Article pubs.acs.org/est
Cite This: Environ. Sci. Technol. 2018, 52, 14088−14098
Land Use Influences Antibiotic Resistance in the Microbiome of Soil Collembolans Orchesellides sinensis Dong Zhu,†,‡ Qing-Lin Chen,†,‡ Hu Li,† Xiao-Ru Yang,† Peter Christie,† Xin Ke,§ and Yong-Guan Zhu*,†,‡,∥
Downloaded via TULANE UNIV on January 11, 2019 at 13:45:20 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.
†
Key Laboratory of Urban Environment and Health, Institute of Urban Environment, Chinese Academy of Sciences, 1799 Jimei Road, Xiamen 361021, China ‡ University of the Chinese Academy of Sciences, 19A Yuquan Road, Beijing 100049, China § Institute of Plant Physiology and Ecology, Shanghai Institute of Biological Sciences, Chinese Academy of Sciences, Shanghai 200032, China ∥ State Key Laboratory of Urban and Regional Ecology, Research Center for Eco-Environmental Sciences, Chinese Academy of Sciences, Beijing 100085, China S Supporting Information *
ABSTRACT: Numerous studies have investigated the composition and diversity of antibiotic resistance genes (ARGs) in multiple environments but the pattern of ARGs in field-collected soil fauna remains poorly understood. In the present study soil collembolans were collected from six sites with three different land use types (parkway land, park land, and arable land) and 285 ARGs and 10 mobile genetic elements (MGEs) in the microbiome of these “wild” collembolans were quantified by high-throughput quantitative PCR. A total of 76 unique ARGs and 5 MGEs were detected. There were significant differences between collection sites in the antibiotic resistome in the collembolans. Land use significantly altered the distribution patterns of collembolan ARGs. Thirty shared ARGs and three shared MGEs were identified. The co-occurrences of shared resistomes were largely random, and more positive relationships were found in the coassociation network. Partial redundancy analysis confirms that the changes in bacterial communities explained 27.77% of the variation in ARGs. These findings suggest that resistance genes are pervasive in the microbiome associated with the field collembolan and the activity of the collembolans may contribute to the spread and dissemination of resistance genes in the environment, an aspect of ARGs that has until now been largely overlooked.
■
INTRODUCTION Antibiotic resistance genes (ARGs) are a product of microbial evolution and selection and are widely distributed in the natural environment, although usually in low abundance.1 The increasing abundance of ARGs at an unprecedented rate in our environment due to the intensive use of antibiotics is well documented2−4 and has aroused widespread public concern.5,6 The emergence and widespread occurrence of antibiotic resistance genes have become a global threat to human health.6 Numerous studies have therefore been conducted to determine the abundance and composition of ARGs in multiple environments (e.g., soils,4,7 rivers,8 landfill sites,9 municipal wastewater treatment plants,10 indoor dust,11 the phyllosphere of vegetables,12 animal faeces,13 and the gut of honey bees14). The widespread use of organic fertilizers derived from animal manures and sewage sludges and irrigation with reclaimed wastewaters has led to the accumulation of ARGs in many soils.7,12,15 Diverse animals dwell in soils at different trophic levels constituting complex food webs and may be important microbial habitats and potential reservoirs of ARGs.16 It is well established that soil © 2018 American Chemical Society
fauna can carry and transport microbes within the soil ecosystem.17 However, little is known about the antibiotic resistome in soil fauna, and there is thus a general knowledge gap regarding the composition and transport of ARGs in soil food webs. There have been several studies of aquatic systems. Daphnia has been found to be a potential refuge for ARGs,18 and the fish gut may be a hotspot of ARG transport and horizontal transfer.19 Previous studies have also revealed that the gut of soil fauna is a hotspot for microorganisms.20−22 It is therefore important to study the composition and diversity of ARGs in the soil fauna. Collembolans are among the most abundant soil fauna and are distributed in diverse soil environments. They play a key role in the decomposition of litter, energy transfer, and the formation of the soil microstructure.23 They can ingest various foods such as plant roots, bacteria, fungi, protozoa, and Received: Revised: Accepted: Published: 14088
September 11, 2018 November 21, 2018 November 27, 2018 November 27, 2018 DOI: 10.1021/acs.est.8b05116 Environ. Sci. Technol. 2018, 52, 14088−14098
Article
Environmental Science & Technology
