Detection of carbapenem-resistance genes in bacteria isolated from wastewater in Ontario

Wastewater (influent) samples (1 L) were collected from Arnprior, a small agricultural city (3 March 2017); Ottawa, a mid-sized city with light industry (10 March 2017); and Toronto, a major industrialized city (17 March 2017). Samples were received on ice at the Ottawa Laboratory Carling (Canadian Food Inspection Agency) where they were filtered and plated on MacConkey agar containing meropenem within 24 h of collection (described below).

Carbapenem-resistance in this study was defined as growth on agar containing ≥4 μg/mL meropenem for Enterobacteriaceae, ≥8 μg/mL for Stenotrophomonas maltophilia, and ≥16 μg/mL for other non-Enterobacterales including Pseudomonas spp. (but not P. aeruginosa), as recommended by Clinical and Laboratory Standards Institute (CLSI; 2020) guidelines. Carbapenem-tolerance was defined as growth on media containing 4 μg/mL for S. maltophilia and Pseudomonas spp. (CLSI 2020).

Water analyses were conducted as per the method of Zurfluh et al. (2013). Briefly, 250 mL of each water sample was filtered through a series of 0.80 μm and 0.22 μm sterile membrane filters (S-Pak Membrane Filters, MilliporeSigma, Ottawa, Ontario, Canada). For each sample all filters were aseptically transferred to a single sterile stomacher bag containing 25 mL modified tryptone soya broth (mTSB) (Oxoid, Ottawa, Ontario, Canada). Filters were rinsed by manually massaging stomacher bags for 5 min to release the collected bacterial cells. An aliquot from each filter rinse was serially diluted in sterile phosphate buffered saline, and dilutions were plated in triplicate on MacConkey agar (MAC) (Sigma, Ottawa, Ontario, Canada) containing 4 μg/mL meropenem (USP, Rockville, Maryland, USA) to select for carbapenem-tolerant gram-negative bacteria. Plates were incubated at 37 °C for 24 h followed by storage at 4 °C until further analysis.

Fifty (Toronto), 51 (Ottawa), and 34 (Arnprior) colony forming units were randomly selected from spread plates for purification via restreaking on MAC containing 4 μg/mL meropenem using a sterile loop. Purified isolates were used for further AMR characterization and β-lactamase ARG detection as described below.

All isolates were tested for carbapenemase-production using the MHT according to CLSI guidelines (CLSI 2011). A 0.5 MacFarland suspension of ATCC 25922 Escherichia coli was diluted 1:10 in saline, and a lawn was spread on Mueller Hinton agar (Thermo Scientific, Oxoid, Ottawa, Ontario, Canada) and left to dry for 3 min. A 10 μg meropenem disc (Thermo Scientific Oxoid, Ottawa, Ontario, Canada) was placed in the centre of the agar plate. Following disc placement, a sterile swab was used to pick three colonies of the purified test organism and was streaked from the edge of the disc to the edge of the agar plate in a straight line. Plates were incubated at 37 °C overnight. If the test organism allowed enhanced growth of the ATCC 25922 lawn towards the disc (into the inhibition zone producing a clover-leaf shape) it was considered to be a carbapenemase producer.

For each purified isolate, one colony was transferred to 1.2 mL of Mueller-Hinton broth (Thermo Scientific Oxoid, Hants, UK) containing 4 μg/mL meropenem using a sterile loop and incubated at 36 °C overnight. Genomic DNA was isolated from overnight broth cultures using the Promega Maxwell 16 Cell DNA purification kit (Promega, Wisconsin, USA). For each sample, 400 μL of broth culture was added to a Maxwell standard elution volume cartridge and the remaining 800 μL of culture was mixed with an equal volume of 30% glycerol and frozen at −80 °C for long-term storage. Extracted DNA was stored at −20 °C.

