Skip to main content
Download PDF

Patterns of antimicrobial resistance in Salmonella isolates from fattening pigs in Spain

BMC Veterinary Research volume 18, Article number: 333 (2022) Cite this article

Abstract

Background

Swine are considered a major source of foodborne salmonellosis, a public health issue further complicated by the circulation of multidrug-resistant Salmonella strains that threaten the safety of the food chain. The current study aimed to identify patterns that can help to understand the epidemiology of antimicrobial resistance (AMR) in Salmonella in pigs in Spain through the application of several multivariate statistical methods to data from the AMR national surveillance programs from 2001 to 2017.

Results

A total of 1,318 pig Salmonella isolates belonging to 63 different serotypes were isolated and their AMR profiles were determined. Tetracycline resistance across provinces in Spain was the highest among all antimicrobials and ranged from 66.7% to 95.8%, followed by sulfamethoxazole resistance (range: 42.5% − 77.8%), streptomycin resistance (range: 45.7% − 76.7%), ampicillin resistance (range: 24.3% − 66.7%, with a lower percentage of resistance in the South-East of Spain), and chloramphenicol resistance (range: 8.5% − 41.1%). A significant increase in the percentage of resistant isolates to chloramphenicol, sulfamethoxazole, ampicillin and trimethoprim from 2013 to 2017 was observed. Bayesian network analysis showed the existence of dependencies between resistance to antimicrobials of the same but also different families, with chloramphenicol and sulfamethoxazole in the centre of the networks. In the networks, the conditional probability for an isolate susceptible to ciprofloxacin that was also susceptible to nalidixic acid was 0.999 but for an isolate resistant to ciprofloxacin that was also resistant to nalidixic acid was only 0.779. An isolate susceptible to florfenicol would be expected to be susceptible to chloramphenicol, whereas an isolate resistant to chloramphenicol had a conditional probability of being resistant to florfenicol at only 0.221. Hierarchical clustering further demonstrated the linkage between certain resistances (and serotypes). For example, a higher likelihood of multidrug-resistance in isolates belonging to 1,4,[5],12:i:- serotype was found, and in the cluster where all isolates were resistant to tetracycline, chloramphenicol and florfenicol, 86.9% (n = 53) of the isolates were Typhimurium.

Conclusion

Our study demonstrated the power of multivariate statistical methods in discovering trends and patterns of AMR and found the existence of serotype-specific AMR patterns for serotypes of public health concern in Salmonella isolates in pigs in Spain.

Peer Review reports

Introduction

The emergence of antimicrobial resistance (AMR) causes a substantial burden on human health and has a strong impact on the economy. Up to 33,000 deaths are attributed or related to multidrug-resistant (MDR, usually defined as resistance to 3 or more antimicrobial classes [1]) bacterial infections every year in Europe, costing 1.5 billion Euros [2]. This phenomenon is well-characterized in the case of non‐Typhoidal Salmonella [3, 4], one of the main foodborne zoonoses worldwide [5], for which resistant isolates have been associated with more severe conditions and higher mortality [6, 7].

Resistant non‐Typhoidal Salmonella strains that show multidrug-resistant phenotypes, such as Typhimurium phage type DT104 (characterized by penta-resistance to ampicillin, chloramphenicol, streptomycin, sulfonamides and tetracycline ─ ACSSuT), are of special concern from a public health perspective [8]. Many phenotypic resistant patterns (resistotypes) have been reported in Salmonella, with their distribution and prevalence changing over time and space [9,10,11]. The interplay between Salmonella serotypes and resistotypes further complicates the picture, since the serotype of Salmonella has been shown to be associated with the pathogenicity, virulence, host range, and, most relevant here, AMR profile [12,13,14], and certain serotypes tend to show higher levels of resistance. For example, the proportion of MDR strains among Derby human clinical isolates in Europe in 2017 was lower than among Typhimurium and 1,4,[5],12:i:- [12]. Liao et al., (2019) also suggested serotype-specific evolutionary patterns of AMR in Typhimurium, Dublin, and Newport [14].

In Spain, approximately 35% of foodborne salmonellosis cases have been attributed to pigs [15], and a high proportion of non‐Typhoidal Salmonella isolates recovered from fattening pig carcases and caecal contents showed a multidrug-resistant phenotype [12]. To enable effective control programmes aiming at reducing multidrug resistance in swine-originated Salmonella, it is essential to understand factors related to the presence of resistance and co-resistance. However, there has been limited research on the distribution of AMR profiles of swine Salmonella in Spain. One study, using faecal samples from fattening pigs across Spain in 2003, reported the resistoypes of 290 Salmonella isolates to 17 antimicrobials, with a high percentage of MDR Typhimurium, including ACSSuT Typhimurium, and 4,5,12:i:-[9].

Here, we aimed to provide information about the phenotypic AMR profiles found in Salmonella isolates recovered from pigs in Spain from 2001 to 2017 through a nationwide sampling system. We investigated (a) the temporal and spatial trends of resistance to individual antimicrobials, (b) relationships between resistance to different antimicrobials and (c) resistotypes, their associations with common serotypes in pigs in Spain and their spatial distribution. The epidemiological evidence generated by the current study is expected to help to identify important relationships between resistance to specific antimicrobials and to shed some light on the underlying mechanisms of the presentation of resistotypes that could help to design preventive and mitigation measures.

Methods

Study population and sample collection

The VISAVET Health Surveillance Centre in Madrid has been involved in the national antimicrobial resistance surveillance for Salmonella detection in food-producing pigs, commissioned by the Spanish Ministry of Agriculture, Food and Fisheries, during the study period. Isolates were recovered from faecal samples from fattening pigs collected at abattoirs of high slaughter capacity (7–20 abattoirs located in ≥ 50% of the provinces in Spain each year) and represented independent epidemiological units (i.e., farms). More details on the sampling strategy can be found in Teng et al., 2020 [16].

Information about Salmonella isolates that were collected between 2001 and 2017 and underwent antimicrobial susceptibility testing (AST) such as the date of sampling, the province and the autonomous community of the originated farm of the sampled pigs, the serotype, and AST results were included in the current study. Bacteriology methods can be found in Teng et al., 2020 [16]. Briefly, isolation of Salmonella was performed according to ISO 6579:2002/Amd 1:2007 before 2017 and ISO 6579–1:2017 for 2017 samples. Serological typing was conducted based on the White-Kauffmann-Le Minor scheme [17]. Minimum inhibitory concentrations (MICs) for various antimicrobials for isolates were determined using the two-fold broth microdilution reference method, according to ISO 20776–1:2006. For each of the tested antimicrobials, a binary outcome was constructed based on the EUCAST epidemiological cut-off value (ECOFF) as suggested by The European Committee on Antimicrobial Susceptibility Testing (EUCAST; accessed on 2020 Mar 17) [18]. More specifically, resistant (i.e., non-wild-type) isolates have a MIC larger than the ECOFF, while susceptible (i.e., wild-type) isolates have a MIC smaller than or equal to the ECOFF. The ECOFFs, as well as the ranges for each antimicrobial and year, are now provided in Supplementary File 1. Although the ranges varied slightly across the study period, this did not affect the interpretation since the ECOFFs were always within the tested ranges.

