Antimicrobial resistance determinants in silage

Animal products may play a role in developing and spreading antimicrobial resistance in several ways. On the one hand, residues of antibiotics not adequately used in animal farming can enter the human body via food. However, resistant bacteria may also be present in animal products, which can transfer the antimicrobial resistance genes (ARG) to the bacteria in the consumer’s body by horizontal gene transfer. As previous studies have shown that fermented foods have a meaningful ARG content, it is indicated that such genes may also be present in silage used as mass feed in the cattle sector. In our study, we aspired to answer what ARGs occur in silage and what mobility characteristics they have? For this purpose, we have analyzed bioinformatically 52 freely available deep sequenced silage samples from shotgun metagenome next-generation sequencing. A total of 16 perfect matched ARGs occurred 54 times in the samples. More than half of these ARGs are mobile because they can be linked to integrative mobile genetic elements, prophages or plasmids. Our results point to a neglected but substantial ARG source in the food chain.

www.nature.com/scientificreports/ regarding the mobility potential of ARGs were also resumed based on genetic characteristics that may play a significant role in HGT.
Bacteriome. By taxon classification, the number of reads aligning to bacterial genomes varied by samples (median: 20.6 × 10 6 , IQR: 2.9 × 10 6 ). The relative abundances of genera that achieved more than 1% of the bacterial hits in any of the samples are shown in Fig. 1.

Mobilome.
We found a total of 53 ARGs that are assumably mobile. Ten of these ARGs are linked to integrative mobile genetic elements (iMGE). A further two ARGs were detected in prophages and forty-one on plasmids. The frequencies of ARGs associated with iMGEs, phages and plasmids are summarized in Fig. 3 by bacterial species of origin.

Figure 1.
Silage core bacteriome. Relative abundances of genera that achieved more than 1% of the bacterial hits in any of the samples. The elements of the PRJNA495415 dataset were taken on days 0, 7, 14 and 28 were classified into groups A, B, C and D, respectively. All items from BioProject PRJNA764355 are assigned to group E.

Discussion
Throughout our study, numerous perfect ARG matches were identified in the metagenome of Medicago silage samples. All but group A of the analyzed subsets had at least one sample containing one or more ARG. Among the PRJNA495415 Bioproject samples, the highest number of ARGs were found in group D. The interpretation of this finding is limited due to the lack of detailed information on the samples. Interestingly, all but one of the PRJNA764355 bioproject samples contained ARGs. Due to the lack of metadata, it is hard to find any reason  www.nature.com/scientificreports/ for this high ARG level. However, one possible cause might be that the PRJNA764355 samples were sequenced deeper and thus contained approximately 1.3 times more reads than the PRJNA495415 samples. It is known from previous studies that deeper sequencing leads to the generation of more complete genes by the de novo assembly 5,7 . In the following, our results will be interpreted from a perspective of bacteriological significance, genomic relevance and furthermore, antimicrobial stewardship and possible clinical aspects.
Taking the microbiome into consideration, bacteria that were predicted to harbor the identified ARGs can be classified according to their presence in silage. In the literature, the following bacteria are mentioned to be characteristic for silage: B. subtilis 8 E. faecium 9 E. coli 10 , L. plantarum 11 , L. lactis 12 , L. brevis [13][14][15] , L. acidipiscis 12 W. paramesenteroides 16 . The genera of these species dominate the bacteriome of the samples. The identified Cronobacter sp. JZ38 17 may be of plant origin. However, it can be assumed that other species may be present as contaminants of the silage: A. amylophilus, E. hormaechei, E. faecalis, Gracilibacillus sp. SCU50, L. manihotivorans, L. paracasei, P. rettgeri, S. aureus, S. carnosus, S. pseudoxylosus, S. saprophyticus, S. suis, T. halophilus. Nevertheless, some of these bacteria are members of the Lactobacillaceae family, the Leuconostoc or Enterobacter genera. Numerous species of these groups are typical for fermented food and feed components.
