Indoor Microbiome, Environmental Characteristics and Asthma Among Junior High School Students in Johor Bahru, Malaysia T

Home , Hyphomicrobiaceae, Rhodomicrobium, Shimwellia

Environment International 138 (2020) 105664

Contents lists available at ScienceDirect

Environment International

journal homepage: www.elsevier.com/locate/envint

Indoor microbiome, environmental characteristics and asthma among junior high school students in Johor Bahru, Malaysia T

Xi Fua,b, Dan Norbäckc, Qianqian Yuanb,d,e, Yanling Lib,d,e, Xunhua Zhub,d,e, Jamal Hisham Hashimf,g, Zailina Hashimh, Faridah Alii, Yi-Wu Zhengj, Xu-Xin Laij, ⁎ Michael Dho Spangfortj, Yiqun Dengb,d,e, Yu Sunb,d,e, a Department of Occupational and Environmental Health, School of Public Health, Sun Yat-sen University, Guangzhou, PR China b Guangdong Provincial Key Laboratory of Protein Function and Regulation in Agricultural Organisms, College of Life Sciences, South China Agricultural University, Guangzhou, Guangdong 510642, PR China c Occupational and Environmental Medicine, Dept. of Medical Science, University Hospital, Uppsala University, 75237 Uppsala, Sweden d Key Laboratory of Zoonosis of Ministry of Agriculture and Rural Aﬀairs, South China Agricultural University, Guangzhou, Guangdong 510642, PR China e Guangdong Laboratory for Lingnan Modern Agriculture, South China Agricultural University, Guangzhou, Guangdong 510642, China f United Nations University-International Institute for Global Health, Kuala Lumpur, Malaysia g Department of Community Health, National University of Malaysia, Kuala Lumpur, Malaysia h Department of Environmental and Occupational Health, Faculty of Medicine and Health Sciences, Universiti Putra Malaysia, UPM, Serdang, Selangor, Malaysia i Primary Care Unit, Johor State Health Department, Johor Bahru, Malaysia j Asia Paciﬁc Research, ALK-Abello A/S, Guangzhou, China

ARTICLE INFO ABSTRACT

Handling editor: Da Chen Indoor microbial diversity and composition are suggested to affect the prevalence and severity of asthma by Keywords: previous home microbiome studies, but no microbiome-health association study has been conducted in a school Bacteria environment, especially in tropical countries. In this study, we collected floor dust and environmental char- Fungi acteristics from 21 classrooms, and health data related to asthma symptoms from 309 students, in junior high Microbial community schools in Johor Bahru, Malaysia. The bacterial and fungal composition was characterized by sequencing 16s Absolute quantity rRNA gene and internal transcribed spacer (ITS) region, and the absolute microbial concentration was quantified Wheezing by qPCR. In total, 326 bacterial and 255 fungal genera were characterized. Five bacterial (Sphingobium, Breathlessness Rhodomicrobium, Shimwellia, Solirubrobacter, Pleurocapsa) and two fungal (Torulaspora and Leptosphaeriaceae) Adolescence taxa were protective for asthma severity. Two bacterial taxa, Izhakiella and Robinsoniella, were positively as- Dampness/visible mold Malaysia sociated with asthma severity. Several protective bacterial taxa including Rhodomicrobium, Shimwellia and Johor Bahru Sphingobium have been reported as protective microbes in previous studies, whereas other taxa were first time Tropics reported. Environmental characteristics, such as age of building, size of textile curtain per room volume, oc- Junior high school currence of cockroaches, concentration of house dust mite allergens transferred from homes by the occupants, were involved in shaping the overall microbial community but not asthma-associated taxa; whereas visible dampness and mold, which did not change the overall microbial community for floor dust, was negatively associated with the concentration of protective bacteria Rhodomicrobium (β = −2.86, p = 0.021) of asthma. The result indicates complex interactions between microbes, environmental characteristics and asthma symptoms. Overall, this is the first indoor microbiome study to characterize the asthma-associated microbes and their environmental determinant in the tropical area, promoting the understanding of microbial exposure and respiratory health in this region.

1. Introduction characteristics have been suggested to be associated with asthma prevalence and severity, including parental asthma, preterm delivery and Asthma prevalence has been rising globally in the past few decades low birth weight, tobacco smoking, respiratory syncytial virus infec- (Eder et al., 2006). Many personal and environmental and tion, exposure to air/traﬃc pollution and indoor mold (de Benedictis

