Antimicrobial resistance phenotypes of E. coli from livestock and wildlife in Wyoming, U.S.A

Abstract

Background

This study measured phenotypic resistance of enteric Escherichia coli (E. coli) obtained from livestock and wildlife across agricultural districts within Wyoming, U.S.A., to antimicrobials used in the cattle industry. It is the first investigation of antimicrobial resistance (AMR) in E. coli from free-ranging wild ungulates in the American West. Although wildlife can harbor resistant bacteria, this host population is generally understudied in AMR research. By identifying overlapping AMR profiles between livestock and free-ranging wild herbivores, transmission dynamics may be better understood.

One hundred and eighty-one enteric E. coli isolates were tested for phenotypic resistance to ampicillin, tetracycline, trimethoprim-sulfamethoxazole, ceftiofur, florfenicol, and tulathromycin using broth microdilution susceptibility testing. For statistical analysis, a logistic regression model was constructed in R using R Studio. A mass spectrometry profile was constructed for each isolate, and protein dendrograms were generated to assess clustering by agricultural district and resistance phenotype.

Result

For the antimicrobials evaluated, a higher number of AMR E. coli isolates were detected in livestock than in wildlife. Wildlife isolates had significantly lower odds of resistance to ampicillin (89.6% decrease in odds of resistance (95% confidence interval (CI) 64.3% − 97.0%), χ21 = 20.251, p = < 0.001), tetracycline (99.1% decrease in odds of resistance (95% CI 93.0% − 99.9%), χ21 = 79.608, p = < 0.001), trimethoprim-sulfamethoxazole (81.4% decrease in odds of resistance (95% CI 14.5% − 96.0%), χ21 = 6.421, p = 0.011), and florfenicol (66.2% decrease in odds of resistance (95% CI 23.1% − 85.1%), χ21 = 7.4272, p = 0.006). There was no significant difference in the odds of resistance between wildlife and livestock for tulathromycin (χ21 = 2.0846, p = 0.15). Ceftiofur could not be statistically evaluated. No significant difference in the odds of resistance was observed across agricultural districts. Simple mass spectrometry did not reveal clustering of isolates by host type, location, or resistance profile.

Conclusions

This study provides a valuable first look into AMR in E. coli from Wyoming’s wildlife and livestock. As AMR remains an ever-increasing threat, it is encouraging that this study found a consistently low number of wildlife isolates exhibited AMR, even to those antimicrobials with high numbers of AMR isolates in livestock. However, the increased resistance in livestock E. coli, and the presence of some wildlife E. coli isolates resistant to multiple drugs is concerning. The findings of this study highlight the need for increased multi-angled AMR research and surveillance efforts in livestock and wildlife in the American West.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12917-026-05386-y.

Background

Antimicrobial drugs are instrumental in modern veterinary medicine. However, as resistance to current antimicrobials increases, livestock producers will face greater challenges when treating infections in their animals, thus reducing the number of healthy animals that can enter the human food supply [1]. Antimicrobial resistance (AMR) is considered one of the top ten global public health threats by the World Health Organization [2]. In this article, the term “AMR” will be used to refer to the resistance of bacteria to antimicrobials. A One Health-inspired, dual-angled approach was used to investigate AMR in Escherichia coli (E. coli) from Wyoming’s livestock, specifically cattle (Bos taurus), and wildlife, specifically mule deer (Odocoileus hemionus) and bighorn sheep (Ovis canadensis). By comparing the prevalence of AMR E. coli in both domestic and free-ranging large herbivore populations, we can gain a comprehensive view of AMR trends.

Livestock may be treated with antimicrobials at any point in their lives, which is one factor in the emergence of AMR. Most livestock AMR research focuses on intensive management systems such as dairies and feedlots, as these environments favor bacterial transmission between animals due to greater opportunities for nose-to-nose and oral-fecal contact [3]. Because bacteria can be shared across host species, AMR E. coli from livestock could also be transferred to untreated free-ranging wildlife [4].

