Characterization of Escherichia coli and Other Enterobacterales Resistant to Extended-Spectrum Cephalosporins Isolated from Dairy Manure in Ontario, Canada

ABSTRACT

Extended-spectrum cephalosporins (ESCs) resistance genes, such as bla_CTX-M, bla_CMY, and bla_SHV, have been found regularly in bacteria from livestock. However, information on their distribution in dairy cattle in Canada and on the associated genome sequences of ESC-resistant Enterobacterales is sparse. In this study, the diversity and distribution of ESC-resistant Escherichia coli throughout manure treatments in six farms in Southern Ontario were assessed over a one-year period, and their ESC-resistance plasmids were characterized. The manure samples were enriched using selective media. The resulting isolates were screened via polymerase chain reaction for bla_CTX-M, bla_CMY, and bla_SHV. No E. coli carrying bla_SHV were detected. Escherichia coli (n = 248) carrying bla_CTX-M or bla_CMY underwent whole-genome sequencing using an Illumina MiSeq/NextSeq. These isolates were typed using multilocus sequence typing (MLST) and their resistance gene profiles. A subset of E. coli (n = 28) were sequenced using Oxford Nanopore Technologies. Plasmids were assembled using Unicycler and characterized via the resistance genes pattern, replicon type, plasmid MLST, phylogenetic analysis, and Mauve alignments. The recovery of ESC-resistant Enterobacterales (18 species, 8 genera) was drastically reduced in manure outputs. However, multiple treatment stages were needed to attain a significant reduction. 62 sequence types were identified, with ST10, ST46, ST58, ST155, ST190, ST398, ST685, and ST8761 being detected throughout the treatment pipeline. These STs overlapped with those found on multiple farms. The ESC-resistance determinants included CTX-M-1, -14, -15, -17, -24, -32, -55, and CMY-2. The plasmids carrying bla_CTX-M were more diverse than were the plasmids carrying bla_CMY. Known “epidemic plasmids” were detected for both bla_CTX-M and bla_CMY.

IMPORTANCE The increase in antimicrobial resistance is of concern for human and animal health, especially when resistance is conferred to extended-spectrum cephalosporins, which are used to treat serious infections in both human and veterinary medicine. Bacteria carrying extended-spectrum cephalosporin resistance genes, including bla_CTX-M and bla_CMY, are frequently found in dairy manure. Manure treatment influences the loads and diversity of bacteria, including those carrying antimicrobial resistance genes, such as Enterobacterales and Escherichia coli. Any bacteria that survive the treatment process are subsequently applied to the environment. Enterobacterales carrying bla_CTX-M or bla_CMY can contaminate soil and crops consumed by humans and animals, thereby increasing the potential for antimicrobial resistance genes to integrate into the human gut microflora through horizontal gene transfer. This furthers the dissemination of resistance. Therefore, it is imperative to understand the effects manure treatments have on ESC-resistance in environmentally applied manure.

INTRODUCTION

Digestate, the end product of dairy manure anaerobic digestion, including any residual antimicrobial-resistant bacteria, is commonly used as a valued fertilizer for crops. This application could potentiate the risk of antimicrobial resistance (AMR) transmission to humans and animals through water runoff and through the consumption of contaminated crops (1, 2). Manure is treated to optimize the nutrient composition, reduce net volume, reduce greenhouse gas emissions, and provide recycled bedding for livestock. This treatment modifies its microflora and incidentally reduces the prevalence and concentration of antimicrobial-resistant bacteria and AMR genes. For instance, composting is an effective method by which to reduce the prevalence of Escherichia coli and Proteobacteria in general (3, 4). Composting at thermophilic temperatures can reduce the absolute abundance of AMR genes, such as sul2 and tet, if done at a consistently high temperature (4). Anaerobic digestion at thermophilic temperatures reduces the amount of sulfonamide-resistant and tetracycline-resistant bacteria in manure (5) as well as the prevalence of extended-spectrum β-lactamase (ESBL)-producing Enterobacterales (6).

Critically important antimicrobials, such as extended-spectrum cephalosporins (ESCs), are widely used in human and veterinary medicine, including the treatment of severe and life-threatening bacterial infections. Resistance to ESCs could result in the failed treatment of infections and in increased mortality. The production of extended-spectrum β-lactamases (ESBLs) is the main resistance mechanism to ESCs, and the most common enzymes among them include the CTX-Ms, SHVs, and TEMs (7–10). Common variants of CTX-M detected in bacteria from dairy cattle include CTX-M-1, -2, -9, -14, -15, and -27 (11, 12). Few studies have described SHV variants in Enterobacterales from dairy cattle in North America (12). Although TEMs have not been frequently investigated in dairy cattle, the most commonly identified variant seems to be the non-ESBL TEM-1. In addition to ESBLs, cephamycinases are frequently responsible for resistance to ESCs, specifically, the CMY enzymes. In North America, the most common CMY variant in Enterobacterales from dairy cattle is CMY-2 (12–16), which is consistent with other geographical regions, such as Europe (17–19) and Japan (20, 21). The few genome-based studies of ESC-resistant Enterobacterales from cattle in Canada show that bla_CTX-M-15 is commonly found in Enterobacterales from dairy cattle (15). This variant and bla_CTX-M- 1, 14, 27, 32, 55, and 65 have been found in Canadian beef cattle (22, 23). In the Enterobacterales from Canadian beef cattle, bla_CMY is more frequent than is bla_CTX-M (22), but the situation is not clear in dairy cattle.

The drop in numbers of resistant bacteria and resistance genes from raw to anaerobic digestion at both mesophilic and thermophilic temperatures in dairy manure has been assessed in vitro (6, 24), but few studies have been performed under field conditions in situ. In general, there is a lack of genomic research on ESC-resistant Enterobacterales and their ESC-resistance plasmids from dairy cattle and their manure in Canada, whereas research on these topics in beef cattle is more frequently reported. To understand how various manure treatments affect ESC-resistant Enterobacterales in dairy manure, manure samples from a variety of treatment stages were used for this study. Our main objectives were twofold. The first was to investigate the diversity and distribution of ESC-resistant E. coli strains through manure treatment processes among six dairy farms in Southern Ontario, using phenotypic analysis, resistance gene identification, and whole-genome sequencing. The second was to assess the diversity of plasmids carrying ESC-resistance determinants and their distribution throughout manure treatment processes, using hybrid genome assemblies and assessments of plasmid genetic relatedness. Due to their recent emergence in farm animals in Canada and their frequent location on multidrug resistant (MDR) plasmids, in addition to their clinical importance in humans, a stronger emphasis was put on ESC-resistant strains and plasmids carrying bla_CTX-M.

RESULTS

Diversity and persistence of ESC-resistant Enterobacterales through manure treatment processes.

(i) Recovery of ESC-resistant bacteria at different stages of manure processing. Although some variation between farms occurred, the proportion of samples with Enterobacterales carrying one of bla_CMY, bla_CTX-M, or bla_SHV followed a downward trend through the treatment pipelines (Table 1; Table S1). This overall reduction of positive samples for ESC-resistant Enterobacterales from raw input manure to the end product of manure treatment was significant for both types of end products (i.e., recycled bedding [OR, 1.0 × 10⁻⁴; 95% CI, 1.53 × 10⁻⁶ to 0.0106; P < 0.001] and environmentally applied manure [OR, 3.0 × 10⁻³; 95% CI, 7.0 × 10⁻⁵ to 0.1258; P = 0.002]) (Table 2). This was not the case for the initial, anaerobically-digested manure step (Table 2). The decrease in samples positive for ESC-resistant E. coli was also significant in the recycled bedding (OR, 1.62 × 10⁻⁵; 95% CI, 8.86 × 10⁻⁹ to 0.0686; P = 0.01) (Table 2). However, this was again not the case for the samples containing the ESC-resistant E. coli from the initial anaerobic digestion, and, this time, also for the samples from the final environmentally-applied manure (Table 2).

