A total of 48 swine farms were sampled from across the state of Iowa in the summer of 2020. At each farm, a single representative manure sample was collected from deep pit storage structures. Specific locations of the farms are not disclosed due to privacy restrictions, but all farms are within the state of Iowa and are geographically independent of each other. Samples were collected at the edge of the pits through a manure pump out via dipping the sample from the top six inches of the manure surface. All farms were deep pit barn facilities where pigs were either grow-finish (GF) or wean-finish (WF) pigs raised on a slatted floor. Pigs were fed commercial production diets consisting primarily of corn, soybean meal, and distillers grains with percentages fed varying by growth stage and price of different feed ingredients. After collection, manure was stored at -20°C for one month until further processing. Each manure sample was subsampled in triplicate prior to DNA extraction. Each farm included was categorized based on originating integrator and production system. Specifically, these categories were company integrator: Integrator 1 (n=24) or integrator 2 (n=24); production system: wean-finish (n=34), or grow-finish (n=14). Ethical review and approval was not required for the study on animals in accordance with the local legislation and institutional requirements. This work was conducted in collaboration with local swine growers who made all animal decisions regarding health and well-being and allowed the collection of manure at their site.

The DNA extraction procedure followed protocols from the MagAttract PowerSoil DNA EP Kit (Qiagen) and an epMotion 5075 automated robot for extraction (Eppendorf). Samples of 0.25 grams wet weight of liquid swine manure were used for DNA extraction. Each manure was sub-sampled into three replicates (“farm replicates”). For each farm replicate, we performed three DNA extractions, resulting in three technical extraction replicates (“extraction technical replicates”) for each farm replicate manure sample. The resulting DNA was cleaned using a DNA Clean and Concentrator kit (Zymo Research). Subsequent DNA concentrations were measured with the Quant-it dsDNA Assay Kit, high sensitivity (Thermo Fisher Scientific). The DNA samples were stored at -80°C until further use.

Conventional qPCR assays were performed on a CFX96 Touch Real-Time PCR Detection System (BioRad) and measured in triplicate using primers targeting the 16S rRNA gene, ermB gene, and tetM gene ( Supplementary Table 1 ). Genes were quantified in all 48 swine manure samples. The DNA template was diluted (1:10) to optimize qPCR detection, to minimize inhibitors, and increase primer efficiency to a target gene. The limit of quantification was determined for each gene using oligonucleotide standards. Standard curves ranged from 10⁷ to 10¹ copies, and all samples measured above the limit of quantification. Outliers in the triplicate were omitted if above 1.5 times the standard deviation in the average of the three values. Efficiencies calculated by standard curves ranged from 82.2 to 100.6% and all R² values were above 0.98 ( Supplementary Table 2 ). All reported absolute abundance (copies/gram) are reported in gene copies per gram of wet weight of manure and were calculated by the equation:

Extracted DNA was analyzed for a wide host of ARGs encoding resistance to a broad spectrum of antibiotics used in swine production; str, aadD, aadA2 (Aminoglycosides), ermB, ermC, ermF, ermQ, ermT, erm(35), erm(36) (Macrolides), sul2, sul1 (Sulfonamides), tetA, tetL, tetM, tetO, tetT, tetW, tetX, tet(36) (Tetracyclines), blaPSE, blaOXA10 (Carbapenems), lnuC, lnuA (Lincosamides), cmlA5, cmlA1, floR (Phenicols), intI3, intI2, intI1F165, and intI1 (Integrons). The high-throughput qPCR primers used for the analysis are originally described in Stedtfeld et al. (2018), Supplementary Table 1 . The high-throughput qPCR assay was performed on the Biomark Fluorescent machine in the 96x96 primer target layout. Each assay was performed in triplicate. The template DNA was diluted in a 1:500 dilution for optimal performance on the Biomark machine and to decrease potential inhibitor effects. Samples reading a cycle threshold value greater than 30 were omitted from further analysis. Cycle threshold detections greater than 30 were assumed to be non-detected. Verification of the high-throughput qPCR machine performance is supported with internal standards for standard curve development of 16S rRNA, ermB, ermF, sul2, tetM, and tetW genes. Each internal standard gene amplified successfully with efficiencies ranging from 80.0 to 104.2% ( Supplementary Table 3 ).

