Analysis of Salmonella enterica Isolated from a Mixed-Use Watershed in Georgia, USA: Antimicrobial Resistance, Serotype Diversity, and Genetic Relatedness to Human Isolates

ABSTRACT

As the cases of Salmonella enterica infections associated with contaminated water are increasing, this study was conducted to address the role of surface water as a reservoir of S. enterica serotypes. We sampled rivers and streams (n = 688) over a 3-year period (2015 to 2017) in a mixed-use watershed in Georgia, USA, and 70.2% of the total stream samples tested positive for Salmonella. A total of 1,190 isolates were recovered and characterized by serotyping, antimicrobial susceptibility testing, and pulsed-field gel electrophoresis (PFGE). A wide range of serotypes was identified, including those commonly associated with humans and animals, with S. enterica serotype Muenchen being predominant (22.7%) and each serotype exhibiting a high degree of strain diversity by PFGE. About half (46.1%) of the isolates had PFGE patterns indistinguishable from those of human clinical isolates in the CDC PulseNet database. A total of 52 isolates (4.4%) were resistant to antimicrobials, out of which 43 isolates were multidrug resistant (MDR; resistance to two or more classes of antimicrobials). These 52 resistant Salmonella isolates were screened for the presence of antimicrobial resistance genes, plasmid replicons, and class 1 integrons, out of which four representative MDR isolates were selected for whole-genome sequencing analysis. The results showed that 28 MDR isolates resistant to 10 antimicrobials had bla_cmy-2 on an A/C plasmid. Persistent contamination of surface water with a high diversity of Salmonella strains, some of which are drug resistant and genetically indistinguishable from human isolates, supports a role of environmental surface water as a reservoir for and transmission route of this pathogen.

IMPORTANCE Salmonella has been traditionally considered a foodborne pathogen, as it is one of the most common etiologies of foodborne illnesses worldwide; however, recent Salmonella outbreaks attributed to fresh produce and water suggest a potential environmental source of Salmonella that causes some human illnesses. Here, we investigated the prevalence, diversity, and antimicrobial resistance of Salmonella isolated from a mixed-use watershed in Georgia, USA, in order to enhance the overall understanding of waterborne Salmonella. The persistence and widespread distribution of Salmonella in surface water confirm environmental sources of the pathogen. A high proportion of waterborne Salmonella with clinically significant serotypes and genetic similarity to strains of human origin supports the role of environmental water as a significant reservoir of Salmonella and indicates a potential waterborne transmission of Salmonella to humans. The presence of antimicrobial-resistant and MDR Salmonella demonstrates additional risks associated with exposure to contaminated environmental water.

INTRODUCTION

Salmonella enterica is the most common cause of foodborne bacterial illnesses in the United States, causing approximately 1.4 million illnesses every year (1, 2). Salmonella has been traditionally considered a zoonotic foodborne pathogen that is associated with food animals; however, cases associated with other sources are on the rise. Outbreaks associated with fresh produce, including melons, sprouts, tomatoes, and lettuce, as well as water sources, such as community water systems and recreational water, have been reported in the United States (3–6). Irrigation water contaminated with feces of infected humans or animals has been pointed out to be one of the potential sources of Salmonella contamination of fresh produce, indicating that there are environmental sources of Salmonella (7, 8). Studies have shown that Salmonella is commonly detected in both fresh and marine surface waters, including rivers, streams, lakes, creeks, and coastal water (7, 9–12). The widespread occurrence of this pathogen in the aquatic environment and its role as an etiology of waterborne outbreaks suggests that water can serve as a reservoir for Salmonella and play a role in its transmission.

Salmonella causes symptoms ranging from self-limiting gastroenteritis to life-threatening invasive systemic diseases such as typhoid and paratyphoid fevers (13). Gastroenteritis does not often require antimicrobial therapy; however, antimicrobial treatment is required for the immunocompromised, who may develop enteric fever (14). Salmonella strains resistant to therapeutic antimicrobial agents are a public health concern, as they limit treatment options and are expected to cause more severe infections than the susceptible bacteria (15). The Centers for Disease Control and Prevention (CDC) estimates antimicrobial-resistant nontyphoidal Salmonella strains to cause 212,500 infections and 70 deaths annually in the United States (16). They represent a hazard for human health, as they are responsible for higher rates of hospitalization, morbidity, and mortality as well as a higher cost of treatment (14, 15, 17, 18). Studies have shown that antimicrobial-resistant Salmonella strains are often detected outside their hosts in the water environment (19–22). These bacteria not only have a potential to be transferred to those exposed to the contaminated water, but also can disseminate their antimicrobial resistance determinants, such as antimicrobial resistance genes that are able to render antimicrobials ineffective, to other bacteria present in the environment.

The Salmonella genus consists of two species, S. bongori and S. enterica, which comprises six subspecies, which in turn include over 2,500 serotypes which vary in host ranges and capabilities to infect hosts (23). However, only a small number of serotypes that belong to S. enterica subsp. enterica account for most human illnesses (24). Due to such differences in their characteristics and behaviors, identifying Salmonella to the serotype level is important for epidemiological investigations. In the current study, all serotypes will be listed in their shortened form (i.e., S. Enteritidis for S. enterica subsp. enterica serotype Enteritidis).

The aquatic environment is a potential contributor of Salmonella transmission to humans; therefore, it is necessary to investigate the distribution of Salmonella in surface water to develop measures to reduce the spread of Salmonella to humans. However, studies of the presence of Salmonella in the aquatic environment are limited compared to those of Salmonella in humans and animals. Furthermore, studies of the environmental isolates investigating antimicrobial resistance genes and associated mobile genetic elements (MGEs) involved in horizontal gene transfer are even fewer. Since resistant Salmonella strains are often detected in the water environment, increasing the risk of antimicrobial resistance dissemination to humans, the prevalence of resistant Salmonella in surface water needs to be studied to understand the role of environmental water in the emergence and spread of antimicrobial resistance in Salmonella. In this study, we isolated Salmonella from a mixed-use watershed in northeast Georgia, USA, over a 3-year period and characterized the isolates phenotypically and genotypically to determine their prevalence and diversity in the aquatic environment. The molecular mechanisms underlying the antimicrobial resistance and the location of antimicrobial resistance genes on MGEs were analyzed in resistant isolates. This study attempts to enhance the overall knowledge of the prevalence, diversity, antimicrobial resistance phenotype, and antimicrobial resistance gene mechanism of Salmonella present in surface water. The results will improve our understanding of the waterborne Salmonella.

RESULTS

Isolation and identification of Salmonella.

(i) Prevalence of Salmonella. A total of 688 water samples were collected from the 12 sampling events. The number of samples positive for Salmonella and the number of isolates recovered from the samples are shown in Table 1. Although only one colony was selected from each positive plate, higher numbers of isolates than the numbers of sites are due to the use of multiple media to recover the bacteria. A maximum of four isolates were collected per sample, adding up to a total of 1,190 Salmonella isolates (see Table S1 in the supplemental material). Overall, Salmonella was detected at all sampling sites (n = 105) during at least one of the 12 sampling events. The recovery rate of Salmonella varied from 30.5% to 92.6% of the total sites for each sampling event. A significantly higher recovery rate was observed during the summer seasons than during the other seasons (P < 0.05) (Table 1). The highest positive correlation between the amount of rainfall (cm) and Salmonella recovery rate was the amount of rain observed for the 3 days preceding each sampling event, although this correlation was not significant (r = 0.4987, P = 0.0939) (Table 2). This correlation weakened for precipitation occurring ≥4 days before each sampling event. A positive correlation was also observed between the water temperature (°C) and Salmonella detection rate (r = 0.7101, P < 0.05) (Table 2).