park land (ZS-P and XD-P), two areas of parkway land (FY-G and ZSY-G) and two of arable land (HX-A and BS-A) representing three typical land use types in the developing city (Figure S1 of the Supporting Information, SI). The history of park ZS-P was longer than park XD-P. Moreover, park ZS-P was located in the heart of the city and park XD-P was located in the development zone. The park land and parkway land were amended with approximately 1 kg m2 organic fertilizer (sewage sludge and pig manure) every year. Compared with the parkway land, the pesticide was only applied in the park land, and higher diversity of vegetation and more human wastes were found in the park land. Organic compound fertilizer (approximately 2 kg m2 organic fertilizer including sewage sludge, animal manure, and green manure) and pesticide were used in the arable land. On April 15, 2017, all soil samples were collected to a depth of 5 cm from each site (three random samples per site, six samples per land use and 1.5 kg of soil per sample). The soil collembolans were separated from the 18 soil samples using a suction sampler. The collembolans collected were identified within 2 h. The dominant species Orchesellides sinensis was selected and the adults obtained were immediately frozen at −20 °C until further analysis. Three to five collembolan specimens collected per 1.5 kg soil sample were treated as one replicate. Sample Pretreatment and DNA Extraction. The frozen collembolan samples were washed three times using 0.5% sodium hypochlorite (0.5 min each time) and rinsed for 1 min with sterile phosphate-buffered saline (PBS) at 4 °C.37,38 Three to five surface-sterilized collembolans per sample were transferred to a sterile 1.5-ml centrifuge tube. The collembolan tissues transferred were disrupted for 0.5 min using a microelectric tissue homogenizer equipped with a sterile 1.5 mL pellet pestle. The DNeasy Blood and Tissue Kit (QIAGEN, China (Shanghai) Co., Ltd.) was used to isolate DNA from the collembolan samples. The collembolans were homogenized and 180 mL tissue lysis buffer (ATL solution, QIAGEN, Hilden, Germany) and 20 mL proteinase K were added to the tube. The tubes were vortexed for 1 min and kept at 56 °C for at least 6 h. After incubation, the procedure of collembolan DNA extraction was followed strictly on the basis of the kit manufacturer’s instructions. Finally, 60 μL elution buffer (AE solution, QIAGEN, Hilden, Germany) was used to elute the extracted collembolan DNA into a sterile 1.5-mL centrifuge tube. The collembolan DNA obtained was stored at −20 °C prior to use. After the morphological identification the information on collembolan species was further confirmed by amplifying the COI barcode region [primers: LCO1490 (GGTCAACAAATCATAAAGATATTGG) and HCO2198 (TAAACTTCAGGGTGACCAAAAAATCA)].39 The PCR system and reaction conditions are based on the description of a previous study.39 We submitted the sequences obtained to the National Center for Biotechnology Information (NCBI) using the Basic Local Alignment Search Tool (BLAST) to identify the species of collembolan. In the BLAST search, the parameters (e.g., percent identity (99%) and e-value (0.0)) were selected to determine the name of collembolan. Antibiotic Resistance Genes Detected Using HighThroughput Quantitative PCR. The composition and abundance of antibiotic resistance genes (ARGs) of collembolan samples were characterized using a SmartChip Real-time PCR system (Wafergen Inc., Fremont, CA) to perform high-throughput quantitative PCR (HT qPCR) in
nematodes. They are also attractive prey for many predators and occupy multiple trophic levels and are therefore important components of soil food webs.23,24 Collembolans are usually colonized by a large number of microorganisms.22 More importantly, they are hosts of pathogenic microbiota and can transport pathogenic microbiota along with their activities.25,26 If pathogenic microbiota present in collembolans contain one or more resistance genes, then the collembolans may promote the dispersal of resistant pathogenic microbiota in the soil ecosystem and this may represent a threat to the soil ecosystem and to human health. Soil collembolans are therefore excellent model organisms for investigation of the composition and diversity of soil faunal ARGs. Moreover, β-lactam resistance genes have been detected in the genome of the model collembolan Folsomia candida.27 Our previous study also indicates that environmental stress such as oral-antibiotic exposure can indeed change the abundance and diversity of ARGs in the gut of the collembolan F. candida.16 Nevertheless, the diversity and abundance of ARGs from field-collected soil animal microbiomes remain poorly characterized. Rapid urbanization often involves changes in land use types.28,29 Land use types of developing cities often include a parkway land, park land, and arable land. The parkway land and park land are the