*Article* **Epidemiological Cut-Off Values and Multidrug Resistance of** *Escherichia coli* **Isolated from Domesticated Poultry and Pigs Reared in Mwanza, Tanzania: A Cross-Section Study**

**Conjester I. Mtemisika 1,2,\*, Helmut Nyawale 1, Ronald J. Benju 3, Joseph M. Genchwere 3, Vitus Silago 1, Martha F. Mushi 1, Joseph Mwanga 4, Eveline Konje 4, Mariam M. Mirambo <sup>1</sup> and Stephen E. Mshana <sup>1</sup>**


**Simple Summary:** The objectives of this study were to determine the prevalence of multidrug resistance phenotypes and the distribution of *Escherichia coli* among poultry and pigs. Laboratory procedures were conducted according to standard operating procedures and international guidelines. Our findings showed that poultry and pigs reared in Mwanza, Tanzania, are colonized with resistant bacterial phenotypes. Further, different populations of intestinal flora, *E. coli*, exist between poultry and pigs.

**Abstract:** Increasing antimicrobial resistance (AMR) colonizing domesticated animals is a global concern threatening food safety. This study aimed at determining the prevalence of multidrug resistance (MDR) and epidemiological cut-off values (ECVs) of *E. coli* isolated from poultry and pigs in Mwanza, Tanzania. This cross-sectional study was conducted between June and August 2021, involving 297 pigs, 191 broilers, and 203 layers. Rectal and cloacal swabs were collected and processed following standard guidelines. ECVs were determined using normalized resistance interpretation (NRI), a computer software, and descriptive analysis was performed using STATA version 13.0. The overall prevalence of MDR *E. coli* was 63.2%, whereas poultry (87.5% layers and 86.3% broilers) were more colonized than pigs (31.8%) (*p* < 0.001). Based on ECVs of antibiotics tested, *E. coli* from broilers, layers, and pigs exhibited different resistance patterns hence different populations. Exotic breed (*p* < 0.001) and recent antimicrobial use (*p* < 0.001) significantly predicted colonization with MDR *E. coli*. Veterinary officers should implement regulations that prohibit the inappropriate use of antimicrobial agents in livestock keeping.

**Keywords:** antimicrobial resistance; epidemiological cut-off values; *Escherichia coli*; poultry; pigs

#### **1. Introduction**

The use of antimicrobials in livestock to maintain health and promote production is increasing [1], resulting in antimicrobial selection pressure leading to the proliferation of antibiotic-resistant bacteria [2]. Generally, the use of antimicrobials in animals is reported higher in poultry and pigs than in cattle, threatening the safe consumption of poultry and pork and increasing environmental contamination with MDR bacterial strains [1]. Moreover, MDR strains may be transmitted to humans directly via contact with live animals or manure

**Citation:** Mtemisika, C.I.; Nyawale, H.; Benju, R.J.; Genchwere, J.M.; Silago, V.; Mushi, M.F.; Mwanga, J.; Konje, E.; Mirambo, M.M.; Mshana, S.E. Epidemiological Cut-Off Values and Multidrug Resistance of *Escherichia coli* Isolated from Domesticated Poultry and Pigs Reared in Mwanza, Tanzania: A Cross-Section Study. *Animals* **2022**, *12*, 835. https://doi.org/10.3390/ ani12070835

Academic Editors: Sangeeta Rao, Paola Roncada and Bruno Tilocca

Received: 25 February 2022 Accepted: 18 March 2022 Published: 25 March 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

and indirectly via the consumption of contaminated animal products [3,4]. This may result in humans being colonized by multidrug-resistant (MDR) bacteria and subsequently MDR bacterial infections [3].

The increasing unregulated use of antimicrobials in livestock production, particularly in Tanzania, lacks AMR data to create evidence-based standard treatment guidelines for animals [5–7]. In Africa, including Tanzania, studies have documented antimicrobials use among domesticated animals ranging from 77% to 100%, whereas carriage of MDR producing bacteria, particularly Gram-negative bacteria were found to range from 20% to 100% [8,9]. The proportion of MDR strains among *E. coli* isolated from poultry and pigs was 55.2% and 44.8%, respectively, along the Msimbazi basin in Dar es Salaam, Tanzania [10]. However, human antimicrobial susceptibility testing disks and guidelines for interpretation of zones of inhibition used among animal surveillances lack veterinary breakpoints. This practice could over and/or under-report the burden of MDR in livestock [11].

In Mwanza, Tanzania, the prevalence of extended-spectrum β-lactamase (ESBL) among companion and domesticated farm animals was 21.7%. ESBL *E. coli* (93.3%) was predominantly isolated, and pigs were more colonized (33.1%) than other animals [8]. Despite the availability of this information, the prevalence and patterns of MDR Gramnegative bacteria (GNB), notably *E. coli,* is not clearly known as the previous study from the same setting used selective culture media to screen for ESBL producing GNB. The lack of this information may underestimate strategic efforts to prevent the emergence and spreading of MDR bacterial strains among livestock, humans, and environments. Therefore, we designed this study to investigate the prevalence and patterns of MDR *E. coli* and establish epidemiological cut-off values (ECVs) of *E. coli* colonizing poultry and pigs reared in Mwanza, Tanzania. The information from this study will not only facilitate the review of empirical treatment guidelines but also necessitate the implementation of MDR control and preventive measures among poultry and pigs reared in Mwanza, Tanzania. *E. coli* is frequently used as indicator bacteria to monitor trends of antimicrobial resistance (AMR) because it can easily acquire and preserve antibiotics resistance genes from other organisms in the environment and animal populations [12–14]. *E. coli* is also considered a good indicator of the selective pressure imposed by antimicrobial use (AMU) in food animals [12,15,16].

#### **2. Material and Methods**

#### *2.1. Ethical Approval*

Ethical approval for this study was obtained from the joint CUHAS/BMC research ethics and review committee with certificate number CREC/474/2021. Permission to conduct this study was requested from Livestock and Fisheries authorities. Livestock keepers (farmers) were requested to sign permission forms before sample collection. Unique identification laboratory numbers were used throughout the study.

#### *2.2. Study Design, Population Setting, and Duration*

This cross-sectional study was conducted between June and August 2021 among domesticated poultry and pigs reared in 16 wards located in 3 districts in Mwanza, Tanzania, namely, Nyamagana (Nyegezi, Buhongwa, Igoma, Busenga, Kilimahewa, Lumala, Mahina alliance, Malimbe, Mkolani, Nyamongolo, and Mabatini), Ilemela (Buswelu, Kiseke PPF, Nyamanoro, and Pasiansi), and Misungwi (Nyashishi).

#### *2.3. Animals and Farms Selection*

Pigs and poultry farms were systematically selected from a list provided by the livestock officers within the study area. Pig farms with pigs aged ≥ 20 weeks and poultry farms with ≥100 poultry were selected. A total of 29 farms (9-pig farms, 9-broiler farms, and 11-layer farms) were visited and enrolled in this study. Poultry farms were selected based on the number of flocks, and 5% of poultry ready to enter the food chain (aged ≥ 12 weeks) in each selected farm were identified for sampling.