TABLE 1.

Salmonella recovered from surface water

Sampling season (no. of samples)	Percentage of positive sites (no. of isolates recovered)a
Winter 2015 (30)	70.0 (59)
Spring 2015 (100)	68.0 (153)
Summer 2015 (33)	81.8 (66)
Fall 2015 (59)	30.5 (37)
Winter 2016 (41)	63.4 (83)
Spring 2016 (87)	77.0 (190)
Summer 2016 (27)	92.6 (67)
Fall 2016 (81)	72.8 (154)
Winter 2017 (53)	58.5 (79)
Spring 2017 (94)	78.7 (147)
Summer 2017 (40)	90.0 (77)
Fall 2017 (43)	72.1 (78)

^aHigher numbers of isolates than the numbers of samples are due to the use of several media to recover Salmonella.

TABLE 2.

Precipitation for 3 days preceding sampling and average water temperature and their correlations with the Salmonella detection rate

Sampling date	Cumulative rainfall 3 days preceding the sampling (cm)	Avg water temp (°C)a
2015 winter	0.85	ND
2015 spring	0.31	ND
2015 summer	0	ND
2015 fall	0	11.72
2016 winter	0	9.15
2016 spring	1.37	16.74
2016 summer	4.97	24.47
2016 fall	0	13.67
2017 winter	0.06	8.35
2017 spring	3.54	13.44
2017 summer	0.31	22.39
2017 fall	0.52	11.24
Correlation	0.4987	0.7101
P value	0.0939	0.0078

^aND, not determined.

(ii) Diversity of Salmonella. The serotype distribution of Salmonella isolates recovered from surface water is shown in Table 3. Of 1,190 S. enterica isolates, 21 isolates belonged to S. enterica subspecies arizonae and houtenae, while the serotype of one isolate was unidentified, which was a rough variant that does not express O antigen. The remaining 1,168 isolates belonged to S. enterica subspecies enterica, and 48 different serotypes were identified among the isolates. The serotypes included 18 of the 20 most frequently identified human clinical Salmonella serotypes in the United States: S. Enteritidis, S. Newport, S. Typhimurium, S. Javiana, S. I 4,[5],12:i:-, S. Infantis, S. Muenchen, S. Montevideo, S. Braenderup, S. Thompson, S. Saintpaul, S. Oranienburg, S. Mississippi, S. Bareilly, S. Berta, S. Agona, S. Paratyphi B var. L(+) tartrate+, and S. Anatum (25). Of the 40 most frequently reported food-producing animal serotypes, 27 were also identified (26). The most prevalent serotype in surface water was S. Muenchen (22.7%; 270/1,190), followed by S. Rubislaw (12.9%; 153/1,190), S. Montevideo (9.5%; 113/1,190), S. Newport (7.7%; 92/1,190), and S. Hartford (7.3%; 87/1,190), accounting for 60.1% (715/1,190) of the total isolates from the watershed. These serotypes were recovered 10 or more times out of the 12 sampling events, and S. Rubislaw was the only serotype that was recovered all 12 times. The highest serotype diversity was seen during the spring (Table 3); the number of serotypes recovered during the spring was significantly higher than those of winter and summer (P < 0.05).

TABLE 3.

Serotypes of Salmonella isolated from surface water

Serotype	No. of isolates													Human ranka	Animal rankb
	2015				2016				2017				Total no. of isolates
	Winter (n = 59)	Spring (n = 153)	Summer (n = 66)	Fall (n = 37)	Winter (n = 83)	Spring (n = 190)	Summer (n = 67)	Fall (n = 154)	Winter (n = 79)	Spring (n = 147)	Summer (n = 77)	Fall (n = 78)
Agona	0	0	0	0	1	0	0	0	0	0	0	0	1	18	13
Anatum	4	0	0	0	17	3	0	8	0	4	0	0	36	20	10
Aqua	0	0	0	0	0	2	0	0	4	0	0	1	7
Baildon	0	0	0	0	0	0	0	0	0	2	0	0	2
Bareilly	7	4	0	1	0	0	1	9	3	2	3	0	30	16
Berta	0	0	0	0	0	1	0	0	2	0	0	0	3	17	26
Braenderup	1	7	0	9	0	5	0	7	3	7	0	0	39	9	25
Brandenburg	0	0	0	0	0	1	0	0	0	0	0	0	1		33
Brazil	0	0	0	0	0	0	0	0	1	0	0	0	1
Cerro	0	0	0	0	0	0	0	0	2	0	0	0	2		18
Cubana	3	0	4	0	0	5	0	0	0	0	0	2	14
Derby	0	0	0	0	0	0	0	0	0	1	0	0	1		16
Enteritidis	0	1	0	0	0	0	0	0	0	0	0	0	1	1	2
Gaminara	0	2	8	2	0	0	4	1	0	4	4	0	25
Give	8	6	0	0	18	8	0	2	7	7	0	1	57		32
Hartford	5	11	1	0	8	16	1	3	10	28	1	2	86
Havana	0	0	0	0	0	0	0	0	0	0	0	3	3
Infantis	3	5	0	2	3	0	0	2	4	0	0	3	22	6	7
Inverness	0	0	3	0	0	1	0	1	0	0	0	0	5
I 4,[5],12:b:-	10	8	0	3	10	6	0	3	7	5	0	1	53
I 4,[5],12:i:-	0	2	0	0	1	0	0	0	0	0	0	0	3	5	8
Javiana	0	0	0	0	0	0	0	0	2	1	0	0	3	4
Kentucky	0	3	0	0	0	0	0	1	0	0	0	1	5		1
Kiambu	0	0	0	0	0	0	0	3	0	0	0	0	3		38
Kintambo	0	0	0	0	0	1	0	0	0	0	0	0	1
Litchfield	0	3	0	0	0	0	0	1	0	0	0	0	4		37
Liverpool	0	0	0	0	0	0	2	0	0	0	0	0	2
Luciana	0	0	0	0	0	1	0	0	0	1	0	0	2
Mbandaka	0	0	0	0	0	2	7	0	0	0	0	0	9		19
Meleagidis	0	0	0	0	0	0	0	0	0	0	2	0	2		24
Mississippi	0	2	0	0	0	3	0	1	0	0	0	1	7	14
Montevideo	4	11	9	0	5	11	5	24	2	12	22	8	113	8	5
Muenchen	0	35	28	8	7	40	11	48	13	30	37	13	270	7	17
Muenster	3	7	2	1	0	3	0	0	0	0	0	1	17		20
Newport	1	9	5	3	4	28	8	18	9	8	0	6	99	2	12
Oranienburg	0	4	0	0	0	0	1	0	0	0	0	27	32	13	35
Orion	0	0	0	0	0	4	0	0	0	0	0	0	4
Ouakam	0	0	0	0	0	0	0	1	0	0	0	0	1
Paratyphi_B_var._L-tartrate+	0	0	0	0	0	0	0	0	0	1	0	0	1	19
Rubislaw	1	23	4	4	6	42	18	14	9	22	7	3	153
Saintpaul	0	4	0	0	0	0	0	0	0	0	0	0	4	11	14
Schwarzengrund	0	3	0	0	1	2	0	0	1	2	0	0	9		9
Senftenberg	0	0	0	0	0	0	2	0	0	0	0	2	4		21
Soerenga	0	0	0	0	0	0	5	0	0	0	0	0	5
Tennessee	0	0	0	0	0	0	0	0	0	0	0	2	2
Thompson	0	0	0	0	0	3	0	0	0	1	0	0	4	10	22
Typhimurium	6	2	0	0	1	0	0	4	0	6	0	0	19	3	3
Worthington	0	0	0	0	0	0	0	0	0	0	0	1	1		31
subsp. arizonae	1	1	2	4	1	1	2	3	0	3	1	0	19
subsp. houtenae	2	0	0	0	0	0	0	0	0	0	0	0	2
Untypeable	0	0	0	0	0	1	0	0	0	0	0	0	1