⁎ Corresponding author at: Guangdong Provincial Key Laboratory of Protein Function and Regulation in Agricultural Organisms, College of Life Sciences, South China Agricultural University, Guangzhou, Guangdong 510642, PR China. E-mail address: [email protected] (Y. Sun). https://doi.org/10.1016/j.envint.2020.105664 Received 11 December 2019; Received in revised form 12 March 2020; Accepted 12 March 2020 Available online 19 March 2020 0160-4120/ © 2020 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/BY-NC-ND/4.0/). X. Fu, et al. Environment International 138 (2020) 105664 and Bush, 2017; Castro-Rodriguez et al., 2016). However, many of sequencing, and thus only dust samples of 21 classes could be se- these risk factors fail to explain the increasing trend of asthma epi- quenced. The numbers of success and failed samples in each school demics. For example, smoking is a strong risk factor for asthma, but the were listed in Table S1. Health data were collected by self-reported prevalence of smoking is significantly reduced in the past thirty years questionnaires from 15 randomly selected students in each class. The (Ng et al., 2014). Recent studies suggested that the changing of lifestyle ethical permission was approved by the Medical Research and Ethics and microbial exposure during the industrialization and urbanization Committee of the National University of Malaysia, and all participants process are associated with the increasing prevalence of asthma gave their informed consent. symptoms (Bello et al., 2018). Nowadays, more people live in the city than the rural area, and they spend most of the time in the indoor 2.1. Assessment of health data environment (Klepeis et al., 2001), thus it is necessary to identify the beneficial and risk exposure in various indoor environments. Progress Questions about doctor-diagnosed asthma and current asthma were in culture-independent microbiome studies reveals the association be- obtained from the European Community Respiratory Health Study tween indoor microbial exposure and human respiratory health in the (ECRHS). The questions included asthma symptoms and related in- home environment. It was reported that high bacterial richness in formation during last 12 months, including wheeze, breathlessness homes of the traditional farm area protected against childhood asthma during wheeze, feeling of chest tightness, shortness of breath during compared with urban families (Ege et al., 2011). Similarly, a high di- rest, shortness of breath during exercise, woken by attack of shortness versity of fungal exposure is protective for childhood asthma develop- of breath, ever had asthma, attack of asthma, and current asthma ment (Dannemiller et al., 2014). There are also studies suggest that the medication use. asthma prevalence is related to the abundance of specific taxa rather A validated asthma score, including eight items, were calculated to than microbial richness (Kirjavainen et al., 2019). For example, two measure asthma severity (Pekkanen et al., 2005) were calculated, and microbiome studies used absolute quantification approaches identified then re-defined as 0, 1, 2, > =3. Questions about current smoking and only one protective or risk microbe for asthma symptoms (Pekkanen parental asthma/allergy were also included. Details about the questions et al., 2018; Dannemiller et al., 2016), and one study in the United were described in a previous study (Norback et al., 2014). States using relative abundance from 16s rRNA identified a few hun- dreds of potentially associated microbes for inner-city children 2.2. Dust sampling and building inspection (O'Connor et al., 2018). Thus, although there are still some dis- crepancies among studies, the association between indoor microbial The detailed dust sampling procedure was reported in a previous exposure and asthma development in the home environment is gen- publication (Norback et al., 2014). Floor dust in the classroom was erally established. collected by a 400 W vacuum cleaner with a dust sampler (ALK Abello, In contrast to the extensive researches in the private home en- Copenhagen, Denmark) through a Millipore filter (pore size 6 µm). The vironment, no study has been conducted in public indoor environments, filter is made of cellulose acetate, which retains 74% of particles of such as schools. Thus, the health effect of microbial assemblage in these 0.3–0.5 mm, 81% of particles of 0.5–1.0 mm, 95% of particles of indoor areas is unclear. Also, these microbiome studies in the home 1–10 mm and 100% of larger particles (> 10 mm). Three dust samples environment are mainly focused on childhood asthma (Ege et al., 2011; were collected at the same time for each classroom. The total vacuum Dannemiller et al., 2014; O'Connor et al., 2018), and the health effect of sampling procedure for each sample lasted 4 min, 2 min on the floor indoor microbial exposure to other age groups, such as adolescents and and 2 min on other surfaces above the floor like chairs and desks. The adults, is not clear. In addition, current home microbiome health as- floor area range from 39 to 82 m2 with a mean of 69 m2. Each class- sociation studies are all conducted in middle and high latitude regions, room was divided into the corridor part and window part, which were mainly from developed countries in Europe and the United States. It has sampled separately as two samples for allergen analysis in the previous been indicated that the indoor microbial composition is geographically publication. The remaining dust was then sieved in the lab, through a patterned, and significant compositional variations can be detected 0.3-mm mesh screen to fine dust, and was stored in the freezer at across different climates, latitudes and geographic regions (Barberan −80˚C. In this study, dust from the two parts of the classroom was et al., 2015; Amend et al., 2010). Thus, the associated-microbes iden- combined together for the amplicon sequencing and quantitative PCR. tified in middle and high latitude provides little implication regarding A third sample was collected by repeating the same procedure for the the microbial exposure and health in the tropical area. Overall, it is whole classroom, and the dust was used to quantify the house dust mite necessary to conduct more microbiome- health association studies and cockroach allergens at the ALK laboratory in Guangzhou. En- covering different geographic regions, age groups and indoor environ- vironmental characteristics, including relative humidity, indoor CO2 ments. and outdoor NO2 concentration, textile curtain factor, the concentra- We conducted a few previous epidemic studies in schools of tion of house dust mite and cockroach allergen, were measured, and Malaysia and identified a list of common risk factors for asthma, in- information about the construction year, visible dampness and mold cluding furry pet allergens, indoor dampness and endotoxins (Norback were noted (Norback et al., 2014). The textile curtain factor was de- et al., 2014, 2017; Cai et al., 2011). But no studies investigated the fined as the area of textile curtain per room volume. The environmental indoor microbiome and the association with asthma symptoms. characteristics showed no difference between the successfully amplified In this study, we conducted the first microbiome survey in a junior samples and those failed in amplification. high school of Johor Bahru, Malaysia, to screen protective and risk microbes associated with asthma symptoms. The absolute concentra- 2.3. DNA extraction and sequencing tion of bacterial and fungal taxa in floor dust from 21 randomly selected classrooms were characterized by amplicon sequencing and quantita- DNA extraction and multiplex sequencing services were provided by tive PCR. Association between microbial taxa, environmental char- GENEWIZ, Suzhou lab. Total genome DNA was extracted for amplicon acteristics and prevalence of asthma were analyzed. sequencing by Soil DNA Kit for all dust samples with bead beating and spin filter technology. Negative controls were added to avoid mass 2. Materials and methods contamination in the amplification process. DNA quality and concentration were evaluated with a NanoDrop One spectrophotometer. Floor dust was collected from 8 junior high schools in Johor Bahru, Amplicons were generated by primers targeting the v3 and v4 regions Malaysia, 4 classes in each school. In total, 32 dust samples were col- on the 16s ribosome RNA (16s rRNA) gene for bacteria, and internal lected, but 11 of them failed to amplify adequate DNA for amplicon transcript space 2 (ITS2) region for fungi. The FASTQ formatted raw

