Host genetics affected the resistome and its expression patterns in the rumen of beef cattle raised without antibiotics used in humans

Background: The rumen microbiome is a potential reservoir of antimicrobial-resistant genes (ARGs), termly resistome. However, the activity of ARGs and what factors affect expression of ARGs in the rumen is unknown. Here, the rumen resistome was evaluated using metagenomic and metatranscriptomic datasets, with the aim to identify the active rumen resistome and whether it can be affected by cattle breed and feed eciency. Results: Genes encoding resistance to 12 ARG classes representing 62 individual ARGs were detected in the rumen metagenomes of Angus, Charolais or Kinsella composite hybrid (KC) beef steers (n = 48) with high and low feed eciency. Three genes encoding tetracycline (tetQ, tetW) and macrolide (mefA) resistance constituted 75.3% of abundance of total ARGs identied in all animals, suggesting they are ‘core’ resistome in the rumen of steers. Only about 20.96% (13/62) of the total ARGs identied were expressed, among which genes encoding resistance to tetracycline, macrolide-licosamide-streptogramin (MLS), aminoglycoside, and multidrug exhibited the highest level of expression. More than half (56.2%) of the ARGs identied were plasmid-associated, while only 5 plasmid-associated ARGs were expressed. The abundance of 17, 14, and 5 individual ARGs were signicantly affected by breed, feed eciency, and breed × feed eciency, respectively, while the expression of ARGs did not differ among breeds or between feed eciency groups. In KC cattle, less number of total ARGs, ARG transcripts, as well as total active bacteria (estimated by 16S rRNA copies) was observed than AN. The total active bacteria were negatively correlated with expression of MLS and tetracycline ARG (mefA, tet40, tetM, tetW, and an unidentied tet), and tended to be negatively correlated with the expression of plasmid-associated tetracycline ARG t only in the rumen of KC cattle. Conclusions: Our results suggest that a large portion of the ARGs are not expressed in the rumen of cattle raised without antibiotics used in humans. The identied less diversied active resistome and total active bacteria, and the signicant correlation between total active bacteria and the abundance of ARG transcripts in KC cattle suggest that the expression of resistome in the rumen may be breed specic and driven by ruminal microbiota.

bacteria, and the signi cant correlation between total active bacteria and the abundance of ARG transcripts in KC cattle suggest that the expression of resistome in the rumen may be breed speci c and driven by ruminal microbiota.

Background
Antimicrobials have been widely used in food producing animals since 1950th to enhance feed e ciency, accelerate growth, and minimize disease [1]. It is estimated that antimicrobials used to prevent/treat disease and/or promote growth in chickens, pigs, and cattle will increase from 63,151 tonnes in 2010 to 105,596 tonnes in 2030 [2]. Consumption of antimicrobials in livestock, as well as antimicrobial residues in food have been proposed to contribute to antimicrobial resistance (AMR) in humans [3,4] and aquatic/soil environments [5,6]. The AMR found in food-producing animals not only reduces the therapeutic e cacy of antimicrobials against diseases but also selects for reservoirs of resistance that could be transferred to humans via the food chain or through the environment [7]. Therefore, reducing antimicrobial resistance and preventing antimicrobial residues from entering the food chain is a priority for livestock industry to address the food safety and public health concerns [8].
In fact, AMR in bacteria is an ancient phenomenon and was present long before the widespread clinical and agricultural use of antimicrobials [9]. Numerous antimicrobial resistance genes (ARGs) encode for resistance to an array of antimicrobials [9] that bacteria produce to compete and survive within complex ecological systems [10]. Recent efforts have documented the rumen resistome, a vast reservoir of ARGs that could be acquired by human commensals and pathogens [11] in beef cattle [12][13][14], dairy cattle [14], and sheep [15], and revealed that ARGs can be affected by diet [13] or antibiotic treatment [12]. However, these studies only assessed ARGs at genomic level using metagenomics based on short-read sequencing [12,13,15] or long-read sequencing [14] and with a limited number of animals were involved. On the other hand, gene expression is a better proxy for assessing functional activity within biological ecosystems [16]. It is largely unknown the extent to which ARGs are expressed in the rumen of cattle that are not under the selective pressure antibiotics used in human is unclear. Recent studies have assessed wastewater treatment plant resistome at both metagenomic and metatranscripome levels and reveals that plant locations not only affected the ARGs but also their transcripts [7,17]. However, comparing to largely identi ed resistomes in the animal system, their functionality and activity in vivo has not been reported.