Genomic DNA was diluted 1:10 (to a concentration of approximately 1–4 ng/μL) and utilized in PCR reactions as described previously (Bogaerts et al. 2013). Twenty-one primer sets (Integrated DNA Technologies, Illinois, USA) were used to screen for select β-lactamase, ESBL, and carbapenemase resistance genes (Table 1). Primers for the OXA PCR were designed in house as follows: (i) nucleotide sequences for multiple alleles of each gene were downloaded from the Center for Genomic Epidemiology (CGE) ResFinder database (accessible at: bitbucket.org/genomicepidemiology/resfinder_db), (ii) alleles for each gene were aligned and analysed for conserved sequence regions, and (iii) conserved regions were input to National Center for Biotechnology Information (NCBI) Primer-BLAST (Basic Local Alignment Search Tool) to design target-specific primers for use in PCR (Ye et al. 2012). The 25 μL reaction mixtures contained 2 μL of diluted DNA extract, 12.5 μL of 2X TopTaq Mastermix (Qiagen, USA), and 0.2 μM of each primer (with the exception of the CARBA multiplex PCR that contained 0.6 μM of bla_IMP primers and 0.2 μM of other primers, Table 1). PCR was performed using a BioRad C1000 Touch Thermal Cycler (BioRad, California, USA) with the following conditions: 15 min at 95 °C, 30 cycles of 30 s denaturation at 94 °C, 90 s annealing at 57 °C, and 90 s elongation at 72 °C, followed by a final elongation at 72 °C for 10 min. PCR amplicons were analysed using capillary electrophoresis on the Qiaxcel DNA Screening system (Qiagen, USA) with QX alignment markers 15/3000 bp, and QX DNA size marker 100–2500 bp.

The species identity of each isolate was determined by matrix-assisted laser desorption/ionization time of flight mass spectrometry (MALDI-TOF). Briefly, colonies were streaked onto tryptic soy agar and incubated overnight at 37 °C. One bacterial colony was applied on duplicate spots of a Biotarget plate, air dried, and overlaid with freshly prepared α-cyano-4-hydroxy-cinnamic acid matrix. Mass spectra were obtained using a MALDI Biotyper mass spectrometer (Bruker Daltonic, Bremen, Germany) and compared with the Biotyper taxonomy database using the MALDI Biotyper 3.1 software. Isolates that received score values of 2.3–3 were identified to the species level. Scores between 2 and 2.299 corresponded to isolates identified to the genus level.

Isolates from a variety of genera with at least one β-lactam resistance gene as determined by PCR were selected for sequencing. A number of PCR-negative isolates were also selected to investigate additional ARGs that may be conferring resistance. From Toronto, Ottawa, and Arnprior 15, 17, and 14, isolates (respectively) were selected for sequencing. Of the selected isolates eight, eight, and one isolate(s) were selected based on PCR-positive results (Toronto, Ottawa, and Arnprior, respectively). The remaining isolates sequenced from each WWTP were PCR-negative. Sequencing libraries were constructed using the Nextera XT DNA sample preparation kit (Illumina, Inc., San Diego, California, USA) and paired-end sequencing was performed on the Illumina MiSeq platform (Illumina Inc.) using 600 cycle MiSeq reagent kits (v3) with 6% PhiX control.

Paired-end Illumina MiSeq sequencing reads were processed with the CFIA-OLC Workflow for Bacterial Assembly and Typing (COWBAT). Source code for COWBAT can be found online: github.com/OLC-Bioinformatics/COWBAT. Quality of raw sequencing reads was assessed with FastQC version 0.11.8 (Andrews 2010). Quality trimming was performed with BBDuk from BBTools version 38.22 (Bushnell 2014) with the following parameters: trim quality of 10 and removal of reads below 50 base pair (bp) long. Error correction was performed using tadpole version 8.22 (Bushnell 2014) in “correct” mode with default parameters. Sequences were checked for contamination using ConFindr 0.5.0 with default parameters (Low et al. 2019). Contigs were assembled from the trimmed and error-corrected reads using SKESA version 2.3.0 with the vector percent argument disabled (Souvorov et al. 2018). Pilon version 1.22 (Walker et al. 2014) was used to perform one round of automatic assembly improvement. Quality was assessed with Qualimap version 2.2.2 (García-Alcalde et al. 2012; Okonechnikov et al. 2016). Multilocus sequence typing (MLST) was conducted using the Center for Genomic Epidemiology MLST-2.0 Server (Larsen et al. 2012), which utilizes MLST allele sequence and profile data from PubMLST.org.