2 X. Fu, et al. Environment International 138 (2020) 105664 sequence data were uploaded in QIITA with accession number 12875 Table 1 (https://qiita.ucsd.edu/study/description/12875). Two separate ex- Prevalence of asthma symptoms and asthma score. tractions of 10 mg dust were conducted for real-time PCR to quantify Symptoms Number (n = 309) Prevalence (%) the absolute concentration of total bacteria and fungi in the dust. The SYBR Green method was used for universal bacterial detection. A 20 µl Wheeze and breathlessness during wheeze 20 6.56 reaction mixture containing 10 µl of Master Mix (Hieﬀ™ qPCR Feeling of chest tight 16 5.18 ® Attack of shortness of breath during rest 28 9.1 SYBR Green Master Mix), 2 µl of template DNA, 0.5 µl of each primer Attack of shortness of breath during 114 36.9 (forward: 5′-GCAGGCCTAACACATGCAAGTC-3′ and reverse: 5′-CTGC exercise TGCCTCCCGTAGGAGT-3′)(Nadkarni et al., 2002). Quantitative PCR Woken by attack of shortness of breath 23 7.4 for fungal DNA was described in a previous publication (Norback et al., Ever asthma 39 12.6 2016). Attack of asthma 7 2.3 Current medication for asthma 11 3.6

2.4. Bioinformatics analysis and statistics Asthma score* 0 148 47.9 1 101 32.7 The forward and reverse reads were joined and assigned to samples 2 28 9.1 by barcoding information, and the quality filter was set as sequence > =3 32 10.3 length > =200 bp. The sequences were then assigned to operational “ ” taxonomic units (OTUs) with a sequence similarity of 97% and anno- * Firstly, an asthma score 8 was calculated by summarizing the eight items listed above. The displayed asthma score is re-defined from “asthma score 8” as tated against Silva and Unite database, respectively. Principle compo- 0, 1, 2, and > =3. nent analysis (PCoA) and Adonis analysis were performed to assess the influence of environmental characteristics to microbial richness and a few classrooms and have more restricted distributions compared to composition based on Bray-Curtis distance matrix. Analyses for mi- bacterial taxa. crobiome dataset were mainly conducted with the Quantitative Insights The major phylum included Proteobacteria (35.0 ± 7.3%, mean Into Microbial Ecology (QIIME, v1.8.0) platform (Caporaso et al., 2010; and standard deviation), Actinobacteria (21.2 ± 6.3%), Cyanobacteria Lawley and Tannock, 2017). Two-level hierarchical ordinal regression (17.6 ± 7.3%) and Firmicutes (17.3 ± 8.6%; Fig. S2 and Table S3). models were performed to analyze associations between microbial The top genus mainly included environmental taxa such as Bacillus richness (number of observed taxonomic units, OTUs) and asthma (4.2 ± 5.8%), Paracoccus (3.2 ± 1.2%), Sphingomonas (2.8 ± 0.9%) score, and between the quantity of single bacterial or fungal phylum, and Saccharopolyspora (2.5 ± 2.0%), and human skin taxa Staphylo- class, and genus (in log10 format) and asthma score. The latter analysis coccus (3.4 ± 2.5%; Fig. 1C and Table S4). Distinct bacterial compo- only included microbial taxa presented in at least four classrooms. In all sitional variation has been observed even for samples collected from the analyses, gender, race, smoking, and parental asthma/allergy were in- same school. For example, Bacillus accounted for 24.3% of the total cluded as adjustment. Parallel line assumption test was performed for bacterial load in classroom No. 1 of school No. 3 (S3C1), whereas ac- the ordinal regression models, and those violated the parallel line as- counted for only 2.4%, 1.4% and 1.5% of total loads in the other three sumption were calculated in a multi-nominal regression model. Asso- classrooms in school No. 3 (Fig. 1C). The fungal phylum was dominated ciation of single environmental characteristics with asthma score were by Ascomycota (72.5 ± 11.6%), followed by Basidiomycota assessed by a hierarchical ordinal regression model. The environmental (17.8 ± 8.9%; Fig. S3, Table S5). The top fungal genus included characteristics potentially related to asthma (p < 0.1) were further common mold taxa such as Aspergillus (16.6 ± 7.9%), Penicillium analyzed with asthma-associated microbes in multivariate linear re- (10.2 ± 15.7%) and Cladosporium (7.8 ± 8.2%), as well as outdoor gression models with a forward stepwise method. All hierarchical environmental fungi such as Hortaea (8.0 ± 7.5%), Wallernia models and parallel line test was conducted by StataSE 15.0 (StataCorp (6.6 ± 9.2%) and Emericella (4.1 ± 9.6%) (Fig. 1D, Table S6). As- LLC), and other statistics were conducted with IBM SPSS software 21.0 pergillus presented in high abundance (> 9%) in all samples. But large- (IBM). Association analysis between environmental characteristics and scale variations were detected for some other genera. For example, microbial richness and community variation were conducted by Adonis Penicillium accounted for 48.1% in S5C2, but less than 5% in S5C1 and in R v3.3. S5C3 (Fig. 1D). The compositional variation of bacterial and fungal community in all samples are illustrated by the principal coordinate 3. Results analysis (PCoA) (Fig. S4A and S4B).

The response rate of the questionnaire was 96% (n = 309). The students were aged from 14 to 16 years, and 52% were girls. The ethnic 3.2. Environmental characteristics associated with overall microbial group of the participants were Malay (43%), Chinese (42%), and Indian richness/composition (15%). The detailed demographic data were described in previous studies (Norback et al., 2014; Cai et al., 2011). The prevalence of In this study, we collected eight environmental characteristics and asthma symptoms and asthma score is presented in Table 1. tested their association with overall microbial richness (Table S7). A high concentration of house dust mite and cockroach allergens and high 3.1. Sequencing statistics and microbial taxa textile curtain factor were negatively associated with the number of fungal observed OTUs (Adonis, p < 0.05; Table S7). No environmental The bacterial 16s rRNA dataset was rarefied to the depth of 27,000 characteristic had a significant association with bacterial richness. reads for each sample, and fungal ITS was rarefied to 35,000 reads. The The overall bacterial community composition was affected by the rarefaction curves indicate that the sequencing depth is deep enough to concentration of house dust mite allergens and textile curtain factor in capture the majority of operational taxonomic units (OTUs) in the floor the classroom; the fungal community was affected by the concentration dust (Fig. S1). In total, 895 bacterial and 1512 fungal OTUs were ob- of house dust mite allergens and cockroach allergens, and age of the tained, and distinct distribution patterns were observed. For bacteria, building (Adonis, p < 0.05; Table 2). We visualize the abundance 36.8% of bacterial OTUs were presented in all samples, whereas only variation of the top bacterial and fungal genera affected by these en- 10.3% of fungal OTUs were presented in all samples and approximately vironmental characteristics. Classrooms with a high textile curtain half of the OTUs were presented in ten or fewer samples (Fig. 1A and B). factor had a higher abundance of bacterial genera Enterococcus, The result suggests that, in floor dust, many fungal taxa are presented in Chroococcidiopisis and an unidentified genus from Nostocaceae, and

