Characteristics of CTX-M Extended-Spectrum β-Lactamase-Producing Escherichia coli Strains Isolated from Multiple Rivers in Southern Taiwan

Abstract

Extended-spectrum β-lactamase (ESBL)-producing Escherichia coli sequence type ST131 has emerged as the leading cause of community-acquired urinary tract infections and bacteremia worldwide. Whether environmental water is a potential reservoir of these strains remains unclear. River water samples were collected from 40 stations in southern Taiwan from February to August 2014. PCR assay and multilocus sequence typing (MLST) analysis were conducted to determine the CTX-M group and sequence type, respectively. In addition, we identified the seasonal frequency of ESBL-producing E. coli strains and their geographical relationship with runoffs from livestock and poultry farms between February and August 2014. ESBL-producing E. coli accounted for 30% of the 621 E. coli strains isolated from river water in southern Taiwan. ESBL-producing E. coli ST131 was not detected among the isolates. The most commonly detected strain was E. coli CTX-M group 9. Among the 92 isolates selected for MLST analysis, the most common ESBL-producing clonal complexes were ST10 and ST58. The proportion of ESBL-producing E. coli was significantly higher in areas with a lower river pollution index (P = 0.025) and regions with a large number of chickens being raised (P = 0.013). ESBL-producing E. coli strains were commonly isolated from river waters in southern Taiwan. The most commonly isolated ESBL-producing clonal complexes were ST10 and ST58, which were geographically related to chicken farms. ESBL-producing E. coli ST131, the major clone causing community-acquired infections in Taiwan and worldwide, was not detected in river waters.

INTRODUCTION

Escherichia coli sequence type ST131 (O25:H4), associated with CTX-M-15 extended-spectrum β-lactamase (ESBL), is the leading cause of community-acquired urinary tract infections (UTIs) and bacteremia worldwide (1–7). This clonal group is virulent and carries a broad range of resistance genes on transferable plasmids (1, 3–9). Studies conducted in southern Taiwan have reported that several clinical isolates of ESBL-producing E. coli belonged to the ST131-O25b lineage. The most prevalent ESBL-encoding gene is bla_CTX-M-14 (10, 11). Most patients with bacteremia or UTIs were healthy previously and did not exhibit any apparent risk factors. Only 30% of infants with UTIs caused by this clone had identifiable potential risk factors (e.g., previous antimicrobial use, hospitalization, neonatal infection, or underlying disease), suggesting that these UTIs were mostly community acquired, not hospital acquired (10–12).

The established reservoirs of the ST131 E. coli clone are humans, companion animals, noncompanion animals, and foods (7). Adequate studies have not been conducted for evaluating whether environmental water is a potential reservoir of this multidrug-resistant clone (13–15). Outbreaks of enteric infections are caused by various bacteria, protozoa, fungi, and viruses present in contaminated drinking water (16, 17). In addition, E. coli in irrigation water contaminates fresh produce (18; http://www.pma.com/∼/media/pma-files/food-safety/cps/cps-research-reportag-water-200813version-11final.pdf). The contamination of rivers can occur through different sources, including human fecal contamination (sewer overflow during heavy rain or inadequate wastewater treatment), animals (particularly livestock), and runoff from pastures and sewage. In this study, we determined whether river water is a potential reservoir of ESBL-producing E. coli in southern Taiwan, focusing on the bla_CTX-M clone found in human infections.

MATERIALS AND METHODS

Sampling of river water.

