Comprehensive Analysis of E. coli, Enterococcus spp., Salmonella enterica, and Antimicrobial Resistance Determinants in Fugitive Bioaerosols from Cattle Feedyards

Abstract

Antimicrobial use in food animals selects for antimicrobial-resistant (AMR) bacteria, which most commonly reach humans via the food chain. However, AMR bacteria can also escape the feedyard via agricultural runoff, manure used as crop fertilizer, and even dust. A study published in 2015 reported AMR genes in dust from cattle feedyards; however, one of the study’s major limitations was the failure to investigate gene presence in viable bacteria, or more importantly, viable bacteria of importance to human health. Our main objective was to investigate the presence and quantity of viable bacteria and antimicrobial-resistant (AMR) determinants in fugitive bioaerosols from cattle feedyards in the downwind environment. Six bioaerosol sampling campaigns were conducted at three commercial beef cattle feedyards to assess variability in viable bacteria and AMR determinants associated with geographic location, meteorological conditions, and season. Dust samples were collected using four different sampling methods, and spiral plated in triplicate on both non-selective and antibiotic-selective media. Colonies of total aerobic bacteria, Enterococcus spp., Salmonella enterica, and Escherichia coli were enumerated. Viable bacteria, including AMR bacteria, were identified in dust from cattle feedyards. Bacteria and antimicrobial resistance genes (ARGs via qPCR) were mainly found in downwind samples. Total suspended particles (TSPs) and impinger samples yielded the highest bacterial counts. Genes encoding beta-lactam resistance (bla_CMY-2 and bla_CTX-M) were detected while the most common ARG was tet(M). The predominant Salmonella serovar identified was Lubbock. Further research is needed to assess how far viable AMR bacteria can travel in the ambient environment downwind from cattle feedyards, to model potential public health risks.

1. Introduction

The continued growth of the global human population has led to increased demand for food and more intensive livestock production [1]. Agricultural animal feeding operations (AFOs) play a vital role in raising livestock for food production. The United States (USA) is the largest beef producer in the world, with beef cattle production occurring on 30% of all USA. farms. Within the U.S, Texas has the largest number of beef cattle, totaling 4,685,000 head [2]. Animals in AFOs may experience stress due to their being reared close to each other, behavioral changes, and altered diets when compared to their extensive rearing origins, all of which increase the risk of disease development and spread [3,4,5,6]. This, in turn, can lead to a greater use of antibiotics to maintain animal health, welfare, and production levels [7].

The use of antimicrobials for the prevention, control, and treatment of infectious diseases in livestock creates selection pressures for antimicrobial resistant bacteria [8,9,10,11]. Resistant bacteria can be transferred from livestock to humans not only via food and animal products, but also through the environment [12,13,14,15,16]. Antimicrobial-resistant bacteria can also escape the feedyard environment through runoff, manure used as fertilizer, or particulate matter in the air [12,13,14,15,16]. There have been multiple research studies on environmental antimicrobial resistant bacteria arising from AFOs [17,18,19]; only a few studies have focused on particulate matter as the pathway of emissions [20,21]. The airborne transmission of AMR bacteria has become an increasing concern as researchers investigate potential sources of bacterial contamination of vegetables and leafy greens [22,23]. Additionally, there are public health concerns regarding the proximity of residential areas to Open Lot Livestock Facilities (OLLFs) and the potential for dust from the feedlots to travel to these residential areas [24,25].

Hoof action and wind scouring on dry animal manure pack in feedlots generate fine particles that can be small enough to form aerosols [24,26] and bacteria can be attached to those particles. In cattle feedyards, the maximum ground-level concentration of dust typically occurs during the evening. Drier pen surfaces, increased animal activity, associated hoof action [27] along with decreased winds, decreased vertical atmospheric dispersion, decreased boundary-layer mixing height, and increased atmospheric stability create favorable conditions for the evening dust peak (EDP) [28]. Airborne bacterial suspension time depends on several factors, such as atmospheric conditions, attachment to particles, and the bacterial species themselves [29,30]. Inhalation and deposition of particulates within the respiratory tract are complex; generally, particulates larger than 10 μm are deposited in the nasal passages, particles between 5–10 μm are primarily deposited in the upper respiratory tract, and particles smaller than 5 μm can reach and settle in the lungs [31]. Therefore, human exposure to airborne bacteria from bioaerosols generated by AFOs may be a concern [21]. To accurately assess the associated risks for human health, we need to determine the quantities of viable bacteria in bioaerosols downwind from cattle feedyards and to further determine associations between viable bacteria and resistance to antimicrobials that are important for public health. A study published in 2015 raised concerns about AMR bacteria escaping cattle feedyards through particulate matter in the air [20]. One of the limitations of that study was that the authors did not verify that the AMR genes were present in viable bacteria [32] and their co-location within the same. The presence of AMR genes is not a risk to public health unless the genes are detected in viable bacteria capable of causing infections in humans. Even then, the risk exists only if a bacterial infection arising is resistant to the same antibiotic chosen by a physician. Additionally, many of the resistance genes reported are naturally found in bacteria in the environment, conferring innate resistance to many classes of antibiotics.

Free DNA fragments, including those that may carry AMR genes from dead bacterial cells, may also be found in the environment and can be taken up by other bacteria via transformation and incorporated into their genome [33]. Transformation was the first discovered mechanism of horizontal gene transfer (HGT) and involves the transfer of DNA to viable bacteria that are in a competent state [34]. Conjugation among two viable bacteria involves specialized and independent replicating genetic elements, like plasmids or integrated conjugative elements (ICEs), including transposons (CTns). Conjugation facilitates the transference of the plasmid or ICE to the same or different species of bacteria [35]; consequently, plasmid-mediated transmission is the most common mechanism of HGT [36]. Although DNA containing AMR genes can be incorporated into bacteria in the environment, the presence of viable bacteria is necessary for this to occur through a HGT mechanism [37].

Various methods, like filtration and impingement, have been employed to measure dust concentrations in feedlots [38], and to detect ARGs [20], yet little is known about the most effective techniques for sampling bioaerosols while preserving bacterial viability. Each bioaerosol sampling method offers distinct advantages and disadvantages, with varying levels of efficacy depending on the specific application. Those methods are classified as active and passive methods according to the type of collection [39]. Active methods (filtration and impingement) use a mechanical component to actively capture particles from the air, while passive methods (deposition) collect particles by allowing them to settle naturally onto a collection surface via gravity.

Filtration methods most often employ a dry filter of varying composition that selectively collects different particulate sizes: Total Suspended Particulates (TSPs) have an aerodynamic equivalent diameter (AED) of up to 45 μm; along with a subset of this, particulate matter of 10 μm (PM₁₀), represent two examples of different particulate sizes targeted by specific collection devices and filter types. Liquid impingers collect particles greater than 2 μm in diameter using a liquid collection medium, which helps reduce damage to microorganisms resulting from physical impact and desiccation [40,41]. Deposition samplers, which measure dry-deposition flux rather than airborne concentration, gather primarily coarse particles of varying sizes, including multiple viable and non-viable bacterial sizes ranges [42].

While considerable research has been conducted on bioaerosol sampling at indoor AFOs, particularly in poultry and swine facilities [25,43,44,45,46,47], similar studies are scarce in the outdoor ambient environments of cattle feedlots. Little is known about (a) the viability of bacteria in particulate matter originating from AFOs and (b) whether AMR genes found in dust are harbored in viable bacteria of concern to human health. The primary aim of this study was to investigate the presence and quantity of commensal and pathogenic bacteria and AMR determinants in bioaerosols upwind and downwind of commercial beef cattle feedyards using three different sampling methodologies (filtration, impingement, and deposition). Additionally, any detected viable bacteria were characterized using phenotypic and genotypic techniques to determine sequence types, serovars, antimicrobial susceptibility, and antimicrobial resistance genes. The outcome of this effort was to improve the understanding of the public health risks associated with fugitive bioaerosols from cattle feedyards.

2. Materials and Methods

2.1. Experimental Design

Three commercial feedyards (FY1, FY2, and FY3) with one-time capacities ranging from 18,000 to 50,000 head of cattle in the high plains of Texas, USA, were sampled for bioaerosols using three different methods. To control dust, all three feedyards used manure harvesting while only FY3 employed a water sprinkler system. Six sampling campaigns were conducted: Campaign 1 (July 2018), Campaign 2 (January 2019), Campaign 3 (February 2019), Campaign 4 (May 2019), Campaign 5 (July 2019), and Campaign 6 (September 2019). Samples were collected for 2.5 h to 4 h (6:30 pm to 10:00 pm) during the EDP. The timing of the sampling was varied depending on the length of the EDP and sampling conditions.

2.2. Bioaerosol Sampling Methods

Downwind samplers were placed along the projected dust plume based on the prevailing wind conditions. Campaigns each included three downwind sampling locations, one at the center of the plume on the property line, and two near the edges of the plume. Downwind samplers were oriented longitudinally at different distances from the property line. Additionally, one sampler was located upwind from the feedyard to represent background aerosols. A schematic of the distribution of the samplers in the upwind and downwind directions is shown in Figure S1. The average distances between the upwind and downwind air monitors were approximately 1.25 km (FY1), 0.5 to 0.8 km (FY2), and 0.75 km (FY3) for the three feedyards.

Hardware

Three distinct sampling methods were used to compare bioaerosols and quantities of viable bacteria. Sampling durations were computed from the optimal flow rate for each sampling device to keep the air volume sampled constant across sampling events. Corresponding sampling durations by campaign and sample type are specified in

Table 1. Sampling duration by campaign and sample type.

