Impact of wastewater treatment processes on antimicrobial resistance genes and their co-occurrence with virulence genes in Escherichia coli

w a t e r r e s e a r c h 5 0 ( 2 0 1 4 ) 2 4 5 e2 5 3

Available online at www.sciencedirect.com

ScienceDirect journal homepage: www.elsevier.com/locate/watres

Impact of wastewater treatment processes on antimicrobial resistance genes and their cooccurrence with virulence genes in Escherichia coli Basanta Kumar Biswal a, Alberto Mazza b, Luke Masson b, Ronald Gehr a, Dominic Frigon a,* a b

Department of Civil Engineering and Applied Mechanics, McGill University, Montre´al, Que´bec H3A 0C3, Canada National Research Council of Canada, Montre´al, Que´bec H4P 2R2, Canada

article info

abstract

Article history:

An increase in the frequency of antimicrobial resistance genes (ARGs) in bacteria including

Received 27 July 2013

Escherichia coli could be a threat to public health. This study investigated the impact of

Received in revised form

activated sludge and physicochemical wastewater treatment processes on the prevalence

29 November 2013

of ARGs in E. coli isolates. In total, 719 E. coli were isolated from the influent and effluent

Accepted 30 November 2013

(prior to disinfection) of two activated sludge and two physicochemical municipal treat-

Available online 12 December 2013

ment plants, and genotyped using DNA microarrays. Changes in the abundance of ARGs in the E. coli population were different for the two treatment processes. Activated sludge

Keywords:

treatment did not change the prevalence of ARG-possessing E. coli but increased the

Antimicrobial resistance genes

abundance of ARGs in the E. coli genome while physicochemical treatment reduced both

Escherichia coli

the prevalence of ARG-carrying E. coli as well as the frequency of ARGs in the E. coli genome.

Activated sludge

Most E. coli isolates from the four treatment plants possessed ARGs of multiple antimi-

Physicochemical

crobial classes, mainly aminoglycoside, b-lactams, quinolone and tetracyclines. In addition

DNA microarray

these isolates harboured DNA insertion sequence elements including integrase and

Insertion sequence elements

transposase. A significant positive association was found between the occurrence of ARGs and virulence genotypes. ª 2013 Elsevier Ltd. All rights reserved.

1.

Introduction

An increase in the prevalence of antimicrobial resistant bacterial strains in the environment is a major concern worldwide because of the likely transmission of antimicrobial resistance genes (ARGs) to both non-pathogenic and pathogenic strains (Bouki et al., 2013). There is evidence indicating that the human intestine is an important habitat for ARG transfer in bacteria (Salyers et al., 2004). Municipal wastewater

treatment plants (WWTPs) are also likely to be important habitats because of the continuous influx of ARGs and antibiotics (Rizzo et al., 2013). Although numerous studies have been conducted on the antimicrobial resistance profile of Escherichia coli originating from municipal wastewaters (e.g. Czekalski et al., 2012; Finch and Smith, 1986; Lefkowitz and Duran, 2009; Mezrioui and Baleux, 1994), little is known on the effects of various wastewater treatment processes on the transfer of ARGs among the E. coli population and on the mechanisms leading to the phylogenetically distant

* Corresponding author. Tel.: þ1 514 398 2475; fax: þ1 514 398 7361. E-mail address: [email protected] (D. Frigon). 0043-1354/$ e see front matter ª 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.watres.2013.11.047

246

w a t e r r e s e a r c h 5 0 ( 2 0 1 4 ) 2 4 5 e2 5 3

dissemination of ARGs. A significant fraction of the antimicrobial agents used in various human and animal treatments is received by WWTPs either unchanged or partially metabolized (Zhang and Li, 2011). Though the concentrations of antimicrobial agents in wastewaters are relatively low (mostly in the order of ng/L to a few mg/L) (Zhang and Li, 2011), these antimicrobial levels are sufficient to exert selective pressure for resistance development among pathogenic and nonpathogenic E. coli (Tello et al., 2012). Therefore, further studies comparing the dynamics of ARGs in wastewater treatment processes are necessary to support the design of new processes capable of controlling the dissemination of ARGs. Metagenomics revealed that Escherichia is among the 30 most commonly observed genera found in the human gut microbiome, but it does not dominate the flora as previously believed (Arumugam et al., 2011). Although normally considered as a commensal, several strains of E. coli cause human and animal diseases, both intra-intestinally (e.g., E. coli O157:H7) and extra-intestinally (e.g., uropathogenic E. coli, UPEC) (Kaper et al., 2004). Pathogenic E. coli are good candidates to carry ARGs from the external environment back to the human gut for several reasons. First, they are well adapted to grow in the gut environment and to grow outside the host (Winfield and Groisman, 2003). Second, ARG-carrying pathogenic E. coli will survive antimicrobial treatment of infections longer than non-resistant strains, which will increase the likelihood of resistance gene transmission to the rest of the microbial community. Third, extraintestinal virulence factors provide competitive advantages for intestinal colonization (Diard et al., 2010). Consequently, extraintestinal pathogens carrying ARGs would more likely transmit these genes to other members of the intestinal flora due to their longer residence time than non- extraintestinal pathogens. Finally, cooccurrence of ARG and virulence genes has been observed in studies of clinical isolates (Lee et al., 2010; Vila et al., 2002). Therefore, understanding variation in co-occurrence of ARGs and virulence genes in E. coli through wastewater treatment systems is important. Phenotypic or genotypic methods have been used to monitor the levels of E. coli-resistant strains. Phenotypic studies indicated that wastewater treatment processes increase the prevalence of E. coli strains resistant to both single and multiple antimicrobial classes (Finch and Smith, 1986; Lefkowitz and Duran, 2009; Mezrioui and Baleux, 1994). To our knowledge, only a few qPCR-based genotyping studies have examined the change of ARG abundance in the microbial biomass of biological treatment systems, usually through the detection of ARGs of some selected antimicrobials (e.g. Czekalski et al., 2012), and no studies have been published on the effect of physicochemical treatment. Furthermore, more than 40% of the municipal wastewater in the province of Quebec is treated by physicochemical processes. The current study goes beyond the previous studies by testing for the presence of 70 antimicrobial resistance genes against 11 antimicrobial classes in E. coli isolates from both biological and physicochemical treatment systems. This study complements a previously published study on the same isolate collection (Frigon et al., 2013). By microarray genotyping, it had been found that extraintestinal