ARG detection in isolate sequences was conducted using both the Comprehensive Antibiotic Resistance Database Resistance Gene Identifier (CARD-RGI) tool version 5.1.0 with default settings (McArthur et al. 2013) and the CGE Bacterial Analysis Pipeline webtool ResFinder version 2.1 (cge.cbs.dtu.dk/services/ResFinder-2.1/) (Zankari et al. 2012) with the default settings for gene identity threshold (90%) and minimum length (60%) unless stated otherwise. Plasmid detection was conducted using the MOB-typer and MOB-recon tools from MOB-suite version 3.0.0 (Robertson and Nash 2018). ARG location determination was conducted by matching the contig encoding the ARG to the matching contig output as plasmid or chromosome by MOB-recon tool.

Whole-genome sequences have been deposited at DDBJ/EMBL/GenBank under bioproject PRJNA489399. Sequence read archive accession numbers, MHT, PCR, WGS-ARG (ResFinder and CARD-RGI), and MALDI-TOF identification data can be found in Table S1. MOB-typer plasmid contents for each sequence can be found in Table S2.

A total of 135 meropenem-tolerant strains (growth on ≥4 μg/mL meropenem) were recovered from WWTP influent samples from Toronto (50), Ottawa (51), and Arnprior (34). Genera identified included Aeromonas, Enterobacter, Klebsiella, Pandoraea, Providencia, Pseudomonas and Stenotrophomonas as determined using MALDI-TOF (Fig. 1b, Table S1).

All strains were tested for carbapenemase-production using the MHT which indicated 12, 7, and 0 carbapenemase-producing strains from Toronto, Ottawa, and Arnprior, respectively. Predicted carbapenemase producing strains included Aeromonas, Klebsiella, Pseudomonas, or Providencia genera (Table 2).

Strains were screened for ARGs using a panel of 21 PCR assays to identify select β-lactamase, ESBL, and carbapenemase resistance genes (Table 1). PCR analysis indicated the presence of β-lactamase genes (bla) with the potential to hydrolyze carbapenems in 35 strains (Fig. 1a, Table S1). The majority of β-lactam-related ARGs were identified in isolates from the Toronto WWTP site, followed by Ottawa (Fig. 1a). Isolates from Arnprior exhibited the lowest number of ARGs, and the identified isolates mostly belonged to species not typically considered to be AMR indicators (e.g., P. fluorescens and P. putida) (Berendonk et al. 2015) (Fig. 1).

The PCR analyses focused on β-lactamase genes reported to confer carbapenem-resistance (Table 3) (Paterson et al. 2003; Monstein et al. 2007; Jacoby 2009; Bush 2010; Bonnin et al. 2011; Nordmann et al. 2011; Bogaerts et al. 2013; van Duin and Doi 2017). Genes for class A carbapenemases are typically found on transferrable plasmids and are highly prevalent among Enterobacteriaceae, with KPC being the most prevalent (van Duin and Doi 2017). The bla_KPC gene was one of the most common ARGs detected by PCR among the isolates with one Klebsiella testing positive from the Ottawa site, as well as eight Klebsiella and a single Providencia from the Toronto site (Fig. 1a). The class A carbapenemase GES enzyme was detected in two Toronto and three Ottawa Aeromonas isolates (Figs. 1a and 2).

The metallo-β-lactamases VIM and IMP were detected in Pseudomonas spp. isolates recovered from both Toronto and Ottawa (Figs. 1 and 2). VIM and IMP are both Class B, metallo-β-lactamases which generally have weak β-lactamase activity against all β-lactams except monobactams but are able to hydrolyze carbapenems (Table 3) (Bush 2010). They are frequently associated with multiple AMR determinants and have spread globally to many of the Enterobacteriaceae (Nordmann et al. 2011; Codjoe and Donkor 2018).

