Antimicrobial Resistance, Virulence Factors, and Pathotypes of Escherichia coli Isolated from Drinking Water Sources in Jordan

The study investigated the prevalence of potentially pathogenic and drug resistant Escherichia coli among drinking water sources in Jordan. A total of 109 confirmed E. coli isolates were analyzed. Antimicrobial susceptibility testing was done using the Kirby Bauer disk diffusion method. Phenotypic identification of extended spectrum beta-lactamase (ESBL) and carbapenemase production was done using the double disk synergy test and the modified Hodge test, respectively. Isolates’ plasmid profiles were determined by gel electrophoresis. PCR was used for detection of virulence and resistance genes. Overall, 22.0% of the isolates were potentially intestinal pathogenic E. coli (IPEC); namely enteroaggregative E. coli (16.5%), enteropathogenic E. coli (2.8%), enteroinvasive E. coli (1.8%), and enterohemorrhagic E. coli (0.9%). A third of the isolates were multi-drug resistant. The highest rates of antimicrobials resistance were observed against ampicillin (93.6%) and sulfamethoxazole/trimethoprim (41.3%). All isolates were susceptible to imipenem, meropenem, doripenem and tigecycline. The prevalence of ESBL and carbapenemase producers was 54.1% and 2.8%, respectively. BlaVIM was the most prevalent resistance gene (68.8%), followed by blaCTX (50.5%), blaTEM (45.9%), blaNDM (11%), blaKPC (4.6%), and blaSHV (0.9%). Fifty-eight (53.2%) isolates contained one or more plasmid ranging from 1.0 to 8.0 kbp. Overall, high prevalence of potentially pathogenic and resistant isolates was observed.


Introduction
Diarrheal diseases are considered major infectious diseases leading to high rates of morbidity and mortality worldwide [1]. Diarrheal diseases occur more commonly in developing countries and particularly among children, in which diarrhea is considered the second-most common cause (after pneumonia) of death under the age of five [2]. The lack of safe water supplies, use of contaminated water sources, inadequate sanitation, and poor hygiene are the main risk factors for acquiring diarrhea in developing countries [2].
Most of the known E. coli strains are members of the gut normal flora. However, some strains are considered true pathogens, capable of causing urinary tract infections, sepsis, meningitis, and enteric or diarrheal diseases [3]. E. coli is an etiologic agent of diarrhea in developing countries. It is responsible for major waterborne bacterial infections and is successfully transmitted through the direct intake of contaminated water or indirectly through food crops exposed to contaminated water sources [3]. Diarrheagenic E. coli strains also known as intestinal pathogenic E. coli (IPEC) have been grouped into six pathotypes: Enteropathogenic E. coli (EPEC), enterotoxigenic E. coli (ETEC), enterohemorrhagic E. coli (EHEC), enteroinvasive E. coli (EIEC), enteroaggregative E. coli (EAEC), and

Antimicrobial Susceptibility
The antimicrobial susceptibility profile of the isolates to various antimicrobial agents is shown in Table 1. The highest rates of resistance were observed against ampicillin (93.6%), followed by sulfamethoxazole/trimethoprim (41.3%), ciprofloxacin (16.5%), levofloxacin (14.7%), ceftriaxone/ cefotaxime (12.9%) and ceftizoxime (9.2%). All isolates were sensitive to imipenem, meropenem, doripenem and tigecycline. Resistance to three or more of the antimicrobial agent groups i.e., multi-drug resistance (MDR) phenotype, was seen among 34.9% of the isolates (Table 2). Based on the results of the five antimicrobial agents used in the double disk synergy test (DDST) to detect ESBL producers, 59 ESBL producers were identified. Augmentation of the inhibition zone which indicates an ESBL producer was observed most frequently with ceftizoxime and cefotaxime (79.7% each; 47/59), followed by aztreonam (72.9%; 43/59), ceftriaxone (67.8%; 40/59), and ceftazidime (49.2%; 29/59). The correlation between the ESBL phenotype and susceptibility to antimicrobial agents is shown in Appendix A. A significant association was observed between the ESBL phenotype and nonsusceptibility to amoxicillin/clavulanic acid, aztreonam, cefotaxime, ceftazidime, ceftizoxime, and ceftriaxone.