The Taiwan Environmental Protection Administration (TEPA) routinely examines river water in southern Taiwan for the presence of heavy metals and E. coli. Figure 1 displays 99 stations located in the upper and lower streams in southern Taiwan that are monitored every month, including the Puzi, Bazhang, Jihsui, Tsengwen, Yanshui, Linbian, Donggang, Fenggang, Fangshan, Gaoping, Erren, Agongdian, and Shihchung Rivers and the Paoli and Gangkou streams. One hundred milliliters of river water was sampled each time and analyzed in accordance with the standard procedures of the Environmental Analysis Laboratory of the TEPA (19). All of the sampling procedures followed the rules established by the Environmental Analysis Laboratory of the TEPA. TPEA method NIEA W104.51C was used as the standard basis for sampling (19). Water samples were analyzed immediately after sample collection; the time required from sample collection to laboratory work completion was <24 h. To perform cluster sampling for each river and county, we used a random-number generator in selecting 40 of the 99 stations. One water sample was collected at each sampled station per month. In total, 280 river water samples were collected between February and August of 2014 in southern Taiwan. The normal mean daily temperature in the study area varied between 20.3°C in February and 28.7°C in August. The monthly precipitation ranged between 20.5 mm in February and 416.7 mm in August (http://www.cwb.gov.tw/V7/climate/monthlyMean/Taiwan_tx.htm).

FIG 1.

Study area in southern Taiwan, including the Puzi, Bazhang, Jihsui, Tsengwen, Yanshui, Linbian, Donggang, Fenggang, Fangshan, Gaoping, Erren, and Agongdian Rivers and the Shihchung, Paoli, and Gangkou streams, with 40 river sampling points. Map created via Paint 2, version 5.3.1 (250) (TryBest Studio), using data from the Taiwan National Land Surveying and Mapping Center.

Parameters of sample collection.

Information on sampling sites, weather, human population grade, rainfall season, the river pollution index (RPI), distance to the river origin, and livestock and poultry densities around the sampling site were evaluated. In addition, the relationship between ESBL-producing E. coli isolates and these parameters was analyzed.

In accordance with the climate statistics provided by the Central Weather Bureau of Taiwan, the period from February to April was defined as the spring (dry) season, during which the mean temperature and precipitation of each month were <25°C and <100 mm, respectively. The period from May to August was defined as the summer (rainy) season, during which the mean temperature and precipitation of each month were >25°C and >100 mm, respectively. The human population around the sampling site and the population grade of each county or town were provided by the Ministry of the Interior of Taiwan. The population of a county or town was categorized as grade 1 for <49,000, grade 2 for between 49,000 and 93,000, grade 3 for between 93,000 and 137,000, and grade 4 for >137,000 people. The RPI was graded according to the definitions provided by the TEPA, which include the concentrations of 4 parameters in water: dissolved oxygen (DO), biochemical oxygen demand (BOD5), suspended solids (SS), and ammonia nitrogen (NH₃-N). We defined the RPI as the average of these 4 parameters, wherein an RPI of <2, 2 to 3, 3 to 6, and >6 was determined as nonpolluted, lightly polluted, moderately polluted, and severely polluted, respectively (20). We defined the upper, middle, and lower streams of a river based on the distance from the sampling site to the river origin, wherein the upper, middle, and lower streams were the first third, middle third, and final third of the river length, respectively. We collected information on livestock and poultry distribution from the Council of Agriculture, Taiwan. We defined totals of <100,000, between 100,000 and 500,000, and >500,000 chickens being raised per month as low, middle, and high densities of chicken farming, respectively. In a similar manner, we defined totals of <10,000, between 10,000 and 50,000, and >50,000 pigs being farmed per month as low, middle, and high densities of pig farming, respectively. We included 128 districts or counties in our study region. After excluding the area without chicken farms, we selected the tertiles as the cut points, which are approximately 100,000 and 500,000 chickens per month, for categorization. For pig farms, we selected 10,000 and 50,000 pigs per month as the cut points, because the chicken-farming density scale is approximately 10 times greater than the pig-farming density scale in southern Taiwan.

Identification and purification of E. coli.

The river water samples were tested for E. coli by using the membrane filtration method. We used the decimal serial dilutions of river water filtered through 0.45-μm-pore-size membranes and followed the official method published by the TEPA (TEPA method NIEA E202.55B) (21). A series of 10-fold dilutions of the water sample were conducted using a sterile phosphate-buffered magnesium chloride solution as dilution blanks. The membranes were placed on an LES (Lawrence Experimental Station) endo agar plate (Bottle M Endo agar LES 500G; Becton Dickson and Company) and incubated at 35°C ± 1°C for 24 h. The E. coli was extracted from colonies that had a green metallic sheen and suspended in 1 ml of phosphate-buffered saline (PBS). We selected as many samples as possible and up to 10 colonies from each LES endo agar plate. One hundred microliters of the highest dilution displaying growth was spread on Luria-Bertani–ampicillin agar plates and incubated at 37°C overnight. Each of these colonies was subcultured 3 additional times. The purified isolates were inoculated on E. coli CHROMagar (ECC) plates (CHROMagar, Paris, France) (22) and further incubated overnight at 37°C to confirm the identification of E. coli.

