Formula-fed and Clostridium difficile infected neonatal piglets carry a highly diverse and abundant set of antimicrobial resistance genes CURRENT STATUS:

Background: Clostridium difficile infection (CDI) is an increasing zoonotic health threat and has also been documented as a cause of enteritis outbreaks in neonatal pigs. Furthermore, CDI in neonatal piglets cause changes in microbial gut colonization. We hypothesized that an imbalanced microbial colonization in piglets with CDI could be associated with an altered abundance of antimicrobial resistance genes. Results: We analyzed fecal metagenomic data of lactating sows (S), their piglets during suckling (SP), the same piglets two weeks after weaning (WP), 5-day old artificially reared and formula-fed siblings (FP) and FP infected with C. difficile (FP-CD) for microbiota composition and antimicrobial resistance gene abundance. FP and FP-CD piglets had an immature-type microbiota and increased abundance of antimicrobial resistance genes. A co-occurrence of genes encoding for resistance against aminoglycosides (e.g. aph(3”)-lb , aph(6)-ld , ant(2”)-la ), β-lactams ( bla CTX-M , bla TEM ), fluoroquinolones ( pat (A) macrolides ( mph (A)), sulfonamides ( sul1, sul2 ), polypeptides (e.g. pmrB, pmrC, arnA, bac (A)) and tetracyclines (e.g. tet (A-D),) was observed. Conclusion: Increased abundance of antimicrobial resistance genes in formula feeding and concomitant CDI may be associated with therapeutic resistance later in life and warrant further studies.

also promote the re-assortment of plasmid-encoded genes (including AR genes) within these microbial communities [5].
In humans, Clostridium difficile (CD) infection (CDI) is a leading cause of nosocomial and community acquired infection, morbidity and fatality. Interestingly, shifts in the intestinal microbiome were recently linked with resistome changes in patients suffering from CDI [6]. In pigs, spontaneous outbreaks of CDI have been increasingly observed in neonatal piglets at < 14 days of age [7].
Moreover, hypervirulent ribotypes can be found in pigs and human patients alike [8]. Although the predisposing factors for CDI in neonatal pigs are largely unknown, diet may play an important role [9].
Because hyperprolifacy in modern pig breeds has increased the average numbers of life-born piglets per litter (often exceeding the nursing capacity of the mother), artificial fostering systems with formula feeding are more and more used in the pig industry, which may increase the prevalence of CDI. However, little is yet known about the consequences of formula feeding and CDI on the intestinal resistome in neonatal piglets.

Results
In order to generate intestinal microbiome data of genetically closely related individuals (full-and half-sib piglets), purebred sows ('German Landrace') and their suckling and weaned piglets were used in the study. Fecal samples were taken from four sows (S) and a pooled sample from their suckling A total of 83 AR genes were identified within 11 antibiotic classes. The AR gene abundance was lower in S, SP, FP and WP groups as compared to FP and FP-CD piglets ( Fig. 2A). Relative abundance of AR genes revealed that all animal groups clustered according to AR gene abundance (Fig. 2B). To determine the association of the 35 most abundant AR genes with the abundance with bacterial genera in the groups, we employed a redundancy analysis (Fig. 2C). The abundance of AR genes related to tetracycline (tet(Q), tet(W)), lincosamide (lnu(C), lsa(E)), and aminoglycoside resistance (ant(6)-Ib) in S, SP and WP metagenomes coincided with the abundance of the bacterial genera such as Megasphaera, Bifidobacterium, Lactobacillus, Clostridium, and Prevotella. RDA analysis further confirmed the strong association between the genera Escherichia, Streptococcus, Shigella, Enterococcus and Fusobacterium with the expansion of the resistome in FP and FP-CD piglets. Cooccurrence of genes encoding for resistance against aminoglycosides (e.g. aph(3")-lb, aph(6)-ld, ant(2")-la), β-lactams (bla CTX−M , bla TEM ), fluoroquinolones (pat(A) macrolides (mph(A)), sulfonamides (sul1), polypeptides (e.g. pmrB, pmrC, arnA, bac(A)) and tetracyclines (e.g. tet(B), tet(D)) was observed.