Most existing studies on AMR in wildlife focus on birds and urban wildlife, because of their high potential for interactions with humans, human waste, and trash [5, 6]. Although free-ranging wildlife do not receive antimicrobial treatments, they have been found to harbor resistant bacteria and resistance genes, and multiple studies suggest that wildlife could act as melting pots for AMR genetic elements [5, 7, 8]. It has also been suggested that wildlife can serve as an important indicator of AMR spillover into non-treated populations [7].

E. coli was chosen as the bacterial species of interest in this study because it is an important AMR surveillance species that can be either a commensal or a pathogen, can colonize many different host species, and may serve as a reservoir of resistance genes [9, 10]. E. coli is also highly adaptable and can rapidly acquire and share AMR genes with other pathogens, particularly with other Enterobacteriaceae species such as Salmonella spp., via horizontal gene transfer [10].

Here, E. coli isolates from both wildlife and livestock were evaluated for phenotypic resistance to ampicillin, tetracycline, trimethoprim-sulfamethoxazole (TMP-SMZ), florfenicol, tulathromycin, and ceftiofur using broth microdilution testing. Ampicillin and tetracycline represent classes of antimicrobials extensively used in the cattle industry, including previous use in livestock feed as animal growth promoters [9]. Florfenicol, tulathromycin, and ceftiofur are often used to treat bovine respiratory disease complex [11]. TMP-SMZ is not licensed for use in cattle, but other sulfonamides are commonly used in the treatment of calf scours [12]. It was hypothesized that, for all antimicrobials tested, a greater number of AMR isolates would be found in cattle in comparison to mule deer and bighorn sheep. It was also hypothesized that higher levels of resistant isolates would be found in state agricultural districts with higher livestock densities.

This study explored the suitability of using MALDI-TOF (Matrix-Assisted Laser Desorption Ionization–Time of Flight) mass spectrometry to quickly differentiate resistant and sensitive isolates by investigating protein dendrogram clustering of E. coli for patterns linked to AMR profiles. It was hypothesized that isolates with similar AMR profiles would cluster together in the protein dendrogram, even without sublethal exposure to antimicrobials during sample preparation for MSP creation.

Methods

Study area

Wyoming provides an ideal study location due to its rural landscape, low human population, and large number of cattle and free-ranging wild ungulates. In 2022, Wyoming’s total human population was 581,381 (Table 1) [13]. In 2020, Wyoming’s cattle population was estimated to be 1,320,000 (Table 1) [14]. The Wyoming Game and Fish Department (WGFD) estimated that there were approximately 330,000 mule deer in Wyoming in 2021, and approximately 5,200 bighorn sheep in Wyoming in 2020 [15, 16].

Table 1.

Percentages of humans and cattle in each agricultural district, and cattle per farm acre

Wyoming Agricultural District	Percentage of Total Humans	Percentage of Total Cattle	Cattle per farm acre
Northeast	17%	25%	0.039
Northwest	16%	17%	0.078
South Central	30%	17%	0.029
Southeast	24%	30%	0.056
West	13%	11%	0.085

To classify samples by location, Wyoming was divided into the five agricultural districts used by the United States Department of Agriculture National Agricultural Statistics Service (USDA NASS) [17, 18] (Fig. 1). The highest density of cattle, calculated as cattle per farm acre, is found in the West district, and the lowest is found in the South Central District (Table 1) [18]. E. coli isolates from livestock and wildlife were assigned to the agricultural district where the animal was sampled.

Fig. 1.

Study area; State of Wyoming, U.S.A. with USDA NASS agricultural districts

Livestock and wildlife sample sources and distribution

Up to twenty E. coli isolates from wildlife and twenty E. coli isolates from livestock were chosen from each agricultural district. When more than twenty samples for each host type were available for a district, samples for the final set were chosen in order they were received (livestock) or collected (wildlife). For wildlife, mule deer samples were chosen first. If 20 mule deer samples were not collected in an agricultural district, bighorn sheep samples from the same agricultural district were used. The exception was the Southeast district, where only one wildlife E. coli sample was able to be obtained. The final sample set included 100 isolates from livestock and 81 isolates from wildlife (Table 2).

