Animal experiments were approved by the Animal Protocol Review Committee of the Institute of Subtropical Agriculture, Chinese Academy of Sciences (Changsha, China), and all experiments were performed according to the institutional animal welfare requirements. Forty-eight healthy weaned piglets, aged 21 ± 1 days (8.26 ± 0.17 kg), were randomly assigned to four groups, with six barrows and six females in each group. The piglets were housed individually and given free access to drinking water and food. The control group was fed a basal diet (Table 1), and the treatment groups were fed a basal diet supplemented with 20, 100, or 200 mg Cu/kg feed, as CuSO₄⋅5H₂O, for 6 weeks. Anal swabs were collected from piglets at days 0, 21, and 42 of the experimental period, for E. coli isolation and antimicrobial drug resistance detection. At days 21 and 42, piglets (n = 6) in the control and 200 mg Cu group were anesthetized with Zoletil (15 mg/kg BW, i.m.) and euthanized. The ileal and cecal contents were collected under sterile conditions, using a 15-ml sterile conical centrifuge tube and submitted for 16S rRNA sequencing, E. coli isolation, and antimicrobial drug resistance evaluation.

The ileal and cecal contents from piglets in the control (without Cu supplementation) group and the group receiving a pharmacological dose of Cu (200 mg Cu/kg feed, n = 6 per treatment) were collected for microbiota analysis, as described previously (Zhou et al., 2018; Xiong et al., 2019). Briefly, total bacterial DNA was extracted using a QIAamp DNA Stool Mini Kit (Qiagen, Hilden, Germany), and the V3–V4 hypervariable region of the microbial 16S rRNA genes were sequenced using the primers, 341F: 5′-CCTAYGGGRBGCASCAG-3′ and 806R: 5′-GGACTACNNGGGTATCTAAT-3′; Illumina adaptors; and molecular barcodes. Libraries were prepared using a TruSeq DNA PCR-Free Sample Preparation Kit (Illumina Inc., San Diego, CA, United States) and were assessed using a Qubit 2.0 Fluorometer (Thermo Scientific, Madison, WI, United States) and a Bioanalyzer 2100 system (Agilent Technologies, Santa Clara, CA, United States). Libraries were sequenced on a HiSeq 2500 platform (Illumina Inc., San Diego, CA, United States). Raw 16S rDNA sequences were assembled using the QIIME (v1.9.0) and FLASH software packages. Operational taxonomic units (OTUs) were analyzed using UPARSE (v7.0.1001), and high-quality sequences were aligned against the SILVA reference database 1 and clustered into OTUs at a 97% similarity level using the UCLUST algorithm 2. Each OTU was assigned to a taxonomic level using the Ribosomal Database Project Classifier program v2.203. The assembled sequences were then submitted to the NCBI Sequence Read Archive (No. PRJNA551517) for open access.

Samples were collected with sterile swabs, and the swab sample was immediately inserted into a 15-ml sterile centrifuge tube containing 5 ml of phosphate-buffered saline. Then, 10 μl of each sample was swabbed onto MacConkey agar and incubated at 37°C for 18–24 h. Three red- or pink-colored colonies were picked per sample and inoculated on eosin methylene blue agar plates at 37°C for 18–24 h to obtain pure cultures. A single colony with a black metallic luster was picked and cultured in Luria–Bertani (LB) liquid medium at 37°C for 8–12 h. Bacterial smears were from the LB liquid culture, and Gram staining and microscopic examination were performed as described (Feyzioglu et al., 2014; Usaini et al., 2015). LB liquid cultures containing Gram-negative bacilli were swabbed onto eosin methylene blue agar plates and incubated at 37°C for 18–24 h. A single colony with a black metallic luster was picked, inoculated on a nutrient agar plate, and cultured at 37°C for 12–18 h. Single colonies from the nutrient agar plate were inoculated in biochemical identification tubes (Hangzhou Microbial Reagent Co., Ltd., Hangzhou, China) for lactose fermentation, and indole and citrate utilization tests, according to the manufacturer’s instructions. Isolates with a Lac+/Ind+/Cit− phenotype were identified as E. coli. Finally, the 16S rRNA gene of isolated E. coli was amplified by PCR. DNA extraction and PCR amplification were performed as previously described (Blyton et al., 2013). Three oligonucleotide primers, 16E1, 16E2, and 16E3, were designed and used in the PCR assays for the identification of E. coli isolates, as previously described (Tsen et al., 1998). The sequences and locations of these primers are shown in Supplementary Table S1. PCR products were analyzed by gel electrophoresis through a 2% agarose gel, as previously described (Tsen et al., 1998) and sequenced by Sangon Biotech Co., Ltd. (Shanghai, China). Homology analysis between the amplified sequences and the E. coli 16S rRNA gene sequences published in GenBank was performed to confirming that the isolates were E. coli. Repetitive element PCR with both ERIC and (CGG)4 primers was also used to discriminate clonal isolates from the same samples (Adamus-Bialek et al., 2009). Isolates producing the same banding pattern were considered to be the same strain.