More and more evidences have revealed the individualized rumen microbiome when animals are fed the same diet and host genetic factors have been reported to drive pan microbiome [18]. We hypothesized that 1) rumen resistome, similar as rumen microbiome can be affected by cattle genetics such as breed and/or feed e ciency; 2) not all ARGs are expressed and the expression of ARGs, including plasmidassociated ARGs, can be affected by active microbiome and/or cattle genetics; 3) as rumen microbiome differs in animals between high and low feed e ciency, and less diverse active rumen microbiome was associated with higher feed e ciency in cattle [19], we also hypothesized that animals selected with high feed e ciency may have higher expression of ARGs because lower bacterial diversity is associated with enriched resistome in human taken antibiotics [20]. Therefore, in this study we assessed the presence and expression of ARGs in the rumen of beef steer differing in breed and feed e ciency raised without antibiotics used in human medicine. Understanding of the expression of ARGs in the rumen could provide important insights into resistome function and if factors such as host genetics or antibiotic usage selective pressures play a role in AMR function in vivo.

Animal experiment and sample collection
The datasets used in the current study were part from our previous study by Li et al. [21]. In brief, ruminal digesta samples were collected from 48 steers, consisting of three breeds and two residual feed intake . H-RFI, which was 1.45 ± 0.17 kg of body weight/day, was considered as ine cient while L-RFI, which was − 1.64 ± 0.21 kg of body weight/day, was considered as e cient as described by Li et al. [21]. All steers were raised in the same feedlot condition and fed the same diet that consisted of 80% barley grain, 15% barley silage, and 5% supplement. The supplement contained 33 ppm monensin, an antibiotic that is not used in human medicine.
Metagenome and metatranscriptome sequencing Rumen digesta was collected at slaughter and snap-frozen in liquid nitrogen as described by Li et al. [21]. Total metagenomic DNA and RNA were isolated from rumen digesta using the methods of Yu and Morrison [22] and Li et al. [23]

Detection of ARGs and plasmid-associated ARGs
The prevalence and abundance of the rumen resistome pro le of each metagenomic dataset was determined using a two-stage (ARG-OAP 2.0) pipeline [24]. Brie y, post quality-controlled reads (pairedend) from each sample were blasted against the Structured ARG database (SARG), comprised of the Antibiotic Resistance Genes Database (ARDB) and the Comprehensive Antibiotic Resistance Database (CARD), to extract ARG-like reads. Those reads were subsequently annotated as ARG-like reads at the cutoff of E value of 10 − 7 , sequence identity of 80% and alignment length more than > 25 amino acids. By using this cut-off, the identi cation accuracy can reach up to 99.5% [25].
Plasmid-associated ARG was determined using a modi ed ARG-OAP 2.0 pipeline. Instead of SARG, reads were blasted against the ACLAME database, a database for identifying mobile genetic elements [26], with a cut-off E value of ≤ 10 − 7 criteria with amino acid identity ≥ 80% and coverage ≥ 70%.

Abundance of ARGs and ARG transcripts
The abundance of ARG classes, total ARGs, and individual ARG were calculated with normalization for the sequence length and the number of 16S rRNA genes, as well as the ARG reference sequence length according to Yin et al. [24], which was de ned as 'copy of ARG per copy of 16S-rRNA gene', following the most recent approaches in the literature [7,27].
