Prevalence, Risk Factors, Antimicrobial Resistance and Molecular Characterization of Salmonella in Northeast Tunisia Broiler Flocks

This study was conducted in northeastern Tunisia to estimate both the prevalence and the risk factors of Salmonella in broiler flocks as well as to characterize the isolated multidrug-resistant (MDR) Salmonella strains. In the present study, a total number of 124 farms were sampled; Salmonella isolates were identified by the alternative technique VIDAS Easy Salmonella. The susceptibility of Salmonella isolates was assessed against 21 antimicrobials using the disk diffusion method on Mueller–Hinton agar using antimicrobial discs. Some antimicrobial resistance genes were identified using PCR. The prevalence rate of Salmonella infection, in the sampled farms, was estimated at 19.9% (64/322). Moreover, a total number of 13 different serotypes were identified. High rate of resistance was identified against nalidixic acid (82.85%), amoxicillin (81.25%), streptomycin (75%), and ciprofloxacin (75%). Alarming level of resistance to ertapenem (12.5%) was noticed. A total of 87.5% (56/64) of isolated strains were recognized as MDR. Three MDR strains were extended-spectrum β-lactamases (ESBL)-producers and three MDR strains were cephalosporinase-producers. The blaCTX-M gene was amplified in all the three ESBL strains. The qnrB gene was not amplified in fluoroquinolones-resistant strains. The tetA and tetB genes were amplified in 5% (2/40) and 2.5% (1/40) of tetracycline-resistant strains, respectively. The dfrA1 gene was amplified in five of the 20 trimethoprim-resistant strains. The mcr-1, mcr-2, mcr-3, mcr-4, and mcr-5 genes were not amplified in any of the phenotypically colistin-resistant strains. In terms of integrase genes int1 and int2, only gene class 2 was amplified in 11% (7/64) of analyzed strains. Risk factors, such as the poor level of cleaning and disinfection, the lack of antimicrobial treatment at the start of the breeding, and a crawl space duration lower than 15 days, were associated with high Salmonella infection in birds. These data should be considered when preparing salmonellosis control programs in Tunisian broiler flocks.


Introduction
According to numerous surveys, undercooked poultry meat is often reported as responsible for non-typhoid Salmonella gastroenteritis outbreaks with Salmonella Typhimurium and Salmonella Enteritidis being the most frequently isolated serotypes [1][2][3]. The Centers for Disease Control and Prevention (CDC) estimated that Salmonella bacteria cause, yearly, 1.35 million infections, 26,500 hospitalizations, and 420 deaths in the United States [4].
Vet. Sci. 2022, 9, 12 2 of 18 Poultry products are the source of most of these cases [5]. In Europe, since 2014, Salmonella spp. has been the second-highest bacterial agent, after Campylobacter, causing gastroenteritis in humans [6]. In Sub-Saharan Africa, non-typhoid salmonellosis, mostly caused by S. Typhimurium and S. Enteritidis, is a major public health problem [2]. In Tunisia, Salmonella food-borne infections are an emerging public health problem [7], with S. Kentucky and S. Anatum being the most identified serotypes in poultry meat [8].
The breeding period represents a critical stage for the development of Salmonella in broiler flocks [9]. Identification of the risk factors associated with broilers' infection with these bacteria is therefore essential in order to prevent its occurrence [9]. In Tunisia, a national program for the control of zoonotic Salmonella infections is performed for only poultry reproductive herds. This program aims to reduce the prevalence of these infections by setting up surveillance. To prevent the spread of the infection outside the poultry farm, all poultry are euthanized when an infection is confirmed in any poultry reproductive farm.
Non-typhoid Salmonella gastroenteritis is expressed by digestive signs (diarrhea, vomiting, and abdominal pain) associated with fever and depression. The first symptom appears approximately 12 to 24 h after ingestion of the contaminated food. The acute phase lasts approximately 24 to 48 h. Symptoms resolve spontaneously, and symptomatic treatment is generally sufficient [10]. In children, elderly, and immune-compromised persons, digestive infections can progress to sepsis and meningitis, leading to death. Antimicrobial therapy is therefore prescribed to particularly sensitive persons [11]. Uncontrolled use of antimicrobials in poultry farming leads to the selection of multidrug-resistant Salmonella strains [12]. The alarming increase of antimicrobial resistance is another aspect of the public health concern of Salmonella infection [12]. It was shown that a proportion of multidrugresistant Salmonella found in humans are of animal origin and have acquired their resistance genes in breeding before being transmitted to humans through food [13].
In Tunisia, data is still lacking, and only one study of Bichiou et al. (2010) [14] is available regarding the infection of broiler chickens by non-typhoid Salmonella. The present study was conducted in northeastern Tunisia to estimate both the prevalence and the risk factors of Salmonella in broiler flocks as well as to characterize the isolated multidrugresistant (MDR) Salmonella strains.