TABLE 1.

Proportion of samples with ESC-resistant Enterobacterales and E. coli isolates across the main successive manure processing steps

Farm	Processing stepa	ESC-resistant enterobacteralesb	ESC-resistant E. colib
1	Raw	100 (8/8)	87.5 (7/8)
	Treated	75.5 (6/8)	62.5 (5/8)
	Environmental	50.0 (2/4)	50.0 (2/4)
	Recycled bedding	57.1 (4/7)	42.9 (3/7)
2	Raw	83.3 (5/6)	83.3 (5/6)
	Treated	100 (6/6)	100 (6/6)
3	Raw	100 (8/8)	100 (8/8)
	Treated	100 (8/8)	100 (8/8)
	Environmental	100 (8/8)	100 (8/8)
	Recycled bedding	0 (0/7)	0 (0/7)
4	Raw	100 (8/8)	100 (8/8)
	Treated	87.5 (7/8)	87.5 (7/8)
5	Raw	100 (8/8)	100 (8/8)
	Treated	87.5 (7/8)	87.5 (7/8)
	Environmental	50.0 (4/8)	50.0 (4/8)
	Recycled bedding	43.0 (3/7)	28.6 (2/7)
7	Raw	100 (8/8)	100 (8/8)
	Treated	100 (8/8)	100 (8/8)
	Environmental	100 (8/8)	100 (8/8)
	Recycled bedding	37.5 (3/8)	37.5 (3/8)
All	Raw	97.8 (45/46)	95.7 (44/46)
	Treated	91.3 (42/46)	89.1 (41/46)
	Environmental	78.6 (22/28)	78.6 (22/28)
	Recycled bedding	34.5 (10/29)	27.6 (8/29)

^aRaw, raw manure input; Treated, manure treated with anaerobic digestion; Environmental, liquid end product of manure treatment to be applied to the environment; Recycled bedding, manure end product dewatered for use as bedding (farms one, three, and five; heat-treated compost on farm seven). For details regarding manure processing and sampling on individual farms, refer to Materials and Methods or Fig. 6.

^bThe numerical value within each cell represents the percentage, with the proportion in parentheses.

TABLE 2.

Multilevel^a logistic regression models examining associations between the recovery of ESC-resistant Enterobacterales, E. coli, and the sampling stage

Sampling stage/independent variableb	Recovery of interest/dependent Variablec	Odds ratio	95% CI	P value
Anaerobically digested manure	ESC-resistant Enterobacterales	0.25	0.0553 to 1.1742	0.079
	ESC-resistant E. coli	0.49	0.1181 to 2.0732	0.336
	E. coli CTX-M	1.22	0.4532 to 3.2704	0.696
	E. coli CMY	0.63	0.1915 to 2.0659	0.445
Environmentally applied manure	ESC-resistant Enterobacterales	3.0 × 10−3	7.0 × 10−5 to 0.1258	0.002
	ESC-resistant E. coli	0.17	1.44 × 10−2 to 1.9571	0.154
	E. coli CTX-M	1.55	0.5383 to 4.4577	0.417
	E. coli CMY	0.23	0.0384 to 1.3464	0.103
Recycled bedding	ESC-resistant Enterobacterales	1.0 × 10−4	1.53 × 10−6 to 0.0106	<0.001
	ESC-resistant E. coli	1.62 × 10−5	8.81 × 10−9 to 0.0686	0.010
	E. coli CTX-M	4.65 × 10−2	1.01 × 10−2 to 0.2139	<0.001
	E. coli CMY	6.96 × 10−3	3.49 × 10−4 to 0.1387	0.001

^aOutcomes were measured at the replicate-level. Models included random intercepts for farm, farm date, and sample. Results in bold are considered to be statistically significant.

^bRefer to the footnote of Table 1. Raw manure is the referent category for these analyses.

^cESC-resistant Enterobacterales carrying bla_CTX-M, bla_CMY, or bla_SHV as well as ESC-resistant E. coli carrying bla_CTX-M or bla_CMY.

(ii) Bacterial species and antimicrobial resistance profiles. The main species of interest for this study was E. coli; however, seven other genera of Enterobacterales were recovered (Fig. 1). Escherichia coli comprised 69.3% of the 707 isolates carrying bla_CTX-M, bla_CMY, or bla_SHV. More E. coli isolates carried bla_CMY (n = 261) than bla_CTX-M (n = 222), and seven isolates carried both genes. The next most frequent species were Klebsiella pneumoniae (11.8%) and Proteus mirabilis (10.3%) (Fig. 1). 79% of the ESC-resistant E. coli isolates were MDR (Table 3). Aside from the β-lactams, the antimicrobials to which E. coli were most frequently resistant included tetracycline (71.6%), sulfonamides (68.6%), sulfamethoxazole-trimethoprim (58.0%), and streptomycin (57.1%) (Table 3). The same trend was seen among the non-E. coli Enterobacterales isolates in general, with resistance to tetracycline being the most common (87.6%). This was followed by resistance to sulfonamides (72.8%), sulfamethoxazole-trimethoprim (70.5%), and streptomycin (65.0%) (Table S2). In E. coli, resistance to gentamicin and ciprofloxacin were the least frequent among E. coli, at 8.6% and 3.4%, respectively, as well as among all other ESC-resistant Enterobacterales, with a 4.1% resistance to gentamicin and no resistance to ciprofloxacin (Table 3; Table S2).

FIG 1.

Distribution of Enterobacterales (n = 707) positive for bla_CTX-M, bla_CMY, or bla_SHV recovered at various manure treatment stages from six farms in Southern Ontario as well as their ESC-resistance genes. The n values represent the numbers of isolates from each genus or species that were recovered from all samples, across all farms. For the genera with multiple species: Escherichia spp. include E. coli (n = 493) and E. fergusonii (n = 1); Proteus spp. include P. mirabilis (n = 73), P. hauseri (n = 17), P. vulgaris (n = 9), and P. penneri (n = 4); Citrobacter spp. include C. sedlakii (n = 6), C. braakii (n = 4), C. freundii (n = 1), C. gillenii (n = 1), C. koseri (n = 1), and C. youngae (n = 1); Providencia spp. include P. rettgeri (n = 4) and P. stuartii (n = 1).

TABLE 3.