To evaluate the expressions of resistome, the extracted ARG-like sequences identi ed in metagenomic datasets were used as a reference database to support extraction of ARG-like transcripts from the metatranscriptome datasets using ARG-OAP 2.0 pipeline. Reads were annotated as ARG-like transcripts using the same cutoffs as described above. The abundance of ARG classes transcript, total ARG transcripts, and individual ARG transcript was reported as 'ppm' (number of ARGs sequences in one million sequences) following a method reported by Yin et al. [24].
Estimation of active rumen microbiome using qPCR Total RNA (1 µg) was reversely transcribed using iScript reverse transcription Supermix for quantitative real-time PCR kit (qRT-PCR; Bio-Rad Laboratories, Hercules, CA) to generate complementary DNA (cDNA). The 20 times-diluted cDNAs were used to evaluate the abundances of total active bacteria by measuring its copy number of 16S rRNA using qPCR with bacterial universal primers bacteria (U2-F: ACTCCTACGGGAGGCAG; U2-R: GACTACCAGGGTATCTAATCC) [28]. The qPCR was performed using SYBR Green chemistry with StepOnePlus Real-Time PCR System (Applied Biosystems) and the programs were as follows: for bacteria, the holding stage at 95 ℃ for 5 min, followed by 40 cycles at 95 ℃ for 20 s and 60 ℃ for 30 s. Standard curves were constructed using serial dilutions of puri ed plasmid containing full length 16S rRNA gene of Butyrivibrio hungatei for total bacteria. The copy number of 16S rRNA of total bacteria/g rumen contents was calculated with an equation previously described by Malmuthuge et al. [29].

Statistical analyses
The cut-off for detected ARG class or individual ARG was abundance > 0 in at least half samples for one breed. Variation in the number (prevalence) and abundance of ARGs and ARG transcripts among breeds, between H-an L-RFI, as well as their interactions (breed × feed e ciency), were analysed using the aligned rank transform (ART) method [30], using 'ARTool' package in R (version 3.6.1). The ART analysis was only conducted when an ARG class/ARG class transcript or individual ARG/ARG transcript was detected in more than 3 samples in all three breeds. The P-value of multiple comparison of breed effect was adjusted into false discovery rate (FDR) using the Benjamini-Hochberg algorithm using 'dunn_test' in R (version 3.6.1). Circos plot analysis was performed in R using the RCircos package [31]. Principle componenet analysis (PCA) of ARGs and ARG trasncripts was conducted using multivariate analysis of variance (MANOVA). Spearman correlation analysis was performed for the abundance of expressed ARGs and log-transformed total active bacteria copy numbers and the results were visulized using 'ggscatter' in R (version 3.6.1). The Spearman's correlation coe cient, known as rho (ρ), ranges from − 1.00 (a perfect negative correlation) to + 1.00 (a perfect positive correlation). Signi cant difference was declared at P ≤ 0.05 and tendencies at 0.05 < P ≤ 0.10. For Spearman correlation, a rho value between 0.40 and 0.65 and P value ≤ 0.05 is considered as signi cant correlation, while P value between 0.05 and 0.10 is considered as tendency towards correlation.

Comparison of class of ARGs and ARG expression pro les among breeds
Sixty-two ARGs belonged to 12 ARG classes were detected in all three breeds, with the exception of genes encoding fosfomycin resistance which were only found in the rumen of 18.7% (3/16) KC animals (Additional le 1: Table S1). Seven of 12 ARG classes were detected with transcriptional activity in all three breeds (Additional le 1: Table S2). Principle component analysis (PCA) showed no separation of the abundance of ARG classes (Fig. 2a) or the expressed ARG class among breeds (Fig. 2b).