CHROMagar ESBL plates (CHROMagar, Paris, France) were used to identify ESBL-producing E. coli according to the manufacturer's instructions.

Detection of O25b-ST131, multilocus sequence typing (MLST), and bla_CTX-M gene groups.

For chromosomal DNA preparation, a single colony of E. coli was suspended in 200 μl of sterile double-distilled water and heated to 100°C for 10 min. After cooling, the samples were centrifuged, and 2 μl of the supernatant containing chromosomal DNA was used as a template for PCR amplification.

The O25b serotype was determined by employing the methods of Clermont et al. (23), and the following screening primers were used: rfb.1bis (5′-ATACCGACGACGCCGATCTG-3′) and rfbO25b.r (5′-TGCTATTCATTATGCGCAGC-3′) (24, 25).

According to the stratification of the O25b status (Fig. 2), we used a random-number generator in selecting 92 isolates for MLST analysis. All isolates with the O25b serotype (12 ESBL-producing and 34 non-ESBL-producing) were evaluated through MLST analysis. Twelve O25-positive, ESBL-positive and 34 O25-positive, ESBL-negative E. coli strains were matched for randomly selecting O25-negative, ESBL-positive and O25-negative, ESBL-negative E. coli isolates.

FIG 2.

Sample selection for detecting the sequence type of E. coli from river waters in southern Taiwan.

The MLST grouping of E. coli was determined by analyzing 8 housekeeping gene sequences (adk, fumC, gyrB, icd, mdh, purA, and recA) on the MLST database website (http://mlst.warwick.ac.uk/mlst/dbs/Ecoli) (26). The eBURST (http://eburst.mlst.net/) algorithm was used to determine the clonal complexes of sequence types (STs) and sequence type complexes (STCs), where STs sharing at least 5 loci in common with at least one other member of the group were assigned to defined STCs; STs with the highest number of single-locus variants (SLVs) were assigned as the founder. In addition, the bla_CTX-M gene of all ESBL-producing E. coli isolates was investigated. The CTX-M groups 1, 2, 8, and 9, as well as SHV and TEM, were detected by performing multiplex PCR with specific primers, as shown in Table 1 (26, 27). Specific PCRs were performed to detect bla_CTX-M-14 and bla_CTX-M-15 by employing the methods of Sidjabat et al. and Chia et al. (25, 26).

TABLE 1.

O25b CTX-M group-specific primers used for multiplex PCR

PCR and primer names	Sequencea (5′–3′)	Reference
Multiplex II, CTX-M group 1, group 2, and group 9
MultiCTXMGp1_for	TTAGGAARTGTGCCGCTGYA	27
MultiCTXMGp1-2_rev	CGATATCGTTGGTGGTRCCAT
MultiCTXMGp2_for	CGTTAACGGCACGATGAC
MultiCTXMGp1-2_rev	CGATATCGTTGGTGGTRCCAT
MultiCTXMGp9_for	TCAAGCCTGCCGATCTGGT
MultiCTXMGp9_rev	TGATTCTCGCCGCTGAAG
CTX-M group 8/25
CTX-Mg8/25_for	AACRCRCAGACGCTCTAC
CTX-Mg8/25_rev	TCGAGCCGGAASGTGTYAT
CTX-M-14
CTX-14F	TACCGCAGATAATACGCAGGTG	26
CTX-14R	CAGCGTAGGTTCAGTGCGATCC
CTX-M-15
CTX-M-15-SF	CACACGTGGAATTTAGGGACT	25
CTX-M-15-SR	GCCGTCTAAGGCGATAAACA
Multiplex I, TEM and SHV
MultiTSO-T_for	CATTTCCGTGTCGCCCTTATTC	27
MultiTSO-T_rev	CGTTCATCCATAGTTGCCTGAC
Multiplex PCR
SHV-F	AACGGAACTGAATGAGGCGCT	26
SHV-R	TCCACCATCCACTGCAGCAGCT

^aY = T or C; R = A or G; S = G or C; D = A, G, or T.

