Environment International 133(2019)105203 Contents lists available at ScienceDirect e environment Environment International ELSEVIER journal homepage: www.elsevier.com/locate/envint Efficient reduction of antibiotic residues and associated resistance genes in tylosin antibiotic fermentation waste using hyperthermophilic composting Hanpeng Liao, Qian Zhao, Peng Cui, Zhi Chen", Zhen Yu, Stefan Geisen, Ville-Petri Friman Shungui Zhoua "Fujian Provincial Key Laboratory of Soil Environmental Health and Regulation, College of Resources and Environment, Fujian Agriculture and Forestry University, Fuzhou, China " Guangdong Key Laboratory of Integrated Agro-environmental Pollution Control and Management, Guangdong Institute of Eco-environmental Science & Technology, Guangzhou 510650. China Department of Terrestrial Ecology, Netherlands Institute of Ecology, Wageningen, Netherlands Department of Biology, University of York, Wentworth Way, YO10 5DD York, UK ARTICLE INFO ABSTRACT Handling Editor: Zhen (Jason) He Insufficient removal of antibiotics and antibiotic resistance genes (ARGs) from waste products can increase the Keywords: risk of selection for antibiotic resistance in non-clinical environments. While composting is an efficient way to Antibiotic resistance genes reduce ARGs, most conventional methods are ineffective at processing highly contaminated antibiotic fermen Antibiotic residue tation waste. Here we explored the efficacy and underlying mechanisms of hyperthermophilic composting at Hyperthermophilic composting removing tylosin antibiotic fermentation residues (TFR) and associated ARGs and mobile genetic elements Horizontal gene transfer (MGEs; plasmids, integrons and transposon). Hyperthermophilic composting removed 95.0% of TFR, 75.8% of Plasmids ARGs and 98.5% of MGEs and this reduction mainly occurred after extended exposure to temperatures above 60"C for at least 6 days. Based on sequencing and culture-dependent experiments, reduction in ARGs and MGEs was strongly associated with a decrease in the number of bacterial taxa that were initially associated with ARGs and MGEs. Moreover, we found 94.1% reduction in plasmid genes abundances (ISCR1 and IncQ-oriV) that sig- nificantly correlated with reduced ARGs during the composting, which suggests that plasmids were the main carriers for ARGs. We verified this using direct culturing to show that ARGs were more often found in plasmids during the early phase of composting. Together these results suggest that hyperthermophilic composting is efficient at removing ARGs and associated resistance genes from antibiotic fermentation waste by decreasing the abundance of antibiotic resistance plasmids and associated host bacteria. 1.Introduction pose a major challenge to global public health through selection for multidrug resistant bacterial "superbugs" that are very difficult to treat Antibiotic fermentation residues are organic solid waste products (Povolo and Ackermann, 2019). Developing effective ways to treat created in the manufacturing process of antibiotics fermentation. They waste that contains antibiotics and ARGs is important for controlling contain fermentation media, antibiotics residues, heavy metals (zinc the development of antibiotic resistance in natural environments and copper etc.) sludge and multiple different antibiotic resistance (Bondarczuk et al., 2016). genes (ARGs)(Zhang et al., 2018a). Globally, millions of tons of anti Treatment of antibiotic fermentation residues using composting biotic fermentation residues are produced every year, and traditionally methods has recently received more research interest (Wang et al., these waste products are disposed in landfills (Chen et al., 2017b) 2016; Zhang et al., 2018a). Besides making waste safer, composting end However, this practice can cause a serious threat to the environment products contain high amounts of organic matter and mineral nutrients, through leaching of antibiotic pollutants into the natural environments which could allow it to be used as organic fertilizer. Despite attempts to including soils, groundwater and nearby waters (Chen et al, 2017a), assess the residual antibiotics and ARGs during the composting in la- where they could select for antibiotic resistant bacteria (Pepper et al. boratory conditions, traditional methods often fail to remove ARGs 2018). ARGs are thus considered emerging human-toxic pollutants that from the composting end products (Zhang et al., 2018a; Zhang et al., 'Corresponding author. E-mail address: sgzhou@soil.gd.cn (S. Zhou). https://doi.org/10.1016/j.envint.2019.105203 Received 14 April 2019; Received in revised form 17 September 2019; Accepted 19 September 2019 Available online 26 October 2019 0160-4120/2019 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/)
