Dominance of Gas-Eating, Biofilm-Forming Methylobacterium Species in the Evaporator Cores of Automobile Air-Conditioning Systems.

Air-conditioning systems (ACS) are indispensable for human daily life; however, microbial community analysis in automobile ACS has yet to be comprehensively investigated. A bacterial community analysis of 24 heat exchanger fins from five countries (South Korea, China, the United States, India, and the United Arab Emirates [UAE]) revealed that Methylobacterium species are some of the dominant bacteria in automobile ACS. Furthermore, we suggested that the predominance of Methylobacterium species in automobile ACS is due to the utilization of mixed volatile organic compounds and their great ability for aggregation and biofilm formation.


RESULTS
Microbial diversities in the ECs of automobile ACS in different countries. A total of 484,328 bacterial sequencing reads were generated from the pyrosequencing of 34 16S rRNA gene PCR amplicons. After removing low-quality and chimeric sequencing reads, 231,648 high-quality bacterial sequencing reads with an average of 6,756 numbers of reads per sample were obtained (Table 1). Because statistical community diversity indices, including the number of operational taxonomic units (OTUs), Shannon-Weaver index, and Chao1 index, are affected by the number of sequencing reads used for diversity index calculations, high-quality bacterial sequencing reads were normalized to the lowest numbers of 1,961 bacterial reads, and their statistical diversity indices were calculated using normalized sequencing reads. OTU, Shannon-Weaver, and Chao1 indices of bacterial sequencing reads representing bacterial community diversity in the ECs of automobile ACS were highly different depending on samples within the countries, but they were generally highest in Indian samples and next in UAE samples (see Fig. S1A in the supplemental material and Table 1). The box plot analysis of Chao1, a representative ␣-diversity index, for the bacterial sequencing reads clearly showed that the bacterial community in the ECs of automobile ACS operated in India was significantly more diverse than those in South Korea, China, and the United States (Fig. S1B). The box plot analysis also showed that the bacterial communities in UAE samples were a little more diverse than those in South Korea, China, and the United States, and the bacterial community diversities in the ECs of automobile ACS operated in South Korea, China, and the United States were relatively similar. However, bacterial community diversity indices were not related to car type or mileage (Table S1). Dominance of Methylobacterium species in ACS. The normalized bacterial sequencing reads were classified at the phylum and genus levels to investigate microbial communities in the ECs of automobile ACS operated in different countries. At the phylum-level analysis of bacterial sequencing reads, members of the phylum Proteobacteria were predominant, accounting for 50.1% to 99.3% of total sequences, in all EC samples regardless of country (Fig. 1A), which was in line with a previous result (21). Members of the phylum Proteobacteria were more abundant in Korean, Chinese, and U.S. samples, with a mean relative abundance of 91.8% Ϯ 6.5%, than in Indian and UAE samples, with a mean relative abundance of 64.3% Ϯ 11.2%. The phylum-level analysis of bacterial sequencing reads showed that the phyla Actinobacteria with 0% to 29.6% abundance and Deinococcus-Thermus with 0% to 18.8% abundance were the next dominant after Proteobacteria, but their abundances were quite varied depending on the samples. The phyla Actinobacteria and Deinococcus-Thermus were identified more abundantly from Indian and UAE samples, with mean relative abundances of 18.3% Ϯ 9.1% and 5.2% Ϯ 6.2%, respectively, than were those from Korean, Chinese, and U.S. samples, with mean relative abundances of 4.6% Ϯ 5.2% and 0.2% Ϯ 0.43%, respectively.
The genus-level analysis of bacterial 16S rRNA gene sequences showed that members of the genus Methylobacterium (phylum Proteobacteria) were predominantly identified from all samples, in agreement with the results from previous studies (3,6). However, their relative abundances were highly varied, with a range of 3.3% to 79.4% of the total bacterial sequences depending on the samples. Methylobacterium members were identified more highly from Korean, Chinese, and U.S. samples than in Indian and UAE samples, with mean relative abundances of 63.6% Ϯ 7.7% versus only 21.7% Ϯ 14.2%, respectively. The community analysis showed that Sphingomonas was a common genus identified from the ECs, and the relative abundance of Sphingomonas spp. was 6.4% Ϯ 4.3%. The genera Roseomonas, Rubrobacter, and Deinococcus were somewhat identified from only Indian and UAE samples. However, the genera Rubrobacter and Deinococcus were barely detected from Korean, Chinese, and U.S. samples (Fig. 1B). A hierarchical clustering analysis was performed using the relative abundances of the bacterial communities (Fig. 2). The bacterial communities of Korean, Chinese, and U.S. and Indian and UAE samples fell into respective clustering groups, and the Korean, Chinese, and U.S. and Indian and UAE groups were distinctly differentiated from each other, which suggests that the air environments or operating conditions of ACS in South Korea, China, and the United States are similar to each other but different from those of India and the UAE. However, 31 out of 34 samples clearly showed that Methylobacterium species were the dominant bacteria in the bacterial community from ACS (Fig. 2).