3 X. Fu, et al. Environment International 138 (2020) 105664

Fig. 1. (A) Frequency spectrum of bacterial OTU presence in samples; (B) Frequency spectrum of fungal OTU presence in samples. Taxonomic composition of (C) bacteria and (D) fungi at the genus level for all samples. Each sample is labelled with school and class numbers.

4 X. Fu, et al. Environment International 138 (2020) 105664

Table 2 results (Dannemiller et al., 2014; Vandeputte et al., 2017). However, Association between outdoor/indoor characteristics and microbial community very few studies compare the absolute abundance with relative abun- a variation. dance in indoor microbiome survey studies. We conducted the asso- Bacteria Fungi ciation analysis between relative abundance and asthma score with the same regression model and found drastic variations between the two Environmental Characteristics R2 pR2 p approaches. Among the seven associated microbes identified by absolute quantification, only three were identified by the relative abun- Relative humidity 0.05 0.39 0.06 0.25 dance approach. Two additional genera, including Wolbachia and No- Indoor CO2 0.02 0.91 0.02 0.99 fi Outdoor NO2 0.09 0.06 0.04 0.6 cardiopsis, were identi ed by the relative approach (Table S9). Building age 0.06 0.29 0.1 0.03 Similarly, only one fungal genus was identified by the relative abun- Visible indoor dampness/mold 0.03 0.84 0.03 0.81 dance (Table S10). Textile curtain factorb 0.16 0.005 0.09 0.06 Concentration of house dust mite allergen in dust 0.2 0.004 0.15 0.009 Concentration of cockroach allergen in dust 0.05 0.4 0.17 0.04 3.4. Environmental characteristics associated with protective microbes

a The calculation is based on Bray-Curtis distance (beta diversity). P-value We investigated the associations between the environmental char- was calculated based on 10,000 permutation bivariate Adonis analysis. acteristics and the protective or risk microbes of asthma and found that Associations with p < 0.05 are formatted with bold font. although the indoor dampness/visible mold was not a significant b The textile curtain factor was defined as the area of textile curtain per room characteristic changing the overall indoor microbial composition for volume. settled dust, it was negatively associated with the concentration of protective microbes, including Rhodomicrobium (β = −2.86, classrooms with a low textile curtain factor had a higher abundance of p = 0.021) and Solirubrobacter (β = −1.62, p = 0.07). Thus, high Deinococcus, Kocuria, Rubellimicrobium, and Paracoccus (Kruskal-Wallis indoor dampness can not only increase the prevalence of asthma by test, p < 0.05; Fig. 2A). Classrooms with a low concentration of house releasing submicron-sized cellular fragments and Microbial Volatile dust mite allergens (which could be transferred from homes) had a Organic Compounds (MVOCs) (Nevalainen et al., 2015), it could also significantly higher abundance of bacterial genera Deinococcus, Kocuria, affect the respiratory health of occupants by suppressing the abundance Acinetobacter, and a fungal genus Candida and an unidentified fungal of protective bacteria of asthma. To our knowledge, this is a new genus from Pleosporales (p < 0.05; Fig. 2B and 2C). The classrooms finding. with no cockroach allergen detected had a higher abundance of an unidentified fungal genus from Pleosporaceae (Fig. 2D). The abundance 4. Discussion of common mold genera, like Aspergillus, Cladosporium, Penicillium, Stschybotrys, were not changed by these factors (Fig. 2). We identified 326 bacterial and 255 fungal genera from the indoor floor dust in seven junior high schools of Johor Bahru, Malaysia. Seven 3.3. Identifying protective and risk microbes for asthma microbial taxa were quantitatively negatively associated with asthma severity, and two microbial taxa were positively associated. Visible The prevalence of asthma symptoms is presented in Table 1. Asthma indoor dampness and mold were not involved in shaping the overall score was calculated based on all eight items to represent the severity of microbial composition but were negatively associated with the con- asthma. There were no associations between the number of OTUs centration of protective bacteria. within the major phylum and class and asthma severity, suggesting microbial richness in the indoor environment was not significantly af- 4.1. Advantages and limitations of the study fecting asthma symptoms (Table S8). Although Proteobacteria (95% CI 0.92–1.01, p = 0.09) and Cyanobacteria (0.95–1.0, p = 0.07) showed This is the first study to investigate the association between bac- marginally protective associations with the asthma score. terial and fungal taxa and asthma symptoms in a tropical region. The We screened associations between the absolute quantity of bacterial study applied culture-free high-throughput sequencing and quantitative and fungal genus and asthma score with a hierarchical ordinal regres- PCR to characterize the absolute concentration of microbial exposure in sion model. Microbes presented in less than five classrooms were not the classroom environment for adolescents in junior high schools. We included in the analyses, and in total 284 bacteria and 202 fungi genera systematically investigated the associations between microbial ex- were examined in the association analyses. P-value < 0.01 was set as a posure and asthma severity, and environmental characteristics, re- cutoff to screen associated microbes. Five bacterial genera were nega- vealing the complex relationship between these factors. There are also tively associated with asthma score (p < 0.01), including Sphingobium, some limitations in our study. Only floor dust was collected and eval- Rhodomicrobium and Shimwellia in Proteobacteria, Solirubrobacter in uated in this study. The active sampling of airborne dust is expected to Actinobacteria, Pleurocapsa in Cyanobacteria. Two genera, including be the most direct way to evaluate the inhalable microbial exposure for Izhakiella in Proteobacteria and Robinsoniella in Firmicutes, were posi- occupants. However, the approach is relatively expensive, and also, the tively associated with asthma score (Table 3). Two fungal genera in microbial composition can vary temporally. Thus, the air sampling Ascomycota phylum were negatively associated with asthma score generally represents a short-term microbial exposure compared with (p < 0.01) (Table 3), including Torulaspora, and an unidentified genus settled dust sampling strategies. We used amplicon sequencing to in Leptosphaeriaceae family. The model for Robinsoniella and Lepto- characterize microbial composition in settle dust. Due to the technical sphaeriaceae violated the parallel line test (p < 0.01), and the asso- limitation of amplicon sequencing, we can only identify the microbes ciations were then assessed by multi-nominal regression. Associations down to the genus level, rather than more taxonomically resolved were observed between Robinsoniella and asthma score 0 to 1 species and strain level. It is common that species within a genus or (RRR = 1.34, p < 0.0001) and 0 to 2 (RRR = 1.39, p = 0.006), and even strains from the same species could have different virulent factors, between Leptosphaeriaceae and asthma score 0 to 1 (RRR = 0.51, thus posing different health effect for human. Thus, more taxonomically p = 0.0001). resolved technique, such as shotgun metagenomics, will improve the Recent studies claimed that absolute abundance approaches for identification accuracy for future indoor microbiome survey. We microbial quantification should be used to link correct microbes to identified several genera quantitatively associated with an asthma phenotypes and quantitative features, and the relative abundance ap- score, but due to the limitation of the cross-sectional study design, we proach can produce some erroneous identification and false-positive can only report the association instead of a conclusion of a causal effect.