H.Liao,et al. Environment International 133 (2019)105203 2018c).For example,it was recently shown that 40 days of composting composition)factors affecting ARG and MGE abundances during hy- only removed 39%of initial tylosin antibiotic residues in swine manure perthermophilic composting. (Zhang et al.,2018c).Moreover,Liu et al.(2018)recently reported that while the abundance of ARGs and mobile genetic elements (MGEs) 2.Materials and methods initially declined during composting of gentamicin fermentation re- sidues,their abundances rapidly recovered and even exceeded the in- 2.1.Full-scale experimental setup for hyperthermophilic composting itial concentrations during the later phases of composting.These fail- ures could be attributed to several factors.For example,MGEs,such as Hyperthermophilic composting experiments were conducted in a plasmids,could mobilize ARGs allowing them to move between suitable full-scale aerobic composting plant located in Henan district,Ningxia, bacterial hosts during the composting.Moreover,antibiotic fermenta- China as described by Liao et al.(2018).The hyperthermophilic com tion waste often also includes antibiotic residues and heavy metals that posting material (approximately 21 tons)consisted a mixture of TFR could reinforce selection for multidrug resistance plasmids that often waste (70%water content)and rice straw husk (15%water content, encode resistance genes for both antibiotics and heavy metals (Song provided by a local farm)in a ratio of 4:1 (w/w).TFR waste was ob- et al.,2017).Here we tested if hyperthermophilic composting that uses tained from a local biological pharmaceutical factory (Ningxia,China). relatively higher composting temperatures to traditional composting The main characteristics of the raw materials used for composting are methods (Liao et al.,2018)could be an effective way to treat tylosin shown in Table S1.Both raw waste materials were mixed thoroughly antibiotic fermentation waste. resulting in final moisture content of approximately 55%before loading Tylosin is one of the main macrolide antibiotics that is globally used into a fermentation compartment with the following dimensions:2.0 m in veterinary medicine and millions of tons of tylosin antibiotic fer- height,8.0 m length,and 4.0 m width.A forced ventilation system at mentation residue (TFR)waste is generated every year.TFR waste ty- the bottom of the compartment was used to ensure aerobic conditions. pically contains a high amount of antibiotic residues and heavy metals To mix the compost substrate well and to reduce pile-edge effects,a (Zhang et al.,2018a)that could act strong selection pressures for the mechanical turning of the hyperthermophilic composting material was transmission and prevalence of ARGs in waste-impacted bacterial performed every seven days using pile-specific forklifts.Fermentation communities (Pal et al.,2015).In addition to macrolide resistance temperature was daily monitored with automatic thermometers placed genes,antibiotic fermentation waste contains resistance genes to other at different depths of the hyperthermophilic composting piles and three antibiotics due to colocalization in multidrug resistance plasmids replicate piles were used for the composting experiment,which was run (Gonzalez-Plaza,2019).Our previous study focusing on composting of for 31 days. sewage sludge using hyperthermophilic composting (periodic tem- peratures reaching almost 90C)demonstrated efficient removal (89%) 2.2.Sample collection and physicochemical analysis of ARGs potentially due to a reduction in the horizontal transfer of ARGs in bacterial communities (Liao et al.,2018).However,it is un- To investigate the effect of time on the removal of ARGs during clear if hyperthermophilic composting is efficient at removing anti- composting,samples were collected at the beginning (DO)and after 4 biotic residues and ARGs in much more concentrated TRF waste.We (D4),7D7),13(D13),18(D18),25(D25)and31D31)days since the also lack a deeper and causal understanding of underlying mechanisms start of composting.This temporal sampling data was divided into early behind ARG removal by hyperthermophilic composting,or which spe- and late phases of composting based on temperature differences as re cific types of MGEs (plasmids,integrons or transposons)or host bac- ported earlier (An et al.,2012).Briefly,the