uropathogenic E. coli (UPEC) and intestinal shiga-toxin producing E. coli (STEC) were the main pathotypes in municipal wastewaters. Comparing influent and effluent samples before disinfection, it had been determined that both activated sludge and physicochemical treatment processes reduced the prevalence of pathogenic E. coli (Frigon et al., 2013). Similarly to the virulence genes, the prevalence of antimicrobial resistance genes may change through wastewater treatment processes, which is of interest in the current study. The objectives of the current study were (1) to compare the impact of conventional activated sludge and physicochemical wastewater treatment processes on ARG prevalence in E. coli isolates, and (2) to assess the co-occurrence of ARGs and virulence genes in these E. coli. To meet these objectives, ARG and virulence gene composition data were generated using a customized DNA microarray containing 70 ARGs representing 11 commonly used antimicrobial classes, 195 virulence genes, as well as 8 genes encoding DNA insertion sequence elements (ISEs) including integron and transposon markers. This rapid microarray genotyping technique allowed us to comprehensively determine the incidence of ARG and virulence genes in E. coli isolates from the influents and effluents of two activated sludge and two physicochemical WWTPs.

2.

Materials and methods

2.1. coli

Wastewater quality analysis and enumeration of E.

The data presented in this study are from the same E. coli isolate collection published previously. A detailed description of wastewater sampling, characterization of samples and enumeration of E. coli was reported in that publication (Frigon et al., 2013). Briefly, grab samples (influent and effluent without disinfection) were collected from two activated sludge treatment plants (called AS1 and AS2) and two physicochemical treatment plants (called PC1 and PC2). E. coli enumeration was performed at 44.5  C on mFC media using the standard membrane filtration technique (APHA. et al., 2005). Confirmation of E. coli was made using Chromocult agar (EMD chemicals, Germany) and Kovac’s reagent (EMD chemicals, Germany). After isolation, between 83 and 93 confirmed E. coli from the influent and effluent of each treatment plant (a total of 719 isolates) were used for DNA microarray genotyping. E. coli isolates which showed the presence of the uidA (b-glucuronidase) gene by DNA microarray were included for the analysis.

2.2. Genomic DNA labelling, microarray hybridization and data analysis The current study used the additional data from the same DNA microarrays that were reported previously (Frigon et al., 2013), where a detailed description on the microarray methodology was also given. Briefly, DNA was extracted from 1 mL of overnight LB cultures then labelled with Cy5-dCTP (GE Healthcare, Little Chalfont, UK) by DNA polymerase followed by proper quality controls. Overnight hybridization of the

w a t e r r e s e a r c h 5 0 ( 2 0 1 4 ) 2 4 5 e2 5 3

microarray was performed at 50  C with supplemented DIG Easy Hyb buffer (Roche Diagnostics, Laval, Quebec, Canada), and three 5-min post-hybridization washes were performed at 37  C in 0.1  SSC/0.1% SDS. Microarray hybridization signals were finally obtained using a microarray scanner and image processing software. Average duplicate signals for each probe were divided by the average background signal, and a ratio greater than 3 was considered positive. Following this approach, the same isolate analysed more than once would show the same gene profile for each hybridization. The current microarray (70-base pair probes) contains 413 probes including 311 probes targeting different alleles and versions of 195 virulence or virulence-related genes, 70 probes detecting ARGs of 11 antimicrobial classes, 8 probes for DNA insertion sequence elements (ISEs), 20 positive-control probes and 5 negative control probes. A complete list of probes encoding ARGs and ISEs is given in the supplementary material (Table S1), while their sequences can be found in Jakobsen et al. (2011). E. coli isolates which did not possess any of the tested ARGs were presumed to be non-antimicrobial resistance isolates. A list of virulence gene probes, their sequences and the rules to identify pathotypes was published previously (Frigon et al., 2013; Hamelin et al., 2007; Jakobsen et al., 2011). Specifically for the current report, shiga toxin carrying E. coli (STEC) were identified as such if one of the two shiga toxin (stx1/stx2) genes were detected, while uropathogenic E. coli (UPEC) pathotypes were identified as such if a combination of five virulence genes among four virulence factors (adherence: 2 genes, capsule: 1 gene, iron uptake: 1 gene and toxins: 1 gene) were detected. The presence of STEC and UPEC pathotypes was assessed simply based on the combination of virulence genes (Frigon et al., 2013). Phylogenetic classification was performed following Clermont et al. (2000). Uropathogenic E. coli (UPEC) pathogenicity islands (PAIs) were determined based on the combination of virulence genes (Table S2).

2.3.

247

Fig. 1 e Proportion of (a) antimicrobial resistance gene (ARG)-carrying E. coli and (b) the average number of ARG classes carried in E. coli from the influents (Inf) and effluents (Eff) of activated sludge (AS1 and AS2) and physicochemical (PC1 and PC2) treatment plants. Error bars represent the standard error as calculated using the loglinear model. The value in parentheses represents the number of E. coli isolates analysed.