main recreational areas for urban residents, especially children,30−32 and people may make direct contact with soil fauna such as collembolans in these areas. The arable land is commonly located at the edge of the city for food production for city dwellers, and soil fauna make an important contribution to the maintenance of the productivity of arable land and food safety.33,34 Moreover, sewage sludges and organic fertilizers are often applied to parkway land, park land, and arable land for nutrient recycling and waste assimilation, and this can lead to increasing abundance of ARGs in these areas which may then become reservoirs of the ARGs.15,32 Reports of the microbiome of collembolans Orchesella cincta35 and Folsomia candida36 suggest that although collembolans may have a conserved microbiome that is different from the environment, the habitat may make an important contribution to their microbiome. It is therefore important to characterize the ARG profiles of soil fauna from different land use types. In the present study soil fauna were sampled from these three land use types in Xiamen, east China, in order to investigate the abundance and diversity of ARGs and mobile genetic elements (MGEs) in the collembolan microbiome by high throughput qPCR (285 primer pairs targeting most classes of ARGs, 8 transposase marker genes, and 2 integrase marker genes), to reveal the effects of land use on the ARG profile in the collembolan microbiome, to present the shared resistome between collembolans from different land use types using network analysis, to investigate the relationship between the shared resistance genes and identify the core roles of specific ARGs by co-occurrence and coassociation analysis, and to explore the relationships between collembolan bacterial communities (using high-throughput sequencing of the 16S rRNA gene), MGEs and ARGs to elucidate the microecological mechanisms by which resistance genes are produced.
■
MATERIALS AND METHODS Field Collection of the Soil Collembolan Orchesellides sinensis. We selected six sites to separate the soil collembolan Orchesellides sinensis in Xiamen, Fujian, China, namely two of 14089
DOI: 10.1021/acs.est.8b05116 Environ. Sci. Technol. 2018, 52, 14088−14098
Article
Environmental Science & Technology
instructions.45 First, clean reads were obtained by merging paired-end reads, removing primer sequences and filtering via Phred quality scores. Then, in QIIME, we used the usearch_qf to remove chimeras and cluster the operational taxonomic units (OTUs) at the 97% similarity level of sequences using de novo OTU picking.47 Only singletons were removed prior to the downstream analysis. In each cluster, the OTU was represented using the most abundant sequence by the default method. On the basis of the Greengenes version 13.8 16S rRNA gene database,48,49 PyNAST aligner was used to align the representative sequence50 and RDP Classifier 2.2 was used to assign the taxonomic status of OTUs. Greengenes lanemask was used to filter the alignment. The phylogenetic tree was structured by the fast tree algorithm.51 OTUs with a total number of reads of 3% read abundance of at least one sample), and was produced using Microsoft Excel 2013. We performed redundancy analysis (RDA) and partial redundancy analysis in the vegan 2.3−1 package of R to reveal the contribution of the collembolan associated with microbiota at the genus level (>0.1 read abundance of at least one sample) and MGEs to the change in ARGs in all samples from each site. We set microbial community and MGEs as independent variables and ARGs as dependent variable. The Euclidean distance between samples was used to reflect the relations between microbial community and MGEs and ARGs. The angles of vectors indicate the strength of the relationship, and a strong correlation was shown by a narrow angle in the RDA. Partial redundancy analysis was used to decompose the total variance of the data matrix into different components to obtain the interpretation of different factors to variables. When other variables were constrained, the contribution of target variables to the change in ARGs was obtained in the partial redundancy analysis. The vegan 2.3−1 package of R was used to calculate the diversity indexes (Shannon, Inverse Simpson, and Pielou evenness) and perform redundancy analysis (RDA), partial redundancy analysis, the Adonis test, the Mantel test, and the Procrustes test to explore the relationships between collembolan ARGs and the bacterial community. To further investigate the cause of ARGs in the collembolan microbiome, we picked out 11 potential antibiotic producers belonging to Bacillus, Streptomyces, Nocardia, and Micromonosporaceae
within the community and used the Mantel test to reveal the correlations between these antibiotic producers and ARGs.