early and late phases of terial taxa are important for the maintenance of ARGs during com- composting were split by samples before (D4 to D7)and after (D18 to posting.We hypothesized that,first,hyperthermophilic composting D31)day 13,which was considered as the 'middle point'based on the could be efficient at reducing ARGs by breaking down tylosin residues, maximum temperature reached during the composting which has been shown to occur faster at high temperatures (Yu et al. Samples were collected using a previously described protocol (Liao 2019).Second,high temperatures are likely to kill most of the non- et al.,2018).To obtain a uniform sampling distribution and re- thermophilic bacteria that carry ARGs leading to a reduction in ARG presentative samples at each time point,each pile was diagonally di- abundances.This process should also lead to a reduction in the abun- vided into 5 domains and each domain was sub-sampled (5000 g)from dance of MGEs,which could further decrease the horizontal transfer of upper,central and lower regions of the composting pile.After sampling, ARGs between remaining surviving bacterial taxa.As a result,we ex- each sample was mixed well and divided into two parts of which one pected to see changes in the composition and abundances of bacterial was shock-frozen in liquid nitrogen for biological analyses and the communities,ARGs and MGEs during the composting. other kept at 4'C for physicochemical analysis.The physicochemical To study these questions,we conducted a replicated full-scale hy- properties including pH,temperature (Temp),water content (WC), perthermophilic composting experiment of TFR waste,where we electrical conductivity (EC),total nitrogen content (TN),total carbon compared the dynamics of tylosin residues,heavy metals,ARGs and content (TC),total organic carbon content (TOC),ammonium (NH4), MGEs and changes in bacterial community composition during early and nitrate (NO3 concentrations were measured as described pre- (4-13 days)and late (18-31 days)phases of hyperthermophilic com- viously (Liao et al.,2018).DTPA-extractable heavy metals (nickel posting.We used temporal sampling followed by quantification of the (Ni2),copper (Cu2+),cobalt (Co2),zinc (Zn2),and plumbum abundance of tylosin residues and bio-active heavy metals,and quan- (Pb2+))were defined as bio-available heavy metals and analyzed as titative PCR to determine the abundances of 27 ARGs and three types of described previously (Guo et al.,2018).More detailed measurement MGEs (plasmids,integrons and transposon).Furthermore,we applied protocol for determining bio-availability of heavy metals is included in 16S rRNA gene amplicon sequencing to determine the composition of the supplementary file. the ARG and MGE-associated bacterial communities and used direct culture assays to validate the presence and location of ARGs in chro- 2.3.LC-MS/MS analysis of tylosin content mosomes and plasmids in the beginning and at the end of the com- posting experiment.The specific objectives of our study were (1)to Changes in tylosin residue quantities were determined using a investigate the feasibility of hyperthermophilic composting in removing previously described Liquid Chromatography-Mass Spectrometry (LC- tylosin residues and associated ARGs and MGEs at an industrial scale, MS/MS)method with some modifications (Zhang et al.,2018a).Briefly, (2)to understand the underlying mechanisms behind ARG removal and tylosin was extracted from 1.0 g composting samples using 5 mL 90% (3)to establish potential links between abiotic (composting properties, acetonitrile aqueous solution (v/v,pH 4.0)as follows.The tubes were tylosin residues,heavy metals)and biotic (bacterial community vortexed for 5 min and then sonicated in an ultrasonic bath for 30 min
H.Liao,et al. Environment Intemational 133 C2019)105203 Subsequently,the mixture was centrifuged at 14,000g for 10 min and performed using a Ribosome Database Project Classifier provided by the the supernatant filtered through a 0.45 um filter.Samples were cleaned- Greengenes 13.8 16S rRNA gene database with 80%sequence similarity up by solid phase extraction (SPE)cartridges (HLB,6 cc/500 mg,Wa- threshold as described previously (McDonald et al.,2012).Differences ters,USA)and SPE eluents were concentrated until dry under a gentle in the diversity (beta-diversity)and composition between microbial nitrogen flow and then dissolved in 1.0 mL of methanol.The final communities (PCoA principal coordinate analysis)were analyzed using samples were filtered through a 0.22 um membrane filter (Millex, weighted Unifrac metric distances.Within microbial community di- Millipore Corp.,Billerica,MA),transferred to 1.5mL amber vials,and versity (alpha-diversity)was quantified using Chaol and Shannon di- stored at -20'C before the LC-MS/MS analysis.Quantity of tylosin was versity indexes.All sequences were deposited in the National Center for measured by liquid chromatography combined with ion trap mass Biotechnology Information Sequence Read Archive under the accession spectrometry system (XEVO-TQS,Waters,USA)as described previously number PRJNA551919. (Zhang et al.,2018a).The tylosin standard was obtained from Dr Eh- renstorfer GmbH (Germany).The average recovery rates of the blank 2.6.Isolation and identification of antibiotic resistant bacteria samples spiked with 10-100 ug/kg of tylosin were 93.6%and the de- tection limit of the tylosin was 1.5 ug/kg.All composting samples were Culture-based methods were used to identify potential ARG hosts at analyzed in triplicate to reduce measurement errors. different phases of composting.Isolation and identification culturable antibiotic resistant bacteria was performed as described by Zhang et al. 2.4.DNA extraction and quantitative PCR (qPCR)for determining (2018b).Samples from D4 to D7 and D25 to D31 were selected as re abundances of genes linked with antibiotic resistance and mobile genetic presentative samples of the early and late phases of composting.Briefly, elements 10 g of mixed sample was suspended into 90 mL phosphate-buffer by shaking at 200 rpm for 30 min.The sample was then serially diluted to Total genomic DNA was extracted using a Fast DNA spin kit (MP different concentrations and 100 uL of all dilutions spread on Luria- Biomedicals,Cleveland,OH,USA)from freeze-dried composting sam- Bertani (LB)agar plates containing four antibiotics:tetracycline,ery- ples according to the manufacturer's instructions.In addition to quan- thromycin,gentamicin and sulfadiazine at final concentrations of tifying tylosin macrolide antibiotic resistance gene abundances,three 16 mg/L,10 mg/L,16 mg/L,512 mg/L,respectively (Ren et al.,2018; other common ARGs found in organic waste (Liao et aL.,2018;Zheng Yang et al,2017).After 48 h incubation at 30'C,the number of colony- et al.,2019)were also measured including resistance genes to tetra- forming units (CFU)was determined for each type of agar plates.Based cycline,aminoglycoside and sulfonamide antibiotics.We chose several on colony color and morphology,a total of 29 culturable antibiotic variants of each type of ARGs including 10 tetracycline resistance genes resistant strains (including 10,9,8 and 2 isolates resistant to tetra- (tetA,tetB,tetC,tetG,tetL,tetM,tetQ,teto,tetw,and tetx),7 macrolide cycline,erythromycin,gentamicin,and sulfadiazine,respectively)were resistance genes (ermB,ermF,ermM,ermT,ermX,mefA,and ereA),7 isolated from the early phase composting samples.A total of 21 cul- aminoglycoside resistance genes (aacA4,aadA,aadB,aadE,aphAl,strA, turable antibiotic resistant strains (including 6,5,6 and 4 isolates re- and strB)and 3 sulfonamide resistance genes (sull,sul2,and sul3).From sistant to tetracycline,erythromycin,gentamicin,and sulfadiazine re- here on,all tetracycline,macrolide,aminoglycoside and sulfonamide sistant isolates,respectively)were isolated from late phase composting resistance genes are abbreviated as Tet,Mac,Amin,and Sul,respec- samples.To identify bacterial colonies,we extracted the genomic DNA tively.To investigate potential changes in the abundance of different from all isolates using Bacteria DNA Kit (Tiangen,Beijing,China)and types of MGEs,we chose the following commonly observed MGEs based amplified the 16S rRNA genes using the primers 27F(5-AGAGTTTGA on previously published studies (Ma et al.,2017):2 integrons (intll and TCCTGGCTCAG-3)and 1492R(5-GGCTACCTTGTTACGACTT-3). int2),2 conjugative plasmids (ISCRI and IncQ-oriv,abbreviated as IncQ)and one transposon (Tn916/1545,abbreviated as Tn916).All 2.7.Extraction of plasmid and genomic DNA from antibiotic resistant information about primers,annealing temperatures,reaction conditions bacteria and amplification used for all target genes were validated in a previous study (Liao et al.,2018)and are listed in the supplementary materials To detect if ARGs were located on chromosomes or plasmids in (Table S2 including information about positive and negative controls isolated antibiotic resistant bacteria,genomic and plasmid DNA were and standard curves).The absolute abundances of target genes are extracted using bacteria and plasmid DNA Kit (Tiangen,Beijing,China) presented as copy numbers per gram of sample.The relative abundance following manufacturer's protocol.Residual linear chromosomal DNA of different types of ARGs (Tet,Mac,Amin,and Sul)and MGEs (in- fragments were removed using plasmid-safe ATP-dependent DNase tegrons,plasmids,and transposon)are presented as the proportion of (Epicentre,Madison,WI,USA)treatment for 24 h to 48 h at 37C fol- all detected target genes. lowing a previous method (Kothari et al.,2019).The presence of chromosomal DNA was tested by PCR using 16S rRNA universal pri- 2.5.High-throughput sequencing and bioinformatics analysis exploring mers(BAC338F,5-ACTCCTACGGGAGGCAG-3,and BAC805R,5-GACT bacterial community diversity and composition ACCAGGGTATCTAATCC-3).If a 16S rRNA genes PCR product was visible on a 1%agarose gel,another overnight digestion reaction was The changes in bacterial community composition and diversity performed until the product could no longer be visualized.The DNase during hyperthermophilic composting were determined using 16S was inactivated at 70'C for 30 min.The chromosomal DNA-free rRNA gene amplicon sequencing on an Illumina HiSeq 2500 platform plasmid and genomic DNA extracted from isolates were used to detect (Guangdong Magigene Biotechnology Co.Ltd,Guangzhou,China).The the ARGs and MGEs using specific primers using PCR (Table S2).Strains V4 region of the bacterial and archaeal 16S rDNA gene was amplified isolated from composting pile replicates at the same phases of com- using the primers 515F (5-GTGCCAGCMGCCGCGGTAA-3)and 806R posting were merged to analyze the data of ARGs and MGEs. (5'-GGACTACHVGGGTWTCTAAT-3)(Caporaso et al,2012).Raw Il- lumina sequence data was quality filtered using a pipeline coupling 2.8.Co-occurrence network analysis between different bacterial taxa,ARGs Trimmomatic (version 0.33)and QIIME (1.8.0)(Caporaso et al.,2010). and MGEs Primer sequences and low-quality reads that contained ambiguous nucleotides,mismatches in primer regions,or had a length shorter than Co-occurrence network analysis was used to explore pairwise cor- 100 bp were removed to obtain better sequence read data.Operational relations between bacterial taxa (based on genus level with abun- taxonomic units (OTUs)were defined at 97%sequence similarity level dance >0.1%)and different ARGs and MGEs during composting. using Uclust (Edgar,2010)and taxonomic assignment of OTUs was Pearson and Spearman correlations were extracted using PAST software
H.Liao,et al. Environment International 133 (2019)105203 v3.04 as described previously (Liu et al.,2019).Only relatively large Ni2+and Zn2+also with integron abundances (intll and intl2; correlation coefficients (p 0.8 and P<0.01)detected with both Spearman,P 0.05,Table 1).Together these results suggest that the methods(Pearson and Spearman)were included in network analyses to reduction in tylosin residues and bio-available heavy metals was linked minimize false-positive correlations.Furthermore,Benjamin Hochberg with elevated composting temperature and reduction in the abundance multiple tests (q-value,q 0.01)were used to adjust P-values to re- of MGEs during hyperthermophilic composting. duce false-discovery rates.Co-occurrence networks were visualized using Gephi v0.9.2. 3.2.The effect of hyperthermophilic composting on the abundance and diversity of ARGs and MGEs 2.9.Statistical analyses All targeted genes,including 27 ARGs and 5 MGEs,were detected in Analysis of variance (ANOVA)and linear regression was used to test the initial tylosin residue waste and all the samples collected during the for differences in the abundances of ARGs and MGEs and the diversity composting (Fig.S1).The mean ARG and MGE abundances in the initial of the bacterial community between different phases of composting (initial TFR vs.early and late phases)using OriginPro 9.4 (OriginLab compost were approximately 5.9 x 1012 and 2.2 x 1012 copies per Corporation,Northampton,MA).Partial Least Squares Path Modeling gram of compost (dry weight),respectively (Fig.2a).Because the initial (PLS-PM)was used to study relationships between physicochemical rice husk contained only very low amounts of ARGs(0.25%of the mean composting properties (WC,TC,TN,EC,TOC,C/N,and NOa )the amount of TFR),it was not included in further analyses.To study these amount of tylosin residues,bio-availability of heavy metals (Ni2, patterns in more detail,we compared changes in ARGs and MGEs Cu2+,Co2+,Zn2+,and Pb2+),bacterial community composition during the early and late phases of composting.We could not detect (based on