Statistical analyses

Data process

Data cleaning, management and analyses were conducted in Microsoft Excel 2013 (Microsoft Corp.) and R programme version 3.5.2 (R Core Team) in RStudio interface version 1.2.1330 (RStudio Team) [19, 20].

Eleven antimicrobials for which AST results were available during most of the study period were considered in the analysis. Because AST was conducted for more antimicrobials in 2008 − 2017, the data were grouped into two datasets: one with more years and fewer antimicrobials being tested, and the other with more antimicrobials being tested and fewer years (Table 1). The first dataset (D_am7) included the AST results for seven antimicrobials [tetracycline (TET), chloramphenicol (CHL), ciprofloxacin (CIP), nalidixic acid (NAL), gentamicin (GEN), florfenicol (FFC), and cefotaxime (CTX)] from 2001 to 2013. The second dataset (D_am10) contained the AST results for the aforementioned antimicrobials except for FFC, as well as sulfamethoxazole (SMX), ampicillin (AMP), trimethoprim (TMP) and ceftazidime (CAZ), which were tested in isolates collected from 2008 to 2013 and 2017 (no sampling was conducted in swine in 2014 and 2016, and no results were available for 2015). The year isolates were recovered and their serotype was also included in the datasets. D_am7 and D_am10 datasets were analysed separately except in the analyses for spatial trends, which were conducted for individual antimicrobials using all data across the years [and including an additional antimicrobial, streptomycin (STR)]. If not otherwise specified, isolates with missing values on the AST results were removed from the analyses.

Table 1 Number of Salmonella isolates from pigs collected through the Spanish Veterinary Antimicrobial Resistance Surveillance Network programme that were tested for susceptibility to each of the eleven antimicrobials from 2001 to 2017
Full size table

Descriptive analysis

Summary and descriptive analyses to determine the resistotypes and their relative frequencies were facilitated by the ‘tidyverse’ package [21]. This was also performed for the top four most common serotypes in pigs in the dataset, namely Typhimurium, 1,4,[5],12:i:-, Derby and Rissen, and Venn diagrams were built to explore the similarity between their resistotypes.

Spatial and temporal trends

The proportion of isolates with information on their geographical origin that was resistant to each of the 12 antimicrobials at the province level was calculated and later adjusted using empirical Bayesian smoothing with the ‘spdep’ package [22]. For the empirical Bayesian smoothing, the neighbouring relationships were described by Queen contiguity, considering two provinces as neighbours when they shared at least one point of their boundaries.

With the aim of jointly modelling the evolution over time of the proportion of resistant strains for each of the antimicrobials, generalized linear models were constructed, denoting by \({{\varvec{\pi}}}_{i}= {\pi }_{i1},\dots\), \({\pi }_{iJ}\), the probability for isolate \(i\) to be resistant to antimicrobials 1…\(J\). The considered model has the following functional form:

$$\text{logit(}{\pi }_{ij}\text{)}={{\varvec{x}}}_{{\varvec{i}}{\varvec{j}}}{\varvec{\beta}}$$

To account for the fact that AST results on the same isolate are not independent, the parameter estimates were not obtained through the classical maximum likelihood theory. Rather, the generalized estimating equations (GEE) approach was followed to incorporate the unknown correlation between the outcomes, facilitated by “gee” package [23]. The unstructured working correlation was chosen in order not to pose any restrictions on the underlying relations. An exhaustive search to obtain the best model describing the temporal trends for the antimicrobials was conducted for each dataset. The search considered the inclusion of an interaction between an antimicrobial and the linear year term and if the interaction held, the interaction with the quadratic year term. In total, 2,186 and 59,045 possible models were built for D_am7 and D_am10, respectively, and the model with the lowest Quasi Information Criterion (QIC) was selected as the best model for each dataset using “MuMIn” package [24, 25].

Bayesian network analysis

Bayesian networks were built for binary AST results using the R package ‘bnlearn’ [26]. Bayesian networks are graphical models where vertices represent random variables and arcs represent probabilistic dependencies between them [27]. Hill-Climbing greedy search was applied to identify directed network structure. Arcs were retained in the networks if their empirical frequency in 100,000 bootstrap samples was ≥ 40%. Four Bayesian networks were built for D_am7 with (a) all data, (b) data from 2001 to 2004, (c) data from 2005 to 2008, and (d) data from 2009 to 2013 to assess if the structure of the network changed over time. Similarly, for D_am10, three were constructed using (a) all information, (b) data from 2008 to 2010, and (c) data from 2011 to 2017, respectively.

Hierarchical clustering

Hierarchical clustering analysis was carried out for both binary logarithms of the MICs and binary AST results by using R packages ‘FactoMineR’, ‘factoextra’ and ‘missMDA’ [28,29,30]. The regularised iterative principal components analysis algorithm and the regularised iterative multiple correspondence analysis algorithm were used to impute missing AST results (2 isolates with one missing AST result each) and serotype information (16 isolates) in the numeric and binary data, respectively [30]. Hierarchical trees were constructed using Ward’s minimum variance method based on Euclidean distances [31], and the number of clusters was determined when the partition had the highest relative loss of inertia [32]. The composition of binary AST results of the antimicrobials and the serotype of the isolates were elicited for each cluster. The percentage of isolates of a cluster belonging to each specific province was depicted on the map, and the temporal trends in the percentage were also generated.

Results

Descriptive results

A total of 1,318 pig Salmonella isolates belonging to 63 different serotypes that were retrieved between 2001 and 2017 were included in at least one of the datasets (Table 1). The yearly number of isolates ranged from 40 to 211. D_am7 consisted of 1,154 isolates from 58 serotypes, of which 23.5% were Typhimurium, 21.2% were Rissen, 16.6% were Derby, 10.1% were 1,4,[5],12:i:-, 5.2% were Anatum, and 4.7% were Bredeney. Among the 1,154 isolates, 318 (27.6%) did not have geographical information. D_am10 included 680 isolates from 46 serotypes of which the most frequent were Rissen (23.9%), 1,4,[5],12:i:- (21.1%), Typhimurium (19.9%), Derby (14.1%), Anatum (2.5%) and Bredeney (2.5%). A total of 281 (41.3%) isolates had missing geographical information in D_am10. Trends in serotypes have been described elsewhere [16].

Twenty-nine out of the 256 (27) possible resistotypes in D_am7 were observed, (including 21 present in Typhimurium, 17 in Rissen, 12 in Derby, and 12 in 1,4,[5],12:i:- isolates) (Fig. 1). Almost 15% (n = 170) of the isolates were susceptible to all seven antimicrobials (Supplementary File 2). Isolates with a MDR resistotype (i.e., resistance to 3 or more antimicrobial classes) accounted for 8.9% (n = 103) of all isolates, and the annual percentage of MDR isolates decreased from 30.1% (n = 22) in 2001 to only 1.6 (n = 1) in 2008 (Table 2). However, it then increased to 7.2 − 10.4% in 2011 − 2013. Among the 607 (52.7%) isolates that were resistant to only one antimicrobial, 592 (51.4%) were resistant to TET, and 9 (0.8%) and 6 (0.5%) were resistant to CHL and CIP, respectively. In D_am10, 65 resistotypes were observed among the 1024 (210) possible combinations of resistances (Supplementary File 2). There were 21 resistotypes in Typhimurium isolates, 30 in Rissen isolates, 15 in Derby isolates, and 25 in 1,4,[5],12:i:- isolates (Fig. 1). More than half of the resistotypes identified among Rissen and 1,4,[5],12:i:- isolates were specific to that serotype. Around 15% of the isolates (n = 103) were susceptible to all the antimicrobials, which is similar to the proportion in D_am7. However, there was a higher percentage of MDR isolates (54.0%, n = 366) due to the inclusion of two antimicrobials (SMX and AMP) against which high levels of resistance were found, and an increasing trend in the annual proportion of MDR isolates was observed (from 37.5% in 2008 to 62.8% in 2017; Table 2). Most isolates that were resistant to TET were also resistant to SMX and/or AMP, and just 14.2% (n = 96) were only resistant to TET.

