Frequency of isolation and phenotypic antimicrobial resistance of fecal Salmonella enterica recovered from dairy cattle in Canada

Salmonellosis is one of the leading causes of gastro-intestinal infections in humans. In Canada, it is estimated that approximately 87,500 cases of salmonellosis occur every year in humans, resulting in 17 deaths. In the United States, it is estimated that 26,500 hospitalizations and 420 deaths occur every year. In dairy cattle, infections caused by nontyphoidal Salmonella enterica can cause mild to severe disease, including enteritis, pneumonia, and septicemia. Our study objectives were to determine the proportion of fecal samples positive for Salmonella in dairy cattle in Canada and determine the resistance pattern of these isolates. We used data collected through the Canadian Dairy Network for Antimicrobial Stewardship and Resistance (CaDNetASR). Pooled fecal samples from preweaning calves, postweaning heifers, lactating cows, and manure storage were cultured for Salmonella, and the isolates were identified using matrix-assisted laser desorption/ ionization-time of flight mass spectrometry. Anti-microbial susceptibilities were determined using the minimum inhibitory concentration test, and resistance interpretation was made according to the Clinical and Laboratory Standards Institute. A 2-level, multivariable logistic regression model was built to determine the probability of recovering Salmonella from a sample, accounting for province, year, and sample source. The proportion of farms with at least one positive sample were 12% (17/140), 19% (28/144), and 17% (24/144) for the sampling years 2019, 2020, and 2021, respectively. Out of the 113 Salmonella isolates, 23 different serovars were identified. The occurrence of Salmonella appeared to be clustered by farms and provinces. The most common serovars identified were Infantis (14%) and Typhimurium (14%). Overall, 21% (24/113) of the Salmonella isolates were resistant to at least one antimicrobial. Resistance to tetracycline was commonly observed (17%); however, very limited resistance to category I antimicrobials (categorization according to Health Canada that includes third-generation cephalosporins, fluoroquinolones, polymyxins, and carbapenems) was observed, with one isolate resistant to amoxicillin and clavulanic acid. The proportion of Salmonella isolates resistant to 2 and 3 antimicrobial classes was 3.5% and 8.8%, respectively. Our study provided valuable information on the proportion of fecal samples positive for Salmonella , the serovars identified, and the associated resistance patterns across CaDNe-tASR herds, at regional and national levels.


INTRODUCTION
Salmonellosis, usually defined as a symptomatic infection caused by nontyphoidal Salmonella enterica in humans, is one of North America's leading causes of enteric diseases.Most infections in humans caused by nontyphoidal Salmonella enterica are self-limiting, and most infected people recover without specific treatment (Thomas et al., 2013).However, some people can develop severe symptoms that require hospitalization.In Canada, it is estimated that approximately 87,500 cases of salmonellosis occur every year, resulting in approximately 925 hospitalizations and 17 deaths (Government of Canada, 2016b).Approximately 1.35 million cases of salmonellosis occur annually in the United States, resulting in 26,500 hospitalizations and 420 deaths per year (CDC, 2022).
Food-producing animals are an important source of nontyphoidal Salmonella infections in humans (Pires et al., 2009;Zeng et al., 2021).Approximately 80% of human salmonellosis cases in Canada are associated with foodborne transmission, with poultry meat being the most common source of infection (Thomas et al., 2013;Christidis et al., 2020).Human infections were previously linked to cattle exposure.For instance, a review on sources of human salmonellosis suggested that infections can also occur through direct contact with infected dairy cattle or the consumption of raw milk (Hoelzer et al., 2011).Dairy cattle can become infected through horizontal transmission via the fecal-oral route by direct contact with other cattle or a contaminated environment, contaminated feed, and through vertical transmission (Hanson et al., 2016;Holschbach and Peek, 2018).
Similar to what is observed in humans, gastrointestinal infections in dairy cattle caused by Salmonella can have no or mild clinical signs, making them a potential reservoir for nontyphoidal Salmonella enterica (Holschbach and Peek, 2018).For instance, Salmonella Dublin is a cattle-adapted serovar with the ability to persist in the cattle host without showing any clinical signs of infection (Nielsen, 2013).However, Salmonella Dublin may cause severe diseases, such as pneumonia and septicemia, especially in dairy calves (Holschbach and Peek, 2018).
Antimicrobial resistance (AMR) in Salmonella can threaten human and animal health, as it might affect the efficacy of antimicrobial treatment.Several studies from North America reported the resistance pattern from Salmonella isolated from dairy cattle.One California study found a higher proportion of Salmonella resistant to tetracycline, followed by ampicillin and ceftriaxone (Pereira et al., 2019).A second California study also reported the resistance pattern in Salmonella isolated from dairy cattle and found a higher proportion of resistance to streptomycin, followed by tetracycline and ampicillin (Davidson et al., 2018).However, the National Antimicrobial Resistance Monitoring System reported low proportions of resistance in Salmonella spp.recovered from dairy cattle cecal samples collected at the abattoir in 2022.For instance, resistance to tetracycline and ampicillin was 3.1% (8/262) and 2.3% (6/262), respectively (FDA, 2022).In a 2018 Canadian study including 8 dairy herds in New Brunswick, Canada, 16 Salmonella isolates were recovered from dairy calves, and all isolates were susceptible to all antimicrobials tested, except for one isolate that was resistant to sulfisoxazole (Awosile et al., 2018).Another Canadian study, including clinical isolates from 27 cattle operations (dairy and beef) in Alberta, reported moderate to high levels of resistance in Salmonella Typhimurium and Salmonella Dublin to the critically important β-lactam antimicrobials ceftiofur, ceftriaxone, and amoxicillin-clavulanic acid (Otto et al., 2018).To the authors' knowledge, no Canada-wide study reports the resistance pattern of Salmonella isolated from dairy cattle.Therefore, the primary objectives of this study were as follows: (1) to estimate the proportion of fecal samples positive for Salmonella and their geographical location, (2) to identify the Salmonella serovars, and (3) to determine the phenotypic antimicrobial resistance patterns of these isolates.