In order to be deemed a successful amplification, we required that the conserved total bacteria gene 16S rRNA was detected in each manure sample. Additionally, we required that detection was observed for each gene in 2 out of 3 farm manure replicates and 2 out of 3 technical extraction replicate detections for each sample.

All statistical analyses were performed using R version 4.0.3. The quantified ermB and tetM gene concentrations (copies/gram wet weight) were log10 transformed to fit a normal distribution. Normality was confirmed with visual inspection of histograms and Q-Q Plots. The linear regression models were fit using the lme4 package (Bates et al., 2015). The two gene responses were analyzed separately. The integrator and production system were treated as fixed effects. Gene concentrations of subsampled triplicates from one representative manure sample per farm were averaged before model building. Model performance was evaluated using the Performance package (Lüdecke et al., 2021) ( Supplementary Table 4 ).

The R package emmeans (Lenth, 2021) was used for calculating the estimated marginal means from the verified models and making pairwise comparisons of fixed effects. All pairwise comparisons were made with a 95% confidence level (P<0.05) and P-values were adjusted using Tukey’s method for multiple comparisons. The main effects refer to the overall effect of the variable while ignoring, or averaging over, the levels of the other predictor variable. The main and interaction effects of each model were analyzed using ANOVA and type-III error.

The number of gene copies of tetM and ermB were quantified in DNA extracted from all manures using targeted amplification of these genes. Additionally, gene copies of the 16S rRNA gene, a phylogenetic marker present in all bacteria, were estimated and used for normalizing total bacterial counts among manure comparisons. Overall, there was a large range of detection of both genes across all 48 farms ( Figure 1 ); the absolute gene concentrations of ermB ranged from 2.20x10⁴ copies gram^-1 to 1.53x10⁸ copies gram^-1 and tetM ranged from 1.33x10⁵ copies gram^-1 to 2.23x10⁸ copies gram^-1. The limit of quantification for each individual qPCR plate are reported in Supplementary Table 2 . The concentrations of ermB and tetM were significantly different across the 48 manure samples (ANOVA, P< 0.0001) ( Supplementary Table 5 ), and a general trend was observed that tetM and ermB concentrations increased with concentrations of 16S rRNA genes.

The company integrator had a significant main effect on observed ermB absolute gene concentrations based on the overall ANOVA with type-III error (P=0.0007) ( Supplementary Table 6 ). The mean concentration of ermB in manures associated with integrator 2 manure was 15% greater than manures from integrator 1. Integrator 2 had an ermB estimated marginal mean of 4.8x10⁶ copies/gram compared to integrator 1 with 6.5x10⁵ copies/gram. This result exists when ermB was normalized to 16S rRNA (P=0.0020) ( Supplementary Figure 1 and Supplementary Table 7 ). Likewise, there is evidence that the integrator had a significant effect on tetM concentrations (P=0.0425), with tetM also being enriched in integrator 2 relative to integrator 1 ( Figure 2 ). However, this result is non-significant when tetM was normalized to 16S rRNA (P= 0.3670). The production system had no significant main effect on ermB or tetM gene concentrations or relative abundance to 16S rRNA ( Supplementary Figures 2, 3 ). Additionally, there was no significant interaction between the two fixed effects in both the absolute copy number model and the 16S rRNA normalized model for each gene ( Supplementary Tables 6, 7 ).

In addition to quantification of tetM and ermB in manures, we also evaluated the presence of 31 ARGs listed in Table 1 and the 16S rRNA gene in manures using methods similar to those previously described (Stedtfeld et al., 2018) to leverage the ability to assay numerous genes simultaneously with high-throughput qPCR (HT-qPCR). Each internal standard gene of 16S rRNA, ermB, ermF, sul2, tetM, and tetW were amplified successfully with efficiencies ranging from 80-104% ( Supplementary Table 3 ). However, while all 48 manures were evaluated against these 32 genes, in total, we detected 22 unique ARGs in 14 independent farm manure samples ( Figure 3 ). In 34 manures, we were unable to amplify the 16S rRNA gene with HT-qPCR assays and thus these samples were removed from further analysis. Within successfully amplified samples, the most frequently detected ARG in manure was tet(36), which was detected in all 14 manures. The second most detected ARG was tetT at 93% detection, followed by erm(35) at 78.6% detection. Genes encoding resistance to tetracycline, tetT, tetM, and tet(36), were present in 13/14, 8/14, and 14/14 farm manure samples, respectively. The macrolide resistance gene class, erm, had the second most detected antibiotic resistance genes with erm(35), ermF and, ermB detected in 11/14, 10/14, and 7/14, respectively. There was no detection of blaPSE, blaOXA10, cmlA5, cmlA1, floR, lnuA, erm(36), tetL, and tetA in any of the manure samples.

