Introduction

The development of sequencing technologies and bioinformatic tools to analyze massive sequence data has accelerated studies on the antibiotic resistome and suggested that antibiotic resistance genes are ancient and ubiquitous in the natural environment (D'Costa et al., 2007;;D'Costa et al., 2007;; Finley et al., 2013; Martinez et al., 2015). Recent studies have focused on the vast diversity of the resistome and the dissemination mechanisms of antibiotic resistance genes from the environment to the clinic (D'Costa et al., 2007; Perry and Wright, 2013; Martinez et al., 2015). Polymorphisms of antibiotic-inactivating enzymes in the environment can contribute to the diversity of the resistome, as genetic polymorphisms can confer plasticity to those proteins and affect their substrate specificities, leading to changes in the resistance spectra (Cambray and Mazel, 2008; Walkiewicz et al., 2012; Galan et al., 2013), as shown for polymorphisms of β-lactamase from soil bacteria (Demaneche et al., 2008). Although shotgun metagenomics can be the most direct method to decipher the complexity of the environmental resistome, revealing polymorphisms of a specific gene in the environment is not easy to due to the limited sequencing depth. Recent attempts to analyze the ARG abundance in various metagenomic data showed that clinically relevant resistance genes were absent or detected in low abundance (Chen et al., 2013; Hu et al., 2013; Nesme et al., 2014). Targeted gene sequencing by high-throughput sequencing technologies can be a good approach to elucidate such polymorphic diversity of functional genes in depth (Iwai et al., 2010; Schmieder and Edwards, 2012; Lee et al., 2014; Zhou et al., 2015). In this study, aminoglycoside 6'-N-acetyltransferase (AAC(6′)-Ib) was chosen for a targeted gene sequencing analysis due to its clinical importance and suitability for interpreting functionally important polymorphisms (Robicsek et al., 2006; Cambray and Mazel, 2008; Maurice et al., 2008; Vetting et al., 2008; Ramirez et al., 2013). The clinically relevant variant AAC(6′)-Ib-cr, which possesses mutations in two amino acid residues (W102R and D179Y) of AAC(6′)-Ib, is known to confer structural plasticity to the protein required for adaptation to new antibiotics, resulting in the extension of resistance to fluoroquinolone, a different class of antibiotics (Robicsek et al., 2006; Maurice et al., 2008; Vetting et al., 2008). In the present study, we explored the diversity of AAC(6′)-Ib in various ecological niches to demonstrate whether such polymorphic diversity occurring in nature is correlated with function: prominent polymorphic variants of AAC(6′)-Ib were identified and their functional implications were studied to reveal the extended resistance spectra of novel AAC(6′)-Ib variants in the natural environment.

Materials and methods

Sampling and DNA extraction

Information on the sampling sites is provided in (Supplementary Table 1). The geographical location of the sampling sites and the population density of the nearest cities are displayed in Supplementary Figure 1. River samples were obtained from four sites of Han river and two sites of Hancheon river. Water samples were also collected from different types of wastewater treatment plants (urban, university and hospital). Soil samples were taken from six mountain and seven agricultural sites. Information on human fecal samples is given in Supplementary Table 2. For the chicken intestine samples, the luminal contents of the ileum and cecum from six broiler chickens (RSS 308) assessed in a previous study were used (Choi et al., 2014). Environmental DNA from water and soil samples, human fecal DNA, and chicken intestinal content DNA were extracted using FastDNA SPIN Kit for Soil (MP Biomedicals, Solon, OH, USA), QIAamp DNA Stool Mini Kit (Qiagen, Valencia, CA, USA) and UltraClean Fecal DNA kit (MO BIO Laboratory, Carlsbad, CA, USA), respectively. The experimental procedures followed the manufacturer’s instructions.

Targeted gene sequencing

A strategy for targeted gene sequencing of AAC(6′)-Ib protein was described in Supplementary Figure 2. PCR primers and conditions for the amplification of the aac(6′)-Ib gene from environmental DNA followed the method of Park et al. (Park et al., 2006). The barcode primers are provided in Supplementary Table 3. Pfu polymerase (Bioneer Co., Daejeon, Korea) was employed for amplification. The obtained amplicon was sequenced using the 454 GS-FLX+ platform (Roche, Branford, CT, USA) at Chunlab (Seoul, Korea). The MOTHUR software platform (v.1.35.1) (Schloss et al., 2009) was used for allocating raw sequences into different samples according to barcode sequences and removing low-quality and shot-length sequence reads (<350 bp). Chimeric sequences were removed using the UCHIME software package (Edgar et al., 2011). Samples with an excessive number of sequence reads were sub-sampled to 3000 sequences per sample. Singleton sequences from each sample were discarded. All non-singleton sequences were aligned with the MAFFT software (v.7) (Katoh and Standley, 2013) and trimmed to defined start and end positions. The trimmed sequences were used for translation to obtain protein sequences. Sequencing errors in homopolymer regions were manually edited by comparison with the reference protein (DQ303918). Frame-shifted sequence reads caused by insertions or deletions and stop codon-containing sequence reads were discarded. After translation, nucleotide sequences that encode the same amino acid sequence were assigned to a unique protein sequence. The relative abundance of each protein sequence was determined by the number of sequence reads that were assigned to a unique protein sequence.