MATERIALS AND METHODS
Data were collected through the Canadian Dairy Network for Antimicrobial Stewardship and Resistance surveillance system in 2019, 2020, and 2021 (Fonseca et al., 2022).The University of Prince Edward Island Research Ethics Board and the Animal Care Committee reviewed and approved the study on March 7, 2019 (file no.6008059).

Farm Selection and Sample Collection
Sample size calculation and farm enrollment were described elsewhere (Fonseca et al., 2022).Briefly, a convenience sample of 140 farms in 2019 and 144 in 2020 and 2021 from 5 Canadian provinces (British Columbia, Alberta, Ontario, Québec, and Nova Scotia) participated in this study.Enrolled farms had a minimum of 50 animals except for Nova Scotia, which had a minimum size of 40 animals.Participating farms also needed to raise their replacement heifers on-site.On each farm, pooled fecal samples (5 fresh fecal pats were selected from different places on the floor and combined) from each of 3 age groups (preweaning calves, postweaning heifers, lactating cows) and a manure storage sample were obtained.Samples were stored in coolers with ice packs and sent to the Atlantic Veterinary College (Charlottetown, PE, Canada) for bacterial isolation and antimicrobial susceptibility testing.All Salmonella isolates were sent to the Animal World Health Organization Salmonella Reference Laboratory at the Public Health Agency of Canada National Microbiology Laboratory (Guelph, ON, Canada) for serovar identification.

Bacterial Isolation and Characterization
Fecal samples were cultured for Salmonella according to the Canadian Integrated Program for Antimicrobial Resistance Surveillance protocol (Government of Canada, 2020b).Briefly, 25 g from each pooled feces sample or manure storage sample was homogenized in 250 mL of buffered peptone water for each sample using a stomacher for 30 s at 230 rpm and incubated at 35°C for 24 h.Modified Semi-solid Rappaport Vassiliadis agar was inoculated with up to 0.1 mL of the buffered peptone water mix using a sterile pipette.The plates were then incubated at 42°C for 24 to 72 h.After 24 h, the length of migration was measured, and if it was ≥20 mm, 2 or 3 colonies from outside the edge of the migration area were streaked onto a MacConkey culture plate and incubated at 35°C for 24 h.If migration was not evident or <20 mm, the plates were incubated for up to 72 h and checked again for migration.A single colony typical for Salmonella (colorless) was subcultured onto a new MacConkey Agar culture plate and incubated at 35°C for 24 h.Isolates were confirmed to be Salmonella enterica by MALDI-TOF MS using the Bruker Microflex MALDI-TOF MS (Bruker Daltonics, GmbH, Bremen, Germany).A single colony was transferred onto the target plate, air-dried at ambient temperature, and overlaid with alfa-cyano-4-hydroxycinnamic acid before being introduced into the MALDI-TOF mass spectrometer for automated measurement of mass spectra and comparison to the reference database (MBT 8468 MSP Library; https: / / www .bruker.com/en/ products -and -solutions/ microbiology -and -diagnostics/ microbial -identification/ maldi -biotyper -library -ruo .html).Identification scores ≥2.0 were required for confident species identification.Reference strain Salmonella Typhimurium ATCC 14028 was used as quality control for each target plate ran.

In Silico Serotyping
All Salmonella isolates were sent to the National Microbiology Laboratory (Guelph, ON, Canada) for serotyping.These Salmonella isolates were serotyped using either the traditional phenotypic serotyping method (isolates from 2019) or the Salmonella In Silico Typing Resource, a whole genome sequencing (WGS)-based alternative method (isolates from 2020 and 2021; Yoshida et al., 2016).The phenotypic serotyping method detects O or somatic antigens of the Salmonella isolates via slide agglutination.The H or flagellar antigens were identified with a microtiter plate well precipitation method (Shipp and Rowe, 1980).Antigenic formulas and serovars of the Salmonella isolates were identified and designated according to the White-Kauffmann-Le Minor scheme.The Salmonella In Silico Typing Resource detects the genes encoding surface O and H antigens and reports the corresponding Salmonella serovar following the existing White-Kauffmann-Le Minor serotyping scheme (Yoshida et al., 2016).