Fig. 1
figure 1

Venn diagrams illustrating the number of resistotypes in Salmonella isolates of serotypes Typhimurium, Rissen, Derby, and 1,4,[5],12:i:- from pigs recovered through the Spanish Veterinary Antimicrobial Resistance Surveillance Network programme. (a) contains 1,154 isolates with seven antimicrobial susceptibility results between 2001 and 2013 (D_am7); (b) contains 680 isolates with ten antimicrobial susceptibility results between 2008 and 2017 (D_am10)

Full size image
Table 2 Percentage of multidrug-resistant Salmonella isolates (the number of multidrug-resistant Salmonella isolates/number of all Salmonella isolates) by year from pigs collected through the Spanish Veterinary Antimicrobial Resistance Surveillance Network programme that were tested for susceptibility to eleven antimicrobials from 2001 to 2017
Full size table

Spatial trends

The empirically adjusted spatial distribution of the proportion of isolates resistant to each of the 12 antimicrobials is shown in Fig. 2 and Supplementary File 3. The percentage of isolates resistant to TET across Spain was the highest among all antimicrobials and ranged from 66.7% to 95.8%, followed by SMX (range: 42.5% − 77.8%), STR (range: 45.7% − 76.7%), AMP (range: 24.3% − 66.7% with a lower percentage of resistance in the South-East of Spain), and CHL (range: 8.5% − 41.1%). A higher proportion of isolates were resistant to TMP (range: 7.1% − 58.3%) in the southwest of Spain. The percentage of resistant isolates for CIP (range: 4.3% − 19.3%) and NAL (range: 3.1% − 18%) was lower in the northwest corner of Spain; in contrast, the value for FFC (range: 0% − 20.0%) was higher there.

Fig. 2
figure 2

Spatial trends adjusted by empirical Bayesian smoothing in the proportion of Salmonella isolates from pigs resistant to twelve antimicrobials from 2001 to 2017, collected through the Spanish Veterinary Antimicrobial Resistance Surveillance Network programme. Provinces in grey indicate where there was no isolate

TET tetracycline, CHL chloramphenicol, CIP ciprofloxacin, NAL nalidixic acid, GEN gentamicin, FFC florfenicol, CTX cefotaxime, SMX sulfamethoxazole, AMP ampicillin, TMP trimethoprim, CAZ ceftazidime, STR streptomycin

Full size image

Temporal trends

Detailed model results and work correlations can be found in Supplementary File 4.

D_am7

A significant tempeoral change in the percentage of resistant isolates during 2001–2013 was found for all the antimicrobials except NAL. While the change for CHL was linear, a quadratic year term was included in the final models for TET, CIP, GEN, FFC and CTX (Fig. 3). According to the GEE estimates, a much higher percentage of isolates resistant to TET was expected (lowest in 2009 at 78.8% and highest in 2001 at 92.1%) than to other antimicrobials (median values <23.9%). From 2001 to 2013, the percentage of resistant isolates for TET, CIP and GEN first decreased and then increased again, while the odds of being resistant to CHL decreased by 8.9% [95% confidence interval (CI): 5.1 − 12.5%] each year. The predicted percentage of CTX and FFC resistant isolates remained below 10% for the most part.

Fig. 3
figure 3

Temporal trends in the percentage of resistant Salmonella isolates from pigs collected through the Spanish Veterinary Antimicrobial Resistance Surveillance Network programme towards seven antimicrobials from 2001 to 2013. Red dots are the observed percentage of resistant isolates; black lines are the fitted values of generalized estimating equation models using the binary results (i.e., resistant and susceptible)

TET tetracycline, CHL chloramphenicol, CIP ciprofloxacin, NAL nalidixic acid, GEN gentamicin, FFC florfenicol, CTX cefotaxime

Full size image

D_am10

A relatively high level of resistance to TET, SMX, AMP (median values >53.7%) compared to other antimicrobials (with median values <27.1%) was shown in the GEE predicted results. The GEE model found temporal changes in the percentage of resistant isolates between 2008 to 2017 for all the antimicrobials except TET, NAL and CAZ (Fig. 4), but the variation in the observed percentage of resistant isolates for TET was rather substantial. The percentage of isolates resistant to CIP, GEN, AMP increased over this period, and the odds of isolates being resistant to AMP was 1.12 times (95% CI: 1.08 − 1.17) higher than the previous year. Although the percentage of resistant isolates to CHL, SMX and TMP seemed to decrease from 2008 to 2013, the values in 2017 increased significantly. These observed and predicted figures for CTX and CAZ have remained below 10% over the years.

Fig.4
figure 4

Temporal trends in the percentage of resistant Salmonella isolates from pigs collected through the Spanish Veterinary Antimicrobial Resistance Surveillance Network programme towards ten antimicrobials from 2008 to 2017. Red dots are the observed percentage of resistant isolates; black lines are the fitted values of generalized estimating equation models using the binary results (i.e., resistant and susceptible)

TET tetracycline, CHL chloramphenicol, CIP ciprofloxacin, NAL nalidixic acid, GEN gentamicin, CTX cefotaxime, SMX sulfamethoxazole, AMP ampicillin, TMP trimethoprim, CAZ ceftazidime

Full size image

The estimates of the pairwise correlations obtained from the GEE modelling using D_am10 are shown in Fig. 5. As expected, the correlation coefficient for CIP and NAL was the highest among all pairs (0.82), and the correlation between CTX and CAZ was also relatively high (0.49). Among antimicrobials from different classes, SMX and AMP had a high correlation coefficient of 0.71, and both had relatively high correlation coefficients with TET, CHL and TMP (range: 0.33–0.51). GEN did not appear to be correlated with any other antimicrobials.

Fig. 5
figure 5

Estimated pairwise correlations obtained from the generalized estimating equations for the binary results (i.e., resistant and susceptible) of antimicrobial susceptibility testing for ten antimicrobials

TET tetracycline, CHL chloramphenicol, CIP ciprofloxacin, NAL nalidixic acid, GEN gentamicin, CTX cefotaxime, SMX sulfamethoxazole, AMP ampicillin, TMP trimethoprim, CAZ ceftazidime

Full size image

Bayesian network analysis

The overall Bayesian networks identified for D_am7 and D_am10 are presented here (Fig. 6), and networks using data from specific periods for both D_am7 and D_am10 are shown in Supplementary File 5. The strength and the direction of the arcs in each of the networks can be found in Supplementary File 6.