Plasmid Profiling
Among the 109 isolates, 58 (53.2%) had one or more plasmid (Table 6). Plasmid size estimates ranged from 1.0 to 8.0 kbp. The plasmid profile of each of the isolates was distinct as none of the isolates shared similar numbers or sizes of plasmids (Appendix C). No significant associations were observed between number of plasmids and number of resistance genes (Table 6), or between number of plasmids and the resistance to antimicrobial agent groups (Table 7). Table 6. Association between the number of resistance genes and number of plasmids among the isolates.  Table 7. Association between resistance to antimicrobial agent groups and number of plasmids among the isolates.

Discussion
E. coli is a biological indicator of fecal contamination of water sources. It is a well-known potential pathogen responsible for a variety of waterborne infections in humans, such as gastrointestinal illness. Furthermore, the presence of antimicrobial resistant pathogenic E. coli strains in water sources may contribute to the spreading of antimicrobial resistance and virulence genes among other bacteria in the environment. In this study, E. coli was recovered from various drinking water sources in Jordan. All isolates were subjected to PCR detection of virulence genes associated with E. coli intestinal pathotypes, plasmid profiling, and phenotypic and genotypic analysis of antimicrobial resistance.
Using PCR to detect virulence genes, 22.0% of the isolates were identified as potentially IPEC. Hence, E. coli obtained from water sources can potentially lead to serious disease. The most prevalent potential pathotype was EAEC accounting for 16.5% of the isolates, followed by EPEC at 2.8%, EIEC at 1.8% and putatively EHEC at 0.9%. The one isolate identified putatively as EHEC, was not positive for eae, as is typical for this pathotype. This may be attributed to mutations in the gene preventing PCR amplification. Alternatively, this could be a non-EHEC strain that acquired stx1 via horizontal gene transfer. Isolates demonstrated highest resistance rates against ampicillin and sulfamethoxazole/trimethoprim. All isolates were sensitive to imipenem, meropenem, doripenem and tigecycline. Many agents also demonstrated high susceptibility. Nonetheless, approximately one third of the isolates had an MDR phenotype.
According to the DDST, 54.1% of the isolates were ESBL producers. Various types of antimicrobials were used for the phenotypic detection of ESBL producers to enhance detection sensitivity. Ceftizoxime and cefotaxime identified the highest number of ESBL isolates, while ceftazidime identified the least. The CLSI (2015) recommends using multiple agents, including aztreonam, cefotaxime and ceftizoxime for ESBL screening, which is consistent with our methodology and findings. All ESBL positive isolates were susceptible to nitrofurantoin, tigecycline, imipenem, meropenem, and doripenem. Similar results have been reported for E. coli in Iran, South Africa, and Bangladesh [16,[19][20][21]. The observed high rates of MDR are likely due to the dissemination of resistance genes via horizontal gene transfer among bacteria and the relative ease for E. coli to acquire genetic material, especially in light of increased selective pressure due to exposure to a wide range of antimicrobials in the human body, as well as in the environment.
A study from Jordan in 2012 on clinical E. coli isolates reported an ESBL rate of 50.3%; (83/165) [22]. A rate surprisingly similar to what we found, suggesting that E. coli obtained from the drinking water sources are potentially a subset of those involved in clinical infections. For comparison, 55.6% of clinical E. coli isolates from India were ESBL positive [23], while lower rates of ESBL producing clinical E. coli were reported in Qatar (34.7%), Iran (21%), Oman (13.3%), and Korea (9.2%) [24][25][26][27]. The higher prevalence of ESBL isolates in Jordan compared to other countries may be attributed to the trend of self-medication, unregulated prescription of antimicrobial agents, and the extensive prophylactic misuse of antimicrobials by Jordanian patients and physicians.
