1. Introduction
Salmonella enterica subsp.
enterica serovar
Dublin (
S. Dublin) is a cattle-adapted serovar that causes significant morbidity and mortality in juvenile and adult animals [
1,
2]. Though infrequent, infection in humans can occur and often leads to clinically severe infection, with septicemia and death as common sequelae [
3]. In recent years
S. Dublin has emerged as one of the most frequently isolated serotypes of
Salmonella in cattle [
2]. Infected animals may shed organisms in feces, milk and colostrum, and chronic asymptomatic shedding may occur in carrier animals [
4]. Once in the environment,
S.
Dublin can persist for months to years [
1,
5,
6]. There is an increasing prevalence of
S. Dublin reported in both cattle and humans, as well as multidrug resistance (MDR) detected in these isolates [
3,
7]. As a host-adapted serovar in cattle,
S. Dublin is highly invasive and commonly associated with septicemia [
8]. Multiple reports suggest that the prevalence of infection, the risk for asymptomatic carrier state with organism shedding and the higher levels of antimicrobial resistance (AMR) are greater in young dairy cattle as compared to adult cattle and beef cattle operations [
4,
7,
9]. Antibiotic treatment is controversial, as
Salmonella spp. are frequently resistant to many antibiotics approved for use and antibiotic exposure may lead to the development of drug resistance in non-target organisms. However, systemic infections and infections in young calves carry a high risk of mortality without the timely administration of effective antibiotics [
2].
The increasing threat that antimicrobial resistance poses to both human and animal health has led to an abundance of multilevel national and international efforts to evaluate, monitor and curb antimicrobial resistance. The 2015 National Antimicrobial Resistance Monitoring System (NARMS) Integrated Report indicates increasing MDR in
S. Dublin with a high level of resistance to ceftriaxone and decreased susceptibility to ciprofloxacin, which are critically important for treating severe
Salmonella infections in humans [
10]. Drug-resistant non-typhoidal (NT)
Salmonella was listed as one of the Serious Threats in the 2019 National Strategy for Combating Antibiotic-Resistant Bacteria (CARB) Antibiotic Threats in the United States report [
11]. Along with
Campylobacter, NT-
Salmonella enterica are the most frequent bacteria causing foodborne illness in humans in the United States, with an estimated 1.35 million illnesses and 420 deaths caused each year by
Salmonella [
12].
The extensive use of antibiotics in food-producing animals is often implicated as one of the major driving forces for the global trends in antimicrobial resistance [
13]. To address this concern, recent legislative actions have led to the increased oversight and regulation of antimicrobials used in food-producing animals, including the categorization and restricted use of medically important antimicrobial drugs (MIADs) by the United States Food and Drug Administration (FDA). In 2017, the FDA implemented the Veterinary Feed Directive (VFD) to require oversight by a licensed veterinarian and a VFD order for all MIADs used in feed or a prescription for MIAD use in water [
14]. California legislation (Livestock: Use of Antimicrobial Drugs, Food and Agricultural Code, FAC 14,400–14,408) enacted in January, 2018, expanded upon these requirements to ensure that any form of administration of a MIAD to livestock is performed in the context of a veterinarian–client–patient relationship [
15].
Information on the long-term patterns of AMR in pathogens of critical importance, such as Salmonella, is scarce, particularly among livestock species. The purpose of this study was to evaluate the minimum inhibitory concentrations (MICs) and antimicrobial susceptibility trends to antimicrobials of clinical relevance to both veterinary and human medicine in S. Dublin isolates recovered from cattle in California over a 27-year timeframe.
3. Discussion
Because clinical breakpoints are not available for S. Dublin in cattle, NARMS interpretive criteria were used to interrogate the trends and patterns of AMR in the study samples. While NARMS interpretations may not accurately reflect clinical efficacy of the drug in the bovine host, monitoring trends and risk factors for resistance and MDR using these available criteria provides information inherent to the bacteria that reveals a trend in decreasing antimicrobial susceptibility that likely is matched by decreasing clinical drug efficacy. The results of this study reveal a trend toward increasing AMR for S. Dublin for the majority of drugs evaluated, as well as increased risk for MDR in later years compared to the baseline 1993–1999 year interval. A number of factors may contribute to the observed changes in resistance; those considered most significant are discussed below.