Fig. 6
figure 6

Bayesian networks for (A) seven and (B) ten binary antimicrobial susceptibility testing results of Salmonella isolates from pigs collected through the Spanish Veterinary Antimicrobial Resistance Surveillance Network programme between 2001 and 2013 for (A) and between 2008 and 2017 for (B). TET: tetracycline; CHL: chloramphenicol; CIP: ciprofloxacin; NAL: nalidixic acid; GEN: gentamicin; FFC: florfenicol; CTX: cefotaxime; SMX: sulfamethoxazole; AMP: ampicillin; TMP: trimethoprim; CAZ: ceftazidime; r: resistant; s: susceptible

Full size image

For D_am7, five relations were built with CHL located in a central position. Except the relation between GEN-TET, all the others were also maintained in at least one of the partial networks (Supplementary File 5). A strong relationship between resistance to certain pairs of antimicrobials was identified in the overall network. The conditional probability for an isolate susceptible to CIP that was also susceptible to NAL was 0.999 but for an isolate resistant to CIP that was also resistant to NAL was only 0.779. An isolate susceptible to FFC would be expected to be susceptible to CHL, whereas an isolate resistant to CHL had a conditional probability of being resistant to FFC at only 0.221.

Regarding D_am10, SMX was located at the centre of the overall Bayesian network with 10 edges. Associations between CIP and NAL and between SMX and TET, CHL, AMP and TMP also persisted in all other partial networks. According to the overall D_am10 Bayesian network, the conditional probability for an isolate that was susceptible to SMX also being susceptible to CHL was 1 and to TMP was 0.972, while one that was resistant to SMX had a 0.310 probability of being resistant to CHL and a 0.413 probability of being resistant to TMP. An isolate resistant to CAZ would be expected to be resistant to CTX. Although networks of D_am7 suggested the occurrence of associations between resistance to CHL and resistance to TET and GEN, in the D_am10-based networks (which included more antimicrobials but fewer isolates), resistance to CHL became independent from TET and more dependent on the patterns for SMX and TMP which were not included in D_am7.

Hierarchical clustering

For D_am7, three and six clusters were generated using binary logarithms of the MICs and binary AST results, respectively, while the number of clusters identified for D_am10 was three and four. Here we present the hierarchical clustering results using binary AST information, and results of the analysis using binary logarithms of the MICs are shown in Supplementary File 7).

D_am7

Six clusters with 185 (16.0%), 720 (62.4%), 58 (5.0%), 61 (5.3%), 21 (1.8%) and 109 (9.4%) isolates, respectively, were identified using the binary AST results in D_am7 (Fig. 7). Isolates in Cluster (n = 185, 16.0% of isolates in D_am7) were susceptible to almost all seven antimicrobials (with a few resistant to either CHL and/or CIP) (Fig. 7). The most abundant serotype in this cluster was Bredeney (36 isolates, accounting for two-thirds of all Bredeney isolates in D_am7), and several other serotypes including Montevideo, Enteritidis or Infantis were also overrepresented (i.e., with a percentage higher than the expected value in the cluster). In contrast, the most prevalent serotypes in the D_am7 collection, including Rissen, Derby, Typhimurium and especially 1,4,[5],12:i:- were under-presented in Cluster 1 (with only between 2 and 22 isolates from each). Cluster 2 was the largest (n = 720, 62.4%) and was characterized by resistance to TET and susceptibility to all other antimicrobials except CHL (17.6% resistance, n = 127). Serotype distribution in Cluster 2 was similar to that observed in the whole D_am7 panel with Rissen (27.8%; n = 200) and Derby (20.3%; n = 146) being overrepresented and Bredeney being underrepresented (1.5%, n = 11).

Fig. 7
figure 7

Hierarchical clusters using the binary antimicrobial susceptibility testing results of seven antimicrobials among 1,154 Salmonella isolates from pigs collected through the Spanish Veterinary Antimicrobial Resistance Surveillance Network programme between 2001 and 2013 (left: the proportion of isolates resistant to seven antimicrobials; right: the composition of serotypes in each of the clusters). Cluster 0 shows the serotype distribution of all isolates in the dataset (D_am7). Only serotypes accounting for ≥5% of the isolates in each particular cluster are shown in the graph

Full size image

Resistance to both TET and GEN was the predominant feature present in all isolates in Cluster 3 (n = 58, 5.0%), with close to 60% of them being also resistant to CHL (n = 35). The most common serotypes in this cluster were 1,4,[5],12:i:- (32.8%; n = 19) and Brandenburg (22.4%; n = 13, representing 41.9% of all Brandenburg isolates in D_am7). Close to all isolates in Cluster 4 (n = 61, 5.3%) were resistant to TET, CHL and FFC and susceptible to CTX, and 85 − 92% of isolates (n = 52 − 56) were susceptible to CIP, NAL and GEN. This cluster was dominated by Typhimurium (86.9%; n = 53) isolates. Isolates in Cluster 5 (n = 21, 1.8%) were resistant to TET and CTX and susceptible to FFC, with varying resistance levels to other antimicrobials. Typhimurium accounted for 57.1% (n = 12) of the isolates in the cluster; 14.3% (n = 3) were Anatum and Rissen, respectively, and 9.5% were Derby (n = 2). Cluster 6 consisted of 109 isolates (9.4% of D_am7), including isolates that were typically resistant to CIP, NAL and TET and susceptible to GEN, FFC and CTX. Wien (9.2%; n = 10), Kapemba (8.3%; n = 9), Brikama (4.6%; n = 5) were overrepresented in Cluster 6, and the only isolates of Essen (n = 3) and Choleraesuis (n = 2) in D_am7 were in this cluster.

The spatial distribution of the proportion of isolates belonging to each of the clusters is shown in Fig. 8. While no obvious spatial patterns were observed for Clusters 1 and 2, a higher proportion of isolates of Cluster 3 was seen in the North and South of Spain. Northern provinces and provinces adjacent to the Mediterranean Sea had a larger proportion of isolates of Cluster 4. Some provinces in the Northeast and the South harbored more isolates belonging to Cluster 5. Last, the proportion of isolates of Cluster 6 seemed lower in the West of Spain, except the province Cáceres. No clear temporal trends in the proportion of isolates in different clusters were noted (Fig. 9).

Fig. 8
figure 8

Spatial distribution of six clusters elicited from hierarchical clustering using binary antimicrobial susceptibility testing results of 836 Salmonella isolates from pigs collected through the Spanish Veterinary Antimicrobial Resistance Surveillance Network programme between 2001 and 2013. Provinces in grey indicate where there was no isolate

Full size image
Fig. 9
figure 9

Temporal distribution of six clusters elicited from hierarchical clustering using binary antimicrobial susceptibility testing results of 836 Salmonella isolates from pigs collected through the Spanish Veterinary Antimicrobial Resistance Surveillance Network programme between 2001 and 2013

Full size image

D_am10

Four clusters with 227 (33.4%), 366 (53.8%), 84 (12.4%) and 3 (0.4%) isolates, respectively, were identified using the binary AST results in D_am10. In Cluster 1, half of the isolates were resistant to TET, and most were susceptible to the rest of the antimicrobials. Compared to the percentage of isolates of different serotypes in the whole D_am10, Cluster 1 had higher percentage of Rissen (30.8%; n = 70) and Derby (20.3%; n = 46) and much lower percentage of Typhimurium (6.6%; n = 15) and 1,4,[5],12:i:- (4.8%; n = 11) (Fig. 10). All the Montevideo (n = 7) and Infantis (n = 5) isolates and 3 out of the 5 Enteritidis isolates in D_am10 belonged to this cluster. Most of the isolates in Cluster 2 were resistant to AMP, SMX and TET, and susceptible to CIP, NAL, GEN, CTX, and CAZ. This cluster was dominated by 1,4,[5],12:i:- (n = 122; 33.3%) and Typhimurium (n = 96; 26.2%), and more than 90% of the Wien isolates (n = 10) were in this cluster. Cluster 3 was different from Cluster 2 in the addition of very high (> 84.5%) levels of resistance to CIP and NAL. The percentage of Rissen, Derby and Typhimurium was similar to the one of Cluster 2, but 1,4,[5],12:i:- only accounted for 11.9% (n = 10) of the isolates. All the Kapemba isolates (n = 6) in D_am10 belonged to this cluster. Finally, Cluster 4, which included only three isolates (all from different serotypes), was mainly characterized by full resistance to CAZ.

