A survey of extended-spectrum beta-lactamase-producing Enterobacteriaceae in urban wetlands in southwestern Nigeria as a step towards generating prevalence maps of antimicrobial resistance

In many countries, emission of insufficiently treated wastewater into water bodies appears to be an important factor in spreading clinically relevant antimicrobial resistant bacteria. In this study, we looked for the presence of Enterobacteriaceae strains with resistance to 3rd generation cephalosporin antibiotics in four urban wetlands in southwestern Nigeria by isolation, whole genome sequencing and qPCR enumeration of marker genes. Genome analysis of multi-drug resistant and potentially pathogenic Escherichia coli isolates (members of the widely distributed ST10 complex) revealed the presence of the extended spectrum beta-lactamase gene blaCTX-M-15 on self-transmissible IncF plasmids. The gene was also present together with a blaTEM-1B gene on self-transmissible IncH plasmids in multi-drug resistant Enterobacter cloacae isolates. A Citrobacter freundii isolate carried blaTEM-1B on an IncR-type plasmid without discernable conjugation apparatus. All strains were isolated from a wetland for which previous qPCR enumeration of marker genes, in particular the ratio of intI1 to 16S rRNA gene copy numbers, had indicated a strong anthropogenic impact. Consistent with the isolation origin, qPCR analysis in this study showed that the blaCTX-M gene was present at an abundance of 1x10-4 relative to bacterial 16S rRNA gene copy numbers. The results indicate that contamination of these urban aquatic ecosystems with clinically relevant antibiotic resistant bacteria is substantial in some areas. Measures should therefore be put in place to mitigate the propagation of clinically relevant antimicrobial resistance within the Nigerian aquatic ecosystems.

Introduction Antimicrobial resistance (AMR) is a global problem that threatens progress in health and the achievement of sustainable development goals [1]. Although resistance to antimicrobials is an ancient phenomenon which evolved without human influence in the past, the current increased global presence of AMR is driven by anthropogenic activities [2]. The introduction of antibiotic resistant bacteria and genes (ARB and ARG) into the environment via waste streams from anthropogenic sources is being increasingly recognized as an important contributing factor for its prevalence in the human population [3,4]. There are indications that in Low and Lower-Middle Income Countries (LIC and LMIC) insufficient sanitation infrastructure and the attendant release of untreated or poorly treated wastewater into the environment contributes more to the prevalence of AMR than antibiotic consumption itself [5]. Consequently, polluted aquatic ecosystems have been recognized as important reservoirs of AMR in recent times and deserve further attention [6].
Next to lacking sufficient sanitation infrastructure, most LIC and LMIC have not yet adopted national AMR surveillance programs. Such programs constitute a basis for implementing best patient-treatment practices, identifying newly emerging threats, and directing actions and resources for the containment of AMR to where they have the highest impact. Although several LMIC are taking steps towards establishing surveillance programs, their implementation is lagging, largely due to insufficient numbers of AMR diagnostic laboratories [7]. However, establishing and operating more laboratories in LIC and LMIC will need substantial economic resources, trained personnel and time. Until these are put in place, monitoring AMR profiles in selected water bodies with anthropogenic impact could be a cost-effective proxy for a national surveillance program [8]. The rational selection of environmental sampling sites is of critical importance for the success of this strategy.
The overall aim of our work is to design and implement approaches for effectively gathering knowledge on AMR prevalence in anthropogenically impacted ecosystems in LIC and LMIC that could be translated into public health-related practices. To this end we have been assessing the level of environmental contamination with clinically relevant ARG in Nigeria [9][10][11], the most populous nation in Africa and a LIMC. Results from these studies provided information about the genetic context of such ARG in bacteria isolated from Nigerian aquatic ecosystems. The level of AMR contamination in four polluted wetlands in Lagos and Ibadan appeared to be moderate to high, based on the relative abundance of three AMR indicator genes (intl1, sul1, sul2 [4]) versus copy numbers of the bacterial 16S rRNA gene [12].