Statistical analysis.

All statistical analyses were performed using the Statistical Package for Social Sciences (version 22.0) software package for Windows (SPSS Inc., Chicago, IL). Data concerning patients with ESBL-producing and non-ESBL-producing E. coli were expressed as percentages of each subgroup of patients and were compared by conducting the χ² test for categorical variables.

RESULTS

Identification of ESBL-producing E. coli in river water.

ESBL-producing E. coli accounted for 186 (30%) of the 621 E. coli strains isolated from 280 river samples in southern Taiwan over the 7-month study period. All 621 E. coli isolates were screened for the O25b serotype through PCR assay. Serotype O25b accounted for 46 (7.4%) of the 621 isolates (Table 2). Although this serotype was observed more frequently among the ESBL-producing E. coli than among the non-ESBL-producing E. coli (12/186 [6.5%] versus 34/435 [2.8%]), the difference was statistically nonsignificant (P = 0.55).

TABLE 2.

Numbers of O25b and non-O25b serotypes in ESBL-producing and non-ESBL-producing E. coli^a

Serotype	No. of E. coli strains		Total
	ESBL (n = 186)	Non-ESBL (n = 435)
O25b	12	34	46
Non-O25b	174	401	575
Total	186	435	621

^aNumbers of O25b and non-O25b serotypes in ESBL-producing E. coli (n = 186) and non-ESBL-producing E. coli (n = 435) are from the 621 E. coli strains isolated from river waters in southern Taiwan.

As shown in Table 3, among the 12 ESBL-producing E. coli isolates belonging to the O25b serotype, 4 strains carried genes encoding CTX-M ESBLs, including 2 carrying the CTX-M group 2 gene, one carrying the CTX-M group 9 gene, and one carrying the CTX-M (3, 15) gene. Among the 186 ESBL-producing E. coli isolates, 18 carried the CTX-M group 1 gene, 3 carried the CTX-M group 2 gene, 6 carried the CTX-M group 8 gene, and 43 carried the CTX-M group 9 gene.

TABLE 3.

Number of CTX-M, SHV, and TEM genes detected in O25b ESBL-producing E. coli and non-O25b ESBL-producing E. coli^a

Gene type	No. of genes in each E. coli serotype		Total
	O25b ESBL (n = 12)	Non-O25b ESBL (n = 174)
CTX-M group 1	0	18	18
CTX-M group 2	2	1	3
CTX-M group 8	0	6	6
CTX-M group 9	1	42	43
SHV	0	23	23
TEM	7	89	96

^aShown are the numbers of genes of CTX-M group 1, group 2, group 8, and group 9 and SHV and TEM detected in O25b ESBL-producing E. coli (n = 12) and non-O25b ESBL-producing E. coli (n = 174) from 186 ESBL-producing E. coli strains isolated from river waters in southern Taiwan.

Among the 24 ESBL-producing E. coli isolates and 68 non-ESBL-producing E. coli isolates selected for MLST analysis, the sequence types were identified for 83 isolates. Genetic relatedness within each clonal complex is shown in Fig. 3 (based on data from the MLST database [http://mlst.warwick.ac.uk/mlst/]). Among the 83 E. coli isolates, the eBURST diagram revealed that the clustering of E. coli STs belonged to 6 main complexes (A, STC10; B, STC58; C, STC46; D, STC4623; E, STC641; and F, STC336); 15 isolates belonged to the ST10 clonal complex, and 18 belonged to the ST58 clonal complex (including 7 ST58 strains and 5 ST155 strains). No ST131 clonal complex was detected among all 92 E. coli isolates.