Statistical Analysis
A descriptive analysis was performed to determine the proportions of Salmonella-positive farms and fecal samples.Frequency distributions of MIC, MIC 50 , and MIC 90 (MIC 50 = the MIC value that inhibits the growth of 50% or more of the isolates; MIC 90 = the MIC value that inhibits the growth of 90% or more of the isolates) were calculated for all isolates and each antimicrobial in the panel.Isolates were also categorized as sensitive or resistant with intermediate sensitivity being categorized as resistant.Additionally, we determined the proportion of different serovars identified, the proportion of isolates resistant to each antimicrobial included in the panel, and the proportion of isolates that were MDR (defined as resistant to ≥3 antimicrobial classes).For all statistical analyses, the unit of analysis was the sample obtained from a given production phase (calves, heifers, lactating cows) or manure storage.Each of these samples was defined by one Salmonella isolate.Given the hierarchical structure of the data, a 2-level, multivariable, logistic regression model with farm as a random intercept was used to determine the probability of recovering Salmonella from a sample.Unconditional associations between the explanatory variables (province, year, and sample source) and the outcome were examined using the previously described multilevel, logistic regression model, but using only one predictor at the time.Chi-squared tests were used to assess associations among these predictors.Only those variables with P ≤ 0.20 were selected for inclusion in the multivariable model.Predictors with P ≤ 0.05 were retained in the final model.Pairwise comparisons using Bonferroni's method were done for categorical predictors with more than 2 categories.Biologically plausible interaction terms were explored (province × sample source; province × year; and sample source × year).Visual examination of the plot of residuals at the higher level did not show any significant pattern that could indicate lack of homoscedasticity.The residuals were also visually inspected for normality.The intraclass correlation coefficient was calculated for the final model using latent variable approximation to estimate clustering within herds (Dohoo, 2009).The odds ratio (OR) was calculated from regression coefficients and adjusted from cluster-specific to the population average using the following formula (Dohoo, 2009): , where βCS is the cluster-specific (CS) estimate to be converted to the population average (PA), and σ2 is the variance of the random effect.Statistical analyses were performed using Stata SE (16.1, StataCorp LLC, College Station, TX).Due to the low number of resistant Salmonella isolates (24/113), no statistical test was conducted.Con-sequently, all comparisons were carried out by considering the numerically higher proportions.

Farm Characteristics
Detailed demographic information of the participating farms was previously described (Fonseca et al., 2022).Briefly, mean herd size was 145 lactating cows, and freestall was the most common barn type (75%).The study farms were comparable to other Canadian dairy farms, except for the proportion of freestall herds in Ontario and Québec, which was higher in the study farms.Additional information on the demographics by province is presented in Table 1.
samples compared with those in British Columbia, Alberta, and Nova Scotia.Farms in Québec also had a higher proportion of Salmonella-positive samples compared with those in British Columbia and Alberta (Table 2).A higher proportion of Salmonella isolates was recovered from manure storage samples compared with the other sample sources (Table 2).Out of the 113 Salmonella isolates, 23 different serovars were identified.No Salmonella Dublin was recovered.The most common serovars identified were Infantis (16/113; 14%), Typhimurium (16/113; 14%), Uganda (11/113; 10%), and Give (9/113; 8%; Figure 3).Salmonella Infantis isolates were recovered from 3 different provinces; however, most were recovered in Ontario (n = 8, from 2 farms) and Québec (n = 5, from 3 farms).Salmonella Typhimurium were also recovered from 3 different provinces, again with most of them being recovered in Ontario (n = 9, from 4 farms) and Québec (n = 6, from 3 farms).Salmonella Uganda were recovered from 4 provinces, most of which were recovered in Québec (n = 7, from 4 farms).Salmonella Give was recovered in Ontario (n = 8, from the same farm) and Alberta (n = 1, from one farm).Salmonella Cerro was recovered in Ontario (n = 4, from the same farm) and Québec (n = 4, from the same farm).The number of each serovar identified in sample source is presented in Table 3.