From a genomic point of view, the following was found in the literature regarding the co-occurrence of the ARGs identified in our study and the bacteria carrying them. AadA2 encoding an aminoglycoside nucleotidyltransferase has been described in A. baumanni in former publications 18,19 . ant(6)-Ia, that is an aminoglycoside nucleotidyltransferase gene, appears in many species, including Lactobacillus spp. 20 . Its species-specific association with A. amylophilus has not been described in any former publications. aph(3')-IIa, an aminoglycoside phosphotransferase 21 , to our knowledge, has not been detected in B. subtilis up untill now. QnrS1 encoding a quinolone resistance protein was originally identified in Shigella flexneri 22 . In line with our results, this gene has recently been mentioned to appear in Cronobacter spp. in a case report 23 . E. hormaechei deriving aadA2 and sul1, a sulfonamide resistant dihydropteroate synthase gene that is described to appear in Gram-negative bacteria 21 have been reported to appear in the genom of Enterobacter spp. and E. hormaechei, respectively in former publications as well 24,25 . Within the Enterococcus genus, two perfect ARG matches were identified, namely aph(3')-IIIa in E. faecalis and poxtA in E. faecium. aph(3')-IIIa is an aminoglycoside phosphotransferase that normally appears in S. aureus 21 and Enterococcus spp. 26 , while poxtA is a gene encoding an ABC-F subfamily (ATP-binding cassette-F) protein that facilitates resistance to tetracycline, phenicol, and oxazolidinone via modification of the bacterial ribosome. First detection of poxtA took place in a methicillin-resistant S. aureus strain 21 , followed by other bacterial species, including E. faecium 27 . Sul2, a sulfonamide resistant dihydropteroate synthase of Gramnegative bacteria is commonly described in E. coli 21,28 . DfrG is a plasmid-encoded dihydrofolate reductase 21 that, to our knowledge, has not been described in Gracibacillus spp. up untill now, but has already appeared in the Bacillaceae family 29 . ant(6)-Ia, an aminoglycoside nucleotidyltransferase gene appears in many species, including Lactobacillus spp. 20 . Its species-specific association with L. manihotivorans has not been described in any publications. PoxtA that was detected in L. paracasei, L. plantarum and L. brevis in the silage samples, has been described to appear in Lactobacillaceae, namely L. acidophilus, but not in these very species 30 . Another www.nature.com/scientificreports/ species that was detected harboring aph(3')-IIIa in the silage samples was L. plantarum. This finding is in line with the ARG-species match results mentioned in former publications 31 . Furthermore, L. plantarum was also associated with vatE that encodes an acetyltransferase conferring resistance against streptogramins 21 . VatE was originally found in E. faecium 21 and has since then been identified in Lactobacillaceae 32 , but not specifically in L. plantarum. L. acidipiscis ant(9)-Ia, an aminoglycoside nucleotidyltransferase gene 21 was associated with this genus for first within this study. Gene qnrD1 encoding a quinolone resistance protein that is normally detected in Salmonella enterica 21 , has already been found in Providencia spp. 33 and was attached to P. rettgeri in our study as well. S. aureus could have been associated with two ARGs, ant(6)-Ia and tetK encoding a tetracycline efflux protein, that are both common findings in Staphylococcus spp. 34,35 . Although, erm (44)v was first detected in the S. saprophyticus 36 , no literature could be found about the appearance of this gene in S. carnosus or in S. pseudoxylosus species. lsaE encoding another ABC-F subfamily protein conferring resistance to pleuromutilin, lincosamide, and streptogramin A is a common finding in Streptococcus spp. 37 and has also been associated with S. suis in previous publications 38 . Besides the bacterial species mentioned above, aph(3')-IIIa was also detected in T. halophilus. This ARG is often appears in Enterococcaceae 21 but has not yet been written down in this species. Furthermore, to our knowledge, W. paramesenteroides associated ant(6)-Ia has first been detected in this study. Throughout our study, several ARGs were predicted to be co-occurring with genetic attributes facilitating mobility. The bioinformatic analysis of the mobility characteristics relied upon the identification of three major