^aThe 20 most frequently reported human clinical serotypes from the CDC report, 2016 (25).

^bThe 40 most frequently reported food-producing animal serotypes from the NARMS report, 2006 to 2015 (49).

When the pulsed-field gel electrophoresis (PFGE) patterns of all Salmonella isolates were analyzed, a high degree of strain diversity was observed within each serotype. Table 4 shows the strain diversity of 16 S. enterica subspecies Enterica serotypes whose number of isolates totaled 10 or more. S. Muenchen and S. Rubislaw exhibited the highest number of PFGE patterns with 141 and 97 unique patterns, respectively, followed by S. Montevideo, S. Newport, and S. Hartford. The highest diversity was observed in 19 S. Typhimurium isolates with 14 unique PFGE patterns, while the lowest diversity was observed in 32 S. Oranienburg isolates with 3 unique PFGE patterns. However, a low diversity seen in S. Oranienburg is due to 27 clonal isolates that were recovered during one sampling event. The next lowest diversity, apart from S. Oranienburg, is S. Anatum with 36 isolates with 7 unique PFGE patterns.

TABLE 4.

Strain diversity of S. enterica subspecies enterica serotypes isolated from surface water

Serotype	Total no. of isolates	Percentage of total isolates	No. of samplings recovered	No. of PFGE patterns
Muenchen	270	22.7	11	141
Rubislaw	153	12.9	12	97
Montevideo	113	9.5	11	31
Newport	92	7.7	10	31
Hartford	87	7.3	11	20
Give	57	4.8	8	31
I 4,[5],12:b:-	53	4.5	9	17
Braenderup	39	3.3	7	11
Anatum	36	3.0	5	7
Infantis	34	2.9	8	13
Oranienburg	32	2.7	3	3
Bareilly	29	2.4	8	14
Gaminara	25	2.1	7	18
Typhimurium	19	1.6	5	14
Muenster	16	1.3	5	6
Cubana	10	0.8	3	6

(iii) Comparison to the PulseNet database. PFGE pattern analysis of 1,190 Salmonella isolates revealed that 549 isolates (46.1%) obtained from surface water had indistinguishable PFGE pattern matches with human isolates reported to the PulseNet database from all 50 states in the United States since 1996 (www.cdc.gov/pulsenet/index.html), while the remaining 53.9% had unique PFGE patterns (Table 5, Table S1). Some serotypes consisted of a low number of isolates with matching PFGE patterns, such as S. Give, S. Muenchen, S. Rubislaw, and S. Anatum, which consisted largely of strains that are genetically distinct from human strains, with only 3.5% (2/57) 12.6% (34/270), 18.3% (28/153), and 22.2% (8/36) of the total isolates, respectively, having PFGE patterns indistinguishable from those of human isolates. On the other hand, some serotypes included a high number of isolates with matching PFGE patterns. For example, S. Oranienburg, S. Hartford, S. Montevideo, and S. Newport had 100% (32/32), 87.2% (75/86), 79.6% (90/113), and 77.8% (77/99), respectively, of their total isolates matching the PFGE patterns of isolates of human origin. They consisted largely of strains that had previously caused human infections. Some serotypes, namely, S. Aqua, S. Brandenburg, S. Brazil, S. Gaminara, S. Kiambu, S. Luciana, S. Orion, S. Ouakam, S. Paratyphi_B_var._L-tartrate+, and S. Worthington, along with S. enterica subspecies houtenae and one untypeable isolate, did not have any matching PFGE patterns in the CDC PulseNet database.

TABLE 5.

Waterborne Salmonella isolates with PFGE patterns indistinguishable from those of human isolates in the CDC PulseNet database