For pigs, 10% of pigs per pig pen ready to enter the food chain (aged ≥ 20 weeks) in each selected farm were randomly identified and sampled. In all 9 farms, a total of 628 pigs were reared, and sampled pigs in each farm totaled: 1st farm 20 pigs, 2nd farm 31 pigs, 3rd farm 28 pigs, 4th farm 41 pigs, 5th farm 33 pigs, 6th farm 45 pigs, 7th farm 49 pigs, 8th farm 31 pigs, and 9th farm 19 pigs, a total of 297 pigs.

#### *2.4. Sample Collection and Transportation*

Fecal samples from the rectum (pigs) or cloaca (chicken) were collected using a sterile cotton swab. Briefly, a sterile swab was gently inserted into the cloaca/rectum and rotated to ensure sufficient sample was collected. Samples were transported to the Microbiology laboratory of the Catholic University of Health and Allied Sciences (CUHAS; Mwanza, Tanzania) in Stuart transport media in a clean, cool box within 8 h following collection.

#### *2.5. Laboratory Procedure*

#### 2.5.1. Isolation of *E. coli*

Swab samples were directly inoculated onto plain MacConkey agar (MCA; HiMedia, India followed by aerobic incubation at 37 ◦C for 18–24 h. After incubation, in the case of mixed growth, a single colony resembling *E. coli* among morphological predominant similar colonies (deep pinkish, round, mid-sized, and flat) was selected for the purity-plate, its sub-culture onto another plain MCA plate, which was incubated aerobically at 37 ◦C for 16–20 h, as reported previously [8,10]. Pure growth of presumptive *E. coli* was further identified by in-house prepared biochemical identification tests to species level.

#### 2.5.2. Physiological and Biochemical Identification of *E. coli*

The presumptive isolates of *E. coli* were preliminarily identified by using conventional in-house prepared physiological and biochemical identification tests, including lactose fermentation, production of CO2 from sugar fermentation, and sulfur production by triple sugar iron (TSI) agar; sulfur production, indole production, and motility by sulfur-indolemotility (SIM) medium; utilization of sodium citrate as the sole source of carbohydrate by Simmons citrate; and urease production by Christensen's urea agar. Identification tests were interpreted as reported previously [17]. Identified isolates of *E. coli* were subjected to antibiotic susceptibility testing (AST) and phenotypic confirmation of ESBL production.

#### 2.5.3. Antibiotics Susceptibility Testing (AST)

All *E. coli* isolates were tested for antibiotics susceptibility by using the disk diffusion method as reported by Kirby-Bauer [18]. Briefly, isolates were suspended in sterile 0.85% normal saline and adjusted to 0.5 McFarland standard solution. Then, MHA plates were inoculated, and antibiotic disks were seeded within 15 min after inoculation of MHA plates. MHA plates were incubated aerobically at 37 ◦C for 16–18 h. The interpretations of zones of inhibitions were performed as recommended by the CLSI 29th Edition guidelines [19]. All *E. coli* that showed intermediate susceptibility to the antibiotics tested were regarded as resistant to such particular antibiotics. Antibiotics tested included ciprofloxacin (CIP 5 μg; HiMedia, Mumbai, India), ampicillin (AMP 10 μg; HiMedia, India), tetracycline (TE 30 μg; HiMedia, India), meropenem (MEM 10 μg; HiMedia, India), ceftazidime (CAZ 30 μg; HiMedia, India), gentamicin (CN 10 μg; HiMedia, India), cefepime (FEP 30 μg; HiMedia, India), and trimethoprim-sulfamethoxazole (SXT 25 μg; HiMedia, India).

#### 2.5.4. Screening and Phenotypic Confirmation of ESBL Production

Isolates from plain MCA were sub-cultured on MCA plates which were supplemented with cefotaxime 2 μg/mL (MCA-C) for the screening of ESBL producing *E. coli* (ESBL-EC) as documented previously [20]. Plates were incubated aerobically at 37 ◦C for 18–24 h. All isolates grown on MCA-C were further confirmed for ESBL production using the phenotypic method, a combination disc method recommended by the Clinical and Laboratory Standards Institute (CLSI) 29th Edition guidelines [19]. Briefly, bacterial suspensions in

sterile normal saline equivalent to 0.5 McFarland standard solution were prepared and inoculated on the entire surfaces of Mueller Hinton agar (MHA; HiMedia, India). Then, disks of ceftazidime 30 μg (CAZ 30 μg; HiMedia, India) with and without clavulanic acid 10 μg (CAZ/CA 30/10 μg; HiMedia, India) were seeded on inoculated MHA plates and incubated aerobically at 37 ◦C for 16–18 h. Isolate exhibiting a difference of ≥5 mm zone of inhibition between CAZ 30 μg and CAZ/CA 30/10 μg were phenotypically confirmed as ESBL-EC.

#### 2.5.5. Quality Control

*E. coli* ATCC 25,922 and *E. coli* ATCC 35,218 were used as control strains to control the performance of culture media, incubation conditions, and antibiotic disks.

#### *2.6. Data Management and Analysis*

Data were entered into Microsoft Excel for cleaning and coding, then into STATA version 13.0 for analysis and NRI computer software, where calculations were performed to define wild type populations by establishing ECVs. All isolates that showed resistance to one or more antibiotic agents in at least three classes were considered multidrugresistant (MDR) strains. Continuous data were presented as mean (±standard deviation: SD)/median (interquartile range: IQR), whereby categorical data were presented as percentages. Chi square analysis was used to show the association between outcome (i.e., MDR colonization) and variables (i.e., antimicrobial exposure, breed of livestock, and species). A *p*-value of <0.05 was considered statistically significant. Epidemiological cut-off values were determined by computer software called Normalized resistance interpretation (NRI; Bioscand AB, Täby, Sweden, International Patent Application WO 02/083935 A1). This method analyzes inhibition zone diameters produced from the disk diffusion technique of antimicrobial susceptibility testing. The NRI software produced a histogram that showed the ECVs and distribution of wild type (sensitive isolates) and non-wild type (resistant isolates) bacteria and the number of SD from the mean [21,22]. In some circumstances where the obtained ECVs were very low, i.e., a zero or negative number, due to the high resistance of *E. coli* to a particular antibiotic, mean was used as a tentative ECV estimate as reported elsewhere [11,23].

#### **3. Results**

#### *3.1. Characteristics of and Antimicrobials Use among Livestock Enrolled in the Study*

A total of 691 livestock, including 27.6% (191/691) broilers, 29.4% (203/691) layers, and 42.9% (297/691) pigs, were sampled from 29 livestock keepers. The majority of livestock were exotic—71.4% (493/691). The recent date of antimicrobial use (AMU) was not known 62.8% (434/691); however, the majority of livestock were administered antimicrobials for therapeutic reasons 86.4% (597/691). About 41.1% (284/691) of livestock were on antimicrobials prescribed by a Veterinary officer (Table 1)


**Table 1.** Characteristics and AMU of livestock.

#### *3.2. Commonly Used Classes of Antibiotics among Livestock Enrolled in this Study*

It was observed from this study that antibiotic agents such as tetracycline, sulfonamides, and quinolones were commonly used in poultry keeping. However, in pigs, antibiotic agents in tetracycline and sulfonamides were common, although classes of quinolones, macrolides, and aminoglycosides were not reported to be used in pigs in this study (Table 2).