Statistical analyses

Statistical significance of the variation in the proportions of ARG-carrying E. coli isolates between treatment plants/processes was evaluated using tools for categorical data. The loglinear model (LLM) was used to test the variations in frequencies through the samples (Sokal and Rohlf, 1994). Computations were performed with the CATMOD procedure of the SAS statistical software (SAS version 9.2, SAS Institute, Inc., Cary, NC). Each treatment plant is considered as an independent sample and all the statistical tests were two tailed. Due to the high variability of wastewater qualities and complex flow patterns in wastewater treatment plants, the statistical significance was evaluated at a P-value <0.10, and all references to statistical significance in the material which follows is with respect to this level.

3.

Results

3.1.

Prevalence of ARG-carrying E. coli

The frequency of ARG-carrying E. coli isolates in the wastewater influents varied between 24 and 65% (Fig. 1a). The slight

increases in the proportions of ARG-carrying E. coli observed through the activated sludge plants were not significant. Conversely, the reductions observed at both physicochemical plants, with an average of 27%, was significant. The difference in proportions (slight increase at activated sludge plants vs decrease at physicochemical plants) between the two types of treatment processes was significant. Furthermore, on comparing changes in the frequency of the average number of ARG classes, it is found that the average number increased in both the activated sludge plants, while it decreased in the two physicochemical plants (Fig. 1b). A majority of ARG-carrying isolates (>68%) possessed several genes conferring resistance to multiple antimicrobial classes (multiple ARG classes isolates), and as high as nine ARG classes were detected in the isolates from influents and effluents of three of the four locations (see Table S3). The average number of ARG classes in those isolates carrying them actually increased by 15.3% in the activated sludge plants, while it decreased by 11.8% in the physicochemical plants (Table 1). Because the experiment was designed to analyse categorical data, it was not possible to test the statistical significance of these differences. Instead, we tested changes in the frequencies of each ARG class (i.e. class-by-

248

w a t e r r e s e a r c h 5 0 ( 2 0 1 4 ) 2 4 5 e2 5 3

Table 1 e Occurrence of antimicrobial resistance genes (ARGs) in various antimicrobial classes. Isolates/ antimicrobial class Isolates ARG-carrying Mean ARG class Aminoglycoside Beta-Lactams Macrolides Phenicols QAC Quinolones Sulfonamides Tetracyclines Trimethoprim Olaquindox a b c d e f g h

No. of E. coli isolates carrying antimicrobial resistance gene a

b

Activated sludge c

d

Physicochemical e

Inf

Eff

% Change

Inf

Eff

% Change

182 58 3.0 25 32 15 15 8 37 13 28 8 0

180 59 3.5 30 35 23 20 6 32 17 35 15 2

NAg þ2.9 þ15.3 þ21.3 þ10.6 þ55.0 þ34.8 24.2 12.6 þ32.2 þ26.4 þ89.6 NDh

174 81 3.3 39 47 24 24 13 46 18 35 22 0

183 66 2.9 25 35 10 15 7 32 17 33 15 1

NA 22.5 11.8 39.1 29.2 60.4 40.6 48.8 33.9 10.2 10.4 35.2 ND

Stat. sig. of difference in frequencyf Overall Inf/ Variation in Inf/Eff changes Eff change between processes NA e NA e e e e e þ e e e ND

NA þ NA þ e þ þ e e e e þ ND

Activated sludge includes both AS1 and AS2 plants. Physicochemical includes both PC1 and PC2 plants. Inf: Influent. Eff: Effluent. % Change: increase (þ) or decrease (). Significant (P < 0.10) changes were italicized. Stat. Sig (statistical significance): P < 0.10: þ, P > 0.10: . NA: Not applicable. ND: Not determined due to the low occurrence (i.e., frequency close to 0).

class) because this provided categorical data in line with the experimental design. This test found significant differences in the variations of ARG classes through the two types of treatment processes (Table 1). Thus, the observed ARG dynamics can be summarized as follows: isolates carrying multiple ARG classes tend to increase through activated sludge plants, while isolates carrying at least one ARG tend to decrease through physicochemical plants. As a complement to this analysis, the gene frequency is presented in Table S4. A closer analysis of the frequency of each ARG class revealed that of the 11 antimicrobial classes tested, 10 were detected; only the ARG of the rifampicin class were not (Table 1). The prevalence of ARG-carrying isolates was higher for four classes: aminoglycoside, beta-lactams, quinolones, and tetracyclines. On assessing the impact of treatment systems on changes in the frequency of the 10 ARG classes, the observed pattern shows that the frequency of seven ARG classes increased (þ10.6% to þ89.6%) through the activated sludge plants whereas it decreased (10.2% to 60.4%) through the physicochemical plants. The differences in behaviour between the processes (increasing at activated sludge plants and decreasing at physicochemical plants) were statistically significant for four of the seven classes: aminoglycosides, macrolides, phenicols and trimethoprim (Table 1). Notably, quinolone and quaternary ammonium compounds (QAC) resistance genes were reduced in both treatment types, and the overall reduction for quinolone ARGs was significant (Table 1). A total of 67 combinations of ARG classes were observed in multiple ARGs-carrying isolates (Table S3). The most common combinations were: aminoglycoside  beta-lactams, betalactams e tetracyclines, beta-lactams e quinolonones, quinolonones e tetracyclines, aminoglycoside  beta-lactams e

tetracyclines, and aminoglycoside  beta-lactams e quinolones. The influents and effluents carried different combinations of multiple class resistance genes (Table S3), therefore WWTP systems may be selecting for different ARG combinations.

