment and molecular sequencing. Unwetted material powders and microbial swab samples were analyzed using reverse phase liquid chromatography with tandem mass spectrometry. Different wood species demonstrated varying susceptibility to microbial growth after 3 weeks and visible coverage and fungal qPCR concentrations were correlated (R2 = 0.55). Aspergillaceae was most abundant across all samples; Meruliaceae was more prevalent on 8 materials with the highest visible microbial growth. A larger and more diverse set of compounds was detected from the wood shavings compared to the microbial swabs, indicating a complex and heterogeneous chemical composition within wood types. Several individual compounds putatively identified in wood samples showed statistically significant, near-monotonic associations with microbial growth, including C11H16O4, C18H34O4, and C6H15NO. A pilot experiment confirmed the inhibitory effects of dosing a sample of wood materials with varying concentrations of liquid C6H15NO (assuming it presented as Diethylethanolamine).


Results
Visible microbial growth and fungal qPCR. A typical example of visible microbial growth observed from an overhead picture, as well as an example of ImageJ processing to quantify visible growth, is shown in Fig. S1. Visible microbial growth was observed within 3 weeks on 13 of the 30 wood species, with different wood species demonstrating widely varying susceptibility to growth (Fig. 1). Of the 13 species with the greatest visible coverage area, Beech (Fagus grandifolia) showed the greatest amount of coverage followed by Ponderosa Pine (Pinus ponderosa) and Basswood (Tilia americana). Similarly, for fungal qPCR outcomes, we observed that Ponderosa Pine had the greatest fungal abundance, followed by Beech, Black Walnut (Juglans nigra), and Hard Maple (Acer saccharum) (Fig. 2). The correlation between visible microbial growth coverage and fungal qPCR concentrations, both at the end of the incubation period, was relatively strong given the uncertainties involved in swabbing, extraction, and PCR reactions (R 2 = 0.55; Fig. 3).
Relative abundance of fungal taxa. Using DADA2 amplicon inference software 45 , 274 amplicon sequence variants (ASVs) were inferred from the ITS amplicon data. These ASVs represent the overall distinct fungal community members identified by the ITS sequencing and Fig. S2 shows the rarefaction curve for each wood type. The ASVs most likely taxonomy classification was determined by mapping their ITS nucleotide Scientific RepoRtS | (2020) 10:14500 | https://doi.org/10.1038/s41598-020-71560-3 www.nature.com/scientificreports/ sequences to the UNITE reference database 46 . These taxonomies were later grouped at the family level for each wood type to identify the specific signatures and patterns across different materials (Fig. 4). Wood species are shown in descending order of qPCR magnitude from bottom to top, and the legend is sorted by taxa prevalence. Despite all being naturally inoculated in the same environment and wetted with the same water, it appears that inherent differences in material composition contributed to differences in which fungal families thrived after wetting and being held at high RH. For example, Aspergillaceae was most abundant across all of the wood samples. However, Meruliaceae was more prevalent on the 8 materials that were observed to have the greatest visible microbial growth coverage, especially for Beech and Maple. Ponderosa Pine, which had both high visible microbial growth coverage and high fungal qPCR concentrations, was dominated by Didymellaceae, while Pleosporaceae appeared in high abundance only on Shedua. Among the identified families in Fig. 4, Aspergillaceae, Cladosporiaceae, Tricholomataceae, Sporidiobolaceae and Nectriaceae were also found commonly in homes with dampness and mold 27,47-50 . Untargeted metabolomics of microbial and material samples. Figure 5a shows a comparison of the microbial and wood metabolites in a volcano plot, which revealed several broad trends. Although there is some overlap in the metabolites detected in these two data sets, there is a large number of metabolites that significantly differentiate the two groups, with differences in relative abundance up to 500-fold. Additionally,  www.nature.com/scientificreports/ a larger and more diverse set of metabolites was detected from the wood shavings compared to the microbial swabs, indicating a complex and heterogeneous chemical composition within the wood types. Data visualization by principal component analysis (PCA) plots revealed further insights. Comparison of the wood shavings and swab extracts on the same plot (Fig. 5b) shows that these groups have distinct signatures, and also that the swab extract samples cluster tightly together, indicating low heterogeneity. These observations mirror the trends seen in Fig. 5a. Furthermore, PCA of either the wood shaving or swab extract samples on    www.nature.com/scientificreports/ their own demonstrates that metabolomics data can be used to distinguish between different wood types. For example, within the swab extract samples (Fig. 5c), Ponderosa Pine was significantly separated from the other wood types; interestingly, this material also exhibited one of the highest levels of fungal growth (Figs. 2, 3) as well as a unique signature of fungal taxa (Fig. 4). Within the wood shaving samples, Sirari and Lovoa wood samples separate significantly from the others, across the PC1 dimension (Fig. 5d). Both of these wood types showed very low levels of microbial growth.