After PCR validation, the E. coli suspensions were subjected to antimicrobial susceptibility testing on Müller–Hinton agar using the Kirby–Bauer disk diffusion method, in accordance with the Clinical and Laboratory Standards Institute guidelines, using E. coli ATCC 25922 as an internal quality control. The antimicrobial disks used for the test included ampicillin (10 μg/disk), amoxicillin/clavulanic acid (20/10 μg/disk), ceftriaxone (30 μg/disk), meropenem (10 μg/disk), aztreonam (30 μg/disk), gentamicin (10 μg/disk), trimethoprim/sulfamethoxazole (1.25/23.75 μg/disk), amikacin (30 μg/disk), and ciprofloxacin (CIP) (5 μg/disk) (Oxoid Ltd., Hampshire, United Kingdom).

All statistical analyses were performed using SPSS 20.0 software (SPSS Inc., Chicago, IL, United States). α and β diversity were analyzed using QIIME (v1.7.0). Data for microbial metabolic functions were analyzed by Student’s t-test only, due to minor changes in the microbiome data. The differences in isolation rates and antimicrobial resistance of E. coli among the groups were analyzed by Chi-square test. A P < 0.05 was considered statistically significant.

The average bodyweights of the weaned piglets at 21 ± 1 days in the control and Cu-supplemented groups were 8.13 ± 0.22 and 8.40 ± 0.26 kg, respectively. There were no significant differences in average daily gain, average daily feed intake, or feed conversion ratio between the control and Cu-supplemented groups in early (1–3 weeks) and late postweaning stages (4–6 weeks, Table 2).

There were no significant differences between treatments for the indices of α diversity, including observed_species, Shannon, Simpson, Chao1, ACE, and PD_whole_tree indices, of ileal (Supplementary Table S2) and cecal microbiota (Supplementary Table S3). In addition, the weighted principal coordinate analysis plots of ileal (Figure 1A) and cecal microbiota (Figure 1C) verified that there were few differences in microbial communities between Cu treatments.

Based on an unweighted pair-group method with arithmetic mean analysis, Firmicutes, Proteobacteria, Tenericutes, and Bacteroidetes were found to be the dominant bacteria in the ileal microbiome (Figure 1B), while Firmicutes, Bacteroidetes, Cyanobacteria, Proteobacteria, and Spirochaetes were the dominant bacteria in the cecal microbiome (Figure 1D). Furthermore, the linear discriminant analysis effect size method was used to determine the variation in microbial composition between groups. Minimal changes were observed after short-term feeding (3 weeks), with differences in abundance of the genera Sarcina, Pasteurella, Veillonella, Streptococcus, Mycobacterium, Lactococcus, and Sphaerochaeta in the ileal microbiome (Figure 2A) and in the family, Acetobacteraceae, and the genera, Veillonella, Pseudoflavonifractor, Defluviitaleaceae, Oribacterium, Ruminococcaceae (UCG008, 014), Lachnospiraceae (UCG003), and Phascolarctobacterium in the cecal microbiome (Figure 2B). Meanwhile, after a longer time of feeding (6 weeks) with Cu supplementation at a pharmacological dose, marked differences were found in the abundance of the family, Clostridiaceae, the phylum Firmicutes, the genus, Sarcina, Lactobacillus paralimentarius, and swine manure bacterium in the ileal microbiome (Figure 2C), and the genera, Lactobacillus and Sporolactobacillus and E. coli in the cecal microbiome (Figure 2D). A heat map of the 40 most different genera showed the taxonomic distributions among groups after 3 and 6 weeks of treatment (Supplementary Figure S1). Next, the impact of Cu treatment on aerobic, anaerobic, mobile element-containing, facultatively anaerobic, biofilm-forming, Gram-negative, Gram-positive, potentially pathogenic, and stress-tolerant bacteria were predicted via BugBase (Zhou et al., 2019). No significant differences were found to occur as a result of Cu treatment (Supplementary Figure S3).