Depending on the interpretation criteria used, 67% or 97% of S. Dublin isolates were classified as resistant to ampicillin. A significant increase in OR for resistance was found for the year interval 2006–2010 compared to the baseline 1993–1999 interval, but no other year group comparisons yielded significant differences. Similar to ampicillin, the majority of isolates (76%) were resistant to tetracycline, with increased odds for resistance seen for 2006–2015 when compared to the baseline 1993–1999 interval. As will be discussed below for MDR in Salmonella, ampicillin and tetracycline resistance is commonly reported among S. Dublin, and trends in mean MIC will be presented in the accompanying second manuscript. For amoxicillin/clavulanic acid (AMC), the odds for resistance were higher in all later year intervals, compared both to the baseline 1993–1999 year group and also to all study years preceding 2011–2019, at which point the odds for resistance did not differ from 2006–2010 year interval.
For year interval comparisons, significant increases in odds for resistance to cephalosporins were found over time from the 1993–1999 year interval until 2006–2010, after which comparisons were no longer significantly different. This trend is similar to what was found for AMC. According to the 2015 NARMS Human Isolates Surveillance report, 67% of
Dublin isolates were resistant to ceftriaxone [
20], similar to the proportion of resistant isolates in this sample set for the third-generation cephalosporins tested (ceftiofur and ceftriaxone) for 2006–2019. The 2012 order issued by the FDA’s Center for Veterinary Medicine to prohibit the extra label use of cephalosporin drugs in cattle (FDA Extra label Use and Antimicrobials) [
21] may explain why levels of resistance did not continue to rise in later year intervals in this study. It is worth noting, however, that while a continued increase in resistance was not evident, the proportion of isolates resistant to cephalosporins also did not significantly decrease in more recent study year intervals following the 2012 regulatory change in the use of these drugs.
Chloramphenicol use is prohibited in food-producing animals, but its fluorinated derivative, florfenicol, is widely used to treat a variety of diseases in cattle and is approved to treat bovine respiratory disease. A number of genes have been shown to mediate resistance in both florfenicol and chloramphenicol [
22], and the same genes may confer resistance for both [
23]. It has also been shown that some bacterial isolates that harbor resistance genes, and are phenotypically resistant to chloramphenicol, are susceptible to the fluorinated derivatives. Additionally, the mechanisms for co-selection of resistance genes may occur in the absence of selection pressure imposed by use of chloramphenicol or florfenicol [
24]. In the absence of clinical breakpoints to florfenicol in cattle, it is impossible to determine if the chloramphenicol resistance approximates florfenicol resistance. In the second accompanying manuscript, trends in MIC values are applied to glean information about susceptibility for drugs on the bovine clinical BOPO6F panel.
The odds ratio for resistance to chloramphenicol was higher for the fall season compared to winter or summer. It is possible that this finding coincides with a history of increased prevalence and treatment of bovine respiratory disease (BRD), for which florfenicol is widely used. A recent study examining the epidemiology of BRD in preweaned calves in CA dairies found a higher prevalence for BRD in the fall [
25].
Significant increases in resistance to nalidixic acid and ciprofloxacin were detected in isolates from 2011–2019. Nalidixic acid is a synthetic quinolone used for surveillance purposes to detect diminished fluoroquinolone susceptibility [
26]. Interestingly, shifts in increasing MIC for nalidixic acid preceded those seen in the fluoroquinolones tested on both the BOPO6F and NARMS panels. The value of monitoring changes in MIC (that are not necessarily reflected by detected changes in measured resistance) will be discussed in the second accompanying paper.
Because S. Dublin infections frequently result in systemic disease that commonly includes respiratory signs, it is possible that empirical treatment is being misdirected at bacterial organisms commonly associated with bovine respiratory disease (BRD), namely Pasteurella multocida, Mannhemia haemolytica, and Histophilus somni. Florfenicol, fluoroquinolones, macrolides, and cephalosporins (ceftiofur) might be selected, as labelled treatment options, and contribute to increased resistance in S. Dublin.