The number of total ARGs detected was lower in KC than in AN (53.2 vs 60.0, P = 0.026) and CH (53.2 vs 63.7, P < 0.001) (Fig. 3a). The number of ARGs belonging to MLS (P = 0.003; Fig. 3b), MDR (P = 0.010; Fig. 3c), and vancomycin (P = 0.004; Fig. 3d) was also lower in KC than CH. The number of ARGs belonging to MLS was also lower (P = 0.003, Fig. 3b) in KC than AN. The number of total ARG transcripts (12.2 vs 15.8, P = 0.012; Fig. 4a), and the number of tetracycline ARG transcripts (6.5 vs 7.8, P = 0.033; Fig. 4b) was lower in KC than in AN.
The abundance of total ARG was higher in KC (P = 0.022) and CH (P = 0.002) than in AN, with no difference between KC and CH (P = 0.684) ( Fig. S1a; Additional le 1: Table S1). The abundance of aminoglycoside resistant genes was higher in KC than AN (P = 0.035) (Fig. S1b) while that of fosfomycin resistant genes was lower in KC than AN and CH (P < 0.001) (Fig. S1c). The abundance of MLS (P = 0.010) (Fig. S1d) and MDR (P = 0.006) (Fig. S1e) resistant genes was higher in CH than AN whereas that of tetracycline (Fig. S1f) resistant genes was lower in AN than KC (P = 0.028) and CH (P = 0.017). At transcripts level, no difference in abundance of any expressed ARG class or individual ARG was detected among breeds (Additional le 1: Table S2).
At transcripts level, no difference in the abundance of any expressed ARG class or individual ARG was detected among breeds (Additional le 1: Table S2).

Differential ARGs and ARG transcripts between RFI
The 12 ARG classes were also detected in both H-RFI and L-RFI beef steers, with the exception that genes encoding polymyxin resistance were only found in the rumen of 33.3% (8/24) of H-RFI animals (Additional le 1: Table S1). Seven of 12 ARG classes were detected with transcriptional activity in both RFI groups (Additional le 1: Table S2). Principle component analysis (PCA) showed no separation of the abundance of ARG classes (Fig. 6a) or the expressed ARG class between H-RFI and L-RFI beef steers (Fig. 6b).
Identi cation of Plasmid-associated ARGs and ARG transcripts and effect of feed and RFI on their abundance Ten classes of ARGs were annotated as plasmid-associated by comparison against the ACLAME database, with abundances of 7.82%, 7.50%, and 7.17% for KC, AN, CH, respectively ( Fig. S2a; Additional le 1: Table S3). A total of 34 of 62 (54.8%) ARGs detected were plasmid-associated (Additional le 1: Table S3). None of the abundance of plasmid-associated ARG class was affected by breed or e ciency. However, breed and e ciency interactively affected chloramphenicol resistance genes (P = 0.016). A total of 28, 27, and 31 plasmid-associated ARGs was detected in KC, AN, and CH, respectively, among which 22 were shared by three breeds (Fig. S2b). The abundance of ant(9)-I (P = 0.006) and tet44 (P = 0.035) was higher, while that of ermG (P = 0.036), and lnuA (P = 0.005) was lower in CH than KC (Fig. S2c). The abundance of ermB (P = 0.008) was higher, while that of vatB (P = 0.005) was lower in AN than CH. The abundance of vatB (P = 0.003) and aadE (P = 0.038) was higher, while that of ermB (P = 0.014) was lower, and lnuA (P = 0.097) tended to be lower in KC than AN (Fig. S2c). Breed and e ciency interactively affected the abundance of lsa (P = 0.038).
Transcripts of resistance to aminoglycoside and tetracycline belonged to plasmid-associated ARGs (Additional le 1: Table S4). Transcript reads belonging to 5 of 34 detected plasmid-associated ARGs in metagenomics data were found in transcriptomic data, among which the expression of tetW was observed in all samples. No effect of breed, e ciency, or their interactions on the abundance of any plasmid-associated ARG transcripts was found.

Total active bacterial population and its relationship with active resistome
The log-transformed total active bacteria copy number was lower (9.52) in KC compared with AN (11.0) and CH (10.6) (P < 0.001; Fig. 8a) and was not differ between H-RFI (10.4) and L-RFI (10.4) (P = 0.963) (Additional le 1: Table S5). No interactive effect of breed and feed e ciency was observed on total active bacterial population (P = 0.993).