Choice of Breeding Sites
The northeast region of Tunisia is one of the main economic regions that covers seven governorates (Tunis, Ariana, Ben Arous, Manouba, Bizerte, Nabeul, and Zaghouan) and represents the most populated region, with 37% of the Tunisian population and a high demographic density (344 inhabitants/km 2 compared to the mean Tunisian density: 69 inhabitants/km 2 ). Northeast Tunisia is characterized by an intense commercial poultry farming, in particular broilers, totaling 1830 broiler farms (41% of broiler Tunisian farms) producing 49 million chickens/year (45% of national production in 2019) [15].

Sampling
The present study was conducted between September 2019 and August 2020 on broiler chicken farms located in Northeast Tunisia (Figure 1).
The required number of farms (n) was estimated using the formula of Thrusfield [16] for a 95% confidence interval (95% CI): n = 1.96 P exp (1 − P exp )/D 2 , where P exp : expected prevalence which was estimated to 17% [14] and D: the precision that was fixed to 5%.
A total number of 124 broiler chicken farms were included in the present study located in Northeast Tunisia (Figure 1). Visited farms were selected according to the willingness of the owners to participate in the current study. In each farm, at least 50% of the poultry house-buildings were randomly included, totaling 322 breeding units. The number of chickens per poultry house-building varied between 10,000 and 25,000. In order to estimate the prevalence of Salmonella presence in the fecal samples of broiler chicken, ten pools of five fresh droppings were collected in sterile flasks per building. Each poultry flock Vet. Sci. 2022, 9, 12 3 of 18 was sampled once during the production cycle when chickens were two to four weeks old. A total number of 224 and 98 samples were collected in the hot season and the cold season, respectively. The number of samples was lower during the cold season because the number of house-buildings occupied by poultry is lower than during the hot season. In fact, due to the high cost of house-building heating, the production cost is higher during the cold season. Information about the main risk factors, including location, environment, infrastructure, hygiene and biosecurity standards, animal welfare, and health control were collected. The risk factors were typed by questionnaire. The risk factors were assessed based on compliance with good practice and meeting the five animal needs, such as need for a suitable environment, need for a suitable diet, need to be able to exhibit normal behavior patterns, need to be housed with, or apart from, other animals, and need to be protected from pain, suffering, injury, and disease [17].
A total number of 124 broiler chicken farms were included in the present study located in Northeast Tunisia (Figure 1). Visited farms were selected according to the willingness of the owners to participate in the current study. In each farm, at least 50% of the poultry house-buildings were randomly included, totaling 322 breeding units. The number of chickens per poultry house-building varied between 10,000 and 25,000. In order to estimate the prevalence of Salmonella presence in the fecal samples of broiler chicken, ten pools of five fresh droppings were collected in sterile flasks per building. Each poultry flock was sampled once during the production cycle when chickens were two to four weeks old. A total number of 224 and 98 samples were collected in the hot season and the cold season, respectively. The number of samples was lower during the cold season because the number of house-buildings occupied by poultry is lower than during the hot season. In fact, due to the high cost of house-building heating, the production cost is higher during the cold season. Information about the main risk factors, including location, environment, infrastructure, hygiene and biosecurity standards, animal welfare, and health control were collected. The risk factors were typed by questionnaire. The risk factors were assessed based on compliance with good practice and meeting the five animal needs, such as need for a suitable environment, need for a suitable diet, need to be able to exhibit normal behavior patterns, need to be housed with, or apart from, other animals, and need to be protected from pain, suffering, injury, and disease [17].