Frequencies of resistance to 12 antimicrobials and of multidrug resistance among E. coli isolates carrying bla_CTX-M or bla_CMY

Farm (number of isolates)	Percentage of Escherichia coli isolates with a resistant phenotypea												MDRa (%)
	CHL	CIP	SUL	SXT	KAN	GEN	STR	TET	AMP	AMC	CTXb	FOX
1 (n = 68)	33.8	7.3	66.1	55.8	29.4	17.6	58.8	58.8	100	47.0	100	45.5	72.1
2 (n = 45)	53.3	6.5	73.3	60.0	28.9	15.2	55.5	86.7	100	65.2	100	65.2	93.3
3 (n = 106)	37.7	1.8	71.6	65.0	22.6	17.9	68.8	80.1	100	54.7	98.1	50.9	85.0
4 (n = 53)	45.2	0	67.9	54.7	9.4	0	56.6	71.6	100	83.0	100	83.0	88.7
5 (n = 103)	17.4	4.8	46.6	34.9	5.8	2.9	37.8	46.6	100	84.4	100	80.5	55.3
7 (n = 115)	34.8	1.7	85.2	73.9	15.3	0.8	62.3	87.8	100	18.2	99.1	18.2	88.7
Overall (n = 490)	34.4	3.4	68.6	58.0	17.6	8.6	57.1	71.6	100	55.5	99.4	53.7	79.0

^aMulti-drug resistant (MDR) is defined as being resistant to at least one antimicrobial in three or more antimicrobial classes. Isolates with an intermediate phenotype were not considered to be resistant. Resistance to AMC was not taken into consideration for multidrug resistance, as isolates carrying bla_CMY are resistant to AMC and would have been considered to be MDR by default. AMP, ampicillin; AMC, amoxicillin-clavulanic acid; CTX, cefotaxime; FOX, cefoxitin; SUL, sulfonamide; SXT, sulfamethoxazole-trimethoprim; TET, tetracycline; STR, streptomycin; KAN, kanamycin; GEN, gentamicin; CIP, ciprofloxacin; and CHL, chloramphenicol.

^bThe total number of isolates with a resistant phenotype was not 100% due to three isolates having an intermediate susceptibility profile.

A comparison of antimicrobial resistance phenotypes found in bla_CTX-M-positive and bla_CMY-positive isolates is reported in Table 4. The odds of bla_CTX-M-positive isolates showing a resistant phenotype for CIP (OR, 42.46; 95% CI 7.36 to +∞; P < 0.0001), SXT (OR, 4.83; 95% CI, 3.19 to 7.41; P < 0.0001), SUL (OR, 3.66; 95% CI, 2.35 to 5.77; P < 0.0001), KAN (OR, 2.15; 95% CI, 1.35 to 3.45; P = 0.0009), TET (OR, 1.93; 95% CI 1.25 to 3.02; P = 0.0024), and STR (OR, 1.83; 95% CI, 1.24 to 2,73; P = 0.0019) were significantly higher than those for isolates carrying bla_CMY. This means, for instance, that the odds of a bla_CTX-M-positive isolate being resistant to ciprofloxacin are approximately 42 times higher than those for a bla_CMY-positive isolate. In contrast, the odds of a bla_CTX-M-positive isolate being resistant to AMC (OR, 0.0024; 95% CI, 0.00 to 0.01; P < 0.0001) and FOX (OR, 0.0007; 95% CI, 0.00 to 0.00; P < 0.0001) were significantly lower than those for a bla_CMY-positive isolate (Table 4). The distribution of the number of antimicrobial classes to which E. coli isolates were resistant was unimodal for those carrying bla_CTX-M, with a mode of five (Fig. 2). Isolates resistant to all seven classes of antimicrobials tested were all E. coli carrying bla_CTX-M. In contrast, E. coli carrying bla_CMY showed a bimodal distribution with modes of two (cephems and ampicillin) and six (Fig. 2).

TABLE 4.

Associations between nonsusceptibility to 12 antimicrobials in E. coli carrying bla_CTX-M in reference to E. coli isolates carrying bla_CMY, using exact logistic regression models^a

	CTX-M versus CMY
Antimicrobial	Odds ratio	95% CI	P value
AMC	0.0024	0.0000 to 0.0142	<0.0001
FOX	0.0007	0.0002 to 0.0024	<0.0001
CHL	0.7932	0.5357 to 1.1711	0.2609
CIP	42.4697 b	7.3647 to +∞	<0.0001
GEN	0.9323	0.4708 to 1.8196	0.9547
KAN	2.1497	1.3503 to 3.4494	0.0009
STR	1.8375	1.2422 to 2.7323	0.0019
SXT	4.8334	3.1919 to 7.4097	<0.0001
SUL	3.6613	2.3543 to 5.7878	<0.0001
TET	1.9341	1.2506 to 3.0223	0.0024

^aFor the abbreviations, refer to the footnote of Table 3. AMP and CTX were not included in the analysis due to their complete resistance for each gene.

^bMedian unbiased estimate. The bolded results are considered to be statistically significant at a P value of <0.05. Isolates with an intermediate phenotype were classified as resistant for these analyses.

FIG 2.

Antimicrobial resistance distribution of E. coli carrying bla_CMY or bla_CTX-M from all sampled farms. The percentages of resistant E. coli were calculated based on the total for each group: E. coli carrying bla_CMY (n = 261) and E. coli carrying bla_CTX-M (n = 222). The number of antimicrobial classes with a resistant phenotype does not include zero, as those isolates would have been excluded during growth in selective media. Additionally, the scale does not include one class, as all of the isolates were resistant to cephems (cefotaxime and cefoxitin) and other β-lactams (ampicillin). Nonsusceptible profiles include the intermediate and resistant ranges.

(iii) ESC-resistance gene variants and other resistance genes. 7 CTX-M variants were identified among the 227 E. coli carrying bla_CTX-M: CTX-M-1 (n = 25), CTX-M-14 (n = 28), CTX-M-15 (n = 125), CTX-M-17 (n = 1), CTX-M-24 (n = 11), CTX-M-32 (n = 4), and CTX-M-55 (n = 34). One isolate carried both bla_CTX-M-14 and bla_CTX-M-32. The CMY-2 variant was the only one identified among the 28 sequenced E. coli carrying a bla_CMY gene. 8 bla_TEM variants were identified among the 138 E. coli isolates sequenced with this gene. These were TEM-1 (n = 124), TEM-34 (n = 4), TEM-104 (n = 3), TEM-135 (n = 1), TEM-198 (n = 1), TEM-207 (n = 1), TEM-216 (n = 2), and TEM-230 (n = 2). The only ESBL among them was TEM-207, which was present in an isolate that was also carrying bla_CTX-M-15. No bla_SHV genes were detected in the E. coli. 11 ESC-resistant isolates tested negative for bla_CTX-M, bla_CMY, and bla_SHV. These were tested for bla_TEM and identified at the species level. Six of them were E. coli, of which four had mutations in the chromosomal ampC promoter region, predicting a stronger promoter (n = 4), and two carried a bla_TEM-1 gene. Additional resistance genes that were otherwise not screened for use in antimicrobial susceptibility testing were identified via genome sequencing (Table S3). The main ones included resistance genes for macrolides: mdf(A) (n = 244), mph(A) (n = 52), mph(E) (n = 5), msr(E) (n = 5), and erm(B) (n = 1). Genes associated with resistance to fosfomycin, rifampin, and lincosamides were also found, but less frequently, and included fosA3 (n = 5), arr-2 (n = 17), and lnu(F) (n = 18), respectively.

(iv) E. coli strain diversity. 62 E. coli STs were identified, of which 12 were novel (Table S4). The most frequent STs included ST58 (n = 30), ST46 (n = 22), ST398 (n = 16), ST10 (n = 14), ST190 (n = 13), ST744 (n = 12), ST685 (n = 12), and ST155 (n = 12). No apparent consistent decrease or increase in the number of STs was visible along the successive manure processing steps (Table S4). Some STs carrying bla_CTX-M were found in more than one farm (Table S4), including ST10 (n = 3), ST38 (n = 2), ST46 (n = 2), ST58 (n = 4), ST69 (n = 2), ST88 (n = 2), ST155 (n = 4), ST162 (n = 2), ST165 (n = 2), ST398 (n = 2), ST540 (n = 2), ST685 (n = 2), ST744 (n = 4), ST746 (n = 2), ST1722 (n = 2), and ST8761 (n = 2). STs that were found in some farms in raw, anaerobically-digested, and environmentally-applied manure overlapped substantially with those found on several farms and included ST10, ST46, ST58, ST155, ST190, ST398, ST685, and ST8761 (Table S4).