Chemical composition of wood material extractives and associations with microbial growth. From the metabolomics analysis of wood shaving samples, a total of 5,375 spectral features were correlated with molecular formulas. Examples of chemical composition results for both Ponderosa Pine and White Oak are shown in Fig. 6 for illustration. Spectra data from the other 28 wood samples are also shown in Fig. S3. www.nature.com/scientificreports/ Multi-compound analysis. Figure 7 and Table 1 show results from the ANOSIM cluster analysis to explore the potential for clusters of multiple compounds to be associated with microbial growth. Results from this dissimilarity test suggest that multiple compound composition factors had only small effects on the microbial growth variables (average ANOSIM R ≈ 0.2) with borderline significance (p ranged from 0.05 to 0.07). This cursory analysis suggests that there were no particular suites of compounds that were clearly associated with the extent  Single-compound analysis. Tables 2 and 3 show the top 35 and 25 compounds identified and quantified that were most strongly correlated with visible microbial growth and fungal growth quantified via qPCR, respectively (e.g., Spearman correlation coefficients > 0.6 and p < 0.001). Compound molecular weights and speculated compound formulas are both shown; note that compound formulas are speculative because compounds identified with a particular molecular weight could represent different compounds depending on molecular struc- www.nature.com/scientificreports/ ture. Also noted in Tables 2 and 3 are those compounds for which correlations were strongly correlated with both microbial growth outcomes, as well as those compounds that showed the clearest monotonic relationships between abundance and microbial growth outcome upon visual inspection (which is approximately 10 compounds per outcome). Correlations are also shown visually in Fig. S4 and S5 for the same 25-35 speculatively identified compounds. Results indicate that the putatively identified C 11 H 16 O 4 had the strongest correlation with visible microbial growth, with a Spearman's rank correlation coefficient of 0.742 (p < 0.0001) (Fig. 8). The compound formula that was most strongly correlated with fungal qPCR concentration was C 18 H 34 O 4 , with a Spearman's rank correlation www.nature.com/scientificreports/ coefficient of 0.768 (p < 0.0001) (Fig. 9). A strong positive correlation in these analyses suggest the potential for the presence/abundance of these compounds may encourage fungal proliferation. Figure 10 shows compound abundance versus microbial growth for several of the identified chemical compounds that were strongly correlated with both visible microbial growth and qPCR fungal growth outcomes. For example, C 18 H 28 O 4 revealed a strong and significant correlation with visible growth (rho = 0.72, P < 0.0001) and fungal qPCR concentrations (rho = 0.68, P < 0.0001), although the relationship appears largely driven by a small number of outliers (Fig. 10a, b). C 12 H 16 N 2 O 12 was strongly correlated with both outcomes, although little information could be found about the potential names of this compound formula (Fig. 10c, d). The compound formula C 6 H 15 NO had a strong negative correlation with both microbial growth outcomes (rho = 0.66, P < 0.0001), with a shape that suggests a near-monotonic nonlinear response below a particular threshold for both growth outcomes (Fig. 10e, f). Because of these combined factors, we conducted a pilot experiment to investigate the inhibitory effects of dosing a sample of wood materials (Ponderosa Pine) with varying concentrations of pure liquid C 6 H 15 NO (assuming it was present in wood samples as Diethylethanolamine) mixed with both tap water and distilled water at varying concentrations.
Inhibitory effect of C 6 H 15 NO on visible microbial growth assessment. Figure 11 shows the mean (± S.D.) visible microbial growth coverage areas on the Ponderosa Pine samples over time after wetted with varying concentrations of C 6 H 15 NO (assuming Diethylethanolamine) solutions mixed with (a) tap water and (b) distilled water. Test coupons that were wetted with tap water and the lowest concentration (0.0001%) of C 6 H 15 NO showed the greatest amount of microbial coverage, ranging from ~ 36% for the first week to ~ 71% for the last week (Fig. 11a). Conversely, coupons that were wetted with tap water and the greatest concentration (1%) of C 6 H 15 NO showed the lowest amount of microbial coverage, with only ~ 4% at the last week of incubation (Fig. 11a). The amount of microbial coverage area on the coupons of the tap water control group was ~ 13% for the first week and increased to ~ 57% for the 6th week, which is similar in magnitude to the coupons that were wetted with 0.001% and 0.01% solutions of C 6 H 15 NO.