Discussion
In this study, metagenomic analysis revealed that the abundance of ARGs in the rumen of beef steers were predominant by tetracycline (77%) and followed by MLS (17%), and aminoglycoside (4%) resistance. These ndings are similar to those reported in fecal samples of beef cattle fed with antibiotics (ionophores, chlortetracycline, or tylosin), where tetracycline resistance was most prevalent (82%), followed by macrolide (14%), and aminoglycoside [32]). In addition, tetracycline, MLS, and aminoglycoside classes of resistance were also predominant in fecal samples of feedlot cattle raised without antibiotic [33], suggesting that the pro les of ARG are consistent in different locations of digestive tract of ruminants. We detected broader ARG pro les (12 classes and 62 individual ARGs) in the rumen of beef cattle not administered antimicrobials used in human medicine than previously reported by Thomas et al. [12] who studied the beef cattle supplemented with tylosin. Based on the analysis of 5 ruminal samples in cattle supplemented with monensin and tylosin, Thomas et al. [12] did not detect aminoglycoside or β-lactam resistance genes in any sample. While Auffet et al. [13] detected a wide range of genes resistant to macrolide, chloramphenicol, β-lactam, and aminoglycoside in the rumen of antimicrobial-free beef cattle under similar feeding condition to our study, however, genes resistant to vancomycin were not detected in their study. The variation could be due to difference in animal, environment and diet. In addition, bioinformatic resources/tools available for resistome analysis may also contribute to the difference in ARG pro les in the rumen among studies. Presently, there are at least 47 bioinformatic resources/tools, but no a 'standard' pipeline has been developed speci cally to characterize the resistome. In this regard, the results are heavily dependent on the analysis methods (assembly-based or read-based) or reference database [34]. In this study, the ARG-OAP (v2) pipeline was applied, which uses a custom database with a hybrid UBLAST and BLASTX algorithm, re ecting the critical need for a comprehensive database combined with lower identity matching for antimicrobial resistance gene annotation of metagenomic data [35]. However, as there is no inclusive ARG database or one speci cally customized for the rumen microbiome, more efforts are needed to construct 'standardized' pipeline for to characterize the rumen resistome and resistomes in other habitats (e.g. soil, water, gastrointestinal tract).
Plasmids are mobile genetic elements found in high abundance in the bacterial populations of bovine rumen [36], which play a major role in the spread of antimicrobial resistance through horizontal gene transfer [37]. Metagenomic approaches have been used to characterize plasmid encoded ARGs in several non-biological habitats such as activated sludge [38,39] as well as the human gut [40]. We found that aadA and tetW were the most abundant plasmid-associated ARG in the rumen of beef steers. In addition, the expression of tetW was highest among all plasmid-associated ARG transcripts. It has been reported that many of the tetracycline resistant genes are associated with mobile plasmids [41], among which tetW has been proven to be transmissible among the ruminal bacteria, Butyrivibrio brosolvens, Selemonas ruminitanium, and Mitsuokella multiacidus [42]. However, the pro les and expressions of plasmid-associated ARGs have not been examined in food-producing animals including ruminants. Considering that the expression of mobile genetic elements such as integrons is a robust strategy of genetic interchange and one of the main drivers of bacterial evolution [43], we speculate that the expression of plasmid-associated ARGs has functions other than transferring antimicrobial resistance in the rumen. Recently, a wide range of bacterial hosts of plasmids in wastewater samples has been revealed by analyzing Hi-C and shotgun metagenomic data [44]. Those approaches can also be applied to investigate plasmid-associated ARG as well as their bacterial host in cattle, which may help understand the contribution of plasmids to the transmission of AMR determinants in the rumen.