Serotype	Total no. of isolates	No. of isolates with PFGE pattern match (%)	PulseNet PFGE pattern names (no. of isolates)
Agona	1	1 (100)	JABX01.0025 (1)
Anatum	36	8 (22.2)	JAGX01.0001 (6), JAGX01.0491 (2)
Aqua	7	0
Baildon	2	2 (100)	TDEX01.0001 (2)
Bareilly	30	18 (60)	JAPX01.0113 (2), JAPX01.0181 (2), JAPX01.0229 (3), JAPX01.0279 (6), JAPX01.0464 (2), JAPX01.0631 (3)
Berta	3	1 (33.3)	JAXX01.0002 (1)
Braenderup	39	31 (79.5)	JBPX01.0002 (19), JBPX01.0039 (4), JBPX01.0101 (3), JBPX01.0126 (2),
			JBPX01.0241 (3)
Brandenburg	1	0
Brazil	1	0
Cerro	2	2 (100)	JCGX01.0001 (2)
Cubana	14	1 (7.1)	JDGX01.0070 (1)
Derby	1	1 (100)	JDPX01.0336 (1)
Enteritidis	1	1 (100)	JEGX01.0021 (1)
Gaminara	25	0
Give	57	2 (3.5)	JEXX01.0699 (2)
Hartford	86	75 (87.2)	JHAX01.0010 (30), JHAX01.0038 (22), JHAX01.0045 (18), JHAX01.0048 (3), JHAX01.0065 (1), JHAX01.0074 (1)
Havana	3	3 (100)	TDLX01.0038 (3)
Infantis	22	15 (68.2)	JFXX01.0023 (4), JFXX01.0081 (3), JFXX01.0314 (2), JFXX01.0560 (3), JFXX01.0731 (3)
Inverness	5	2 (40)	JRLX01.0051 (1), JRLX01.0063 (1)
I 4,[5],12:b:-	53	40 (75.5)	JKXX01.0059 (22), JKXX01.0120 (1), JKXX01.0258 (2), JKXX01.0316 (1), JKXX01.0402 (6), JKXX01.0484 (1), JKXX01.0487 (1), JKXX01.0524 (3), JKXX01.0804 (2), JKXX01.1240 (1)
I 4,[5],12:i:-	3	3 (100)	JPXX01.0621 (1), JPXX01.1212 (1), JPXX01.1393 (1)
Javiana	3	3 (100)	JGGX01.1513 (2), JGGX01.2069 (1)
Kentucky	5	1 (20)	JGPX01.0027 (1)
Kiambu	3	0
Kintambo	1	1 (100)	JRNX01.0004 (1)
Litchfield	4	4 (100)	JGXX01.0004 (4)
Liverpool	2	2 (100)	TDPX01.0004 (1), TDPX01.0005 (1)
Luciana	2	0
Mbandaka	9	9 (100)	TDRX01.0005 (1), TDRX01.0011 (6), TDRX01.0232 (2)
Meleagidis	2	2 (100)	JHXX01.0060 (2)
Mississippi	7	6 (85.7)	JIPX01.0007 (1), JIPX01.0052 (1), JIPX01.0073 (1), JIPX01.0099 (1), JIPX01.0483 (2)
Montevideo	113	90 (79.6)	JIXX01.0011 (2), JIXX01.0013 (2), JIXX01.0027 (52), JIXX01.0063 (2), JIXX01.0076 (2), JIXX01.0077 (1), JIXX01.0080 (1), JIXX01.0081 (3), JIXX01.0084 (2), JIXX01.0086 (2), JIXX01.0090 (8), JIXX01.0182 (2), JIXX01.0253 (1), JIXX01.0525 (2), JIXX01.0542 (1), JIXX01.0599 (2), JIXX01.0878 (3), JIXX01.0995 (1), JIXX01.1080 (1)
Muenchen	270	34 (12.6)	JJ6X01.0301 (2), JJ6X01.0319 (5), JJ6X01.0483 (6), JJ6X01.0498 (4), JJ6X01.0712 (1), JJ6X01.0955 (3), JJ6X01.1427 (1), JJ6X01.2068 (2), JJ6X01.2220 (2), JJ6X01.2597 (2), JJ6X01.2850 (1), JJ6X01.2951 (1), JJ6X01.3588 (3), JJ6X01.3871 (1)
Muenster	17	13 (76.5)	TDSX01.0037 (3), TDSX01.0142 (1), TDSX01.0188 (6), TDSX01.0246 (3)
Newport	99	77 (77.8)	JJPX01.0025 (9), JJPX01.0030 (4), JJPX01.0039 (2), JJPX01.0041 (1), JJPX01.0061 (2), JJPX01.0062 (5), JJPX01.0094 (13), JJPX01.0267 (2), JJPX01.0273 (2), JJPX01.0422 (1), JJPX01.0438 (1), JJPX01.0507 (12), JJPX01.0534 (6), JJPX01.0611 (5), JJPX01.0869 (7), JJPX01.3320 (1), JJPX01.6171 (4)
Oranienburg	32	32 (100)	JJXX01.0020 (27), JJXX01.0028 (1), JJXX01.0117 (4)
Orion	4	0
Ouakam	1	0
Paratyphi_B_var._L-tartrate+	1	0
Rubislaw	153	28 (18.3)	JLPX01.0020 (2), JLPX01.0030 (1), JLPX01.0062 (2), JLPX01.0108 (3), JLPX01.0125 (3), JLPX01.0139 (3), JLPX01.0184 (3), JLPX01.0243 (2), JLPX01.0244 (6), JLPX01.0457 (2), JLPX01.0558 (1)
Saintpaul	4	4 (100)	JN6X01.0143 (4)
Schwarzengrund	9	5 (55.6)	JM6X01.0221 (1), JM6X01.0226 (2), JM6X01.0385 (2)
Senftenberg	4	4 (100)	JMPX01.0028 (2), JMPX01.0090 (2)
Soerenga	5	5 (100)	SRNX01.0003 (5)
Tennessee	2	2 (100)	JNXX01.0002 (2)
Thompson	4	4 (100)	JP6X01.0001 (3), JP6X01.0216 (1)
Typhimurium	19	16 (84.2)	JPXX01.0023 (1), JPXX01.0038 (2), JPXX01.0324 (2), JPXX01.0359 (3), JPXX01.0586 (4), JPXX01.0681 (1), JPXX01.1212 (1), JPXX01.1494 (2)
Worthington	1	0
Subsp. arizonae	19	1 (5.3)	ZATX01.0004 (1)
Subsp. houtenae	2	0
Untypeable	1	0
Total	1,190	549 (46.1%)

(iv) Antimicrobial resistance of Salmonella. While most (95.6%; 1,138/1,190) of the Salmonella isolates were susceptible to the 14 antimicrobial drugs tested, 4.4% (52/1,190) of the isolates were resistant to at least one of the drugs, out of which 43 isolates exhibited multidrug-resistant (MDR) phenotypes (Table 6). A total of 28 isolates, one S. Newport isolate and 27 S. Oranienburg isolates, were resistant to 10 antimicrobials, including the third-generation cephalosporin drugs ceftiofur and ceftriaxone. All 27 S. Oranienburg isolates were recovered from the same sampling event (2017 fall) from 10 different sampling sites (Fig. 1). Resistance to neither azithromycin nor ciprofloxacin was detected in this study.

TABLE 6.

Antimicrobial-resistant Salmonella isolated from surface water

AR profilesa	No. of resistances	No. of isolates	Serotype(s) (no. of isolates)
Pan-susceptible	0	1,138
Nal	1	3	Muenster (3)
Sul	1	1	Montevideo (1)
Tet	1	5	Muenster (3), Muenchen (1), Gaminara (1)
StrTet	2	1	Kentucky (1)
SulTet	2	2	Typhimurium (2)
StrSulTet	3	6	Muenster (6)
SulTetTri	3	5	Montevideo (5)
AmpChlSulTetTri	5	1	Derby (1)
AmoAmpFoxTioAxoChlGenStrSulTet	10	27	Oranienburg (27)
AmoAmpFoxTioAxoChlStrSulTetTri	10	1	Newport (1)

^aAmo, amoxicillin/clavulanic acid; Amp, ampicillin; Fox, cefoxitin; Tio, ceftiofur; Axo, ceftriaxone; Chl, chloramphenicol; Gen, gentamicin; Nal, nalidixic acid; Str, streptomycin; Sul, sulfisoxazole; Tet, tetracycline; Tri, trimethoprim/sulfamethoxazole.

FIG 1.

Map of water sampling sites in the Upper Oconee watershed, Georgia, USA. Sampling sites where antimicrobial-resistant (AR) Salmonella were isolated are symbolized as blue circles, sampling sites where multidrug-resistant (MDR) S. Oranienburg were isolated are symbolized as red circles, a sampling site where S. Thompson with the PFGE pattern JP6X01.0001 was isolated is symbolized as a green circle, and all other sampling sites are represented as orange circles. The North Oconee water reclamation facility (WRF) is shown as a triangle. The inset map shows the Upper Oconee watershed in blue, counties within which the water sampling sites are located in yellow, and metro Atlanta counties in orange.

Antimicrobial-resistant Salmonella.