Campaign	Sample Type	Δt (min) 4	Δt (h) 4
1	Impinger, QF 1 (PM10 2, TSP 3), and marbles	215–249	3.58–4.15
2	QF (TSP)	149–215	2.48–3.58
3	Impinger, QF (PM10, TSP), and marbles	184–192	3.07–3.20
4	Impinger, QF (PM10 and TSP), and marbles	142–148	2.37–2.47
5	Impinger, QF (PM10 and TSP), and marbles	no recorded	no recorded
6	Impinger, QF (PM10 and TSP) and marbles	195–205	3.25–3.42

¹ QF: Quartz filter. ² PM₁₀: particulate matter with aerodynamic diameter <= 10 μm. ³ TSP: total suspended particulate. ⁴ Δt (): Sampling measurement time in minutes (min) and hours (h).

. Before each sampling, a field and a trip blank were prepared for each sample type to serve as controls using 1X PBS Gibco^® phosphate-buffered saline (Thermo Fisher Scientific, Gaithersburg, MD, USA). The field blank was exposed to the environment during sample collection and processed in the laboratory as an environmental sample. The trip blank was maintained within the sterilization pouch and was not exposed to the environment during sample collection. Additional measurements collected during the sampling included near-real-time mass concentrations of particulate matter (PM), near-real-time particle counts, and standard meteorological data (air temperature, wind speed/direction, relative humidity, solar radiation, barometric pressure, and precipitation).

Critical-flow liquid impingers (BioSampler, 20 mL) (SKC, Inc., Eighty Four, PA, USA) containing a previously validated solution of 95% 1X PBS and 5% glycerol were used to capture bioaerosol samples. The impinger consisted of (a) a glass vessel containing approximately 20 mL of a sterile liquid, and (b) a glass air inlet with three roughly tangential nozzles that induced swirling in the sampling liquid. The swirling action reduced the mechanical violence of particle impact on the liquid surface. The pump flow rate from the BioSampler was adjusted to induce gentle but easily visible swirling without the liquid splashing up the glass vessel wall, and the airflow rate was measured with an integrated rotameter. Following the sampling interval, the sampling liquid and its suspended particles were extracted with a sterile syringe and injected into a sterile, 50 mL conical tube (Falcon, Corning, NY, USA).

The filtration method is an active method that combines impaction and filtering. The filtration method included high-volume gravimetric aerosol samplers for both total suspended particulate (TSP) (Staplex^® TSP; The Staplex^® Company, Brooklyn, NY, USA) and particles ≤ 10 μm AED (PM₁₀) (Andersen Model RAAS10-100 PM₁₀ Single Channel PM₁₀ Sampler; Manual Reference Method: RFPS-0699-130) using 20.3 × 25.4 cm A/E Glass Fiber Filters (quartz filter = QF) (Cytiva, Marlborough, MA, USA) following the Environmental Protection Agency (EPA) guidelines [48]. Tapered Element Oscillating Microbalance (TEOM) (Series 1400a, Thermo Fisher Scientific, East Greenbush, NY, USA) were used to measure particle mass concentrations in the air.

Deposition samples were collected using a marble sampling technique. Marble sampling was based on passive methods previously described but with modifications [49]. Sterilized opaque, white glass marbles (0.91 kg) (Moon Marble Company, Bonner Springs, KS, USA) were placed in a sterilized 22.9 cm aluminum pie plate and set out for the sampling duration. After exposure, the marbles and pan were rinsed using 1X PBS.

Liquid from the impingers and rinsate from the marbles were dispensed into sterile 50 mL conical tubes (Falcon); meanwhile, the quartz filters (QFs) were packed in a sterilization pouch. All of the samples were stored at 4 °C, packaged with ice, and shipped overnight to the laboratory at Texas A&M University in College Station, TX for further processing and microbiological analysis.

2.3. Microbiological Processing

The liquid samples (impinger liquid aliquot and marble rinsate) and quartz filters (TSP and PM₁₀) from the three feedyards were processed to determine both bacterial prevalence and bacterial quantification. The liquid samples were vortexed for 20 s before processing, if there was more than one conical tube per sample, the liquid from both tubes was combined into a single conical tube and vortexed. A 20 mL aliquot of the mixed liquid was used for microbiological processing. Each packaged quartz filter was opened in a sterilized biological safety cabinet, removed from the envelope with sterile forceps, and placed into a sterilized glass 22.5 × 33 cm Pyrex^® pan (Pyrex, South Greencastle, PA, USA). A total of 200 μL of 1X PBS was used to wash the filter. Each glass pan, with a filter and 1X PBS, was covered and gently shaken by hand for 1 min, then held at room temperature for 2 h. After this holding period, the liquid was collected with a sterile serological pipette (VWR, Radnor, PA, USA) and distributed into four 50 mL conical tubes (Falcon) that were centrifuged at 4700 rpm for 3 min. After centrifugation, 25 mL of each sample from the top of each tube was discarded. The remaining 25 mL from each of the two falcon tubes were combined to obtain 100 mL in total. The suspension of one of the tubes was used for microbiological processing. The remaining liquid was aliquoted into 5 mL polypropylene culture tubes (VWR): one tube with 50% glycerol at a 1:1 ratio for microbiological analysis, and a second tube without glycerol for molecular analysis. Both polypropylene tubes were stored at −80 °C until further use.

2.3.1. Crude Bacterial Quantification

All samples were processed in triplicate. Fifty microliters from each sample (including the blank controls) were spiral plated using an Eddy Jet^® 2 spiral plater (Neutec Group Inc., Farmingdale, NY, USA) with the E-Mode 50 μL setting onto Tryptic Soy Agar (TSA) (Difco, Becton Dickinson, Franklin Lakes, NJ, USA) for total aerobic bacterial counts (colony forming units—CFU). Three types of MacConkey Agar (MAC) (Difco) were used for E. coli quantification; plain (MAC), MAC supplemented with ceftriaxone (MAC-AXO) (Sigma-Aldrich, St. Louis, MA, USA) at 4 μg/mL, and MAC supplemented with tetracycline (MAC-TET) at 16 μg/mL. Three types of m-Enterococcus agar (ME) (Difco) were used for Enterococcus spp. quantification; plain ME, ME supplemented with tetracycline at 16 μg/mL (ME-TET) (Sigma-Aldrich), and ME supplemented with erythromycin at 8 μg/mL (ME-ERY) (Sigma-Aldrich). All antimicrobial supplemented plates were prepared at the Clinical and Laboratory Standards Institute (CLSI) human clinical breakpoint concentrations for bacterial resistance [50]. The TSA and MacConkey plates were incubated at 37 °C for 18 h, and the ME plates were incubated at 42 °C for 48 h. After incubation, all aerobic bacteria from TSA, presumptive E. coli (pink, lactose fermenting) colonies from MAC, and Enterococcus spp. (dark red to maroon with a cream halo) colonies from ME, all based on phenotypic characterization, were counted using the automated Flash & Go colony counter (Neutec Group Inc., Farmingdale, NY, USA).

Additionally, 1 mL of each sample type (impinger, marbles, quartz filter) was inoculated onto three different CompactDry™ plates in triplicate. CompactDry™ Total Count (CDTC) (all CompactDry™ products: Hardy Diagnostics, Santa Maria, CA, USA) for the aerobic bacterial counts, and CompactDry™ Escherichia coli (CDEC) for E. coli and coliforms counts. The CDTC and CDEC CompactDry™ plates were incubated at 37 °C for 18 h. For Salmonella, 1 mL of the sample was inoculated into 9 mL of trypticase soy broth (TSB) (Difco) and incubated at 37 °C for 18 h. Following incubation, 10 μL of the TSB culture was inoculated onto the CompactDry™ Salmonella (CDSL) plate and incubated at 42 °C for 24 h. Colonies were counted using a light box with a grid sheet (1 cm × 1 cm). Specifications on the CompactDry™ plates, incubation time and temperature, bacteria cultured, and morphology of the presumptive colonies are described in

Table 2. Specifications for compact dry plates.

Plate Type	CDTC 1	CDEC 2	CDSL 3
Bacteria	Total Aerobic	Coliforms and E. coli	Salmonella
Temperature (°C)	37 °C	37 °C	42 °C
Incubation time (h)	18	18	24
Expected phenotypic results	Most colonies will be red	Pink and purple colonies are coliforms and blue-colored colonies are E. coli	Green or blue colonies with and without the black center on a yellow background

¹ CDTC= CompactDry™ Total Count. ² CDEC= CompactDry™ Escherichia coli count. ³ CDSL = CompactDry™ Salmonella count.

. Figure S2 displays the morphological characteristics of bacterial colonies cultured on the various media.

2.3.2. Quantitative Real-Time PCR (qPCR) for Antimicrobial Resistance Genes

Community DNA Extraction for qPCR

Community DNA from samples stored without glycerol was extracted using the DNeasy^® PowerSoil ^®Pro Kit (Qiagen, Valencia, CA, USA) following the manufacturer’s protocol with some modifications. Samples were centrifuged at 4700 RPM for 15 min, the supernatant was discarded, and approximately 250 μL of the sediment was used for DNA extraction. The CD1 solution (Qiagen) was prepared by adding 0.3 g of sodium dodecyl sulfate (SDS) (Sigma, St. Louis, MO, USA) to 10 mL of CD1 buffer. Then, 800 µL of the solution (CD1 + SDS) was added to the sample and vortexed. Proteinase K (Qiagen) (40 µL) was added to each tube. The tubes were incubated at 60 °C for 1 h at 900 RPM in a ThermoMixer™ (Eppendorf North America, Enfield, CT, USA), and a heat shock was performed for 10 min at 95 °C. The suspension was cooled at room temperature and 4 µL of RNAse A (Qiagen) was added to the suspension. The total volume was added to one PowerBead pro tube (Qiagen) to mechanically break the cells using the Qiagen TissueLyser system™ (Qiagen) at 25 Hz, for 10 min. The remaining steps of the DNeasy^® PowerSoil ^®Pro Kit protocol was continued on the Qiacube robot (Qiagen), following the manufacturer’s instructions for the DNeasy^® PowerSoil ^®Pro Kit protocol. DNA quality and quantity were measured at 260/280 nm wavelengths on the FLUOrstar Omega microplate reader, program version 3.02 R2 (BMG Labtech, Cary, NC, USA). DNA was quantified using a Qubit 3 fluorometer using the Quant—iT™ Pico Green ™ ds DNA Assay kit (Thermo Fisher Scientific, Gaithersburg, MD, USA), and the DNA was stored at −30 °C for further genotypic analysis.