The rapid reversal of resistance seen with gentamicin, apparently preceding downward trends in other antimicrobials and greater than the changes observed for other aminoglycoside drugs in the panels, suggests an abrupt and widespread decrease in its use in California. It is possible that the extra label use of parenteral gentamicin in cattle had previously been more widespread but essentially came to a halt as FDA, the American Veterinary Medical Association, and others issued warnings about misuse and residues in the late 1990s and early millennium. [
27,
28]. Around that time there were FDA warnings against its illegal use [
27]. Furthermore, there were emerging data from the Food Animal Residue Avoidance Databank (FARAD), showing the prolonged detection of aminoglycoside residues that posed a risk for violative residues in food animal tissues and the prohibition of extra label use of drugs in feed by the Animal Medicinal Drug Use Clarification Act (AMDUCA) [
1]. Interestingly, the OR for resistance for gentamicin was higher in the south compared to both north and central regions, possibly reflective of regional practices, perceptions, or adherence to new restrictions; or animal products from those with higher resistance to gentamicin may have been family animals or otherwise not destined for human consumption.
Increased levels of MDR were observed for all time intervals when compared to the baseline year interval 1993–1999. This is similar to what was reported for
S.
Dublin for the year interval 1996–2004 compared to 2005–2013 [
3] and comparisons of 2002–2009 and 2010–2016
S.
Dublin isolates from dairy cattle in California that similarly showed an increase in MDR [
7]. The whole genome sequencing of
S. Dublin isolates points to a plasmid-mediated acquisition of MDR [
29,
30,
31]. Plasmid-mediated MDR has been described elsewhere as well [
24]. Significant differences in phenotypic AMR profiles and the presence/absence of plasmid replicons were found by geographic location [
31], which aligns with the findings of a study in Denmark that utilized WGS data to describe the epidemiology of
S.
Dublin revealing persistence of a single strain within herds for years [
30]. The WGS of the isolates in this study, which is currently in progress, would further elucidate the basis of phenotypic resistance and any patterns related to geography or the year that the isolate was recovered.
Frequently encountered MDR patterns reported for
S. Dublin isolated from cattle include ampicillin, chloramphenicol, streptomycin, sulfonamides, and tetracycline, with increasing levels of ceftriaxone resistance in more recent years [
20,
32]. Utilizing NARMS criteria, 67% of all isolates were resistant to sulfisoxazole but only 3.2% were resistant to TMS (
Table 3 For the other drugs typically associated with MDR in
S. Dublin listed above, between 67% and 81% of isolates in this study were classified as resistant.
In November 2019, an outbreak of
S. Dublin in humans occurred that was traced back to contaminated ground beef from California. This recent outbreak highlights the seriousness of risk to humans: 82% of sick individuals required hospitalization and one person died; and
S. Dublin was recovered from the blood of 46% of infected individuals [
33]. In addition, the outbreak led to the recall of more than 34,000 pounds of ground beef, also highlighting the economic and food safety impacts from this serovar. Interestingly, whole-genome sequencing on the outbreak isolates did not identify any known antibiotic resistance mechanisms. AMR has been reported to be higher in clinical vs. non-clinical infections in cattle [
26], which might explain this finding. A study that compared AMR between bovine and human isolates of
S. Dublin found that resistance was generally higher in bovine isolates and was associated with the presence of an IncA/C2 plasmid that genetically distinguished the bovine from human isolates [
34].
With the exception of gentamicin, the odds ratio for resistance to each of the drugs tested on the NARMS panel was higher for central and southern California regions compared to northern California. Possible explanations for this include differences in the sizes and numbers of production facilities distributed throughout California, with larger numbers of large-scale productions being located in central and southern California [
35]. It is also possible that regional variation in production types and their respective management practices regionally influence antimicrobial exposure and selection pressures for resistance. As an example, counties within the San Joaquin Valley, (the majority of which are in the central and southern regions of California), support the largest dairy productions in the state [
35].