mobility determination groups, namely iMGEs, phages and plasmids. We found aph(3')-IIa linked to an integrated mobile genetic element in B. subtilis that is in line with similar findings of E. coli. 21 While aadA2 and sul1 have both been described to appear on plasmids in E. hormaechei 39 , we found them associated with iMGEs. Our finding on iMGE flanked poxtA in E. faecium is in line with the current literature 40 . We found the same co-occurrence, namely poxtA and an iMGE, in L. paracasei. This phenomenon has not been published in that species to the best of our knowledge. Gene erm (44)v and lsaE were associated with prophages in S. pseudoxylosus and S. suis. While a similar linkage can be found in the literature in connection with erm(44)v 41 , no details of mobility characteristics are mentioned in a recent report of the latter gene 38 . All other mobile ARGs were detected on contigs that were predicted to derive from plasmids. In case of aadA2 in A. baumannii 21 ; aadA2 and sul1 in E. hormaechei 21 43 . Previous findings confirm the occurrence of vat(E) on plasmids 44 . Nevertheless, in spite of its frequent presence in enterococci 45 there is no evidence of its former plasmid-associated appearance in L. plantarum. We found that gene aph(3')-IIIa of T. halophilus was encoded on a plasmid that is consistent with the fact that aph(3') IIIa is often identified on high molecular weight plasmids and chromosomes of the enterococcal species 46 . Nonetheless, to the best of our knowledge, a description of the aph(3')-IIIa gene in T. halophilus is a pioneer finding.
The mobility characteristics of the ARGs may not only provide us with information regarding the public health risk that may be associated with the samples, but also point to the possible origins of the genes. Regardless of human intervention, ARGs are present in the microbial communities 47 . However, antimicrobial use and abuse intensifies the horizontal transfer of ARGs and thus contributes to the spread of AMR. In the animal production sector, the use of antibiotics is common, thus bacteria appearing in the feces and in the surroundings of the animals (e.g. in farm air, on tools, vehicles or other settings related to animals) often harbor bacteria with an advanced ARG set. Silage may get in direct physical contact with these bacteria at the farms and thus get contaminated with a few ARGs. Consequently, the presence of ARGs in the silage samples was well-expected, but the abundance of resistance genes and MGEs may increase due to the application of antibiotics.
Examining further aspects of antimicrobial stewardship and possible clinical relevance, phenotypical manifestations and public health considerations associated with the detected ARGs are both important. Intense antimicrobial use (AMU) can be associated with the headway of AMR, as antibiotic pressure selects for bacteria carrying ARGs that facilitate bacterial survival. Quantifically, the majority of AMU around the globe occurs in agricultural settings 48,49 . Intensive farming, that serves to fulfill the high global demand for animal proteins relies on an antibiotic infrastructure to treat and prevent disease and occasionally, to increase feed efficacy. In order to maximize economic gains, few countries still apply regulations that facilitate the use of low doses of antibiotics as growth promoters 50 , while other regions, like the U.S. or Europe, have banned this practice. Nevertheless, besides the treatment of symptomatic infectious diseases, antibiotics are still widely used in the livestock sector for metaphylactic and prophylactic purposes in higher doses 51,52 . Even though, compared to the poultry and pig production sector, average antibiotic usage has relatively lower rates by cattle 53 , antimicrobial compounds are often chosen in this species as well. In cattle farming, mastitis is the most predominant reason for the administration of antibiotics by adult cattle, while enteritis and pneumonia is the most common reason for calves 54,55 . According to various reports and studies from around the world 54,56,57 tetracyclines are of inevitable significance in the medication of cattle, while beta-lactams, macrolides, sulfonamides, lincosamides and ionophore antibiotics are also very widely used. Of the European Medicines Agency (EMA) Highest-Priority Critically Important Antibiotics (HPCIAs), namely third and fourth generation