**Table 2.** Antimicrobial classes commonly used in livestock keeping.


#### *3.3. Culture Results*

A total of 95.1% (657/691) *E. coli* were isolated from pigs and poultry, of which a total of 63.2% (415/657) were MDR *E. coli*. The MDR proportions in broilers, layers, and pigs were 86.3% (164/190), 87.5% (161/184), and 31.8% (90/283), respectively. A total of 17.8% (117/657) screened positive for potential ESBL production by MCA-C plates. Furthermore, all presumptive ESBL producing *E. coli* 100% (117/117) were phenotypically confirmed to be ESBL producers. ESBL production was significantly high among *E. coli* from layers (30.9%, 57/184) compared to pigs (17.7%, 50/283) and broilers (5.3%, 10/190) *p* < 0.001) (Figure 1).

**Figure 1.** Percentages of MDR and ESBL *E. coli* from broilers, layers, and pigs.

#### *3.4. Resistance Patterns of E. coli to Antibiotics Tested*

Percentage resistance for ciprofloxacin, tetracycline, ampicillin, and trimethoprim/sulfamethoxazole, was high among *E. coli* isolated from poultry compared to pigs. However, tetracycline was the most resistant antibiotic among *E. coli* from pigs compared to other antibiotics tested (Table 3).


**Table 3.** Percentage susceptibility of *E. coli* to antibiotic agents tested.

**Key:** CIP, ciprofloxacin; AMP, ampicillin; MEM, meropenem; TET, tetracycline; CAZ, ceftazidime; SXT, trimethoprim/sulfamethoxazole; CN, gentamicin; FEP, cefepime.

#### *3.5. ECVs of Tested Antibiotics against E. coli*

The ECV of antibiotics tested against *E. coli* isolated from broilers ranged from 11 mm for TET to 18 mm for FEP and CIP; in layers, it ranges from 9 mm for TET to 18 mm for MEM; and from 9 mm for TET to 30 mm for SXT in pigs. In general, *E. coli* from poultry and pigs exhibited different resistance patterns to antibiotics tested and, therefore, ECVs. This observation indicates that *E. coli* from poultry and pigs belong to different populations, whereby *E. coli* from pigs were more susceptible to antibiotics tested than *E. coli* from poultry (Table 4 & Figure 2).



**Key:** CIP, ciprofloxacin; AMP, ampicillin; MEM, meropenem; TET, tetracycline; CAZ, ceftazidime; SXT, trimethoprim/sulfamethoxazole; CN, gentamicin; FEP, cefepime. \* Mean used as ECV tentative estimate. SD = standard deviation.

**Figure 2.** The ECVs of AMP tested against *E. coli* isolated from layers, broilers, and pigs. The arrow indicates where the ECV is located in a histogram, and distribution of WT and non-WT *E. coli* as determined by NRI. Subfigure (**A**) shows ECV of AMP tested against *E. coli* from layers which is 15 mm; subfigure (**B**) shows ECV of AMP tested against *E. coli* from broilers which is 17 mm; and subfigure (**C**) shows ECV of AMP tested against *E. coli* from pigs which is 24 mm.

There is a varying proportion of wild type (WT) distribution when a human clinical breakpoint is used compared to when ECVs are used. Notably, clinical breakpoints underreport percentages of WT-*E. coli* for CIP, TE, MEM, FEP, and CAZ in poultry, whereas clinical breakpoints over-report percentages of WT *E. coli* for AMP, CN, and SXT in pigs (Table 5).

**Table 5.** Comparison of ECVs and CLSI clinical breakpoints of antibiotics tested against *E. coli* isolated from poultry and pigs.


Key: CIP = ciprofloxacin; TE = tetracycline; AMP = ampicillin; MEM = meropenem; CAZ = ceftazidime; FEP = cefepime; CN = gentamicin; SXT = trimethoprim-sulfamethoxazole; ECVs = epidemiological cutoff values; WT = wild type as per ECVs; CBs = clinical breakpoints; and S = susceptible as per clinical breakpoints.

#### *3.6. Factors Associated with MDR Colonization*

The chi square test showed poultry (broilers and layers) are significantly colonized with MDR *E. coli* (*p* < 0.001), furthermore, exotic breeds were significantly colonized with MDR *E. coli* (*p* < 0.001), and the recency of the antimicrobial use predicts the colonization of MDR (*p* < 0.001) (Table 6).

**Table 6.** Factors associated with MDR colonization.


#### **4. Discussion**

Our findings showed different levels of resistance and different ECVs to commonly used antimicrobials in livestock among *E. coli* isolated from poultry and pigs. Most of the animals enrolled in this study (95.1%) were colonized with *E. coli,* similar to a study by Kimera et al., which observed colonization in 80.5% of animals [10] The majority of livestock enrolled were exotic breeds (71.4%) compared to a study done by Seni et al., who enrolled more local breeds (66.7%) [8]. Furthermore, exotic breeds were exposed to antimicrobial use more than local breeds. Generally, about two-thirds (63.2%) of *E*. *coli* isolated from livestock in our study were MDR, similar to a study done in Morogoro, which reported a prevalence of 65.1% [24]. A study done in Dar es Salaam reported low MDR prevalence (51.6%), which is lower than in the current study [10] The high prevalence of MDR in our study might be due to variations in farming conditions and antibiotic use. Furthermore, the level of MDR in the present study is lower than what was observed in China (83%) [25]. This might be explained by the fact that there is different antimicrobial exposure time and frequency, and China is one of the largest users of livestock antimicrobials in the world, increasing the antimicrobials selection pressure as a result of high MDR prevalence [25,26].

It was observed that over three-quarters and one-third of *E. coli* isolated from poultry and pigs were MDR, respectively. As documented previously [10] the level of MDR was significantly higher in poultry (in broilers and layers) than in pigs because there is a higher antimicrobial use in poultry keeping than in keeping pigs, as observed in this and other studies [24,27,28].

We observed more resistance to antibiotics of class quinolones, penicillin, tetracycline, and sulfonamides. High resistance to antibiotics of these classes was not surprising because these were the antibiotics reported to be commonly used by the livestock keepers in this study. MDR patterns observed in this study are in line with what was previously reported in Tanzania (quinolones, penicillin, tetracycline, and sulfonamides), China (tetracycline, sulfonamides, penicillin, quinolones), and Nigeria (tetracycline, sulfonamides, penicillin) [7,10,25,29].

