Data sources and Sampling.
Data sets on air quality and antibiotic resistance (percent resistant) in 13 common pathogens i.e., carbapenem resistant Escherichia coli (CREC), Pseudomonas aeruginosa (CRPA), Acinetobacter baumannii (CRAB), and Klebsiella pneumoniae (CRKP), as well as erythromycin resistant Streptococcus pneumoniae (ERSP), methicillin resistant Staphylococcus aureus (MRSA), and coagulase negative staphylococci (MRCNS), penicillin resistant Streptococcus pneumoniae (PRSP), quinolones resistant E. coli (QREC), third-generation cephalosporin resistant E. coli (3GCREC) and Klebsiella pneumoniae (3GCRKP), and vancomycin resistant Enterococcus faecalis (VREFS) and Enterococcus faecium (VREFM) were downloaded from publicly available data from the National Bureau of Statistics(http://www.stats.gov.cn/) and Antimicrobial Resistance Surveillance System of China(http://www.carss.cn/), respectively. The final data sets of antibiotic resistance bacteria represented over 10 million clinically-relevant bacterial pathogens from over 1.3 thousand hospitals across thirty-two provincial administrative regions of China from 2014 to 2017. The used clinically-relevant bacterial pathogens were mainly derived from sputum specimens (averagely accounting for 41.6%), urine specimens (averagely accounting for 17.6%) and blood specimens (averagely accounting for 9.5%). In China, air quality index (AQI) is divided into six levels, ranging from first-level excellent (0–50), second-level good (51–100), third-level mild pollution (101–150), fourth-level moderate pollution (151–200), to fifth-level severe pollution (201–300), and sixth-level severe pollution (more than 300). The number of days with the AQI below 100 can more effectively represent the air quality of a year than the average concentration of pollutants. To characterize ARGs, MGEs, and bacteria community distribution in clinical aerial environment, 70 samples were collected from a hospital during July 27 to August 2 in 2018, including 17 patient airway (P) sputum samples (the samples were collected from volunteers in Inpatient Respiratory Department) and 11 healthy human airway (H) sputum samples, and respiratory-related environmental samples such as 7 indoor dust (ID), 7 outdoor dust (OD), 7 indoor PM2.5 (I-PM2.5), 7 outdoor PM2.5 (O-PM2.5), 7 indoor PM10 (I-PM10), and 7 outdoor PM10 (O-PM10) samples (detailed sampling protocols, ethics approval, and DNA extraction procedures are described in Supplementary Information Text S1).
High-throughput qPCR.
ARGs and MGEs were analyzed by high-throughput qPCR (HT-qPCR) using the Takara SmartChip™. For this, 285 antibiotic resistance gene primer sets for all major classes of antibiotics, 9 transposase gene primer sets, one clinical class 1 integron-integrase gene primer set, and one 16S rRNA gene primer set were evaluated (Table S1)[6, 26]. The antibiotics resistance genes to aminoglycoside, beta-lactams, fluoroquinolone/quinolone/florfenicol/chloramphenicol/amphenicol (FCA), macrolide/lincosamide/streptogramin B (MLSB), sulfonamide, and tetracycline, vancomycin, and genes coding multidrug efflux pumps or antibiotic deactivation protein resistance to other antibiotic and bactericide (multidrug) resistance genes were targeted by the 285 ARGs primer sets [5, 27] (detailed HT-qPCR procedures are described in Supplementary Information Text S2).
Bacterial 16S rRNA gene sequencing.
Bacterial community structures were determined by 16S rRNA gene sequencing on a HiSeq PE250 platform (Illumina Inc., San Diego, CA). V4 to V5 regions of bacterial 16S rRNA were amplified with the universal primer set 515F (5’- GTGCCAGCMGCCGCGG-3’) and 907R (5’-CCGTCAATTCMTTTRAGTTT-3’)[28] labeled with unique barcodes (7-nucleotide barcodes) for each sample. All sequences of each sample were screened by filtering adaptor sequences and removing low-quality reads, ambiguous nucleotides, and barcodes. High-quality sequences were analyzed using Quantitative Insights into Microbial Ecology (QIIME) and clustered into operational taxonomic units (OTUs) at 97% similarity level, using UCLUST.
Physiochemical characterization of PM 2.5 and PM10 samples.
PM2.5 and PM10 sample size distributions were evaluated in Zetasizer Nano ZS90 and Mastersizer 2000 equipment (Malvern, UK). Metal elements were analyzed by Inductively coupled plasma emission spectrometer (Perkin Elmer, Norwalk, CT), after acid digestion.
Isolation of airborne conjugative plasmids and nanopore sequencing.