Sequence analysis

The adequacy of the sequencing depth was verified by monitoring the saturation of the rarefaction curve. The composition, abundance and phylogenetic distance of unique protein sequences at each sampling site were used for the principal coordinates analysis (PCoA) based on the weighted pairwise UniFrac method (Lozupone and Knight, 2005) implemented in the QIIME software (Caporaso et al., 2010). Shannon entropy values (H′) and the relative abundance of polymorphic amino acids at 133 residues in the AAC(6′)-Ib protein were calculated as previously described (Iwai et al., 2010). Statistical tests were performed using analysis of similarity (ANOSIM) (Clarke, 1993) and non-parametric multivariate analysis of variance (Adonis) (Anderson, 2001) implemented in ‘vegan’ R package (Dixon, 2003).

Construction of phylogenetic tree with relative abundance

The software package Beast (v.2.4.0) (Bouckaert et al., 2014) was used to construct a Bayesian maximum clade credibility phylogenetic tree for all unique AAC(6′)-Ib protein sequences. The Beauti software was employed to create an XML file to run Markov Chain Monte Carlo (MCMC) analyses for ten million states in Beast. The Bayesian analysis was performed using the Yule model and a strict clock model with default parameters. Samples of trees and parameters were collected every 1000 steps after discarding a burn-in of 10%. The log file obtained from the run was subjected to the Tracer program (v.1.6.0) to visualize convergence. A maximum clade credibility tree was generated by TreeAnnotator (v.2.4.0) from the output of the MCMC run after the initial trees were discarded (50%) as burn-in. The type information and the relative abundance of branches in the tree were added using Interactive Tree of Life (iTOL v.3) (Letunic and Bork, 2016).

Construction of AAC(6′)-Ib variants for susceptibility testing and protein expression

The major variants used for the functional analyses were generated by site-directed mutagenesis (Stratagene, La Jolla, CA, USA). A gene encoding AAC(6′)-Ib-cr from the plasmid pAC3 (KM204147) of Aeromonas sp. strain C3 was used to produce the variant genes. Variants IbL, RD, RG and WY were constructed with specific primers according to the manufacturer’s instructions (Supplementary Table 3) and the mutated sequences were confirmed by sequencing analysis (Solgent, Daejeon, Korea). To analyze their contribution to resistance in E. coli, all variants were cloned into the pBR322 and pUC18 plasmids under the control of the natural promoter of AAC(6′)-Ib-cr from pAC3 (Supplementary Table 3). Briefly, the promoter region was cloned into the pBR322 plasmid, and the variant genes were then inserted downstream of the promoter to produce pBR322 constructs. For the pUC18 constructs, the variant genes together with the promoter region were amplified with Pro_F and aac6_R2 primers by PCR and inserted into the pUC18 plasmid (Supplementary Table 3). All variant genes were cloned into the pET-28a plasmid (Novagen, Madison, WI, USA) for heterologous expression and purification of the variant proteins (Supplementary Table 3). Protein expression and purification were performed according to the manufacturer’s instructions (Novagen).