In this study, we investigated the same urban wetlands for the occurrence of members of the Enterobacteriaceae that are resistant to 3 rd generation cephalosporins (3GC) via production of extended spectrum beta-lactamases (ESBL) or AmpC beta-lactamases. The presence of such microorganisms in the environment is of high concern since they can cause difficult-totreat infections which will place a high morbidity burden on the society [13]. ESBL genes in Enterobacteriaceae are frequently located on self-transmissible plasmids together with various other ARGs [14-16], while the ampC gene can be located on the chromosome or on a plasmid and can confer resistance to 3GC when overexpressed due to mutations [13,17]. Most common ESBL fall into one of three groups named CTX-M, TEM, and SHV, of which the CTX-M enzymes are the most prevalent [14,18,19]. The linkage of bla CTX-M with plasmids may be a key contributor to its epidemiology in the human population, for which there are no distinct boundaries between clinical settings and community [15]. Here, 3GC-resistant Enterobacteriaceae were isolated from a wetland in Ibadan with high anthropogenic impact. The prevalence of these microbes correlated with qPCR-based enumeration of the relative abundance of intI1 copy numbers, indicating that measuring this abundance alone could be sufficient for assessing the distribution of AMR in a watershed.

Isolation of 3 rd Generation Cephalosporin-Resistant Bacteria from wetland sediments
Bacteria showing resistance to 3 rd generation cephalosporins (3GC-RB) were isolated from the sediments of four polluted wetlands in Lagos [Abule Agege (06.5145˚N, 03.4002˚E), Ogbe Creek (06.5135˚N, 03.3937˚E)] and Ibadan [Awba (07.4468˚N, 03.8763˚E), Apete (07.4577˚N, 03.8828˚E)] in southwestern Nigeria. The wetlands receive untreated wastewater and raw sewage from student hostel facilities of two universities (Awba, Apete and Abule Agege), waste streams from fish farms (Awba and Abule Agege), a Zoological Garden (Awba), and seepages from upland solid waste dumpsites (Apete and Ogbe Creek) (see [10] for a map). Detailed knowledge on quantity and quality of the various water sources in the regions is not available. Sediment samples were collected monthly between October 2014 and January 2015 and processed for the isolation of bacteria as described previously [10] except that the agar plates were supplemented with ceftriaxone (4 μg/ml). Briefly, the samples were serially diluted in saline and 3GC-RB isolated from the samples by selective enrichment on ceftriaxone-supplemented Eosine Methylene Blue agar plates. Morphologically distinct colonies selected from the agar plates were purified by re-streaking on fresh ceftriaxone-supplemented agar plates. All isolates were stored in glycerol broth (15% v/v) at -20˚C for further processing. Phenotypic detection of ESBL production was carried out by the double disc synergy test (DDST) as described by the Clinical and Laboratory Standards Institute [20].
Total genomic DNA extracted from DDST-positive 3GC-RB (DNeasy Blood and Tissue Kit, Qiagen) were used as templates in PCR with primers 27F and 1492R [21] targeting the 16S rRNA gene for bacterial identification. PCR products were purified (QIAquick PCR Purification Kit, Qiagen) and sequenced with primer 27F (GATC, Konstanz, Germany).