While the CLSI (2015) only recommends using ertapenem and meropenem for carbapenemase screening, this study utilized four carbapenem drugs for screening of carbapenemases using the MHT. Based on this, three (2.8%) isolates were identified as carbapenemase producers. Carbapenemase positive isolates were only identified using ertapenem and imipenem. Surprisingly, these isolates were still susceptible to all four carbapenem drugs. This could be attributed to the production of low carbapenemase levels, or a false positive MHT mostly generated by ESBL production in association with decreased drug permeability [28]. For comparison, in Lebanon the reported prevalence rate of carbapenemase production was 2.2% among clinical E. coli isolates [29]. MDR E. coli harboring ESBLs and carbapenemases in water sources is alarming, especially considering that carbapenems are usually the antimicrobials of last resort to treat ESBL-producing pathogens.
Different beta-lactamase genes were detected among the isolates; namely, bla VIM (68.8%) followed by bla CTX (50.5%), bla TEM (45.9%), bla NDM (11%), bla KPC (4.6%), and bla SHV (0.9%). Bla VIM was the most prevalent gene. This may be due to its presence on highly mobile genetic elements that facilitate its spread among bacteria and that it is the most common ESBL gene distributed worldwide [30]. The presence of bla VIM in Jordan was documented previously in clinical isolates [31].
The presence of ESBL-producing E. coli possessing bla CTX genes in the environment was documented in different parts of the world [17,32,33]. Salah et al. reported a very high rate of 94.2% of E. coli producing CTX-M-type ESBLs in the intestines of Jordanian infants [34]. The high rate of E. coli harboring CTX-M-type ESBLs reported here is consistent with its high prevalence among the intestinal flora reported by Salah et al. and may be attributed to improper clinical waste management leading to contamination of water sources. Consistent with the findings of our study, 50% of the isolates from water sources in Bangladesh harbored bla CTX [21]. The frequency of bla TEM among the isolates was similar to those reported from Qatar and Malaysia among water isolates [18,24].
Significant associations were observed between stx1 and bla KPC , aaic and bla TEM , and aat with bla VIM . However, only one stx1 positive isolate carried bla KPC . Therefore, this association may not be necessarily typical. The co-presence of virulence genes and resistant genes in pathogenic strains likely favors their survival and persistence in the host and the environment [35]. Similarly, the co-presence of bla CTX and stx1 was described in E. coli isolates from Japan and France [36,37]. A study from Iran also reported the presence of bla CTX and bla TEM in EAEC [38].
Significant co-association was observed between bla KPC and bla TEM , which may reflect co-existence of the genes, their dissemination on mobile genetic elements via horizontal gene transfer, and the ability of bacteria to acquire several resistance genes. Similarly, the co-association of bla KPC and bla TEM was reported in E. coli clinical isolates from Taiwan and Greece [39,40].
In the present study, about half of the isolates carried one or more plasmid. This was similar to findings of a study from India in which 52.6% of E. coli isolates carried plasmids [41]. Plasmid size estimates ranged from 1.0 to 8.0 kbp. These findings are in agreement with studies from Bangladesh and Pakistan [42,43].
No significant associations were observed between the number of plasmids and number of resistance genes, suggesting that resistance genes might have translocated from plasmids to the chromosome, and/or that they were carried on a limited number of plasmids. Furthermore, there were no significant associations between resistance to various antimicrobial agent groups and number of plasmids. This may be due to the presence of genes encoding antibiotics resistance on plasmids as well as the bacterial chromosome. Similarly, some strains containing more than one plasmid were reported to be only resistant to one antimicrobial agent, and vice versa [44]. Other studies also reported no correlation between the number of plasmids and resistance patterns [45].
Due to the identification of potentially IPEC having high rates of drug resistance among drinking water sources, we suggest implementation of wider-scale studies to characterize their presence and the presence of other potential pathogens, and the identification of isolates' virulence and antimicrobial resistance determinants. Furthermore, periodical monitoring of E. coli and other pathogens among water sources is required to assess their contribution to waterborne infections and health of the population, and to evaluate if water sources may serve as reservoirs for these pathogens.