Multilevel Logistic Regression
Descriptive data and the unconditional association results for the 3 variables considered in the model are presented in Tables 2 and 4. All the 3 variables (province, year, and sample source) met the criteria to be included in the multivariable model.The interaction terms "province × year" (P = 0.9), "province × sample source (P = 0.8), and "sample source × year" (P = 0.2) were not statistically significant at 5% level.Logistic regression final model results are presented in Table 5.The variable "year" was not significant in the final model (P = 0.09).As demonstrated by the descriptive statistics, on average and from any herd, samples from Ontario dairy farms had a higher odd to be positive for Salmonella than British Columbia (OR = 6.05, 95% CI: 2.25−16.51),Alberta (OR = 13.53,95% CI: 2.26−80.91),and Nova Scotia (OR = 4.28, 95% CI: 1.62−11.30).Samples from Québec had a higher odd to be positive for Salmonella than British Columbia (OR = 4.27, 95% CI: 1.58−11.59)and Alberta (OR = 9.47, 95% CI: 2.72−32.94).Manure storage samples had a higher odd to be positive for Salmonella than the other sample sources (Table 5).The intraclass correlation of herd level was 0.51, indicating a high clustering effect among farms.
In our study, the MDR pattern tetracyclinesulfonamide-aminoglycoside was the most frequently observed, accounting for 5.3% (6/113) of the isolates.This pattern was found in the serovars Uganda (n = 5) and Schwarzengrund (n = 1).The tetracycline-only resistance pattern was also common, accounting for 4.4% (5/113) of the isolates, and was most commonly associated with the serovar Infantis.Similarly, the tetracycline-sulfonamide resistance pattern was observed in 3.5% (4/113) of the isolates and was most commonly associated with serovar Uganda.One Salmonella isolate serovar Typhimurium was resistant to 5 antimicrobial classes (6 antimicrobials), whereas one Salmonella isolate serovar I 1,4,[5],12:i:-was resistant to 4 classes (5 antimicrobials), as presented in Figure 6.