Fig. 10
figure 10

Hierarchical clusters using the binary antimicrobial susceptibility testing results of ten antimicrobials among 680 Salmonella isolates from pigs collected through the Spanish Veterinary Antimicrobial Resistance Surveillance Network programme between 2011 and 2017 (left: the proportion of isolates resistant to ten antimicrobials; right: the composition of serotypes in each of the clusters). Cluster 0 shows the serotype distribution of all isolates in the dataset (D_am10). Only serotypes accounting for ≥5% of the isolates in each particular cluster are shown in the graph

Full size image

Figure 11 shows the proportion of isolates originating from different provinces in Spain in different clusters. Although there were no isolates from many of the provinces in the West, some patterns might be observed. A higher proportion of isolates in Cluster 1 was in the East than in the West, which seemed to be the opposite for Cluster 2. The proportion of isolates in Cluster 3 was higher in the far East and the far West of Spain, and all the isolates in Cluster 4 were in a single province of the west. The temporal trends in the proportion of isolates in the clusters seemed to oscillate over the period (Fig. 12).

Fig. 11
figure 11

Spatial distribution of six clusters elicited from hierarchical clustering using binary antimicrobial susceptibility testing results of 399 Salmonella isolates from pigs collected through the Spanish Veterinary Antimicrobial Resistance Surveillance Network programme between 2011 and 2017. Provinces in grey indicate where there was no isolate

Full size image
Fig. 12
figure 12

Temporal distribution of six clusters elicited from hierarchical clustering using binary antimicrobial susceptibility testing results of 399 Salmonella isolates from pigs collected through the Spanish Veterinary Antimicrobial Resistance Surveillance Network programme between 2011 and 2017

Full size image

Discussion

The current study demonstrates the usefulness of a range of methods that can be easily applied to transform phenotypical AST data into information helpful in assessing the evolution of AMR over time, including temporal and spatial trends, and pairwise and multiple associations [33]. Some studies have previously presented multidrug-resistant patterns for Salmonella isolates recovered from pigs in (part of) Spain and elsewhere [9, 11, 34,35,36,37]. Although some AMR studies using multi-sourced Salmonella isolates have applied some multivariate analyses to explore the relations between the isolates [38, 39], to our best knowledge, reports of longitudinal AMR data from swine Salmonella isolates in Spain using multivariate analyses to explore links between resistances to specific antimicrobials were missing.

Two multivariate analyses, Bayesian network analysis and hierarchical clustering were conducted to investigate potential relationships between and patterns among the resistances to antimicrobials, whose existence was also demonstrated by the pairwise correlations from the GEE. Bayesian network analysis provides conditional probabilities of resistance to an antimicrobial given one or more resistances to other antimicrobials. This allows to uncover co-resistance patterns among the antimicrobials of interest efficiently and to discern the degree to which the resistance of an antimicrobial is associated with another one. However, although directionality was indicated in our Bayesian networks, causality should not be implied. Changes in Bayesian network structures and conditional probabilities between or among antimicrobials may indicate shifts in mechanisms of co-resistance among antimicrobials over time or space. Some of these changes were observed in the networks built with subsets of data of different periods while certain patterns were more stable. In any case, the volume of our data might have been insufficient to exhibit meaningful temporal trends. On the other hand, hierarchical clustering provides a rather straightforward way to explore common resistotypes whose spatial and temporal trends and dominant serotypes can be subsequently inspected. When this method was applied to both the binary logarithm of MIC and binary AST results, the results showed that, although the optimal number of clusters found might be different, the composition of clusters was similar if the number of clusters was the same. For example, the composition of three clusters generated using both types of data of D_am10 was very alike. However, when the number of clusters increased to four for D_am10, a CAZ-resistant cluster and a GEN-AMP resistant cluster were found using the binary logarithm of MICs and the binary AST results, respectively.

In our data, a large proportion of the isolates (>53.7%) were resistant to TET, SMX or AMP, fewer isolates (9.4 − 27.1%) were resistant to CHL, CIP, NAL or TMP, and very few isolates (<6.9%) were resistant to GEN, FFC, CTX, or CAZ. The resistance profile of Salmonella revealed by the current study was notably different from the ones in other countries in Europe [12], the United States [40] and China [41], although TET resistance seemed to be the most prominent in all the countries. There was a 4-year gap between 2013 and 2017, and the percentage of resistant isolates to some antimicrobials such as SMX, AMP and TMP increased substantially during this period. It is not clear whether this increase reflects the true situation in the field as the results of 2015 and 2019 were unavailable to us (The Spanish Veterinary Antimicrobial Resistance Surveillance Network programme started to survey AMR in Salmonella in pigs every odd year since 2013). When comparing with the AST results of the isolates from clinical human salmonellosis cases in Spain from 2010 to 2017, the resistance levels to all antimicrobials except FFC (not tested in human isolates), CTX, and CAZ were substantially higher [42,43,44,45,46,47,48]. Still, similar temporal trends in the percentage of resistant isolates from swine (this study) and humans [42,43,44,45,46,47,48] were observed for TET and CHL (decreasing trend) and CIP (increasing trend) but not for other antimicrobials, which may suggest the existence of other sources of foodborne salmonellosis for the general public in Spain in which antimicrobial resistance dynamics may be different.

The values of and trends in the percentage of isolates with MDR resistance varied substantially between D_am7 and D_am10, which could be explained by the inclusion of SMX and AMP, two antimicrobials in which high levels of resistance were observed in D_am10. Except in 2008, the proportion of MDR isolates in D_am10 was always between 50–60% (Table 2), while a clear decreasing trend in the percentage of clinical human MDR isolates (about 50% in 2012 to about 25% in 2017) has been reported in the same time frame [42,43,44,45,46,47].

The relationships revealed by Bayesian networks largely agreed with the clustering patterns found by hierarchical clustering and were expected in certain cases, such as the CIP and NAL, CHL and FFC, and CTX and CAZ links. Interestingly, although different antimicrobials were used in D_am7 and D_am10, similar patterns were still discovered by hierarchical clustering. These included a cluster of mostly pan-susceptible isolates (although about half of the isolates were resistant to TET in this D_am10 cluster), a cluster of isolates resistant to CIP and NAL (and TET) and a cluster of isolates with resistance to GEN (and TET and to a lesser extent CHL) (for D_am10, this pattern was shown in the results using binary AST information with a pre-set cluster number). Clusters of resistance to CHL and FFC (which was not in D_am10) and to CTX were also observed in D_am7, and a cluster with isolates resistant to CAZ (which was not in D_am7) was identified in D_am10. Antimicrobials that did not show clear patterns in the hierarchical clusters─ CHL in D_am7 and SMX, AMP, and TMP in D_am10─ were located in a more central position in the Bayesian networks.