ESBL production is one of the commonest MDR phenotypes. In this study, it was observed that about one-fifth of *E. coli* isolates from poultry and pigs were ESBL producers. Similar to a previous study done in the same region, Mwanza, Tanzania, reported a prevalence of 21.7% [8]. However, the proportion of ESBL producing *E. coli* in our study is lower than in a study done in another region in Tanzania, which reported a prevalence of 65.3% [10]. This might be due to different exposure times to beta-lactam and cephalosporin in livestock keeping, and different *E. coli* populations in different geographic locations.

ECVs determined from NRI are based on the assumption that *E. coli* isolated from broilers, layers, and pigs carrying resistance mechanism/non-wild type exhibit smaller inhibition zone diameters (IZD) than *E. coli* without resistance mechanism/wild type. To the best of our knowledge, this is the first study in Mwanza, Tanzania, to report the ECVs to the commonly used antibiotics tested against *E. coli* isolated from poultry and pigs.

According to this study, the majority of *E. coli* from poultry and pigs were wild type meropenem, ceftazidime, cefepime, and gentamicin. This could be explained by the fact that these antibiotics are not readily available over the counter; that they are expensive (meropenem), while the gentamicin, ceftazidime, and cefepime mode of administration is intravenous (IV), which is not easy for the livestock keepers. This is further supported by the fact that livestock keepers did not report using these antibiotics.

The majority of *E. coli* from broilers were non-wild type to ciprofloxacin, trimethoprimsulfamethoxazole, and ampicillin, while the majority of *E. coli* from non-wild type layers to tetracycline and trimethoprim-sulfamethoxazole. The majority of *E. coli* from pigs were non-wild type to ampicillin. This is not surprising since these antibiotics are commonly used, readily available over the counter, and the cheapest [9]. Furthermore, most livestock keepers from this study reported commonly using antibiotics from tetracycline, quinolones, sulfonamides categories.

A study in China reported 92.3% of *E. coli* from chicken respiratory tract infections were non-wild type to danofloxacin (quinolones), while 22.3% were non-wild type to apramycin (aminoglycosides). Similar observations were seen in this study in *E. coli* from broilers, where the majority were non-wild type to ciprofloxacin (quinolones), and the majority of *E. coli* from poultry were wild type to gentamicin (aminoglycosides) [27]. A study done by Yang et al. reported ECV of danofloxacin (quinolones) tested against *E. coli* from pigs to be 8 μg/mL using MIC [28].

However, contrary to poultry, most *E. coli* from pigs were wild type to most antibiotics tested. This can be explained by the fact that pigs are not exposed to high antimicrobial use compared to poultry. Furthermore, some antibiotics reported to be used in poultry were not reported to be used in pigs in this study, such as quinolones, macrolides, and aminoglycosides.

This study observed different ECVs to the same antibiotics tested to *E. coli* from poultry and pigs, meaning different *E. coli* population, this is not astonishing since poultry and pigs are exposed to different antibiotics with different frequencies of use. Contrary to what was observed in Sweden, the wild type distribution of *E. coli* isolated from human and wild birds in the same area was identical [23].

This study observed that human clinical breakpoints (CLSI) could lead to both over and under-reporting antimicrobial resistance burdens. Similar observations were made by Doidge et al. [11] in the UK, although it was in sheep and beef.

MDR colonization can be attributed to different factors. The most common is antimicrobial pressure which causes the selection of resistant bacteria. In this study, it was observed that exotic breeds were found to predict MDR colonization among poultry and pigs, and there was a significant association of MDR colonization to broilers and layers. This is not surprising as it might be contributed by the fact that there is high antimicrobial use in exotic breeds and poultry keeping. A study done by Seni et al. observed that exotic breeds were more at risk of being colonized with ESBL bacteria than local breeds [8]. ESBL is one of the common MDR phenotypes. Similar to what was observed in the current study, it was observed in a study done by Nonga et al., Katakweba et al., and Kimera et al. that poultry farming is associated with uncontrolled use of both veterinary and human antimicrobials [5,24,28]. A systematic review by Mshana et al. also reported that poultry production accounts for high antimicrobial use in Africa [7]. Recent antimicrobial use was found to predict MDR colonization in the present study. This can be explained by the fact that antimicrobial use creates selection pressure allowing resistant bacteria to multiply and propagate. It was reported that uncontrolled use of antimicrobials in livestock keepings as a growth factor, prophylaxis, and/or therapeutics had been associated with the emergence of MDR bacteria [29].

#### **5. Conclusions**

*E. coli* from layers and broilers are more resistant to commonly used antibiotics than *E. coli* isolates from pigs. Distinct populations of *E. coli* were circulating in layers, broilers, and pigs based on ECVs of different antibiotics tested, which was likely due to differences in antibiotic exposure and breeding type. Furthermore, recent antimicrobial use and exotic breeds predicted MDR colonization which might result from high antimicrobial use. Further studies involving other laboratories are needed to establish the ECVs for commonly used antibiotics and the data used to monitor resistance and further research to establish veterinary breakpoints. Veterinary officers should implement regulations that prohibit the inappropriate use of antimicrobial agents in livestock keeping. More studies to establish the genotypes of *E. coli* circulating in these species are warranted to provide data to monitor the emergence of new *E. coli* strains/genotypes.

#### **6. Study Limitation**

Limited knowledge of livestock keepers on antimicrobial used in their livestock may impact statistical analysis, particularly associated with ESBL carriage. In addition, the ECVs data are from a single laboratory, and, therefore, should be carefully interpreted to reflect the local settings. The resistance frequency may be underestimated, as only one colony per sample was analyzed.

**Author Contributions:** C.I.M., H.N., M.M.M. and S.E.M. conceptualized this study; C.I.M., R.J.B. and J.M.G. worked on the process of sample collection; C.I.M., H.N. and V.S. performed sample processing; C.I.M., V.S., H.N., M.F.M., J.M., E.K., M.M.M. and S.E.M. performed data analysis, C.I.M. wrote the first draft of the manuscript, M.F.M., E.K., M.M.M., J.M. and S.E.M. critically reviewed the manuscript and all authors approved the final manuscript. All authors have read and agreed to the published version of the manuscript.

**Funding:** This study was in-parts funded by SNAP-AMR (MR/S004815/1) and HATUA (MR/S004785/1) projects at CUHAS.

**Institutional Review Board Statement:** The study was conducted in accordance with the Declaration of Helsinki, and approved by the Institutional Review Board of the joint CUHAS/BMC research ethics and review committee (CREC/474/2021).

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** The data presented in this study are available on request from the corresponding author.

**Acknowledgments:** We would like to thank the livestock keepers (farmers) for granting permission for their livestock to participate in the study.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


## *Article* **Metabolomic Profiles of Multidrug-Resistant** *Salmonella* **Typhimurium from Humans, Bovine, and Porcine Hosts**

**Jessie M. Overton 1, Lyndsey Linke 1, Roberta Magnuson 1, Corey D. Broeckling <sup>2</sup> and Sangeeta Rao 1,\***


**Simple Summary:** The global threat that is imposed by the resistance the pathogens develop to antimicrobial drugs is escalating. Tools to detect the resistance (with evidence on molecular and cellular outcomes) would reveal intricate mechanisms through which novel drugs could be developed. Approaches such as metabolomics, which involve metabolite detection, provide scientific evidence of metabolite expression of antimicrobial-resistant pathogens. The current study involved metabolomics of antimicrobial-resistant *Salmonella* Typhimurium collected from various hosts (human, porcine, bovine) and were exposed to antimicrobial drugs—ampicillin, chloramphenicol, streptomycin, sulfisoxazole, and tetracycline—as one set of the experiment. The same isolates were also cultured with no drug exposure as a comparison. There are certain pathways of metabolite expression that are impacted by drug exposure when compared to no drug exposure, meaning that the expressed metabolites could be potential targets for drug companies for the treatment of antimicrobial-resistant pathogens.