Similarly, the coupons that were wetted by distilled water solutions mixed with C 6 H 15 NO showed less overall microbial growth compared to the coupons wetted with tap water (Fig. 11b). However, the inhibitory effect of C 6 H 15 NO is still clearly demonstrated, as the distilled water control group and lowest concentration of C 6 H 15 NO (0.0001%) showed the greatest visible microbial growth, followed by small amounts of visible growth at 0.001% C 6 H 15 NO and negligible or no visible growth at 0.01% C 6 H 15 NO and higher.

Discussion
Results of this study indicate that the extractive chemical composition of the tested wood material samples has a significant effect on the magnitude and dynamics of microbial growth on wetted surfaces. Moreover, results suggest that a better understanding of fungal colonization susceptibility of different materials commonly used in construction could be used to limit adverse health outcomes caused by exposure to pathogenic fungal species and lessen the chance of fungal-associated wood rot.
The identified fungal families are mainly composed of saprobic species that decompose and digest plant matter. However, several contain members that are mycoparasitic or pathogenic, either to plants or animals. The fungal families Aspergillaceae, Aureobasidiaceae, Cladosporiaceae, and Nectriaceae all contain one or more pathogenic species that can cause disease in humans. The pathogenic species from these fungal families can all be characterized as opportunistic pathogens, most often infecting those with weakened immune systems such as cancer patients, people with disorders of their immune systems, or those taking drugs that intentionally suppress the immune system [51][52][53] .
Invasive aspergillosis, caused by certain species of the Aspergillaceae family, is a disease that can occur in several organs of the body but is most often associated with pulmonary infections initiated by the inhalation www.nature.com/scientificreports/ of fungal spores [54][55][56][57] . A devastating illness for immunocompromised patients, mortality rates linked to invasive aspergillosis can range from 40 to 90% in certain cases 58 . Chronic human exposure to one species of Aureobasidiaceae, Aureobasidium pullulans, can induce hypersensitivity pneumonitis 59,60 . Commonly referred to as "humidifier lung", this disease is characterized by maladies of the lungs, including coughing, dyspnea, and acute inflammation. Species from the Cladosporiaceae family, although rarely pathogenic to humans, produce airborne spores that if one is exposed to over an extended period can cause adverse health effects, especially for people with asthma or those suffering from other respiratory diseases 61,62 . For Nectriaceae, the majority of species from this family are innocuous, soil-borne saprobes but several have been reported as opportunistic pathogens while others produce harmful mycotoxins 63 .   64 . Brown rot, associated with members of Fomitopsidaceae, results from the breakdown of the structural compounds cellulose and hemicellulose by fungal enzymes and alters the wood into shrunken, brown-discolored cubical pieces 65 . Differing from Fomitopsidaceae, certain species of Meruliaceae and Schizophyllaceae families cause another form of wood rot called white rot, wherein the lignin of moist wood is broken down, leaving behind a stringy, light-colored material composed mostly of undigested cellulose 66,67 .
Of the individual chemical compounds that were most strongly correlated with the extent of visible microbial growth and/or fungal qPCR concentrations, several may have some plausible explanations. For example, possible names for the putatively identified C 11 H 16 O 4 , which was positively correlated with visible growth, could be methylenolactocin and DETOSU. According to a compound search, methylenolactocin is apparently known as an isolate of Penicillium with anti-cancer activity. Similarly, a possible name for the putatively identified C 6 H 15 NO, which had a strong negative correlation with both microbial growth outcomes, includes Diethylethanolamine, which is used as a corrosion inhibitor and a precursor in the production of a local anesthetic. Subsequent testing of the application of this individual compound at varying concentrations in tap water confirmed its inhibitory effects. Given that the compound is an alcohol with some known toxicity, the association is not particularly surprising, and the compound may not be a viable candidate for use as an anti-fungal agent in materials. However, the process used herein to identify individual and/or clusters of compounds intrinsic to materials that correlate with microbial growth outcomes can be utilized and expanded to isolate other candidate compounds from woods and other material and could potentially inform their integration into other types of materials to increase microbial growth resistance under wetting conditions. While these data provide novel insights into the chemical drivers of differential microbial growth susceptibility within this otherwise homogenous set of material samples, it is important to note several limitations to this work, including: (1) only one test coupon for each type of wood was used, which does not allow for capturing variability inherent in microbial dynamics and growth on surfaces of the same material composition; (2) the material chemical composition analysis targeted a specific range of compounds (i.e., 70-1,050 m/z); other compositional www.nature.com/scientificreports/ analysis approaches could uncover additional insights outside of these bounds; and (3) the underlying mechanisms explaining observed associations between material compositions and microbial growth are not explored in detail. Future work with other materials and analytical approaches should focus on overcoming these limitations.