The strong linkage between resistance to CIP and NAL (found in all analyses carried out) was independent of the resistance to other antimicrobials according to the Bayesian network analyses, and isolates resistant to CIP and NAL were classified in the same cluster in both D_am7 (Cluster 6) and D_am10 (Cluster 4). These results indicate that mechanisms of the resistance of CIP and NAL are independent of the resistance to other antimicrobials, which points out the existence of chromosomal mutations in the quinolone resistance determinant regions (not related to the horizontal transfer of resistance determinants along with resistance genes to other antimicrobial families). However, a small proportion (22.1%) of NAL-susceptible or CIP-resistant isolates were also found. This atypical quinolone resistance phenotype has been linked to the carriage of plasmid‐mediated quinolone resistance genes [49, 50]. The association in the presentation of fluoroquinolone resistance with resistance to aminoglycosides (as suggested in the Bayesian network for D_am10) and β-lactams and sulfonamides (as indicated in the clustering analysis) could suggest the involvement of mechanisms other than chromosomal mutations in at least a proportion of the strains, although this hypothesis should be confirmed using molecular tools such as whole-genome sequencing.

A strong correlation between the resistance phenotype of FFC and CHL, both belonging to the phenicol family, was also found as expected, and could be mediated by genes conferring resistance to both antimicrobials such as the floR gene. However, approximately 80% of the CHL-resistant genes were susceptible to FFC, suggesting the presence of other genes whose presence do not result in resistance to FFC, including cat and cmlA [10, 51,52,53].

In D_am10, resistotypes with a core pattern of TET, SMX, and AMP were most common in Typhimurium (83.0%) and its monophasic variant (78.3%) (and this could also be true for isolates in D_am7 for which SMX and AMP were not tested). This could indicate isolates carried the resistance background typical of DT104 as, among AMP-SMX-TET resistant isolates that were also tested for STR, 93.5% were also resistant to STR. The AMP-SMX-TET resistance pattern was also found in more than 40% of the Rissen isolates, in agreement with previous AMR results of Salmonella from pigs in Spain from 2003 to 2004 [9]. A similar AMP-CHL-SMX-TET-TMP pattern has been also described as very prevalent among Derby isolates from poultry [54], while the majority of the core resistant pattern of our Derby isolates was TET-SMX, thus suggesting the circulation of different strains/resistance genes in the different host species as previously suggested for Derby isolates circulating in pig and poultry populations in France [55].

The results of hierarchical clustering demonstrate the existence of serotype-specific differences in the presentation of several resistance patterns: For example, 1,4,[5],12:i:- isolates were more likely associated with MDR profiles, with no isolate presenting pan-susceptible profiles and most (92.3%) of the isolates in D_am10 being classified in highly resistant clusters (e.g., Cluster 2 and 3). In contrast, all Montevideo isolates were pan-susceptible, and most 65 – 100%) of the isolates belonging to serotypes Enteritidis, Infantis and Bredeney were classified in the “susceptible” clusters in D_am7 (Cluster 1) and D_am10 (Cluster 1). These results provide directions for further exploration of patterns of AMR within isolates of specific serotypes.

There are some limitations to the current study. First, due to the difference in the antimicrobials tested, we could not analyse the data as a whole nor every test result. Second, as the panels used for AST did not remain the same across the years, the range of inhibitory concentrations of some antimicrobials changed. This made the results of hierarchical clustering using the binary logarithm of MIC less reliable. Also, the graphical information of some Salmonella isolates was missing, compromising the reliability of the results of the spatial analysis. Furthermore, since samples were collected at abattoirs of high slaughter capacity (adding up to >50% of the national slaughter capacity), the results might not be representative of the farms that did not usually work with these abattoirs. Nevertheless, this strategy is in fact in line with the EU regulations regarding sampling strategies for monitoring and reporting of AMR in zoonotic and commensal bacteria, and the inclusion of a large number of samples throughout the study period and the consistency in the tests performed for at least several periods allowed the detection of certain robust trends in our collection.

Conclusion

Our study demonstrated the power of multivariate statistical methods such as Bayesian network analysis and hierarchical clustering in combination with spatial and temporal analyses applied to phenotypical AMR data for generating valuable insights about patterns of and associations between AMR phenotypes. Besides spatial and temporal trends in the percentage of isolates being resistant to various antimicrobials, we also found pairwise relationships between antimicrobials and patterns of resistotypes. The existence of serotype-specific AMR patterns for serotypes of public health concern in Salmonella isolates in pigs in Spain allows us to draw hypotheses on their possible genetic background. Future research should focus on testing these hypotheses using highly discriminatory molecular tools to decipher the relationship between strains and resistance patterns found in Salmonella from different origins.

Availability of Data and Material

The individual-level data is provided as Supplementary File 8.

References

  1. Magiorakos A-P, Srinivasan A, Carey RB, Carmeli Y, Falagas ME, Giske CG, et al. Multidrug-resistant, extensively drug-resistant and pandrug-resistant bacteria: an international expert proposal for interim standard definitions for acquired resistance. Clin Microbiol Infect. 2012;18:268–81.

    Article  CAS  PubMed  Google Scholar 

  2. European Union. EU Action on Antimicrobial Resistance. Antimicrobial Resistance - European Commission. https://ec.europa.eu/health/amr/antimicrobial-resistance_en. Accessed 17 Apr 2020.

  3. Gieraltowski L, Higa J, Peralta V, Green A, Schwensohn C, Rosen H, et al. National outbreak of multidrug resistant Salmonella Heidelberg infections linked to a single poultry company. PLoS One. 2016;11(9):e0162369.

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  4. Ethelberg S, Lisby M, Torpdahl M, Sørensen G, Neimann J, Rasmussen P, et al. Prolonged restaurant-associated outbreak of multidrug-resistant Salmonella Typhimurium among patients from several European countries. Clin Microbiol Infect. 2004;10:904–10.

    Article  CAS  PubMed  Google Scholar 

  5. Li M, Havelaar AH, Hoffmann S, Hald T, Kirk MD, Torgerson PR, et al. Global disease burden of pathogens in animal source foods, 2010. PLoS ONE. 2019;14:e0216545.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  6. Helms M, Vastrup P, Gerner-Smidt P, Mølbak K. Excess mortality associated with antimicrobial drug-resistant Salmonella Typhimurium. Emerg Infect Dis. 2002;8:490–5.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  7. Helms M, Simonsen J, Mølbak K. Quinolone resistance is associated with increased risk of invasive illness or death during infection with Salmonella serotype Typhimurium. J Infect Dis. 2004;190:1652–4.

    Article  PubMed  Google Scholar 

  8. Threlfall EJ. Epidemic Salmonella Typhimurium DT 104–a truly international multiresistant clone. J Antimicrob Chemother. 2000;46:7–10.