A linear discriminant analysis (LDA) effective size (LEfSe) algorithm for the identification of statistically differential bacterial groups between Korean, Chinese, and U.S. and Indian and UAE samples clearly showed that the genera Methylobacterium and Sphingomonas were more abundant in Korean, Chinese, and U.S. samples than in Indian and UAE samples, whereas the genera Roseomonas, Rubrobacter, Deinococcus, and Barrientosiimonas were more abundant in Indian and UAE samples than in Korean, Chinese, and U.S. samples ( Fig. 2 and dS1C). In particular, members of Methylobacte- rium, the most predominant bacterial genus group in the ECs, were significantly more abundant in Korea-China samples than in Indian and UAE samples, and the genera Roseomonas and Rubrobacter were significantly more abundant in Indian and UAE samples than in Korea-China samples ( Fig. 2 and S1C and D).
Superior growth of Methylobacterium species from the EC of ACS in mixed VOCs. Atmospheric VOC profiles of four large cities in the Republic of Korea demonstrated that toluene is the highest concentration of VOC in the air from all four cities and ethyl acetate is the second most serious air pollutant in Daegu and Seoul (Table 2). In addition, washer liquid used in automobiles contains a high concentration of methanol (range, 30% to 90%) (22). Thus, the growth of bacteria derived from automobile ACS in the above-mentioned three VOCs was assessed. A total of 91 species of the 628 species isolated from car evaporators were selected in the following phyla: Alphaproteobacteria (34 species), Betaproteobacteria (10 species), Gammaproteobacteria (1 species), Bacteroidetes (5 species), Firmicutes (8 species), and Actinobacteria (33 species), without duplicated species (Table S2). Among them, 12 species of Methylobacterium were identified. Gordonia species showed the highest growth under all three single VOCs at each time point (Fig. S2A). However, the growth of Methylobacterium species was not surprisingly increased in toluene-supplemented minimal salt basal (MSB) medium, except for Methylobacterium aquaticum ( In addition, these species started to grow at 96 h, which is relatively slow compared to the growth under toluene and ethyl acetate. Because complex VOCs are present in the real world, a mixture of three VOCs was provided to Methylobacterium species (M. currus, M. radiotolerans, M. frigidaeris, M. dankookense, M. brachiatum, and M. organophilum) which grew poorly under a single-VOC-supplied medium. Surprisingly, enhanced growth was observed in all tested Methylobacterium species under the mixed three-VOC-added conditions compared to the growth when a single-carbon source was added ( Fig. 3A and S2B). The most dramatically increased growth was approximately eight times at 144 h in M. dankookense under mixed conditions compared to that in a single-carbon source-added medium (0.5% mixture [0.17% toluene, 0.17% ethyl acetate, 0.17% methanol] and 1.5% mixture [0.5% toluene, 0.5% ethyl acetate, 0.5% methanol]) ( Fig. 3 and S2B). It is notable that mixed VOCs do not have toxicity to but rather promote the growth of Methylobacterium species. Biofilm former versus aggregator lifestyles of Methylobacterium groups. The growth test of 91 species on R2A medium revealed that most bacteria reached stationary phase within 60 h. In addition, the optical density values of Methylobacterium species at 60 h were close to or lower than the mean OD 600 value of all tested bacteria (Fig. S4A). To compare dramatic biofilm formation, we performed a biofilm assay of automobile ACS-derived bacteria grown at 60 h in R2A medium (Fig. 4). As a result, a wide range of biofilm formation values (OD 595 /OD 600 ratio) among automobile-isolated bacteria was observed from 7.7 to 0.07. To evaluate whether the classification of biofilm formers was valid, they were grouped as extreme (OD 595 /OD 600, Ͼ3), high (OD 595 / OD 600 , 0.9 to 3), and normal (OD 595 /OD 600 , Ͻ0.9) biofilm formers ( Fig. 4 and S3B and Table S3). Statistically valid classifications (DF ϭ 2, F ϭ 347.8, P Ͻ 0.001) showed that 9   Table S3). It was intriguing that dominance in the bacterial population is associated with the biofilm formation ability. Thus, we retrieved the community profiles, and a correlation analysis between biofilm formation and bacterial population size was conducted. The result revealed that high biofilm formers occupied a large population under automobile ACS niches (Spearman correlation coefficient, 0.23; P Ͻ 0.05; Fig. S3C). However, there was no statistically significant difference in biofilm formation between the two groups, Methylobacterium and non-Methylobacterium species (DF ϭ 1, F ϭ 1.54, P ϭ 0.11; Fig. S3D). Taken together, the dominance in bacterial community is associated with the ability for biofilm formation, but this does not explain the prevalence of Methylobacterium species under automobile ACS.