■
RESULTS Abundance, Diversity, and Composition of Antibiotic Resistance Genes. A total of 76 unique ARGs and 5 MGEs (2 integrase genes containing 1 class1 integron-integrase gene and 3 transposase genes) were detected in all collembolan samples. The average numbers of ARGs detected in the collembolan microbiome of each site ranged from 17 to 30 and were significantly different between different sites (F(5, 12) = 6.19, P = 0.005; Figure 1a). According to the classification of antibiotic resistance, we divided the ARGs detected into 9 types (i.e., aminoglycoside, macrolide-lincosamide-streptogramin B (MLSB), sulfonamide, β-lactams, multidrug, vancomycin, tetracycline, phenicol, and others). The numbers of aminoglycoside and tetracycline resistance genes in collembolans from the sites HX-A and BS-A were significantly higher than those from other sites (one-way ANOVA, P < 0.01). More β-lactamase (7 and 6), MLSB (3 and 4), and vancomycin (4 and 5) resistance genes were detected in the sites HX-A and BS-A than in the sites FY-G (4, 2, and 3) and ZSY-G (2, 2, and 3), respectively (P < 0.05). The normalized abundances of ARGs in collembolans collected ranged from 0.19 to 2.68 copies per bacterial cell, and the abundances of MGEs were between 0.015 and 0.36 (Figure 1b). The samples from different collection sites followed the sequence in the normalized abundance of ARGs: ZS-P > ZSY-G > FY-G > BSA > HX-A > XD-P (one-way ANOVA, P < 0.01). The Venn diagram showed that the soil microbiome could contain the same ARGs as the collembolan gut (Figure S2). The distribution patterns of ARGs presented a significant difference in the collembolan microbiome across collection sites (Adonis tests, P < 0.005). Principal coordinates analysis (PCoA) using Bray−Curtis distances based on the relative abundance of ARGs of collembolan samples also reveals that the ARG patterns of collembolans were clustered by the collection sites (Figure 2). Almost half of the variance (48%) of collembolan ARGs was explained via the first two principal components. In addition, Inverse Simpson and Shannon 14091
DOI: 10.1021/acs.est.8b05116 Environ. Sci. Technol. 2018, 52, 14088−14098
Article
Environmental Science & Technology
Figure 2. Principal coordinates analysis (PCoA) of collembolan ARGs samples using Bray-Curtis distances based on relative abundance of ARGs. Each point represents a distribution pattern of ARGs of one sample. Collection sites are represented by different colors. FY-G and ZSY-G, parkway land; ZS-P and XD-P, park land; and HX-A and BS-A, arable land. The variation explained by the PCoA axes is listed in parentheses.