Primers and Reaction Setup

Quantitative real-time PCR (qPCR) was performed on community DNA from QF and marble samples to identify target gene copies [i.e., 16S rRNA, tet(A), tet(B), bla_CMY-2, bla_CTX-M group 1, bla_CTX-M group 9_, tet(M), and ermB]. The 16S rRNA gene quantities were used as a proxy index of the total bacterial population and to normalize the target genes. The tet(A), tet(B), bla_CMY-2, bla_CTX-M group 1, and bla_CTX-M group 9 genes were used to quantify the antimicrobial resistance genes of importance among Gram-negative bacteria, while the tet(M) and ermB genes were used to quantify antimicrobial resistance genes of clinical importance in Gram-positive bacteria. Primer sets are listed in Table S1. Reaction volumes included 10 µL of Ultrafast Brilliant III SYBR green (Agilent Technologies, Santa Clara, CA, USA), 3 µL nuclease-free water (Invitrogen, Waltham, MA, USA), 2.5 µL each of forward and reverse primers (Integrated DNA Technologies, Coralville, IA, USA), and 2 µL of total community DNA.

The reactions were run on an AriaMx Real-Time PCR system (Agilent Technologies) with thermocycler conditions adjusted for each gene (Table S2). The reaction plates were set up on a QIAgility™ (Qiagen) and samples were run in duplicate. The assays were subjected to PCR performance analysis following the Minimum Information for Publication of Quantitative Real-Time PCR Experiments (MIQE) [51].

Standard Curve Generation

Genomic DNA arising from previously sequenced isolates in our laboratory were used as templates for standard curve generation, including an E. coli NCBI Bioproject PRJNA625742/Biosample SAMN14842426 with genome size of 4,761,741 bp, E. coli NCBI Bioproject PRJNA625742/Biosample SAMN14842436 with genome size of 5,043,385 bp, E. coli NCBI Bioproject PRJNA625290/Biosample SAMN14596646 with genome size of 5,265,300 bp, Enterococcus faecium NCBI Bioproject PRJNA1197524/Biosample SAMN45828174 genome size of 2,926,163 bp, and Enterococcus faecium NCBI Bioproject PRJNA595617/Biosample SAMN45828633 with genome size of 2,946,289 bp. Standard curve templates were prepared by performing ten serial dilutions for each bacterial strain that contained the target gene for each primer pair. Gene copy numbers were estimated using the DNA concentration (ng/µL), Avogadro’s number = 6.022 × 10²³, size of the reference genome (bp), and the molar mass per base pair set as 650 (g/mol)/bp (Equation (1)).

$$\mathrm{c}\mathrm{o}\mathrm{p}\mathrm{y} \mathrm{n}\mathrm{u}\mathrm{m}\mathrm{b}\mathrm{e}\mathrm{r}=\mathrm{a}\mathrm{m}\mathrm{o}\mathrm{u}\mathrm{n}\mathrm{t} \mathrm{b}\mathrm{y} \mathrm{Q}\mathrm{u}\mathrm{b}\mathrm{i}\mathrm{t} \left({\displaystyle \frac{\mathrm{n}\mathrm{g}}{\mathrm{\mu }\mathrm{L}}}\right)\times {\displaystyle \frac{\mathrm{A}\mathrm{v}\mathrm{o}\mathrm{g}\mathrm{a}\mathrm{d}\mathrm{r}\mathrm{o}\mathrm{’}\mathrm{s} \mathrm{n}\mathrm{u}\mathrm{m}\mathrm{b}\mathrm{e}\mathrm{r}}{\left(\mathrm{b}\mathrm{p}\times 650\times 1\times {10}^9\right)}}\times 2 \mathrm{\mu }\mathrm{L}\times \# \mathrm{o}\mathrm{f} \mathrm{g}\mathrm{e}\mathrm{n}\mathrm{e} \mathrm{c}\mathrm{o}\mathrm{p}\mathrm{i}\mathrm{e}\mathrm{s} \mathrm{p}\mathrm{e}\mathrm{r} \mathrm{g}\mathrm{e}\mathrm{n}\mathrm{o}\mathrm{m}\mathrm{e}$$

(1)

2.3.3. Phenotypic Characterization of Bacteria

One presumptive, phenotypically identified bacterial isolate per each selective agar plate was further characterized using biochemical tests. E. coli were confirmed by the indole test, Salmonella were confirmed using agglutination with O-antigen polyvalent antisera (Difco), and Enterococcus spp. were confirmed by the bile esculin test. Additionally, one presumptive E. coli, Enterococcus spp., and Salmonella colony per plate each were plated onto tryptic soy agar (TSA) with 5% sheep blood agar (RemelTM, Lenexa, KS, USA) and incubated at 37 °C for 24 h (E. coli and Salmonella) or 48 h (Enterococcus spp.), respectively, for genus and species confirmation using matrix-assisted laser desorption/ionization-time of flight mass spectrometry (MALDI-TOF) (Bruker Daltonik GmbH, Billerica, MA, USA). A detailed description of the MALDI-TOF methods in our laboratory was previously published [52]. After confirmation of genus and species, isolates were preserved in cryoprotectant beads (Scientific Device Laboratory, Des Plaines, IL, USA) at −80 °C for further characterization.

Antimicrobial Susceptibility Testing (AST)

Antimicrobial susceptibility was performed for all the Salmonella isolates and E. coli resistant to tetracycline or ceftriaxone using the broth microdilution method via the Sensititre^TM system (TREK, Thermo Scientific Microbiology, Oakwood Village, OH, USA) using Sensititre™ NARMS Gram-negative CMV3AGNF plates (TREK) to identify the minimum inhibitory concentration (MICs) of E. coli and Salmonella. A detailed description of the protocol in our laboratory was previously published [53]. The results were interpreted as susceptible, intermediate, or resistant according to Clinical and Laboratory Standards Institute (CLSI) guidelines [50] using SWIN software version 3.4 (TREK) or else breakpoints established by the National Antimicrobial Resistance Monitoring System for enteric bacteria (NARMS) for Salmonella and E. coli (

Table 3. Interpretive categories used for susceptibility testing of Salmonella and E. coli for the 14 antimicrobials included on the CMV3AGNF Trek Sensititre^TM Gram-negative NARMS plate.

Antimicrobial Class	Antimicrobial Agent	Breakpoints (µg/mL)
		Susceptible	Intermediate	Resistant
Aminoglycosides	Gentamicin	≤4	8	≥16
	Streptomycin 1	≤16	N/A	≥32
β-Lactam/β-Lactamase Inhibitor Combinations	Amoxicillin–Clavulanic Acid	≤8/4	16/8	≥32/16
Cephems	Cefoxitin	≤8	16	≥32
	Ceftriaxone	≤1	2	≥4
	Ceftiofur	≤2	4	≥8
Folate Pathway Inhibitors	Sulfisoxazole	256	N/A	≥512
	Trimethoprim–Sulfamethoxazole	≤2/38	N/A	≥4/76
Macrolides	Azithromycin 1	≤16	N/A	≥32
Penicillins	Ampicillin	≤8	16	≥32
Phenicols	Chloramphenicol	≤8	16	≥32
Quinolones	Ciprofloxacin 2	≤0.06	0.12–0.5	≥1
	Nalidixic acid	≤16	N/A	≥32
Tetracyclines	Tetracycline	≤4	8	≥16

¹ Breakpoints were adopted from the CLSI (Clinical and Laboratory Standards Institute) M100-Ed30 document, except for streptomycin and azithromycin, which have no CLSI breakpoint. ² In 2012, the Clinical and Laboratory Standards Institute (CLSI)’s M100-S27 expanded the Minimum Inhibitory Concentration (MIC) range that defines the intermediate susceptibility category for ciprofloxacin. Decreased susceptibility to ciprofloxacin (DSC, MIC >= 0.12 µg/mL) is used as a marker for emerging fluoroquinolone resistance (CLSI, 2017).

). Intermediate isolates were later reclassified as susceptible for statistical analyses. Isolates resistant to three or more classes of antimicrobials were considered multidrug-resistant (MDR).

2.3.4. Genotypic Characterization of Bacterial Isolates

DNA Extraction for Whole Genome Sequencing (WGS)

DNA was extracted from stored Salmonella and a selection of resistant E. coli isolates using the QIAamp 96 DNA QIAcube HT Kit™ (Qiagen) on the QIAcube HT™ instrument (Qiagen). Fresh cultures from cryoprotectant beads were streaked onto tryptic soy agar (TSA) with 5% sheep blood and incubated at 37 °C for 18–24 h. After incubation, one colony per isolate was suspended into 5 mL of trypticase soy broth (TSB) (Difco) and incubated overnight at 37 °C. DNA extraction methods have been previously described in detail [53].

Whole Genome Sequencing (WGS)

WGS was performed using the Illumina MiSeq platform using the Miseq V2 Reagent Kit (500 cycles) with paired-end reads (Illumina, San Diego, CA, USA). Libraries for 32 isolates were multiplexed with the Swift 2s Turbo DNA Library Kit with Adapters (Integrated DNA Technologies), Swift Normalase Kit (Integrated DNA Technologies), and Swift Normalase Combinatorial Dual Indexing Primer Kit (Integrated DNA Technologies) following the manufacturer’s instructions using the Eppendorf Flexlid Mastercycler nexus gradient thermocycler (Eppendorf North America). The quality and quantity of the libraries were checked using the Fragment Analyzer instrument system (Advanced Analytical, Ankeny, IA, USA).