(i) Antimicrobial resistance genes. Among the 52 resistant Salmonella isolates, 13 different antimicrobial resistance genes were detected (Fig. 2). Three tetracycline resistance genes (tetA, tetB, tetC) were detected in 95.8% (46/48) of the tetracycline-resistant isolates, with tetA being the most frequently detected tet gene (41/48; 85.4%), followed by tetC (3/48; 6.25%) and tetB (2/48; 4.2%). Two sulfisoxazole resistance genes, sul1 and sul2, were detected in 16.3% (7/43) and 83.7% (36/43) of the sulfisoxazole-resistant isolates, respectively, with an isolate positive for both the resistance genes. Detection of both strA and strB (35/35; 100.0%) was observed in all of the streptomycin-resistant isolates. Only 6.9% (2/29) of the chloramphenicol-resistant isolates contained floR. Approximately 97% (28/29) of the ampicillin-resistant isolates were positive for bla_TEM-1, while all of the 28 isolates that were resistant to third-generation cephalosporins contained bla_CMY-2. Among the seven trimethoprim/sulfamethoxazole-resistant isolates, dhfr1 (5/7), dhfr12 (2/7), and dhfr13 (1/7) were detected, with one isolate positive for both dhfr12 and dhfr13. The remaining 10 resistance genes tested in this study, aac(3)-Iva, aacC2, aadA1, aadA2, bla_CTX-M, cat1, cat2, dhfr5, tetG, and tetM, were not detected. None of the three Salmonella isolates that displayed resistance to nalidixic acid contained mutations in the quinolone resistance-determining regions (QRDRs) of gyrA and parC.

FIG 2.

PFGE analysis of the antimicrobial-resistant Salmonella isolates recovered from surface water. Black represents resistance to the antimicrobial drugs or the presence of plasmid replicon types or resistance genes. Gray represents susceptibility to the antimicrobial drugs or the absence of plasmid replicon types or resistance genes.

(ii) Phenotypic and genotypic detection of ESBL. Based on the phenotypic resistance to ceftriaxone, 28 isolates (one S. Newport and 27 S. Oranienburg) were selected to be tested for extended-spectrum β-lactamase (ESBL) production. The result of the phenotypic assay of ESBL production was negative for the presence of ESBL producers. Upon amplification and sequencing of β-lactamase genes, these third-generation cephalosporin-resistant isolates were positive for bla_CMY-2 and bla_TEM-1, except for the S. Newport isolate, which was positive for bla_CMY-2 only.

(iii) PFGE analysis. Among the 52 resistant Salmonella isolates, 13 unique PFGE patterns were identified with XbaI. PFGE revealed that there were isolates with indistinguishable PFGE patterns, many of which also had other identical characteristics, including antimicrobial resistance patterns and plasmid replicon types, indicating that they may be clones (Fig. 2). Some of these isolates were recovered from the same locations during the same seasonal collections, e.g., 514 TB and 514 TX, while some of them were recovered from multiple locations during the same or different seasons, e.g., 55 GB, 55 TB, 62 GB, 62 TB, 139 TB, and 139 TX.

(iv) Mobile genetic elements. Approximately 71% of the resistant Salmonella isolates (37/52) were positive for plasmid replicons; all of the isolates contained one replicon each, except for one isolate with three replicons (Fig. 2). Out of the 30 replicons tested by PCR, 6 types (A/C, FIB, FII, HI2, N, X1) were identified; A/C was the most common, carried by 81% (n = 30) of the resistant Salmonella isolates with plasmid replicons. N was the second most common replicon (n = 5), while FIB, FII, HI2, and X1 were represented only once. Class I integrons (intI) were detected in 67.4% (34/52) of the isolates. All of the isolates negative for plasmid replicons were also negative for class I integrons, while 3 out of the 37 isolates positive for plasmid replicon were negative for class I integrons.

Whole-genome sequence analysis of MDR representative Salmonella.

When whole genomes of the four selected isolates were sequenced and analyzed, it was found that all of their antimicrobial resistance genes were associated with a specific replicon type. The isolate 78 TX (S. Newport) contained an A/C plasmid with aadA2, bla_CMY-2, dfrA12, floR, strA, strB, sul1, sul2, tetA, and tetR, with aadA2, dfrA12, and sul1 on an integron within the A/C plasmid. In the isolate 256 GB (S. Montevideo), an N plasmid was associated with dfrA15, sul1, tetA, and tetR. In 561 TX (S. Derby), an HI2 plasmid contained aadA2, bla_TEM-1, dfrA12, floR sul1, and tetB, while 647 GB (S. Oranienburg) contained an A/C plasmid with aadB, aph3-Ia, bla_CMY-2, bla_TEM-1, cmlA5, strA, strB, sul2, tetA, and tetR. The genome statistics of the four Salmonella isolates are presented in Table 7.

TABLE 7.

Genome assembly statistics of the Salmonella isolates from surface water

Isolate ID	Serotype	Season recovered	Antimicrobial resistance phenotypes	Genome
				Length (bp)	No. of contigs	N50 (bp)a	GC (%)	Median coverage (×)	GenBank accession no.
78TX	Newport	2015 spring	AmoAmpFoxTioAxoChlStrSulTetTri	4,982,902	40	699,801	52.1	51	JAJQJE000000000
256GB	Montevideo	2016 winter	SulTetTri	4,763,406	65	511,247	52.1	59	JAJQJD000000000
561TX	Derby	2017 spring	AmpChlSulTetTri	5,149,436	63	314,449	51.7	46	JAJQJC000000000
647GB	Oranienburg	2017 fall	AmoAmpFoxTioAxoChlGenStrSulTet	4,748,896	44	473,596	52.1	47	JAJQJB000000000

^aN₅₀ length is defined as the length for which the collection of all scaffolds of that length or longer contains at least half of the total of the lengths of the scaffolds.

DISCUSSION

Salmonella isolates were recovered from 70.2% of total samples from surface water of the Upper Oconee watershed. In previous studies, a wide range of Salmonella recovery rates from surface water has been reported throughout the United States and Canada (as low as below 10% to as high as above 90%), while similar rates (70% to 80%) have been reported by the studies conducted in the southeastern United States (10, 19, 27–31). These different Salmonella recovery rates could be due to geographical differences, such as climatic factors, agricultural practices, and nutrient availability, or due to different methods used for the isolation of Salmonella, including sample volumes. Environmental parameters, such as precipitation and water temperature, were shown to be correlated with the prevalence and concentration of Salmonella as well as the diversity of serotypes in the water environment (7, 27). In the current study, the presence of Salmonella in water was positively correlated with water temperature, and while there was a weak positive correlation with rainfall, this was not statistically significant. The results from the previous and current studies suggest that environmental factors affect the diversity and survival of Salmonella in the environment to a certain extent.