Methods
Thirty (30) different wood species were selected to study microbial growth susceptibility and community structure upon wetting. The wood samples were all purchased new in a kit from an online retailer in an attempt to collect a wide variety of wood species that were likely to experience a diversity of microbial growth patterns upon wetting. The wood samples were naturally seeded with environmental microbes, wetted, and then incubated at high relative humidity (RH) conditions. Microbial growth was assessed over time using a combination of visual assessment and quantitative polymerase chain reaction (qPCR). The same samples were also swabbed at the end of the experiments for ITS (fungal) rRNA amplicon sequencing. Additionally, powders from unwetted duplicates of each type of wood were shaved and collected for material chemical composition analysis using reverse phase liquid chromatography with tandem mass spectrometry (RPLC-MS-MS).
preparation of materials. The names and categorical classification of rot resistance of each wood type utilized are listed in Table 4 68-70. All tested wood materials were cut to 5 cm × 7.5 cm coupons and sterilized by UV. The samples were then naturally inoculated by environmental microbes by leaving the sterilized coupons unprotected in a laboratory setting for ~ 30 days. Next, to simulate what happens after a material comes in direct contact with bulk liquid from a flood or leak, all 30 wood coupons were submerged in tap water for ~ 12 h. Tap water (as compared to distilled water) was chosen to provide more realistic growth from a wetting event. The same laboratory tap water source was used to wet all coupons at the same time. After wetting, each coupon was then placed in individual petri dishes and incubated at room temperature (20-25 °C) inside a static airtight chamber at high RH for ~ 3 weeks to encourage fungal growth. Potassium nitrate salt solutions were used to maintain RH at ~ 94% for the duration of the experiment. www.nature.com/scientificreports/ Microbial growth assessment. Microbial growth was assessed using multiple methods. Visual microbial growth was assessed by taking overhead images on a weekly basis during a ~ 3-week incubation period. Image analyses were conducted using ImageJ to estimate the percentage of microbial growth coverage over time using the area fraction option. Additionally, at the end of the incubation period, the surfaces of the coupons were swabbed using sterile polyester swabs for subsequent sequencing and analysis, including ITS for fungal communities and qPCR for fungal quantification using universal primers. Finally, coupon surfaces were also swabbed with cotton-tipped swabs that were dipped in ethanol for subsequent surface chemistry and metabolomics analysis.
DNA extraction, sequencing, and qPCR. The microbial community from each sample was collected by rubbing the tips of sterile polyester swabs along the surface of the coupons. Following sample collection, the swab tips were cut off into DNA extraction tubes (DNeasy Powersoil, Qiagen), and the DNA extracted following the manufacturer's protocol 71 with the following modification to reduce sample loss: combine steps 7 through 10 by adding 150 µL each of solutions C2 and C3 to the tube that the lysed sample was transferred to in step 6. Following a 5-min incubation period at 4 °C, the manufacturer's protocol is resumed as normal at step 11. Amplification of the ITS region used the Illumina Earth Microbiome ITS protocol 72 . Reactions were pooled, cleaned with Agencourt AMPure beads, and then the clean amplicon pool was sequenced at Argonne National Laboratory's Environmental Sample Preparation and Sequencing Facility, following the Earth Microbiome Protocol 73 . For PCR cycling the following reaction mix was used: 9.5 μL of molecular biology grade H2O, 12.5 μL of Accustart II PCR Toughmix, 1 μL each of forward and reverse primers at 5 μM, and 1 μL of sample DNA for a total reaction volume of 25 μL. The following PCR program was used: Initial denaturing step at 94 °C for 3 min, followed by 35 cycles of: 94 °C for 45 s, 50 °C for 60 s, and 72 °C for 90 s, followed by a final extension step of 72 °C for 10 min. Sequencing was performed on an Illumina Miseq using V3 chemistry. Fungal amplicons were sequenced using 2 × 300 nt reads.