Bioinformatics Analysis of WGS Data

Raw reads (fastq files for reverse and forward reads) were obtained and used for further analysis on the Terra cluster of the High Performance Research Computing system (HPRC) at Texas A&M University. Fastq files were trimmed using Trimmomatic v0.39 [54]. The quality of the raw reads was checked with FastQC v0.11.9 [55]. The results from FastQC were aggregated in MultiQC [56]. Trimmed reads were de novo assembled by Spades v3.14.0 [57]. The assembly quality was assessed on QUAST v5.0.2 [58], which analyzes the number of contigs, and total length of the genome by bp, GC content (%), N50, and L50. The minimum required contig length was set to 250 bp. The SISTR v1.1.1 pipeline was used for serovar identification [59]. Serovar predictions were determined from WGS assemblies by antigen and cgMLST gene alleles using BLAST Sequence type v2.2.30 [59]. Multilocus sequence types (MLSTs) were identified for Salmonella and E. coli isolates using BIO-MLST-Check v2.1.1706216 [57], which identifies the combination of 7 house-keeping genes for each Salmonella (aroC, dnaN, hemD, hisD, purE, sucA, and thrA) and E. coli (adk, fumC, gyrB, icd, mdh, purA, and recA). The genomes were analyzed through Abricate v0.9.9 [60], which includes the Resfinder database [61,62] to determine acquired antimicrobial resistance genes, PointFinder [63] to determine chromosomal point mutations, and PlasmidFinder [64] to identify plasmid replicon types.

2.3.5. Dust Mass Calculations

To calculate the dust mass recovered by the quartz filter (Equation (2)) and impinger (Equations (3) and (4)) samples, the following equations, and parameters were applied.

Aerosol mass concentration (AMC) in grams per m³ of air sampled in quartz filter:

$$\begin{gathered} \mathrm{A}\mathrm{M}\mathrm{C}={\displaystyle \frac{\mathrm{M}\mathrm{e}-\mathrm{M}\mathrm{o}}{\mathrm{Q}\times \Delta \mathrm{t}}} \\ \mathrm{Sampler-specific\; equation}: \\ \mathrm{Aerosol} \mathrm{mass} \mathrm{concentration} (\mathrm{AMC}) \mathrm{in} \mathrm{grams} \mathrm{per} {\mathrm{m}}^3 \mathrm{of} \mathrm{air} \mathrm{in} \mathrm{TSP}: {\mathrm{A}\mathrm{M}\mathrm{C}}_{\mathrm{T}\mathrm{S}\mathrm{P}}={\displaystyle \frac{\mathrm{T}\mathrm{S}\mathrm{P}\left(\mathrm{g}\right)}{\mathrm{a}\mathrm{i}\mathrm{r}\left({\mathrm{m}}^3\right)}} \\ \mathrm{Aerosol} \mathrm{mass} \mathrm{concentration} (\mathrm{AMC}) \mathrm{in} \mathrm{grams} \mathrm{per} {\mathrm{m}}^3 \mathrm{of} \mathrm{air} \mathrm{in} \mathrm{PM}10: {\mathrm{A}\mathrm{M}\mathrm{C}}_{\mathrm{P}\mathrm{M}10}={\displaystyle \frac{\mathrm{P}\mathrm{M}10\left(\mathrm{g}\right)}{\mathrm{a}\mathrm{i}\mathrm{r}\left({\mathrm{m}}^3\right)}} \end{gathered}$$

(2)

where,

• Mo = mass of filter before sampling (g)
• Me = mass of filter after sampling (g)
• AMC_TSP = concentration of bioaerosol captured in TSP filter in grams per m³ of air
• AMC_PM10 = concentration of bioaerosol captured in PM10 filter in grams per m³ of air
• TSP(g) = Me_TSP − Mo_TSP
• PM₁₀(g) = Me_PM10 − Mo_PM10
• Q = Air sampling flow rate (m³/min)
• $\Delta \mathrm{t}$ = Sampling time (min)

Aerosol mass concentration (AMC) in grams per m³ of air sampled in impinger:

$${\mathrm{A}\mathrm{M}\mathrm{C}}_{\mathrm{i}\mathrm{m}\mathrm{p}}={\displaystyle \frac{{\mathrm{V}}_{\mathrm{A}}\left(\mathrm{L}\right)\times {\mathrm{A}\mathrm{M}}_{\mathrm{T}\mathrm{S}\mathrm{P}}\left(\mathrm{g}\right)}{1000\left(\frac{\mathrm{L}}{{\mathrm{m}}^3}\right)}}$$

(3)

where:

• V_A = Volume of air sampled by impinger (L)
• AM_TSP = Aerosol mass (AM) derived from co-located high-volume TSP sampler (TSP (g))
• AMC_imp = concentration of bioaerosol captured in impinger in grams per m³ of air

$${\mathrm{V}}_{\mathrm{A}}={\mathrm{Q}}_{\mathrm{i}\mathrm{m}\mathrm{p}}\times {\Delta \mathrm{t}}_{\mathrm{i}\mathrm{m}\mathrm{p}}\left(\mathrm{L}\right)$$

(4)

where,

• Q_imp = Air sampling flow rate through impinger (L/min)
• Δt_imp = Impinger sampling time (min)

Marble sampling required a different approach to measure the dust mass. The transmittance of the rinsate was measured using the Spectronic 200^® spectrophotometer (Thermo Fisher Scientific, Gaithersburg, MD, USA) after generating a standard curve by weighing 3 g of dried manure sample and preparing ten 2-fold serial dilutions in 1X PBS. Values were plotted in Microsoft^® Excel^® for Microsoft 365 MSO (Version 2402 Build 16.0.17328.20346) 64-bit using the exponential function to model the data (Figure S3).

Mass of dust (g) was estimated for each sample from the standard curve and subsequently used in statistical analyses. The sampling time for marbles was held constant throughout the campaigns in which this method was deployed.

2.4. Statistical Analyses

Statistical analyses were conducted using Stata/BE version 17.0 (StataCorp LLC, College Station, TX, USA).

2.4.1. Bacterial Count Analyses

For bacterial count analyses, the numbers of aerobic bacteria, coliforms, E. coli, Enterococcus spp., and Salmonella in the dust samples were each expressed in (1) log₁₀ CFU per g of dust, (2) log₁₀ CFU per m³ of air, and (3) log₁₀ CFU per mL of raw sample. As a result of the expected high number of zero counts in several of the log₁₀ CFU bacterial endpoints from the dust samples, Cragg Hurdle regression (churdle) statistical models were used to address overdispersion. The hurdle model has two subsets of populations. In the first subset, there are observations at the boundary values (i.e., zero), and the second subset has observations above the lower boundary. There are two models: (1) the selection model—modeling the probability of an observation being in the non-boundary subset; and (2) the outcome model—modeling the value of the dependent variable for non-boundary observations. The final model used a linear regression analysis. Three types of models were generated to evaluate the log₁₀ CFU (colony-forming units) per g of dust, per m³ of bioaerosol, and per mL of sample. The dependent variable in the first model used CFU/g of dust using the standardized aerosol mass concentrated in grams (AMstd) specified in Equation (5) for impingers, in Equation (6) for marbles, and in Equation (7) for QF. The second model used the CFU/m³ of dust aerosols calculated using the CFU/g and multiplying by the mass of bioaerosol concentrated in air AMC (g/m³) Equation (2). The third model simply used the CFU/mL of the raw sample. The AMstd per sample type was calculated as follows:

$$\mathrm{A}\mathrm{M}{\mathrm{s}\mathrm{t}\mathrm{d}}_{\mathrm{i}\mathrm{m}\mathrm{p}}={\displaystyle \frac{{\mathrm{A}\mathrm{M}}_{\mathrm{i}\mathrm{m}\mathrm{p}}}{{\mathrm{V}}_1}} \mathrm{f}\mathrm{o}\mathrm{r} \mathrm{i}\mathrm{m}\mathrm{p}\mathrm{i}\mathrm{n}\mathrm{g}\mathrm{e}\mathrm{r}\mathrm{s}$$

(5)

where,

• AM_imp = Aerosol mass (AM) derived from co-located high-volume TSP sampler (TSP (g))
• V₁ = 100 mL of raw sample

$${\mathrm{A}\mathrm{M}\mathrm{s}\mathrm{t}\mathrm{d}}_{\mathrm{m}\mathrm{b}}={\displaystyle \frac{{\mathrm{A}\mathrm{M}}_{\mathrm{m}\mathrm{b}}}{{\mathrm{V}}_2}} \mathrm{f}\mathrm{o}\mathrm{r} \mathrm{m}\mathrm{a}\mathrm{r}\mathrm{b}\mathrm{l}\mathrm{e}\mathrm{s}$$

(6)

where,

• AM_mb = Aerosol mass (AM) derived from transmittance measure
• V₂ = 150 mL of raw sample

$${\mathrm{A}\mathrm{M}\mathrm{s}\mathrm{t}\mathrm{d}}_{\mathrm{Q}\mathrm{F}}={\displaystyle \frac{{\mathrm{A}\mathrm{M}}_{\mathrm{Q}\mathrm{F}}}{{\mathrm{V}}_3}} \mathrm{f}\mathrm{o}\mathrm{r} \mathrm{q}\mathrm{u}\mathrm{a}\mathrm{r}\mathrm{t}\mathrm{z} \mathrm{f}\mathrm{i}\mathrm{l}\mathrm{t}\mathrm{e}\mathrm{r}\mathrm{s}(\mathrm{T}\mathrm{S}\mathrm{P} \mathrm{o}\mathrm{r} {\mathrm{P}\mathrm{M}}_{10})$$

(7)

where,

• AM_QF = Aerosol mass (AM) derived from quartz filter sampler (g)
• V₃ = 200 mL of raw sample

Marbles were excluded from the second model (CFU/m³) because of the missing AMC records due to the sampling methodology differences.