5 X. Fu, et al. Environment International 138 (2020) 105664

Fig. 2. Relative abundance of bacterial and fungal genera in different environmental factors. The abundance of bacterial genera in different (A) textile curtain factor and B) concentration of house dust mite allergens. The abundance of fungal genera in different (C) concentration of house dust mite allergens and (D) the presence of cockroach allergens. Only genera with relative abundance differences > 0.5% are plotted. Error bars represent the standard error, and a Kruskal-Wallis test was conducted to calculate p values (*** p < 0.001, ** p < 0.01, * p < 0.05).

Further longitudinal studies are needed to disentangle the causal effect environment, which had protective effect for asthma symptoms and temporal dynamic of the indoor microbiome. (Kirjavainen et al., 2019). The consistent result in tropical and Eur- opean countries indicates a possible universal protective effect of these taxa across large geographic regions and in various climate conditions. 4.2. Bacterial and fungal taxa associated with asthma score The positively associated microbes identified in this study were not reported to be associated with respiratory health in previous studies. It In our study, we observed five bacterial genera protectively asso- is possible that the presence of these microbes is geographically re- ciated with asthma score (p < 0.01) among students, including stricted in tropical areas. Izhakiella is a newly identified genus, and Sphingobium, Rhodomicrobium, Shimwellia in Proteobacteria, recently isolated form mired bug and Australian desert soil (Ji et al., Solirubrobacter in Actinobacteria, Pleurocapsa in Cyanobacteria. The 2017). The genus of Robinsoniella belongs to the class of Clostridia. We protective effect of these taxa has been previously reported in a mi- found no research articles about this genus, but many other taxa in crobiome study from the farm and non-farm rural homes in Finland and Clostridia class associated with asthma and human health. For example, Germany. In this study, the relative abundance of family Clostridium cluster XI in the home environment was shown to be pro- Hyphomicrobiaceae (including Rhodomicrobium), Enterobacteriaceae tectively associated with asthma prevalence among adults (Pekkanen (including Shimwellia), Sphigomonadaceae (including Sphingobium), the et al., 2018). Several families in Clostridia, including Phascolarcto- class Thermoleophilia (including Solirubrobacter), and the phylum bacterium, Mogibacterium and Proteiniclasticum, were more abundant Cyanobacteria (including Pleurocapsa) were higher in the farm home

6 X. Fu, et al. Environment International 138 (2020) 105664

in rural farm homes, where there were lower asthma prevalence and healthier indoor microbiomes as compared to non-farm rural homes p value (Kirjavainen et al., 2019). In addition, Clostridium butyricum was suggested to be a potential therapeutic microbe combined with specific ff 0.724) 0.004 0.939) 0.002 0.813) 0.002 1.342) 0.006 0.923) 0.003 0.788) 0.005 1.326) 0.004 0.870) 0.004 0.893) 0.006 immunotherapy for asthma treatment. However, the harmful e ect of – – – – – – – – – * Clostridia taxa has also been reported, such as Clostridium difficile,