**Abstract:** Antimicrobial resistance (AMR) is a global public health threat, yet tools for detecting resistance patterns are limited and require advanced molecular methods. Metabolomic approaches produce metabolite profiles and help provide scientific evidence of differences in metabolite expressions between *Salmonella* Typhimurium from various hosts. This research aimed to evaluate the metabolomic profiles of *S.* Typhimurium associated with AMR and it compares profiles across various hosts. Three samples, each from bovine, porcine, and humans (total *n* = 9), were selectively chosen from an existing library to compare these nine isolates cultured under no drug exposure to the same isolates cultured in the presence of the antimicrobial drug panel ACSSuT (ampicillin, chloramphenicol, streptomycin, sulfisoxazole, tetracycline). This was followed by metabolomic profiling using UPLC and GC–mass spectrometry. The results indicated that the metabolite regulation was affected by antibiotic exposure, irrespective of the host species. When exposed to antibiotics, 59.69% and 40.31% of metabolites had increased and decreased expressions, respectively. The most significantly regulated metabolic pathway was aminoacyl-tRNA biosynthesis, which demonstrated increased expressions of serine, aspartate, alanine, and citric acid. Metabolites that showed decreased expressions included glutamate and pyruvate. This pathway and associated metabolites have known AMR associations and could be targeted for new drug discoveries and diagnostic methods.

**Keywords:** antimicrobial resistance; metabolomics; metabolites; *Salmonella* Typhimurium; resistance markers

## **1. Introduction**

Antimicrobial resistance (AMR) in bacteria isolated from animal hosts is a major global public health threat. The Centers for Disease Control and Prevention have determined

**Citation:** Overton, J.M.; Linke, L.; Magnuson, R.; Broeckling, C.D.; Rao, S. Metabolomic Profiles of Multidrug-Resistant *Salmonella* Typhimurium from Humans, Bovine, and Porcine Hosts. *Animals* **2022**, *12*, 1518. https://doi.org/10.3390/ ani12121518

Academic Editors: Paola Roncada and Bruno Tilocca

Received: 1 April 2022 Accepted: 7 June 2022 Published: 10 June 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

that AMR is "one of the greatest public health challenges of our time" [1]. One of the key goals for slowing or decreasing AMR identified by the White House in the National Strategy for Combatting Antibiotic-Resistant Bacteria is to "accelerate basic and applied research and development for new antibiotics ... " [2]. However, current tools for detecting phenotypic resistance patterns are limited and require advanced molecular methods to reveal associations with AMR patterns [3,4].

Metabolomics is a relatively new tool that can be used to construct metabolite profiles and these metabolite patterns provide evidence of metabolite regulation at the cellular level. When bacteria are exposed to antibiotics, this exposure can trigger cellular changes within the bacterial cell that results in specific metabolic patterns that can help predict antimicrobial drug resistance profiles. Such predictability can have an immediate impact on human and animal health by leading to advancements in drug discoveries by targeting the expression of certain metabolites, as well as diagnostic tools to screen large numbers of samples for AMR. Identifying possible new drug targets would help pharmaceutical companies develop more specific and effective antibiotics to combat AMR bacterial infections [4].

Recent studies have shown that bacteria produce specific metabolic fingerprints when exposed to different classes of antibiotics. These fingerprints can help predict the mode of action used by antibiotics [4] to help develop any novel therapies. For many years, the development of antibiotic resistance could be partly explained by the synthesis of novel analogues of existing compounds [4]. However, such chemical modifications are finite, to keep pace with the remarkable adaptability of the bacteria when exposed to these selective drug pressures in the environment. To combat the prevalence of multidrugresistant (MDR) pathogens, novel antibiotics that target distinct cellular functions are needed [5]. Better understanding the metabolic patterns in AMR bacteria to expose new cellular functions associated with drug resistance and susceptibility is one way to identify new drug targets. One of these studies also suggested that a core metabolic profile for each bacterium is identifiable regardless of the environmental condition, suggesting bacteria could be identified using in vitro metabolic profiles whether in a wound, on surgical equipment, or in the environment [6].

*Salmonella* is a rod-shaped, Gram-negative bacillus that belongs to the Enterobacteriaceae family. This organism is of high public health importance due to its ability to cause several syndromes in both animals and humans. Enteritis, septicemia, abortion, and asymptomatic miscarriages are the major syndromes that present in animals. Enteric fever, gastroenteritis, septicemia, and focal infections are the major syndromes that present in humans infected with *Salmonella* bacteria [7]. *Salmonella enterica* serovar Typhimurium is of utmost importance to public health due to its ability to infect human hosts via contaminated foods. It is one of the most identified serovars in cattle, humans, and pigs and has displayed resistance to ampicillin, chloramphenicol, sulfamethoxazole, and tetracycline [8].

The aim of this project was to identify cellular biomarkers (metabolites) associated with mechanisms of AMR in *Salmonella* Typhimurium using metabolomics and investigate the diversity of those markers among established genetic patterns of resistance in *S*. Typhimurium isolated from humans, bovine, and porcine samples. Metabolomics can be used as a tool to identify the cellular effects of AMR in this pathogen of public health importance. This research has two specific aims (1) to establish metabolomic profiles of *Salmonella* Typhimurium isolated from humans, porcine, and bovine and cultured in the presence and absence of an ACSSuT panel of drugs, and (2) to evaluate similarities and differences in these metabolomic profiles in *Salmonella* Typhimurium across isolates originating from humans, porcine, and bovine hosts. The hypothesis of the study is that non-targeted metabolite profiling will identify biomarker profiles distinctive of AMR in *S*. Typhimurium and, more specifically, the metabolite patterns will differ across various host species.

#### **2. Materials and Methods**

#### *2.1. Isolate Collection, Screening, Identification, and Growth*

*Salmonella* isolates from various institutes (*Salmonella* Typhimurium isolates were contributed by the Colorado Department of Public Health and Environment, CSU-Veterinary Diagnostic Laboratory, Ohio State University, University of Illinois, University of Pennsylvania, and Washington State University) in the US were shipped directly to the Animal Population Health Institute laboratory. A total of 88 human, 33 bovine, and 36 porcine isolates were screened to verify proper serovar typing belonging to *Salmonella* Typhimurium. Briefly, samples were streaked for isolation onto blood agar plates containing 5% sheep blood and incubated overnight at 37 ◦C. A single colony was first tested with the *Salmonella* O Antiserum group Poly A-I, & Vi, and then *Salmonella* O Antiserum Group B, factors 1, 4, 5, 12 (BD Diagnostic Systems, Fisher Scientific, Hampton, NH, USA). After antibody confirmation, the *Salmonella* Typhimurium isolates were grown in 1 mL of trypticase soy broth (TSB), and generated stocks were frozen at −80 ◦C in 10% sterile glycerol.