Although it is generally believed that aggregation is one of the essential steps toward biofilm formation, biofilm formation requires attachment to a surface, resulting in the production of a slimy extracellular matrix (15,23). However, we observed that certain Methylobacterium species produced a small amount of biofilm but strongly aggregated with each other rather than attached to the surface ( Fig. 4 and S4A). Like M. currus, aggregate formations of M. radiotolerans and M. frigidaeris were observed during their growth. Conversely, Methylobacterium dankookense and M. organophilum were freely distributed in the medium, like M. brachiatum (Fig. S4A), which belonged to the high biofilm formers (Fig. 4). To quantify aggregation, aggregation percentages were measured during the growth of the above-mentioned six Methylobacterium species on tryptone-glucose-yeast extract (TYG) medium (Fig. S4B). As a result, M. frigidaeris is the highest autoaggregator, showing 75.4%, 90.0%, and 86.8% aggregation in 12, 24, and 48 h, respectively. M. currus was also aggregated, with high percentages of 29.1%, 53.6%, and 50.8% at the same time intervals, respectively. Thus, M. currus, M. radiotolerans, and M. frigidaeris were grouped as aggregators but those that form a small amount of biofilm. M. brachiatum, M. dankookense, and M. organophilum were designated biofilm formers, which are relatively high-biofilm-forming species rather than aggregators (Fig. 4).
Main factors of aggregation and biofilm formation in Methylobacterium species. For a deeper understanding of aggregation and biofilm formation, comparative genome analysis was performed with genomes of M. currus, M. radiotolerans, M. frigidaeris, M. brachiatum, and M. organophilum. Aggregation often occurs between cells conjugated by type IV pili to exchange genomic substances (24). The numbers of genes encoding pili were varied between Methylobacterium species. Nine pilusencoding genes of the trb operon (trbBCDEJLFGI) were present in the second contig of M. currus (Table 3). Furthermore, there were 8 genes (e.g., tra and trb genes) for the synthesis of pili in M. radiotolerans and 11 genes (e.g., pilZ and cpa genes) in M. frigidaeris. In contrast, only four genes (trwC, traC, traG, and trbI) are located in the major contig of M. brachiatum, and few genes could be observed in the genome of M. organophilum. Field emission scanning electron microscopy (FE-SEM) analysis revealed that M. currus was conjugated by pili that could be also observed in other aggregators (M. radiotolerans and M. frigidaeris) (Fig. S5A). However, this pilus network was not detected in the biofilm former group, including in M. brachiatum, M. organophilum, and M. dankookense.
Previous research studies have suggested that motility is important for aggregation, because nonmotile cells tend to form aggregates and hold each other (25).  (Table S4). Likewise, Methylobacterium species synthesize similar or different EPS among species.
Enhanced biofilm formation during coculture of aggregators and biofilm formers. Confocal laser scanning microscopy (CLSM) analyses of M. currus and M. brachiatum during single culture or coculture showed that the biofilm production was enhanced when they were grown together rather than singly at 48 and 72 h (Fig. 5A). The investigation of the biofilm formation and cell-forming units between M. currus and M. brachiatum during coculture suggested that a low abundance of M. currus was maintained until 48 h (6.25% at 24 h and 5.56% at 48 h; Fig. 5A). At 96 h, M. currus still remained, although the amount of biofilm in mixed culture of M. currus and M. brachiatum was reduced, implying that the biofilm produced by M. brachiatum enables M. currus aggregates to attach to the surface. Furthermore, EPS from M. brachiatum and pili from M. currus could be also observed when the two species were cocultured (Fig. 5B). Taken together, biofilm formation was enhanced due to the abilities of aggregation and biofilm formation in mixed culture.