and Leptosphaeriaceae, which can cause severe diarrhea to life-threatening colitis (Smits et al., 2016; Borali and De Giacomo, 2016). Moreover, we tested the health 0.368 (0.187 OR (95% CI) 0.837 (0.749 0.560 (0.385 1.187 (1.051 0.787 (0.672 0.465 (0.275 1.182 (1.054 0.643 (0.476 0.672 (0.505 association for some previously reported pathogen genera. A pathogen genus Bacillus was detected in the classroom microbiome with a high abundance, but there was no association with asthma severity. A Robinsoniella common nosocomial pathogen genus Staphylococcus was also common in the classroom, but with no health association observed. Previous culture-dependent studies have reported Aspergillus, Penicillium, Alternaria and Cladosporium as risk fungal taxa for asthma symptoms (Sharpe et al., 2015). However, very few studies apply culture-independent approaches to systematically screen fungal microbes for asthma symptoms. We examined the health associations for the mold genera in classrooms, and there were no associations for asthma severity. Dannemiller et al. identified that Volutella was positively and Kondoa was protectively associated with asthma severity among atopic 21 Number of classrooms with(N presence = 21) 17 21 5 19 21 13 7 7 and nonatopic children in the Northeast of the United States (Dannemiller et al., 2016). In this study, we identified two protective fungi taxa, Torulaspora and an unidentified genus from the family of Leptosphaeriaceae. In a previous study, Torulaspora delbrueckii has been shown to have probiotic potential that can be used as a supplement in food production to regulate intestine response and promote human health (Zivkovic et al., 2015). The family of Leptosphaeriaceae has not been previously reported to be associated with human health. 0.15 1.65 0.26 0.13 0.13 0.04 0.11 0.01 0.12 – – – – – – 0.02 – 0 Range of relative abundance (%) 0.01 – 0 0 0.01 – 0 0 0 4.3. Indoor dampness/mold affected protective bacteria for asthma cant. The ORs displayed here are all from the ordinal regression model. Taxonomic information of the fi Among the eight environmental characteristics examined in this study, only indoor dampness/visible mold was associated with asthma- related genus. Indoor dampness/visible mold was negatively associated with bacterial genera protective to asthma severity, Solirubrobacter and Rhodomicrobium, which is new to our knowledge. Dampness and mold have been proved as risk factors for respiratory health, including asthma (Castro-Rodriguez et al., 2016; Quansah et al., 2012). A recent study has reported that indoor dampness and mold increase the onset of asthma symptoms and reduce remission from asthma among adults (Wang et al., 2019). Previous studies on dampness and mold in buildings established the direct association between mold species, fungal cellular fragment and MVOCs and related airway inflammation Absolution abundance GM ± GSD (copies / g 156 ± 1.4 53 ± 7.5 dust) 167 ± 2.0 3 ± 8.0 66 ± 4.1 140 ± 1.6 17 ± 9.4 2 ± 2.2 2 ± 2.5 (Nevalainen et al., 2015; An and Yamamoto, 2016; Zhang et al., 2016). Our results suggest that, despite the direct harmful effect from fungi, mold growth may suppress the concentration of beneficial bacteria that

ed ﬂ

ﬁ are protective for asthma symptoms. It has been reported that in oor dust, the absolute concentration of most mold taxa, including Asper- gillus, Penicillium and Alternaria, keeps increasing with elevated hu- Genus Sphingobium Rhodomicrobium Shimwellia Izhakiella Solirubrobacter Pleurocapsa Robinsoniella Torulaspora g_unidenti midity, whereas the concentration of speciﬁc bacterial taxa, including Pasteurellaceae, Prevotella and Cytophaga, decreases with elevated humidity (Dannemiller et al., 2017). However, as the dynamics of fungal and bacterial growth is a complex issue, the detailed interaction among cance level: p < 0.01.

ﬁ microbes are still unclear. As the majority of fungal and bacterial species are non-culturable, new study designs such as shotgun metagenomic sequencing strategy combined with in silico growth rate analysis, Gammaproteobacteria Dothideomycetes f_Leptosphaeriaceae such as tools like GRiD (Emiola and Oh, 2018), holds a promising so- lution for the issue.

4.4. Absolute and relative quantiﬁcations in microbiome phenotype association analysis Actinobacteria Thermoleophilia Cyanobacteria Oxyphotobacteria Firmicutes Clostridia

In our study, we used absolute taxonomic quantities to assess the associations with asthma score, while some previous studies used re- * The associations were calculated in ordinal regression models. Sex, race, smoking, and parental asthma or allergy were adjusted for in the association analyses. The microbes Bacteria Proteobacteria Alphaproteobacteria Kingdom Phylum Class Fungi Ascomycota Saccharomycetes Table 3 Microbial taxa associated with asthma severity status in junior high school. which failed for parallelassociated line microbes test, are were listed. Signi tested by nominal regression model, and the associations were signi lative abundance (Kirjavainen et al., 2019; O'Connor et al., 2018). An