3.2. Co-occurrence of ARGs with DNA insertion sequence elements (ISEs) The occurrence of ISEs such as class 1, 2 and 3 integrons (intI13), transposon Tn 21 (tnpM), and other markers included in the class 1 integron (qacED1-sulI and 30 conserved sequence [30 CS]) was determined in the E. coli isolates, and their enrichment in non-ARG and ARG-carrying (single ARG and multiple ARG classes) E. coli isolates was evaluated. Among the 719 E. coli isolates, the integrons and the transposon were detected in 12.8% and 5.8% of the isolates, respectively, while the qacED1sulI and 30 CS integrons were found in 7.2% and 11.7%, respectively. Among the integrases, the prevalence of classes 1 and 2 was the same (10%), while the class 3 integrase was detected in only 2.1% of the isolates. The frequency of integrase genes increased on average by 2.9% (1.1%e6.1%) through the activated sludge processes, while it decreased on average by 32.4% (23.9%e38.2%) through the physicochemical plants. The transposase, qacED1-sulI genes and 30 CS marker decreased in both types of treatment systems by an average of 21.8% (data not shown). For the transposase gene only, the difference between the activated sludge and physicochemical processes was statistically significant. Of the 719 E. coli isolates analysed, 455 did not contain any ARGs (“zero ARG isolates”) while 101 were single ARG class and 163 were multiple ARGs classes. On examining the distribution of ISEs, very few zero ARG isolates (<1%) carried ISEs,

w a t e r r e s e a r c h 5 0 ( 2 0 1 4 ) 2 4 5 e2 5 3

249

single ARG isolates carried low levels (<6%) while ISEs were abundant (24e47%) in the multiple ARG isolates (Fig. 2). The difference in the abundance between multiple ARG and zero ARG/single ARG isolates was statistically significant.

3.3.

Co-occurrence of ARG and virulence genotypes

The current dataset had been previously analysed for extraintestinal and intra-intestinal pathotypes (Frigon et al., 2013). Of the 719 isolates, 29% were assigned a pathotype. The extra-intestinal uropathogenic E. coli (UPEC) was by far the most abundant pathotype at 18%, while shiga toxin producing E. coli (STEC) was the most abundant intra-intestinal pathotype, accounting for 9%. In addition, almost all isolates carried a virulence gene, with average virulence gene abundance per isolate of 30.3 (out of 195 virulence genes). Most of the detected genes were associated with extra-intestinal pathotypes. Therefore, in testing for the co-occurrence of ARGs with virulence determinants, based on those results, we focused on UPEC and STEC pathotypes and genes. Of the 719 isolates tested, over one third (37%) carried at least one ARG. Among them, 39% had a designated pathotype (UPEC and STEC), while the proportion of designated pathotypes among non-ARG isolates was significantly lower at 24% (Table S5). The UPEC and STEC pathotypes were more prevalent among ARG-carrying isolates (Fig. 3a). Beyond the designated pathotypes, the average number of virulence genes in non-ARG isolates was 27.9, while it increased to 34.5 among the ARG-carrying isolates (Table S5). Focussing on the frequency of UPEC and STEC genes encoding specific virulence factors, the frequency of all virulence factors was higher among ARG-carrying isolates, and was significant for three out of six virulence factors (Fig. 3a). A detailed complementary analysis of the prevalence of all 25 UPEC genes and 2 STEC genes is presented in the supplementary material (Table S5). The occurrence of 25 of the 27 genes was higher in all ARGcarrying isolates, and this increase was statistically significant in 14 cases. Furthermore, our data revealed that

Fig. 3 e The co-occurrence of (a) antimicrobial resistance genes (ARGs) and virulence genes (VGs) encoding virulence factors (VFs) of uropathogenic E. coli (UPEC) and shiga toxin producing E. coli (STEC) pathotypes, and (b) ARGs and UPEC PAIs in E. coli isolates carrying no ARGs (non-ARG, 455 isolates) and ARGs (ARG, 264 isolates). Significant occurrences are indicated by an asterisk.

pathogenic E. coli on average carried ARGs of 3e4 antimicrobial classes whereas non-pathogenic E. coli harboured resistance genes of 2e3 antimicrobial classes, and this difference was statistically significant. Virulence genes are often clustered in specific DNA insertion sequence elements (ISEs) called pathogenecity islands (PAIs) (Schmidt and Hensel, 2004). Thus, we also analysed the co-occurence of UPEC PAIs and ARGs. Although UPEC virulence factors and virulence genes were significantly more abundant in the ARG-carrying isolates (Fig. 3a), the same pattern was only observed for the PAI IV536 (Fig. 3b). This PAI was most prevalent and was enriched significantly within the ARG-carrying isolates.

3.4.

Fig. 2 e Prevalence of integrons and transposon in E. coli isolates carrying no antimicrobial resistance genes (NARG, 455 isolates), ARGs of a single antimicrobial class (SARG, 101 isolates) and ARGs of multiple antimicrobial classes (MARG, 163 isolates).

Phylogenetic clustering of E. coli isolates

Phylotyping (A, B1, B2 and D) was performed for E. coli isolates carrying ARGs of zero, one and multiple antimicrobial classes as well as for isolates carrying integrase/transposase (intI/ tnpM) genes (Fig. 4). Overall, E. coli isolates were clustered into four phyloclasses, but the frequency of E. coli isolates affiliated with group D was lower (714%) than the frequency for the other three groups (21e46%). On comparing the distribution of

250

w a t e r r e s e a r c h 5 0 ( 2 0 1 4 ) 2 4 5 e2 5 3

Fig. 4 e Phylotyping of E. coli isolates containing zero antimicrobial resistance genes (NARG, n [ 455), ARGs of single antimicrobial class (SARG, n [ 101) and ARGs of multiple classes (MARG, n [ 164) as well as E. coli isolates without (intI/tnpM negative, n [ 589) or with (intI/tnpM positive, n [ 130) integrase/transposase genes. “n” refers to the number of E. coli isolates in each class.