Based on the detection of ARGs, we have developed recommendations of the most commonly detected ARGs in Iowa swine manures ( Table 2 ). Importantly, we also identify the ARGs that were not strongly present in manure holding pits, and these ARGs include tetL, tetA, erm(36), floR, cmlA5, cmlA1, blaPSE, and blaOXA10 (no detection), sul1, inti1, and inti1F165, (7.1%), aadA2 (14.2%), inti2 and sul2 (21.4%). In general, we observed that the two main resistance mechanisms of ARGs present in the manures studied were associated with target protection and target alteration.

Consistent with previous observations of the association and enrichment of ermB and tetM genes with manures (Whitehead and Cotta, 2013; Joy et al., 2014; Zalewska et al., 2021) and adjacent soils and waters (Peng et al., 2017; Zhang et al., 2021), we detected these genes in all 48 manures in this study. The concentrations measured in our study were consistent with those detected in manure holding pits measured in other studies (Mackie et al., 2006; Joy et al., 2013; Hall et al., 2020; Alt et al., 2021) and also demonstrate the wide variations of ARGs that can be observed within manures, with variations up to three-fold. The wide ranges of measured ermB and tetM in these manures may be caused by covariates in manure holding pits that have yet unknown implications on ARG concentrations after long-term exposure such as concentrations of heavy metals, manure pit additives, or changes in chemical properties such as pH or organic substrates (Hölzel et al., 2012; He et al., 2020). While it is clear that these ARGs are consistently observed between swine manures, it is less clear what the implications are of the magnitude and variability of these gene concentrations (ermB and tetM varying between 2.20x10⁴ and 1.53x10⁸ copies/gram in our samples). We speculate that the concentration of ARGs may be associated with the time spent in storage, with manure sampled right at defecation presumably containing different concentrations of ARGs than in manure stored for up to six months (Joy et al., 2014). Future studies of the relationship between these gene concentrations and to risks antibiotic resistance are much needed (Gullberg et al., 2011; Hughes and Andersson, 2017), and the results of this study provide some insight the variability of these concentrations in varying manures.

The abundances of these genes also followed observable patterns based on their farm of origin. We observed significant differences of ermB and tetM gene concentrations among farms with different company integrators, with both genes consistently largest in the same integrator. Integrators generally manage piglet source, feedstock, and veterinary practices (Tsoulouhas and Vukina, 1999; McBride and Key, 2003; Reimer, 2006). Our observations that different integrators have different concentrations of these genes suggest that these management decisions may affect ARG concentrations in manures (Lu et al., 2017; Ghanbari et al., 2019; Cheng et al., 2021). We did not observe any significant differences in tetM or ermB in association to the production system, or whether manure originated from wean or grow-finished pigs. This finding is consistent with previous studies who investigated the differences of ARGs in swine from the same farm over time and found that similar genes were consistently observed among samples from different stages in the production process (Petrin et al., 2019) and also at similar concentrations (Wen et al., 2019). Our results combined with these previous studies suggest that despite higher quantities of antibiotics administered to younger weanling pigs than mature growers (Dunlop et al., 1998; Dewey et al., 1999), the concentrations of these ARGs in manure do not change significantly. Overall, our results also indicate that the integrator is a larger source of variation among these genes than production stage and highlight the opportunity to engage in AMR stewardship towards integrators in partnership with farms (Hayes, 2022; Mitchell et al., 2022).