7 X. Fu, et al. Environment International 138 (2020) 105664 issue of relative abundance is that the abundance change of one of the original draft, Writing - review & editing. Dan Norbäck: taxa will lead to abundance changes of all other taxa, which may lead to Conceptualization, Methodology, Writing - review & editing. Qianqian over-identification or misidentification. O’Connor et al. have identified Yuan: Data curation, Formal analysis, Visualization. Yanling Li: Data 201 risk and 171 protective microbes from household dust associated curation, Formal analysis, Visualization. Xunhua Zhu: Investigation. with prevalent asthma among children (O'Connor et al., 2018). Among Jamal Hisham Hashim: Resources. Zailina Hashim: Resources. these microbes, some common human skin bacteria have been identi- Faridah Ali: Data curation. Yi-Wu Zheng: Investigation. Xu-Xin Lai: fied as risky microbes for childhood asthma, including Staphylococcus, Investigation. Michael Dho Spangfort: Investigation. Yiqun Deng: Corynebacterium, Haemophilus and Sphingomonas, but they are unlikely Funding acquisition, Project administration. Yu Sun: to be risk agents since these taxa are universally present around human Conceptualization, Writing - review & editing, Visualization, occupants. One study on quantification profiling of gut microbiome has Supervision, Funding acquisition. reported that absolute abundance of microbes significantly differs from the rank by relative abundance, which affects the result of association Declaration of Competing Interest analysis for phenotypes (Vandeputte et al., 2017). In this study, only 3 fi bacteria identi ed by the relative approach were consistent with the The authors declare that they have no known competing financial fi absolute quanti cation approach, indicating the discrepancy between interests or personal relationships that could have appeared to influ- the two approaches. From our results, we observed that most the ence the work reported in this paper. identified taxa related to health were low-frequency taxa with a relative abundance < 0.2% in all samples (Tables 3). The relative abundance of Acknowledgements these low-frequency microbes can be impacted by the variation of dominant microbes, and their concentration is more appropriately We thank South China Agricultural University and Department of presented by the absolute approach. Education of Guangdong Province (2018KTSCX021) for financial sup- port. 4.5. Abundance and distribution of fungal taxa Appendix A. Supplementary material In this study, we found that Ascomycota is the most abundant fungal fl phylum in oor dust. The results are consistent with several previous Supplementary data to this article can be found online at https:// fungal microbiome composition surveys. For example, a global survey doi.org/10.1016/j.envint.2020.105664. across multiple continents revealed that Ascomycota is the most abundant fungal phylum in various indoor environments (Amend et al., References 2010). Another intensive sampling of 1200 households in the United States revealed that Ascomycota, including Cladosporium, Tox- Adams, R.I., Miletto, M., Taylor, J.W., Bruns, T.D., 2013. Dispersal in microbes: fungi in icocladosporium and Alternaria, dominated the indoor environment. indoor air are dominated by outdoor air and show dispersal limitation at short dis- However, all of these results were based on dust analysis from floor, tances. ISME J. 7 (7), 1262–1273. Amend, A.S., Seifert, K.A., Samson, R., Bruns, T.D., 2010. Indoor fungal composition is door-frame or passive air dust collection with petri-dish. A recent air- geographically patterned and more diverse in temperate zones than in the tropics. borne microbiome study in Singapore revealed that Basidiomycota is PNAS 107 (31), 13748–13753. the most dominant phylum in the near-surface atmosphere, and its An, C., Yamamoto, N., 2016. Fungal compositions and diversities on indoor surfaces with visible mold growths in residential buildings in the Seoul Capital Area of South Korea. community structure was temporally stable (Gusareva et al., 2019). The Indoor Air 26 (5), 714–723. claim was supported by another study with active sampling for the Barberan, A., Dunn, R.R., Reich, B.J., Pacifici, K., Laber, E.B., Menninger, H.L., et al., Amazon rainforest air (Womack et al., 2015). However, as the active air 2015. The ecology of microscopic life in household dust. Proc. Biol. Sci. 282(1814). ff sampling is much more expensive than the floor dust sampling, these Barberan, A., Ladau, J., Le , J.W., Pollard, K.S., Menninger, H.L., Dunn, R.R., et al., 2015. Continental-scale distributions of dust-associated bacteria and fungi. Proc. Natl. two studies collected samples in only one site, and thus the geographic Acad. Sci. USA 112 (18), 5756–5761. distribution of Basidiomycota taxa is unclear. In this study, we found Bello, M.G.D., Knight, R., Gilbert, J.A., Blaser, M.J., 2018. Preserving microbial diversity. – that, in floor dust, the fungal taxa had more restricted geographic dis- Science 362 (6410), 33 34. Borali, E., De Giacomo, C., 2016. Clostridium difficile infection in children: a review. J. tribution compared with bacterial taxa. Two studies that sampled settle Pediatr. Gastroenterol. Nutr. 63 (6), e130–e140. dust also reported the same pattern y (Adams et al., 2013; Barberan Cai, G.H., Hashim, J.H., Hashim, Z., Ali, F., Bloom, E., Larsson, L., et al., 2011. Fungal et al., 2015). Future studies incorporating active air sampling with DNA, allergens, mycotoxins and associations with asthmatic symptoms among pupils in schools from Johor Bahru, Malaysia. Pediatric Allergy Immunol.: Off. Publ. Eur. geographic distribution are needed to address whether the pattern still Soc. Pediatric Allergy Immunol. 22 (3), 290–297. holds for airborne Basidiomycota. Caporaso, J.G., Kuczynski, J., Stombaugh, J., Bittinger, K., Bushman, F.D., Costello, E.K., et al., 2010. QIIME allows analysis of high-throughput community sequencing data. Nat. Methods 7 (5), 335–336. 5. Conclusion Castro-Rodriguez, J.A., Forno, E., Rodriguez-Martinez, C.E., Celedon, J.C., 2016. Risk and protective factors for childhood asthma: what is the evidence? J. Allergy Clin. In this study, we reported a list of low frequency bacterial and Immunol. Pract. 4 (6), 1111–1122. Dannemiller, K.C., Mendell, M.J., Macher, J.M., Kumagai, K., Bradman, A., Holland, N., fungal taxa that were associated with asthma severity in classrooms in et al., 2014. Next-generation DNA sequencing reveals that low fungal diversity in junior high school students, in Johor Bahru, Malaysia. Environmental house dust is associated with childhood asthma development. Indoor Air 24 (3), characteristics associated with the overall microbial community were 236–247. not associated with the protective or risk microbes, but indoor damp- Dannemiller, K.C., Gent, J.F., Leaderer, B.P., Peccia, J., 2016. Indoor microbial communities: Influence on asthma severity in atopic and nonatopic children. J. Allergy Clin. ness/ visible mold, which did not associate with microbial community Immunol. 138 (1), 76–83.e1. variation, was associated with the asthma-associated bacteria. This is Dannemiller, K.C., Weschler, C.J., Peccia, J., 2017. Fungal and bacterial growth in floor – the first study to reveal the complex interaction between the micro- dust at elevated relative humidity levels. Indoor Air 27 (2), 354 363. de Benedictis, D., Bush, A., 2017. Asthma in adolescence: is there any news? Pediatr. biome, environmental characteristics and asthma symptoms in a tro- Pulmonol. 52 (1), 129–138. pical area. The study contributes to new knowledge on how to promote Eder, W., Ege, M.J., von Mutius, E., 2006. The asthma epidemic. New Engl. J. Med. 355 – the establishment of a healthy building microbiome in this region. (21), 2226 2235. Ege, M.J., Mayer, M., Normand, A.C., Genuneit, J., Cookson, W.O., Braun-Fahrlander, C., et al., 2011. Exposure to environmental microorganisms and childhood asthma. New CRediT authorship contribution statement Engl. J. Med. 364 (8), 701–709. Emiola, A., Oh, J., 2018. High throughput in situ metagenomic measurement of bacterial replication at ultra-low sequencing coverage. Nat. Commun. 9 (1), 4956. Xi Fu: Conceptualization, Methodology, Formal analysis, Writing -