DISCUSSION
Our study focused on Salmonella recovered from feces and manure on Canadian dairy farms and the   antimicrobial resistance phenotypes of the isolates.The proportion of farms with at least one Salmonellapositive sample from 2019 to 2021 was relatively low, ranging from 12% to 19%.In addition, the multilevel logistic model indicated a clustering effect on the occurrence of Salmonella in our study farms, meaning that if a Salmonella was isolated from a farm, it was more likely to be isolated from multiple sample sources and on multiple years.
In contrast to the low proportion of farms positive for Salmonella found in our study, according to the report from the National Animal Health Monitoring System, 39.7% (48/121) of the dairy operations in the United States included in the survey had a fecalculture positive for Salmonella in 2007 (USDA/ APHIS, 2007).A study from Pennsylvania, published in 2019, which also collected pooled fecal samples from 80 dairy farms, reported that 64% of the farms had at least one sample positive for Salmonella (Cao et al., 2019).This disparity might be partially attributed to methodological differences, as the Pennsylvania study collected one pooled fecal sample from multiple sites within a pen (6-8 sites), which could increase the sensitivity in detecting a Salmonella isolate.Furthermore, according to a study published in 2019 using survey data to describe the management of dairy herds in Pennsylvania, it was possible to identify differences in demographic characteristics between our farms and those in Pennsylvania, which could influence the occurrence of Salmonella (Holly et al., 2019).For instance, mean herd size in Pennsylvania farms was smaller (71 lactating cows) compared with our farms (145 lactating cows).Additionally, the proportion of barn types differed, with tiestall barns being the most common in Pennsylvania (34%), whereas freestall barns were more common for the herds in our study.These variations in methodology, herd characteristics, and management practices might have contributed to the differences observed in Salmonella occurrence between the studies.Furthermore, the higher proportion of Salmonella-positive farms in both the Pennsylvania study and the National Animal Health Monitoring System report could potentially be attributed to the specific serovars detected within these research studies.A higher proportion of serovars Cerro and Kentucky was identified, both of which are commonly distributed across dairy operations within the United States (USDA/APHIS, 2007; Cao et al., 2019).In the present study, the overall proportion of Salmonella recovered from fecal samples was 6.6%.The proportion of Salmonella-positive fecal samples in our study was similar to that reported in a 2005 study conducted in the United States, which included dairy farms from 4 states and reported an overall proportion of 4.9% (Fossler et al., 2005).The proportion of Salmonella-positive fecal samples was also similar to what was described in a meta-analysis published in 2019.This meta-analysis included studies from various countries; however, most studies were conducted in the United States.The meta-analysis reported an overall proportion of 9% (95% CI: 7%-11%) for Salmonella in apparently healthy cattle (beef, dairy, and mixed), with a higher proportion of 16% (95% CI: 12%-20%) when considering only studies from North America (Gutema et al., 2019).Other studies reported slightly higher proportions of Salmonella recovered from fecal samples in dairy cattle.A study using data from 4 states in the United States reported 10% of samples were positive for Salmonella (Callaway et al., 2005).Similarly, another study from Pennsylvania reported an overall proportion of 11.5% (Cao et al., 2019).These varia-tions in proportions might be attributed to differences in study design, geographical location, herd management practices, sampling, and laboratory techniques.Overall, the proportion of Salmonella recovered from fecal samples in our study seemed comparable to other studies conducted in North America.
Regional factors might also influence the occurrence of Salmonella in dairy cattle in Canada.To our knowledge, this study is the first to report the proportion of Salmonella recovered from fecal samples collected on-farm in dairy herds across 5 provinces in Canada.In our study, the proportion of positive samples for Salmonella was higher from dairy farms located in Ontario and Québec than in British Columbia, Alberta, and Nova Scotia.The province with a lower proportion of positive samples was Alberta (1.1%).A low proportion of Salmonella from dairy cattle in Alberta was also reported in a study from 2003, where pooled fecal samples were collected (2 to 3 cows for each sample) in 50 dairy farms across the province.That study reported a proportion of 0.7% of positive samples out of the 750 pooled samples collected during the study (Sorensen et al., 2003).In another Canadian study, which col- lected data from 8 dairy farms in New Brunswick, 3.3% of 488 fecal samples collected from dairy calves, were Salmonella-positive (Awosile et al., 2018).In addition, 3 serovars were identified in this latter study: Salmonella Senftenberg, Salmonella Typhimurium DT02, and Salmonella Derby (Awosile et al., 2018).In the present study, no Salmonella was recovered from pre-weaned calf samples in Nova Scotia (another Maritime province).Additionally, none of the serovars identified by Awosile et al. were detected in the positive samples for Nova Scotia (Table 3).The demographics of the study farms were found to be similar to those of other Canadian dairy farms, with one exception.It was observed that the study farms from Ontario and Québec had a higher proportion of freestall herds in comparison to the general distribution among dairy farms across Canada.The variation in the proportion of Salmonellapositive samples across provinces could be attributed to other management factors not controlled by this study.
Although not reported, we explored herd size and barn type in the analysis, and neither was associated with the outcome (data not shown).
The effectiveness of Salmonella detection in dairy cattle varies depending on the type of sample collected, which can influence the sensitivity of detection for surveillance purposes.In our study, a higher proportion of isolates were recovered from manure samples, specifically the manure storage area.A study in the United States collected data from 5 states, including sampling from various sources such as the farm environment and direct rectal swabs from animals, including swine, dairy and beef cattle, and poultry.They observed a higher occurrence of Salmonella in samples collected from the environment than in rectal swabs.Specifically, the proportion of Salmonella recovered from rectal swabs in dairy cattle was 0.4%, whereas the environment samples ranged from 10% in feed to 15% in bedding material (Rodriguez et al., 2006).The findings for environmental samples aligned with our study, which also found a higher proportion of Salmonella in manure samples (14%).As our study was part of a national surveillance focused on AMR, our primary goal was not to estimate the prevalence of Salmonella at the farm level.Van Kessel et al. (2008)  more suitable methods for this purpose.In addition, certain serovars were exclusively isolated from fresh fecal pats of animals.Consequently, the current sampling scheme, incorporating fresh fecal pats from animals and manure samples, offered advantages over relying solely on manure samples.Finally, although manure samples are more representative of the herd than fecal samples from a given production age, they might be influenced by external contamination from other animal sources or the surrounding environment.As a result, the Salmonella recovered from manure samples may not solely originate from the dairy farm itself.