To quantify the total abundance of fungi, qPCR was performed using a Roche Lightcycler 480 and the SYBR Green I Master kit. Primers targeting the ITS1f.-ITS2 priming sites, without the Illumina adapters or barcodes, were used during amplification. To calculate the abundance for the fungal qPCR, the Femto Fungal DNA Quantification Kit from Zymo was used as a standard control.
Microbial taxonomy identification. ITS Amplicon produced 797,766 reads from 29 of the 30 woods; Leopardwood did not yield any ITS reads. Using DADA2 45 , the reads were clustered and 274 amplicon sequence variants (ASVs) were identified. The closest taxonomy groups were identified by mapping ASVs to the UNITE database 46 .
Microbial metabolomics and chemical composition analysis. Unwetted samples of each type of wood were shaved into ~ 100 mg of powder using sterile scalpels on the surface. For each sample, 20 mg wood shavings were transferred to an epitube, then 500 μL of 50/50 methanol/water was added, and the sample was vortexed. The mixture was left overnight at room temperature, and on the following day, samples were centrifuged and supernatant was transferred to a fresh epitube. The supernatants were evaporated to dryness and reconstituted in 100 µL LCMS buffer (5% acetonitrile/water + 0.1% formic acid). Ethanol extracts from surface swabs were evaporated to dryness and reconstituted in 200 µL of LCMS buffer. High resolution mass spectrometry analysis was performed according to previously reported procedures 23 . Analysis of LC-MS/MS data from the microbial swab extracts and the wood shavings was performed with Compound Discoverer 3.0 (Thermo Fisher), which identified metabolites as unique spectral features based on a combination of molecular weight and retention time. The predicted compound hits from Compound Discoverer were filtered to include only those compounds confirmed by both intact mass and MS/MS spectral matching via the mzCloud database. Compounds were identified by their molecular weight and quantified using the compound integrated area.
Statistical analysis. Nonparametric Spearman rank correlations were first used to explore associations between microbial growth and individual compounds identified and quantified in the material shavings. Over 5,000 identified compounds were first ranked by their Spearman correlation coefficients and p-values using Stata version 15 (StataCorp SE, College Station, TX, USA). The top ~ 30 compounds that showed the strongest correlations with microbial/fungal growth outcomes (positive or negative) were then visually inspected, and those with the clearest monotonic relationships were chosen for further analysis.
Additionally, a dissimilarity analysis was conducted using ANOSIM with the entire suite of 5,000 + compounds from the 30 wood species by using PAST (PAleontological Statistics) version 3 74 to determine if chemical signatures of multiple compounds (instead of just single compounds) were also correlated with microbial growth. In our visible microbial growth dataset, the 30 wood materials were clustered into three groups based on a histogram of visible coverage level (i.e., heavy growth, light growth, and no growth). The ANOSIM statistic, which compares the mean of ranked dissimilarities between groups to the mean of ranked dissimilaristies within groups, was used to determine whether distances between samples of the same growth level were significantly lower than distances between samples of different growth levels.
Single compound inhibitory effects: pilot experiment. Following the analysis of the 30 wood samples, a single compound found in the material composition analysis that was strongly and monotonically correlated with fungal growth was then tested in a pilot chemical dosing experiment to evaluate the potential inhibitory effect of the compound. A new batch of samples of ponderosa pine wood (one of the tested woods shown to Scientific RepoRtS | (2020) 10:14500 | https://doi.org/10.1038/s41598-020-71560-3 www.nature.com/scientificreports/ be most susceptible to fungal growth) was purchased and cut into 5 cm × 7.5 cm coupons and sterilized by UV. The coupons were naturally inoculated again by leaving them unprotected in a laboratory setting for ~ 30 days (same as before). Pure liquid chemical of a single putatively identified compound (C 6 H 15 NO) was purchased and mixed with both tap water and distilled water at varying concentrations (by volume), including 0.0001%, 0.001%, 0.01%, 0.1%, and 1%, and controls with 0%. Next, wood coupons were wetted by submerging them into each of the prepared varying concentration mixtures (triplicate coupons at each concentration), covered with aluminum foil, and placed inside a biosafety hood to soak overnight (~ 12 h). The coupons, including a control coupon wetted with sterile water and with none of the dosing compound, were then incubated at room temperature inside the same static airtight chamber at high RH conditions for several weeks (same as before).
Microbial growth was evaluated visually on a weekly basis during a ~ 6-week incubation period using the same image processing procedures as described previously.

Data availability
The datasets generated during the current study are available from the corresponding author on reasonable request.