cephalosoprines, fluoroquinolones and polymixins, polymixins and fluoroquinolones are the most applied, although their sales rates are still far below the most frequently administered antibiotic groups by livestock species 57 . In our samples E. faecium, L. paracasei, L. plantarum, L. brevis, S. aureus and S. suis harbored genes, namely poxtA and lsaE that may confer resistance against multiple antibiotic groups, including tetracyclines. Moreover, poxtA was detected in the proximity of iMGEs in L. paracasei and E. faecium and harbored on a plasmid in L. brevis. In line with our findings, Enterococcus species related www.nature.com/scientificreports/ to cattle were heavily associated with tetracycline resistance by other authors too 58 . At some species poxtA and lsaE were even predicted to co-occur with more than one MGE type. In the genome of L. plantarum poxtA was predicted to be positioned on a plasmid and associated with an iMGE, while S. suis associated lsaE was located on a plasmid attached to a phage. Such genetic features may contribute to the horizontal transfer of ARGs among bacteria which is of outstanding clinical relevance in the case of such a commonly applied antibiotic group in cattle medicine, as tetracyclines. Perfect matches of genes conferring resistance against other clinically significant antibiotic groups, such as macrolides and sulfonamides were also identified in the genome of E. hormaechei, E. faecium, E. coli, L. paracasei, L. plantarum, L. lactis, L. brevis, S. carnosus, S. pseudoxylosus, S. saprophyticus and S. suis. Of these genes, lsaE, sul1, sul2 and poxtA were even predicted to have enhanced mobility due to their association with multiple MGE groups. The only perfect match for an ARG against HPCIAs, qnrS1, that can confirm resistance against fluoroquinolones, has been detected in a Cronobacter spp. The presence of several ARGs presumably associated with iMGEs in the feed of dairy cows harbors the potential to affect the resident microbiota of the animals. As B. subtilis and E. faecium frequently appear in probiotics for cattle 59,60 it is possible that some microorganisms colonize niches in the foregut and proliferate the ARGs they possess. However, even if they cannot reproduce in the ruminal environment, ARGs can still be disseminated through horizontal gene transfer, especially in the presence of antibiotic therapy. Furthermore, ARGs can possibly spread further, to lower gastrointestinal (GI) regions. Fecal microbiota transfer administered to the stomach could restore the microbial population of the colon in human patients 61 , indicating a high volume of viable bacteria reaching the distal regions. Similar results were found in cattle with rumen microbiota transplantation affecting the microbial population of the hindgut 62 . If ARGs spread all around the GI tract, serious animal and public health concerns could be raised. Among enteral diseases, salmonellosis is the major indication of antibiotic therapy in dairy cattle. Enhanced antibiotic resistance of these bacteria could contribute to the economic loss from the disease as many strains already exhibit resistance to several antibiotics 63 . Furthermore, during pathological conditions, like ruminal acidosis, bacteria can translocate to distant locations in the host's body. Interestingly bacterial translocation was even described in the absence of GI diseases in case of specific microorganisms in humans and rodents 64,65 .
As a consequence of the colonization and possible ARG proliferation processes, pathologies caused by phenotypically resistant bacteria can induce animal welfare and economic issues. Animals harboring ARGs in their gut can contaminate their environment with ARGs through fecal matter as well as farm workers who get in direct contact with the animals, even consumers of dairy products can be affected, as farm animal-borne bacteria that harbor potentially mobile ARGs 66-68 might be distributed by products intended for human consumption. For instance, we have previously found ARGs in raw milk samples provided for human consumption 69 . Fecal contamination during milking 70 is a possible way of ARG transfer into raw milk, however other routes are also possible. In humans and in rodents for instance, maternal mononuclear cells transfer microorganisms to milk during lactation 66,67 . The possibility for this phenomenon was described in cattle as well 68 .