Salmonella isolated from surface waters of the Upper Oconee watershed included a high diversity of serotypes, including those frequently found in humans. A total of 48 S. enterica subspecies enterica serotypes were identified, besides a single untypeable isolate and 21 isolates that belonged to S. enterica subspecies arizonae and houtenae. This number is substantially higher than those noted in other studies, which have identified 11 to 33 serotypes (19, 22, 27, 28, 32). The current study encountered a larger number of isolates than most of the previous studies, and this difference in the sample size may have led to the difference in serotype diversity. Differences in Salmonella serotype diversity could also be governed by geographical differences. Haley et al. reported S. enterica subsp. arizonae, which is often associated with reptiles, to be the predominant species (40.6%) in their study area with a high density of wetland, while Baudart et al. reported S. Typhimurium, one of the most common serotypes found in humans, to be the predominant species (33.1%) in a Mediterranean coastal watershed (27, 33). The most prevalent serotypes recovered in the mixed-use Upper Oconee watershed were S. Muenchen, S. Rubislaw, S. Montevideo, S. Newport, and S. Hartford, which is consistent with the observations seen in other studies (9, 19, 27, 28, 32). These serotypes were detected in almost every water collection, and their continuous presence and persistence in surface water show that certain serotypes survive and thrive better in aquatic environments than the others or are continuously shed into these surface waters. On the other hand, 19 serotypes were identified only once throughout the study, suggesting their transient nature in the water environment or the rarity of their introduction into the waters.

The findings of this study suggest that Salmonella diversity in the aquatic environment may be higher than previously appreciated. This is likely due to the use of a modified USDA method for Salmonella isolation from animal samples, which is based on the Standard Method 9260 B (34, 35). This method helps to recover multiple Salmonella serotypes from a single sample by using several enrichment and plating media; up to four Salmonella isolates can be isolated when one colony is picked from each positive plate. This method ensures the recovery of diverse Salmonella serotypes with various characteristics, as different selective media tend to favor selective growth of certain serotypes. A recent finding using a new sequencing tool, CRISPR-SeroSeq, supported this high diversity of Salmonella in freshwater environments by recovering up to 10 different serotypes in a single water sample. This method described by Deaven et al. (36) found that 80% of the sites positive for Salmonella harbored more than one serotype, while our method demonstrated that 33.5% of the positive sites contained more than one serotype. Overall, these results suggest that water may serve as a reservoir for diverse Salmonella serotypes, even more diverse than those that are associated with human and animal infections, the majority of which are caused by a few serotypes.

PFGE typing suggests that isolates present in surface water include large numbers of isolates that are indistinguishable from human isolates. The PFGE patterns of all Salmonella isolates were compared to the PFGE patterns of human isolates in the CDC PulseNet database, and it was found that approximately half of the isolates were indistinguishable from CDC XbaI patterns. This shows that Salmonella strains present in the aquatic environment were genetically associated with those from human sources. The prevalence of Salmonella that are often linked to human cases of salmonellosis is of epidemiological significance, as this might suggest that Salmonella found in the environment could have originated from humans or that humans could have potentially acquired Salmonella infections from the environment.

To further investigate the possibility of human-environment transmission of Salmonella, we compared the PFGE patterns of our environmental isolates with those of Salmonella isolates recovered from patients within the state of Georgia. Although we have considered the survival period of Salmonella in surface water to be 30 days, it should be noted that Salmonella can potentially persist longer due to biofilm formation and the presence of invertebrates, such as free-living protozoa, and sediments, which can enhance the survival of Salmonella in aquatic environments (37). There were several incidences where the same Salmonella strains with the same PFGE patterns were simultaneously recovered from both surface water and humans in the surrounding area. A particularly strong example is S. Thompson. An S. Thompson isolate with PFGE pattern JP6X01.0001 was recovered from surface water in April 2016 (Fig. 1). Of the 60 S. Thompson human clinical isolates reported in the state of GA between 2001 and 2018 with the same PFGE pattern, 47 cases occurred during the study period of 2015 to 2017, out of which 36 cases occurred between September and November of 2016. Furthermore, while S. Thompson is not one of the most common serotypes that cause human infections, and only a median of eight S. Thompson cases occur annually in Georgia, two S. Thompson outbreaks occurred in a restaurant in Georgia in 2016 (38). The first outbreak, which occurred in April 2016, was associated with JP6X01.0002, and the second outbreak, which occurred in October 2016, was associated with JP6X01.0001, but the two patterns were very similar and differ by only one band and are highly related by whole-genome sequencing (WGS). While epidemiological investigations are needed to track and identify the initial sources of the infections, the fact that most of the human infections associated with S. Thompson with the JP6X01.0001 pattern, including an outbreak, occurred at about the same time an indistinguishable strain of Salmonella was present in the environment suggests a potential epidemiologic association between the aquatic environment and human infections. However, whether this indicates shedding of Salmonella into surface water through wastewater or human infections due to exposure to contaminated surface water or consumption of produce irrigated with contaminated water needs more and differently designed investigations.

The remaining half of the isolates from surface water had unique PFGE patterns that had not been documented in human isolates in the United States, representing strains genetically distinct from those of human origin. Salmonella isolates of certain serotypes recovered from surface water did not have any matching PFGE patterns in the PulseNet database, and while this may be because only a few isolates were recovered for these serotypes, this was not the case for Gaminara. A total of 25 S. Gaminara isolates, which were recovered from 7 sampling seasons, had 18 different PFGE patterns, but none of them were indistinguishable from XbaI PFGE patterns in the PulseNet database. Gaminara is one of the serotypes that are found in humans and animals, and this result shows that Gaminara isolates found in the environment are different strains from the ones that cause infections in humans, suggesting nonhuman sources of these Salmonella isolates found in the environment.

The majority of Salmonella isolates were pan-susceptible to all of the 14 antimicrobial drugs tested, with a low antimicrobial resistance rate of 4.4%. This rate is much lower than that of many of the previous studies that conducted susceptibility testing on waterborne Salmonella (20–22, 28). Resistance to third-generation cephalosporins that are used to treat bacterial infections, including invasive salmonellosis in children (39), was detected in some isolates. However, no resistance to ciprofloxacin and azithromycin was detected, which are antimicrobial drugs that are commonly used for the treatment of severe Salmonella infections (40). Most of the serotypes that were identified among the antimicrobial-resistant isolates (S. Derby, S. Kentucky, S. Montevideo, S. Muenchen, S. Muenster, S. Newport, S. Oranienburg, and S. Typhimurium) are those that are frequently reported in humans and animals. Interestingly, no resistance was detected among the most common environmental serotypes, including S. Rubislaw, S. Give, and S. Hartford. This indicates that these antimicrobial-resistant human- and animal-associated serotypes may have been exposed to a wide range of antimicrobials prior to their contamination of surface water.

One isolate was identified as S. Newport MDR-AmpC, which is an S. Newport strain with resistance to nine or more antimicrobials, including extended-spectrum cephalosporins, and that is responsible for infections in both humans and animals on an epidemic scale (41). The MDR S. Newport from this study carried bla_CMY-2 on an A/C plasmid, a large plasmid that is usually associated with MDR (42). This was somewhat expected, as bla_CMY genes, which encode cephamycinases (CMY), ampC plasmid-mediated β-lactamases, are responsible for the majority of resistance to extended-spectrum cephalosporins among Salmonella in the United States and are commonly carried on plasmids that belong to the incompatibility groups A/C and I1 (39, 43–47). Although the XbaI PFGE pattern of the S. Newport isolate matched the outbreak pattern of JJPX01.0422, human cases associated with the same strain of Salmonella were infrequent in Georgia, and only one case was reported during the collection period in 2015, 6 months after the water sample was collected. Since S. Newport is known to be largely associated with cattle and other food animals, many of which harbor the MDR-AmpC strains, the S. Newport isolate could be from an animal source.