phylogroups among no ARG and single ARG/multiple ARG classes-carrying E. coli isolates, the number of isolates affiliated with group A was significantly lower among single ARG/ multiple ARG classes-carrying E. coli isolates, and nearly onethird to half of single ARG/multiple ARG classes-carrying isolates were clustered in groups B2 and D. Similarly, on assessing the distribution of phylogenetic classes among intI/ tnpM-negative and intI/tnpM-positive E. coli isolates, the proportion of isolates belonging to B2 and D groups was significantly higher (40.8%) in the intI/tnpM-carrying E. coli than in the intI/tnpM-missing E. coli (29.7%). Previous investigators have reported similar types of phylogenetic distributions among ARGs or ISEs-carrying E. coli isolates from environmental samples (Mokracka et al., 2011).

4.

Discussion

4.1. Impact of wastewater treatment processes on the prevalence of ARGs The potential for activated sludge and physicochemical wastewater treatment processes to change the abundance of ARGs and ARG-carrying E. coli isolates was evaluated. The plants which were sampled differed not only in their basic treatment processes, but also in key operational parameters such as hydraulic and solids residence times, which could affect the dynamics of genetic exchange. These parameters may impact the distribution of ARGs in E. coli passing through the system (Novo et al., 2013; Novo and Manaia, 2010). Due to the large quantity of genes that needed to be screened on a suitably high number of isolates, it was not possible to characterize more than one influent sample and one effluent sample for each plant. This limits the statistical analyses and the generalizations that can be done. However, the average

trends reported for each process type were observed at the two plants utilizing the same treatment process, which provide some indication of the reproducibility of the results. The observed trends in the prevalence of ARGs can be summarized as follows. In our study, the activated sludge process was found to have no effect on the removal of ARG-carrying E. coli, but it increased the average frequency of multiple ARG classes in the ARG-carrying isolates. The physicochemical process, however, reduced both the number of ARG-carrying E. coli (by an average of 27%) and the average frequency of multiple ARG classes in the ARG-carrying isolates. Thus, overall prevalence of ARGs had increased in the E. coli population which passed through the activated sludge process, and it had decreased through the physicochemical process. It was not possible to find data in the literature on the impact of physicochemical treatment on the prevalence of antimicrobial resistance, but a number of studies have examined variations in resistance levels through biological processes. When comparing different studies of E. coli and Enterobacteriaceae or coliforms, the trends vary greatly with a majority of studies having found increases in the prevalence of antimicrobial resistance or ARGs in these populations (Ferreira da Silva et al., 2007; Finch and Smith, 1986; Lefkowitz and Duran, 2009; Mezrioui and Baleux, 1994; Novo et al., 2013; Reinthaler et al., 2003) through biological treatments, while others have found the reverse (Harris et al., 2013; Novo and Manaia, 2010). The reasons for these differences are not clear. A recent study in Portugal found that, in the treated effluent, the prevalence of Enterobacteriaceae resistant to four antimicrobials was positively correlated with temperature (i.e., the season) and the concentration of tetracyclines in the influent (Novo et al., 2013). The samples analysed in the current study were obtained in the summer; thus, the observed increase in the prevalence of ARGs through activated sludge treatment is consistent with the correlation between temperature and resistance prevalence found in the Portuguese study. The evaluation of any mechanism to explain the variations of ARG prevalence through the treatment processes observed in the current study and the differences observed between processes need to consider the correct time-scale. For the physicochemical process, the residence time of E. coli found in the effluent is the hydraulic retention time (HRT) because biomass is not accumulated. We argue that it is also likely the case in activated sludge systems because E. coli are usually not detectable by high-throughput sequencing of activated sludge microbial communities, suggesting low abundance in the community (Xia et al., 2010). E. coli are typically rapidly removed from wastewater by adsorption and by grazing from protozoa (van Der Drift et al., 1977; We´ry et al., 2008).Thus, the E. coli in the effluent were probably not released from the activated sludge flocs, but rather were either brought in the influent flow or multiplied as planktonic cells in the mixed liquor. Consequently, we considered that the average residence time of E. coli in the effluent of the physicochemical plants was 2.3 h, and it was 18 h for the activated sludge plants. Both of these residence times are long enough for E. coli to grow for a few generations. From these time-scale considerations, two mechanisms often mentioned in the literature can be evaluated: (1) the

w a t e r r e s e a r c h 5 0 ( 2 0 1 4 ) 2 4 5 e2 5 3

selection of antimicrobial resistant strains though differential growth rates, and (2) the lateral transfer of genes (Rizzo et al., 2013). The density of bacterial biomass harbouring potential ARG donors for lateral gene transfer was a few orders of magnitude lower in physicochemical systems. This would have reduced the rates of lateral gene transfers considerably in physicochemical systems compared to activated sludge systems. Furthermore, physicochemical treatment has been shown to be capable of removing certain antimicrobial agents at a high rate (Xu et al., 2007). In this context, it is possible that E. coli which carry ARGs may be counter selected through growth in physicochemical treatment processes. In this view, the ARG dynamics observed herein for the two treatment types would be the result of the balance between selection rates and the lateral gene transfer rates. Nevertheless, more work is needed to reproduce the observed dynamics and identify a satisfactory mechanism. On average, 34% of E. coli isolates collected from the effluents of the four treatment plants contained ARGs against a mean of 3.1 antimicrobial classes (Table 1). Also, genes from nine of the eleven antimicrobial classes tested had average incidences between 5% and 21%. Most of these antimicrobials were among those most widely prescribed to humans in Canada, and were detected in high frequencies in the effluents of Canadian WWTPs (Loeb et al., 2001; Miao et al., 2004). In addition, quaternary ammonium compounds (QACs) are widely used as antiseptics, and resistance to these antimicrobials is believed to favour the resistance to prescription antimicrobials (Hegstad et al., 2010). Notably, only rifampicin ARGs were not detected, while the average frequency of olaquindox ARGs was very low (<0.5%). For rifampicin, it is possible that the lack of detection is due to the main use of this antimicrobial for the treatment of Mycobacterium infections such as tuberculosis and leprosy (Zenkin et al., 2005), rare diseases in Canada. Alternatively, direct mutation of RNA polymerase may also be an important resistance mechanism (Telenti et al., 1993) that would not have been detected by our microarray assay. Olaquindox antimicrobials are less likely to be present in municipal wastewater because they are mainly used as growth promoters for livestock rather than for chemotherapy in humans (Hansen et al., 2004), which may explain the lower frequencies of the associated ARGs.