Salmonella spp. Screening
All samples of fresh droppings were screened for Salmonella by the alternative technique VIDAS Easy Salmonella (bioMérieux SA, Lyon, France). This technique is based on the detection of specific Salmonella proteins by immune-fluorescence after primary and secondary enrichment followed by lysis of the bacteria by heating at 100 • C. Primary enrichment consists of incubating, at 37 • C for 22 h, a mixture of 25 g of fresh droppings samples and 225 mL of Buffered Peptone Water (Biokar Diagnostics, Beauvais, France). After incubation, 100 µL of the mixture was transferred into 10 mL of Salmonella Xpress 2 (SX2) broth (bioMérieux SA) and incubated at 41.5 • C for 24 h. One milliliter of the mixture was boiled for 15 min, then cooled to room temperature. To detect Salmonella target Vet. Sci. 2022, 9, 12 4 of 18 proteins, a volume of 0.5 mL of the already-cooled mixture was transferred to an SLM chip and deposited at VIDAS (bioMérieux SA). The unheated SX2 broth is used for the isolation of Salmonella, from positive samples, on selective agar (XLD and SS) (Biokar Diagnostics, Beauvais, France). Confirmation is performed by urease test followed by API 20E system test (bioMérieux SA) according to the reference technique ISO 6579, 2017 [18].

Salmonella Strains Serotyping
The rapid agglutination on the slide with specific immune sera against the O, H, and Vi antigens of Salmonella (BioRad, Paris, France) was the serotyping technique performed in the present study at Pasteur Institute of Tunis (Tunis, Tunisia). The identification of serotypes was based on the White-Kauffmann-Le Minor scheme [19].

Molecular Study
The extraction of Salmonella genomic DNA was carried out from the colonies present in the unheated SX2 broth (bioMérieux SA) using the ONE-4-ALL GENOMIC DNA Mini kit. Preps (Bio Basic, Markham, ON, Canada). After centrifugation of the SX2 broth for 1 min at 10,000 rpm, the pellet was mixed with washing solution and proteinase K (20 mg/mL). A second centrifugation of the mixture (already heated to 56 • C for one hour) was performed at 12,000 rpm for 2 min. After washing the pellet, the DNA was eluted by incubation at room temperature for 2 min, followed by centrifugation at 9000 rpm for 2 min, then stored at −20 • C until used.
Molecular screening for Salmonella was performed by PCR targeting a 1 kb DNA fragment [20], using specific primers (F5 ACCACGCTCTTTCGTCTGG3 and R-5GAACTGACT ACGTAGACGCTC3 ) [20]. In addition, 12 virulence genes (Table 1) [21][22][23] were screened in Salmonella isolates. PCR was performed with an Esco Swift Max Pro thermal cycler (Horsham, PA, USA) in a total volume of 25 µL containing 1 U Taq Polymerase (Bio Basic, Markham, ON, Canada), 1 × PCR buffer (5 mM KCl Tris-HCl, pH 8.5), 1.5 mM MgCl 2 , 0.1 mM dNTP (Bio Basic, Markham, ON, Canada), 1 µM forward and reverse primer (Bio Basic, Markham, ON, Canada) and 1 µL DNA. Denaturation was carried out first at 95 • C for 10 min followed by 35 cycles (denaturation at 95 • C for 1 min, hybridization for 1 min at temperatures depending on the target gene and elongation at 72 • C for 1 min) and a final extension at 72 • C for 10 min. A total number of 12 primers were used for the detection of the target genes. The amplicons were visualized in an agarose gel. Negative (sterile distilled water) and positive (Salmonella strains isolated in a previous published and unpublished studies [24]) controls were added in each PCR cycle.

