Detection of Legionella Species and the Inuence of Precipitation on the Amount of Legionella DNA Present in Aerosols from Outdoor Sites Near Asphalt Roads in Toyama Prefecture, Japan, with Microbiome Analysis

Background: Legionellosis can be caused by the inhalation of aerosolized water contaminated with Legionella. In this study, we investigated the prevalence of Legionella species in aerosols collected from outdoor sites near asphalt roads, bathrooms in public bath facilities, and other indoor sites such as buildings and private homes using culture methods, quantitative PCR with ethidium monoazide treatment (EMA-qPCR), and 16S rRNA gene amplicon sequencing. Results: Legionella species were not detected in culture. However, Legionella DNA was detected in 114/151 (75.5%) air samples collected near roads (geometric mean ± standard deviation was 1.80 ± 0.52 log 10 copies/m 3 ); these numbers were comparable to those obtained from bathrooms [15/21 (71.4%), 1.82 ± 0.50] and higher than those obtained from other indoor sites [11/30 (36.7%), 0.88 ± 0.56] (P < 0.05). By EMA-qPCR, Legionella DNA was detected in 20/30 (66.7%) samples collected near roads, indicating the presence of membrane-intact Legionella cells in the air. The amount of Legionella DNA correlated with the monthly total precipitation (r = 0.25, P < 0.01). It was also directly and inversely correlated with the daily total precipitation for seven days (r = 0.21, P = 0.01) and one day (r = −0.29, P < 0.01) before the sampling day, respectively. In addition, 16S rRNA gene amplicon sequencing revealed that the three most abundant bacterial genera in the samples collected near roads were Sphingomonas (21.1%), Streptococcus (14.6%), and Methylobacterium (1.6%); Legionella species were detected in 9/30 samples (30%) collected near roads (mean proportion of reads, 0.11%). At the species level, L. pneumophila was detected in 2/30 samples collected near roads (mean proportion of reads, 0.11% and 0.09% in the detected samples). Conclusions: DNA from Legionella species, including Legionella To the relationships between the rates of DNA and the sampling source type and the detection rates and sampling site (locations A ‒ L), Fisher’s exact test followed by post hoc Holm test were performed. We also investigated the relationships between the amount of Legionella DNA and the sampling source type, and between the amount of Legionella DNA and sampling site (locations A ‒ L) by using the one-way ANOVA test, followed by Tukey–Kramer multiple comparisons. The Student's t-test was performed to compare the amounts of Legionella DNA between samples associated with high and low total precipitation. Fisher’s exact test was performed to compare the rates of Legionella DNA detection between samples associated with high and low total precipitation. These tests were performed using the R statistical software package (version 3.0.0). Pearson’s correlation coecients (r) between the climatic conditions (air temperature, relative humidity, total precipitation, and wind speed) and the amount of Legionella DNA were investigated using Microsoft Excel (Microsoft, Tokyo, Japan). An absolute value of Pearson’s r ≥ 0.20 was considered indicative of a correlation, and a P value < 0.05 was considered signicant.

serve as potential environmental reservoirs of L. pneumophila [11]. Legionella present in puddles on roads could be aerosolized by splashing or by wind. Recently, several studies have attempted to detect Legionella species in aerosols released from hot tap water in bathrooms, shower water, and compost [12][13][14]. These studies revealed that Legionella species were present in the aerosols derived from these environments. However, to date, the prevalence of Legionella species in aerosols from outdoor sites near asphalt roads has not been analyzed.
Typically, conventional plate culture methods have been used to detect Legionella in clinical and environmental samples. In some cases, amoebic co-culture has been used because of its high sensitivity [15]. In addition, some rapid detection methods, such as PCR/real-time quantitative PCR (qPCR), have been applied in the eld [16,17]. However, these methods lack the ability to differentiate between viable and dead cells. Ethidium monoazide (EMA) is a dye that allows the differentiation of viable and dead cells [18]. EMA can penetrate compromised cell walls and membranes, form covalent links with DNA, and cleave it into small fragments after photoactivation. Thus, viable Legionella cells can be selectively quanti ed through the combined use of photoactivated EMA and qPCR (EMA-qPCR) [19]. Furthermore, metagenomic analysis is a powerful tool for analyzing microbial communities and 16S rRNA gene amplicon sequencing has been widely used to detect bacterial pathogens in environmental samples [20].
The main objective of this study was to determine whether Legionella species present in aerosols derived from outdoor sites near asphalt roads could be a source of Legionella infection. We investigated the prevalence of Legionella species in aerosols from outdoor sites near asphalt roads, bathrooms next to bathtubs in public bath facilities, and other indoor sites such as buildings and private homes using culture methods, EMA-qPCR, and 16S rRNA gene amplicon sequencing.