N-Acetyltransferase activity assay

For ciprofloxacin, 2 μg of purified protein, 0.5 mm acetyl-CoA, 0.3 mm ciprofloxacin (Sigma-Aldrich, St. Louis, MO, USA) and 1 mm magnesium chloride were mixed in 200 μl of reaction buffer (50 mm Tris-HCl, pH 8.0). For gemifloxacin, 20 μg of purified protein, 0.5 mm acetyl-CoA, 0.2 mm gemifloxacin (Santa Cruz Biotechnology, Santa Cruz, CA, USA), and 1 mm magnesium chloride were mixed in 200 μl of reaction buffer (50 mm Tris-HCl, pH 8.0). The reaction was stopped by the addition of 5% (v/v) of a 2 m hydrochloride solution. The chemical structure of N-acetylated gemifloxacin was determined by LC–MS/MS and proton nuclear magnetic resonance (1H-NMR) analyses. An LTQ Velos mass spectrometer (Thermo Scientific, Waltham, MA, USA) and Accela PDA detector (Thermo Scientific) were used for the identification of N-acetylated fluoroquinolones. The chemical structures of the metabolites were predicted through a comparison of the full-scan and product-ion mass spectra. 1H-NMR analysis was performed using a 600 MHz NMR spectrometer (VNS, Varian Inc., Walnut Creek, CA, USA) and deuterated acetonitrile was used as a solvent. The obtained spectra were analyzed by MestReNova software v.8.1.0 (Mestrelab Research, Spain) according to those from the previous studies (Al-Hadiya and Mahmoud, 2011; Kim et al., 2017). The N-acetylation of fluoroquinolone was analyzed by HPLC. The analysis was performed using an Atlantis dC-18 column (4.6 mm × 250 mm; Waters Corp., Milford, MA, USA) and the Varian ProStar HPLC (Varian Inc., Walnut Creek, CA, USA) system set at 280 nm with a diode-array detector. The mobile phase consisted of a linear gradient from 10% to 95% acetonitrile containing 0.1% formic acid at a flow rate of 1 ml/min. For kanamycin, 0.9-3.5 μg of purified protein, 0.1 mm acetyl-CoA, 0.005% (w/v) kanamycin sulfate (Sigma-Aldrich), and 1 mm 5,5′-dithio-bis-(2,2)-nitroenzoic acid (Sigma-Aldrich) were mixed in 3 ml of reaction buffer (89 mM 4-morpholineethanesulfonic acid buffer, pH 5.7) (Benveniste and Davies, 1971). The increase in A412 was measured. All reactions were performed in duplicate or triplicate and statistical analysis was performed by Student’s t-test. Enzyme units were defined as the amount of enzyme that catalyzed the conversion of 1 nmol of antibiotic substrate to its N-acetylated form per minute at 37 °C.

Antibiotic susceptibility test

The disk-diffusion and micro-broth dilution assays were performed according to protocols based on the CLSI guide (Wiegand et al., 2008). M9 minimal and Mueller Hinton media (Difco Laboratories) were used for the disk-diffusion and broth dilution assays, respectively. pBR322 clones were employed for the broth dilution assay for ciprofloxacin and kanamycin. pUC18 clones were used for the disk-diffusion assay against ciprofloxacin, gemifloxacin and kanamycin and for the broth dilution assay against gemifloxacin.

Nucleotide sequence accession numbers

The nucleotide sequences described in this study have been deposited in the NCBI Sequence Read Archive (www.ncbi.nlm.nih.gov/sra) with Bioproject accession no. PRJNA360018.

Results

Targeted gene sequencing for AAC(6′)-Ib

Forty-one samples were collected from a variety of ecological niches in South Korea, including rivers, wastewater treatment plants, mountain and agricultural soils, chicken intestines and human feces, and subjected to targeted gene sequencing (Supplementary Figure 1; Supplementary Tables 1 and 2). A partial fragment (482 bp) of the gene (600 bp) was amplified and sequenced using the 454 GS-FLX+ platform. The obtained non-singleton nucleotide sequences (68 667 sequence reads) were processed to produce well-curated protein sequences for diversity analyses. The processed sequences (133 amino acids) covered the complete acetyltransferase domain from residues 54 to 186 of the reference protein (DQ303918). A total of 460 unique protein sequences consisting of 44 026 sequence reads were finally obtained and regarded as environmental variants that represented the diversity of AAC(6′)-Ib (Supplementary Table 4).

Ecological niche-specific diversity and abundance of AAC(6′)-Ib

PCoA of weighted UniFrac distances for the AAC(6′)-Ib protein sequence data showed that the AAC(6′)-Ib communities clustered into two distinctive groups according to their ecological niches: (1) soil and intestine and (2) water samples (Figure 1). The water samples were further sub-categorized not by sample source (river and wastewater treatment plants) but instead by geographical location (for example, the metropolitan city Seoul and other rural areas; Figure 1; Supplementary Table 1). The soil samples showed relatively high diversity and the intestine samples were further sub-clustered by their origin (human and poultry). These results suggest that the polymorphic diversity of a single resistance gene can also be determined by ecology, as shown in previous studies on the close correlation between the resistome and ecology (Gibson et al., 2015).

Figure 1
figure 1

Principal coordinates analysis of weighted UniFrac distances based on 460 AAC(6′)-Ib protein sequences in various environmental samples. All samples are displayed as follows: river (blue circle), wastewater treatment plants (green circle), mountain soil (black square), agricultural soil (white square), chicken intestine (purple triangle) and human feces (red triangle). The main and sub-categories with statistically significant differences (Supplementary Table 5) are indicated by solid and dashed lines, respectively.