Statistical Analysis
Eight factors were selected according to known Salmonella infection risk factors in poultry farms. In addition, 21 variables that represent used antimicrobials were included in the statistical analysis. Data were analyzed with XLSTAT ® (Addinsoft, New York, NY, USA) and SPSS (Version 23.0, IBM Corp; New York, NY, USA). The Pearson chi-square test aims to assess relation between the data collected on poultry farms and the infection of the broiler flock with Salmonella at a threshold value of 5%. Logistic regression was used to assess the effect of the parameters studied on Salmonella infection prevalence; both odds ratios and 95% confidence intervals were estimated. The retrospective likelihood ratio selection method was used, and a p-value cut-off of 0.1 was considered for selecting the risk factors to be included in the model. The goodness of fit was assessed using Hosmer and Lemeshow χ2 test [38]. The area under curve (AUC) of the receiver operating characteristic (ROC) was plotted as an estimation of the predictive ability of the model. Principal component analysis (PCA) was performed to investigate correlations between the isolated strains and the antimicrobial resistance profiles.

Risk Factors Associated with the Presence of Salmonella in the Fecal Samples of Broiler Chicken
The logistic regression based on univariate analysis used to evaluate the impact of eight factors selected according to known Salmonella infection in poultry farms revealed that broiler infection with Salmonella spp. was significantly related to at least six risk factors, namely, hot season (T ≥ 20 °C), duration of crawl space lower than 15 days, absence

Risk Factors Associated with the Presence of Salmonella in the Fecal Samples of Broiler Chicken
The logistic regression based on univariate analysis used to evaluate the impact of eight factors selected according to known Salmonella infection in poultry farms revealed that broiler infection with Salmonella spp. was significantly related to at least six risk factors, Vet. Sci. 2022, 9, 12 8 of 18 namely, hot season (T ≥ 20 • C), duration of crawl space lower than 15 days, absence of treatment with an antimicrobial at the start, wet litter, no cleaning and disinfection around the breeding units, and number of chicks higher than 25/m 2 (p < 0.05) ( Table 3).
Seasonal fluctuation of broiler contamination by Salmonella showed a significant difference between cold and hot season that was characterized in Northeast Tunisia by an average ambient temperature of ≥20 • C (OR = 3.218; 95% CI = 1.520-6.813) ( Table 3). Unlike other serotypes, S. Kentucky, S. Mbandaka, S. Anatum, S. Zanzibar, and S. Enteritidis were present during the whole hot season. They showed a prevalence peak during hot season (78.1%; 50/64) (p < 0.05) (Figure 2).
The eight factors selected according to recognized Salmonella infection risk factors in poultry farms were included in multivariate logistic regression analysis. Using backward likelihood ratio selection method, three factors were left in the final model, namely, no cleaning and disinfection around the breeding units, no treatment with an antimicrobial at the start, and duration of crawl space lower than 15 days (Table 3).
Salmonella broiler flock contamination risk was significantly higher in breeding units where no cleaning and disinfection around the breeding units were performed (OR = 8.642; 95% CI = 1.770- 42.196), no treatment with an antimicrobial at the start was performed (OR = 4.675; 95% CI = 1.720-12.703), and duration of crawl space was lower than 15 days (OR = 3.562; 95% CI = 1.436-8.835). The goodness of fit of the model was assessed using Hosmer and Lemeshow χ2 test. The area under the ROC curve (AUC) was estimated to 0.825 (Figure 3), allowing qualifying the model as good. This means that the obtained results regarding risk factors are reliable. of treatment with an antimicrobial at the start, wet litter, no cleaning and disinfection around the breeding units, and number of chicks higher than 25/m 2 (p < 0.05) ( Table 3). Seasonal fluctuation of broiler contamination by Salmonella showed a significant difference between cold and hot season that was characterized in Northeast Tunisia by an average ambient temperature of ≥20 °C (OR = 3.218; 95% CI = 1.520-6.813) ( Table 3). Unlike other serotypes, S. Kentucky, S. Mbandaka, S. Anatum, S. Zanzibar, and S. Enteritidis were present during the whole hot season. They showed a prevalence peak during hot season (78.1%; 50/64) (p < 0.05) (Figure 2).
The eight factors selected according to recognized Salmonella infection risk factors in poultry farms were included in multivariate logistic regression analysis. Using backward likelihood ratio selection method, three factors were left in the final model, namely, no cleaning and disinfection around the breeding units, no treatment with an antimicrobial at the start, and duration of crawl space lower than 15 days (Table 3).
Salmonella broiler flock contamination risk was significantly higher in breeding units where no cleaning and disinfection around the breeding units were performed (OR = 8.642; 95% CI = 1.770- 42.196), no treatment with an antimicrobial at the start was performed (OR = 4.675; 95% CI = 1.720-12.703), and duration of crawl space was lower than 15 days (OR = 3.562; 95% CI = 1.436-8.835). The goodness of fit of the model was assessed using Hosmer and Lemeshow χ2 test. The area under the ROC curve (AUC) was estimated to 0.825 (Figure 3), allowing qualifying the model as good. This means that the obtained results regarding risk factors are reliable.
All of the strains were resistant to at least two antimicrobials, and 87.5% (56/64) of strains were multidrug-resistant (MDR). The MDR strains were selected based on resistance to over three classes of antimicrobials.   All of the strains were resistant to at least two antimicrobials, and 87.5% (56/64) of strains were multidrug-resistant (MDR). The MDR strains were selected based on resistance to over three classes of antimicrobials.