ARGs that are transferred to the human body through these routes might decrease the efficacy of antibiotic therapy. In order to gain a deeper insight into the exact role of silage in possible ARG transmission processes, many points still need to be examined and clarified. It would be essential to analyse the colonization success of ARG harboring silage-borne bacteria that enter the body of animals and the extent of ARG transfer of invasive donor bacteria to recipient bacteria living in the gastrointestinal tract. The silage involved in the study is of Medicago origin, and our results are based on data from only two projects. Hence, it would also be necessary to investigate the ARG content of other alfalfa and corn silages.
Antimicrobial resistance is an emerging global threat to public health that both affects agriculture and the healthcare sector. The usage of antibiotics in livestock species exceeds the rate of human applications 71 . Antibiotic use in food animal medicine is also considered a risk as it may provide an indirect transfer route of antibiotic residual 72 ARGs via the food chain 73 . Even though antimicrobials administered for veterinary use, may exert an undesired effect on the food chain, the presence of ARGs in dairy cattle nutrition research is still underrepresented in the literature. According to our results, microbial mass contained in fermented feeds have other medical risks than transmitting contagious diseases, like listeriosis 74 . The bacterial content of these mass feeds, that is either, required for the fermentation processes or collected from various sources of contamination on the farms, could play an essential role in the ARG shift through the food chain.

Materials and methods
Data. We searched appropriate datasets in the National Center for Biotechnology Information (NCBI) Sequence Read Archive (SRA) repository. In December 2021, only two shotgun metagenomic BioProjects (PRJNA495415 75 , PRJNA764355) could have been found that had adequate depth for the de novo assembly that our study is based on. The median read count (interquartile range, IQR) of the samples was 26.5 × 10 6 (3.0 × 10 6 ) and 34.7 × 10 6 (1.5 × 10 6 ) in datasets PRJNA495415 and PRJNA764355, respectively. There is limited metadata available of the samples in the NCBI SRA database (Table 2). Nevertheless, it can be assumed from the metadata that the samples of PRJNA495415 were taken at different fermentation periods. Samples were taken on days 0, 7, 14 and 28 were classified in groups A, B, C and D, respectively. Based on metadata of PRJNA764355 samples, no such stratification was possible, so all samples were classified as group E.
Bioinformatic analysis. Quality based filtering and trimming of the raw short reads was performed with TrimGalore (v.0.6.6, https:// github. com/ Felix Krueg er/ TrimG alore), setting 20 as a quality threshold. Only reads longer than 50 bp were retained and taxonomically classified using Kraken2 (v2.1.1) 76 and a database created (24/03/2021) from the NCBI RefSeq complete archaeal, bacterial, viral and plant genomes. For this taxon assignment the -confidence 0.5 parameter was used to obtain more precise species level hits. The taxon classification data was managed in R 77 using functions of the packages phyloseq 78 and microbiome 79 6 . Following the distance concept of Johansson et al. 6 for each bacterial species, only those with a distance threshold defined within iMGEs and ARGs were considered associated. In the MobileElementFinder database (v1.0.2) for E. hormaechei, the longest composite transposon (cTn) was the Tn3000. In case of this species, its length (11,823 bp) was taken as the cut-off value. For E. faecium, this threshold was the length of the Tn6246 transposon, namely 5,147 bp. As the database neither contains species-level, nor genus-level cTn data for Bacillus, Lactiplantibacillus and Lacticaseibacillus species, a general cutoff value was chosen for the contigs of these species. This value was declared as the median of the longest cTns per species in the database (10,098 bp). The plasmid origin probability of the contigs was estimated by PlasFlow (v.1.1) 84 The phage content of the assembled contigs was prediced by VirSorter2 (v2.2.3) 85 . The findings were filtered for dsDNAphages and ssDNAs. All data management procedures, analyses and plots were performed in R environment (v4.1.0) 77 .

Data availability
The datasets analysed in the current study are available in the National Center for Biotechnology Information (NCBI) Sequence Read Archive (SRA) repository and can be accessed through the PRJNA495415 and PRJNA764355 BioProject identifiers.  www.nature.com/scientificreports/