Table 2.

E. coli isolate sources

	Livestock - Cattle		Wildlife – Mule deer		Wildlife – Bighorn sheep
Total samples	100		65		16
			81
District	Collected by	Number of isolates	Collected by	Number of isolates	Collected by	Number of isolates
South Central	WSVL	20	WGFD	15	WGFD	5
Southeast	WSVL	20	-	-	WGFD	1
Northeast	WSVL	20	WGFD	20	-	-
Northwest	WSVL	20	WGFD	14	WGFD	6
West	WSVL	15	WGFD	16	WGFD	4
	Bangoura Research	5

Livestock E. coli isolates were primarily obtained from cattle diagnostic cases submitted to the Wyoming State Veterinary Laboratory (WSVL) for testing between October 2021 and April 2023. Of the 95 livestock E. coli isolates obtained from WSVL, all may be presumed to be from clinically ill animals with unknown antimicrobial treatment histories. Five additional cattle E. coli isolates from presumably healthy animals were collected in April of 2022 from a herd located within the West agricultural district and were provided to this study by the Bangoura Research Group in the University of Wyoming Department of Veterinary Sciences (Table 2). Cattle ages range from newborn calves to adult cattle, with up to five animals sampled from any one herd. Because complete information regarding the health status and antimicrobial treatment status of all cattle sampled for this study was not available, no distinction was made between healthy and clinically sick cattle, nor between antimicrobial-treated and untreated cattle.

Wildlife fecal pellets were primarily collected from apparently healthy, adult mule deer and bighorn sheep by the WGFD during routine surveillance between December 2021 and March 2022. Two additional wildlife fecal samples were collected from the rectum post-mortem during necropsies in 2022. Of the final 81 wildlife samples, 65 originated from mule deer, and 16 originated from bighorn sheep (Table 2). All wildlife E. coli isolates were assumed to have been collected from healthy animals with no antimicrobial treatment histories.

Bacterial cultures

Please note, all culture media was purchased from Hardy Diagnostics (Hardy Diagnostics, Santa Maria, California, U.S.). For cattle samples obtained from WSVL diagnostic cases, E. coli isolates were cultured from fresh feces, intestinal tissue, or intestinal contents by WSVL Bacteriology laboratory personnel following standard operating procedures (SOP-BACT-73 and SOP-BACT-90) for culturing E. coli from feces. For livestock samples obtained from the Bangoura Research Group, a researcher collected fecal samples from the ground, and they were cultured for E. coli following the same WSVL Bacteriology standard operating procedures. For wildlife samples, WGFD personnel collected fecal pellets from the rectum of each animal and placed them in 5 mL conical tubes. The samples were stored at -80 °C until they were cultured following WSVL Bacteriology standard operating procedures. Any fecal samples that did not yield E. coli growth by completion of the workflow were labeled “no E. coli growth” and were not used. After bacterial growth was obtained, all samples followed the same workflow regardless of host type or source (Fig. 2).

Fig. 2.

Diagram of sample handling workflow

Identification and purification of E. coli isolates

Colonies were identified using matrix-assisted laser desorption ionization – time of flight (MALDI-TOF) mass spectrometry. MALDI-TOF applies a laser to a prepared bacterial sample in a vacuum tube, and the protein peptides separate and hit a detector, creating a unique protein fingerprint [19]. One representative colony of each morphologically unique type was spotted onto a clean well of a Bruker MALDI-TOF reusable 96-well metal target plate using a sterile toothpick. Each target spot was overlayed with 1 µl of 70% formic acid and allowed to dry. Then, each spot was overlayed with 1 µl of α-Cyano-4-hydroxycinnamic acid (HCAA) Matrix (Bruker Daltonik GmbH, Bremen, Germany) and allowed to dry. The spots were analyzed in a Bruker microflex LT MALDI-TOF using MBT Compass 4.1 software (Bruker Daltonik GmbH, Bremen, Germany), which runs flexControl version 3.4 (Bruker Daltonics GmbH & Co. KG, Bremen, Germany) to acquire spectra. MBT Compass provides an identification and a score for each sample by comparing the generated sample spectra to a Bruker reference database. A score above 2.0 indicates that the software is confident in both the genus and species identification. A score between 1.7 and 1.9 indicates that the software is confident in the genus identification. A score of under 1.7 indicates no reliable identification could be made for that sample [20]. MALDI-TOF can have difficulty differentiating between E. coli and Shigella spp. Thus, a colony was considered positively identified as E. coli if it was pink on MacConkey agar, which indicates lactose fermentation, and received a MALDI-TOF identification of E. coli with a score above 2.0.