The dependent variable of the model was the log₁₀ CFU per g of dust, per m³ of bioaerosol, or per mL of raw sample, and the independent variables were the season (baseline was summer), and the interaction between direction (baseline downwind) and sample type (baseline TSP). The same variables were included in the selection equation. Sample replicate was included as a random effect to account for the potential clustering of the samples due to processing in triplicate using the clustered robust function -vce- (cluster).

For aerobic bacteria isolated on TSA, a full factorial model was used; however, for lower counts, like aerobic bacteria on CDTC and bacteria on selective media (E. coli and Enterococcus spp.), only the two-way interaction term was included to address model convergence issues caused by the large number of zero counts. The feedyard variable was not significant (p > 0.05), and so it was excluded from the model for posterior analysis. The log₁₀ CFU from Campaign 5 was excluded from the analysis due to missing dust mass records.

To assess the potential total human exposure to bacteria, the Human Hourly Intake (HHI) of the targeted bacteria via PM₁₀, TSP, and impinger samples was calculated using a previously reported equation [65]. The daily inhalation rate for a human body weight of 70 kg is 20 m³/day [66], which corresponds to 0.83 m³/h. The calculation of the HHI was performed using Equation (8).

$$\mathrm{H}\mathrm{H}\mathrm{I}={\displaystyle \frac{{\mathrm{l}\mathrm{o}\mathrm{g}}_{10}\frac{\mathrm{C}\mathrm{F}\mathrm{U}}{{\mathrm{m}}^3}}{\Delta \mathrm{t}\left(\mathrm{h}\right)}}\times 0.83 {\mathrm{m}}^3/\mathrm{h}$$

(8)

2.4.2. Antimicrobial Susceptibility Analysis

Antimicrobial susceptibility MIC data were imported into Stata/BE in the Trek SWIN (CSV) format. Intermediate resistance was collapsed into the “susceptible” category and resistance was analyzed as a binary variable. The presence or absence of a known resistance gene was compared with the phenotypic interpretation for the corresponding antimicrobial using descriptive statistics.

2.4.3. Quantitative Real-Time PCR (qPCR) for Antimicrobial Resistance Genes

qPCR reactions below the limit of quantification (LOQ) were recorded as missing values. Multiple imputations were assessed for datasets with a high number of missing values in the data (>5%). Models incorporating fixed effects for the campaigns were built. The models were compared across feedyards, seasons, and meteorological variables. Marginal mean estimates from the final model and bar graphs were generated for visualization. In a multiple imputation model, each missing value (i.e., value below the LOQ) was replaced by ‘m’ suitable values. Using typical methods without missing data, each filled dataset was separately examined, producing m sets of estimates and their variances. A comprehensive estimate was generated by integrating these estimations, considering both the observed data and the presence of missing values [67].

A three-step process was followed to impute the missing values. First, the data were explored using histograms and summary statistics (number of observations, missing, mean, minimum, and maximum counts) to understand the data distribution. Second, the appropriate auxiliary variables were selected to improve the predictive missingness and correlation with the observed data, as recommended [68]. Most of these auxiliary variables were not used in the estimation analysis. Third, an independent imputation model was employed that best matched the distribution and requirements of the data [69]. The Multiple Imputation by Chained Equations (MICE) model was employed choosing the appropriate model for each gene in Stata/BE version 17.0 (StataCorp LLC). For the 16S rRNA, bla_CTX-M group 9, tet(A), tet(B), tet(M), and ermB genes the imputation model used a truncated regression whereas bla_CMY-2 used an interval regression model. The number of imputations (77) was based on the maximum missing qPCR data for any one gene (77%). A multivariable regression model was then built for each of the log₁₀ gene copy endpoints. The variables in the full factorial model included log₁₀ gene copies as the dependent variable, the independent variable feedyard (FY2 as baseline), and the full interaction of sample type (TSP as the baseline) and wind direction (downwind as the baseline). The model included only the sample types analyzed by qPCR (PM₁₀, TSP, and marbles). Marginal mean estimates from the final models were generated and displayed with bars corresponding to the 95% confidence intervals. For the analysis of the bla_CTX-M group 1 gene, the hurdle model was used instead of multiple imputation. Methods for the hurdle model followed those previously described for the CFU data.

The log₁₀ gene copies were further analyzed in bivariable models with meteorological data. Eight variables were analyzed (relative humidity, air temperature, vapor pressure, total wind run, maximum wind just, total solar load, precipitation, and reference evapotranspiration (ETo)). Evapotranspiration refers to the total water loss to the atmosphere through the combined processes of evaporation and transpiration [70] and it is influenced by radiation, air temperature, relative humidity, wind speed, and vapor pressure deficit [71]. For our purposes, ETo represents the atmosphere’s propensity to evaporate water from surfaces and living tissues.

3. Results

3.1. Bacterial Prevalence and Quantification

3.1.1. Descriptive Statistics

A total of 91 raw samples were processed in triplicate (i.e., 273 total processed samples) for bacterial quantification. These included impinger (n = 22), quartz filter: TSP (n = 25) and PM₁₀ (n = 22), and marble (n = 22) samples. Bacterial prevalence varied by campaign (Table S3), feedyard (Table S4), and sample type (Table S5) with aerobic bacteria having the highest prevalence and antimicrobial-resistant E. coli and Enterococcus spp. having lower prevalences (Tables S3–S5). Raw bacterial CFU (log₁₀) for each bacterial type as tabulated by campaign (Table S6), feedyard (Table S7), and sample type (Table S8) also varied. The quantities of dust by direction and sample type are described in

Table 4. Concentration of bioaerosol captured in grams per m³ of air sampled by sample type and direction.

Direction	Sample Type	Observations	Mean	Std. Dev.	Min	Max
downwind	PM10	33	0.00080	0.00111	0.00003	0.00380
downwind	TSP	42	0.00227	0.00279	0.00006	0.01006
downwind	impinger	33	0.00276	0.00297	0.00006	0.01006
upwind	PM10	21	0.00008	0.00009	0.000004	0.00023
upwind	TSP	21	0.00009	0.00005	0.00003	0.00018
upwind	impinger	21	0.00009	0.00005	0.00003	0.00018

.

3.1.2. Hurdle Regression Models

Aerobic Bacteria

Viable aerobic bacterial counts were higher in downwind than upwind samples for both TSA and CDTC plates (Figure 1 and Figure 2). Downwind aerobic bacterial quantities on TSA plates were significantly higher (p < 0.05) in impinger samples for log₁₀ CFU/g and log₁₀ CFU/m³ of dust when compared to TSP samples. Downwind aerobic bacterial quantities for log₁₀ CFU/mL were significantly less for PM₁₀ and marble sample types compared to TSP and impinger sample types on TSA plates.

Downwind aerobic bacterial quantities for log₁₀ CFU/g of bioaerosol on CDTC plates were significantly higher for marble samples (p < 0.05) compared to PM₁₀, TSP, and impinger samples. In contrast, downwind aerobic bacterial quantities for log₁₀ CFU/m³ of bioaerosol on CDTC plates were significantly higher for impinger samples than PM₁₀ and TSP samples. Considering downwind aerobic bacterial quantities for log₁₀ CFU/mL of sample, TSP samples were significantly higher than PM₁₀, impinger, and marble samples.

E. coli 

E. coli counts were higher downwind than upwind for all sample types (Figure 3). Downwind E. coli counts for log₁₀ CFU/g and log₁₀ CFU/mL did not differ significantly between sample types (p < 0.05); however, downwind E. coli quantities for log₁₀ CFU/m³ were significantly higher (p < 0.05) in impinger samples compared to PM₁₀ and TSP samples.

Tetracycline-resistant E. coli were only present in downwind samples and no tetracycline-resistant E. coli were isolated from marble samples (Figure 4). There were no significant differences in downwind tetracycline-resistant E. coli quantities for log₁₀ CFU/g or log₁₀ CFU/mL when comparing sample types; however, for log₁₀ CFU/m³ E. coli quantities were significantly higher (p < 0.05) in impinger samples compared to PM₁₀.

Ceftriaxone-resistant E. coli were only present in downwind samples and were found in very low quantities (Figure 5). There were no significant differences among sample types for log₁₀ CFU/g, log₁₀ CFU/mL or log₁₀ CFU/m³, although TSP samples had numerically higher quantities of ceftriaxone-resistant E. coli compared to PM₁₀, impinger, and marble samples (Figure 5).

Enterococcus spp.

For upwind samples, Enterococcus spp. were only recovered from TSP samples and the quantity was significantly less than for downwind TSP samples for log₁₀ CFU/g of bioaerosol, log₁₀ CFU/mL of raw sample, and log₁₀ CFU/m³ of bioaerosol (Figure 6). Downwind Enterococcus spp. quantities for log₁₀ CFU/g were significantly lower in PM₁₀ samples compared to TSP and impinger samples; for log₁₀ CFU/mL and log₁₀ CFU/m³, quantities were significantly lower in PM₁₀ compared to TSP.

Tetracycline-resistant Enterococcus spp. were only recovered from downwind samples (Figure 7). There were no significant differences in downwind tetracycline-resistant Enterococcus spp. among sample types for log₁₀ CFU/g, log₁₀ CFU/mL, or log₁₀ CFU/m³; however, quantities were numerically higher in TSP and impinger samples compared to PM₁₀ and marble samples.

Erythromycin-resistant Enterococcus spp. were only recovered from downwind samples and found in low quantities (Figure 8). There were no significant differences among sample types concerning downwind erythromycin Enterococcus spp. quantities for log₁₀ CFU/g, log₁₀ CFU/mL, and log₁₀ CFU/m³.

Salmonella 

For upwind samples, Salmonella were only recovered from impinger samples and although not significant, the quantities were numerically less than the downwind impinger samples (Figure 9). As was reported in the methods section (Section 2.3.1), to increase the probability of recovering Salmonella, the bacteria were exposed to a pre-enrichment step which inflated the CFUs of Salmonella per sample. Salmonella downwind quantities were significantly higher (p < 0.05) in TSP samples compared to PM₁₀ samples for log₁₀ CFU/g, log₁₀ CFU/mL, and log₁₀ CFU/m³.