4.2. Co-occurrence of ARGs and DNA insertion sequence elements (ISEs) Our findings show that those E. coli isolates possessing genes that encode DNA insertion sequence elements (integrons and transposons) also harboured multiple ARGs (Fig. 2). To date, five classes of integrons have been detected in gram-negative bacteria; however, only three of these classes (intI1, 2, and 3) have been reported in environmental E. coli isolates, with class 1 integrons being the most common (Stalder et al., 2012). The abundance of integrase genes is usually lower in environmental (Mokracka et al., 2011) than in clinical strains (Japoni et al., 2008). The proportion of integron-carrying E. coli isolates detected in our study was comparable to those found by Mokracka et al. (2011). In the current work, along with integrase genes, a low number of E.

251

coli isolates carried the transposon Tn21 (tnpM) gene, and these isolates harboured multiple ARGs at significant levels (Fig. 2). The findings of our study suggest that ARGs mobility through ISEs play a role in the acquisition of ARGs by E. coli during biological wastewater treatment, but not during physicochemical treatment.

4.3.

Co-occurrence of ARG and virulence genotypes

Our analysis revealed a significant co-occurrence between ARG and virulence genotypes, i.e. ARG-possessing E. coli carried more virulence genes encoding virulence factors associated with both uropathogenic E. coli (UPEC) and shiga toxin carrying E. coli (STEC) pathotypes (Fig. 3a) than non-ARGcarrying isolates. The associations between b-lactam or quinolone resistance and the prevalence of virulence factors have been the most widely studied among clinical isolates. Production of the b-lactamase enzyme (bla-CTX-M-15 gene) correlates with the enrichment in certain E. coli virulence genes (Lee et al., 2010), while resistance to quinolones is linked to reduced E. coli virulence potential (Vila et al., 2002). Similar relationships were observed in our study for the same antimicrobial resistance classes. The occurrence of pathogenic isolates was significantly higher (48% pathogenic) in isolates carrying b-lactam ARGs than in those isolates not carrying these genes (only 27% pathogenic). In contrast, only 37% of the isolates carrying quinolone ARGs were pathogenic while 63% were non-pathogenic. Only one of the known quinolone resistance genes was tested, but it had been observed previously that, among the isolates carrying that gene, they seemed to have an enhanced loss of virulence genes through the treatment processes, compared to the rest of the E. coli population (Frigon et al., 2013). Except for quinolone resistance, significant co-occurrence of UPEC and STEC virulence factors and ARGs has been generally observed in our dataset (Fig. 3a). However, any correlation between the PAIs of UPECs (the most prevalent pathotype in our tested isolates) and ARGs was not observed for most of the seven PAIs (Fig. 3b), with the exception of the more genomically stable PAI IV536 (Middendorf et al., 2004). This is probably due to the incomplete description of UPEC PAIs in the literature. The enrichment of ARGs among strains carrying UPEC and STEC virulence genes in wastewater isolates suggests that pathogenic E. coli strains are a good potential vector to carry ARGs back to the human microflora. Not only is it clear that antibiotic resistant pathogens will have an advantage over non-resistant ones during chemotherapy, but also UPEC virulence factors are known to favour the establishment of E. coli strains in the intestine (Diard et al., 2010). Further study needs to be conducted to clarify the impact of the treatment processes on the co-occurrence of ARGs and virulence genes. In a previous study, we showed that UPEC virulence genes were lost through both activated sludge and physicochemical treatment processes (Frigon et al., 2013). However, it remains difficult to properly quantify the changes in co-occurrence of ARGs and virulence genes through the processes despite the relatively large isolate collection. This points to the complexity of the association between ARG and virulence genes, which likely depends on various factors including the habitat and ARG type.

252

5.

w a t e r r e s e a r c h 5 0 ( 2 0 1 4 ) 2 4 5 e2 5 3

Conclusions

 Based on the limited number of samples tested, the prevalence of ARGs in the E. coli population appears to increase through the activated sludge treatment system and decrease through physicochemical treatment.  A high percentage of E. coli isolates harboured ARGs of multiple antimicrobial classes (up to nine classes), and these isolates carried a greater number of insertion sequence elements.  ARG-carrying E. coli isolates contained a significantly higher number of virulence genes encoding UPEC and STEC virulence factors. This co-occurrence provides a potential vector for ARGs to become re-associated with the human microflora.

Acknowledgements The authors thank Dana Zheng and Katherine Warren for assistance with enumeration of E. coli and determination of the wastewater sample characteristics. We also acknowledge the plant operators of the four wastewater treatment plants for their support in collecting samples. This study was funded by the Natural Sciences and Engineering Research Council of Canada (NSERC) (grant no. STPGP 35117-07) in partnership with Trojan Technologies Inc.