If E. coli growth was present on more than one plate type for the same sample, preference was given to growth on MacConkey agar because it is selective and differential for E. coli. If more than one visually unique colony morphology type on a single plate was identified as E. coli with a score of over 2.0, a single colony of the most abundant morphology was selected for further analysis. One isolate was chosen from each animal, except for one animal from the Northeast district, where two isolates were chosen from the same animal. The isolates appeared phenotypically different and had different resistance profiles, so they were both included in the final sample set and treated the same as the other isolates.

A sterile loop was used to streak the same selected colony used for MALDI-TOF identification onto Columbia Blood Agar (CBA). The CBA was incubated in a 37 °C incubator for 24 h. The resulting pure bacterial growth was transferred to Hardy Diagnostics freezing media and stored at -80 °C. If only one unique colony morphology type was present on a plate, colonies were saved directly from the original plate.

For further processing, each pure E. coli isolate was thawed, and a sterile swab was used to sub-culture from the freezing media to a CBA plate. The CBA was incubated for 24 h in a 37 °C incubator. The resulting pure stock was used to perform susceptibility testing and to generate a mass spectrometry Main Spectra Profile (MSP) for the MALDI-TOF database.

Broth microdilution antimicrobial resistance phenotype testing

Broth microdilution testing was performed to determine the minimum inhibitory concentration (MIC) of each isolate. Sensititre™ BOPO7F plates were used (Thermo Fisher Scientific, Waltham, Massachusetts, U.S.A.). A sterile wooden transfer device was used to transfer bacteria to sterile water to reach a turbidity of 0.5 McFarland standard. Then, 10 µL of the inoculated water was transferred to Cation Adjusted Mueller-Hinton Broth w/ TES (Thermo Fisher Scientific, Waltham, Massachusetts, U.S.A.). Fifty microliters of the inoculated broth was transferred to each well of the Sensititre™ BOPO7F plate, and the plate was incubated at 37 °C for 18–24 h. The Sensititre™ plates were read using the Biomic^®, Clinical Microbiology Laboratory, and Antibiotic Disk Diffusion Susceptibility Test System Version 7.9.2.2022 (Giles Scientific, Santa Barbara, California, U.S.A.). The plates were read on the “No UV” setting, and the MIC results were compiled into a table and visualized as bar charts, which were constructed in R version 4.3.3 using R Studio version 2023.06.1 + 524 [21].

Statistical analysis

For each antimicrobial, a logistic regression model was constructed in R version 4.3.3 using R Studio version 2023.06.1 + 524 [21]. To estimate the odds of resistance, each MIC result was categorized as resistant or sensitive using the best available breakpoints (Table 3), with resistant being considered a “success” for the purpose of the model. An intermediate result indicates the isolate is not fully sensitive to that antimicrobial. To avoid underestimating the odds of resistance, intermediate MIC results were classified as resistant for logistic regression. For example, isolates were only considered sensitive to ampicillin if they were inhibited by a concentration of less than or equal to 8 µg/mL (Table 3). All other results were considered resistant.

Table 3.