Prevalence of Antimicrobial Resistance Genes
The bla CTX-M gene was identified in all the three ESBL strains. The qnrB gene was not identified in fluoroquinolone (enrofloxacine and ciprofloxacine)-resistant strains. The tetA and tetB genes were identified in 5% (2/40) and 2.5% (1/40), respectively, of tetracyclineresistant strains. The dfrA1 gene was identified in five of the 20 trimethoprim-resistant strains. The genes mcr-1 to mcr-5 were not identified in any of the five colistin-resistant strains. Integrase gene (int2) was identified in only 11% (7/64) of the Salmonella spp. strains.

Discussion
The results of our study showed that broiler infection prevalence with Salmonella spp. was 19.9% (64/322). These results of the current study corroborate those of Chaiba et al. (2016) [9], who reported that 24% (18/75) of chicken farms were infected with Salmonella in Morocco. Moreover, our result was significantly higher than Salmonella chicken infection rate in China (11.38%; 33/290) (p < 0.05) indicated by Cui et al. (2016) [42]. In addition, our study revealed that the prevalence of Salmonella broiler infection was significantly higher during the hot season compared to the cold season (p < 0.05). High prevalence of infection during the hot season could be due to the high temperature and hygrometry, both favorable for Salmonella growth.
On the other hand, a total of 13 serotypes were identified in the current study. Five serotypes were predominant (S. Kentucky, S. Mbandaka, S. Anatum, S. Zanzibar, and S. Enteritidis) (p < 0.05) ( Table 4). Our results show slight similarities with studies from other countries, such as Canada, the USA, Spain, and China, where S. Enteritidis and S. Typhimurium were the most prevalent [43]. In South Korea, Vietnam, and Cambodia, S. Hadar, S. Infantis, and S. Anatum were the most prevalent [44][45][46]. This suggests that there are geographical differences in the occurrence and dominance of Salmonella serotypes. In addition, the presence of S. Kentucky (20.3%; 13/64) revealed by our study represents a real threat to human health as it is often associated with multidrug resistance to several antimicrobial families, as indicated by Turki et al. (2012) [7].
Our study revealed that the presence of Salmonella spp. in broiler farm buildings is significantly favored by a lack of biosecurity standards, as well as a lack of compliance with animal welfare rules. Our results are in agreement with those of Chaiba et al.  [50] indicated that chicken density greater than 25 subjects per square meter in broiler farm buildings, wet litter, and crawl space < 15 days are important risks of Salmonella broiler infection, respectively.
The results of the present study are in agreement with those of Karraouan et al. (2010) [20] who reported that Salmonella (39 isolates) were positive for the invA gene in turkey meat in Morocco. In addition, most of the Salmonella serotypes isolated were negative for the spvC and Hli genes, while four strains (S. Kentucky) were positive for the Hli gene. The spvC gene was amplified only in a strain of S. Gallinarum. Moreover, Abouzeed et al. (2000) [27] amplified the invA gene in 75 Salmonella isolates of food and human origins. The results of the present study concerning the spvC gene differ from those of Abouzeed et al. (2000) [27]. These authors reported that 28% of Salmonella isolates were positive for the spvC gene (21/75). Our results agree with those of Turki et al. (2012) [7], who indicated that all strains (57) of Salmonella Kentucky isolated from different sources (animals, food, and human) were negative for the spvC gene.