Appendix A. Supplementary data Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.watres.2013.11.047.

references

APHA, AWWA, WEF, 2005. Standard Methods for the Examination of Water and Wastewater, twenty-first ed. American Public Health Association, Washington, D.C. Arumugam, M., Raes, J., Pelletier, E., Le Paslier, D., Yamada, T., Mende, D.R., Fernandes, G.R., Tap, J., Bruls, T., Batto, J.M., Bertalan, M., Borruel, N., Casellas, F., Fernandez, L., Gautier, L., Hansen, T., Hattori, M., Hayashi, T., Kleerebezem, M., Kurokawa, K., Leclerc, M., Levenez, F., Manichanh, C., Nielsen, H.B., Nielsen, T., Pons, N., Poulain, J., Qin, J., SicheritzPonten, T., Tims, S., Torrents, D., Ugarte, E., Zoetendal, E.G., Wang, J., Guarner, F., Pedersen, O., De Vos, W.M., Brunak, S., Dore´, J., Weissenbach, J., Ehrlich, S.D., Bork, P., 2011. Enterotypes of the human gut microbiome. Nature 473 (7346), 174e180. Bouki, C., Venieri, D., Diamadopoulos, E., 2013. Detection and fate of antibiotic resistant bacteria in wastewater treatment plants: a review. Ecotoxicol. Environ. Saf. 91, 1e9. Clermont, O., Bonacorsi, S., Bingen, E., 2000. Rapid and simple determination of the Escherichia coli phylogenetic group. Appl. Environ. Microbiol. 66 (10), 4555e4558. Czekalski, N., Berthold, T., Caucci, S., Egli, A., Buergmann, H., 2012. Increased levels of multiresistant bacteria and resistance genes after wastewater treatment and their

dissemination into Lake Geneva, Switzerland. Front. Microbiol. 3 (106), 1e18. Diard, M., Garry, L., Selva, M., Mosser, T., Denamur, E., Matic, I., 2010. Pathogenicity-associated islands in extraintestinal pathogenic Escherichia coli are fitness elements involved in intestinal colonization. J. Bacteriol. 192 (19), 4885e4893. Ferreira da Silva, M., Vaz-Moreira, I., Gonzalez-Pajuelo, M., Nunes, O.C., Manaia, C.M., 2007. Antimicrobial resistance patterns in Enterobacteriaceae isolated from an urban wastewater treatment plant. FEMS Microbiol. Ecol. 60 (1), 166e176. Finch, G.R., Smith, D.W., 1986. The survival of antibiotic resistant Escherichia coli in an activated sludge plant. Environ. Technol. Lett. 7 (1e12), 487e494. Frigon, D., Biswal, B.K., Mazza, A., Masson, L., Gehr, R., 2013. Biological and physicochemical wastewater treatment processes reduce the prevalence of virulent Escherichia coli. Appl. Environ. Microbiol. 79 (3), 835e844. Hamelin, K., Bruant, G., El-Shaarawi, A., Hill, S., Edge, T.A., Fairbrother, J., Harel, J., Maynard, C., Masson, L., Brousseau, R., 2007. Occurrence of virulence and antimicrobial resistance genes in Escherichia coli isolates from different aquatic ecosystems within the St. Clair River and Detroit River areas. Appl. Environ. Microbiol. 73 (2), 477e484. Hansen, L.H., Johannesen, E., Burmølle, M., Sørensen, A.H., Sørensen, S.J., 2004. Plasmid-encoded multidrug efflux pump conferring resistance to olaquindox in Escherichia coli. Antimicrobial. Agents Chemother. 48 (9), 3332e3337. Harris, S., Morris, C., Morris, D., Cormican, M., Cummins, E., 2013. The effect of hospital effluent on antimicrobial resistant E. coli within a municipal wastewater system. Environ. Sci.: Process. Impacts 15 (3), 617e622. Hegstad, K., Langsrud, S., Lunestad, B.T., Scheie, A.A., Sunde, M., Yazdankhah, S.P., 2010. Does the wide use of quaternary ammonium compounds enhance the selection and spread of antimicrobial resistance and thus threaten our health? Microb. Drug Resist. 16 (2), 91e104. Jakobsen, L., Garneau, P., Kurbasic, A., Bruant, G., Stegger, M., Harel, J., Jensen, K.S., Brousseau, R., Hammerum, A.M., Frimodt-Møller, N., 2011. Microarray-based detection of extended virulence and antimicrobial resistance gene profiles in phylogroup B2 Escherichia coli of human, meat and animal origin. J. Med. Microbiol. 60 (10), 1502e1511. Japoni, A., Gudarzi, M., Farshad, S., Basiri, E., Ziyaeyan, M., Alborzi, A., Rafaatpour, N., 2008. Assay for integrons and pattern of antibiotic resistance in clinical Escherichia coli strains by PCR-RFLP in Southern Iran. Japanese J. Infect. Dis. 61 (1), 85e88. Kaper, J.B., Nataro, J.P., Mobley, H.L.T., 2004. Pathogenic Escherichia coli. Nature Rev. Microbiol. 2 (2), 123e140. Lee, S., Yu, J.K., Park, K., Oh, E.J., Kim, S.Y., Park, Y.J., 2010. Phylogenetic groups and virulence factors in pathogenic and commensal strains of Escherichia coli and their association with blaCTX-M. Ann. Clin. Lab. Sci. 40 (4), 361e367. Lefkowitz, J.R., Duran, M., 2009. Changes in antibiotic resistance patterns of Escherichia coli during domestic wastewater treatment. Water Environ. Res. 81 (9), 878e885. Loeb, M., Simor, A.E., Landry, L., Walter, S., McArthur, M., Duffy, J., Kwan, D., McGeer, A., 2001. Antibiotic use in Ontario facilities that provide chronic care. J. Gen. Intern. Med. 16 (6), 376e383. Mezrioui, N., Baleux, B., 1994. Resistance patterns of E. coli strains isolated from domestic sewage before and after treatment in both aerobic lagoon and activated sludge. Water Res. 28 (11), 2399e2406. Miao, X.S., Bishay, F., Chen, M., Metcalfe, C.D., 2004. Occurrence of antimicrobials in the final effluents of wastewater treatment plants in Canada. Environ. Sci. Technol. 38 (13), 3533e3541.