Enterobacteriaceae resistance breakpoints and sources [22]

¹Breakpoints used to classify MIC results for each antimicrobial

²Isolates with an intermediate (if available) or resistant MIC were categorized as resistant for logistic regression. The thick vertical line indicates the cutoff between the resistant and sensitive categories for logistic regression

For ampicillin, tetracycline, and TMP-SMZ, CLSI breakpoints for E. coli were available through the Biomic^®, Clinical Microbiology Laboratory and Antibiotic Disk Diffusion Susceptibility Test System Version 7.9.2.2022 (Giles Scientific) (Table 3). For florfenicol, ceftiofur, and tulathromycin, CLSI breakpoints for E. coli were not available. For florfenicol and ceftiofur, CLSI breakpoints for Salmonella enterica were used, as S. enterica is also a Gram-negative, rod-shaped, facultative anaerobe in the Enterobacteriaceae family [23] (Table 3). No established breakpoints were available for tulathromycin for any enteric bacterial species. To perform statistical analysis, isolates were categorized as sensitive to tulathromycin if they were inhibited by the lowest concentration tested, and all other isolates were categorized as resistant (Table 3) [22].

Once the E. coli isolates were categorized as sensitive or resistant, a logistic regression model using host type and agricultural district as categorical predictor variables was constructed for each antimicrobial. There was no ability to estimate variable interactions because of the n = 1 sample size for wildlife from the southeast agricultural district.

Likelihood ratio tests (LRT) were used to determine the significance of the difference between host types and between agricultural districts in the logistic regression models. If the difference between agricultural districts was found to be not significant, it was dropped from the model. The model was then refit using only host type as a categorical predictor. For all tests, significance was evaluated using an alpha level of 0.05.

Mass spectrometry

To evaluate protein expression, a mass spectrometry Main Spectra Profile (MSP) was generated for each isolate. An MSP is a composite spectrum created from 18 or more spectra of the same sample. All MSPs were created following Bruker’s sample preparation protocol [20]. Briefly, after 24 h of incubation, a loop of E. coli from pure culture was added to a 1.5 mL microcentrifuge tube with 1.2 mL 75% ethanol prepared with ultra-pure water. The suspension was centrifuged at 13.4 × 1000 RPM for 2 min to form a pellet. The supernatant was removed, and the pellet was air-dried for approximately 5 min. 25 µL 100% acetonitrile and 25 µL 70% formic acid were added to the pellet and it was vortexed and centrifuged again. 1 µL of the resulting supernatant was spotted onto 8 spots on a metal MALDI target plate and allowed to dry. Each spot was overlayed with 1 µL of HCAA Matrix and allowed to dry, before placing the target plate in the MALDI-TOF instrument.

The software flexControl (version 3.4, Bruker Daltonik GmbH, Bremen, Germany) on AutoXecute (MBT AutoX method) was used to create 3–4 spectra from each of the eight spots containing the sample. Once generated, all the spectra from a single sample were loaded into flexAnalysis (version 3.4, Bruker Daltonik GmbH, Bremen, Germany) for analysis. Background subtraction and smoothing settings were applied, and any flat line spectra, unusually low or high peaks, or peaks inconsistent with the rest of the spectra were manually removed until a minimum of 18 good-quality spectra remained. The cleaned spectra were loaded into MALDI Biotyper^® Compass Explorer 4.1 (Bruker Daltonik GmbH, Bremen, Germany), where Bruker’s default settings were used to create an MSP for each isolate.

The MSPs were used to create dendrograms showing relatedness in protein expression between isolates. ATCC reference E.coli MSPs from the Bruker MSP database were included as controls. Dendrograms were created using the default Creation Standard Method, which uses the correlation distance measure, the average linkage measure, and a score threshold value for a single organism of 300.

Results

Broth microdilution susceptibility testing

The MIC results for ampicillin, tetracycline, TMP-SMZ, ceftiofur, florfenicol, and tulathromycin were organized in a table (Table 4) and visualized as bar charts (Fig. 3, A-F). For each antimicrobial tested, more resistant isolates originated from livestock than from wildlife. Resistance levels did not reflect cattle stocking densities, which are highest in the West district and lowest in the South Central district.

Table 4.