OTUs abundance composition value)and absolute abun- statistically significant reductions in ARGs and MGEs during the early phase of composting relative to the initial TFR waste (all P 0.05). dances of MGEs and ARGs.PLS-PM is a powerful statistical method to However,75.8%and 98.5%reduction in the total abundances of ARGs study relationships among observed and latent variables (Wagg et al., and MGEs were observed between the early and late phases of com- 2014),where path coefficients (i.e.standardized partial regression posting (Fig.2a-b,P<0.01).Even though the absolute ARG abun- coefficients)represent the direction and strength of linear relationships between variables (direct effects).Indirect effects are the multiplied dances decreased significantly,the proportion and diversity of different path coefficients between a predictor and a response variable,adding types of ARGs did not change drastically (Fig.2c);while some temporal dynamics were observed in the proportion of all ARGs,the Mac and Tet the product of all possible paths excluding the direct effect.Models with different structures were evaluated using the goodness of fit (GoF) type ARGs were the most dominant accounting for 49%to 80%of all ARGs during the composting (Fig.2c).In contrast,the proportion of statistic,a measure of their overall predictive power of a given model (Cui et al,2016;Wagg et al.,2014).PLS-PM was also chosen instead of MGEs changed more drastically during the composting.Specifically,the percentage of plasmid genes decreased from 98%to 9%,while integron structural equation modeling because it is more appropriate for data sets with small sample sizes and less sensitive to the sampling dis- and transposon genes became the dominant MGEs during the com tribution of PLS weights.The R package plspm (v 0.4.7)was used to posting accounting for 91%of all MGEs (Fig.2d).Furthermore,we construct the final PLS-PM model.Canonical correspondence analysis found a strong correlation between MGEs,specifically ISCRI plasmid gene,and total ARGs abundances (R=0.7,P 0.0001),which im- (CCA),Adonis test and Procrustes analysis were performed in R 3.5.1 plies that ARGs were likely carried in plasmids (Fig.S2). using the vegan package v2.4-3 and labdsv v1.8.Linear discriminant analysis effect size analysis (LEfSe)was used to determine differentially abundant taxa between different stages of composting using the Galaxy 3.3.Changes in the bacterial community composition during web application (Segata et al.,2011).All data are presented as mean hyperthermophilic composting values+1 standard error. We next compared changes in bacterial community composition and 3.Results diversity between initial TFR waste and early and late phases of hy- perthermophilic composting.Based on the PCoA analysis (weighted 3.1.Hyperthermophilic composting is effective at removing tylosin residues Unifrac distances),no difference was found between initial and early and ARGs phase composting samples (Adonis test,P >0.05,Fig.S3a).However, early and late composting phase samples showed distinct clustering The initial composting TFR waste contained high levels of tylosin indicative of a difference in bacterial community composition (Adonis residues(85.0 mg/kg)and bio-available heavy metals (987.4 mg/kg). test,P=0.01,Fig.S3a).Similarly,total bacterial abundances (ob The hyperthermophilic composting method was effective at removing served OTUs),community richness (Chaol index)and community di- 95.0%of tylosin residues and 88.9%of bio-available heavy metals in versity (Shannon index)differed between early and late phases of 31 days (Fig.1a).The maximum composting temperature increased to composting (P 0.01,Fig.S3b-d),while no differences were found about 80'C after 13 days (Fig.1b)which coincided with the reduction between initial and early phase samples (P>0.05).Early and late of both tylosin residues and bio-available heavy metals (Fig.la-b).To phase composting samples clustered distinctly also at phylum and further understand how tylosin residues and heavy metals influenced genus levels (unsupervised hierarchical clustering based on the relative ARG and MGE abundances during composting,we explored their re- abundance of most prevalent taxa (>1%in any given sample)).Spe- lationships using correlation analysis.We found no significant corre- cifically,early phase communities had typically high relative abun- lation between the amount of tylosin residues and total ARGs abun- dances of Proteobacteria,Chloroflexi,OP11 and Thermi phyla and dances (P 0.05,Fig.1c).However,the amount of tylosin