FIG 3.

eBURST diagram showing the clustering of E. coli STs belonging to the 6 main complexes isolated from river waters in southern Taiwan: A, clonal complex of sequence type 10 (STC10; n = 15); B, STC58 (n = 18); C, STC46 (n = 11); D, STC4623 (n = 4); E, STC641 (n = 9); F, STC336 (n = 4). E. coli strains from the same ancestor based on the entire MLST data set were grouped and are indicated with black circles. Each ST is represented as a node. Single-locus variants (SLV) are connected with a purple line, and double-locus variants (DLV) are connected with a blue line. Strains that shared 5 of the 7 alleles were grouped into a single clonal complex.

Among the identifiable 83 isolates detected from MLST analysis, 7 isolates belonged to the ST10 or ST18 clonal complex in the ESBL-producing group, and 26 isolates belonged to the ST10 or ST58 clonal complex in the non-ESBL-producing group.

Relationship between river water samples and evaluated parameters.

The geographic distribution of the proportion of ESBL-producing E. coli in relation to the number of chickens being raised per month and RPI in southern Taiwan is displayed in Fig. 4. As shown in Table 4, no significant differences in the proportion of ESBL-producing E. coli strains were observed among the seasons, stream sites, populations, and numbers of pigs. The proportion of ESBL-producing E. coli was significantly higher in the low-RPI group (P = 0.025). In addition, the proportion of ESBL-producing E. coli strains was significantly higher in regions with a large number of chickens being raised (P = 0.013). No significant difference was observed in the proportion of the ST10 or ST58 clonal complex between the high- and low-density chicken farming areas (P = 0.67).

FIG 4.

Geographic distribution of the proportion of ESBL-producing E. coli in relation to the number of chickens being raised and river pollution index (RPI) in southern Taiwan. The background of the map with the color gradient presents the number of chickens raised in each county; the dots with a size display the proportion of ESBL-producing E. coli in each sampling point, and the color of the dots represents the RPI. Map created via Paint 2, version 5.3.1 (250) (TryBest Studio), using data from the Taiwan National Land Surveying and Mapping Center.

TABLE 4.

Characteristics of ESBL-producing and non-ESBL-producing E. coli of 621 E. coli strains isolated from river waters in southern Taiwan

Characteristic	No. (%) of isolates		P value
	Non-ESBL (n = 435)	ESBL (n = 186)
Seasona			0.80
Dry (2–4 mo)	133 (70.7)	55 (29.3)
Rain (5–8 mo)	302 (69.7)	131 (30.3)
Regionb			0.999
Upper	44 (69.8)	19 (30.2)
Middle	166 (70.0)	71 (30.0)
Lower	225 (70.1)	96 (29.9)
Populationc			0.19
≤137,000	423 (70.5)	177 (29.5)
>137,000	12 (57.1)	9 (42.9)
No. of animals being raised
Chickend			0.013
<100,000	131 (78.9)	35 (21.1)
100,000–500,000	192 (66.2)	98 (33.8)
>500,000	112 (67.9)	53 (32.1)
Pige			0.55
<1,000	205 (71.7)	81 (28.3)
10,000–50,000	124 (67.0)	61 (33.0)
>50,000	106 (70.7)	44 (29.3)
RPIf			0.025
≤6	327 (67.8)	155 (32.2)
>6	108 (77.7)	31 (22.3)

^aDry season is defined from February to April and wet season from May to August.

^bWater collected in the region is classified as upper, middle, and lower stream by the distance of the sampled station to the river origin.

^cClassified by the population density at the sampled station.

^dClassified by the number of chickens produced in the region of the sampled station in a month.

^eClassified by the number of pigs produced in the region of sampled station in a month.

^fCalculated by the average of the concentrations of four parameters in water: dissolved oxygen (DO), biochemical oxygen demand (BOD5), suspended solids (SS), and ammonia nitrogen (NH3-N).