Considering that the presence of a gene does not directly correlate with the activity of the gene a certain environment, direct measurements of transcripts based on metatranscriptomics may be an important complementary approach to metagenomics. Our results indicated that the expression of ARGs is also not directly linked to the presence of ARGs as previously shown in environmental microbiome [17]. We found that about only 20.96% (13/64) of ARGs were expressed, suggesting that around 80% of ARGs were not functional in the rumen of these steers at the time of sampling. Among the 13 ARGs expressed, the prevalence of tet40, tetM, tetO, tetW, mefA was 100%, while that of aadA, tetO, and vatB was 77.1%, 70.8%, and 58.3%, respectively. This suggests that these eight ARGs may constitute the 'core' active resistome in the rumen of the steers studied in our study. In particular, the average abundance (ppm) of tetW, mefA, tetQ, and tet40 was 19.84, 13.61, 7.64, and 6.38, respectively, which were the predominant ARG transcripts in our study. The mechanisms of action of these resistant genes have been well characterized. Both tet40 [45] and mefA [46] encode for e ux pumps which render antimicrobials ineffective by pumping them out of the cell, while tetQ [47] and tetW [48] encode for tetracycline ribosomal protection proteins. Among these ARGs, the expression of tet40 has been detected in Clostridium species in human [49] and swine [45] gut. In our previous study, we observed active Clostridium genus (the relative abundance averaged 0.15%) across three breeds of beef cattle [21], and we thus speculate that it may contain certain Clostridium species that carry tet40 gene. To our knowledge, our study reported for the rst time the presence and active ARGs simultaneously for foodproducing animals in vivo with a large dataset. Although Sabino et al. [27] analyzed the expression of rumen ARGs, only 15 metatranscriptomic samples were used (5 dairy cattle, 5 beef cattle, and 5 sheep) and their aim was to con rm the expression of ARGs found in 435 reference genomes of ruminal bacteria and archaea in silico, but not to link the expression of ARGs back to the presence of those ARGs using metagenomic data. More recently, the resistome in chicken and pig gut were analyzed using both metagenomic and metatranscriptomic data, but only 6 fecal samples were used as representative of gut samples for each species [50]. In this regard, more efforts are needed to detect and validate our ndings based on both metagenomic and metatranscriptomic analysis. We speculate that besides acting against antimicrobial present in the environment, the detected ARGs in the rumen may have functions in addition to antimicrobial resistance, which deserves further investigation.
It has been reported that the prevalence and abundance of ARGs in the gut of cattle is affected by diet.
For example, dietary transition from milk replacer to starter led to alternation in the fecal resistome of dairy calf [7]. In addition, the diversity and abundance of total ARGs were higher in the rumen of beef cattle fed high concentrate than those fed high forage diet, with chloramphenicol and aminoglycoside resistance genes being predominant in forage-and concentrate-fed cattle, respectively [13]. A recent study also suggested that the dietary supplementation of tulathromycin, a macrolide antimicrobial drug used as metaphylaxis, signi cantly affected the temporal development of fecal microbiota and associated resistome in feedlot cattle [51]. To our knowledge, there is no study reporting how host genetic factors affect the active gut resistome in mammalian species. In food-producing animals such as beef cattle, understanding the 'host-resistome' association may be a prerequisite to select breeds with high feed e ciency and low risk of ARG transmission to the environment, as the gut microbiome that harbor ARGs has been proved to be largely host-driven [21,[52][53][54]. In this study, all beef steers were raised under the same dietary and environmental conditions, suggesting that the prevalence and expressions of ARGs were driven by host genetic factors such as breed and feed e ciency. On the contrary to the ndings by Auffret et al. [13], who didn't observe breed effect on the abundance of rumen microbiota and abundance of ARGs in beef cattle, we not only observed a signi cant difference in both prevalence and abundance of ARGs, but also the prevalence of ARG transcripts among three breeds. Speci cally, we detected less type of ARG transcripts, especially tetracycline resistant gene transcripts, in the rumen of crossbred (KC) compared with purebred (AN), which may be explained by less copy number of total active bacteria in the rumen of KC than AN animals. Besides, our previous study also indicated that the active phylum Bacteroidetes, which account for a high proportion of the microbial genomes (e.g. species belonging to Prevotella and Bacteroides) that harbor resistance genes in the rumen [27], was less abundant in KC compared with the other two breeds [21]. The signi cant correlation between total active bacterial population and the abundance of ARG transcripts observed for KC cattle only further support that the expression of resistome in the rumen may be host breed speci c and driven by ruminal microbiota. However, it is not clear why copy number of total active bacteria is negatively correlated with the abundance of multiple tetracycline and macrolide ARG transcripts. In this regard, the active bacterial host of those ARG transcripts deserves further investigations using the pure cultures.