Increased odds for resistance to tetracycline, streptomycin, and MDR were encountered for isolates obtained from animals in the Early PW and Early HF age groups compared to adult cows. Because younger animals are more susceptible to a variety of neonatal disease agents, the increased therapeutic use of antimicrobials in this age group may account for some of the higher levels of resistance seen in younger animals. As reviewed by Springer et al., increased risk for harboring resistant
Salmonella enterica and
Escherichia coli among younger calves has been documented in numerous studies; however, the factors that contribute to the relative resistance remain poorly understood [
36].
The present study has several limitations. S. Dublin isolates originated from sick or dead animals, creating a biased sampling. Specimens (i.e., feces) or necropsy tissues from which the S. Dublin was isolated may not represent the levels of resistant isolates that would be encountered during surveillance of healthy animals. It can be assumed that most of the source animals had likely received antibiotic treatment(s) prior to testing at CAHFS; however, treatment history is only rarely provided at the time of sample or carcass submission and so cannot be accurately established. Additionally, S. Dublin is frequently isolated from multiple tissues; the specimen source isolate that was selected for further characterization was at the discretion of the diagnostician and potentially subject to lack of equal representation in the study dataset. Clinical history provided on submission forms was variable and generally limited; therefore, information regarding clinical history, management practices (including prior antimicrobial administration), and production type were not available to better analyze risk factors. This study does not provide information about the prevalence of S. Dublin among cattle in California.
4. Materials and Methods
4.1. Sample Source
S.Dublin isolates were recovered from cattle specimens submitted to the California Animal Health and Food Safety Laboratory (CAHFS) between January 1993 and December 2019. Isolates were obtained from clinically ill animals either at the time of necropsy or from diarrhea samples. Approximately ten isolates per year were selected to increase the representation of the diversity of isolates over time. Criteria for inclusion included having a confirmed serovar as being Dublin and having an association with the clinical disease based on laboratory submission form information; furthermore, to reduce the risk of oversampling, a same location, only one isolate what collected from the same farm in the same year. Additionally, whenever possible, isolates within each year period were selected from distinct counties to avoid over-representation from particular regions in the state.
4.2. Salmonella Identification and Serotyping
Lyophilized or cryopreserved isolates stored at −80 °C were utilized in this study. Following a standard approach for samples submitted to CAHFS, all isolates had previously been isolated and identity confirmed by conventional aerobic culture and classical serotyping methods. Following the recovery of preserved isolates, Salmonella sp. identification was confirmed by matrix-assisted laser desorption–ionization mass spectrometry (MALDI-TOF; Bruker Daltonics, Fremont, CA) and serovar was confirmed using the Luminex nucleic acid bead-based suspension array and xMAP® Salmonella serotyping assay (Luminex; Austin, TX, USA).
4.3. Antimicrobial Susceptibility Testing
In vitro MIC determination was performed in accordance with criteria provided in Clinical and Laboratory Standards Institute (CLSI) documents [
18,
37] using the Bovine/Porcine (BOPO6F) and the NARMS Gram-negative panels (Thermo Fisher Scientific, Sensititre). Antimicrobial resistance interpretation was defined using the Veterinary CLSI-defined (when available for BOPO6F panel drugs) or NARMS consensus breakpoints [
17,
18]. Drugs tested on the BOP6F panel include: ceftiofur (0.25–8 µg/mL), tiamulin (0.5–32 µg/mL), chlortetracycline (0.5–8 µg/mL), gentamicin (1–16 µg/mL), florfenicol (0.25–8 µg/mL), oxytetracycline (0.5–8 µg/mL), penicillin (0.12–8 µg/mL), ampicillin (0.25–16 µg/mL), danofloxacin (0.12–1 µg/mL), sulphadimethoxine (256 µg/mL), neomycin (4–32 µg/mL), trimethoprim/sulfamethoxazole (2/38 µg/mL), spectinomycin (8–64 µg/mL), tylosin (0.5–4 µg/mL), tulathromycin (1–64 µg/mL), tilmicosin (4–64 µg/mL), clindamycin (0.25–16 µg/mL), and enrofloxacin (0.12–2 µg/mL). Drugs tested on the NARMS panel include: cefoxitin (0.5–32 µg/mL), azithromycin (0.125–16 µg/mL), chloramphenicol (2–32 µg/mL), tetracycline (4–32 µg/mL), ceftriaxone (0.25–64 µg/mL), amoxicillin/clavulanic acid (1/0.5–32/16 µg/mL), ciprofloxacin (0.015–4 µg/mL), gentamicin (0.25–16 µg/mL), nalidixic acid (0.5–32 µg/mL), ceftiofur (0.12–8 µg/mL), sulfisoxazole (16–256 µg/mL), trimethoprim/sulfamethoxazole (0.12/2.38–4/76 µg/mL), ampicillin (1–32 µg/mL), and streptomycin (2–64 µg/mL). Quality control testing was performed on
Escherichia coli ATTC 25922,
Pseudomonas aeruginosa ATTC 27853,
Enterococcus faecalis ATTC 29212, and
Staphylococcus aureus ATTC 29213.