DISCUSSION

Several studies have been conducted on the isolation of ESBL-producing E. coli from river water (13–15). The difference in the abundance of human and animal ESBL-producing E. coli clones in various river waters may be related to the nature of their drainage basins, the intensity of antibiotic use, and the relative amount of fecal contamination from humans and farm animals. Clonal groups similar to those found in human infections have been isolated from contaminated rivers adjacent to large cities (13–15). In contrast, nonhuman clonal groups more frequently are isolated from rural streams (28–30). The isolation rate of ESBL-producing E. coli was approximately 30% in our study, indicating their high numbers in environmental water. Zurfluh et al. reported an isolation rate of 36.2% for ESBL-producing E. coli from river and lake water in Switzerland (15). In a Chinese study, an isolation rate of 14.8% was reported for ESBL-producing E. coli from downstream water (31). However, research investigating river water as a potential reservoir of ESBL-producing E. coli is limited.

In our previous study, we observed that most ESBL-producing clinical E. coli strains isolated from infants belonged to CTX-M-14 in either ST131 or non-ST131 clones (10). In the current study, the most predominant clones observed were ST10 and ST58 clonal complexes. These ESBL-producing clones are associated with chickens and are unrelated to those found in human infections.

Colomer-Lluch et al. isolated quinolone-resistant E. coli belonging to the clonal groups O25b-H4 B2-ST131 and O25b:H4-D-ST69 from raw sewage and river water in Spain (13). In contrast, Dhanji et al. did not detect bla_CTX-M-15, the most common ESBL-encoding gene in clinical E. coli strains in the United Kingdom, in the Thames in London (14, 32). Geser et al. (33) identified CTX-M-1 ESBL-encoding genes in Enterobacteriaceae isolates from food-producing animals in Switzerland. However, the predominant ESBL-producing clinical isolates in Switzerland belonged to CTX-M-15 (34).

As shown in Table 5, various ESBL-producing E. coli clones have been isolated from animals, although ST131 clones have not been detected in certain studies (28, 30, 35–40, 42–45). The ST10, ST58, and ST155 clones have been detected in animals, water, and human isolates in the Netherlands, Canada, Chile, and China (28, 30, 35, 38–45). We detected numerous ST10 or ST58 clonal complexes in high-density chicken-farming areas. Additional studies are warranted to elucidate the virulence potential of E. coli ST10 and ST58 clonal groups.

TABLE 5.

Summary of 13 major animal or human E. coli sequence type studies conducted after 2000

Author	Yr	Country	Sample origin (no.)	Most common serotype (%)	Dominant ST	ST131 (no.)	ST10 (no.)	ST58/ST155 (no.)	Reference
Overdevest et al.	2009	Netherlands	Chicken meat (89), beef (85), pork (57), mixed meat (22), other meat (9)	Chicken (CTX-M-1; 58.1), all other meat (CTX-M-1; 62.5)	ST117, ST10	Not found	Found	ST155	35
Ma et al.	2006	China	Duck feces (230), water (15)	CTX-M-27/55/24/105/14 (27.7/22/21.1/16.2/12.2)	NA	NA	NA	NA	36
Dierikx et al.	2009	Denmark	Broiler feces (1,066), farmer's feces (18)	CTX-M-1, CMY-2, SHV-12	ST93	Not found	Not found	Not found	37
Rashid et al.	2010	Bangladesh	OBSa feces (170), water (8)	CTX-M-15 (100)	ST156	Not found	ST10CC (1)	Not found	28
Hernandez et al.	2009	Chile	Gull feces (370), human feces (49)	CTX-M group 1 (94.6)	ST10	Not found	ST10 (24)	ST58 (4), ST155 (1)	30
Bergeron et al.	2005–2007	Canada	Human urine (475), meat of chicken (253), beef (242), pork (242), animal feces (349)	NAb	ST117, ST10	Not found	ST10(13)	Not found	38
Blaak et al.	2011	Netherlands	Flies (87), poultry manure in broiler farm (10), laying-hen farm (10)	SHV-12, CTX-M-1, TEM-52	ST58, ST10	Not found	ST10 (4)	ST58 (6), ST155 (2)	39
Cortes et al.	2003	Spain	Floor of poultry (59) and pig (27) farms	Poultry, CTX-M-9 group (64.9) and CTX-M-1 group (5.3); pig, CTX-M-1 group (69.0)	ST155, ST10	ST131 (1)	ST10 (3)	ST155 (4)	40
Peirano et al.	2000–2010	Canada	Human blood (4,698)	CTX-M-15, 119 (60); CTX-M-14, 58 (29)	ST131	ST131 (117)	ST10CC (13)	Not found	41
Leverstein-van Hall et al.	2006–2010	Netherlands	Chicken meat (98), floor of poultry farm (35), human (516)	Poultry, CTX-M-1 (49); chicken meat, CTX-M-1 (49); human, CTX-M-1 (24)	ST10, ST58	Not found	Human (3), meat (4)	Human, ST58 (3)	42
Hu et al.	2010–2013	China	Swine feces (31), water (26), human urine, blood, sputum, and body secretion (82)	CTX-M-1 group (33), CTX-M-9 (99)	ST131, CC10	ST131 (20)	CC10 (19)	NA	43
Ding et al.	2006–2007	China	Pig tissue (81)	NA	CC10	Not found	CC10 (26)	CC58 (8)	44
Borjesson et al.	2010	Sweden	Broiler feces (10)	CTX-M-1 (10)	ST155	Not found	Not found	ST155 (3)	45