    Article  CAS  PubMed  Google Scholar 

  9. García-Feliz C, Collazos JA, Carvajal A, Herrera S, Echeita MA, Rubio P. Antimicrobial Resistance of Salmonella enterica Isolates from apparently healthy and clinically ill finishing pigs in Spain. Zoonoses Public Health. 2008;55:195–205.

    Article  PubMed  Google Scholar 

  10. Hur J, Jawale C, Lee JH. Antimicrobial resistance of Salmonella isolated from food animals: A review. Food Res Int. 2012;45:819–30.

    Article  CAS  Google Scholar 

  11. Jiang Z, Anwar TM, Peng X, Biswas S, Elbediwi M, Li Y, et al. Prevalence and antimicrobial resistance of Salmonella recovered from pig-borne food products in Henan. China Food Control. 2021;121: 107535.

    Article  CAS  Google Scholar 

  12. European Food Safety Authority, European Centre for Disease Prevention and Control. The European Union summary report on antimicrobial resistance in zoonotic and indicator bacteria from humans, animals and food in 2017. EFSA J. 2019;17:5598.

    Google Scholar 

  13. Jajere SM. A review of Salmonella enterica with particular focus on the pathogenicity and virulence factors, host specificity and antimicrobial resistance including multidrug resistance. Vet World. 2019;12:504–21.

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  14. Liao J, Orsi RH, Carroll LM, Kovac J, Ou H, Zhang H, et al. Serotype-specific evolutionary patterns of antimicrobial-resistant Salmonella enterica. BMC Evol Biol. 2019;19:132.

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  15. De Knegt LV, Pires SM, Hald T. Attributing foodborne salmonellosis in humans to animal reservoirs in the European Union using a multi-country stochastic model. Epidemiol Infect. 2015;143:1175–86.

    Article  Google Scholar 

  16. Teng KT, Martinez Avilés M, Ugarte-Ruiz M, Barcena C, de la Torre A, Lopez G, et al. Spatial trends in Salmonella infection in pigs in Spain. Front Vet Sci. 2020;7:345.

    Article  PubMed  PubMed Central  Google Scholar 

  17. Grimont PA, Weill FX. Antigenic formulae of the Salmonella serovars. WHO collaborating centre for reference and research on Salmonella. 2007;9:1–166.

    Google Scholar 

  18. EUCAST. Antimicrobial wild type distributions of microorganisms. https://mic.eucast.org/Eucast2/. Accessed 17 Mar 2020.

  19. R Core Team. R: A Language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. 2019.

    Google Scholar 

  20. RStudio Team. RStudio: Integrated development for R. Boston, MA: RStudio, Inc; 2019.

    Google Scholar 

  21. Wickham H. Tidyverse: Easily install and load the “tidyverse.” 2017.

    Google Scholar 

  22. Bivand R, Wong DWS. Comparing implementations of global and local indicators of spatial association. TEST. 2018;27:716–48.

    Article  Google Scholar 

  23. Halekoh U, Højsgaard S, Yan J. The R Package geepack for generalized estimating equations. J Stat Softw. 2006;15:1–11.

    Article  Google Scholar 

  24. Pan W. Akaike’s Information Criterion in generalized estimating equations. Biometrics. 2001;57:120–5.

    Article  CAS  PubMed  Google Scholar 

  25. Bartoń K. Package ‘MuMIn.’ 2020.

    Google Scholar 

  26. Scutari M. Learning Bayesian Networks with the bnlearn R Package. J Stat Softw. 2010;35:1–22.

    Article  Google Scholar 

  27. Korb KB, Nicholson AE. Bayesian artificial intelligence. Boca Raton: Chapman & Hall/CRC; 2004.

    Google Scholar 

  28. Le S, Josse J, Husson F. FactoMineR: An R package for multivariate analysis. J Stat Softw. 2008;25:1–18.

    Article  Google Scholar 

  29. Kassambara A, Mundt F. Package “factoextra”: Extract and visualize the results of multivariate data analyses. 2016.

    Google Scholar 

  30. Josse J, Husson F. missMDA: A Package for handling missing values in multivariate data analysis. J Stat Softw. 2016;70:1–31.

    Article  Google Scholar 

  31. Ward J. Hierarchical grouping to optimize an objective function. J Am Stat Assoc. 1963;58:236–44.

    Article  Google Scholar 

  32. Husson F, Josse J, Pages J. Principal component methods-hierarchical clustering-partitional clustering: why would we need to choose for visualizing data. 2010.

    Google Scholar 

  33. Jaspers S, Ganyani T, Ensoy C, Faes C, Aerts M. Development and application of statistical methodology for analysis of the phenomenon of multi-drug resistance in the EU: demonstration of analytical approaches using antimicrobial resistance isolate-based data. EFSA Supporting Publications. 2016;13:1084E.

    Google Scholar 

  34. Márquez RJA, Salaberria AE, García AM, Jimenez SV, Martinez AC, García AA, et al. Surveillance and antimicrobial resistance of Salmonella strains isolated from slaughtered pigs in Spain. J Food Prot. 2007;70:1502–6.

    Article  Google Scholar 

  35. Mejía W, Casal J, Zapata D, Sánchez GJ, Martín M, Mateu E. Epidemiology of salmonella infections in pig units and antimicrobial susceptibility profiles of the strains of Salmonella species isolated. Vet Rec. 2006;159:271–6.

    Article  PubMed  Google Scholar 

  36. Agustín AI, Carramiñana JJ, Rota C, Herrera A. Antimicrobial resistance of Salmonella spp. from pigs at slaughter in Spain in 1993 and 2001. Lett Appl Microbiol. 2005;41:39–44.

    Article  PubMed  CAS  Google Scholar 

  37. Wu B, Ed Dra A, Pan H, Dong C, Jia C, Yue M. Genomic Investigation of Salmonella Isolates Recovered From a Pig Slaughtering Process in Hangzhou China. Front Microbiol. 2021;12:704636.

    Article  PubMed  PubMed Central  Google Scholar 

  38. Pan H, Paudyal N, Li X, Fang W, Yue M. Multiple Food-Animal-Borne Route in Transmission of Antibiotic-Resistant Salmonella Newport to Humans. Front Microbiol. 2018;9:23.

    Article  PubMed  PubMed Central  Google Scholar 

  39. Wang X, Biswas S, Paudyal N, Pan H, Li X, Fang W, et al. Antibiotic Resistance in Salmonella Typhimurium Isolates Recovered From the Food Chain Through National Antimicrobial Resistance Monitoring System Between 1996 and 2016. Front Microbiol. 2019;10:985.

    Article  PubMed  PubMed Central  Google Scholar 

  40. Schmidt JW, Brichta-Harhay DM, Kalchayanand N, Bosilevac JM, Shackelford SD, Wheeler TL, et al. Prevalence, Enumeration, Serotypes, and Antimicrobial Resistance Phenotypes of Salmonella enterica Isolates from Carcasses at Two Large United States Pork Processing Plants. Appl Environ Microbiol. 2012;78:2716–26.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  41. Jiang Z, Paudyal N, Xu Y, Deng T, Li F, Pan H, et al. Antibiotic Resistance Profiles of Salmonella Recovered From Finishing Pigs and Slaughter Facilities in Henan. China Front Microbiol. 2019;10:1513.