To further assess strain relatedness, cgMLST minimum spanning trees were created (Fig. 3). These trees demonstrated that most bla_CTX-M variants, including bla_CTX-M-1, bla_CTX-M-14, bla_CTX-M-15, bla_CTX-M-55, and bla_CMY-2, are not restricted to single clonal lineages but have spread horizontally and reside in a variety of unrelated STs (Fig. 3). Groups of multiple tightly related isolates were found at different processing stages in farms one, three, and seven (annotated with red arrows in Fig. 3), and this suggests the spread of these clonal lineages through several manure processing steps. Two other clusters of tightly related isolates (annotated with black arrows in Fig. 3), including ST155 and ST744, were found in more than one farm. The relatedness of the ST744 isolates from four different farms is further supported by the presence of the bla_CTX-M-24 gene in all of them and in none of the other isolates in this study. Core trees using an SNP analysis with WGS short reads were created for the isolates from farms three and seven (Fig. S1 and S2). Clusters identified in cgMLST (Fig. 3) were generally conserved in the SNP-based trees, and isolates which were not clustered in cgMLST remained so in the individual SNP trees (Fig. S1 and S2).

FIG 3.

Comparison using the core genomes of the ESC-resistant E. coli isolates recovered from dairy farms in Southern Ontario. Minimum spanning trees using a cgMLST comparison of 2,513 core genes among ESC-resistant E. coli carrying bla_CTX-M and/or bla_CMY. Color codes represent the (A) farm source, (B) manure treatment source, (C) ESC-resistance gene variant, and (D) MLST profile. Red arrows denote clusters of isolates from the same farm source, whereas black arrows denote clusters of isolates from different farm sources.

Diversity of ESC-resistance plasmids.

A subsample of E. coli isolates from farms three and seven were chosen for long read sequencing and hybrid assembly. 24 isolates carrying bla_CTX-M and 9 isolates with bla_CMY were sequenced (Bioproject PRJNA833106 and GenBank accession numbers SAMN27959348 through SAMN27959371). Of these, 32 ESC-resistance genes were carried on plasmids, and in two instances, bla_CTX-M-15 and bla_CTX-M-55 were found on the chromosome, with the latter isolate also carrying an additional bla_CTX-M-55 on a plasmid (Table S3). Most of the plasmids were fully assembled using Unicycler (n = 29). However, two plasmids could not be assembled using Unicycler and were instead assembled using Flye and polished with Pilon (isolates 260-1c and 331-3a). One plasmid could not be assembled using either approach (isolate 329-1a), and it was excluded from the downstream SNP analysis (Table S3). Additionally, the ESC-resistance plasmid in isolate 260-1c was also excluded from downstream SNP analysis due to discrepancies between the Unicycler and Flye assemblies.

The 30 successfully and reliably assembled plasmids were first compared and grouped using a presence and absence of genes. They showed a wide diversity, but the clusters correlated with a predominant incompatibility group in most instances (Fig. 4). The four bla_CTX-M-1 plasmids were of two different Inc groups, and they were found in isolates of different STs. The two bla_CTX-M-1 IncI1/ST3 plasmids were found independently in the two farms (Table 5). The bla_CTX-M-14 gene was found in farm three only and was located on IncI1 plasmids in all five investigated isolates. Four of these plasmids were ST166 and were found in three different ST strains in two types of samples on two different dates (Table 5). Plasmids carrying bla_CTX-M-15 were of various IncF replicon types and of one IncY replicon type, and each was found in isolates of different STs (Table 5). The plasmids carrying bla_CTX-M-55 were the most diverse, and each was located in a different strain (Table 5). As mentioned above, one of the isolates (279-2a) carried two separate bla_CTX-M-55 copies, one of which was on the chromosome and the other of which was on an IncHI2/IncN plasmid. This isolate had an intermediate susceptibility for cefoxitin, despite its lack of the bla_CMY gene, with no mutations in the chromosomal ampC promoter region and no additional β-lactamase genes. The plasmids of three incompatibility groups carried bla_CMY-2, with the main ones being IncC and IncI1 (Table 5). The IncC plasmids were all of the pMLST profile ST3, but each was found in a different E. coli ST. They all carried the resistance genes floR, tet(A), aph(6)-Id, aph(3″)-Ib, sul1, sul2, dfrA12, and aadA2 in addition to bla_CMY-2 (Table 5). The IncI1 plasmids did not carry any resistance genes beyond bla_CMY-2. Three of these are IncI1/ST20, and they are structurally similar (Fig. 5). Two of them were recovered from different farms but were located in related ST162 isolates.

FIG 4.

Comparison of plasmids of diverse replicon types carrying bla_CTX-M or bla_CMY among E. coli from dairy cattle manure. Similarity tree, based on the presence and absence of 1,357 genes among ESC-resistance plasmids (n = 30) from the E. coli isolates (n = 28) that were recovered from manure samples across two dairy farms in Southern Ontario. Each plasmid name includes the isolate ID, the ESC-resistance gene and variant, and the isolate MLST type. IncI1 (red nodes), IncC (orange), Inc-/p0111 (yellow), IncN (light green), IncY (dark green), IncFII (light blue), IncFIB/IncFIC (medium blue), IncFIB/IncFIB(K) (dark blue), IncFIB (purple), IncHI2/IncN (pink), and IncHI2 (magenta). Isolate 329-1a is excluded from this figure due to its inability to close the plasmid.

TABLE 5.

Characteristics of ESC-resistant Escherichia coli genomes and their ESC-resistance plasmids isolated from dairy manure at various treatment stages on six farms in Southern Ontario^a