^aOBS, open bill stork (Anastomus oscitans).

^bNA, not applicable.

We observed a significant association of ESBL-producing E. coli clones isolated from river water with the presence of chicken farms but not with pig farms (P = 0.013). This finding may be explained by the shorter period and higher antibiotic usage per average body weight for poultry compared to those for pigs. In southern Taiwan, antibiotic usage is 10 times higher in chickens than in pigs (42), although dosing is higher in pigs (43). According to the Annual Report of Food and Drug Research of 2014, released by the Food and Drug Administration, Taiwan, antibiotic residues were detected with greater frequency in chicken meat than in pig meat (46). In addition, a significant difference was observed in the proportion of ESBL-producing E. coli among different RPIs in our series (P = 0.025); this difference may be due to chicken farms being located mostly in suburban areas, where the river water is less polluted. The relationship between resistant bacteria and the RPI remains controversial. Multidrug-resistant bacteria in rivers may have developed because of the deterioration of water quality resulting from the flow of untreated sewage from domestic wastewater or intensive agricultural and industrial activities (47). However, the detection of multidrug-resistant bacteria in chlorinated water also has been reported (48, 49).

Our study has several limitations. Data from southern Taiwan may be inapplicable to other countries because of the geographic diversity of E. coli strains and their variable virulence and resistance genes. For example, ESBL-producing E. coli strains expressing bla_CTX-M-14 are prevalent in Taiwan, whereas those expressing bla_CTX-M-15 are more common in other parts of the world (11). Although we selected sample stations based on the surveillance data from the TEPA, water samples may not be independent because certain stations are located on the same river; however, the distribution of sampling stations is essential for the analysis of regional characteristics. Furthermore, the replication of E. coli in a single water sample is possible. We did not isolate the DNA of all E. coli strains, as has been conducted in previous studies; therefore, it is difficult to evaluate the replication (13, 15). Moreover, we did not detect ESBL-associated genes, including bla_CTX-M, bla_SHV, and bla_TEM, from certain ESBL-producing E. coli isolates in our study. We attribute this finding to the difference between the genes identified through PCR assay and the design of CHROMagar ESBL plates or other ESBL-associated genes, such as bla_OXA, which we did not detect. In addition, we did not detect several resistant genes reported in other studies; for instance, bla_CMY-2 may contribute to cefotaxime resistance. In an Asian surveillance study on ESBL-producing Enterobacteriaceae, isolates were reported to have a higher prevalence of AmpC bla_CMY-type β-lactamases, particularly bla_CMY-2, in Taiwan (50). Additional epidemiological studies are required to correlate river water, animal contact, poultry consumption, and other environmental exposures with the emergence of ESBL-producing E. coli.

In conclusion, this study investigated ESBL-producing E. coli in various river waters in southern Taiwan. We isolated several ESBL-producing E. coli clones found in poultry and other farm animals but not the O25b-ST131 clone, which causes human infections. These environmental ESBL-producing clones potentially can spread among humans, rendering treatment difficult. The continued surveillance of food and environment is essential to anticipate emerging infections.

Funding Statement

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.