w a t e r r e s e a r c h 5 0 ( 2 0 1 4 ) 2 4 5 e2 5 3

Middendorf, B., Hochhut, B., Leipold, K., Dobrindt, U., BlumOehler, G., Hacker, J., 2004. Instability of pathogenicity islands in uropathogenic Escherichia coli 536. J. Bacteriol. 186 (10), 3086e3096.  ska, L., Kaznowski, A., 2011. Mokracka, J., Koczura, R., Jabłon Phylogenetic groups, virulence genes and quinolone resistance of integron-bearing Escherichia coli strains isolated from a wastewater treatment plant. Antonie van Leeuwenhoek, Int. J. Gen. Mol. Microbiol. 99 (4), 817e824. Novo, A., Andre´, S., Viana, P., Nunes, O.C., Manaia, C.M., 2013. Antibiotic resistance, antimicrobial residues and bacterial community composition in urban wastewater. Water Res. 47 (5), 1875e1887. Novo, A., Manaia, C.M., 2010. Factors influencing antibiotic resistance burden in municipal wastewater treatment plants. Appl. Microbiol. Biotechnol. 87 (3), 1157e1166. Reinthaler, F.F., Posch, J., Feierl, G., Wu¨st, G., Haas, D., Ruckenbauer, G., Mascher, F., Marth, E., 2003. Antibiotic resistance of E. coli in sewage and sludge. Water Res. 37 (8), 1685e1690. Rizzo, L., Manaia, C., Merlin, C., Schwartz, T., Dagot, C., Ploy, M.C., Michael, I., Fatta-Kassinos, D., 2013. Urban wastewater treatment plants as hotspots for antibiotic resistant bacteria and genes spread into the environment: a review. Sci. Total Environ. 447, 345e360. Salyers, A.A., Gupta, A., Wang, Y., 2004. Human intestinal bacteria as reservoirs for antibiotic resistance genes. Trends Microbiol. 12 (9), 412e416. Schmidt, H., Hensel, M., 2004. Pathogenicity islands in bacterial pathogenesis. Clin. Microbiol. Rev. 17 (1), 14e56. Sokal, R.R., Rohlf, F.J., 1994. Biometry: The Principles and Practice of Statistics in Biological Research. W.H. Freeman, San Francisco. Stalder, T., Barraud, O., Casellas, M., Dagot, C., Ploy, M.C., 2012. Integron involvement in environmental spread of antibiotic resistance. Front. Microbiol. 3, 119. Telenti, A., Lowrie, D., Matter, L., Imboden, P., Cole, S., Schopfer, K., Marchesi, F., Colston, M.J., Bodmer, T., 1993.

253

Detection of rifampicin-resistance mutation in Mycobacterium tuberculosis. Lancet 341 (8846), 647e650. Tello, A., Austin, B., Telfer, T.C., 2012. Selective pressure of antibiotic pollution on bacteria of importance to public health. Environ. Health Perspect. 120 (8), 1100e1106. van Der Drift, C., van Seggelen, E., Stumm, C., Hol, W., Tuinte, J., 1977. Removal of Escherichia coli in wastewater by activated sludge. Appl. Environ. Microbiol. 34 (3), 315e319. Vila, J., Simon, K., Ruiz, J., Horcajada, J.P., Velasco, M., Barranco, M., Moreno, A., Mensa, J., 2002. Are quinoloneresistant uropathogenic Escherichia coli less virulent? J. Infect. Dis. 186 (7), 1039e1042. We´ry, N., Lhoutellier, C., Ducray, F., Delgene`s, J.J., Godon, J.J., 2008. Behaviour of pathogenic and indicator bacteria during urban wastewater treatment and sludge composting, as revealed by quantitative PCR. Water Res. 42 (1e2), 53e62. Winfield, M.D., Groisman, E.A., 2003. Role of nonhost environments in the lifestyles of Salmonella and Escherichia coli. Appl. Environ. Microbiol. 69 (7), 3687e3694. Xia, S., Duan, L., Song, Y., Li, J., Piceno, Y.M., Andersen, G.L., Alvarez-Cohen, L., Moreno-Andrade, I., Huang, C.L., Hermanowicz, S.W., 2010. Bacterial community structure in geographically distributed biological wastewater treatment reactors. Environ. Sci. Technol. 44 (19), 7391e7396. Xu, W., Zhang, G., Li, X., Zou, S., Li, P., Hu, Z., Li, J., 2007. Occurrence and elimination of antibiotics at four sewage treatment plants in the Pearl River Delta (PRD), South China. Water Res. 41 (19), 4526e4534. Zenkin, N., Kulbachinskiy, A., Bass, I., Nikiforov, V., 2005. Different rifampin sensitivities of Escherichia coli and Mycobacterium tuberculosis RNA polymerases are not explained by the difference in the b-subunit rifampin regions I and II. Antimicrobial. Agents Chemother. 49 (4), 1587e1590. Zhang, T., Li, B., 2011. Occurrence, transformation, and fate of antibiotics in municipal wastewater treatment plants. Crit. Rev. Environ. Sci. Technol. 41 (11), 951e998.