Subsequently, to further investigate the changes in the function of the gut microbiota, the metagenome was generated, based on 16S rDNA sequencing results, using Picrust. The Kyoto Encyclopedia of Genes and Genomes annotation results were then submitted for principal component analysis. Microbial metabolism-related pathways at Kyoto Encyclopedia of Genes and Genomes level 3 were specifically filtered. No obvious segregation of ileal and cecal microbiomes was found in the third week (Supplementary Figures S2A,B). Furthermore, a t-test showed that Cu supplementation (200 mg Cu/kg feed) increased the abundance of microbiota related to the transcriptional machinery and tuberculosis in the ileal microbiome, while an increase in the abundance of microbiota related to methane metabolism was observed in the cecal microbiome at the third week (Supplementary Figures S2C,D). An obvious segregation was found in the ileal microbiome in the sixth week (Figure 3A), but there was still no obvious segregation in the cecal microbiome at this time point (Figure 3B). Cu supplementation decreased the abundance of microbiota related to energy metabolism, lysine biosynthesis, amino acid metabolism (alanine, aspartate, glutamate, histidine, glycine, serine, and threonine), butanoate metabolism, carbon-fixation pathways in prokaryotes, terpenoid backbone biosynthesis, nitrogen metabolism, benzoate degradation, and the metabolism of cofactors and vitamins in the ileal microbiome (Figure 3C). Meanwhile, Cu supplementation increased the abundance of microbiota related to amino acid biosynthesis (valine, leucine, isoleucine, and lysine), lipid biosynthesis protein, amino acid metabolism (glycine, serine, and threonine), and C5-branched dibasic acid metabolism and decreased the abundance of microbiota related to peptidases in the cecal microbiome (Figure 3D).

Based on microbial composition analysis, we further determined the antimicrobial resistance profiles of E. coil isolated from the ileal and cecal contents. Isolates that were resistant to three or more classes of antimicrobials were defined as multidrug-resistant strains. A total of 119 E. coli isolates from ileal (n = 50) and cecal contents (n = 69) were included in disk diffusion testing, with 20 ileal isolates and 34 cecal isolates from control pigs and 30 ileal isolates and 35 cecal isolates from pigs fed a pharmacological dose of Cu (Supplementary Table S4).

Of the 119 isolates, 10 (8.40%) were not resistant to any of the antimicrobial agents tested, 9 (7.56%) were resistant to one antimicrobial agent, and 21 (17.65%) were resistant to two antimicrobial agents. Furthermore, 79 (66.39%) were resistant to at least three antimicrobial agents, and 47 (39.50%) of these were isolated from pigs fed a pharmacological dose of Cu. In particular, seven isolates (5.88%) were resistant to more than six classes of antimicrobials, and six (5.04%) of these were isolated from pigs fed a pharmacological dose of Cu (Table 3). The antimicrobial resistance rates of E. coli in the ileum and cecum are summarized in Table 4. Resistance to CIP in ileal isolates and resistance to chloramphenicol (C) in cecal isolates were found to significantly respond to Cu dose.

On the basis of these results, anal swabs were collected from four groups of pigs and used for E. coli isolation to investigate the development of resistance in response to Cu dose and feeding period. No significant differences were found in the rate of E. coli response to Cu dosage or feeding period (Supplementary Table S5). However, a dose-dependent effect on resistance to C and CIP was found at day 42 (P < 0.05, Table 5).