Differences in the proportion of Salmonella recovered from different production ages in dairy cattle were previously reported.A study from Pennsylvania that collected fecal samples from different production ages in dairy farms reported a higher proportion of positive samples for Salmonella in lactating cows (64%), followed by dry cows (61%).In the same study, the proportion of Salmonella recovered from pre-weaned calves was 13%, indicating a lower proportion when compared with other production ages (Cao et al., 2019).However, in our study, we did not observe any major variation in the proportion of Salmonella-positive samples among the different production ages.To the authors' knowledge, no studies have reported production age as a risk factor for Salmonella.Additionally, the probability of recovering Salmonella from lactating cows and dry cows can be affected by other risk factors.For instance, access of lactating or dry cows to surface water and herd size, have been previously identified as potential risk factors for Salmonella shedding in dairy cattle (Kabagambe et al., 2000;Fossler et al., 2005).Therefore, the variation in the proportion of Salmonella observed in different production ages across studies could be attributed to other confounding factors beyond age alone.
In addition to determining the proportion of Salmonella recovered from dairy herds, identifying the specific serovars is important for a comprehensive understanding of the distribution of different Salmonella strains within the dairy industry.It is well known that Salmonella Dublin is a concern for the dairy industry in North America (Fritz et al., 2022).This serovar is host-adapted for cattle, is associated with subclinical shedding over animal life, and is usually associated with MDR (Hsu et al., 2019).In our study, the serovar Dublin was not recovered; however, it is important to mention that our sampling scheme was not designed to detect Salmonella Dublin.We collected pooled fecal samples, which have been previously reported to exhibit a lower sensitivity in detecting this particular serovar (Nielsen et al., 2007;Frye, 2021).Other validated methods, such as testing bulk milk samples for the presence of antibodies, have been previously described as a convenient and cost-effective strategy for detecting Salmonella Dublin.However, the sensitivity of ELISA testing to detect Salmonella Dublin in bulk milk can vary, ranging from 40% to 100% in different studies (Nielsen, 2013;Um et al., 2022;Nobrega et al., 2023).
Although no Salmonella Dublin was identified, 2 other serovars of concern, Typhimurium and Newport, were identified.These 2 serovars can also be associated with subclinical shedding in dairy cattle, posing a contamination risk for people with direct contact with the animals, their feces, or ingesting contaminated raw milk (Holschbach and Peek, 2018).Additionally, Salmonella Typhimurium is commonly associated with MDR and mortality in dairy calves (Carroll et al., 2020;Casaux et al., 2023).In our study, Salmonella Newport represented 2.7% of recovered isolates (2 isolates from Ontario and one from Québec), and all 3 Salmonella Newport isolates came from manure storage samples.A study published in 2018 using WGS suggested that an outbreak from 21 states in the United States caused by Salmonella Newport could possibly be linked to dairy cattle sources (contaminated ground beef produced from slaughtered dairy cows; Marshall et al., 2018).Salmonella Typhimurium was one of our study's most commonly isolated serovars (n = 16), accounting for 14% of the recovered isolates (half of the isolates came from animal samples).Our findings suggested that this serovar was clustered within certain provinces and farms, as out of the 16 identified Salmonella Typhimurium, 9 were identified in Ontario (across 4 farms) and 6 in Québec (across 3 farms).Another study conducted in Alberta, Canada, analyzed clinical Salmonella isolates from dairy cattle and reported that Salmonella Typhimurium corresponded to 58.3% of the isolates (28/48;Otto et al., 2018).A WGS analysis from Salmonella recovered from humans and dairy cattle in New York and Washington states suggested that 2 Salmonella Typhimurium isolates from humans were remarkably similar based on their AMR gene sequences with an isolate from dairy cattle (Carroll et al., 2017).The AMR genes in all 3 isolates displayed 100% sequence identity, except for one gene, tet(RG), which exhibited a nucleotide difference at position 73 (Carroll et al., 2017).Additionally, a report documented an outbreak of MDR Salmonella Typhimurium in 2018 linked to the consumption of soft cheese in Mexico and beef in the United States (Plumb et al., 2019).These combined findings might suggest that dairy cattle and the dairy farm environment could serve as a potential source of Salmonella Typhimurium and Newport infections in humans.
Salmonella Infantis was the other most frequent serovar identified in our study (15%).Similar to Salmonella Typhimurium, our results suggested that this serovar was also clustered within specific provinces and farms, with 8 isolates recovered from 2 farms in Ontario and 5 from 3 farms in Québec.According to the literature, this serovar is usually associated with poultry, and in 2016, an outbreak associated with exposure to raw chicken was reported in Canada, including cases across 9 provinces (Government of Canada, 2016a).Previous studies have highlighted the variability in the occurrence of Salmonella Infantis in dairy cattle among different geographic locations and farming systems (Cummings et al., 2009;Davidson et al., 2018;Aleri et al., 2022;Abraham et al., 2022).In North America, the proportion of isolation for this serovar was reported to be 1% and 8% in 2 different studies from the United States (Cummings et al., 2009;Davidson et al., 2018).In addition, at national level, less than 1% (1/121) of the dairy operations in the United States were positive for Salmonella Infantis (USDA/APHIS, 2007).In pasturebased farms in Australia, the occurrence of Salmonella Infantis was particularly high, accounting for 61% of the serovars recovered according to a study published in 2022 (Aleri et al., 2022).Out of the 16 Salmonella Infantis isolates, 7 were recovered from animal samples.