Results
Prevalence of Legionella species in air samples Legionella species were not isolated by direct culture or amoebic co-culture methods from any of the 202 air samples collected from roads or bathrooms and other indoor sites (Table 1). Although 49 Legionellasuspected colonies grew on glycine-vancomycin-polymyxin B-cycloheximide (GVPC) agar plates, they also grew on blood agar plates, indicating that they do not belong to the genus Legionella. There was low background growth on the GVPC agar plates.

Geographic And Meteorological Characterizations
The numbers of air samples collected near roads at the 12 sampling sites were as follows: At these sampling sites, the detection rates of Legionella DNA ranged from 60.0-93.3%. The geometric means ± SD (log 10 copies/m 3 ) of Legionella-speci c 16S rRNA genes in the Legionella DNA-positive samples ranged from 1.54 ± 0.66 to 2.03 ± 0.42. The differences in the detection rates according to sampling sites and the amount of Legionella-speci c 16S rRNA genes according to sampling sites were not signi cant by Fisher's exact test followed by post hoc Holm test and the one-way ANOVA test, respectively.
We assessed the correlation between the climatic conditions (air temperature, relative humidity, total precipitation, and wind speed) and the amount of Legionella-speci c 16S rRNA genes (log 10 copies/m 3 ) in the Legionella DNA-positive samples ( Table 2). The amount of Legionella DNA correlated with the monthly total precipitation (r = 0.25, P < 0.01). It was also directly and inversely correlated with the daily total precipitation for seven days (r = 0.21, P = 0.01) and one day (r = − 0.29, P < 0.01) before the sampling day, respectively. The scatter plots of total precipitation and the amount of Legionella DNA are shown in Fig. 1. The geometric mean ± SD of the Legionella DNA-positive samples at a monthly total precipitation of > 200 mm (2.00 ± 0.40 log 10 copies/m 3 ) was higher than that at a monthly total precipitation of ≤ copies/m 3 ) were also higher than those at a daily total precipitation of ≤ 10 mm (72.6%, 98/135 samples; 1.75 ± 0.53 log 10 copies/m 3 ) (P < 0.05; Fisher's exact test and Student's t test, respectively). However, the geometric mean of Legionella DNA one day before the sampling day at a daily total precipitation of ≤ 10 mm (1.83 ± 0.49 log 10 copies/m 3 ) was higher than that at a daily total precipitation of > 10 mm (1.55 ± 0.64 log 10 copies/m 3 ; P < 0.05, Student's t-test); the detection rates of Legionella DNA one day before the sampling day at daily total precipitation of ≤ 10 mm (75.0%, 99/132 samples) and > 10 mm (78.9%, 15/19 samples) were not signi cantly different (P ≥ 0.05, Fisher's exact test). At the genus level, 485, 421, and 368 bacterial genera were detected in samples collected from roads and from bathrooms and other indoor sites, respectively (Fig. 2). The top three most abundant bacterial genera in the samples collected near roads were Sphingomonas (21.1%), Streptococcus (14.6%), and Methylobacterium (1.6%); those in the samples collected in bathrooms were Sphingomonas (17.6%), Pseudomonas (5.4%), and Methylococcus (4.3%); and those in the samples collected from indoor sites other than bathrooms were Sphingomonas (19.2%), Achromobacter (5.0%), and Arthrobacter (3.8%).
Reads from Legionella species were detected in 9/30 samples (30%) collected near roads (mean proportion of reads, 0.11%), 5/21 samples (24%) in bathrooms (mean, 0.04%), and 1/30 samples (3%) from indoor sites other than bathrooms (mean, 0.03%). The rate of positivity for the samples collected near roads was signi cantly higher than the positivity rate for samples collected from indoor sites other than bathrooms (P < 0.05; Fisher's exact test followed by post hoc Holm test).
At the species level, L. pneumophila was detected in 2/30 samples collected near roads (mean proportion of reads, 0.11% and 0.09% in the detected samples) and in 1/21 samples collected in bathrooms (mean, 0.15% in the detected sample). Furthermore, L. birminghamensis and L. geestiana were detected in 1/30 samples collected from roads (mean, 0.81% and 0.02% in the detected samples, respectively), and L. nautarum was detected in 1/30 samples collected from indoor sites other than bathrooms (mean, 0.79% in the detected sample).
Alpha diversity indices in air samples are shown in Fig. 3. Shannon diversity indexes (a quantitative measure of bacterial community richness) and Pielou's evenness (a measure of community evenness) were not signi cantly different between the three sampling source types (P ≥ 0.05, Kruskal-Wallis test).
However, Faith's phylogenetic diversity indices, a qualitative measure of bacterial community richness that incorporates phylogenetic relationships between the features associated with samples from road exhibited a much higher value than those from other indoor sites (FDR-adjusted P < 0.05; Kruskal-Wallis test, Benjamini-Hochberg correction). The principal coordinates plot showed that a portion of the air samples collected from roads and indoor sites other than bathrooms plotted separately from those from bathrooms (Fig. 4). Linear discriminant analysis (LDA) effect size (LEfSe) analysis revealed 15 genera with an LDA score of at least 3.0, which were signi cantly more abundant in the three sampling source types ( Fig. 5; P < 0.05, pairwise Wilcoxon test). Speci cally, we found the following four genera to be more abundant in the samples (LDA score > 4.0): Pseudomonas, Vibrio, and Staphylococcus from bathrooms (red color), and Achromobacter from other indoor sites (green color).