We also studied the detection of other ARGs to expand this study beyond ermB and tetM by leveraging high-throughput qPCR (HT-qPCR) methods which allow simultaneous testing of multiple gene probes. ARG targets were selected based on published primers (Stedtfeld et al., 2018) of ARGs previously observed to be present in swine manures ( Table 1 ). Between manures, the tetracycline resistance gene class was the most prevalently detected in our samples, which is consistent with its wide use in swine production (Center for Veterinary Medicine, 2020). Likewise, the macrolide resistance gene class, erm, had the second most detected antibiotic resistance genes and is consistent with previous literature (Whitehead and Cotta, 2013; Joy et al., 2014). For instance, a study by Wen et al. (2019) studied nine ARGs at 18 different swine farms and found tetO as the predominant gene in manure and tetQ, tetW, ermB, and ermF were identified as having the highest risk of spread to the soil and water environment through manure application. Moreover, a study by Mu et al. (2015) took manure samples right after defecation from swine in nine feedlots in China finding oqxB (plasmic mediated quinolone) as the highest detected ARG followed by sul1, sul2, tetO, tetM, and ermB. Surprisingly, sul1 and sul2 were only detected 11.1% and 23% respectively, in the manure storage pits from the current study, suggesting a temporal shift in ARG presence between fresh manure and stored manure. Finally, a study of manure from three swine farms in China measured 28 tetracycline resistance genes and reported detection of 22 with the most common genes tetA, tetL, tetM, and tetG (Zhu et al., 2013), whereas in the current study, tetA and tetL were not detected in any of the 14 farms. These variations in detected classes of ARGs among studies and farms are speculated to be caused by differing antibiotic treatments, legacy resistance in piglets passed down by the maternal gut (Pärnänen et al., 2018), and co-selection of resistance genes (Looft et al., 2012).

Compared to conventional qPCR, fewer detections of ARGs were observed on HT-qPCR, most likely due to a combination of both the significantly reduced reaction volume (and thus lower limit of quantification) and presence of inhibitors (Funes-Huacca et al., 2011; Sandberg et al., 2018; Luo et al., 2021; Keenum et al., 2022). Specifically, we observed ermB and tetM gene detection in 100% of manure samples with conventional qPCR but 50% with HT-qPCR. To better understand these results, we compared the lower limit of quantification for ermB and tetM for traditional qPCR and HT-qPCR and found that traditional qPCR was 63 (ermB) and 94 (tetM) times more sensitive than HT-qPCR ( Supplementary Tables 2, 3 ), suggesting that limit of quantification contributed to the inconsistency among ARG detections. Additionally, the DNA for the HT-qPCR assays were diluted 500:1 to balance measuring high 16S-rRNA gene copies, enabling the detection of low concentration ARGs, and reducing inhibitor effects. We conclude that the combination of diluting DNA and the HT-qPCR’s significantly reduced reaction volume contributed to the inconsistent detection of ARGs. This observation should be considered in selecting monitoring methods for ARG detection in future studies. Although HT-qPCR is not as robust as conventional qPCR, the advantages of this method are its ability to simultaneously measure multiple gene targets, use of much less reagent per sample, and significantly reduced labor. We recommend that HT-qPCR be used to screen the presence or absence of diverse ARG targets in environmental samples, and conventional qPCR be used for more rigorous quantification.

While ermB and tetM were inconsistently detected with HT-qPCR methods, there were specific genes that were broadly present using this method. Specifically, the tet36 and erm35 genes, encoding resistance to tetracyclines and macrolides respectively, were detected more frequently with the HT-qPCR than their counterpart tetM and ermB. This suggests that tet36 and erm35 are consistently associated with swine manure and able to be detected with current high throughput methods. The tet36 gene was first discovered in swine manure pits, and is yet unclear whether it is enriched or persists in the environment upon manure application (Whittle et al., 2003; Kang et al., 2018; He et al., 2019). Less is known about the erm35 gene, except that it was detected in poultry manure with metagenomics (Błażejewska et al., 2022; Wang and Chai, 2022). The erm35 gene may have potential as a swine indicator since it was detected so frequently with HT-qPCR in the current study. One major difference between the two sets of genes is their association with mobile genetic elements (MGEs) where ermB and tetM are highly associated with MGEs while erm35 and tet36 are not (Zhang et al., 2022). MGEs are associated with the mobility of ARGs, which may be a significant variable for the dissemination of the gene after manure application. The class-3 MGE inti3 was present in half of the manure samples tested in the ARG survey, and this is significant as this gene has the potential for horizontal gene transfer (Martínez et al., 2015). We highlight these genes tet36 and erm35 as potential targets for swine manure borne resistance.