In addition to the occurrence of Salmonella, we described the antimicrobial resistance phenotypes of the Salmonella isolates.Out of the 47 Salmonella-positive farms, only 16 had Salmonella-resistant isolates.Antimicrobial-resistant Salmonella were isolated from all the production phases and the manure storage.All provinces had at least one Salmonella resistant to at least one antimicrobial, except Alberta, where no resistance was observed.Among the antimicrobials tested, resistance to tetracycline was highest, accounting for 17% of the isolates, followed by sulfisoxazole (13%) and streptomycin (12%).In contrast to our study, a study from Pennsylvania reported that all the 124 Salmonella isolates recovered from different production ages were pan-susceptible according to the National Antimicrobial Resistance Monitoring System antimicrobial panel (Cao et al., 2019).However, higher proportions of resistance to streptomycin and tetracycline were previously reported in North America.A study from California found that 51% of the Salmonella isolates from dairy cattle were resistant to streptomycin, and 50% were resistant to tetracycline (Davidson et al., 2018).A second study from California also reported higher proportions of resistance to tetracycline (39%) in Salmonella isolated from cull dairy cows (Pereira et al., 2019).Although resistance to sulfisoxazole in our study was the second most commonly identified, both studies from California found no resistance to sulfisoxazole.The absence of resistance to sulfisoxazole in California studies could be due to the fact that trimethoprim-sulfamethoxazole is not approved to be used in dairy cattle in the United States (FDA, 2023).In Australia, a study from 2022 reported a higher proportion of non-wild-type Salmonella isolates that were resistant to streptomycin (57%); however, all isolates were susceptible to all 16 tested antimicrobials using Clinical and Laboratory Standards Institute clinical breakpoints (Aleri et al., 2022).The differences in the resistance pattern could be attributed to different herd management, such as the choice of antimicrobials being used and other management that could affect the burden of AMR on those farms.
Antimicrobial resistance involving antimicrobials considered highly important for human medicine are of particular concern, especially in food-producing animals (Scott et al., 2019).In Canada, antimicrobials are classified into 4 categories; those very high important for human medicine are considered category I (Government of Canada, 2009).Our study observed lower resistance to category I antimicrobials, as only one isolate was resistant to amoxicillin/clavulanic acid.A study conducted in Alberta, Canada, reported higher proportions of resistance to category I antimicrobials.Resistance to ceftiofur, ceftriaxone, and amoxicillin/ clavulanic acid was 43.6% for Salmonella Typhimurium.In addition, resistance to ceftiofur and ceftriaxone was 68.8% in Salmonella Dublin (Otto et al., 2018).However, these Salmonella isolates were obtained from clinical samples, and these serovars are commonly associated with resistance to category I antimicrobials (Davidson et al., 2018;Hsu et al., 2019).Other studies from the United States have also reported resistance to these antimicrobials.For instance, a study from California reported 10, 9, and 5% of Salmonella resistant to ceftriaxone, amoxicillin/clavulanic acid, and ciprofloxacin, respectively (Davidson et al., 2018).Another study from California, including Salmonella isolates recovered from healthy and diseased animals, found that 40 and 46% of Salmonella was resistant to amoxicillin/clavulanic acid and ceftriaxone, respectively (Davidson et al., 2018).These findings highlight the variability in antimicrobial resistance patterns in North America and emphasize the need for continued surveillance.
Some Salmonella serovars are more frequently resistant to antimicrobials than others (Davidson et al., 2018).In our study, we identified MDR isolates representing 4 Salmonella serovars: Typhimurium, Uganda, Schwarzengrund, and I 1,4,[5],12:i:-.The predominant MDR pattern observed was tetracycline-streptomycinsulfisoxazole, accounting for 5.3% (6/113) of the isolates, with Uganda being the most commonly associated serovar with this pattern of resistance.The serovar I 1,4,[5],12:i:-was associated with 2 different MDR patterns: tetracycline-sulfisoxazole-ampicillin and tetracycline-sulfisoxazole-ampicillin-chloramphenicol-trimethoprim/sulphamethoxazole.Although only one Salmonella Typhimurium isolate was identified as MDR, it displayed resistance to 6 different antimicrobials: tetracycline, sulfisoxazole, streptomycin, ampicillin, chloramphenicol, and amoxicillin/clavulanic acid.A study conducted on dairy cattle in 2009 reported MDR proportions of 60% and 39% for the I 1,4,[5],12:i:-and Typhimurium serovars, respectively (Cummings et al., 2009).A second study highlighted that all Typhimurium isolates were MDR (Davidson et al., 2018).Additionally, this same study reported 50% of Salmonella isolates being MDR, and the most common MDR pattern was amoxicillin/clavulanic acidampicillin-cefoxitin-ceftiofur-ceftriaxone-chloramphenicol-streptomycin-tetracycline (16%), which included resistance to category I antimicrobials.Furthermore, in 2019, a study from California reported an MDR proportion of 12% among Salmonella isolates, with certain MDR patterns displaying resistance to category I antimicrobials, such as third-generation cephalosporins and fluoroquinolones (Pereira et al., 2019).These findings highlighted the concerning presence of resistance to critically important antimicrobials in Salmonella isolates; however, our results suggested that such resistance is lower in Canadian dairy herds.Furthermore, it is worth noting that Salmonella Dublin is commonly associated with MDR (Hsu et al., 2019); thus, our study's absence of Salmonella Dublin might have influenced the lower proportion of MDR observed.
Some limitations applied to the present study.Convenience sampling could introduce selection bias if the farms in the present study did not represent the source population.Participants in our study were not blinded to the study's objective, and farmers who agreed to participate in the study could be the ones with higher levels of biosecurity, underestimating the burden of Salmonella and resistance associated with these isolates.In addition, the proportion of freestall herds in Ontario and Québec study farms slightly differed from other commercial dairy herds in Canada as demonstrated in Table 1.Therefore, interpretation of results must take that into account, which could affect the external validity of the results to other dairy farms in Canada.Another limitation was the low number of Salmonellaresistant isolates which precluded the logistic regression analysis.Finally, the ability to recover Salmonella from fecal samples can vary according to the culture method employed (Eriksson and Aspan, 2007;Love and Rostagno, 2008).Therefore, comparisons of results among different studies might be difficult when different culture methods are used.
Our study findings revealed a clustering pattern in the occurrence and resistance of Salmonella among farms and provinces.Understanding the diversity and occurrence of Salmonella serovars in dairy cattle and the re-sistance pattern associated with these serovars provides important information to mitigate infections related to foodborne transmission and improve animal health.We identified a very low proportion of Salmonella isolates resistant to highly important antimicrobials compared with previous studies from North America.
Figure 1.Number of positive fecal samples for Salmonella per farm from 2019 to 2021.On each farm, up to 4 samples were collected (preweaning calves, postweaning heifers, lactating cows, and manure storage).