Additionally, a total of 27 MDR S. Oranienburg isolates, which were resistant to 10 different antimicrobials, were isolated during the fall 2017 sampling. They are very likely to be clones with identical phenotypic and genotypic characteristics, including PFGE and antimicrobial resistance gene profiles. These S. Oranienburg isolates all harbored an A/C plasmid and belonged to the PFGE pattern JJXX01.0020. Habing et al. recovered two Salmonella isolates with the same PFGE pattern from Michigan dairy farms in 2000, but interestingly, these isolates were pan-susceptible to all the antimicrobial drugs tested (48). S. Oranienburg may have acquired an A/C plasmid with antimicrobial resistance genes, including bla_CMY-2, later in time, or our clone may have been from a different ancestral source. The MDR S. Oranienburg isolates were recovered from 10 different sites, most of which were located along McNutt Creek, starting with a site potentially influenced by runoff from chicken houses (Fig. 1). Even though their exact association with the S. Oranienburg isolates is not clear, chicken houses could be a potential contamination source, as S. Oranienburg is often isolated from chicken (49). There were only three reported human cases associated with S. Oranienburg with the particular PFGE pattern in Georgia during the study period, indicating a lower frequency of the human clinical cases associated with the given Salmonella strain. Three other distinct drainages were identified that had this contamination, including one just downstream of a wastewater treatment plant and two on unconnected creeks, both of which are in low-density population areas (Fig. 1).

Conclusions.

Salmonella populations in surface water were highly prevalent and diverse, underlining the role of environmental water as a source of Salmonella with the potential to disseminate to humans. The detection of Salmonella of human clinical serotypes and of PFGE patterns indistinguishable from those of human isolates supports possible transmission between the aquatic environment and humans, thus playing a role in the epidemiology of Salmonella infections. Furthermore, the presence of antimicrobial-resistant Salmonella strains, especially those with MDR phenotypes, shows a potential public health safety risk for those exposed to surface water. Surface water is not only used for recreational purposes but also applied to irrigate fresh produce that is often consumed uncooked. Thus, the microbial quality of the water needs to be ensured for the protection of the human populations that are exposed to the water or consume the fresh produce. Therefore, persistent contamination of surface water with Salmonella that has potential to infect humans is a risk factor that needs to be monitored continuously.

MATERIALS AND METHODS

Isolation and identification of Salmonella.

(i) Collection of water samples. Water samples were collected from rivers and streams in the Upper Oconee watershed (USGS cataloging unit 03070101) located in the Appalachian Piedmont region of northeastern Georgia, USA (Fig. 1). Sampling sites (n = 105) were located along the Middle Oconee River (MIDO), North Oconee River (NORO), and their tributaries (also indicated with a prefix of MIDO or NORO). Sampling sites were selected to represent a range of land uses/land covers, including residential, agricultural, industrial, recreational, and forested. The 1-L water samples were collected seasonally for 3 years from 2015 to 2017 (12 seasons) with the assistance of Upper Oconee Watershed Network (UOWN) volunteers. The number of water samples collected each season ranged from 30 to 100, dependent on the available volunteers and access to the sampling sites. Samples were stored at 4°C and processed within 24 h. Water temperature (°C) was recorded in situ using water-resistant thermometers (VWR, Radnor, PA, USA). Daily precipitation data were obtained from the U.S. climate data website (www.usclimatedata.com).

(ii) Isolation and identification of Salmonella. As previously described, water samples were mixed with 0.5 g of cellulose filter powder (Aqua Dew; Aqua Dew Water Purification Products, Lahore, Pakistan) and filtered onto 47-mm glass fiber filters (Pall Corporation, Ann Arbor, MI, USA), which were preloaded with another 0.5 g of cellulose filter powder suspended in 15 mL of sterile water (50). The filter, along with the filter powder, was incubated in 25 mL of 1× buffered peptone water (BD Difco, Franklin Lakes, NJ, USA) for nonselective preenrichment of samples. All overnight incubations were carried out at 37°C for 18 to 20 h. All media used for the isolation of Salmonella were purchased from BD Difco. For Salmonella isolation, 1 mL of each peptone broth enrichment was transferred to Gram-negative (GN) Hajna and tetrathionate (Tet) broths for selective enrichment and then processed as previously described (34). Briefly, after the secondary selective enrichment using Rappaport-Vassiliadis (RV) broth and then selective isolation using brilliant green sulfa (BGS) and xylose lysine Tergitol 4 (XLT4) agar plates, one colony with the typical appearance of Salmonella was picked from each plate positive for Salmonella and subjected to biochemical assays: triple sugar iron (TSI) and lysine iron agar (LIA) slants. Salmonella suspects, positive by the biochemical assay, were serogrouped using serogroup-specific typing sera and serotyped using Salmonella multiplex assay for rapid typing (SMART) PCR (51). Salmonella isolates that could not be serotyped using the SMART PCR method were sent to the National Veterinary Services Laboratories (NVSL; Ames, IA, USA) for serotyping.

(iii) Antimicrobial susceptibility testing. The susceptibility of all Salmonella isolates was determined by broth-microdilution using the Sensititre semiautomated antimicrobial susceptibility system (TREK Diagnostic Systems, Inc., Cleveland, OH, USA) and the Sensititre custom National Antimicrobial Resistance Monitoring System (NARMS) plate CMV3AGNF according to the manufacturer’s directions. MICs of the isolates for the 14 antimicrobials tested were determined, and using the breakpoints set by the Clinical and Laboratory Standards Institute (CLSI) (52), each isolate was classified as resistant, intermediate, or susceptible to the antimicrobials tested. Where no CLSI interpretative criteria were available, NARMS guidelines were used (53). For the analysis, antimicrobials identified as intermediate were considered susceptible. For azithromycin, which lacks CLSI-approved breakpoints, the epidemiological cutoff value for wild-type Salmonella was used (54). The 14 antimicrobials and the breakpoints (μg/mL), for determining resistances are as follows: amoxicillin/clavulanic acid (≥32/16), ampicillin (≥32), azithromycin (>16), cefoxitin (≥32), ceftiofur (≥8), ceftriaxone (≥4), chloramphenicol (≥32), ciprofloxacin (≥1), gentamicin (≥16), nalidixic acid (≥32), streptomycin (≥32), sulfisoxazole (≥512), tetracycline (≥16), and trimethoprim/sulfamethoxazole (≥4/76). Escherichia coli ATCC 25922, Pseudomonas aeruginosa ATCC 27853, Enterococcus faecalis ATCC 29212, and Staphylococcus aureus ATCC 29213 were used as control strains for MIC determination.