PCR analysis was also conducted to investigate non-carbapenemase class A ESBL genes (Table 3), of which nine isolates were predicted to harbour bla_PER (n = 1) and bla_CTX-M (n = 8) based on PCR results (Fig. 1a). However, WGS analysis of four CTX-M PCR-positive isolates did not detect the bla_CTX-M gene but instead detected the class A ESBL bla_OXY-6. The chromosomal class A enzymes of Klebsiella oxytoca (bla_OXY) have been documented as one of the β-lactamases most closely related to the CTX-M family, sharing 70%–75% amino acid identity (Tzouvelekis et al. 2000; Bonnet 2004). Analysis of the universal CTX-M-U primers used in this study revealed that the primers used annealed to regions of conserved sequence in both the bla_CTX-M and bla_OXY genes (Fig. S1). To our knowledge, this is the first report of false positive binding of CTX-M universal primers with OXY-type enzymes.

The bla_OXY encoded enzyme is a chromosomal class A β-lactamase discovered and reported in K. oxytoca isolates, conferring amino- and carboxy-penicillin resistance (Fevre et al. 2005). K. oxytoca is an opportunistic pathogen and documented causative agent of antibiotic-associated hemorrhagic colitis (Högenauer et al. 2006). Eight K. oxytoca were isolated from the Toronto WWTP during this study. Each of the K. oxytoca isolates that were sequenced encoded both KPC-2 and OXY-6-2 enzymes (Fig. 2). Reports from Brazil and the United States have also characterized clinical K. oxytoca isolates producing the K. pneumoniae carbapenemase enzyme KPC-2 (Yigit et al. 2003; Almeida et al. 2013). Shared environments between K. pneumoniae and K. oxytoca may facilitate horizontal gene transfer and permit adaptation to new environments across species (Moradigaravand et al. 2017). This suggests K. oxytoca may be an equally suitable 1MR indicator and is of concern as shared habitats such as wastewater may permit transfer of AMR to clinically relevant bacterial species and possibly facilitate entrance of multidrug-resistant (MDR) organisms into the environment.

As this study did not include PCR analyses for all possible β-lactamase-related resistance genes, WGS was conducted for a subset of 46 isolates from each WWTP including isolates with and without PCR-identified resistance genes to confirm ARGs and to investigate additional bla genes that may have contributed to carbapenem tolerance. WGS analyses detected additional β-lactamase encoding genes including ampS, bla_ACT, bla_CARB, bla_CEPH, cphA, bla_FOX, bla_MOX, bla_OXY, bla_L1, and bla_PAM (Fig. 2). No ARGs were detected in eleven of the sequenced isolates using ResFinder-2.1 or CARD-RGI (Fig. 2, isolates: OLC2865, 2866, 2868, 2691, 2871, 3144, 3145, 3147, 3634, 3635, and 3636). ResFinder also failed to detect any ARGs in both Pandoraea apista (Burkholderiaceae) isolates (OLC2679 and 3043); however, both had PCR positive results for bla_OXA-10 and CARD-RGI detected bla_OXA-153 at 99.6% identity (Fig. 2, Table S1). Lowering ResFinder-2.1 thresholds to 80% identity (ID) enabled detection of bla_OXA-62 at 83.46%. This suggests the identity thresholds used with ResFinder may have been too stringent, and some ARG-alleles may not be included in the ResFinder database. Similarly, blastn analysis using bla_L1, a chromosomal metallo-β-lactamase from Stenotrophomonas maltophilia (Accession: EF126059), resulted in positive hits for the S. maltophilia isolates OLC2691, OLC2871, OLC3147, and OLC3635 (Fig. 2). ResFinder analysis of the S. maltophilia isolates with the %ID threshold lowered to 80 resulted in positive hits for bla_L1 at 89.57%, 88.69%, 89.70%, and 89.70%, respectively. In contrast, CARD-RGI did not detect the bla_L1 gene in any of these isolates.