#### *2.2. Integron and AMR Testing*

A portion of each *Salmonella* Typhimurium stock isolate was scraped into a separate microcentrifuge tube, thawed, and centrifuged for 5 min at 5000× *g*. The supernatant was removed, and each pellet was resuspended in molecular grade water in a 1:3 ratio (10 μL cell pellet suspended in 30 μL of water). A total of 5 μL of each washed, resuspended isolate was used as a template and added to the following PCR mastermix for a 25-μL total reaction volume: 2.5 μL 1× Amplitaq Gold Buffer II and 1.5 mM MgCl2 (Applied Biosystems, Foster City, CA, USA), 0.8 mM dNTPs (0.2 mM each) (Roche Applied Sciences, Indianapolis, IN, USA), 0.4 μM of each primer (Int forward primer sequence: 5 -GGC ATC CAA GCA GCA AGC-3 ; Int reverse primer sequence: 5 -AAG CAG ACT TGA CCT GAT-3 ), 1.875 U Amplitaq Gold polymerase (Applied Biosystems, Foster City, CA, USA), and 2.5 μL 5× Q-Solution (Qiagen, Valencia, CA, USA).

The primers amplify the variable region between the 5 CS to 3 CS region of class 1 integrons [9]. Each reaction was overlaid with 30 μL of Chill Out wax (Bio-Rad, Hercules, CA, USA) to prevent evaporation and placed into an MJ Research 60 place thermal cycler (Bio-Rad). Thermal cycling conditions consisted of an initial incubation at 94 ◦C for 10 min to activate the polymerase and lyse cells, followed by 35 cycles of 94 ◦C for 30 s, 54 ◦C for 1 min, 72 ◦C for 1.5 min, and a final extension incubation at 72 ◦C for 10 min.

PCR products were analyzed by agarose gel electrophoresis using the FlashGel® DNA System (Lonza Group, Ltd., Basel, Switzerland) and visualized by UV light transillumination. A 100 bp–4 kb molecular weight marker (Lonza Group, Ltd., Basel, Switzerland) was concordantly run on the gel as a ladder to aid in the calculation of the size of the amplified DNA fragments. A positive control sample generated from purified DNA from two isolates previously analyzed [9] for class 1 integrons and containing integron sizes of 1000, 1200, and 1600 was included (5 pg total) with each PCR and gel. Samples containing integron sizes of 1000, 1200, 1600, 1800, or both 1000 + 1200 bp were recorded and subsequently re-run on a 1% agarose gel containing a marker and a positive control for proper band size identification. Integron bands were excised from the gel and submitted for DNA purification using the QIAquick PCR Purification kit (Qiagen, Hilden, Germany).

All *Salmonella* Typhimurium isolates used in this study were tested for susceptibility to 16 antimicrobial agents by the disk diffusion assay according to CLSI standard procedures. The AMR testing panel consisted of the following sixteen antimicrobial drugs—amoxicillin–clavulanate (AMC-30), cephalothin (CF-30), chloramphenicol (C-30), ampicillin (AM-10), ceftiofur (CTO-30), enrofloxacin (ERF-5), streptomycin (S-10), triple sulfa (SSS-0.25), tetracycline (TE-30) sulfamethoxazole/trimethoprim (SXT 23.75–1.25), cefoxitin (FOX-30), ciprofloxacin (CIP-5), florfenicol (FFC-30), gentamicin (GM-10), kanamycin (K-30), and nalidixic acid (NA-30). *Escherichia coli* (*E. coli*) ATCC 25922 and *Staphylococcus aureus* ATCC 25923 were used as quality controls.

#### *2.3. Isolate Growth and Extraction for Proteomic and Metabolomics Profiling*

Nine *S.* Typhimurium isolates (three human, three porcine, and three bovine) were selected to undergo an antimicrobial drug growth challenge followed by a non-targeted metabolomics analysis. Criteria for selection were the presence of both 1000 and 1200 base pair integrons, and matching susceptibility/resistance profiles across the 16 drugs tested. Five drugs, ampicillin, chloramphenicol, streptomycin, sulfisoxazole, and tetracycline (ACSSuT panel; Sigma Aldrich, St. Louis, MO, USA), were selected for the *S.* Typhimurium antimicrobial drug challenge.

Ampicillin, chloramphenicol, tetracycline, and streptomycin were each dissolved in water to the desired stock concentration. Sulfisoxazole was added to 10% HCl and heated at 80 ◦C until dissolved. The sulfisoxazole–acid mix was added to TSB, the broth was neutralized to pH 7.0 using NaOH, and the other antibiotics were subsequently added. The final concentration of each antibiotic was based on the recommended minimum inhibitory concentration (MIC) recommended by the Clinical Laboratory and Standards Institute, as shown in Table 1.

**Table 1.** Recommended MIC values for *Salmonella enterica* serotype Typhimurium for the ACSSuT pattern according to 2014 CLSI standards.


Isolates were processed using standard laboratory procedures. They were thawed and streaked for isolation on sheep blood agar plates. One resulting colony from each selected isolate was suspended in 0.5 mL of TSB; 100 μL was inoculated into 20 mL of normal TSB (no drug = ND) and 100 μL was inoculated into 20 mL of ACSSuT TSB (Drug = D). The only difference between the ND and D was that the ND group of cultures were without antimicrobials. Cultures were then incubated with shaking at 37 ◦C for 24 h. After pelleting at 4300× *g* for 10 min at 4 ◦C and supernatant removal, the wet weight of each culture pellet was recorded and adjusted to 20 mg. Pellets were washed with phosphate-buffered saline (PBS) and centrifuged again as above; after discarding PBS supernatant, the pellets were frozen at −20 ◦C. Each sample pellet was thawed at 4 ◦C, suspended in methyl tert-butyl ether (MTBE), and sonicated for 30 s intervals for a total of 6 cycles, with a 30 s cooling on ice between cycles. The sonicated lysates were then centrifuged at 2500× *g* for 5 min at 4 ◦C, and 150 μL of LC-MS grade water and an additional 100 μL MTBE was added to the cleared supernatants. After sealing with Parafilm, sample tubes were vortexed at room temperature for 15 min, incubated at −80 ◦C for 15 min, and centrifuged at 15,890× *g* for 15 min at 4 ◦C. Samples were then divided by a non-polar supernatant, a polar supernatant, and protein lysates. Each layer was dried via nitrogen gas and stored at −80 ◦C for metabolomics analysis.