As it was reported in several countries (France, Belgium, Slovak Republic, Morocco, and Ethiopia) [51][52][53], we noticed high resistance rates to nalidixic acid, amoxicillin, streptomycin, and ciprofloxacin, and a high multidrug-resistance (MDR) rate of 87.5% (56/64) in Salmonella isolated in the current study. Three MDR strains were extended-spectrum β-lactamase (ESBL)-producers, and three MDR strains were cephalosporinase-producers. Moreover, the antimicrobials often used in the visited poultry farms belong to the families of fluoroquinolones, beta-lactamines, second-and third-generation cephalosporin, sulfamides, and tetracycline. The high rates of resistance revealed by our study correspond to antimicrobials that belong to these families. Colistin is used very little in the visited poultry farms. It is considered the treatment of last resort prescribed for severe infections caused by bacteria resistant to commonly used antimicrobials. This therapeutic approach could explain the relatively low resistance rate to colistin.
The bla CTX-M gene was amplified in all the three ESBL strains. The tetA, tetB, and dfrA1 genes were identified in a few resistant strains, but qnrB and mrc genes were not identified in any of the MDR Salmonella isolates. Genes of integrase class 2 were identified in 11% (7/64) of resistant Salmonella strains. Our results are comparable to those from the study of Lapierre et al. (2020) [54], who studied antimicrobial resistance of 87 S. Infantis isolates from chicken meat for sale in supermarkets in Santiago, Chile. In this study, high levels of multidrug-resistant and ESBL strains were indicated at the rate of 94 and 63%, respectively. The bla CTX-M-65 gene was identified in 15% (13/87) of isolates. In addition, three isolates were resistant to fluoroquinolones, with the presence of the qnrB gene in two strains. The rate of resistance to colistin was high (29%), but the mcr genes were absent in the 87 studied strains. In addition, class 1 integrons were amplified in 7% (6/87) of the isolates [54].In the present study, some resistance genes were not detected in disc diffusion test positive isolates because resistance to an antimicrobial can be linked to different genes, but we cannot test all these genes.
An interesting result was reported in our study indicating high resistance rate to ertapenem. Carbapenem resistance, mainly among Gram-negative pathogens, is an ongoing public health problem of global dimensions. This type of antimicrobial resistance, especially when mediated by transferable carbapenemase-encoding genes, is spreading rapidly, causing serious outbreaks and dramatically limiting treatment options [55].In addition, the results of our study corroborate those of Ben Hassena et al. (2019) [56], who reported the emergence of multidrug-resistant Salmonella strains isolated from food with decreased susceptibility to fluoroquinolones and third-generation cephalosporin in Tunisia. Then, our findings show the importance of embracing the "One Health" approach that was promulgated by the World Health Organization in its action plan against antimicrobial resistance, which mobilizes various actors in three main sectors: animal husbandry, environment, and human health. In fact, the World Health Organization has named antimicrobial resistance as one of the three most important public health threats of the 21st century [57].

Conclusions
Twenty percent of the broiler flocks were infected with different serotypes of Salmonella. High resistance rates to nalidixic acid, amoxicillin, streptomycin, and ciprofloxacin were detected. An alarming level of resistance to ertapenem (12.5%) was noticed. Nearly ninety percent of strains were identified as multidrug-resistant (MDR). Extended-spectrum β-lactamases (ESBL)-producer and cephalosporinase-producer Salmonella strains were identified. The bla CTX-M gene was amplified in three ESBL strains. Therefore, vigorous control measures are needed in the first step of the poultry chain. In addition, the association of multidrug resistance and virulence genes in Salmonella isolates justifies the need for surveillance systems for both human and animal health sectors.
Institutional Review Board Statement: Ethical approval was not required in this study since neither live animals nor human samples were used in the experiment.

Informed Consent Statement: Not applicable.
Data Availability Statement: The study did not report any data.