Whole genome sequencing of ESBL-producing Enterobacteriaceae
Whole genome sequencing (WGS) was carried out as previously reported [10]. In total, 12 members of the Enterobacteriaceae (7 Escherichia coli, 4 Enterobacter cloacae and 1 Citrobacter freundii isolate) were selected for WGS. Genomic DNA (500-1000 ng) of each isolate was sheared with a Covaris S220 sonication device (Covaris Inc.; Massachusetts, USA) with the following settings: 55 seconds, 175 W, 5% Duty factor, 200 cycles of burst, 55.5 μL input volume. Sequencing libraries were prepared using the NEBNext1 Ultra™ DNA Library Prep Kit for Illumina1 (New England Biolabs, Frankfurt, Germany) following manufacturer's instructions and the libraries were sequenced with an Illumina1 MiSeq machine using v3 chemistry and paired-end approaches with 301 cycles per read. Raw sequences were subjected to adapter clipping and quality trimming using Trimmomatic [22], and processed reads were assembled with SPAdes v3. 6

Reference-based in silico plasmid assembly
Plasmid draft assemblies were generated by using reference plasmids as previously described [32]. Here, contigs of apparent plasmids and of uncertain replicon origin were sorted from those of putative chromosomal origin via MAUVE alignments between the reference strains (GenBank accession numbers in parenthesis) E. coli K12 substr. MG1655 (NC_000913), E. cloacae subsp. cloacae ATCC 13047 (CP001918), C. freundii CFNH1 (NZ_CP007557) and the concatenated contigs of the new isolates [33]. Annotations of the genes on potential plasmid contigs were manually inspected and contigs containing at least one rep, tra or trb gene were used as query sequences for BLASTn searches in the GenBank. Full plasmid sequences harboring close homologs to the aforementioned genes were downloaded from GenBank and used in MAUVE alignments with the contigs of potential plasmid origin from the novel strains. Reference plasmids belonged to the incompatibility groups IncF, IncH and IncR, respectively, and are listed in S1 Table. Furthermore, the Illumina reads of the novel strains were mapped against the plasmid references using default parameters on Geneious Prime 2019 (https:// www.geneious.com). These mapped reads were re-mapped against the contigs of apparent chromosomal origin of the strains isolated in this study in order to check that no potential plasmid sequences had been overlooked. Then, initial draft assemblies were generated from all contigs of putative plasmid and of uncertain replicon origin. The drafts were iteratively refined by a combination of contig ordering and omission of some contigs based on MAUVE alignments with the reference plasmids, Illumina read mapping, and in silico replicating insertion sequences present in multiple copies on the respective plasmid.

Conjugation assay
The presence of plasmids in the novel isolates and their self-transmissibility were tested via conjugation assays with Aeromonas aquatica AE235 (DSMZ 100827) as recipient strain. The donor isolates and A. aquatica were cultured in LB broth and incubated at 30˚C on a shaker (120 rpm). When donor and recipient cell counts reached 10 5 cells/ml and 10 6 cells/ml, respectively, 100 μl from each culture were mixed in a 1.5 ml Eppendorf tube and incubated at 37˚C for 1 hr. The cell suspension was vortexed and then plated on Aeromonas Isolation Agar plates with ceftriaxone (50 μg/ml) but without ampicillin typically applied to this medium. The type strain A. aquatica AE235 did not grow with ceftriaxone concentrations of 12.5 μg/ml and above. The plates were incubated at 30˚C overnight after which A. aquatica colonies were counted to calculate transformation efficiencies. Three randomly selected colonies were then re-streaked and incubated at 30˚C overnight on Aeromonas Isolation Agar plates with 50 μg/ ml ceftriaxone to obtain presumptive A. aquatica transconjugants.
The presence of the respective plasmids in the presumptive transconjugants were confirmed by PCR-based detection of plasmid backbone genes (targeting traA, traN, traU, trbB) and ARGcontaining contigs [targeting bla  , aph(6)-Id, aph(3")-Ib, aadA5, aac(6')-Ib-cr, aac3-IIa, tet(B)]. Primers were designed using the NCBI primer blast tool (https://www.ncbi.nlm.nih.gov/ tools/primer-blast/) (S2 Table). DNA was extracted by the microwave boiling method [34] from randomly selected single colonies of transconjugants grown overnight. The following reaction mixture was used for PCR amplification: 1 μl extracted DNA, 0.50 μl primers (10 μM, 0.25μl each), 6.25 μl Red Taq 2 X Mastermix diluted to 12.5 μl with ddH 2 O. Initial denaturation at 95˚C for 5 min was followed by 30 cycles of amplification (denaturation at 95˚C for 30 sec, annealing for 30 sec at the primer pair-specific temperature shown in S2 Table, and extension at 72˚C for 30 sec) and ending with final extension at 72˚C for 5 min. DNA isolated from the donor strains and A. aquatica AE235 were used as positive and negative controls, respectively.