Discussion
We demonstrate that DNA from Legionella species is widely present in aerosols derived from outdoor sites near asphalt roads, especially during the rainy season, regardless of the sampling site. Aerosols collected near roads contain viable Legionella cells, as shown by EMA-qPCR. In addition, 16S rRNA gene amplicon sequencing revealed that DNA from several Legionella species, including L. pneumophila, which is a major causative agent of Legionnaires' disease, were detected in aerosols collected near roads.
Legionella DNA was detected in more than 70% of the air samples collected near roads. The positivity rate for samples collected near roads was almost the same as that for samples collected in bathrooms in public bath facilities, which are a major source of legionellosis in Japan, and the rate was signi cantly higher than that for samples collected from other indoor sites (P < 0.05). Montagna et al. reported that Legionella DNA was detected in 72.7% (8/11) of air samples from bathrooms in healthcare facilities by qPCR using a Coriolis µ air sampler [12]. In this study, although the sampling height was not at 150-180 cm above ground level, which is the average height of an adult, owing to the stability of the air sampler, our results showed the existence of aerosols, including Legionella species near asphalt roads. Further investigation is required concerning the relationship between the sampling height and the prevalence of Legionella species in aerosols to reveal the risk of inhalation of aerosols, including Legionella species.
However, no Legionella species were isolated from any of the collected air samples (roads, bathrooms, and other indoor sites) by culture or amoebic co-culture methods. This may be due to the relationship previously reported, which indicated that a high concentration of Legionella species (> 300,000 CFU/l) was required to be present in a water sample near an air sampling site to isolate colonies from aerosols by culture [14]. In this study, the highest concentration of Legionella species in bath water near air sampling sites was not high (8,100 CFU/l in 16/21 tested samples; Kanatani, unpublished observations).
In puddle water, the concentration of Legionella species was not high; the highest concentration of Legionella species in puddle water was 75,200 CFU/l based on the results of our previous study [11]. The reason Legionella species could not be isolated may be related to the stress encountered during aerosolization and the air sampling process, which may have led to a loss of culturability [21]. Montagna et al. reported that viable Legionella cells could not be isolated by the culture method from air samples collected using a Coriolis µ air sampler [<link rid="bib12">12</link>]. Sampling solution may be another factor that in uenced the viability of the Legionella cells collected. Chang et al. reported that the samples collected with deionized water yielded greater recovery of viable L. pneumophila by EMA-qPCR than those collected with the tween mixture [19]. However, bacteria cells may become damaged in deionized waster due to differences in osmotic pressure; this in uences the survival activity of Legionella cells as stated in the manufacturer's instruction of Coriolis µ air sampler. Additional improvements in culture methods, such as strains of amoeba used for co-culture and incubation periods with the amoeba, may also allow the isolation of Legionella from aerosols. Therefore, new sampling and/or culture methods to isolate Legionella species from aerosols need to be established.
It is also necessary to consider the possibility of VBNC Legionella cells. Although various stress factors may induce Legionella cells to enter a VBNC state, these cells can still directly infect human macrophages and amoebae, indicating that VBNC Legionella cells are able to cause disease in humans [22,23]. Several studies have shown that VBNC Legionella in water samples regain culturability in amoebic co-culture [24,25]. Our study also showed that amoebic co-culture increased the amount of Legionella DNA in one air sample collected from a bathroom, but not in any sample collected near roads. To eliminate free DNA released from dead or membrane-impaired cells [21], EMA-qPCR was performed on air samples collected near roads, and Legionella DNA was detected in about 70% of the samples. This indicates that aerosols from roads contain viable Legionella cells. Although Legionella is sensitive to environmental stresses such as desiccation and UV irradiation, amoeba cysts, which are resistant amoebal forms that can survive under stress conditions [26], may play a role in airborne survival and transmission. Further investigations are needed to obtain a detailed understanding of the state of Legionella in aerosols.
It seems plausible that precipitation is positively associated with the occurrence of legionellosis [27]. Our study also showed a weak but positive correlation between the monthly total precipitation and the amount of Legionella DNA present in aerosols collected near roads. According to previous studies, the number of legionellosis cases peaked in July, the second half of the rainy season, in Japan [28,29]. Thus, there is a risk of contracting legionellosis near asphalt roads, especially during the rainy season. The amount of Legionella DNA was also directly and inversely correlated with the daily total precipitation of seven days and one day before sampling, respectively. Our results suggest that Legionella may have multiplied within amoeba present in the environment after precipitation at the seven-day-before sampling time point, whereas the dry condition during the day before the sampling time point may have favored the release of aerosols of small particle sizes from the ground into the atmosphere because of splashing or the effect of wind. Alternatively, the survival rate of Legionella in aerosols may be different due to climatic conditions.
16S rRNA gene amplicon sequencing revealed that Sphingomonas was the most frequently detected genus in the air samples. This genus has been found in various environments such as soil, water, clinical specimens, air, and other locations [30][31][32], indicating the opportunity for the bacteria to be released into the air. Pseudomonas and Achromobacter were signi cantly more abundant in air samples from bathrooms and other indoor sites, respectively. They have frequently been detected in bathwater and moist indoor environments [33,34]. Thus, our results were congruent with those reported in other studies. In this study, reads from Legionella species were detected in 30% of the air samples collected near roads, which was equivalent to those in bathrooms and signi cantly higher than those from other indoor sites.