Discussion
The complex intestinal microbial ecosystem is characterized by a close microbe-microbe-host relationship. However, this homeostasis can be shifted into a state of dysbiosis by e.g. antimicrobial agents, diet, immune deficiency or infection, and thereby possibly facilitate the expansion of the resistome [5]. Despite the well-studied and highly diverse microbial ecosystem in older individuals, the colonization of the neonatal gut in both humans and pigs is a de novo establishment of a microbial ecosystem undergoing several consecutive steps under the influence of genetic, maternal, and environmental (dietary and medical) factors [10,11,12]. In pigs, the very early colonizers between birth and 2 days of age are mainly enterobacteria, streptococci and clostridia [10,11]. This is supported by our findings here, where Escherichia, Streptococcus, Shigella, Enterococcus, Collinsella, Fusobacterium and Ruminococcus were the dominating genera in FP and FP-CD piglets.
Under normal conditions, these genera become later outcompeted by others such as Alistipes, Bacteroides, Fusobacterium and Lactobacillus [10]. Because FP and FP-CD piglets were transferred into the artificial rearing units within 6 h after birth and had no further contact to the mothers surroundings, it seems likely that these piglets haboured an 'immature-like' microbiome as compared with SP and WP piglets, accompanied by a further expansion of these early colonizers through the formula diet. For example, formula with high lactose content promoted the proliferation of enterobacteria in neonatal piglets [13]. We have previously shown that formula feeding also promotes CDI in neonatal piglets [9]. Moreover, formula composition promoted the expansion of naturally colonizing CD in FP piglets in the present study. This is in line with studies in formula-fed infants showing a higher intestinal colonization with CD and E. coli as compared to breast-fed infants [14,15]. It seems likely that maternal antibodies in the milk may provide protection against CDI in the suckling offspring. For example, we have recently shown that sow milk and serum contains antibodies against CD toxin A [16]. A formula-induced altered or delayed intestinal colonization and low abundance of lactobacilli or clostridia may decrease colonization resistance thereby promoting CD expansion [17,18,19]. The expansion of Escherichia, Streptococcus, Shigella, Enterococcus phylotypes through the formula diet is of particular relevance because they represent a presumed source of persisting multi-drug resistant pathogens through further forward selection under medical treatment or other stress factors in the gut environment later in life. In pigs, for example, reports have linked the use of supra-nutritional levels of trace elements such as Cu and Zn in swine diets with an increased prevalence of multi-resistant E. coli strains [20]. A co-selection for AR genes within the intestinal microbial communities during or after an initial state of dysbiosis could even further promote an expansion of the bacterial resistome.
In the present study, a total of 83 AR genes were identified, conferring resistance to members of 11 classes of antimicrobial agents. A higher abundance of these genes was determined in FP as well as FP-CD piglets as compared to the other groups. However, the abundance of certain AR genes such as tet(Q), tet(W), lnu(C), lsa(E), or ant(6)-Ib in S, SP and WP groups was apparently linked to Megasphaera, Bifidobacterium, Lactobacillus, or Prevotella thereby confirming these genera as common carriers of these AR genes [21]. Moreover, the observed co-occurrence of some AR genes among all groups could be attributed to microbial community composition and co-occurrence of certain bacterial taxa in general but also to gene co-localization in bacterial genomes. Plasmid colocation of AR genes such as bla CTX−M genes and mph(A) was previously reported in extendedspectrum β-lactamase-producing E. coli isolates from farm animals (including pigs) in Germany [22].A recent study showed clusters of co-occurring AR genes in pig environments including such as tet(W), tet(Q), tet(B), mefA, bla TEM [23]. The present study shows that such co-occurrences are rather dynamic and that the abundance of AR genes largely depends on the microbiome composition. As shown, the latter is influenced by age (e.g. SP vs. WP vs. S), diet and rearing environment (e.g. SP vs. FP), and CDI (FP vs. FP-CD). However, our study also supports the hypothesis that metagenomic approaches may help to identify and monitor antibiotic resistance patterns in gut microbial communities in the pig [24].