Gene	ID	F	Sample	SM	Plasmid				Isolate
					Replicon type	pMLST	Plasmid size (bp)	Additional resistance genes on ESC-plasmid	MLST
CTX-M-1	280-3a	7	DWS	Apr	IncN	ST1	42, 484	None	ST190
	280-2c	7	DWS	Apr	IncN	ST1	42, 481	None	ST398
	302-1b	7	DWS	Jun	IncI1	ST3	105, 449	sul2, tet(A)	ST3580
	254-3a	3	DWS	Mar	IncI1	ST3	104, 156	sul2	ST162
CTX-M-14	200-2a	3	Raw	Jan	IncI1	ST166	95, 660	None	ST58
	202-3b	3	DWOS	Jan	IncI1	ST166	93, 630	None	ST58
	200-3b	3	Raw	Jan	IncI1	ST166	91, 896	None	ST4429
	270-3a	3	DWOS	Apr	IncI1	ST166	90, 813	None	ST398
	162-3b	3	DWOS	Nov	IncI1	Novel	90, 590	None	ST13743
CTX-M-15	329-1a	3	Raw	Aug	IncI1/FIB/FIC	ST31/F46:A-:B20	208, 008	aadA17, aph(3′)-1a, blaTEM-1, cmlA1, dfrA12, floR, sul2, sul3, tet(A), tet(M)	ST8761
	238-2b	7	DEW	Feb	IncFIB	Unknown	111, 929	None	ST683
	331-3a	3	DWOS	Aug	IncFIB	Unknown	110, 943	None	ST155
	190-2a	7	DEW	Dec	IncFIB/FIB(K)	F-:A-:B53	104, 690	blaTEM-1, dfrA14, qnrS1, tet(A)	ST162
	343-3b	7	DEW	Aug	IncY	NA	98, 488	aph(3″)-Ib, aph(6)-Id, blaTEM-1, dfrA14, qnrS1, sul2, tet(A)	ST540
CTX-M-55	260-3b	7	DEW	Mar	IncY	NA	130, 626	aadA5, aph(3″)-Ib, aph(6)-Id, dfrA17, floR, mdf(A), sul2, tet(A)	ST11633
	342-3a	7	DWOS	Aug	IncFIB/FIC	F18:A-:B58	129, 958	aadA1, ant(3”)-1a, aph(3′)-Ia, floR, lnu(F), sul3	ST744
	254-2c	3	DWS	Mar	IncFIB/FIC	F18:A-:B58	112, 048	aph(3′)-Ia, tet(A)	ST7978
	259-1a	7	DWOS	Mar	IncFII	F33:A-:B-	70, 625	blaTEM, fosA3	ST155
	233-3a	3	DWOS	Feb	IncHI2	ST2	270, 799	aadA22, aac(3)-IId, aph(3′)-Ia, ARR-2, blaTEM-1, cmlA1, dfrA14, floR, mph(A), qnrS1, sul3, tet(A)	ST8761
	344-2a	7	HTC	Aug	IncHI2/IncN	ST2b	267, 447	aadA22 aac(3)-IId, aph(3′)-Ia, ARR-2, blaTEM-1, dfrA14, floR, lnu(F), mph(A), qnrS1, sul3, tet(A)	ST10
	279-2a	7	Raw	Apr	IncHI2/IncN	ST2b	261, 342	aadA23, ant(3”)-1a, aph(3′)-Ia, ARR-2, dfrA14, floR, lnu(F), mph(A), qnrS1, sul3, tet(A)	ST219
	181-3a	3	Raw	Dec	-/p0111	NA	127, 570	aadA5, aph(3″)-Ib, aph(6)-Id, dfrA17, floR, sul2, tet(A)	ST224
CMY-2	292-1b	3	DWOS	Jun	IncC	ST3	164, 259	aadA2, aph(3″)-Ib, aph(6)-Id, dfrA12, floR, sul1, sul2, tet(A)	ST657
	162-3b	3	DWOS	Nov	IncC	ST3	160, 272	aadA2, aph(3″)-Ib, aph(6)-Id, dfrA12, floR, sul1, sul2, tet(A)	ST13743
	280-2c	7	DWS	Apr	IncC	ST3	155, 827	aadA2, aph(3″)-Ib, aph(6)-Id, dfrA12, floR, sul1, sul2, tet(A)	ST398
	259-2c	7	DWOS	Mar	IncC	ST3	116, 140	aadA2, aph(3″)-Ib, aph(6)-Id, dfrA12, floR, sul1, sul2, tet(A)	ST10
	259-1a	7	DWOS	Mar	IncY	NA	162, 914	aph(3″)-Ib, aph(6)-Id, floR, sul2, tet(A)	ST155
	183-2b	3	DWOS	Dec	IncI1	ST12	102, 760	None	ST165
	344-1a	7	HTC	Aug	IncI1	ST20	97, 674	None	ST162
	331-2a	3	DWOS	Aug	IncI1	ST20	97, 660	None	ST162
	290-1a	3	Raw	Jun	IncI1	ST20	95, 120	None	ST609

^aID, isolate identification; F, farm; SM, sampled month; MLST, multilocus sequence type; DWS, digestate with solids; DWOS, digestate without solids; DEW, dewatered; HTC, heat-treated compost. Isolate 329-1a is mentioned in this table, as the plasmid details regarding its Inc type and resistance genes were identified. However, it was excluded from the plasmid sequence comparison in Fig. 4 due to its inability to close the plasmid.

^bThe pMLST for IncHI2 was identified, but the pMLST for IncN was not identified.

FIG 5.

Comparison of IncI1 plasmids carrying bla_CTX-M or bla_CMY in E. coli from dairy cattle manure. (A) Phylogenetic maximum likelihood tree, using SNP comparisons of IncI1 plasmids carrying bla_CTX-M or bla_CMY from E. coli, using a total of 189 genes, including core genes (n = 63). (B) pMLST profile in the middle panel and (C) Mauve alignments of the IncI1 plasmids in the right panel. Plasmid names are comprised of the isolate identification along with the ESC-resistance gene variant.

IncI1 plasmids carrying bla_CTX-M-1, bla_CTX-M-14, or bla_CMY-2 were compared using a core SNP analysis (Fig. 5A) and a Mauve alignment (Fig. 5C). These plasmids clustered in accordance with their pMLST and ESC-resistance genes (Fig. 5A and B). Despite the observed variability in resistance genes, the pMLST types, and the recovery from two farms at different points in time in a variety of samples, the overall structure of the IncI1 plasmids was relatively conserved (Fig. 5).

DISCUSSION

The persistence of ESC-resistant bacteria and plasmids through manure treatment could contribute to the transmission of ESC-resistance within agricultural environments and, consequentially, into the food chain (25, 26). Using sensitive enrichment methods, this study attempted to obtain a better understanding of the diversity and dynamics of ESC-resistant bacteria during the treatment of dairy manure by using samples from six farms of variable sizes in Ontario.

Our results showed a significant and drastic decrease in the frequency of environmentally-applied manure samples containing ESC-resistant Enterobacterales at the end of manure treatment (estimated OR = 3 × 10⁻³), compared to raw manure before processing. Despite the persistence of some ESC-resistance in these end products, this decrease is indicative of the effectiveness of manure treatments at reducing the prevalence of ESC-resistant Enterobacterales. However, the first mesophilic anaerobic digestion step alone was not sufficient to achieve a significant decrease. This strongly supports the need for multiple treatment steps for the effective removal of ESC-resistant Enterobacterales from manure before it is released into the environment. Four of the farms investigated produced recycled bedding, also known as solid composite, in addition to environmentally-applied manure. The reduction in the prevalence of the ESC-resistant Enterobacterales was even more drastic (estimated OR = 1 × 10⁻⁴) in recycled bedding than in environmentally-applied manure. This has beneficial implications for livestock and husbandry purposes by limiting the recirculation of resistant bacteria and associated genes in the animal environment. It also suggests that emphasis should be put on liquid manure as a target to reduce the spread of resistant bacteria in the environment. Contrary to the case of ESC-resistant Enterobacterales, the trend toward a decrease in samples that were positive for ESC-resistant E. coli during manure processing was not significant (estimated OR = 0.17; P = 0.154). This could be an issue of a lack of statistical power, and larger sample sizes may be needed in future studies to assess whether there is a significant effect of treatment on ESC-resistant E. coli populations and to more precisely estimate its magnitude.

The Enterobacterales, other than E. coli, that were recovered from our manure samples included several bacterial species that have been less thoroughly characterized with regard to ESC-resistance in the literature (Fig. 1). Other than E. coli, the two most frequently recovered ESC-resistant species of Enterobacterales were K. pneumoniae and P. mirabilis. These species may play an important role in the epidemiology of ESC-resistance in dairy manure and in the surrounding environment. They certainly warrant further investigation, including comparative studies on ESC-resistance plasmids and how these relate to E. coli plasmids.