Materials and Methods
This study was approved by the Jordan University of Science and Technology (JUST) research committee. Requirement for approval by the institutional review board of JUST was waived as the study did not involve the study of human subjects, human data or tissue, or animals.

Bacterial Isolates
A total of 157 suspected Escherichia coli isolates were obtained from the water testing authority's microbiology laboratory, Amman, Jordan from February to June 2015. Water samples processed at this laboratory originated from all sources of the drinking water supply in Jordan. Isolates were obtained from fluorescent wells of a Colilert Quanti-Tray (IDEXX Laboratories, Westbrook, ME, USA) and streaked onto MacConkey agar, followed by subculture on eosin-methylene blue agar. The identity of all isolates was confirmed using conventional methods, and PCR via detection of the E. coli-specific gene uspA. Of the total 157 isolates, 109 were confirmed as E. coli. For long term storage, overnight colonies from agar media were suspended in LB broth supplemented with glycerol (16% final concentration), and stored at −80 • C. Bacterial colonies from overnight cultures on Muller Hinton agar (MHA) were resuspended in 2.0 mL of sterile normal saline, to create a bacterial suspension having a turbidity equivalent to 0.5 McFarland for antimicrobial susceptibility testing (AST).

Detection of ESBLs
The production of ESBLs was detected by the double disk synergy test (DDST) according to the CLSI (2015) guidelines using a disk of amoxicillin/clavulanic acid along with aztreonam, cefotaxime, ceftazidime, ceftizoxime, and ceftriaxone. An MHA plate was inoculated with each isolate. Next, an amoxicillin/clavulanic acid disk was placed in the center of the plate, and aztreonam, cefotaxime, ceftazidime, ceftizoxime, ceftriaxone disks were placed 25 mm (center to center) from the amoxicillin/clavulanic acid disk. After overnight incubation at 37 • C, any distortion or increase in the zone of inhibition (i.e., augmentation of inhibition) towards the amoxicillin/clavulanic acid disk was considered a positive result for ESBL production. Klebsiella pneumoniae ATCC 700603 was used as positive control. E. coli ATCC 25922 was used as negative control.

Detection of Carbapenemase Production
Carbapenemase production was evaluated using the Modified Hodge test (MHT) according to the CLSI (2015) guidelines. Briefly, a 0.5 McFarland-equivalent suspension of E. coli ATCC 25922 in 5 mL of sterile saline was prepared. A 1:10 dilution was prepared in sterile saline. The diluted suspension was used to inoculate the full surface of an MHA plate. The plate was dried for 5 min and a disk of either doripenem (10 µg), ertapenem (10 µg), imipenem (10 µg), or meropenem (10 µg), was placed in the center of the plate. Two-to-four colonies of the test organism were selected and streaked in a straight line, from the edge of the disk, up to the edge of the plate. After overnight incubation at 36 • C, carbapenemase production was identified by observing a clover leaf-like indentation of E. coli ATCC 25922 growing along the test organism's growth streak within the disk diffusion zone [46]. K. pneumoniae ATCC BAA-1705 was used as a positive control. K. pneumoniae ATCC BAA-1706 was used as a negative control.

Plasmid Profiling
Plasmid DNA was extracted from all isolates by the alkaline lysis method using the Zyppy™ Plasmid Miniprep Kit (Zymo Research, Irvine, CA, USA) according to the manufacturer instructions. Seven microliters of each plasmid DNA sample were mixed with 1.5 µL of loading dye (6X), and the mixture was loaded on 0.7% agarose. GeneRuler TM one kilo-base DNA linear Ladder (Fisher Scientific, Loughborough, UK) was used as a reference marker. Samples were electrophoresed for one hour at 150 V in a 1X Tris-borate-EDTA buffer, and the gel was photographed using a UV transilluminator. No restriction enzymes were utilized before electrophoretic separation. Hence, plasmid sizes measured were estimates based on the size of supercoiled plasmid DNA. A clinical isolate of E. coli previously identified to possess plasmids was used as a positive control. A clinical isolate previously identified to have no plasmids was used as a negative control.

Total DNA Extraction
Total genomic (chromosomal and plasmid) DNA was extracted from the isolates using a simple boiling method. Briefly, two pure bacterial colonies were inoculated into 5 mL of LB broth and the tubes were incubated overnight at 37 • C for 16 h. Next, 1.5 mL of overnight culture was transferred to an Eppendorf tube and centrifuged at 13,000× g for 10 min. The bacterial pellet was suspended in 300 µL sterile water and heated at 100 • C for 10 min to lyse the cells. Debris were removed by centrifugation at 13,000× g for 10 min and the supernatant was transferred into an Eppendorf tube. The extracted DNA was stored at −20 • C until used for PCR.