Conclusions
Here, we demonstrated that DNA from Legionella species, including L. pneumophila, were widely present in aerosols collected from outdoor sites near asphalt roads, especially during the rainy season, regardless of the sampling site. The detection rate and amount of Legionella DNA in the aerosols collected near roads were similar to those in aerosols collected from bathrooms and were signi cantly higher than those collected from other indoor sites. Our ndings suggest that there may be a risk of exposure to Legionella species in the areas surrounding asphalt roads, especially during the rainy season.

Sample collection
We Under these conditions, spiked L. pneumophila was recovered by approximately 60% recovered by culture (Kanatani, unpublished observations). The volume remaining after each sampling was 6-14 mL, and each sample solution was vortexed for 1 min.
Isolation of Legionella spp.
We attempted to detect Legionella species through culture and amoebic co-culture methods. For the culture investigations, 0.1 mL of the sample was spread onto a GVPC agar plate (Nissui Pharmaceutical Co., Tokyo, Japan). The agar plate was incubated at 35 °C for seven days in a humidi ed chamber. Any candidate colonies that were smooth gray with characteristic outward structures that were cut-glass-like or mosaic-like in appearance were viewed under a stereomicroscope with oblique illumination [35] and were sub-cultured on a buffered charcoal-yeast extract (BCYE) agar plate with L-cysteine (bioMérieux, Lyon, France) and a blood agar plate (Eiken Chemical, Tokyo, Japan). Colonies growing only on the BCYE agar plate and not the blood agar were presumed to belong to the genus Legionella.
Amoebic co-culture was carried out as described previously [36]. Acanthamoeba species isolated from cooling tower water were incubated in Proteose peptone-Yeast extract-Glucose-Cysteine medium at 30 °C for 5-7 days. Cells were washed and resuspended in phosphate-buffered saline, and then 0.5 ml of the suspension (approximately 1.0 × 10 5 cells) was added to the remaining amount of the sampling solution (4-12 mL) after culture (0.1 mL) and qPCR (2 mL). To prevent evaporation, the sample was incubated at 35 °C in 50 ml screw-cap tubes. After four weeks, the sample was mixed with equal volumes of 0.2 mol/L KCl-HCl buffer (pH 2.2) for 15 min at room temperature, and 0.2 mL was spread onto a GVPC agar plate. Then, the candidate Legionella colonies were analyzed as described above.