DISCUSSION
Water vapor condensation by air cooling on ECs causes microbial growth by providing water, an indispensable life-giving factor (4). In addition, airflow through tiny air conduits of automobile ACS to efficiently cool down the air temperature leads to the deposition of microorganisms and airborne particles on the EC surface. Therefore, microbial growth, as well as the deposition of airborne microorganisms, may be an important factor in determining microbial communities in the ECs of automobile ACS. Temperature, humidity, dust particles, and VOCs that have a strong influence on the microbial growth in the ECs and microbial community in the air are different depending on the country. Therefore, in this study, microbial communities in automobile ACS that were operated in seven places in five countries were investigated using a barcoded 454-pyrosequencing strategy (26).
Microbial diversity analysis showed that bacterial communities in the ECs of automobile ACS operated in India and the UAE, especially in India, were more diverse than those in South Korea, China, and the United States ( Fig. 1 and S1), which might be caused by the differences in environmental and climate factors such as temperature, humidity, and VOCs (3,4). Because the average temperatures for the year in India (Delhi) and the UAE (Dubai) are higher than those in South Korea (Namyangju and Ulsan), China (Beijing and Shanghai), and the United States (Irvine, CA), the operation times of automobile ACS in India and the UAE will also be longer than those in South Korea, China, and the United States. The long operation time of automobile ACS will cause thick biofilm formation in the ECs with diverse microenvironments, probably giving rise to more diverse bacterial communities, which may be one of the most important for malodor production from automobile ACS (3). The EC of automobile ACS has unique environmental conditions, experiencing drastic fluctuations in temperature and water availability. It encounters high temperature by engine operation with no air-conditioner operation, whereas its temperature drops to almost zero during airconditioner operation. When the air conditioner is not in operation, water, a necessity for microbial growth, is not provided. Dust particles or VOCs in the air may be a sole nutrient or carbon source for microbial growth in automobile ACS. Therefore, microorganisms that are able to grow in the ECs of automobile ACS may be tolerant to several stresses and have the ability to metabolize dust particles or VOCs.
The VOC profiles in the air of four Korean cities suggested that benzene-tolueneethylbenzene-xylene and ethyl acetate were abundantly present, which is compatible with the previous literature (27). In addition, bacterial community analysis revealed that Methylobacterium species were one of the dominant bacteria in all ECs of automobile ACS used in this study ( Fig. 1 and 2). Members of the genus Methylobacterium are typically strictly aerobic methylotrophic bacteria and grow well on single-carbon compounds as the sole sources of carbon and energy, as well as on a wide range of multicarbon growth substrates (28). Recently, novel Methylobacterium species have been isolated from ACS (29,30), and 12 species of Methylobacterium were identified in the ECs of automobile ACS in this study, implying that Methylobacterium members might have the ability to metabolize VOCs as the nutrient or carbon source to grow in the ECs of automobile ACS. Acetylesterase (EC 3.1.1.6) is the key enzyme that mediates ethyl acetate to acetate and ethanol in the classical ethyl acetate metabolism (31). The enzyme has been mainly discovered and studied in Pseudomonas species, so that acetylesterase (NCBI RefSeq accession no. AAM16269) in Pseudomonas putida as a query sequence was used to search the homologs with the reference genomes of M. currus, M. frigidaeris, M. radiotolerans, M. organophilum, and M. brachiatum. A BLASTP search showed that all Methylobacterium species possess the ␣/␤-hydrolase that has more than 30% similarity and 70% query cover with the query sequence (NCBI RefSeq accession numbers WP_099953624.1, WP_099900826.1, WP_116656825.1, WP_053621636.1, and WP_091857735.1), implying that ethyl acetate metabolism could occur via acetylesterase in Methylobacterium species. Among five toluene-degrading pathways that have been revealed in Pseudomonas putida F1, P. putida PaW15, Burkholderia cepacia, Ralstonia pickettii PKO1, and Pseudomonas mendocina KR1 (32), Methylobacterium species possibly oxidize toluene to acetyl-coenzyme A (acetyl-CoA), because homologs of XylA (xylene monooxygenase electron transfer), XylB (benzyl alcohol dehydrogenase), and XylC (benzaldehyde dehydrogenase) were present in all Methylobacterium species (Fig. S5E). Furthermore, all five Methylobacterium species possess multiple copies of methanol dehydrogenases. In the case of M. frigidaeris, there are two copies of PQQ-dependent dehydrogenase encoding gene in the genome, which was the lowest copy number among Methylobacterium species. Thus, all above-mentioned homolog searches indicated that Methylobacterium species is a potential VOC utilizer.