Frequency of E. coli isolates from wildlife and livestock by MIC (μg/mL)

¹Thick vertical lines indicate resistance breakpoints used for logistic regression, with sensitive MICs to the left of the thick lines and resistant MICs to the right. Isolates with an intermediate MIC were considered resistant for logistic regression. Dashes indicate concentrations not tested for that antimicrobial

²Breakpoints from Biomic^®, Clinical Microbiology Laboratory and Antibiotic Disk Diffusion Susceptibility Test System Version 7.9.2.2022 (Giles Scientific) for enteric E. coli from cattle

³Breakpoints based on Salmonella enterica breakpoints in VET01S, 6th Edition, Table 2A

⁴Breakpoints for tulathromycin and E. coli are missing. For this study, isolates inhibited by the lowest antimicrobial concentration available on the plate are considered “sensitive,” and all isolates not inhibited by the lowest antimicrobial concentration are considered “resistant.”

Fig. 3.

MIC results of livestock and wildlife E. coli isolates for each antimicrobial tested (A-F)

Logistic regression

For ampicillin, LRT found no significant difference in the odds of resistance between agricultural districts (χ²₄ = 4.9636, p = 0.29). The model was therefore refitted with only host type. There is a significant difference in the odds of resistance between wildlife and livestock (χ²₁ = 20.251, p = < 0.001). Specifically, there is an estimated 89.6% (95% confidence interval (CI) 64.3% − 97.0%) decrease in odds of resistance to ampicillin when the host type is wildlife versus when the host type is livestock (Table 5).

Table 5.

Odds of resistance decrease for E. coli isolated from wildlife versus livestock

Antimicrobial	Decrease in odds of resistance for wildlife vs. livestock	95% CI
Ampicillin	89.6%	64.3% − 97.0%
Tetracycline	99.1%	93.0% − 99.9%
TMP-SMZ	81.4%	14.5% − 96.0%
Ceftiofur	Difference not possible to estimate	-
Florfenicol	66.2%	23.1% − 85.1%
Tulathromycin	No significant difference	-

For tetracycline, LRT found no significant difference in the odds of resistance between agricultural districts (χ²₄ = 2.225, p = 0.69). The model was then refitted with only host type. There is a significant difference in the odds of resistance between wildlife and livestock (χ²₁ = 79.608, p = < 0.001). Specifically, there is an estimated 99.1% (95% CI 93.0% − 99.9%) decrease in odds of resistance to tetracycline when the host type is wildlife versus when the host type is livestock (Table 5).

For TMP-SMZ, LRT again found no significant difference in the odds of resistance between agricultural districts (χ²₄ = 7.1180, p = 0.13). The model was then refitted with only host type. There is a significant difference in the odds of resistance between wildlife and livestock (χ²₁ = 6.421, p = 0.011). Specifically, there is an estimated 81.4% (95% CI 14.5% − 96.0%) decrease in odds of resistance to TMP-SMZ when the host type is wildlife versus when the host type is livestock (Table 5).

A logistic regression model could not be constructed for ceftiofur. No wildlife isolates were resistant in any of the agricultural districts, and 3% (n = 100) of all livestock isolates were resistant (Fig. 3; Table 3). The lack of variability for the wildlife host type does not allow the model to be estimated, and we therefore cannot estimate changes in odds of resistance.

For florfenicol, LRT found no significant difference in the odds of resistance between agricultural districts (χ²₄ = 7 3.5981, p = 0.46). The model was refitted with only host type. There is a significant difference in the odds of resistance between wildlife and livestock (χ²₁ = 7.4272, p = 0.006). Specifically, there is an estimated 66.2% (95% CI 23.1% − 85.1%) decrease in odds of resistance to florfenicol when the host type is wildlife versus when the host type is livestock (Table 5).

For tulathromycin, LRT found no significant difference in the odds of resistance between agricultural districts (χ²₄ = 5.2487, p = 0.26). When the model was refitted with only host type as a predictor variable, LRT found no significant difference in the odds of resistance between host types (χ²₁ = 2.0846, p = 0.15).

MALDI-TOF mass spectrometry

AMR phenotypes were overlaid on the protein dendrograms. Dendrograms of isolates from each host type revealed no distinct clustering by AMR profile (Fig. 4). For each host type, resistance was distributed evenly throughout the trees and not confined to certain nodes.