Coliforms

For upwind samples, coliforms were only recovered from impinger and marble samples, meanwhile, coliform quantities were numerically higher in downwind impinger and marble samples (Figure 10). Downwind coliform quantities were significantly higher in TSP samples compared to PM₁₀ samples (p < 0.05) for log₁₀ CFU/g, log₁₀ CFU/mL, and log₁₀ CFU/m³.

Modeled marginal means and 95% confidence intervals of all types of log₁₀ CFU bacteria per gram, per mL, and per m³ of dust are described in Tables S9–S11, respectively.

3.2. Bacterial Resistance Gene Quantification in Dust

Higher bacterial resistance gene counts were found in downwind samples compared to upwind samples, similar to the viable bacterial quantification results (Figure 11). The one exception was where slightly higher counts of bla_CTX-M group 9 genes were found in the upwind marble samples compared to the downwind marble samples. Impinger samples did not yield large enough quantities of community DNA to allow for bacterial quantification via qPCR.

The 16S rRNA gene represented both the viable and non-viable bacterial counts in the samples. A significant difference (p < 0.05) was found between downwind and upwind gene copies in PM₁₀ samples. No other significant differences (p > 0.05) were found between sample types or directions for the quantity of 16S rRNA gene copies. Interestingly, the highest quantity of 16S rRNA genes was found in the downwind marble samples.

Among the antimicrobial resistance genes targeted in this study, the highest quantity gene was tet(M), followed in decreasing order by ermB, tet(A), tet(B), bla_CMY-2, bla_CTX-M group 1, and lastly bla _CTX-M group 9. The highest quantities of bla_CMY-2, bla_CTX-M group 1, bla_CTX-M group 9, tet(B), ermB, and tet(M) genes were found in downwind TSP samples; however, the only significant difference (p < 0.05) found was for the bla_CTX-M group 1 when comparing downwind and upwind TSP samples. The highest quantity of tet(A) genes were found in the downwind marble samples; however, there were no significant differences in tet(A) gene quantities by sample type or direction.

The log₁₀ gene copies were further analyzed with meteorological data. ETo which represents the atmosphere’s propensity to evaporate water from surfaces and living tissues influenced by temperature, humidity, wind speed, solar radiation, soil type, and vegetation cover was the meteorological variable that was significantly higher with the most prevalent genes: log₁₀ 16S rRNA (p = 0.011), tet(A) (p = 0.005), ermB (p < 0.001), and tet(M) (p = 0.031). Other variables, such as relative humidity, air temperature, vapor pressure, total wind run, maximum wind gust, total solar load, and precipitation, were not significantly associated.

3.3. Phenotypic Characterization

Evaluation of Antimicrobial Resistance in E. coli and Salmonella

The minimum inhibitory concentrations (MIC) to 14 antimicrobials were determined for all 27 Salmonella isolates cultured from the bioaerosol samples and a subset of 26 E. coli strains cultured from MAC supplemented with tetracycline or ceftriaxone (Figure 12). The proportion of Salmonella isolates resistant to an individual antimicrobial ranged from a low of 0% for ciprofloxacin and gentamicin to a high of 48.1% for tetracycline. The proportion of tested E. coli isolates resistant to an individual antimicrobial ranged from a low of 0% for gentamicin to a high of 92.9% for tetracycline.

For the 27 Salmonella isolates, six (22.2%) were MDR (≥3 classes of antimicrobials), six (22.2%) were classified as resistant to fewer than three classes of antimicrobials, and 14 (51.9%) were classified as pan-susceptible (i.e., susceptible to all the antimicrobials evaluated). For the 26 E. coli isolates grown on TET or AXO-antimicrobial-selective media, 16 (61.5%) were MDR, nine (34.6%) were classified as resistant to fewer than three classes of antimicrobials, and one (3.6%) was classified as pan-susceptible. Among the 13 Salmonella isolates exhibiting resistance to at least one antimicrobial, seven distinct antimicrobial resistance patterns were found among eight serovars (Table S12). Salmonella Reading was resistant to most antimicrobials (nine antimicrobials), while S. anatum, S. lubbock, and S. montevideo were resistant only to tetracycline.

Among the 25 E. coli isolates exhibiting resistance to at least one antimicrobial, fifteen distinct antimicrobial resistance patterns were found among seventeen different sequence types (STs) (Table S13). ST 224 and ST 278 were resistant to most antimicrobials (10 antimicrobials), while ST 20, ST 58, ST 109, ST 730, ST 1148, ST 5204, and ST 6643 were resistant to only one antimicrobial (tetracycline or ampicillin). It is important to note that all E. coli subjected to phenotypic antimicrobial susceptibility testing were cultured on antibiotic selective media containing either ceftriaxone or tetracycline at CLSI breakpoint concentrations.

3.4. Genotypic Characterization

3.4.1. Antimicrobial Resistance Genes

Whole genome sequencing identified a total of 10 ARGs in the Salmonella isolates and 15 ARGs in the E. coli isolates. It is important to note that genes not typically associated with phenotypic resistance (fosA7) and genes encoding multidrug efflux pumps (mdf(A)) were not included in the analysis. Out of 27 isolates, 13 Salmonella (48.1%) carried at least a single antimicrobial resistance gene (ARG) and 14 (51.8%) did not harbor any ARGs (Table S11). Out of 26 E. coli isolates, 25 (96.4%) carried at least one ARG (Table S12), and only one isolate (3.5%) did not harbor any ARGs. Once again, it is important to note that all E. coli subjected to WGS were cultured on antibiotic selective media.

3.4.2. Salmonella Serovars

A total of eight different Salmonella serovars were identified and all serovars represented a single ST. The most prevalent serovars were Lubbock (ST 413, 25.9%, 7/27), Anatum (ST 64, 22.2%, 6/27), followed by Cerro (ST 2398, 18.5%, 5/27), Meleagridis (ST 463,14.8%, 4/27), Montevideo (ST 138, 7.4%, 2/27), Lille (ST 297, 3.7%, 1/27), Muenchen (ST 83, 3.7%, 1/27), and Reading (ST 1628, 3.7%, 1/27). Interestingly, the proportion of Salmonella serovars differed by sample type (Figure 13). Salmonella was isolated from only two of the six campaigns and from two of the three feedyards (Figure 14 and Figure 15).

3.4.3. Plasmid Analysis

A total of 25 (89%) E. coli and 12 (44.4%) Salmonella isolates harbored at least one plasmid. Six different plasmid incompatibility (Inc) types were identified in the 27 Salmonella isolates (IncA/C2, IncFII, IncI1, IncN, IncR, and IncX) and the predominant type was IncI1 (22.2%, n= 6) (Table S12). Eight different plasmid incompatibility (Inc) types were identified in E. coli (IncFIA, IncFIB, IncFIC, IncFII, IncI1, IncR, IncX, and IncY) and the predominant type was IncFIB (64.3%, n= 18) (Table S13).

4. Discussion

This study was conceived and designed to determine the presence and quantities of viable Gram-negative and Gram-positive indicator bacteria in addition to pathogenic bacteria (indicators of antimicrobial resistance) [72] and antimicrobial resistance genes in bioaerosol samples collected from three feedyards during six seasonal sampling campaigns using several sampling methods. The study revealed significant findings regarding the presence of viable antimicrobial-resistant (AMR) bacteria and antimicrobial-resistant genes (ARGs) in bioaerosol samples from cattle feedyards. Viable AMR bacteria were detected, which suggests a potential route for dissemination within the feedyard environment and to the surrounding area. These findings are similar to results reported in other studies regarding dust emissions from animal production facilities [24,73]. The identification of AMR bacteria, particularly those resistant to commonly used antimicrobials, such as cephalosporins, macrolides, and tetracycline, aligns with previous studies from animals in feedlots [74,75,76,77,78]. The viability of bacteria in bioaerosols suggests these bacteria can survive the harsh environmental conditions of feedyards and become aerosolized. A recent study from our own group found bacteria in manure samples were able to survive UV-B exposure and desiccation under strict laboratory conditions [79]. The potential for bacteria to persist in bioaerosols raises potential health concerns, particularly regarding the risk of respiratory transmission to humans working with animals residing in the feedyard environment.

The mean concentrations of bioaerosols were higher downwind in TSP, impinger, and PM₁₀ samples than in upwind samples. The concentration of bioaerosols in PM₁₀ samples was three times higher than what has previously been reported from feedyards [80]; however, for TSP, similar studies reported ten times more dust than our study [81,82]. The ratio of upwind to downwind bioaerosols was between 0.03 to 0.10 depending on the sampling method, which was within the range (0.004 to 0.74) previously reported [80,81,82,83,84,85,86,87]. These concentrations were measured directly downwind and onsite at the feedyards. Previous studies have reported a rapid decay of particulate matter with increasing downwind distance, sometimes reporting a reduction to 8.5% at 3.5 km beyond the edge of the feedyard [88].

4.1. Bacterial Quantification in Dust

Similar to bioaerosol mass and concentrations, bacterial prevalence and concentrations were higher in downwind samples when compared to upwind samples, indicating that the majority of bacteria in the bioaerosols likely originated from the feedyard. Previous studies in cattle feedyards have also found that bacterial concentrations were higher downwind than upwind [26,87,89]. Although bacterial concentrations were consistently higher in downwind samples across all sample types, we detected variations in bacterial levels among the different sample types.

A variety of different selective and differential agars was used to quantify bacteria in the bioaerosol samples. TSA agar, MAC agar (with and without antibiotics), and ME agar (with and without antibiotics) were used with spiral plating techniques to determine the concentrations of total aerobic bacteria, E. coli, and Enterococcus spp., respectively. Additionally, CompactDry™ plates, which are commonly used in a variety of industries to test food, cosmetics, and other raw materials, were utilized to quantify coliforms and Salmonella. The inclusion of a variety of different media types allowed for the investigation of both broad and narrow groups of viable aerobic and enteric bacteria in bioaerosols and allowed determination of differences within direction, sample type, feedyard, and campaign.