#### *2.4. Metabolomic Profiling by UPLC- and GC–MS*

An ultra-performance liquid chromatography–mass spectrometry (UPLC-MS) analysis was performed on a Waters Xevo G2-TOF MS coupled with a Waters Acquity UPLC [10]. Separation was performed on a UPLC T3 reverse phase column and data were collected in MSE mode (alternating low and high collision energy) [11]. For the gas chromatography– mass spectrometry (GC–MS) analysis, cell extracts were dried and derivatized using a standard protocol. Briefly, GC–MS data were acquired on a Thermo Scientific Trace-ISQ GC–MS system (Waltham, MA, USA) with separation using a 30 m TG-5MS column. Data from both UPLC-MS and GC–MS acquisitions were processed using XCMS (https: //www.bioconductor.org/packages/release/bioc/html/xcms.html, accessed on 31 March

2022) for peak detection, retention time alignment, and normalization [12]. Metabolite annotation of GC–MS data was performed by grouping molecular features into peak groups using AMDIS software (http://www.amdis.net/, accessed on 31 March 2022) and screening spectra against the CSU in-house spectral library, NIST GC–MS spectral library, and the Golm Metabolite Database (http://gmd.mpimp-golm.mpg.de/, accessed on 31 March 2022). Annotations of UPLC-MS data were performed by an unbiased grouping of molecular features into spectra based on correlational clustering across the dataset [10] and screening spectra against the CSU in-house spectral library (consisting of approximately 1100 compounds), NIST LC-MS spectral library, and MassBank spectral library [12].

#### *2.5. Statistical Analysis (MetaboAnalyst 4.0)*

Data analysis of the biomarkers was completed using MetaboAnalyst 4.0 (MetaboAnalyst 4.0 is available at https://www.metaboanalyst.ca/ (accessed on 26 April 2021) and its R packages are available at https://github.com/xia-lab/MetaboAnalystR, (accessed on 26 April 2021)). The UPLC and GC–MS spectra were combined, normalized, and scaled. To determine the statistically significant (S.S.) metabolites, a pairwise analysis was conducted, including a non-parametric Wilcoxon rank-sum test and fold-change analysis. A two-way analysis of variance (ANOVA) followed by a principal component analysis (PCA) and heatmapping were used to determine and visualize the species and drug effects and interactions. A pathway analysis was then conducted to match S.S. metabolites to known metabolic pathways and determine the biological significance of those pathways.

Multiple features of this program were used, including "Two-factor", "Statistical Analysis", and "Pathway Analysis", to conduct multiple statistical tests, including Wilcoxon rank-sum, fold-change, two-way ANOVA, PCA, and heatmapping. Conducting the "Pathway Analysis" in MetaboAnalyst required all metabolites to have an HMDB identifier. The Human Metabolome Database (HMDB) is a website that compiles detailed information about metabolites and their roles in human metabolic pathways and assigns HMDB identifiers or numbers.

#### **3. Results**

#### *3.1. AMR Patterns and Integrons*

The most common AMR pattern among all resistant samples (23/126 = 18.3%) was ampicillin, amoxicillin–clavulanate, streptomycin, sulfonamides, tetracycline, chloramphenicol, and florfenicol (coded as AMC-AM-S10-SSS-TE-C-FFC). All isolates with this AMR pattern carried both the 1000 and 1200 bp integrons.

#### *3.2. Metabolite Expression by Drug Treatment and Host Species*

Visualization by the principal component analysis (Figure 1) and the two-way ANOVA heatmap (Figure 2) showed that a greater effect on metabolite production was apparent when the samples were exposed to the full drug (ACSSuT panel) treatment, irrespective of species.

**Figure 1.** PCA chart derived from two–way ANOVA showing clustering of samples by drug treatment.

**Figure 2.** Heatmap derived from two–way ANOVA showing clustering of metabolite concentrations based on drug treatment.

#### *3.3. Metabolite Expression and Matched Metabolic Pathways*

Wilcoxon rank-sum showed 653 metabolites that had an S.S. concentration difference (59.69% increased and 40.31% decreased expressions) when the sample was exposed to the ACSSuT antibiotic panel versus when it was not. Of those 653 metabolites, 23 unique metabolites were annotated by the PMF, identifiable by HMDB, and matched to one or multiple of the 9 statistically significant metabolic pathways in MetaboAnalyst. Of these, 60.87% of metabolites had an increased expression when exposed to antibiotics and 39.13% had a decreased expression (Figure 2).

Methionine, nicotinamide, nicotinate, pantothenate, phenylalanine, proline, pyroglutamic acid, pyruvate, serine, threonine, tryptophan, tyrosine, uracil, and valine significantly increased with full drug treatment. Conversely, alanine, aspartate, citrate, cysteine, glutamate, glycerate, glycerone phosphate, glycine, and leucine decreased with full drug treatment.

The metabolic pathways matched to the significantly different metabolites include glycine, serine, and threonine metabolism; alanine, aspartate, and glutamate metabolism; aminoacyl-tRNA biosynthesis; pantothenate and CoA biosynthesis; glutathione metabolism; valine, leucine, and isoleucine biosynthesis; nicotinate and nicotinamide metabolism; glyoxylate and dicarboxylate metabolism; and beta-Alanine metabolism, in order of descending pathway impact scores (Table 2).



**Table 2.** *Cont.*

metabolite expression when exposed to the full drug treatment.

Non-significant

 metabolites.

#### *3.4. Univariate Analysis*

A between-subject, two-way ANOVA identified 297 metabolites (Table 3) that were statistically significant only for the treatment factor. No metabolites were found to be significant for the host species factor or the interaction between host species and treatment.

**Table 3.** Significant identifiable metabolites found via univariate, between-subject, two-way ANOVA.


FDR = false discovery rate; FDR helps control for falsely positive significant features; FDR < 0.05 has less than a 5% probability of being a falsely significant feature.

#### **4. Discussion**

In this study, we investigated metabolite expression patterns in AMR *Salmonella* Typhimurium isolated from human, bovine, and swine when exposed to antibiotics. We were able to demonstrate a greater difference in metabolite expression when the isolates were exposed to the full drug challenge compared to no drug exposure, irrespective of host species. The univariate analysis further confirmed that metabolite expression changes were significant only according to the treatment factor, not according to the host species or interaction of the host species and treatment. Metabolite expression being non-host specific suggests that AMR *Salmonella* Typhimurium drug targets are consistent across human, bovine, and swine hosts. This finding has great significance when considering that future drug testing on AMR *Salmonella* Typhimurium in swine and bovine could be translated to human treatments.

While the expression of 23 specific metabolites significantly changed when exposed to the full drug treatment and these upregulated metabolites each matched significant metabolic pathways, a specific resistance mechanism remains unclear. These isolates were exposed to multiple antimicrobial drugs and each drug has a different mechanism of action. Therefore, there are potentially many mechanisms of resistance that have developed in these isolates [4]. As per Hoerr et al. (2016), the metabolic profiles could be separated in a fingerprint, and based on the specific fingerprints obtained for different classes of antibiotics, the mode of action of several antibiotics could be predicted. The profiles could also be used as potential drug targets for pharmaceutical companies. Over the past few decades, there has been a decline in approvals of new antibiotic drugs in the market by the US Food and Drug Administration (FDA) [13,14]. The number of new antibiotics being developed every year decreases due to the challenges of effectively dispatching both antibiotic-resistant bacteria and novel infectious bacteria [15]. If we want to reverse these trends and facilitate new approaches to overcoming resistance, we must first understand the microbial forces responsible for developing resistance [16]. Metabolomics in particular offers a unique strategy to detect metabolic changes that occur in an organism in response to drugs and the outcomes of such studies can provide insights into their corresponding modes of action [17,18].