Figure 3. Bipartite network analysis depicting the shared ARGs and MGEs between arable land, park land, and parkway land in the microbiome of the soil collembolan Orchesellides sinensis. The collection sites were classified according to land use type before analysis. The prevalent genes within a land use type were used as edge weights. The meanings of the letters are (a) seven ARGs and one MGE unique to park land; (b) 11 ARGs shared via park land and arable land; (c) one transposase and 20 ARGs were exclusively found in collembolans collected from arable land; (d) only two shared ARGs were detected between arable land and parkway land; (e) parkway land contained only three unique ARGs; (f) parkway land and park land together shared three ARGs; and (g) 30 ARGs and three MGEs were found across all collection sites. The prevalent genes within a land use type were used as edge weights.
indices of collembolan ARGs shifted significantly due to the different collection sites (P < 0.05, Figure S3). The Inverse Simpson index indicates that the diversity of ARGs from collection site HX-A was significantly lower than from sites BSA, ZS-P, and XD-P, which is consistent with a change in Simpson index (P < 0.05). The relative abundance of ARGs was positively correlated with MGE abundance (R2 = 0.816, P < 0.001) in collembolans from the field (Figure S4). Influences of Land Use on Antibiotic Resistance Genes. Collembolans collected from arable land harbored the maximum numbers of ARGs (30) (Figure 1a). In the park land ARGs normalized abundance of collembolans from the ZS-P were approximately 14.1 times higher than from the XD-P (Figure 1b). Overall, normalized abundances of MGEs in park land and parkway land were higher than in arable land (Figure 1b). The results of PCoA showed that the ARG patterns of collembolans from each land use type were clustered together and were differentiated from other land use types (Figure 2). Along with PCo 1, parkway land and arable land were separated, and collection site ZS-P was separated from collection site XD-P. Although the significant effects of land use on collembolan ARGs were not found at the abundance, alpha diversity and PCoA1 of ARGs as dependent variable, respectively, the results of the multilevel linear model indicate that land use significantly altered the detected number (F(1, 3.568) = 18.45, P = 0.01) and distribution patterns of collembolan ARGs (F(1, 3.435) = 16.84, P = 0.02) in dimension 2 of the PCoA (explaining 16.9% of the variation). Shared and Unique Resistome between Different Land Use Types. A bipartite network analysis was adopted to present shared ARGs between different land use types and those unique ARGs in each sample (Figure 3). We obtained seven categories, and the categories referring to shared ARGs are of major concern due to their prevalence. They were (summarized clockwise): (a) seven ARGs and one MGE (IS613) unique to park land; and (b) 11 ARGs shared via park land and arable land. These shared ARGs are classified as aminoglycoside, phenicol, MLSB, tetracycline, β-lactamase, and multidrug based on their resistance; (c) one transposase (Tp614) and 20 ARGs were exclusively found in collembolans
collected from arable land; (d) only two shared ARGs (ampC and cmx(A)) were detected between arable land and parkway land; (e) parkway land contained only three unique ARGs (pncA, putitive multidrug and tolC); (f) parkway land and park land together shared three ARGs (acrF, mtrD, and aphA1(aka kanR)); and (g) 30 ARGs and three MGEs (two integrase and one transposase) were found at all collection sites. These shared ARGs consisted of multidrug (7), aminoglycoside (6), β-lactamase (6), vancomycin (4), tetracycline (2), MLSB (2), sulfonamide (2), and others (1). Between these shared genes, the four genes with the highest normalized abundance were integrase intI-1 (0.05) and ARGs (vanXD (0.17), blaCTX-M (0.09) and blaTEM (0.08)). Overall, the shared genes held 60.5% of all the genes detected. These shared genes were used for further analysis. The 49 shared resistance genes (detected in at least two land use types) were used to analyze relationships between classes of ARGs in collembolans. 1023 gene pair combinations were compared in which only 41 (∼4%) pairs showed significant cooccurrence, consisting of 4 negative and 37 positive relationships (Figure S5). Co-occurrences of most shared gene classes were predominantly random. The ARGs tetB had higher positive correlations than other genes, and more negative correlations were observed in ARGs tetG. Both tetB and tetG are tetracycline resistant genes. Interestingly, the integrase cIntI-1 also showed a relatively high positive correlation. Coassociation patterns of 49 shared ARGs were investigated by network analysis in collembolans. Eleven of 49 shared ARGs were removed in the network analysis and these were not correlated with other shared ARGs. The coassociation network 14092
DOI: 10.1021/acs.est.8b05116 Environ. Sci. Technol. 2018, 52, 14088−14098
Article
Environmental Science & Technology
Figure 4. Co-association network of the antibiotic resistance genes in the microbiome of the soil collembolan Orchesellides sinensis. Nodes reveal the 29 shared resistance genes that occurred in at least two land use types. (a) The nodes are colored based on ARG types. Node labels refer to ARG classes, and node size indicates their relative abundance. For all plots, red edges represent positive coassociations, blue edges indicate negative coassociations, and edge thickness is related to the correlation coefficient. (b) Modules are presented with the colors of nodes.