Conclusion
Taken together, we show that artificial rearing and formula feeding alone or combined with CDI and an immature-like microbiome promote a highly abundant bacterial resistome in neonatal piglets.
Following the "One Health" concept, the relevance for therapeutic resistance or the role of formulafed neonatal pigs as potential AR gene shedders later in life needs to be further elucidated.

Materials And Methods Animals, experimental setup and sampling
The aim of the present study was to determine a possible link between the intestinal microbiome development and the resistome in neonatal piglets fed formula and infected with CD, as compared to their mothers, suckling littermates, and weaned littermates later in life.
Purebred sows ('German Landrace') and their suckling and weaned piglets were used in the study.
Sows received a standard lactation diet and were not treated with antibiotics before or during the experiments. During the suckling period, sow milk was the only nutrient source and piglets had no access to creep feed. Immediately after birth, a total of six piglets were randomly selected and transferred into artificial rearing units [9,13]. Three of these piglets were fed formula (formula piglets, FP), whereas three animals were fed formula and infected at 6 h after birth with 10 8 CFU Clostridium difficile ribotype 078 (FP-CD) spores. At 28 days of age, the remaining suckling piglets were weaned and fed a standard cereal-soy-based weaning diet without antibiotics or special feed additives (e.g. organic acids or probiotics). All diets met the nutrient requirements according to current recommendations. For microbiome analysis, fecal samples were taken from four sows (S) and a pooled sample from their suckling piglets (n = 4) at day 5 (suckling piglets; SP) and two weeks after the weaning at day 42 (n = 8 weaned piglets, WP). Fecal samples from FP (n = 3) and FP-CD (n = 3) animals were taken on day 5.

DNA extraction and sequencing
Total genomic DNA was extracted using the QIAamp Fast DNA Stool Mini Kit (Qiagen, Hilden, Germany) with bead-beating to improve DNA recovery from spore-forming bacteria [25]. After DNA quality control using the Genomic DNA Analysis ScreenTape on a 2200 TapeStation Instrument

Data analyses
The FLEXBAR tool was used to remove adapter sequences from the metagenomic reads and discard reads with length shorter than 100 nucleotides [26]. We mapped quality checked sequences (89% of the original sequencing reads) against reference genomes of 13193 different bacterial and archaeal species taken from NCBI (ftp://ftp.ncbi.nlm.nih.gov/genomes/genbank/). We used the yara ("yet another read aligner") tool to perform read mapping [27]. The resulting mapping files were further analyzed using the Species Level Identification of Microorganisms from Metagenomes (SLIMM) tool to produce for taxonomic profiles of each sample [28]. We calculated relative abundances of individual taxa using the SLIMM tool. SLIMM defines relative abundance as the number of reads assigned uniquely to a taxon divided by the total number of mapped reads then corrected for genome sequence length. We used a relative abundance cut-off of 0.01% to consider a given taxon of bacteria present.
For resistome analysis, we mapped metagenomic reads against version 1.1.2 of the Comprehensive Antibiotic Resistance Database (CARD; http://arpcard.mcmaster.ca) using yara read mapper. Then we used the number of reads mapped to each resistance gene per 1000 bp as an indicator for the relative sequence abundance for each gene. Only genes coding for a true resistance were used for further analysis, whereas genes related to multidrug resistance, transcription factors or regulating factors were deleted from the initial list. Within AR genes coding for β-lactam resistance, 'bla CTX−Mtype' and 'bla TEM -like' genes were summarized as one group, whereas the genes related to vancomycin (glycopeptide) resistance were summarized under the term 'vanB gene cluster'. Using the relative abundance of AR genes (normalized by total metagenomic sequences per sample), a heatmap was calculated using the 'heatmap2' package in R. The hierarchical clustering was

Consent for publication
Not applicable.

Availability of data and materials
The metagenomic sequences generated during and analyzed during the current study have been deposited in the European Nucleotide Archive (ENA) under accession code PRJEB19642 (https://www.ebi.ac.uk/metagenomics/projects/ERP021695).

Competing interests
The authors declare that they have no competing interests.