Quantitative PCR analysis
Total community DNA samples previously extracted from the wetland sediments [12] were pre-screened to query for the presence of beta-lactamase genes bla CTX-M , bla TEM , bla SHV , and ampC using primers and conditions reported before [37,38]. Based on the results of the PCR screening, the relative abundances of Group 1 bla CTX-M [18] and ampC genes in the total sediment community DNA of the four wetlands were determined by SYBR Green-based real-time PCR with four technical replicates per sample. The assay was run on a StepOne Plus Cycler (Applied Biosystems) in a 20 μl reaction volume. The primers and conditions for qPCR assays were described before for Group 1 bla CTX-M [37] and ampC [38]. The abundance of intI1 and 16S rRNA genes in the samples were determined in our previous study [12] using published primers [39,40]. Here, standards were PCR-amplified fragments of the bla CTX-M and ampC genes from environmental isolates obtained in this study. Abundance of total E. coli was carried out by measuring the uidA gene by qPCR with the primers and conditions reported previously [41]. DNA concentrations of the standards were measured by Nanodrop spectrophotometry and the copy number (CN) per gram of wetland sediment was calculated using the relation: where c is the measured DNA concentration (μg/μl) and N is the DNA fragment length in bp.

Ethics statement
There are no specific permits required for sample collection in the field studies. The wetlands are not protected and not privately owned, hence there are no regulations that restrict collection of sediment samples for research purposes from the four wetlands. The field study does not involve any endangered or protected species, only sediment samples were taken from the sites.

Third Generation Cephalosporin-Resistant Bacteria were isolated from all investigated wetlands
A total of 90 colonies representing all different morphotypes (based on size, colour, surface texture and colony edge) growing on the ceftriaxone-supplemented agar plates from all wetland sediment samples were subjected to DDST to confirm ESBL production (41 colonies from Awba, 28 from Apete, 17 from AbuleAgege and 4 from Ogbe Creek). Thirty-five of the isolated 3GC-RB tested positive for ESBL production and were identified by partial 16S rRNA sequencing as belonging to the Enterobacteriaceae genera Escherichia (7 strains [9,50], in particular from sites were insufficiently treated wastewater is released into the aquatic environment. In the present study, ESBL production was more frequent among isolates from Awba (n = 23) compared to Apete (n = 5), Abule Agege (n = 7) and Ogbe Creek (n = 1). All ESBL-producing Enterobacteriaceae were from Awba wetland located within the campus of the University of Ibadan. In Abule Agege and Ogbe Creek all ESBL producers were identified as Caulobacter sp., while Caulobacter sp. and Pseudomonas spp. were identified as ESBL producers in Apete. This isolation pattern of 3GC-RB might reflect the different anthropogenic impact on the four wetlands. Enumerations of intI1, sul1, and sul2 by qPCR in a previous study indicated that Awba and Abule Agege were more impacted with AMR contamination than Apete and Ogbe Creek[12].

Overview of the draft genomes of the ESBL-producing Enterobacteriaceae
In order to analyze the genetic basis of 3GC resistance in the isolated Enterobacteriaceae their genomes were sequenced. The remaining DDST-positive isolates were not further investigated since they were outside the target phylogenetic group and PCR screening did not detect any of the tested beta-lactamase genes. Full-length 16S rRNA gene sequences showed that the sequenced strains affiliated with E. coli (n = 7), E. cloacae (n = 4) and C. freundii (n = 1) with draft genome assemblies of 4.8 Mb, 5.1 Mb, and 5.2 Mb in size, respectively. Further genome assembly characteristics are provided in Table 1.