Fig. 4.

Dendrogram of livestock (A) and wildlife (B) E. coli MSPs with corresponding AMR

Discussion

From veterinary health and One Health perspectives, the high number of livestock isolates that are resistant to commonly used antimicrobials is concerning. Also concerning is that 34 of the 100 livestock isolates were resistant to two or more antimicrobials (Fig. 4). It may be assumed that the main source of AMR in cattle is selective pressure from the direct use of antimicrobials in livestock production. However, there may be more complex mechanisms involved than simple selective pressure, as other researchers have found that cattle never directly exposed to antimicrobials may still harbor high levels of AMR bacteria [24, 25].

Although the number of resistant isolates from wildlife was lower than the number from livestock for each antimicrobial tested, wildlife E. coli are not free from AMR. This begs the question, “Where is resistance in wildlife coming from?” The literature offers a few theories. Other researchers have found that, even in the absence of selective pressure, resistance genes may be maintained in untreated populations, especially if they are contained on plasmids that provide other advantages to the bacteria and have low energetic costs to maintain [26]. It has also been suggested that bacteria in wildlife are exposed to selective pressure via the environment [27, 28]. This exposure may originate from sources such as agricultural runoff, landfill leachates, and municipal, manufacturing, and hospital wastes [27, 28].

The hypothesis that a higher number of AMR isolates would be found in livestock was supported for tetracycline, ampicillin, TMP-SMZ, and florfenicol, but not for tulathromycin. The results for ceftiofur could not be statistically evaluated. Resistance to ceftiofur was relatively low, with 3% (n = 100) of livestock isolates and no wildlife isolates exhibiting phenotypic resistance. Over half (57%, n = 100) of the livestock isolates showed phenotypic resistance to tetracycline (Table 2), which most likely stems from the long history of tetracycline use in cattle for therapeutic and growth promotion purposes [29]. However, only one resistant isolate was obtained from randomly sampling wildlife (Table 2). Similarly, 27% (n = 100) of the livestock isolates had phenotypic resistance to ampicillin, and less than 4% (n = 81) of wildlife isolates were resistant (Table 2). The low number of wildlife isolates resistant to tetracycline and ampicillin is an encouraging finding, as it indicates that high levels of resistant isolates in livestock do not necessarily lead to high levels of resistant isolates in wildlife.

Similarly, significantly more livestock isolates than wildlife isolates were resistant to florfenicol. The higher number of resistant livestock isolates is most likely attributable to florfenicol’s use for the treatment of bovine respiratory disease. The difference between host types was also significant for TMP-SMZ, with livestock isolates having significantly higher odds of resistance. Although TMP-SMZ has no on-label use in cattle, the higher number of resistant livestock isolates could stem from treatment with other sulfonamides or from off-label use of TMP-SMZ.

Tulathromycin is the newest antimicrobial tested, and resistance in both livestock and wildlife was low (Fig. 3; Table 4). There was no significant difference in the number of tulathromycin-resistant isolates between livestock and wildlife. Based on our findings from the other antimicrobials tested in this study, resistance may remain low in wildlife, even if resistance to tulathromycin eventually rises in livestock.

Agricultural district was not found to significantly affect the odds of resistance to any of the antimicrobials tested. Differences between agricultural districts may not be large enough to affect E. coli AMR. The average stocking density in each district is relatively low, ranging from 0.029 to 0.085 cattle per farm acre, and does not vary greatly throughout the state [14]. Additionally, the movement of animals between agricultural districts was not controlled for or measured in this study. Cattle may be moved throughout the year, and free-ranging wildlife migrate with the seasons. The movement of animals may have contributed to the weakness of using agricultural districts to predict resistance.

Dendrogram clustering of MSPs was not able to detect AMR differences based on protein profile. Since isolates were not exposed to antimicrobials during sample preparation for MSP creation, constitutively expressed resistance proteins may not produce major enough peaks to affect dendrogram clustering. The MALDI-TOF algorithm uses only the most major peaks to determine isolate relatedness.