To quantify total aerobic bacteria, we employed two different types of media, TSA and CompactDry™ CDTC plates. Interestingly, we found differences between the media and sample types. Aerobic bacterial counts on CDTC plates were highest for marbles, followed by impinger samples. In contrast, for TSA plates, bacterial counts were higher for TSP and impinger samples. Although similar aerobic bacterial counts have been reported between TSA and CDTC plates in samples with a low quantity of bacteria [90], the differences that were found in our study could be due to the high concentrations and variability of bacteria in the feedyard environment, along with differences in sample types.

Marbles are passive samplers and particles are collected passively via gravity, deposition, and electrostatic attraction [91]. Previous studies suggested that marbles should be treated only as a qualitative measure as the sampled air volume is unknown [91]. Passive methods can yield very different species based on the size of the microorganisms, deposition rates, ventilation, and other parameters [91]. The TSP method collects all airborne particles regardless of size, and that broader range allows TSP samples to capture a more comprehensive array of particles, including larger ones that may carry bacteria. The impinger is capable of preserving the physiological state of bacteria (viability, cultivability, metabolic activity, ice-nucleation activity), even during long sampling periods due to the gentle collection method [92,93,94].

For comparison with other studies, it is important to note that we reported bacterial concentrations in log₁₀ CFU, while other studies present their data in raw CFU values. For downwind samples, our mean quantity of aerobic bacteria was 4.0 on TSA and 2.9 log₁₀ CFU/m³ on CDTC. Purdy, et al. (2007), reported a mean CFU of 1383 ± 223 CFU/m³ (3.1 log₁₀ CFU/m³) in non-respirable particles (particles sizes 5 to 10 μm) [80], and Wilson, et al. (2002), reported a mean of 1146 CFU/m³ (3.1 log₁₀ CFU/m³) aerobic bacteria [87]. For upwind samples, our mean quantity of aerobic bacteria was 1.1 on TSA and 1.6 log₁₀ CFU/m³ on CDTC, while Purdy et al. (2007), reported a mean of 376 ± 66 CFU/m³ (2.6 log₁₀ CFU/m³) in non-respirable particles. It is important to clarify that the time of sampling in Purdy et al. (2007), was very short (5 min) in comparison with our study, which was between 3 h to 4 h. Additionally, Purdy et al. (2007), used impactor samplers to collect the bacteria [80]. Differences in CFU between our study and previous studies could be due to sampling time or different sampling methods.

E. coli were primarily recovered from downwind samples; however, a small quantity of E. coli were recovered from upwind impinger and marble samples. Antimicrobial resistant (TET and AXO) E. coli were only recovered from downwind samples; overall, ceftriaxone-resistant E. coli were found in very low quantities. Downwind impinger samples had the highest quantity of E. coli and tetracycline-resistant E. coli, similar to previously published studies [95]. A study that compared different bioaerosol sampling methods on poultry farms, found that impingers were more effective in collecting E. coli than ACD-200 Bobcat (dry surface collection) and similar to the Andersen six-stage impactor [95]. This is most likely due to the liquid media in the impinger that reduced the stress of the bacteria during the collection process, maintaining viability and facilitating recovery and quantification of E. coli [39]. Moreover, impingers capture a wide range of particle sizes, including larger particles (>7 μm), that may be more likely to carry bacteria [95].

Previous studies have focused on the potential impact of bioaerosols from cattle feedyards by studying E. coli and E. coli O157:H7 concentrations on spinach and leafy greens grown in close proximity to cattle feedyards [23]. The samples collected in those studies would be expected to be similar to the marble samples in the current study. In the study by Berry et al. (2015), E. coli O157:H7 were not detected in downwind samples from a feedlot using MAS-100 Eco microbial air samplers [23]; however, on a very dusty day, 837.2 CFU/m³ non-type specific E. coli were found at the edge of the feedlot. Additionally, in the leafy green plots; E. coli quantities were 16.7 CFU/m³ at 60 m, 10 CFU/m³ at 120 m, and 5.3 CFU/m³ at 180 m, showing a decrease in E. coli concentrations with increasing distance from the feedyard [23].

Enterococcus spp. were also primarily recovered from downwind samples and only recovered from TSP samples upwind. This contrasts with E. coli which were recovered in impinger and marble samples upwind. This most likely can be explained by the thick cell wall of Enterococcus spp. (and other Gram-positive organisms) that allowed the bacteria to remain viable after the harsh impact and desiccation when using filters in the TSP samplers. Additionally, the high-volume collection of the TSP samplers favor finding bacteria in lower abundance, and a lower abundance of Enterococcus spp. compared to E. coli in the upwind environment may explain why Enterococcus spp. were not recovered from the impinger or marble samples. Similar to the antimicrobial-resistant E. coli, antimicrobial-resistant Enterococcus spp. were only recovered from downwind samples. A previous study in feedyards isolated Enterococcus spp. from downwind samples using a high-volume air sampler; however, that the study did not test for antimicrobial resistance [96].

Salmonella were also primarily recovered from downwind samples except for upwind impinger samples. Similar to E. coli, Salmonella are Gram-negative organisms and most likely were less able to survive the high impact of the TSP and PM₁₀ samplers. Similar to E. coli and Enterococcus spp., Salmonella were found in higher quantities in the downwind TSP and impinger samples. However, care should be taken when comparing the Salmonella findings with the other bacteria because the Salmonella were subjected to a pre-enrichment step prior to plating. Additional growth curve experiments were conducted to determine the effect of pre-enrichment in TSB for 24 h on Salmonella quantities; however, at 24 h, Salmonella had reached the stationary phase and were no longer increasing in CFUs. Future studies on Salmonella quantification from bioaerosols would benefit from testing pre-enrichments at various intervals to obtain a better understanding of the effect of pre-enrichment on Salmonella quantification.

Similar to E. coli, coliforms were found primarily in downwind samples but were recovered from upwind impinger and marble samples. We would expect the coliform results to be similar to the E. coli results because coliforms include E. coli and other Gram-negative bacilli, such as Citrobacter spp., Enterobacter spp., and Klebsiella spp. Previous studies have reported marbles to be an effective sampling method for coliforms and the interstitial spaces between the marbles can trap particles and coliforms effectively and protect them from harsh environmental conditions (e.g., heat, desiccation, UV light) [97]. Marbles have been reported to capture different Enterobacteriaceae bacteria, like Enterobacter spp., Klebsiella spp., and Citrobacter spp., that are more resistant to environmental conditions than E. coli and Salmonella [98], and this may explain why we recovered coliforms in upwind marble samples in addition to impinger samples. Additionally, since coliforms encompass genera in addition to Escherichia spp., we would expect higher bacterial quantities in comparison to E. coli alone.

The risk for exposure to bioaerosols containing antimicrobial-resistant bacteria was evaluated using the Human Hourly Intake (HHI) model. The raw total count of aerobic bacteria on TSA and CDTC media had the highest level of HHI (at 7 to 10 CFU/h) and if totaled across the 4 h (EDP) we have 40 CFU of aerobic bacteria. The HHI for raw total E. coli was 3 CFU/h, tetracycline-resistant E. coli was 2 CFU/h, and Enterococcus spp. was around 3 CFU/h. For the entire duration of the EDP, this resulted in a cumulative total of 12 CFU for total E. coli and Enterococcus spp., and 4 CFU for tetracycline-resistant E. coli. Bai et al. (2022), reported a daily intake (DI) of 1769 to 2160 CFU/day of Staphylococcus spp. from a chicken house, and 1512 to 3522 CFU/day from a dairy barn [73], which is a higher intake in comparison to what we found. An important consideration is that we found lower HHI for PM₁₀ in comparison with TSP, which is logical being PM₁₀ a fraction within TSP.

4.2. Bacterial Gene Quantification in Dust

Similar to the viable bacterial quantities, gene quantities were consistently higher in downwind samples, though such genes were consistently recovered both upwind and downwind in all sample types. In concert with our inability to recover viable bacteria in several upwind sample types, this provides evidence that many of these genes are likely harbored in non-viable bacteria. As expected, the highest copy number gene was the 16S rRNA gene. The number of bacteria in the sample was estimated using this gene as a baseline, but it is crucial to remember that this number includes both viable and non-viable bacteria, and bacteria can harbor more than one copy of this gene. For most antimicrobial resistance genes, the highest quantity was found in the downwind TSP samples. Since the results include both viable and non-viable bacteria, we would anticipate seeing the highest concentrations of genes from the TSP samplers, because these samplers collected the highest concentrations of dust. The most prevalent antimicrobial resistance gene was tet(M), followed by ermB, tet(A), tet(B), bla_CMY-2, and then bla_CTX-M variants.

The tet(M) gene is predominant in Gram-positive bacteria like Enterococcus spp. [99]. This gene is responsible for clinically important resistance to tetracycline, which is a highly important antimicrobial family in human medicine [100]. In cattle, tetracycline is used to prevent, control, and treat bacterial pneumonia associated with shipping fever complex or bovine respiratory disease (BRD) [101]. The tet(M) gene has been reported at a prevalence of 30.7% in Enterococcus hirae isolates from beef production systems [102]. McEachran et al. (2015), found tet(M) was the most abundant tetracycline gene in downwind samples [20], which is consistent with our results. The tet(M) gene has also been reported in bioaerosols from pig farms [47,103] and feedyards [82].