Results from the PlasmidFinder pipeline indicated that the isolated E. coli strains carried an IncF plasmid (replicon type II-IA-IB). The conjugative plasmids of the IncF family range in size from 45 to 200 kb [16,51,52]. They seem to have a narrow host range specific to the Enterobacteriaceae, being mostly found in E. coli. Furthermore, they are the predominant plasmids involved in the dissemination of bla CTX-M-15 and are widely reported in bacteria isolated from humans, animals and the environment [52]. IncFII-IA-IB plasmids carrying bla CTX-M have been implicated in community-associated infections on several continents, emphasizing the epidemic nature and global distribution of these plasmids [53,54]. The E. cloacae genomes contained contigs with similarities to IncH-like plasmids of replicon types IncH12 and H12A. These are low copy number plasmids with a wide host range, varying in size from 75 to 400 kb. They have been found in various Gram-negative microbes, and frequently carry ESBL genes alongside various other ARG [16]. The C. freundii genome contained sequences with similarities to IncR plasmids, which range in size from 40 to 340 kb and are less frequently detected although they appear to have broad host range [16]. Detailed information on these plasmid types in the novel isolates is provided in the next section.
The E. coli and E. cloacae strains were assigned by MLST to sequence type (ST) 617 and ST976, respectively, and C. freundiiCC12 was assigned to ST323. E. coli ST617 is a member of the ST10 complex, which has emerged as important carrier mediating the spread of antimicrobial resistance genes including bla CTX [62], and other environmental sources [48,49,63]. We are not aware of any published information on E. cloacae ST976 and C. freundii ST323 other than their mentioning in EnteroBase [64].
Several virulence genes (ast, capU, gadAB, iss, and genes for the biosynthesis and uptake of the siderophore aerobactin) were found in the genomes of the sequenced E. coli strains, which suggest that the isolates could be pathogenic. The astA gene codes for an aggregative heat-stable enterotoxin [65]; capU is predicted to code for a hexosyltransferase in enteroaggregative E. coli [66]; gadAB code for a glutamate decarboxylase which is involved in acid resistance in habitats such as the stomach [67]; the iss gene product promotes immune evasion by increasing serum survival [68]; and aerobactin is a virulence factor during urinary tract infections [69]. No potential virulence gene was detected in E. cloacae CC14, CC80, CC81, CC90 and C. freundii CC12.
All isolates were multi-drug resistant based on their ARG content, and several identical resistance genes were present in members of the three species ( Table 2). The E. coli isolates harbored ARG against beta-lactams (bla CTX-M-15 and bla OXA-1 ) as well as aminoglycosides, fluoroquinolones, macrolide-lincosamide-streptogamin B, phenicols, sulfonamides, tetracycline, and trimethoprim. The E. cloacae genomes contained ARG against beta-lactams (ampC, bla CTX-M-15, bla OXA-1 and bla TEM-1B ), and aminoglycosides, fluoroquinolones, fosfomycin, phenicols, sulfamethoxazole, tetracycline, and trimethoprim. The C. freundii genome contained ARG against beta-lactams (ampC and bla TEM-1B ) as well as against aminoglycosides, sulfonamides, and trimethoprim. The genomic context of the ESBL gene bla CTX-M-15 and the broad-spectrum beta-lactamase genes bla OXA-1 and bla TEM-1B in the isolates are described further in the subsequent section.
In the E. cloacae isolates and the C. freundii strain, ampC and the genes ampR, ampE, ampG, and ampD involved in ampC induction were at conserved loci on contigs with high synteny and sharing up to 98% nucleotide sequence similarity with chromosomal regions in several fully sequenced reference genomes [e.g. E. cloacae strain SBP-8 (CP016906) and C. freundii strain CFNIH1 (NZ_CP007557)], respectively. The ampC sequences in the E. cloacae reference and our E. cloacae isolates were identical; while in C. freundii CC12 the gene was 99.65% similar to bla CMY-89 (NG_048886). A phylogenetic tree of representative AmpC amino acid sequences in our isolates and reference strains is shown in S1 Fig. We could not find sequence-based evidence for constitutive expression of ampC in our strains [70][71][72], indicating that the presence of this gene did not contribute to 3GC resistance in the isolates.