4.4. Statistical Analysis
For all statistical analyses conducted, clindamycin, penicillin, sulphadimethoxine, tiamulin, tilmicosin, trimethoprim sulfamethoxazole, and tylosin in the BOPO6F panel were not included because their MIC distribution had more than 95% of isolates within one same dilution distribution or the MIC distributions were within less than 3 dilutions. Furthermore, for the BOPO6F panel, only 2 drugs could be classified using the SIR systems either because the MIC breakpoints were not contained within the range of concentrations tested or because cattle-specific clinical breakpoints for the drug and organism combination are lacking. The two drugs in the BOPO6F panel for which SIR classification was conducted were ampicillin and ceftiofur.
A table with MIC distribution for each antimicrobial was created for each drug in BOPO6F and NARMS panel. For the BOPO6F panel, sulfadimethoxine, trimethoprim sulfamethoxazole, and tylosin were not included because all of their isolates were within one MIC distribution, 246, 2, and 32 µg/mL, respectively.
Epidemiological cut-off points (ECOFFs) were calculated with three endpoint criteria (95%, 97.5%, and 99%) using ECOFFINDER version 2.1, which is based on the methodology described by Turnidge et al. (2006) [
19]. Using ECOFFS, isolates with a MIC above ECOFF were categorized as resistant and those with and MIC below the ECOFF were categorized as susceptible. Assumptions and standard for using ECOFFINDER were not fulfilled by every antimicrobial tested and because of that only select ECOFFS were determined for select antimicrobial drugs [
38]. Some of the standards not fulfilled, resulting in the exclusion of antimicrobial forms using ECOFFS, include that the necessity of a single peak in the putative wild-type MIC distribution and a wild-type MIC distribution that follow a log-normal distribution 3 to 5 two-fold dilutions wide.
Multiple Pearson chi-squared tests were first used to evaluate the association between the descriptive variables (year group, region, clinical signs, source group, and age group) and the binomial variable for each S. Dublin isolate based on their classification as resistant to the referred antimicrobial drug based on CLSI breakpoints for each antimicrobial drug in the NARMS panel. No significant association between categorical variables and the azithromycin, sulfisoxazole, and trimethoprim sulfamethoxazole was observed, and they were excluded from further analysis. Following this, for each antimicrobial drug a logistic regression model in SAS was used with the antimicrobial drug resistance classification as the dependent variable and descriptive variables with a p value less than 0.1 from the Chi square analysis being offered to the model as explanatory variables. Variables with a p value greater than 0.05 were removed from the model. The Wald statistics and the Akaike information criterion (AIC) were also used for model selection and to assure a more parsimonious model was selected. The goodness of fit of each model was evaluated by performing the Hosmer–Lemeshow test. The odds ratio was calculated in the model to evaluate the odds for isolating a S. Dublin isolate resistant to the referred drug according to different response levels of the descriptive variables. Pairwise comparisons for the least square means of the proportion of resistance isolated to relevant variables were conducted, adjusting for multiple comparisons using the Tukey–Kramer approach.