It has been proved that rumen microbiome differs in beef cattle with high and low feed e ciency [21], which may explain the difference in the prevalence (e.g. ARGs belonging to vanconmycin, aminoglycoside, and MLS) and abundance of several ARGs (e.g. tetX, vatE, and lnuA) between H-RFI and L-RFI beef cattle based on metagenomic data. However, H-and L-RFI steers share a similar ARG transcript pro les, suggesting that ruminal fermentation capacity may not be a main factor driving the expression of ARGs. Our results also showed that breed × feed e ciency interactions only affect the abundance of ARGs, but not ARG transcripts. Taken together, the lack of feed e ciency and interaction effect suggest that host breed is the main drive of rumen resistome of beef steers.

Conclusions
In the current study, we not only detected a comprehensive ARG pro le but also discovered 'active' resistome in the rumen of beef steers that were not administered those classes of antimicrobials used in human medicine based on both metagenomic and metatranscriptomic analysis. Our major ndings include, rst, not only the existence but also the expression of ARGs in the rumen are not necessarily associated with the use of antimicrobials, suggesting that the detected ARGs in the rumen may have functions in addition to antimicrobial resistance. Second, comparing with the diverse ARGs detected in the rumen, their expression level of both number and abundance of transcript is relatively low. It is plausible that there is direct relationship between the active rumen bacterial population and the active resistome. The bacterial origin, function, as well as the mechanisms of action of the active resistome needs to be veri ed in future studies. Third, breed exhibits a stronger effect on ARGs and their expressions compared with feed e ciency. In particular, KC has a less diversi ed ARG transcripts, which may be explained by the lower copy number of total active bacteria in the rumen compared with AN. Beef steers differing in RFI steers share a similar ARG transcript pro les regardless of breed, suggesting that the rumen resistome may not be a concern for future selection of beef steers with high feed e ciency.
One potential limitation of the current study is that resistome was only analyzed in the rumen of monensin supplemented beef steer only, and it is unclear whether rumen resistome would have been the same if no antibiotic was administered. In this regard, comparative analysis of transcriptional pro le of samples originating from cattle raised with and without antibiotics is warranted to compare the impact of the presence of antimicrobial residue on the expression of ARGs. Regardless, the ndings from this study provide new insight into the active rumen resistome without the antibiotic selective pressure, which may be essential to develop strategies for limiting the spread of antimicrobial resistance from rumen to the environment.       Principle component analysis (PCA) showed no separation of the abundance of ARG classes (Fig. 6a) or the expressed ARG class between H-RFI and L-RFI beef steers (Fig. 6b).   Total active bacteria copy number and its relationship with the abundance of expressed ARGs. a Comparison of log-transformed total active bacteria copy number in the rumen of KC, AN, and CH beef steers. b Correlation between the abundance of mefA transcript and log-transformed total active bacteria copy number in KC beef steers. c Correlation between the abundance of tet40 transcript and logtransformed total active bacteria copy number in KC beef steers. d Correlation between the abundance of tetM transcript and log-transformed total active bacteria copy number in KC beef steers. e Correlation between the abundance of tetW transcript and log-transformed total active bacteria copy number in KC beef steers. f Correlation between the abundance of unidenti ed tet transcript and log-transformed total active bacteria copy number in KC beef steers. KC, Kinsella composite hybrid.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.