The top three resistances, other than those to β-lactams, that were found in the Enterobacterales in general, and in E. coli in particular, were to tetracyclines, sulfonamides, and potentiated sulfonamides. These drugs have been used for a long time in animal husbandry, and they are the most frequently used antimicrobials on dairy farms in Canada, other than β-lactams (27). These findings were consequently expected. Likewise, it is not surprising that ESC-resistant strains are multidrug resistant, as plasmids carrying bla_CTX-M or bla_CMY often carry other AMR genes. This is also illustrated by the subsample of plasmid sequences that we assembled (Table 5). However, the lack of positive associations between bla_CMY and resistance to non β-lactam antimicrobials, compared to bla_CTX-M, suggests that plasmids carrying bla_CMY in E. coli from dairy manure are less prone to be a cause of MDR than are those carrying bla_CTX-M. In this context, the strong association of ciprofloxacin resistance with isolates carrying bla_CTX-M, compared to those carrying bla_CMY, was particularly striking. This observation was further supported by the presence of qnr genes on plasmids carrying bla_CTX-M-15 and bla_CTX-M-55 as well as by the lack thereof in plasmids carrying bla_CMY-2. This is an important observation with regard to both human and veterinary medicine, as quinolones, like ESCs, are critically important antimicrobials (28).

A large diversity of ESC-resistant E. coli strains (62 STs) was identified in manure from the six farms investigated. Despite this diversity, some STs were isolated repeatedly in some farms and at different treatment stages. A variety of scenarios can explain these repeated findings. Seven of the eight CTX-M-positive E. coli STs recovered from raw to environmentally-applied manure were also found in multiple farms. These certainly represent widespread clonal lineages in the environment or in dairy farms, and their recovery at multiple time points and treatment stages is therefore not surprising. More detailed analyses using cgMLST and SNPs demonstrated that in some cases (e.g., ST46 on farm 3) (Fig. 3; Fig. S1), the majority of isolates of one ST found in a farm at different treatment stages were closely related. This suggests a clonal spread within the local farm environment and manure treatment system over a relatively short time period. In other cases (e.g., ST10 in farms 3 and 7 [Fig. 3] and ST398 in farm 7 [Fig. 3; Fig. S2]), the isolates of these STs that were recovered through the entire treatment process were genetically diverse. In the latter scenario, unrelated subclones of these widespread strains were likely to have entered the farm environments on multiple occasions or to have persisted within farms over much more extended periods of time. These two examples represent extreme cases, and a combination of both scenarios is visible, with the majority of STs being recovered repeatedly. Thus, both clonal expansion within the manure treatment system and repeated entry into the system seem to work concomitantly to maintain ESC-resistant E. coli in farm manure treatment systems. These findings strongly support the use of detailed genome analyses, such as cgMLST and core gene SNP analysis, instead of the less informative, conventional MLST in order to understand in detail the dynamics of specific strains or clonal lineages over time and through the stages of manure treatment.

ESC-resistance genes were found on a large variety of plasmids. The bla_CTX-M-1 gene was not detected frequently, but it was located on IncN/ST1 and IncI1/ST3 plasmids. Both of these plasmids represent “epidemic” plasmids (29) that have been recently detected in bacteria from animals in North America (30–32) as well as in humans and animals in Europe (33, 34). Not surprisingly, they were found in genetically unrelated isolates. The bla_CTX-M-14 gene was found repeatedly but in only one of the two farms investigated for plasmid diversity. It was located on the IncI1 plasmids of the same pMLST ST166 in three different strains at different points in time and processing stages. IncI1/ST166 plasmids have been found in Salmonella from human clinical samples and from animal sources in Asia, but these plasmids do not seem to have been described yet in human or in animal sources in North America (35, 36). They are probably not frequent plasmids and are likely to represent examples of plasmids spreading actively via horizontal gene transfer within the manure treatment process. Plasmids carrying bla_CTX-M-15 and bla_CTX-M-55 were more diverse and were found on both farms. This fits the known mobility of these two ESBL genes, with locations on diverse plasmids with various host ranges, conjugation abilities, and advantageous accessory genes (37). Fewer replicon types carrying bla_CMY were identified than for the latter two genes, and in contrast to the plasmids carrying bla_CTX-M, we did not find any IncF plasmids carrying bla_CMY. The IncC plasmids with bla_CMY (classified until recently as IncA/C) all carried the typically associated resistance genes for phenicols, tetracyclines, aminoglycosides, and folate pathway inhibitors (38). As expected (39, 40), the IncI1 plasmids carrying bla_CMY did not carry any additional resistance genes. The mixed population of IncI1 and IncC plasmids that we sequenced, with their respective resistance gene profiles, explains the bimodal distribution that we observed in the number of non-β-lactam resistances in strains carrying bla_CMY. In Canadian beef cattle, plasmids carrying bla_CMY-2 have essentially been IncA/C (41, 42), and our results suggest that the situation may be different in dairy cattle, with a larger proportion of IncI1 plasmids. However, further studies are needed to assess the proportions of IncI1 and IncC plasmids, not only in dairy manure but also in feces from dairy cattle. Finally, IncI1/ST20 plasmids carrying bla_CMY-2 were found in both farms investigated for plasmid diversity (in the same E. coli ST162). This type of plasmid has not been described in animals from Canada, but it has occasionally been found in human clinical samples in Canada and the USA (41, 43). The potential relations between these plasmids in bacterial populations from dairy cattle and humans may warrant further investigation. In the course of this study, we also identified three IncY plasmids carrying bla_CTX-M-15, bla_CTX-M-55, and bla_CMY-2, respectively. These are phage-like plasmids that likely transfer through transduction (44). Other phage-like plasmids carrying bla_CTX-M-15 have been described (45), and IncY plasmids have been identified in ESC-resistant strains (13). However, to our knowledge, IncY plasmids encoding ESC-resistance have not been described in E. coli from dairy cattle. Our results show that these phage-like plasmids carry diverse ESC-resistance genes and may represent an alternate route of ESC-resistance dissemination, besides the well-known conjugative IncI1, IncC, and IncF plasmids. These findings warrant further studies on the role of IncY plasmids in the epidemiology of ESC-resistance.

In conclusion, previous investigations on dairy manure treatments in the same farms that were observed this study showed a decrease in coliform bacteria and E. coli in general, as well as in the overall load of some antimicrobial resistance genes (46). The present study expands upon these results by demonstrating a drastic reduction in the frequency of samples containing ESC-resistant Enterobacterales in the end products of manure treatment. It also demonstrates the need for multiple treatment steps in order to achieve such a significant reduction. However, ESC-resistant Enterobacterales and ESC-resistance plasmids were still recovered in environmentally-applied manure. Thus, risk assessments to address downstream implications for human health are needed. The presence of several ESC-resistant E. coli clones throughout the treatment process was observed. In some cases, this seems to be related to the widespread occurrence of clonal lineages in animals and the environment in general, as well as repeated entry into the manure treatment pipeline. Our data also strongly suggest the persistence of some clones in the manure treatment system (e.g., ST10, ST46, ST58, ST155, ST398, ST685, and ST8761). Further investigations are needed to assess whether these latter clones have adapted to the manure processing environment. Our findings of multiple bacterial species carrying bla_CTX-M and bla_CMY in addition to E. coli warrants further comparative studies on plasmids from other Enterobacterales, such as K. pneumoniae or P. mirabilis, to better understand the ESC-resistance plasmid dynamics and epidemiology. This may be particularly relevant, since the most frequent plasmid (Inc-untypeable, bla_CTX-M-15 plasmid) recovered via conjugation in our sister study (46) was not found among the bla_CTX-M-15 plasmids from E. coli in this study. The presence of related plasmids in multiple clonal lineages and at different stages of the manure treatment environment further supports the role of horizontal transfer in the spread and maintenance of ESC-resistance determinants within manure treatment facilities. Finally, the “epidemic” plasmids IncI1/ST3/bla_CTX-M-1 and IncI1/ST20/bla_CMY-2 were identified in our samples, and the role of dairy manure in their global spread will also need further investigation.