However, contrary to expectations, many Methylobacterium spp., such as M. dankookense, M. frigidaeris, M. organophilum, M. longum, and M. currus, did not grow sufficiently in single-VOC-supplied media, except for two Methylobacterium species (M. aquaticum and M. rhodesianum; Fig. S2A). Methylobacterium species are well-known slow growers even in rich medium (9), suggesting that their metabolisms are slow. However, the growth of all tested Methylobacterium species was increased when a three-VOC mixture (toluene, ethyl acetate, and methanol) was supplied (Fig. 3). Even when the total concentrations of single and mixed VOCs (0.5%) were adjusted, the growth of Methylobacterium species was enhanced in the three-VOC mixture, implying that growth improvement was not due to the amount of carbon availability under VOC-mixture-added conditions ( Fig. 3 and S2B). Based on VOC metabolic pathways of Methylobacterium species, acetyl-CoA enters the ethyl malonyl cycle after it is generated through toluene, ethyl acetate, and methanol metabolism (Fig. S5D). Thus, the activation of several genes, including those of the ethyl malonyl-CoA pathway and the tricarboxylic acid cycle, might happen during toluene and ethyl acetate metabolism, supported by the superior growth of Methylobacterium spp. in the sole toluene or ethyl acetate rather than in methanol-supplemented medium (Fig. 3). This activation might accelerate the growth of Methylobacterium species in the coexistence of VOCs (33,34). Because it is obvious that a variety of VOCs are present in the air, improved growth under mixed-VOC-added conditions can be a reasonable explanation for the predominance of Methylobacterium species in automobile ACS as well as in other natural or artificial environments.
As suggested above, ACS is a harsh environmental condition due to the existence of several stresses, and one possible feature required for growth in ECs is that bacteria have the ability to form biofilms by producing EPS (6). Although the growth of Methylobacterium species is not above average among bacteria, even in R2A-rich medium, the profiles of the biofilm assay showed that 9 of 12 Methylobacterium species belong to the extreme-or high-biofilm-former groups. However, the comparison of biofilm data between Methylobacterium and non-Methylobacterium spp. proved to be statistically invalid due to the large gap between extreme biofilm and normal biofilm formers in Methylobacterium species ( Fig. 4 and S3D). Nevertheless, the combination of bacterial population data and the biofilm assay profile revealed that high biofilm formers predominantly occupy the microbial population in the ECs of automobile ACS (Fig. S3B).
Interestingly, all three aggregators possessed many pilus synthesis-associated genes, whereas M. brachiatum and M. organophilum did not. In addition, SEM analysis showed that M. currus cells were connected by pili, and other aggregators also displayed conjugation between cells, indicating that pili seem to be the main factor for aggregation (Fig. S5A). Additionally, genome analysis of M. currus revealed that flagellin-encoding genes were divided into three large clusters, and the genes also occupy a similar location in the genome of M. frigidaeris (Fig. S5B). These distant three clusters possibly delay the assembly of flagella, resulting in the lack of motility (Fig. S5B). M. radiotolerans showed two genotypical features of aggregators and biofilm formers. A succinoglycan synthesis operon and one large gene cluster in a flagellar synthesis operon were located in the genome of M. radiotolerans ( Fig. S5B and D). However, pilus synthesis-related genes were present in M. radiotolerans. In addition, this species showed an aggregator phenotype (nonmotile and conjugation) ( Table 3 and Fig. S4 and S5A and C); thus, it seems that pilus synthesis function is more important in the lifestyle of M. radiotolerans. Coculture of the aggregator M. currus and the biofilm former M. brachiatum, a low biofilm former, increased biofilm formation compared to that with the single culture, and EPS could also be observed ( Fig. 5A and B). It is expected that M. brachiatum may provide the surface for M. currus to aggregate and, as a result, produce increased biofilm, although a low abundance of M. currus remains in the biofilm, which is the other reason for the dominance of Methylobacterium species in automobile ACS.