Likewise, ResFinder v2.1 and v3.2 with lower ID threshold reported bla_ACT-4 at 85.08% and bla_ACT-9 at 98.86%, respectively, in the Enterobacter asburiae isolate (OLC3633). Jousset et al. (2019) recently described an Enterobacter kobei lineage with weak carbapenemase activity due to the chromosomally encoded AmpC gene bla_ACT-28. CARD-RGI analysis of OLC3633 detected ACT-28 (accession: NG_048614), which is not present in the ResFinder database, at 98.95% identity. In contrast to Jousset et al. (2019) who found isolates were resistant or less sensitive to many β-lactams including ESBLs and expanded spectrum cephalosporins, but not imipenem and meropenem, the Enterobacter asburiae isolated in this study exhibited resistance to meropenem at 4 μg/mL. In addition to AmpCs and ESBLs, porin mutations or modified expression has also been found to confer resistance in Enterobacter species (Codjoe and Donkor 2018; Hao et al. 2018) and may have contributed to resistance in this isolate.

The remaining seven isolates with no identifiable β-lactam ARGs were either Pseudomonas fluorescens or Pseudomonas putida (Fig. 2). Neither CARD-RGI, ResFinder analysis with %ID and minimum length lowered to minimum values, nor blastn (BLAST with default settings) analyses of sequences to known Pseudomonas spp. metallo-β-lactamase genes (Thaller et al. 2011; Suzuki et al. 2014; Borgianni et al. 2015) found any identifiable carbapenem-resistance genes in these isolates. It is possible that these Pseudomonas isolates exhibit carbapenem-tolerance due to reduced outer membrane permeability and (or) increased efflux as has been described elsewhere (Livermore 2001, 2002; Tacão et al. 2015; Kittinger et al. 2016; Codjoe and Donkor 2018; Jousset et al. 2019).

MOB-recon predicted a combined 111 plasmids in 24 of the 46 sequenced isolates. Of the 111 plasmids 17 were predicted as conjugative, 34 were predicted as mobilizable, and 24 included contigs which encoded ARGs (Table S2). Plasmid incompatibility groups included ColRNAI (n = 5), IncC (n = 2), IncFIA (n = 3), IncFIB (n = 6), IncFII (n = 7), IncP (n = 8), IncR (n = 2), IncQ1 (n = 2), IncQ2 (n = 2), and IncX3 (n = 1) (Fig. 2, Table S2).

All five of the sequenced Klebsiella spp. contained plasmids encoding ARGs. In contrast, while each of the six sequenced Aeromonas spp. isolates encoded plasmids, only three of these Aeromonas isolates’ plasmids included contigs encoding ARGs (Table S2). However, short read sequences such as those generated in this study can be challenging to assemble, preventing complete plasmid sequences from being accurately reconstructed (Orlek et al. 2017). MOB suite is dependent on a plasmid database to provide accurate results, and the authors acknowledge that it will not perform well on novel plasmids lacking sequence similarity to those in the database (Robertson and Nash 2018). The database contents are biased towards plasmids from Enterobacteriaceae (Robertson and Nash 2018), presenting an additional obstacle. Although some of the resistance genes identified in this study are known to be chromosomal, it is likely that MOB suite was unable to accurately reconstruct all plasmids for these wastewater isolates as there may not have been matches in current public databases. To determine ARG location more accurately in these isolates (Table S2), additional work to isolate and transfer plasmid(s) and (or) long-read sequencing will need to be conducted.

In addition to monitoring specific ARGs, AMR surveillance programs also recommend monitoring for the presence of bacterial indicators of the Gammaproteobacteria class as these are the most frequent ARG carriers (Berendonk et al. 2015). Pseudomonas aeruginosa and Aeromonas spp. have been proposed as indicators in environmental samples where fecal contamination is not expected. Bacterial indicators where fecal contamination is expected include Escherichia coli, Enterococcus faecalis, Enterococcus faecium, and Klebsiella spp. The recommended indicator organisms are ubiquitous as well as carriers of clinically relevant ARG(s) in multiple environments. A variety of different bacterial species were identified using MALDI-TOF (Fig. 1b) and WGS (Fig. 2). Suggested AMR-indicators of the genera Aeromonas and Klebsiella were isolated from both the Toronto and Ottawa samples tested (Fig. 1b, Table S1), which correlated with an increased number of carbapenem-resistance genes (Fig. 1a). Every Klebsiella and Aeromonas spp. isolated and identified during this study harboured multiple ARGs, some of which were located on plasmids.