This study provides a valuable basis for future AMR studies in Wyoming, but further studies with larger sample sizes would allow more conclusions and generalizations to be made. Overall, this study sampled a small number of livestock and wildlife, relative to their respective populations within the state. Although the n = 1 sample size for Southeast wildlife E. coli is not ideal, the one sample obtained was included to provide the most comprehensive data set possible. Future work would benefit from a larger sample size to equalize the number of samples from each agricultural district, capture more variation, and enable further statistical analysis, such as examining variable interactions. A larger sample size may also reduce the size of the confidence intervals in logistic regression.

In addition to expanding the sample size, there are a multitude of other future directions for this study. One would be to seek out livestock fecal samples directly from producers so healthy versus clinically sick cattle could be compared, and information such as antimicrobial treatment history could be obtained. A bias of this study was that 95 (n = 100) of the livestock isolates came from diagnostic cases submitted to WSVL from presumably sick animals, while wildlife isolates were assumed to be collected from apparently healthy animals. Sourcing samples directly from producers would avoid this bias in future studies. It would also be beneficial to test multiple isolates from each animal and not simply the most abundant, as one animal may harbor multiple E. coli isolates, each with its own AMR phenotype.

Another future direction would be to integrate livestock and wildlife AMR data with data from human and wastewater samples to provide a more complete One Health view of AMR. Future studies could also include samples from a wider range of host species, such as domestic sheep and elk. With the addition of more samples from a wider range of host species, samples could be separated by species and not merely categorized as livestock or wildlife. This study could also be expanded to include other states.

Conclusions

This study provides the first insights into AMR in E. coli from Wyoming’s wildlife and livestock. The hypothesis that more resistant isolates would be found in livestock E. coli than in wildlife E. coli was supported for four of the five antimicrobials that could be statistically evaluated. In this dataset, agricultural district did not significantly affect the odds of resistance for any of the antimicrobials evaluated. This study also found that simple MALDI-TOF mass spectrometry, though a valuable tool for identifying bacteria, was not effective in predicting phenotypic resistance.

For all antimicrobials tested, a low number of wildlife isolates exhibited resistance, even to those antimicrobials for which a high number of resistant isolates were found in livestock. While this is an encouraging finding, the increased resistance in livestock E. coli, and the presence of some wildlife E. coli isolates resistant to multiple drugs is concerning. The results of this study provide a valuable first snapshot into the prevalence of AMR in Wyoming’s wildlife and livestock. Our findings highlight the need for increased multi-angled AMR research and surveillance efforts in livestock and wildlife in the American West, and for responsible antimicrobial use in livestock.

Abbreviations

AMR

Antimicrobial Resistance

USDA-NASS

United States Department of Agriculture–National Agricultural Statistics Service

CBA

Columbia Blood Agar

WGFD

Wyoming Game and Fish Department

WSVL

Wyoming State Veterinary Laboratory

HE agar

Hektoen enteric agar

MALDI-TOF

Matrix–Assisted Laser Desorption Ionization–Time of Flight

MIC

Minimum inhibitory Concentration

TMP-SMZ

Trimethoprim sulfamethoxazole

CLSI

Clinical Laboratory Standards Institute

LRT

Likelihood Ratio Test

MSP

Main Spectra Profile

ATCC

American Type Culture Collection

CI

Confidence interval

Authors’ contributions

O. B. finalized the study design, performed the experiments, analyzed the data, and prepared the manuscript. H. K. helped with laboratory protocols and obtaining samples. K. K. helped develop the study design and perform broth microdilution testing. K. B. helped perform broth microdilution testing and create MSPs. J. S. helped with data analysis. K. S. developed the original idea and helped with study design and manuscript preparation. All authors approved the final manuscript.

Funding

Y Cross Ranch Endowment, Riverbend Endowment.

Data availability

The dataset used for statistical analysis is available in the supplementary information file.

Declarations

Ethics approval and consent to participate

No permissions were necessary to collect the data used in this study.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.