In this study, the microbial populations in EC samples as part of automobile ACS from five different countries (South Korea, China, the United States, India, and the UAE) were analyzed by performing high-throughput pyrosequencing. The next-generation sequencing data analyses and taxonomic classification data clearly showed the microbial community of the EC surface, and Methylobacterium species are commonly dominant. This result is expected due to the utilization of mixed VOCs and the great ability for biofilm formation. However, aggregator Methylobacterium species, such as M. currus, tend to form a small amount of biofilm; in contrast, biofilm formers, such as M. brachiatum, did not show aggregation. Comparative genome analysis showed that flagella, pili, and large membrane proteins are correlated with the aggregation of Methylobacterium species. Finally, when aggregators and nonaggregators are cultured together, biofilm production was promoted. These data support why Methylobacterium species are predominant in automobile ACS; furthermore, the data are applicable for the removal of malodor in a car so that a pleasant vehicle indoor environment can be created.
Aggregation assays of Methylobacterium species. The measurement of aggregation was performed for Methylobacterium species grown in TYG medium at 48 h. At each time point at 12, 24, and 48 h, 1 ml bacterial cells was transferred to a 1.7-ml EP tube and harvested by preprophase band (PPB) centrifugation at 800 rpm for 1 min. Then, the supernatant was measured immediately by spectrophotometry (Eppendorf, Germany). The aggregation percentage was determined using the equation where A t represents the absorbance at different time points (t ϭ 24, 48, and 72 h) and A 0 represents the absorbance before the centrifugation.
EPS purification and HPAEC analysis. EPS purification was conducted using the ethanol precipitation method. M. currus and M. brachiatum were incubated in 1 liter TYG medium for 48 h and centrifuged to obtain a clear supernatant. The volume of the ethanol supernatant was added three times and incubated at 4°C overnight. Then, the sample was centrifuged, and the pellets were washed three times with an 80% ethanol solution, followed by three washes with PBS to remove residues and then freeze drying for 3 days. The composition of freeze-dried EPS samples was analyzed via HPAEC (ICS-5000; Dionex Co., USA). The CarboPac PA-1 column (Dionex) was used with 18 mM NaOH as the solvent at a flow rate of 1.0 ml/min at 25°C. HPAEC was performed by Biosystems, South Korea.
FE-SEM and CLSM analysis. For observation of phenotypical morphology, M. currus and M. brachiatum cells were incubated in TYG medium for 48 h (for single culture, OD 600 , 0.04; for mixed culture, OD 600 , 0.02 for each strain). To prepare FE-SEM samples, 5 ml cells was harvested by centrifugation (1 min, 13,000 ϫ g), and the pellets were fixed using Karnovsky's fixation method at 4°C overnight. After cold incubation, the pellets were washed with 0.05 M potassium phosphate buffer three times for 10 min each at 4°C. Dehydration of cells was performed at room temperature using 30%, 50%, 80%, and 100% ethanol serially. The samples were coated with platinum before FE-SEM (FEI, Japan) analysis. Biofilm cells were stained with FilmTracer Sypro Ruby biofilm matrix stain for 30 min at room temperature and observed by CLSM (Carl Zeiss, Germany). Confocal images of stained biofilm were observed under red fluorescent light (excitation wavelength, 450 nm; emission wavelength, 610 nm) to evaluate the height and density of the biofilm.
Statistical analysis. The LEfSe algorithm was used to identify significantly different genus or KEGG pathway abundances between the Korean, Chinese, and U.S. and Indian and UAE groups (44). Relative abundance information of genera or KEGG metabolic pathways was imported into LEfSe (version 1.0) on the Web-based Galaxy, and only genera or KEGG metabolic pathways with logarithmic LDA scores of Ͼ4.0 (genus) or Ͼ3.0 (KEGG) were included. The relative abundances of significantly different taxa between the Korea-China-UAE and Indian and UAE groups were compared through box plot analyses using the package ggplot2, and a Wilcoxon signed-rank test was performed to determine statistical significance. Validation of high, medium, and low biofilm former classification and comparative means of biofilm formation for Methylobacterium and non-Methylobacterium species were evaluated by Tukey's test using PROCGLM (SAS). Furthermore, the correlation coefficients and E values between biofilm formation and population size were calculated by Spearman rank correlation.
Data availability. The sequenced data for M. brachiatum TX0642 were deposited in GenBank (accession numbers CP033231 to CP033234). Pyrosequencing data were also deposited in the Sequence Read Archive (SRA; https://www.ncbi.nlm.nih.gov/sra) under accession numbers SRR7769850 to SRR7769883.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.