residues Psychrobacter,T78,and Methanosarcina,Ignatzschineria genera,whereas correlated positively with total MGE abundances and especially with late phase samples were enriched with Firmicute and Actinobacteria ISCRI and IncQ plasmid gene abundances (P 0.001,Fig.1d).Simi- phyla and Georgenia,Yaniella,Alcaligenes,Pseudomonas,Staphylococcus, larly,all heavy metals (Ni2+,Cu2+,Co2+,Zn2+,and Pb2+)sig- Bacillus genera (Fig.3).These bacterial community composition dif- nificantly correlated with MGE abundances (Spearman,P 0.05).In ferences were further confirmed using linear discriminant analysis ef- contrast,the majority of heavy metals (Ni2+,Cu2+,Co2+,Zn2+,and fect size analysis (Fig.S4).Together,these results suggest that changes Pb2)did not significantly correlate with total ARGs abundances in ARG and MGE abundances were linked with a reduction in bacterial (Spearman,P>0.05,Table 1).However,all detected heavy metals abundances and changes in the diversity and taxonomic composition of correlated significantly with plasmid abundances (ISCRI and IncQ)and composting communities
H.Liao,et al. Environment Internarional 133 (2019)105203 (a) (b) 100 100 --Tylosin residues Composting pile ·-Heavy metals 1000 ●-Amibent 80 E 80 800 E 60 600 (6x/6w) 40 40 400 6 20 20 200 0 0 (c) 0 510152025 30 (days) 0 510 15202530(days) Early phase Late phase d Early phase Late phase 6o 14.0 15 ●1SR1,R2=0.7100,P<0.0001 13.5 (601) ● ● 14 ▲ncQ,R2=0.6815.P<0.0001 13.0 13 12.5 seue6 plwseld ●●● 12 ●● ●● ● 12.0 44 11.5 10 AA 11.0 91 10.5 A Total MGEs,R2=0.7855,Pe0.0001 Total ARGs R2-0.2422,Pa0.C601 10.0 0 20 0 60 80 100 0 20 40 60 80 100 Amount of tylosin residues(mg/kg) Amount of tylosin residues(mg/kg) Fig.1.Changes in tylosin residues and bio-available heavy metals(a)and composting temperature during hyperthermophilic composting(b).Correlation analyses between tylosin content and the abundance of total ARGs and total MGEs (c)and plasmid genes(d)based on linear regression models.All absolute abundances of the target genes were normalized using logarithmic transformation (log10)and bars in panels (a)and (b)show1 standard error of mean. 3.4.Correlation between ARG,MGE and bacterial taxa abundances The co-occurrence patterns between ARGs,MGEs and bacterial taxa abundances were further compared using correlation-based co-occur- Based on procrustes analysis,changes in resistomes (based on all rence network analysis.The networks showed a clear shift between ARGs and MGEs)were significantly correlated with bacterial commu- early and late phases of composting mainly due to differences in bac- nity composition at genus level (P=0.0017,M2 0.5537,R 0.6681. terial diversity and community composition (Fig.4).The co-occurrence 999 permutations,Fig.S5a).Similarly,changes in MGEs (based on all network constructed at early phase of composting was larger and more types of MGEs)and bacterial community composition were sig- connected compared to the late phase network(Fig.4).Also,a larger nificantly correlated (P 0.0185,M=0.6940,R 0.5531,999 per- number of nodes and edges were included in early versus late com- mutations,Fig.S5b).As expected,all types of ARGs (tetracycline,sul- posting phase network and several network indices such as network fonamide,aminoglycoside and macrolide gene)and MGEs (plasmids, diameter,network density,network modularity,average path length, integrons and transposon)were significantly correlated with their as- and average degree were greater for early compared to late phase sociated bacterial community (all P 0.05,999 permutations,Fig. network.Based on a previous study (Li et al.,2015),we hypothesized S6). that non-random co-occurrence patterns between ARGs and microbial Table 1 Correlation analysis between the amount of bio-available heavy metals,individual MGEs and total MGEs and ARGs during hyperthermophilic composting. MGEs type Ni2+ Cu2+ Co2+ Zn2+ P%3+ T-metals i Integron -0.68*t 0.33 -053 0.55* 0.38 0.55 int2 Integron 0.5744 -0.24 0.41 -0.66* -0.02 -0.66 Tn916 Transposon -0.6 0.63 -0.62 -0.49 0,81 0.49 ISCR1 Plasmid 0.774 -0.63 0.684 0.864 -0.48 0.86 IncQ Plasmid 0.78* -0.62* 0.71* 0.86* -0.52 0.85* T-MGEs 0.774 -0.64* 0.69*4 0.86*4 -0.5 -0.8644 T-ARGs -0.30 0.20 -0.27 0.49 0.05 0.39 Notes:Spearman's rank order correlation analysis was used based on absolute target gene abundances with following significances: Significant at P 0.05. *Significant at P 0.01. **Significant at P<0.001.T-metals,T-MGEs and T-ARGs denote for total concentration of bio-available heavy metals,total abundance of MGEs and total abundance of ARGs,respectively