    Article  PubMed  Google Scholar 

  42. European Food Safety Authority, European Centre for Disease Prevention and Control. The European Union summary report on trends and sources of zoonoses, zoonotic agents and food‐borne outbreaks in 2017. EFSA  J. 2018;16:5500.

    Google Scholar 

  43. European Food Safety Authority, European Centre for Disease Prevention and Control. The European Union summary report on antimicrobial resistance in zoonotic and indicator bacteria from humans, animals and food in 2016. EFSA J. 2018;16:5182.

    Google Scholar 

  44. European Food Safety Authority, European Centre for Disease Prevention and Control. The European Union summary report on antimicrobial resistance in zoonotic and indicator bacteria from humans, animals and food in 2015. EFSA J. 2017;15:4694.

    Google Scholar 

  45. European Food Safety Authority, European Centre for Disease Prevention and Control. The European Union summary report on antimicrobial resistance in zoonotic and indicator bacteria from humans, animals and food in 2014. EFSA J. 2016;14:4380.

    Google Scholar 

  46. European Food Safety Authority, European Centre for Disease Prevention and Control. The European Union summary report on antimicrobial resistance in zoonotic and indicator bacteria from humans, animals and food in 2012. EFSA J. 2014;12:3590.

    Google Scholar 

  47. European Food Safety Authority, European Centre for Disease Prevention and Control. The European Union summary report on antimicrobial resistance in zoonotic and indicator bacteria from humans, animals and food in 2013. EFSA J. 2015;13:4036.

    Google Scholar 

  48. European Food Safety Authority, European Centre for Disease Prevention and Control. The European Union summary report on antimicrobial resistance in zoonotic and indicator bacteria from humans, animals and food in 2010. EFSA J. 2012;10:2598.

    Google Scholar 

  49. Jacoby GA, Strahilevitz J, Hooper DC. Plasmid-mediated quinolone resistance.  Microbiol Spectr. 2014;2:10.

    Google Scholar 

  50. Gunell M, Webber MA, Kotilainen P, Lilly AJ, Caddick JM, Jalava J, et al. Mechanisms of resistance in nontyphoidal Salmonella enterica strains exhibiting a nonclassical quinolone resistance phenotype. Antimicrob Agents Chemother. 2009;53:3832–6.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  51. Arcangioli M-A, Leroy-Sétrin S, Martel J-L, Chaslus-Dancla E. A new chloramphenicol and florfenicol resistance gene flanked by two integron structures in Salmonella Typhimurium DT104. FEMS Microbiol Lett. 1999;174:327–32.

    Article  CAS  PubMed  Google Scholar 

  52. Murray IA, Shaw WV. O-Acetyltransferases for chloramphenicol and other natural products. Antimicrob Agents Chemother. 1997;41:1–6.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  53. Chuanchuen R, Padungtod P. Antimicrobial resistance genes in Salmonella enterica isolates from poultry and swine in Thailand. J Vet Med Sci. 2009;71:1349–55.

    Article  CAS  PubMed  Google Scholar 

  54. Alvarez J, Lopez G, Muellner P, Frutos C de, Ahlstrom C, Serrano T, et al. Identifying emerging trends in antimicrobial resistance using Salmonella surveillance data in poultry in Spain. Transbound Emerg Dis. 2020;67:250-62.

  55. Sévellec Y, Vignaud M-L, Granier SA, Lailler R, Feurer C, Le Hello S, et al. Polyphyletic nature of Salmonella enterica serotype Derby and lineage-specific host-association revealed by genome-wide analysis. Front Microbiol. 2018;9:891.

    Article  PubMed  PubMed Central  Google Scholar 

Download references

Acknowledgements

The authors thank the European Union’s Horizon 2020 Research and Innovation program and the Community of Madrid for financial support, as well as María García-Martín for the logistics, and all the laboratory staff who have performed laboratory work.

Funding

The current study was supported by the European Union’s Horizon 2020 Research and Innovation program under Grant Agreement No 773830 (One Health European Joint Program) and the Community of Madrid project S2013/ABI-2747.

Author information

Authors and Affiliations

  1. VISAVET Health Surveillance Centre, Universidad Complutense, Madrid, Spain

    Kendy Tzu-yun Teng, Maria Ugarte-Ruiz, Miguel A. Moreno, Lucas Dominguez & Julio Alvarez

  2. Department of Veterinary Medicine, College of Veterinary Medicine, National Chung Hsing University, Taichung City, Taiwan

    Kendy Tzu-yun Teng

  3. Interuniversity Institute for Biostatistics and Statistical Bioinformatics, Hasselt University, Diepenbeek, Belgium

    Marc Aerts & Stijn Jaspers

  4. Data Science Institute, Hasselt University, Diepenbeek, Belgium

    Marc Aerts & Stijn Jaspers

  5. Department of Animal Health, Faculty of Veterinary Medicine, Universidad Complutense, Madrid, Spain

    Miguel A. Moreno, Lucas Dominguez & Julio Alvarez

  6. Subdirección General de Sanidad e Higiene Animal y Trazabilidad, Dirección General de La Producción Agraria, Ministerio de Agricultura, Pesca y Alimentación, Madrid, Spain

    Jose Luis Saez & Soledad Collado

  7. Laboratorio Central de Veterinaria (LCV Algete), Ministerio de Agricultura, Pesca y Alimentación, Madrid, Spain

    Cristina de Frutos

Authors
  1. Kendy Tzu-yun Teng
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Marc Aerts
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Stijn Jaspers
    View author publications

    You can also search for this author in PubMed Google Scholar

  4. Maria Ugarte-Ruiz
    View author publications

    You can also search for this author in PubMed Google Scholar

  5. Miguel A. Moreno
    View author publications

    You can also search for this author in PubMed Google Scholar

  6. Jose Luis Saez
    View author publications

    You can also search for this author in PubMed Google Scholar

  7. Soledad Collado
    View author publications

    You can also search for this author in PubMed Google Scholar

  8. Cristina de Frutos
    View author publications

    You can also search for this author in PubMed Google Scholar

  9. Lucas Dominguez
    View author publications

    You can also search for this author in PubMed Google Scholar

  10. Julio Alvarez
    View author publications

    You can also search for this author in PubMed Google Scholar

Contributions

KTT, MA, SJ, and JA contributed conception and design of the study; KTT, MU, MM, JLS, SC, CF, LD and JA participated in the generation, collection and curation of the data; KTT performed the statistical analysis; KTT, MA, SJ and JA interpreted the results; MA and JA provided supervision; MU and JA facilitated project administration; KTT wrote the first draft of the manuscript. All authors contributed to manuscript revision, read and approved the submitted version.

Corresponding author

Correspondence to Kendy Tzu-yun Teng.

Ethics declarations

Ethics approval and consent to participate

There is no human participant, human information, live animal, and human or animal cell involved in the current study. Only secondary data were used. Therefore, no ethical approval was needed for the current study.

Consent for Publication

Not applicable.

Competing interests

The authors declare that this current work was carried out without the presence of any conflict of interest.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Teng, K.Ty., Aerts, M., Jaspers, S. et al. Patterns of antimicrobial resistance in Salmonella isolates from fattening pigs in Spain. BMC Vet Res 18, 333 (2022). https://doi.org/10.1186/s12917-022-03377-3

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/s12917-022-03377-3

Keywords

BMC Veterinary Research

ISSN: 1746-6148

Contact us