MATERIALS AND METHODS

Sample processing and enrichment.

164 manure samples were obtained from 6 farms located in Southern Ontario from November of 2018 to October of 2019 (Table S5). A seventh farm (farm six) was initially sampled but had to be excluded from the study due to its low turnover rate and lack of anaerobic digestion. Farms one through four had two mesophilic anaerobic digesters. They were sampled before digestion and at the secondary digester during digestion, with the exception of farm on which was sampled in the holding pit immediately following digestion (Fig. 6). On farms one and three, manure from the secondary digester is separated into solids and liquids, which were both sampled (Fig. 6). Farm five utilizes a three-step digestion process involving two mesophilic anaerobic digesters that are followed by a third thermophilic anaerobic digester. Raw manure, digesters one and two, and output from digester three, which was separated into a solids and liquids tank, were sampled (Fig. 6). Farm seven is the most complex and utilizes a primary mesophilic anaerobic digester that is followed by two screw presses and a rotating kiln that performs at thermophilic temperatures to produce compost. Samples were taken before digestion, at the digestate pit, after the first screw press, after the second screw press, and following heat treatment (Fig. 6). Coproducts were added to the primary digestion stages on each farm. Further details regarding the sampling procedures used for each farm were previously described (46). The manure samples were immediately refrigerated and processed within 24 hours. The samples were enriched by adapting a previously described protocol that selects for ESC-resistant Enterobacterales (47). Each sample was tested systematically in independent triplicates by inoculating 10 mL (or a 10 g equivalent) of manure into 90 mL of Escherichia coli (EC) broth (Becton, Dickinson and Company, Franklin Lakes, NJ, USA) that contained 2 mg/L of cefotaxime. Each sample was incubated overnight with constant shaking at 37°C. After incubation, 10 μL were plated onto MacConkey agar (Becton, Dickinson and Company) that contained 1 mg/L ceftriaxone. The samples were incubated overnight at 37°C and examined for growth.

FIG 6.

Flow of manure treatment for each of the six farms in Southern Ontario that were sampled for this study. Farms one through four use mesophilic temperatures (approximately 35°C) for anaerobic digestion, farm five uses thermophilic anaerobic digestion at 51 to 52°C, and farm seven uses 65°C for thermophilic composting in addition to mesophilic anerobic digestion. Raw manure is exposed to the elements and is affected by seasonal weather and temperature. Red arrows direct the flow of manure to a proceeding process. Black solid arrows indicate the destination of the final product being in the barn. Black dashed arrows indicate the final product being applied to land as fertilizer. The colored circles in the top right corners of the various manure treatment stages indicate where the sampling occurred. The n values indicate the herd size on each farm.

Selection of ESC-resistant bacteria.

Three or more colonies were selected per plate, based on lactose fermentation and morphology: (i) lactose fermenter with nonmucoid morphology; (ii) lactose fermenter with mucoid morphology; and (iii) lactose fermenters with a different morphology or nonlactose fermenters with various morphologies were selected until each unique morphology was subcultured. Selected colonies were plated onto MacConkey agar (Becton, Dickinson and Company) (1 mg/L ceftriaxone) and incubated overnight at 37°C. Subcultures were stored at –70°C in brain heart infusion (BHI) broth (Becton, Dickinson and Company) containing 20% glycerol until further use.

Detection of antimicrobial resistance genes and bacterial species identification.

Polymerase chain reaction (PCR) was performed for bla_CTX-M, bla_CMY, and bla_SHV with reagents and thermocycler conditions as previously described (Table 6). The bla_TEM gene was not screened for systematically on all isolates, as the most frequent variant in Enterobacterales from animals in Canada is bla_TEM-1, which does not confer resistance to ESCs. Bacterial lysates were obtained via the boiling method. Bacterial isolates carrying bla_CTX-M, bla_CMY, or bla_SHV were identified using matrix-assisted laser desorption/ionization-time of flight (MALDI-TOF) (Bruker Daltonik GmbH, Bremen, Germany) at the Animal Health Laboratory, University of Guelph, Guelph, ON. A subsample of ESC-resistant bacteria (n = 42) that tested negative for bla_CTX-M, bla_CMY, or bla_SHV and originated from all six farms and covered all of the manure processing stages and various time periods, were identified using MALDI-TOF, and they were tested for bla_TEM using the primers described in Table 6. Among the latter, E. coli isolates that remained negative for bla_TEM were further investigated via the Sanger sequencing of the PCR products of their chromosomal ampC promoter region, using the primers in Table 6 and previously described thermocycler conditions (48).

TABLE 6.

Primers used for the detection of β-lactamase genes

Target	Primer sequence	Product size (bp)	Reference
bla CTX-M	F: ATGTGCAGYACCAGTAARGTKATGGC R: TGGGTRAARTARGTSACCAGAAYCAGCGG	593	62
bla CMY	F: GACAGCCTCTTTCTCCACA R: TGGACACGAAGGCTACGTA	1,000	63
bla SHV	F: AGGATTGACTGCCTTTTTG R: ATTTGCTGATTTCGCTCG	393	63
bla TEM	F: TTCTTGAAGACGAAAGGGC R: ACGCTCAGTGGAACGAAAAC	1,150	64
ampC	F: AATGGGTTTTCTACGGTCTG R: GGGCAGCAAATGTGGAGCAA	191	48

Antimicrobial susceptibility testing.

Enterobacterales carrying bla_CTX-M, bla_CMY, or bla_SHV underwent antimicrobial susceptibility testing via the disk diffusion method, according to the Clinical and Laboratory Standards Institute guidelines (49, 50). The following antimicrobials were used for susceptibility testing: ampicillin (AMP), amoxicillin-clavulanic acid (AMC), cefotaxime (CTX), cefoxitin (FOX), compound sulfonamides (SUL), sulfamethoxazole-trimethoprim (SXT), tetracycline (TET), streptomycin (STR), kanamycin (KAN), gentamicin (GEN), ciprofloxacin (CIP), and chloramphenicol (CHL). The inhibition zone diameter data from each isolate were used to create phenetic trees in BioNumerics v7.6, using the Pearson correlation coefficient and the unweighted pair group method with the arithmetic mean. Clusters were determined from these phenetic trees, using a 95% similarity cutoff to select a representative subset of E. coli carrying bla_CMY for sequencing.

Illumina and MinION sequencing.

Since the CTX-M diversity was our main interest, E. coli carrying bla_CTX-M from all six farms were sequenced (n = 227). A subset of E. coli carrying bla_CMY from farms three and seven were selected for sequencing using stratified random sampling, with one or two isolates being selected per phenetic cluster (n = 28). Seven of these isolates carried both bla_CMY and bla_CTX-M. DNA extractions were completed using an Epicentre MasterPure DNA Purification Kit (Epicentre, Madison, WI, USA), following the manufacturer’s instructions. The DNA sequencing was performed on 188 isolates using an Illumina NextSeq system (Nextera XT library prep, PE150; Illumina, San Diego, CA, USA) at the National Microbiology Laboratory in Winnipeg, MB. The DNA sequencing was performed on the remaining isolates (n = 60) using an Illumina MiSeq (PE300) system at the Advanced Analysis Centre, University of Guelph, Guelph, ON, CA.