The significant changes observed in our study include increases in methionine, nicotinamide, nicotinate, pantothenate, phenylalanine, proline, pyroglutamic acid, pyruvate, serine, threonine, tryptophan, tyrosine, uracil, and valine, and decreases in alanine, aspartate, citrate, cysteine, glutamate, glycerate, glycerone phosphate, glycine, and leucine. These metabolites were matched to nine significant metabolic pathways, including glycine, serine, and threonine metabolism; alanine, aspartate, and glutamate metabolism; aminoacyltRNA biosynthesis; pantothenate and CoA biosynthesis; glutathione metabolism; valine, leucine, and isoleucine biosynthesis; nicotinate and nicotinamide metabolism; glyoxylate and dicarboxylate metabolism; and beta-Alanine metabolism.

Lin et al. (2019) [19] found biosynthesis of amino acids, biosynthesis of phenylpropanoids, and purine metabolism were commonly enriched in MDR strains of *E. coli*, and the results concurred that antibiotic resistance affects the metabolite profiles of MDR bacteria. Several related metabolites, such as glycerol, were increased in MDR strains, while citric acid and succinic acid were decreased in MDR strains [19].

An established resistance mechanism against β-lactams, such as ampicillin, includes the production of metallo-β-lactamases, which inactivate the drug through a cleavage process. The metallo-β-lactamases are especially threatening due to their ability to inactivate multiple β-lactams and their insensitivity to β-lactamase inhibitors that target the acyl serine transferases. This resistance mechanism has been identified in extended-spectrum β-lactamases where two amino acid substitutions are critical, a serine-for-arginine and a lysine-for-glutamate [20]. This substitution may explain the increased expression of serine and the decreased expression of glutamate observed when isolates are exposed to the ACSSuT drug panel in our study. Aspartate has also been identified as a critical component

of the metallo-β-lactamases; thus, the increased expression of aspartate may support this mechanism [20].

Perhaps one of the largest resistance mechanisms is through decreasing TCA cycle flux. Previous studies have shown that exogenous alanine and/or glucose increase susceptibility to antibiotic treatment by increasing TCA flux and thereby increasing drug uptake by the cell [21]. Therefore, it is possible that decreased TCA flux could contribute to decreased drug susceptibility. Decreased concentrations of pyruvate and glutamate in our study support this conclusion, as pyruvate directly feeds the TCA cycle and glutamate is converted to pyruvate by α-ketoglutarate [22].

These data from our study suggest that another resistance mechanism utilized by these AMR isolates may be initiated from the aminoacyl-tRNA pathway. AminoacyltRNA biosynthesis is responsible for changing cell membrane properties and increasing a pathogen's resistance. It has previously been identified as an attractive drug target [22]. This pathway likely acts by decreasing cell permeability and, thus, inhibiting drug entrance into the cell. The aminoacyl-tRNA biosynthesis pathway in our study is significantly altered when isolates are exposed to the ACSSuT antibiotic panel.

Alanine is a required component of cell wall peptidoglycan and it has been demonstrated that inhibition of alanine transport results in increased susceptibility to drugs [23]. Increased concentrations of alanine may indicate that the cell wall has undergone peptidoglycan remodeling, resulting in decreased susceptibility.

In our study, citrate (citric acid) has the highest fold change of any of the metabolites matched to a significant pathway, but its possible role in antimicrobial resistance is less clear. Citrate has previously been described as having a role in the regulation of cell division and gene expression and is known to be a chelator, which may allow bacteria to manage intracellular concentrations of cations. Previous research has shown an increase in citrate concentrations when *Salmonella* aureus is exposed to cold temperatures, as well as upregulated cell division proteins [24]. Therefore, increased citrate concentrations may suggest that *S.* Typhimurium depends on this metabolite to maintain intracellular Ca++ concentrations and increases the rate of cell division. An increased rate of cell division would also increase the chances of DNA mutation occurring and antibiotic resistance developing. Further examining the role of citrate in bacterial survival and AMR is warranted.

In future research, exposing isolates to only one antibiotic or one class of antibiotics would allow for a more specific interpretation of the expressed metabolites and potentially provide more robust evidence on resistance mechanisms. Interpretation of these data is limited due to the multiple mechanisms by which the ACSSuT panel targets bacteria. Resistance mechanisms against one class of antibiotics differ from those against another class, hence why bacteria resistant to one class may be susceptible to a different one [25]. This explains why a distinct resistance mechanism was not identifiable in this project. Exposure to a single antibiotic class may create a more easily identifiable profile of metabolites attributable to a specific resistance mechanism.

#### **5. Conclusions**

The findings of this study suggest that exposing AMR *Salmonella* Typhimurium to an ACSSuT panel significantly alters metabolic pathways and, thus, metabolite expression by the bacteria. This research supports the continuation of using metabolomics to study AMR and identify resistance mechanisms, which could become future drug or testing targets. However, further studies are necessary to identify specific resistance mechanisms for different classes of antibiotics.

**Author Contributions:** Conceptualization, S.R., C.D.B., L.L. and R.M.; Methodology, C.D.B., L.L., S.R. and R.M.; software, J.M.O. and S.R.; validation, J.M.O. and S.R.; formal analysis, J.M.O. and S.R.; resources, C.D.B.; data curation, J.M.O.; writing—original draft preparation, J.M.O.; writing—review and editing, S.R., C.D.B., L.L. and R.M.; visualization, J.M.O.; supervision, S.R.; project administration, S.R.; funding acquisition, S.R., C.D.B., L.L. and R.M. All authors have read and agreed to the published version of the manuscript.

**Funding:** This study was funded by the USDA Animal Health & Disease (USDA COLV-2015-36100- 06008) through the College of Veterinary Medicine and Biomedical Sciences College Research Council at Colorado State University.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** All data generated or analyzed during this study are included in this published article.

**Acknowledgments:** We are thankful to all the institutes who have contributed the *Salmonella* Typhimurium isolates for this study: Colorado Department of Public Health and Environment, Washington State University Paul G Allen School for Global Animal Health, Ohio State University College of Veterinary Medicine, University of Illinois Urbana-Champaign Veterinary Diagnostic Laboratory, University of Pennsylvania Veterinary Diagnostic Services, and Colorado State University Veterinary Diagnostic Laboratory. Metabolomic profiling was performed at the Proteomics and Metabolomics Facility (currently known as Analytical Resource Core) at Colorado State University.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Animals* Editorial Office E-mail: animals@mdpi.com www.mdpi.com/journal/animals

MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel: +41 61 683 77 34

www.mdpi.com

ISBN 978-3-0365-6343-5