Klebsiella spp. have been recommended as possible indicators to assess AMR status in environmental settings as they are considered early indicators of new AMR emergence (Berendonk et al. 2015; Navon-Venezia et al. 2017). We identified eight K. oxytoca and one K. pneumoniae in the Toronto and Ottawa samples (respectively) (Sievert et al. 2013; Moradigaravand et al. 2017). The K. pneumoniae isolated from Ottawa was notable for its MDR genotype. It was predicted to contain several plasmids and multiple ARGs including resistance to aminoglycosides, fluoroquinolones, macrolides, phenicols, sulphonamides, and trimethoprim, in addition to the β-lactamases bla_TEM-1, bla_SHV-11, bla_KPC-3, bla_CARB-2, and bla_OXA-9 (OLC2685, Fig. 2). Netikul and Kiratisin (Netikul and Kiratisin 2015) found carbapenem-resistant K. pneumoniae isolated from a university hospital in Thailand were resistant to carbapenems at a higher ratio than other CREs, harboured multiple β-lactamase and ESBL genes, and (or) expressed altered porin expression, yet only two isolates (out of 36 they tested) encoded commonly associated carbapenemase genes (e.g., bla_KPC), suggesting the need to investigate other mechanisms of carbapenem-resistance in isolates.

Other than Klebsiella spp. a number of isolates belonging to the AMR-indicator genus Aeromonas were also isolated from the Toronto and Ottawa sites (Figs. 1b and 2), although this genus is suggested as an indicator where fecal contamination is not expected. Members of the genus Aeromonas are ubiquitous opportunistic pathogens primarily found in aquatic environments but have also been isolated from clinical, soil, and animal environments (Varela et al. 2016; Chenia and Duma 2017). Several studies have investigated members of Aeromonas as a reservoir of AMR in aquatic environments (Figueira et al. 2011; Trudel et al. 2016; Varela et al. 2016). Trudel et al. (2016) found 37% of tested Canadian A. salmonicida isolates contained at least one resistance plasmid and could pose a potential risk for transfer of resistant genetic material to other waterborne bacteria (including Salmonella spp.). MALDI-TOF identified three isolates from each of the Toronto and Ottawa samples as Aeromonas spp. (Table S1). WGS analysis identified six Aeromonas isolates from Toronto (n = 4) and Ottawa (n = 2), four and three of which encoded bla_GES and bla_OXA β-lactamase genes, respectively. WGS detected β-lactamase ARGs bla_FOX, ampS, bla_OXA, bla_GES, bla_CEPH, cphA2, and bla_MOX-6 in aeromonads isolated during this study (Fig. 2). Overall, 10 Aeromonas spp. isolates were identified using a combination of MALDI-TOF and WGS from Toronto (n = 5) and Ottawa (n = 5) (Table S1).

Species identification of Arnprior isolates, which had few known carbapenem or β-lactam resistance genes, found a large proportion of pseudomonads that were not P. aeruginosa (Figs. 1 and 2). AMR in P. aeruginosa has been attributed to expression of modifying enzymes, efflux pumps, lower membrane permeability, and chromosomal mutations (Moriyama et al. 2009; Chatterjee 2016). While P. aeruginosa has been well studied and is documented to be inherently resistant and exhibit decreased susceptibility to many antibiotics (Livermore 2002; Breidenstein et al. 2011), there are relatively few studies regarding AMR mechanisms in P. fluorescens and P. putida. It is possible that carbapenem resistance in the pseudomonads isolated from Arnprior is due to presence of chromosomal ampC genes and (or) altered membrane permeability as has been observed in other Pseudomonas species (Breidenstein et al. 2011; Molina et al. 2014; Suzuki et al. 2014; Borgianni et al. 2015).