The ermB gene was the second most common gene and is also predominant in Enterococcus spp. in cattle, the environment, and humans, [104,105]. This gene encodes for resistance to erythromycin with cross-resistance to other macrolides like azithromycin [106]. Macrolides (azithromycin and erythromycin) are critically important antimicrobials for treating Legionella spp., Campylobacter spp., MDR Salmonella, and Shigella spp. in humans [100]. Tilmicosin, tulathromycin, and tylosin are macrolides used to control and treat bovine respiratory disease in cattle [107], and tylosin is also commonly used to prevent liver abscesses [108]. Another study on bioaerosols detected the ermB gene in samples collected near a conventional cattle farm, although quantification data were not provided [109]. The ermB has also been reported as an abundant gene in bioaerosols from poultry and pig production environments [110].

The tet(A), tet(B), bla_CMY-2, and bla_CTX-M genes were found in lower quantities. Interestingly, these genes are predominant in Gram-negative bacteria, such as E. coli and Salmonella, rather than in Gram-positive organisms, such as Enterococcus spp. The finding of higher gene copies for resistance genes predominant in Gram-positive organisms would suggest that Gram-positive organisms are more prevalent in bioaerosols downwind from cattle feedyards. Alternatively, Gram-positive organisms may be more prevalent in the cattle feedyard environment, or they may have a greater ability to survive in feedyard environment and become aerosolized. Further research is needed to determine the differences in viable Gram-positive and Gram-negative bacteria in the feedyard environment and the ability of these organisms and their resistance genes to survive and become aerosolized.

4.3. Phenotypic and Genotypic Characterization of E. coli and Salmonella Isolates

4.3.1. Evaluation of Antimicrobial Resistance in E. coli and Salmonella spp.

Many of the Salmonella isolates were pan-susceptible; however, the most prevalent resistance phenotypes were tetracycline, sulfisoxazole, and trimethoprim-sulfamethoxazole. These results are similar to previous studies of Salmonella from cattle feces, which found that the majority of the Salmonella were pan-susceptible, and the highest resistance was for tetracycline and sulfisoxazole [111]. In contrast to Salmonella, most of the E. coli isolates were phenotypically MDR, and the most prevalent resistance phenotypes were of tetracycline, streptomycin, and sulfisoxazole. It is important to note that E. coli isolates selected for phenotypic and genotypic characterization in our study were chosen from antimicrobial selective plates and, therefore, we would not expect any of the E. coli isolates to be pan-susceptible. One E. coli isolate selected from antimicrobial selective media was pan-susceptible when subjected to Sensititre testing. This result could be explained by the bacterium exhibiting transient tolerance mechanisms, such as efflux pumps, which allow survival under selective pressure [112], Additionally, the isolate might belong to a heterogeneous population (mixed population), where not all cells carry specific resistance determinants [113].

A study comparing airborne bacteria from conventional and organic beef cattle farms found that bacteria from conventional farms were more resistant to beta-lactams, tetracycline, and sulfonamides [109], which is similar to the resistance profiles reported in this study for Salmonella and E. coli. According to the NARMS 2023 data, tetracycline was the highest reported phenotypic resistance for E. coli (38.9%) and Salmonella (15.8%) from the cecal content of beef cattle [114]. The higher proportion of tetracycline resistance in E. coli versus Salmonella concurs with the results in our study. Additionally, the resistance types found in the E. coli and Salmonella isolates from bioaerosols were consistent with antimicrobials commonly administered to feedlot cattle [115].

Some E. coli and Salmonella isolates were also resistant to third-generation cephalosporins, aminoglycosides, macrolides, phenicols, quinolones, and sulfonamides. These antibiotics are of varying importance in human medicine. Macrolides, third-generation cephalosporins, and quinolones are considered to be the highest priority and critically important antimicrobials by the World Health Organization [100]. It is important to note that no gentamicin-resistant E. coli were found, and no ciprofloxacin-resistant Salmonella were detected.

4.3.2. Salmonella Serovars

The most common Salmonella serovar identified was Lubbock. Salmonella Lubbock has been identified in feedlot cattle populations across Texas [116,117,118,119,120]. Additional serovars cultured included Anatum, Cerro, Lille, Meleagridis, Montevideo, Muenchen, and Reading, which also have been reported in Texas feedlots [53,116,117,121,122,123].

The TSP samples had a wider variety of serovars in comparison with marbles, from which only serovar Lubbock was found. This is unsurprising, as TSP samples tend to capture higher dust concentrations, thereby increasing the likelihood of detecting Salmonella serovars that may be present in lower abundance. As previously discussed, the Salmonella culturing process included a pre-enrichment step in TSB, which could promote the growth of specific serovars over others. This pre-enrichment step likely increased the chances of culturing certain serovars by providing favorable conditions for their expansion, even if they were initially present in low numbers. Salmonella was only isolated from two campaigns (1 and 5), both of which occurred in the summer. Earlier studies have reported seasonal differences in Salmonella prevalence [124], with Salmonella being more prevalent during the summer and early fall compared to the winter and spring [117,125,126]. Additionally, Salmonella was only cultured from two of the three feedyards. Studies have also reported differences in Salmonella prevalence between feedyards, which is most likely due to differences in management strategies [127].

4.3.3. Bacterial Antimicrobial Resistance Genes

The phenotypic tetracycline resistance found in Salmonella and E. coli isolates was consistent with the presence of the tet genes found through WGS. The tet(A) gene was the most common gene encoding for tetracycline resistance in the Salmonella (n = 13) and E. coli (n = 27) isolates. In a study of E. coli from air samples and exudates on dairy cattle farms, 10% of the isolates harbored the tet(A) gene [128]. This finding was also consistent with what we found in the log₁₀ gene counts detected by qPCR, in which the tet(A) gene counts were higher than the tet(B) gene. Additionally, the tetracycline genes were most likely harbored in additional bacterial species besides E. coli, Salmonella, and Enterococcus spp. Unfortunately, we cannot comment on the tet(M) gene, as we did not further characterize the Enterococcus spp. in this study by using WGS. The Enterococcus species primarily identified in our study was Enterococcus hirae, which is associated with cattle, but it is very rarely associated with infections in humans [129].

The bla_CMY-2 gene was the only beta-lactam resistant gene found in Salmonella isolates and was also detected by qPCR analysis. Previous studies in feedlot cattle animals in the USA have reported the presence of bla_CMY-2, bla_CTX-M, bla_TEM, and bla_SHV [130]. In E. coli, the bla_TEM gene was the most common beta-lactam resistance gene (n = 12), and the ESBL resistance genes detected were bla_CTX-M-55 and bla_CTX-M-1. The bla_CTX-M gene was also detected by qPCR. In a study by Schmid et al. (2013), ESBL E. coli were also isolated from dust samples in beef cattle feedyards, highlighting the potential for the airborne transmission of antimicrobial-resistant bacteria [74]. It is important to note that there was a small number of Salmonella isolates recovered from bioaerosol samples and subsequently sequenced and only a subset of phenotypically resistant E. coli isolates were sequenced. Therefore, phenotypic and genotypic AMR results may not be representative of the overall population of Salmonella and E. coli in bioaerosol samples from cattle feedyards.

4.3.4. Plasmids

The most common plasmid incompatibility (Inc) type in Salmonella was IncI, which is the most common Inc type reported in Salmonella [131,132]. IncI has been found to harbor beta-lactamase genes like bla_CMY-2 [133,134]. The most common plasmid Inc type in E. coli was the IncFIB, as reported in previous studies [135,136]. IncFIB has been associated with multiple ARGs including aminoglycosides (aph(3″)-Ib and aph(6)-Id), beta-lactams (bla_TEM-1B), folate synthesis inhibitors (dfrA5, sul2, and sul1), and tetracyclines (tet(A) and tet(B)) [135]. The MDR Reading, Cerro, Meleagridis, and Lille Salmonella serovars were harboring either A/C2, X1, or R Inc plasmids. The IncA/C2 plasmid is commonly associated with MDR and is widely distributed among Salmonella [137]. This plasmid often carries multiple resistance genes, including bla_CMY-2, which confers resistance to second and third generation cephalosporins and potentiated beta-lactam antibiotics complicates treatment options [137,138].

Our study provided data on the quantification of viable E. coli, Enterococcus spp., and Salmonella and phenotypic and genotypic resistance patterns of these bacteria in bioaerosols from cattle feedyards. Several studies have been conducted on the quantification of dust in feedyards, and additional studies have reported on antimicrobial resistance in the dust from feedyards. However, the majority of these studies have focused on molecular methods that are more sensitive to identifying antimicrobial resistance genes but cannot differentiate between viable bacteria or attribute the genes to particular bacteria. To our knowledge, this study is one of the most comprehensive investigations focusing on the identification and quantification of a viable indicator and pathogenic bacteria, along with their associated ARGs, in feedlot dust using various sampling methods. The extensive comparison of bioaerosol sampling methods will be invaluable for future studies on bacterial populations in bioaerosol samples in agricultural environments. Limitations related with the spatial and temporal variation in studies that capture dust in open environments in the feedyards are inevitable because of the required meteorological conditions which may affect the sampling frequency. Additionally, in our study, the use of different equipment to compare the sampling methods required a complex, labor-intensive set up. Further exploration is required to determine additional spatial and temporal variations between feedyards.

This study provides crucial data to guide the monitoring of antimicrobial resistance in the feedyard environment and to facilitate the development of mitigation strategies to decrease the risk of antimicrobial resistance via bioaerosols.

5. Conclusions

Our study provides valuable insights into viable bacteria and antimicrobial resistance in bioaerosols from beef cattle feedyards. The results highlight the potential public health implications and the need for continued monitoring and surveillance of antimicrobial resistance in feedyards. Further research and collaborative efforts are crucial to better understand the dynamics of ARG dissemination through the dust in feedyards and to develop strategies for mitigating the spread of antimicrobial-resistant bacteria in the feedyard environment as well as in landscapes further downwind.

These findings emphasize the need for management practices to mitigate the spread of AMR bacteria from agricultural settings. This includes improving manure handling and dust-control measures, stewardship in the use of antimicrobials, as well as exploring alternatives for the use of antibiotics in beef cattle production. By addressing these factors, the dispersion risk of AMR bacteria in the environment, especially via bioaerosols, can be reduced.