The genes bla CTX-M-15 , bla TEM-1B and bla OXA-1 were present on IncF, IncH and IncR plasmids
Draft assemblies of the various plasmids in the E. coli, E. cloacae and C. freundii isolates were generated via iterative MAUVE alignments with IncF, IncH, and IncR reference plasmids and mapping of the Illumina sequencing reads against the references and draft assemblies. The plasmid assemblies of the E. coli and E. cloacae isolates, respectively, were identical. Representative assemblies from the Awba isolates E. coli CC85 (pAWCC85_draft), E. cloacae CC14 (pAWCC14_draft), and C. freundii CC12 (pAWCC12_draft) were therefore used for further analysis. The pAWCC85_draft of about 155 kb in size is most similar to the IncF plasmid pRCS57 from E. coli strain 690 isolated from a urine sample collected in France [73] (Fig 1). The backbone region of pAWCC85_draft encompassing genes for the conjugative apparatus, the arginine deiminase pathway (ADI pathway), metal acquisition, and plasmid maintenance together with various genes of unknown function had 99.9% nucleotide sequence identity (99% coverage) with pRCS57. Such high levels of sequence identities in plasmids from environmental isolates have rarely been found before [74]. The presence of genes coding for the ADI pathway and for metal acquisition may contribute to the virulence of the host cell and foster the epidemiological success of IncF plasmids [73,75]. The accessory region contained all 13 ARG detected in the WGS conferring resistance to beta-lactams including 3GC (bla CTX-M-15 , bla OXA-1 ), aminoglycosides, fluoroquinolones, macrolide-lincosamide-streptogamin B, phenicol, sulfonamides, and tetracycline. The entire ARG region harbors multiple copies of genes coding for transposases, integrases, and other insertion elements, due to which the positional order of the ARG-containing contigs was uncertain and four assembly gaps remained at insertion elements encoding for a transposase domain protein, integrase core domain protein, uncharacterized insertion element, and ISEcp18 transposase, respectively.
There are a few differences between pAWCC85_draft and the closest known relative, pRCS57. In the latter as in many other IncF plasmids the bla CTX-M-15 gene is inserted into a target site in the tnpA of a bla TEM-1 -Tn2 transposon [76]. This transposon is not present in the entire WGS of the E. coli isolates from Awba. Further, the tetracycline resistance gene tetA is coupled to transposon Tn1721 in pRCS57 whereas pAWCC85_draft carries a tet(B)-Tn10 transposon. In pAWCC85_draft there is a class I integron with dfrA17, aadA5, and a conserved 3' region whereas pRCS57 harbors a class I integron which is truncated after dfrA14 by tnpA. Furthermore, pAWCC85_draft carries aph (3")-lb, aph(6)-ld, and sul2, all of which are not present in pRCS57. Next to these differences, the genes for aerobactin biosynthesis and uptake of the iron-aerobactin complex are present in pAWCC85_draft as well as other IncF plasmids used as reference (e.g. E. coli strain AR_0014 plasmid unitig_1_pilon; CP024860) but not in pRCS57.
To support the overall structure of the draft assembly and to test for transmissibility of the presumptive IncF plasmid in the E. coli isolates, conjugation assays were carried out using A. aquatica AE235 as recipient strain. Conjugation frequencies were 2 x 10 −4 . The A. aquatica transconjugants grew in LB broth and on agar plates with 50 μg/ml ceftriaxone or100 μg/ml kanamycin while growth of the parent strain AE235 was completely inhibited with 12.5 μg/ml of either antibiotic. PCR assays confirmed the presence of plasmid backbone genes (traA, traN, traU, trbB) and all ARG included in the assembly in the transconjugants.