Figure 5. Role of the bacterial community and mobile genetic elements (MGEs) in the changes in ARGs in the microbiome of the soil collembolan Orchesellides sinensis; (a) redundancy analysis (RDA) of the quantitative correlation between ARGs and the major bacterial genus (>0.1 read abundance of at least one sample) and abundance of MGEs in all samples from each site; and (b) partial redundancy analysis differentiating effects of bacterial community (BC) and MGEs on the variations in ARGs.
consisted of 40 edges and 29 nodes, and the modularity and clustering coefficient of the network were 0.56 and 0.44, respectively (Figure 4). More positive correlations (37) were shown in the coassociation network compared to negative correlations (3) (Figure 4), and this is consistent with the results of the co-occurrence analysis. Three ARGs (vanXD, blaCTX-M, and blaTEM) with the highest normalized abundances were all located in module 1 and these were also highly connected with other genes. Module 6 contained two MGEs (cIntI-1(class1) integrin and tnpA transposase), and cIntI-1 was positively connected with aadA, aadA2, and aac(6′)-Ib(aka aacA4) which were divided into aminoglycoside. Relationships between Collembolan ARGs, MGEs, and Bacterial Communities. Using the Procrustes and Mantel tests we explored whether the ARG profiles were correlated with the collembolan microbial OTU composition
(Figure S6). The results of the Procrustes analysis reveal clustering of collembolan bacterial OTU data and ARGs by collection site and present a goodness-of-fit test (M2 = 0.476, P < 0.01, 9999 permutations) based on the Bray−Curtis dissimilarity metric. Moreover, the Mantel test also shows that the collembolan bacterial 16S rRNA gene OTU data were significantly correlated with the ARG profiles (r = 0.334, P < 0.01). Moreover, we found a significant relationship between 11 potential antibiotic producers and the ARGs by the Mantel test (r = 0.512, P = 0.001). Relationships between collembolan bacterial communities (at the genus level), MGEs and ARGs were studied using redundancy analysis (RDA) (Figure 5a). The results reveal that 85.6% of the total variation was explained by the first two PCs of selected variables (Figure 5a). Apart from the genus Citrobacter, other genus all positively correlated with the first axis (explaining 75.6% of total variance). The MGEs and the 14093
DOI: 10.1021/acs.est.8b05116 Environ. Sci. Technol. 2018, 52, 14088−14098
Article
Environmental Science & Technology
Figure 6. Composition and abundance of the microbiota of soil collembolan Orchesellides sinensis. Heatmap revealing the 27 most abundant OTUs (>3% of the total number of reads) detected across land use sites, which are displayed by the genus level or higher possible taxonomic classification. The individuals from each site are sorted by site and land use.
first axis were negatively correlated. Partial redundancy analysis indicates that our selected factors explained 80.66% of total ARG variation (Figure 5b). The interaction between collembolan bacterial community and MGEs contributed 39.88% of total ARG variation, and the single collembolan bacterial community explained 27.77% of the variation in the changes in ARGs. Characterization of the Collembolan Microbiome from Different Collection Sites. Across all 18 samples we obtained 2 270 840 nonsingleton reads, and each sample included at least 48 953 sequences. Overall, 890 OTUs were used for downstream analysis and other OTUs were removed due to