Filter mating assays were applied to capture airborne conjugative plasmids using Escherichia coli NK5449 (nalidixic acid and rifampicin-resistant) as recipients and airborne sample isolations as donors. Airborne sample isolations cultured from collected air sample filters using LB media. Donor and recipient strains initial bacterial concentrations were approximately 108 CFU/mL, which were monitored at an optical density (OD) at 600 nm (OD600 value was approximately 0.8). After culturing for 12 h at pH 7.0 and 37 oC, suspension was plated on LB agar supplemented with 50 mg/L kanamycin, 50 mg/L rifampicin, 100 mg/L cycloheximide, 100 mg/L ampicillin, 4 mg/L ciprofloxacin, 20 mg/L chloramphenicol, 50 mg/L gentamicin, 30 mg/L streptomycin, 10 mg/L tetracycline, and 50 mg/L erythromycin, respectively. Transconjugants showed more resistance than recipients, grew on selective LB agar plates, and were identified using denaturing gradient gel electrophoresis (DGGE), following transconjugant plasmids extraction by the Qiagen Midi kit (Qiagen, Germany). Harvested plasmids DNA were detected by agarose gel electrophoresis and quantified by Qubit® 2.0 Fluorometer (Thermo Fisher Scientific, Waltham, MA). The whole plasmid genome was sequenced using nanopore sequencing on PromethION platform and Illumina NovaSeq PE150 equipments at Beijing Novogene Bioinformatics Technology Co., Ltd.; NCBI reference sequence database, comprehensive antibiotic resistance database, and antibiotic resistance genes database were used to identify antibiotic, biocide, and metal resistance genes. Presence of genomic islands within plasmids was predicted using Island-Viewer 4 and plasmid profiles were created using the SnapGene software 3.2.1.
Conjugation experiments after PM 2.5 and PM10 exposure.
An optimized conjugation model to evaluate the conjugative transfer of ARGs between E. coli HB101 and E. coli RP4 was used, as previously reported [29]. E. coli HB101, which harbors RP4 plasmid carrying resistance to tetracycline, ampicillin, and kanamycin, was selected as the donor strain in this study; E. coli NK5449 represented the plasmid recipient. Donor and recipient strains were mixed and treated with PM2.5/PM10 at 31.25, 62.5, 125, 250, and 500 µg/mL for 12 h at pH 7.0 and 37 oC [30]. PM2.5 and PM10 solutions were prepared as previously reported [30]. Amount of recipient bacteria and transconjugants were counted on LB-selective solid medium as colony-forming units per milliliter (CFU/mL). Transconjugants were identified using DGGE and amplified by PCR with RP4F 5’- AAAGCGGACAGCATCAGTAACGAA-3’) and RP4R (5’- GAGCTTGGTGGCCGCATAGTGTAG − 3’) primers. The conjugative transfer frequency was calculated as the amount of transconjugant cells to recipient cells ratio.
ROS levels evaluation.
Intracellular ROS levels were determined using the 2′,7′- dichlorofluorescein diacetate (DCFH-DA) probe (Invitrogen, Carlsbad, CA), according to manufacturer’s instructions. A CytoFLEX flow cytometer (Beckman Coulter, Brea, CA) was used to detect ROS at 488 nm excitation and 525 nm emission wavelengths, and evaluated by the fluorescence intensity of treated sample and control group ratio. All assays were performed in triplicate.
Transmission electron microscopy (TEM) detection.
Cell membrane permeability was visualized using a JEM-1010 transmission electron microscope (JEOL, Tokyo, Japan) at 75 kV with a CCD camera (detailed procedures are described in Supplementary Information Text S3).
RNA extraction, genome-wide RNA sequencing, and bioinformatics.
Conjugation mating systems were established as described above, using PM2.5 and PM10 at 0.0 (control), 31.25 (low-dosage), 125 (medium-dosage), and 500 µg/mL (high-dosage). After a mating period, QIAGEN miRNeasy Mini Kit (QIAGEN, Germany) was used to extract the total RNA from the mixture, following manufacturer’s instructions. RNA samples were analyzed in a Qubit 2.0 (Thermo Fisher Scientific) and quality controlled on an Agilent 2100 bioanalyzer (Agilent Technologies, Palo Alto, CA). cDNA libraries were constructed using the rRNA-depleted RNA by NEBNext® Ultra RNA Library Prep Kit (New England Biolabs, Inc., Ipswich, MA). According to supplier’s instructions. Sequencing was performed on Illumina NovaSeq PE150 (Illumina Inc.). Gene expression was calculated as fragments per kilobase of a gene per million mapped reads (FPKM). Differences in fold changes between different groups were calculated by log2 fold-change (LFC) between control and airborne PM-treated samples.
Statistical analysis.
Diversity index, Mantel test, and Procrustes analysis were generated by R with ‘vegan’ packages. Heatmaps of China were generated by R with ‘mapdata’, ‘maptools’, ‘plyr’, ‘mapproji’, ‘sp’, ‘maps’, and ‘ggplot2’ packages. Linear regression analysis graphs were generated by R with ‘ggplot2’ package. Heatmap was generated by R with ‘pheatmap’ packages. PCoA (Bray–Curtis distance based) was generated by CANOCO 5. Box and bar charts were generated by OriginPro 9.0 (OriginLab Corp., Northampton, MA). Network analysis was conducted in the Gephi platform and only statistically robust correlations of Spearman’s correlation coefficient (ρ) > 0.9 and significance level P < 0.01 were used to form the final networks by Frucherman Reingold algorithms [26, 31, 32]. Phylogenetic tree was extracted from the taxa in MEGA7 program, using the maximum-likelihood tree with 500 bootstraps. SourceTracker 1.0.1 was used to estimate the relative contribution of the microbial taxa and ARGs from the source environments to the sink environments by Bayesian methods [33] (details are described in Supplementary Information Text S4).