(iv) PFGE. All Salmonella isolates were subjected to pulsed-field gel electrophoresis (PFGE) using the previously described method (55). In brief, plugs were prepared by embedding bacterial genomic DNA in 1.0% SeaKem gold agarose (BioWhittaker Molecular Applications, Rockland, ME, USA) and digested with XbaI restriction enzyme (Roche Molecular Biochemicals, Indianapolis, IN, USA). Digested DNA was separated by electrophoresis in 0.5× Tris-borate-EDTA (TBE) buffer at 6 V for 19 h with pulse times of 2.16 to 54.17 s using the CHEF-DRII PFGE system (Bio-Rad, Hercules, CA, USA). S. Braenderup H9812 was used as a standard strain. A dendrogram for cluster analysis was generated in BioNumerics (Applied Maths, Inc., Austin, TX, USA) using the Dice coefficient and the unweighted pair group method of arithmetic averages (UPGMA) with 1.5% optimization and 1.5% tolerance (56). PFGE patterns of the Salmonella isolates were compared to the PFGE patterns in the PulseNet database, a national molecular subtyping network run by the CDC to surveil foodborne diseases and identify outbreak sources (www.cdc.gov/pulsenet/index.html). PFGE patterns of the Salmonella isolates were also compared with those of human isolates recovered within the state of Georgia, specifically from the counties where the watershed is located (Barrow, Clarke, Jackson, and Oconee counties) and the surrounding metro-Atlanta counties (Cherokee, Clayton, Cobb, Dekalb, Douglas, Fayette, Forsyth, Fulton, Gwinnett, Henry, and Rockdale counties) (Fig. 1). Only those human isolates recovered 1 month before and after each water sampling event were included for the analysis.

Antimicrobial-resistant Salmonella.

(i) Antimicrobial resistance gene PCR and sequence analysis. Antimicrobial-resistant Salmonella isolates were screened for the presence of 23 genes associated with the phenotypic resistance exhibited. Resistant isolates were tested for genes encoding resistance to β-lactams (bla_CMY, bla_CTX-M, bla_TEM), tetracycline (tetA, tetB, tetC, tetG, tetM), trimethoprim/sulfamethoxazole (dhfr1, dhfr5, dhfr12, dhfr13), sulfisoxazole (sul1, sul2), chloramphenicol (cat1, cat2, floR), and aminoglycosides [aacC2, aac(3)-Iva, aadA1, aadA2, strA, strB]. Nalidixic acid-resistant isolates were screened for mutations in the quinolone resistance-determining regions (QRDR) of gyrA and parC. PCR assays were performed using previously described primers and thermal cycling conditions (57). Amplified PCR products were then analyzed by electrophoresis on a 2% agarose gel and stained with ethidium bromide for visualization.

For the screening of mutations in the QRDR, the PCR products of gyrA and parC were purified using a QIAquick PCR purification kit (Qiagen, Germantown, MD, USA) according to the manufacturer’s directions and then used as templates in the following sequencing reactions. The sequencing PCR mixture was prepared with 10 μL of water, 8 μL of BigDye ready reaction mix (Applied Biosystems, Foster City, CA, USA), 1 μL of 3.2 pmol forward or reverse primer, and 1 μL of the purified PCR product. Sequencing was performed using an ABI Prism 3130x genetic analyzer (Applied Biosystems), and the sequences were analyzed against the GenBank database using the Basic Local Alignment Search Tool (BLAST), available at the National Center for Biotechnology Information website (www.ncbi.nlm.nih.gov).

(ii) Phenotypic and genotypic detection of ESBL. Isolates resistant to ceftriaxone (MIC ≥ 2 μg/mL) were considered potential producers of extended-spectrum β-lactamase (ESBL) and selected for further testing. These isolates were assayed phenotypically using cefotaxime/clavulanic acid and ceftazidime/clavulanic acid and classified as ESBL producers if there was at least a 3- to 2-fold concentration decrease in the MIC for ceftazidime or cefotaxime in combination with clavulanic acid compared to the MIC when tested without clavulanic acid (52). The potential ESBL producers were also genotypically characterized by the amplification of β-lactamase genes (bla_CMY, bla_CTX-M, bla_TEM), followed by sequencing of the PCR products using the ABI Prism 3130x genetic analyzer as described above. The gene sequences were analyzed using BLAST as described above.

(iii) Replicon typing and integron analysis. The presence of 30 plasmid replicons (A/C, BO, FIA, FIB, FIB-KN, FIB-KQ, FIB-M, FII, FIIK, FIIS, HI1, HI2, HIB-M, I1-α, I1γ, I2, K, L, M, N, N2, P1, R, T, U, W, X1, X2, X3, and X4) was determined using the PCR-based replicon typing (PBRT version 2.0) kit (Diatheva, Fano, Italy) according to the manufacturer’s directions.

Antimicrobial-resistant Salmonella isolates were screened for the presence of class 1 integrons by the amplification of the conserved segment of the integrase gene, intI, using previously described primers and thermal cycling conditions (57, 58). Amplified PCR products were then analyzed by electrophoresis on a 2% agarose gel and stained with ethidium bromide for visualization.

(iv) WGS of select isolates. Four selected Salmonella isolates (78 TX, 256 GB, 561 TX, and 647 GB) representing multidrug resistance (MDR; resistance to two or more classes of antimicrobials) phenotypes were subjected to whole-genome sequencing (WGS) analysis. Genomic DNA of overnight grown cultures was extracted using the GenElute bacterial genomic DNA kit (Sigma-Aldrich, St. Louis, MO, USA). The DNA libraries were prepared using the Nextera XT DNA library preparation kit (Illumina, Inc., San Diego, CA, USA) according to the manufacturer’s instructions. The concentration and fragment size distribution of the DNA libraries were determined on a Qubit fluorometer, using the double-stranded DNA (dsDNA) high-sensitivity (HS) assay kit (Life Technologies, Inc., Carlsbad, CA, USA) and Bioanalyzer 2100 using an Agilent high-sensitivity DNA kit (Agilent Technologies, Santa Clara, CA, USA), respectively. Paired-end sequencing (2 × 250 bp) of the DNA libraries was run on a MiSeq sequencer using the 500-cycle MiSeq reagent kit version 2 (Illumina, Inc.). The generated reads were assembled de novo using the A5-miseq assembler (59), and the assembled sequence was annotated using the NCBI Prokaryotic Genome Annotation Pipeline (60). Plasmid replicon-associated genes were identified using PlasmidFinder from the Center for Genomic Epidemiology (CGE) server (accessed on 21 March 2018) to determine the locations of the antimicrobial resistance genes (61). Locations of antimicrobial resistance genes on contigs not containing a replicon were determined by homology using BLAST.

(v) Statistical methods.

Data were analyzed using either Microsoft Excel or GraphPad Prism. Seasonal comparisons of Salmonella recovery rates and serotype diversity were carried out using the Kruskal-Wallis test. Correlations between Salmonella recovery rate and environmental parameters, i.e., the amount of precipitation (cm) and water temperature (°C), were made using Pearson linear correlation. P values of <0.05 were considered significant.

Data availability.

This whole-genome shotgun project has been deposited at NCBI GenBank under BioProject accession number PRJNA787236, and the accession numbers are as follows: JAJQJB000000000 (S. Oranienburg 647 GB), JAJQJC000000000 (S. Derby 561 TX), JAJQJD000000000 (S. Montevideo 256 GB), and JAJQJE000000000 (S. Newport 78 TX).