A subset of E. coli carrying bla_CTX-M (n = 24) and bla_CMY (n = 9) were sequenced using a MinION MIN-101B device (Oxford Nanopore Technologies, Oxford, UK). These isolates were selected based on the Illumina MiSeq data in addition to the phenetic trees so as to cover the highest diversity of ESC-resistance gene variants over as many manure treatment stages and time periods as possible. These isolates were from farms three and seven only, as these two farms had the highest recoveries of E. coli carrying bla_CTX-M and a large diversity of phenetic clades from which to select, and these encompassed all of the manure treatment stages and time periods being assessed. The DNA extractions were the same as those described above. The sequencing libraries and barcoding preparation were done using the SQK-LSK109 and EXP-NBD104/114 ligation and native barcoding kits (Oxford Nanopore Technologies), according to the manufacturer's instructions. Flow cells (version FLO-MIN106 R9.4) were used and run for 72 h each or until exhausted. The basecalling of the fast5 files and the demultiplexing were performed using Guppy Basecaller v3.3 (Oxford Nanopore Technologies) with barcode trimming enabled.

Genomic assembly and analysis.

Whole-genome sequences using short reads only were assembled using BioNumerics v7.6 (Applied Maths, Saint-Martens-Latem, Belgium), using the SPAdes de novo assembler and assembly-free and assembly-based allele calling. Resistance genes, plasmids, and virulence genes were identified using the E. coli functional genotype scheme in BioNumerics at a 90% similarity threshold, utilizing the PlasmidFinder, ResFinder, and VirulenceFinder databases from the Center for Genomic Epidemiology, Technical University of Denmark, DTU. The multilocus sequence types (MLSTs) were assigned using the cgMLST application for BioNumerics and the BioNumerics E. coli/Shigella Enterobase scheme with Achtmann Sequence Types (https://www.applied-maths.com/applications/wgmlst). Minimum spanning trees (MSTs) were generated using the wgMLST (core enterobase) function for cgMLST trees in BioNumerics with 10× bootstrapping and no multithreading, using 2,513 core loci.

Long and short reads were assembled using hybrid assembler Unicycler v0.4.8 (51) in parallel with Flye v2.6 (52) and polished with short reads using Racon v1.4.0 (53) and Pilon v1.23 (54). The assemblies were visualized using Bandage v0.8.1 (55) and annotated using Abricate (https://github.com/tseemann/abricate). The sequenced genomes were analyzed using Geneious v9.1.8 (Biomatters, Auckland, New Zealand), and the plasmids were aligned and visualized using Mauve plug-in v2.3.1 (56). Plasmids that were assembled and closed with Unicycler and had consistent long read coverage were used for analysis. If Unicycler was unsuccessful at assembling the complete plasmids, the polished Flye assemblies were used for the downstream analysis (Table S3).

Phylogenetic analysis.

Core phylogenetic SNP analyses were performed using paired-end Illumina short reads. These analyses were completed using Snippy v4.4.5, according to the developer’s guidelines, under the Core SNP Phylogeny (https://github.com/tseemann/snippy), using E. coli K-12 substrain MG1655 (GenBank accession number U00096) as the reference sequence. The clean.full.aln files were analyzed using Gubbins v2.4.0 (57), and the output clean.core.aln files were analyzed using SNP-sites v3.0 (58). FastTree v2.2.11 (59) was used to create maximum likelihood phylogenetic trees, which were analyzed in Geneious. Strains with >25% missing data were excluded from the analysis and are summarized in Table S3.

SNP analyses were performed on the hybrid-assembled ESC-resistance plasmids, using an adaptation of a predetermined protocol (32). A gene presence/absence analysis was first conducted on all of the ESC-resistance plasmids, independently of their replicon types. The ESC-plasmids (n = 30) were analyzed using Prokka v1.14.6 (https://github.com/tseemann/prokka), and the output.gff files were analyzed using Roary v3.13.0 (60). A SNP analysis was then performed on plasmids of the same replicon type. The same steps that were previously mentioned with Prokka and Roary were used; however, the SNP-sites were run on the core.gene.alignment output file from Roary. These SNP sites (14601, 13750, 14124, and 12621), along with the core.gene.alignment file, were used to create a maximum likelihood tree in IQtree v2.0.3 (61). All of the trees were visualized in Geneious.

Statistical analysis.

Exact logistic regression models were constructed to assess the associations between reduced susceptibility (i.e., intermediate susceptibility and complete resistance combined) to each of the 12 antimicrobials and the predictor variable bla_CTX-M, in comparison to bla_CMY. Keeping consistent with previous literature, only resistant phenotypes were included in the consideration of multidrug resistance. Multidrug resistance was defined as being resistant to at least one antimicrobial in three or more classes, as determined using the CLSI guidelines. The classes were categorized as follows: aminoglycosides (KAN, GEN, and STR), tetracyclines (TET), quinolones (CIP), folate pathway inhibitors (SUL and SXT), phenicols (CHL), cephems (CTX and FOX), and other β-lactams (AMP). For the purpose of MDR classification, AMC was excluded from the analysis, as resistance to this β-lactam and inhibitor combination was mainly due to the presence of bla_CMY and correlated strongly with FOX resistance. The exact logistic regression models were constructed in Stata v16.1 (StataCorp LP, College Station, TX).

Multilevel logistic regression models were fitted at the replicate level so as to identify any significant associations between the stages of manure processing (i.e., raw versus processed) and the recovery of the four categories of Enterobacterales (i.e., present versus absent): (i) ESC-resistant Enterobacterales carrying bla_CTX-M, bla_CMY, or bla_SHV, (ii) ESC-resistant E. coli carrying bla_CTX-M and/or bla_CMY, (iii) E. coli carrying bla_CTX-M, and (iv) E. coli carrying bla_CMY. The changes from raw to anaerobically digested manure were assessed for all six farms, which was digestate with solids on all of the farms, except for farm five, in which digestate one with solids was used. Due to the diversity of the manure treatment processes on the six farms (Fig. 6), the changes in the environmentally-applied manure and in the recycled bedding were assessed for farms one, three, five, and seven only. The environmentally-applied manure samples were final digestate without solids for all four farms, and the recycled bedding included dewatered manure on farms one, three, and five, as well as heat-treated compost on farm seven. All of the samples were systematically processed in triplicate.

The individual models included each manure process (treated, environmentally-applied manure, or bedding) in reference to the raw input manure as well as the recovery group of interest as fixed effects. The random intercepts included in the models accounted for the clustering and autocorrelation among the samples from the same farm on the date of sampling and contained an additional random intercept to account for the clustering at the sample level. All of the variables were forced into each model. The multilevel models were fitted using an adaptive quadrature with the “melogit” or “meqrlogit” commands in Stata v16.1 (StataCorp LP, College Station, TX). Descriptive statistics alone were reported for outcomes with a prevalence of 100% or 0%, as such data are not suitable for multilevel modeling. A significance level of 5% (i.e., α = 0.05) was used for all of the statistical analyses.

Data availability.

The Escherichia coli assemblies can be found within BioProject PRJNA833106, under the GenBank accession numbers SAMN27959348 through SAMN27959371.