8 X. Fu, et al. Environment International 138 (2020) 105664

Gusareva, E.S., Acerbi, E., Lau, K.J.X., Luhung, I., Premkrishnan, B.N.V., Kolundzija, S., Penang, Malaysia in relation to signs of dampness at school and fungal DNA in school et al., 2019. Microbial communities in the tropical air ecosystem follow a precise diel dust. Sci. Tot. Environ. 577, 148–154. cycle. Proc. Natl. Acad. Sci. USA 116 (46), 23299–23308. O'Connor, G.T., Lynch, S.V., Bloomberg, G.R., Kattan, M., Wood, R.A., Gergen, P.J., et al., Ji, M., Tang, S., Ferrari, B.C., 2017. Izhakiella australiensis sp. nov. isolated from an 2018. Early-life home environment and risk of asthma among inner-city children. J. Australian desert soil. Int. J. Syst. Evol. Microbiol. 67 (11), 4317–4322. Allergy Clin. Immunol. 141 (4), 1468–1475. Kirjavainen, P.V., Karvonen, A.M., Adams, R.I., Taubel, M., Roponen, M., Tuoresmaki, P., Pekkanen, J., Sunyer, J., Anto, J.M., Burney, P., 2005. Operational definitions of asthma et al., 2019. Farm-like indoor microbiota in non-farm homes protects children from in studies on its aetiology. Eur. Eespiratory J. 26 (1), 28–35. asthma development. Nat. Med. 25 (7), 1089–1095. Pekkanen, J., Valkonen, M., Täubel, M., Tischer, C., Leppänen, H., Kärkkäinen, P.M., Klepeis, N.E., Nelson, W.C., Ott, W.R., Robinson, J.P., Tsang, A.M., Switzer, P., et al., et al., 2018. Indoor bacteria and asthma in adults: a multicentre case–control study 2001. The National Human Activity Pattern Survey (NHAPS): a resource for assessing within ECRHS II. Eur. Respir. J. 51 (2), 1701241. exposure to environmental pollutants. J. Expo. Anal. Environ. Epidemiol. 11 (3), Quansah, R., Jaakkola, M.S., Hugg, T.T., Heikkinen, S.A., Jaakkola, J.J., 2012. Residential 231–252. dampness and molds and the risk of developing asthma: a systematic review and Lawley, B., Tannock, G.W., 2017. Analysis of 16S rRNA gene amplicon sequences using meta-analysis. PLoS ONE 7 (11), e47526. the QIIME software package. Meth. Mol. Biol. (Clifton, NJ). 1537, 153–163. Sharpe, R.A., Bearman, N., Thornton, C.R., Husk, K., Osborne, N.J., 2015. Indoor fungal Nadkarni, M.A., Martin, F.E., Jacques, N.A., Hunter, N., 2002. Determination of bacterial diversity and asthma: a meta-analysis and systematic review of risk factors. J. Allergy load by real-time PCR using a broad-range (universal) probe and primers set. Clin. Immunol. 135 (1), 110–122. Microbiology (Reading, England). 148 (Pt 1), 257–266. Smits, W.K., Lyras, D., Lacy, D.B., Wilcox, M.H., Kuijper, E.J., 2016. Clostridium difficile Nevalainen, A., Taubel, M., Hyvarinen, A., 2015. Indoor fungi: companions and con- infection. Nat. Rev. Dis. Primers 2, 16020. taminants. Indoor Air 25 (2), 125–156. Vandeputte, D., Kathagen, G., D'Hoe, K., Vieira-Silva, S., Valles-Colomer, M., Sabino, J., Ng, M., Freeman, M.K., Fleming, T.D., Robinson, M., Dwyer-Lindgren, L., Thomson, B., et al., 2017. Quantitative microbiome profiling links gut community variation to et al., 2014. Smoking prevalence and cigarette consumption in 187 countries, microbial load. Nature 551 (7681), 507–511. 1980–2012. JAMA 311 (2), 183–192. Wang, J., Pindus, M., Janson, C., Sigsgaard, T., Kim, J.L., Holm, M., et al., 2019. Norback, D., Markowicz, P., Cai, G.H., Hashim, Z., Ali, F., Zheng, Y.W., et al., 2014. Dampness, mould, onset and remission of adult respiratory symptoms, asthma and Endotoxin, ergosterol, fungal DNA and allergens in dust from schools in Johor Bahru, rhinitis. Eur. Respirat. J. 53(5). Malaysia- associations with asthma and respiratory infections in pupils. PLoS ONE 9 Womack, A.M., Artaxo, P.E., Ishida, F.Y., Mueller, R.C., Saleska, S., Wiedemann, K.T., (2), e88303. et al., 2015. Characterization of active and total fungal communities in the atmo- Norback, D., Hashim, J.H., Cai, G.H., Hashim, Z., Ali, F., Bloom, E., et al., 2016. Rhinitis, sphere over the Amazon rainforest. Biogeosciences 12 (21), 6337–6349. ocular, throat and dermal symptoms, headache and tiredness among students in Zhang, Z., Reponen, T., Hershey, G.K., 2016. Fungal exposure and asthma: IgE and non- schools from Johor Bahru, Malaysia: associations with Fungal DNA and mycotoxins IgE-mediated mechanisms. Curr. Allergy Asthma Rep. 16 (12), 86. in classroom dust. PLoS ONE 11 (2), e0147996. Zivkovic, M., Cadez, N., Uroic, K., Miljkovic, M., Tolinacki, M., Dousova, P., et al., 2015. Norback, D., Hashim, J.H., Hashim, Z., Cai, G.H., Sooria, V., Ismail, S.A., et al., 2017. Evaluation of probiotic potential of yeasts isolated from traditional cheeses manu- Respiratory symptoms and fractional exhaled nitric oxide (FeNO) among students in factured in Serbia and Croatia. J. Intercult. Ethnopharmacol. 4 (1), 12–18.