PCR for Isolates' Identification and Detection of Resistance Genes
PCR was performed to confirm the identity of the 157 isolates by detecting the E. coli-specific uspA, as described previously [47]. The 25 µL PCR contained 12.5 µL 2X PCR master mix solution, 2 µL template bacteria DNA solution, 1.5 µL of both forward and reverse primers for each of the targeted genes (6.25 pmoles/µL) and 7.5 µL nuclease free water. The following PCR conditions were used; 5 min at 94 • C, followed by 30 cycles of 94 • C for 2 min, 70 • C for 1 min, and 72 • C for 1 min. The 109 isolates confirmed as E. coli (uspA positive) underwent PCR for detection of virulence and resistance genes. Multiplex PCR was performed to detect eae of EPEC, in addition to aaiC and aat of EAEC, as described previously [19]. Three pairs of primers, were used for amplification of these genes. The 25 µL PCR contained 12.5 µL 2X PCR master mix solution, 3 µL template bacteria DNA solution, 0.4 µL of both forward and reverse primers (6.25 pmoles/µL) for each of aaiC and aat, 0.44 µL of both forward and reverse primers (6.25 pmoles/µL) for eae, and 7.02 µL nuclease free water. The following PCR conditions were used; 4 min at 96 • C, followed by 34 cycles of 95 • C for 20 s, 57 • C for 20 s, and 72 • C for 1 min. Touchdown multiplex PCR was performed to detect stx1 and stx2 of EHEC as described previously [19,48]. The 25 µL PCR contained 12.5 µL 2X PCR master mix solution, 3 µL template bacteria DNA solution, 0.5 µL of both forward and reverse primers (6.25 pmoles/µL) of each targeted gene, and 7.5 µL nuclease free water. The following PCR conditions were used; 94 • C for 5 min, followed by 40-cycle each of 94 • C for 30 s, 64 • C for 30 s, and 72 • C for 60 s, for two cycles followed by eight cycles having a 2 • C decrease of annealing temperature after every two cycles. When the annealing temperature of 54 • C was reached at cycle 10, the PCR continued with these cycling parameters followed by a final extension at 72 • C for 10 min. PCR was performed to detect ipaH of EIEC as described previously [19]. The 25 µL PCR contained 12.5 µL 2X PCR master mix solution, 2 µL template bacteria DNA solution, 2 µL of both forward and reverse primers for the ipaH gene (6.25 pmoles/µL), and 6.5 µL nuclease free water. The following PCR conditions were used 95 • C for 5 min, 30 cycles of 95 • C for 50 s, 55 • C for 1.5 min, and 72 • C for 2 min, with a final extension at 72 • C for 7 min. Table 8 lists PCR primers used. Multiplex PCR assay was performed to detect carbapenem resistance genes (bla KPC , bla NDM , and bla VIM ), as described previously [49]. The 25 µL PCR contained 12.5 µL 2X PCR master mix solution, 1 µL template bacteria DNA solution, 0.75 µL of both forward and reverse primers (6.25 pmoles/µL) for each of the targeted genes, and seven µL nuclease free water. The following PCR conditions were used; 10 min at 94 • C; 36 cycles of 30 s at 94 • C, 40 s at 52 • C, and 50 s at 72 • C, and final extension at 72 • C for 5 min. PCR was performed to detect ESBL genes (bla CTX , bla TEM , and bla SHV ), as described previously [50]. The 25 µL PCR contained 12.5 µL 2X PCR master mix solution, 3 µL template bacteria DNA solution, 0.83 µL of both forward and reverse primers (6.25 pmoles/µL) of each targeted gene, and 7.84 µL nuclease free water. The following PCR conditions were used; 5 min at 94 • C, followed by 35 cycles of 30 s at 94 • C denaturation, 30 s at (50 • C for bla CTX , 52 • C for bla TEM , and 56 • C for bla SHV ) for annealing, 30 s (60 s for bla SHV ) at 72 • C for extension, and a final elongation step of 5 min at 72 • C. Controls for PCR were K. pneumoniae ATCC BAA-1706 (bla KPC and bla NDM negative control), K. pneumoniae ATCC BAA-1705 (bla KPC positive control), K. pneumoniae ATCC BAA-2146 (bla NDM positive control), K. pneumoniae ATCC 700603 (bla SHV positive control), and E. coli ATCC 35218 (bla TEM positive control). Amplification products were detected by electrophoretic separation on 1.5% agarose. For genes having no positive control strains, representative PCR bands were sequenced for verification of amplified genes. Table 8 lists PCR primers used.

Statistical Analysis
The SPSS software version 21 (IBM, Armonk, NY, USA) was used to generate the descriptive analysis of raw data, including generation of all frequency tables and cross tabulations. Fisher's exact test was used to compare frequency data. Associations between plasmid profiles with resistance genes and phenotypes were investigated using Pearson's Chi-square test. A p value equal to or less than 0.05 was considered statistically significant. Full study data are available online as supplementary material (Table S1).

Conclusions
In conclusion, 22% of the E. coli isolates recovered from drinking water sources in Jordan were potentially intestinal pathogenic strains, most likely being EAEC. Many isolates harbored beta-lactamase resistance genes, had an MDR phenotype, and an ESBL phenotype. Carbapenemase production was infrequent. None of the isolates had similar plasmid profiles.