The IncH-type assembly pAWCC14_draft (scaffold with 15 contigs) from the E. cloacae isolates has a size of about 221 kb (Fig 2), which is within the range of sizes reported for these plasmids [16]. Organization and sequence similarity among the IncH reference plasmids and pAWCC14_draft were not as high as with the IncF plasmids. The backbone region containing the genes for conjugation and plasmid maintenance in pAWCC14_draft was most similar to pAPEC-O1-R (NC_009838) from the extraintestinal pathogenic E. coli strain APEC O1 [77]. Similar to other IncH plasmids, pAWCC14_draft contains genes for resistance against copper (pco), nickel/cobalt (rcn), and tellurite (ter). Close homologs to silver resistance genes (sil) present on some IncH plasmids [78] including pAPEC-O1-R were however found on a 293 kb contig of apparent chromosomal origin in the E. cloacae isolates. The assembly pAWCC14_draft harbored 15 out of the 19 ARG identified in the E. cloacae genomes, namely the ESBL genes bla CTX-M-15 , bla OXA-1 , bla TEM-1B , as well as those conferring resistance to aminoglycosides (aadA1, acc(3')-IIa, aph(3")-lb, aph(6)-ld), fluoroquinolones (acc(6)lb-cr, oqxA, oqxB, qnrB1), phenicols (catA1, catB4), sulfonamide (sul2), and trimethoprim (dfrA14). The bla CTX-M-15 gene together with genes coding for a cupin-fold metalloprotein and a fragment of a Tn3 family transposase were identical in the E. cloaceae and E. coli isolates as well as many plasmid sequences in the GenBank, e.g. pRCS57. Likewise, the bla OXA-1 gene together with catB4 and acc(6)lb-cr were identical in the pAWCC14_draft and pAWCC85_draft, and 99% similar to the homologues in many sequenced IncF and IncH plasmids. In successful conjugation assays with A. aquatica as recipient strain, conjugation frequencies were 6 x 10 −5 . The presence of IncH backbone genes and all ARG on the contigs were confirmed by PCR.
The IncR-type pAWCC12_draft (16 contigs) in C. freundii CC12 is 85 kb in size (Fig 3), which is similar to the size of 9 out of the 23 reference plasmids used for generating the assembly scaffold. The size of pAWCC12_draft is a conservative estimate as many described IncR plasmids contain prophages. At least10 prophages appear to be present in the WGS of isolate CC12 but there was no sufficient evidence that any of these are part of pAWCC12_draft. The draft assembly contained genes for plasmid replication, maintenance and stability but no recognizable conjugation system could be detected. Consistent with this, conjugation assays with A. aquatica were not successful. The inability to be transferred via conjugation together with the absence of a discernable conjugation system was already described for IncR plasmids [79]. Based on plasmid sequence comparisons and extensive read mapping the pAWCC12_draft assembly contains the bla TEM-1B gene and ARG conferring resistance against aminoglycosides (aadA2, aph (3")-lb, aph (6)-ld), sulfonamides (sul1, sul2) and trimethoprim (dfrA19). The bla TEM-1B gene and an adjacent Tn3 family transposase gene (tnpR) were identical in pAWCC12_draft and pAWCC14_draft, and several plasmid sequences deposited in the GenBank such as pEC-IMPQ [80].

Abundances of marker genes in the wetlands
ESBL genes bla CTX-M-1 , bla TEM , bla SHV , and ampC were assayed as a marker for clinically relevant ARG contamination in the wetland sample. Only bla CTX-M and ampC were detected in samples from Awba, samples from the other sampling sites tested negative for all assayed genes. The abundances of bla CTX-M-1 and ampC were then enumerated in the AW samples using qPCR.
bla CTX-M-1 could only be reliably quantified in the January 2015 samples with an absolute abundance of 2x10 4 ±9x10 3 copy numbers per gram of wetland sediment and normalized abundance relative to bacterial 16S rRNA gene copy numbers of 1x10 -4 . Measured abundances in the other Awba samples were not considered due to low amplification efficiencies (around 70%). Isolation of bla CTX-M containing strains from samples taken in November 2014, December 2014 and January 2015 together with amplification in standard PCR demonstrated that microbes harboring this gene were present at Awba at several time points. The absolute and relative abundances of bla CTX-M-1 measured at the site were similar to that reported for the sediment of Lake Brêt, Switzerland [81], an urban river in Kinshasa, Democratic Republic of Congo [82], and in waste water treatments plants and urban wetlands in the United States [83], but slightly lower than that measured in a river receiving hospital effluents in Tamil Nadu, India [84]. In all of those studies primers targeting Group 1 bla CTX-M were used.