Fonseca
Figure 2. Dairy farms in which Salmonella was recovered in at least one sample from 2019 to 2021.On each farm, up to 4 samples were collected (preweaning calves, postweaning heifers, lactating cows, and manure storage) each year.Each farm is represented by the identifier H0XXX.

Fonseca
Figure 3. Distribution of serovars among 113 isolates recovered from fecal and manure samples from 2019 to 2021.

Fonseca
Figure 4. (A) Proportion of resistant Salmonella isolates for each antimicrobial (AM) tested in the panel by province from 2019 to 2021.AMC = amoxicillin/clavulanic acid; AMP = ampicillin; FOX = cefoxitin; CHL = chloramphenicol; NAL = nalidixic acid; STR = streptomycin; SOX = sulfisoxazole; SXT = trimethoprim/sulphamethoxazole; TET = tetracycline.No resistance was observed for azithromycin, ceftriaxone, ciprofloxacin, gentamicin, and meropenem.(B) Proportion of Salmonella isolates pan-susceptible or resistant to one or more AM classes by sample source from 2019 to 2021.No resistance was observed in isolates from Alberta.

Fonseca
Figure 5. (A) Proportion of resistant Salmonella isolates for each antimicrobial (AM) tested in the panel by sample source from 2019 to 2021.AMC = amoxicillin/clavulanic acid; AMP = ampicillin; FOX = cefoxitin; CHL = chloramphenicol; NAL = nalidixic acid; STR = streptomycin; SOX = sulfisoxazole; SXT = trimethoprim/sulphamethoxazole; TET = tetracycline.No resistance was observed for azithromycin, ceftriaxone, ciprofloxacin, gentamicin, and meropenem.(B) Proportion of Salmonella isolates pan-susceptible or resistant to one or more AM classes by sample source from 2019 to 2021.

FonsecaFigure 6 .
Figure 6.Phenotypic resistance pattern for the 24 of 113 Salmonella isolates recovered from fecal samples that were resistant to at least one antimicrobial tested.(A) Year in which the isolates were recovered; (B) province where the isolates were recovered; (C) sample source where the isolates were recovered; (D) Salmonella serovar; (E) number of isolates and their respective phenotypic resistance pattern.The different colors for the connected dots represent how many antimicrobials an isolate was resistant to: green = resistant to 6 antimicrobials; blue = resistant to 5 antimicrobials; red = resistant to 3 antimicrobials; black = resistant to 2 antimicrobials; light blue = resistant to 1 antimicrobial.TET = tetracycline; SOX = sulfisoxazole; STR = streptomycin; AMP = ampicillin; CHL = chloramphenicol; SXT = trimethoprim/sulphamethoxazole; NAL = nalidixic acid; FOX = cefoxitin; AMC = amoxicillin/clavulanic acid.No resistance was observed in isolates from Alberta.

Fonseca
et al.: FECAL SALMONELLA FROM CANADIAN DAIRY CATTLE Fonseca et al.: FECAL SALMONELLA FROM CANADIAN DAIRY CATTLE

Table 1 .
Fonseca et al.: FECAL SALMONELLA FROM CANADIAN DAIRY CATTLE Demographics from the 144 dairy farms, by province, enrolled in the Canadian Dairy Network for Antimicrobial Stewardship and Resistance surveillance (study farms) compared with the Canadian dairy farms' statistics (without robotic system) in 2022 1

Table 2 .
Number and proportion (%) of Salmonella-positive samples over provinces by sample source from 2019 to 2021; a total of 1,709 fecal samples were collected from Canadian dairy farms from 2019 to 2021 (140 dairy farms in 2019 and 144 dairy farms in 2020 and 2021)

Table 3 .
Number of Salmonella serovars identified in the 113 positive samples for Salmonella enterica from each sample source from 2019 to 2021; each sample source was represented by only 1 Salmonella isolate per farm and per year

Table 4 .
Fonseca et al.: FECAL SALMONELLA FROM CANADIAN DAIRY CATTLE Unconditional association of herd-level (n = 144) categorical predictors with the probability of Salmonella isolation in fecal samples collected from Canadian dairy farms

Table 5 .
Final multilevel logistic regression model for the isolation of Salmonella from 1,709 samples (144 farms) collected from calves, heifers, lactating cows, and manure storage from 2019 to 2021 1 Fonseca et al.: FECAL SALMONELLA FROM CANADIAN DAIRY CATTLE