Qpcr
For qPCR, DNA was extracted from 2 mL of sample solution. The suspension was centrifuged at 20,000 × g at room temperature for 5 min and then resuspended in 100 µl of 5% (w/v) Chelex-100 solution (37; Bio-Rad Laboratories, CA, USA). The suspension was boiled at 100 °C for 10 min and then centrifuged at 20,000 × g at room temperature for 5 min. The supernatant was used as the DNA template, and qPCR was carried out using the CycleavePCR Legionella (16S rRNA) Detection Kit (Takara Bio, Shiga, Japan) and a Thermal Cycler Dice Real Time System II (Takara Bio) according to the manufacturer's instructions.

Ema-qpcr
To detect DNA derived from viable Legionella cells, air samples were subjected to EMA-qPCR. For this investigation, another set of 30 samples was collected from eight outdoor sites near asphalt roads. These sampling sites were different from those described in the sample collection subsection (locations A-L), although the tra c volume at the sites was approximately the same as locations A-L. Air samples were collected in 15 mL of sterile deionized water at a ow rate of 300 l/min for 30 min, and distilled water was added every 10 min to bring the sampling solution volume up to 15 mL. Sample solution (2 mL) was centrifuged at 20,000 × g at room temperature for 5 min and then resuspended in 40 µL of sterile deionized water. After treatment with EMA using the Viable Legionella Selection Kit for PCR Ver. 2.0 (Takara Bio) and LED CrossLinker 12 (Takara Bio), the DNA was extracted with Lysis Buffer for Legionella (Takara Bio) [38]. All protocols were carried out according to the manufacturer's instructions. qPCR was carried out as described above in the qPCR subsection.

Sequencing of the mip gene
For samples in which the amounts of Legionella DNA increased after amoebic co-culture, the species of Legionella was determined. DNA was extracted from a 2 ml sample of the amoebic co-culture with Chelex-100 solution as described above. The mip gene was directly ampli ed and sequenced as described previously [39]. Microbiome bioinformatics was performed using QIIME 2 version 2019.7 [40]. Imported demultiplexed sequence data were denoised with DADA2 [41] (via q2-dada2). All ASVs were aligned with mafft [42] (via q2-alignment) and used to construct a phylogeny with fasttree2 [43] (via q2-phylogeny). Taxonomy from kingdom to species was assigned to ASVs using the q2-feature-classi er [44] based on the classifysklearn naïve Bayes taxonomy classi er against the Greengenes 13_8 99% OTU reference sequences [45].

Meteorological Data Collection
During the study period (2016-2018), meteorological data were obtained from two main weather stations in Fushiki and Toyama, Toyama Prefecture, Japan [51]. The meteorological data used in this study were monthly or daily values of air temperature (mean, °C), relative humidity (mean, %), total precipitation (mm), and wind speed (mean, m/s).

Statistical analysis
To investigate the relationships between the rates of Legionella DNA detection and the sampling source type and between the detection rates and sampling site (locations A-L), Fisher's exact test followed by post hoc Holm test were performed. We also investigated the relationships between the amount of Legionella DNA and the sampling source type, and between the amount of Legionella DNA and sampling site (locations A-L) by using the one-way ANOVA test, followed by Tukey-Kramer multiple comparisons. The Student's t-test was performed to compare the amounts of Legionella DNA between samples associated with high and low total precipitation. Fisher's exact test was performed to compare the rates of Legionella DNA detection between samples associated with high and low total precipitation. These tests were performed using the R statistical software package (version 3.0.0). Pearson's correlation coe cients (r) between the climatic conditions (air temperature, relative humidity, total precipitation, and wind speed) and the amount of Legionella DNA were investigated using Microsoft Excel (Microsoft, Tokyo, Japan). An absolute value of Pearson's r ≥ 0.20 was considered indicative of a correlation, and a P value < 0.05 was considered signi cant.

Declarations
Ethics approval and consent to participate Not applicable.

Consent for publication
Not applicable.

Availability of data and materials
The datasets analyzed during the current study are available from the DNA Data Bank of Japan (http://www.ddbj.nig.ac.jp/) under the accession number LC472487 for the mip gene, and DRA008310 and DRA009422 for the 16S rRNA gene amplicons.