The ampC gene was detected in the AW sediment samples collected November 2014 and January 2015, where all the sequenced ampC-bearing bacteria were isolated from, with absolute abundance (copy number per gram of sediment) of 3.8x10 5 ±2.9x10 4 and 1.3x10 5 ±8.8x10 3 , and normalized abundance relative to the bacterial 16S rRNA gene of 2.1x10 -2 and 1.5x10 -4 respectively. The absolute abundance was similar to what was observed in a river estuary [85] in Shanghai, China. The normalized abundance (3.9x10 -6 to 5x10 -6 ) was 2 to 4 orders of magnitude higher in the Nigerian sediments than in the sediments from the Chinese estuary and 2-6 orders of magnitude higher than in sediments from the Saudi Arabian red sea coast where normalized abundances of 7.5x10 −8 to 9.1x10 −6 were measured [86].
Occurrences of bla CTX-M-1 and ampC correlated with the qPCR-based enumeration of E. coli abundance using the uidA gene as marker, which was detected in the AW samples at all time points but not at the other sites. Copy numbers of uidA per gram of sediment at AW ranged from 4.3x10 4 to 2.6x10 5 , while the relative abundance normalized against bacterial 16S rRNA gene copy numbers ranged from 3.8x10 -5 to 2.8x10 -4 .

Conclusions
In this study, we found multi-drug resistant and potentially pathogenic E. coli belonging to the globally distributed ST10 complex harboring bla CTX-M-15 on a self-transmissible IncF plasmid in a polluted urban wetland (Awba) in Ibadan, southwestern Nigeria. The wetland also harbored multi-drug resistant E. cloacae with bla CTX-M-15 on a self-transmissible IncH plasmid, and C. freundii with bla TEM-1B on an IncR plasmid. Previously, carbapeneme-resistant Pseudomonas spp. harboring bla VIM-5 on novel class I integrons were isolated from the same wetland [10], further indicating a substantial level of contamination of this wetland with clinically relevant ARG. The wetland drains into a reservoir that is used as source of domestic water supply and may potentially contribute to the spread of AMR in the human population. Given the insufficient status of local sanitation infrastructure these findings on AMR prevalence are not overly surprising, but it is hoped that these results will be a catalyst for infrastructure improvements to prevent dissemination of ARG into the aquatic ecosystem in this region.
We further note that the copy number of intI1 to bacterial 16S rRNA gene abundance correlated with the observed level of AMR prevalence in the four investigated urban wetlands [12] (Fig 4). The intI1 gene is among several indicators suggested for monitoring anthropogenic AMR pollution in the environment [4,87]. Although we analyzed only a small number of samples taken over a short time frame, results from this and our previous studies on the four Nigerian wetlands suggest that determining the intI1/16S rRNA gene copy number ratio may be sufficient for generating maps of AMR distribution in the environment, including the presumptive identification of local hotspots. Such maps could guide were to carryout detailed analyses involving cultivation or metagenomics as described by Hendricksen et al., [8] to provide a picture of the AMR profile in the human community. In LIC and LMIC, detailed knowledge on environmental prevalence of AMR might stand as proxy for a more-costly national surveillance program which involves extensive sampling in the clinical setting. It could also indicate which antibiotics are no more useful for treatment due to high levels of resistance as well as identify newly emerging resistance determinants in the region. The availability of prevalence maps could also be helpful to direct resources for sanitation infrastructure improvements to where they would be most cost-effective for combating the spread of AMR.
Supporting information S1