Full size image

Polymorphisms, abundance and phylogeny of AAC(6′)-Ib variants in various ecological niches

As shown through Shannon entropy analysis of all AAC(6′)-Ib sequences (Figure 2a), most of the amino acid residues were conserved, with the exception of three residues: 102, 117 and 179. When the degree of polymorphism at each AAC(6′)-Ib residue was expressed as a heatmap for the different ecological niches (Figure 2b), these three residues were greatly altered in the soil and intestine samples. Similar patterns of polymorphism in these residues were also shown in the 1395 AAC(6′)-Ib sequences reported in the NCBI GenBank database (Supplementary Figure 3). Remarkably, these residues were the mutation sites in the clinically important variant AAC(6′)-Ib-cr, which was prevalent in the soil and intestinal environments (Figure 2c). Other variants harboring different combinations of amino acid substitutions at these residues were also present in these environments (Figure 2c). Among the 460 unique protein sequences, seven polymorphism-based sequences identified due to polymorphisms at these residues accounted for 84% of the total sequence reads (94.5% in water, 93.9% in soil and 69.2% in intestine; Table 1). Out of 460 unique protein sequences, 444 sequences were assigned to one of seven types, including AAC(6′)-31 type, which was also detected in the amplicon sequencing (Mendes et al., 2007), and the following six polymorphic types of AAC(6′)-Ib: Ib (wild-type AAC(6′)-Ib), IbL (S117L), RD (W102R and S117L), RG (W102R, S117L and D179G), WY (S117L and D179Y) and cr (W102R, S117L and D179Y) (Figure 2c; Table 1). Only 16 unique protein sequences consisting of 0.2% of total sequence reads were not assigned to one of the seven polymorphism-based sequence types and these sequences were designated as ‘others’ in Figures 2c and 3. The composition of the seven major types in each sample was also ecology-specific (Figure 2c), as shown in the PCoA results. The wild-type Ib was dominant in all water samples, whereas the cr type was dominant in the soil and intestine samples. The RD, RG and WY types were mainly detected in the soil and intestine samples, whereas type 31 was found only in the water samples. A Bayesian maximum clade credibility phylogenetic tree of all AAC(6′)-Ib variants showed the diversity and abundance of entire sequences in the different ecological niches (Figure 3). The tree also indicated that 460 unique protein sequences formed distinctive sequence-based lineages congruent with the seven major polymorphism-based sequence types, which represented almost all sequences. Furthermore, PCoA results of weighted Unifrac distances based on seven major types of AAC(6′)-Ib (Supplementary Figure 4) also showed a similar coordination among the different ecosystems to those based on 460 unique protein sequences (Figure 1). The most abundant variant sequences representing each type are shown in the heatmap (arrows in Figure 3), and their relative abundances in the water, soil and intestine samples are presented in Table 1.

Figure 2
figure 2

Polymorphism of AAC(6′)-Ib and the relative abundance of representative types in various ecological niches. (a) Shannon entropy at 133 amino acid residues of AAC(6′)-Ib among the entire sequences obtained from all samples. (b) Heatmap of the degree of polymorphism at 133 amino acid residues of AAC(6′)-Ib in each sample. (c) Major types of AAC(6′)-Ib and their relative abundance in each sample. Others (black) indicate sequences that were not assigned to the seven major types.

Full size image
Table 1 Representative variant sequences of seven major types and their characteristics obtained in this study
Full size table
Figure 3
figure 3

Phylogenetic diversity of the AAC(6′)-Ib variants and their relative abundance in various ecological niches. Bayesian maximum clade credibility phylogeny and abundance profiles of 460 unique protein sequences of AAC(6′)-Ib were presented for 41 environmental samples. Each phylogenetic lineage is assigned to one of seven major types of AAC(6′)-Ib and is colored differently. Others (black) indicate sequences that were not assigned to the seven major types. Representative variant sequences of each type (Table 1) are indicated with arrows.

Full size image

N-Acetylation activity and altered resistance of the AAC(6′)-Ib variants

Polymorphisms of TEM β-lactamase have previously been implicated in causing the extended spectrum of antibiotic resistance (Salverda et al., 2010). A total of 5411 protein sequences obtained from the NCBI GenBank database exhibited variations in critical residues (Supplementary Figure 5), and these variations were closely related to their extended resistance spectra (Salverda et al., 2010). In the case of AAC(6′)-Ib, the cr variant was previously shown to have extended substrate specificity conferring resistance to both kanamycin and ciprofloxacin (Robicsek et al., 2006). Considering that the present study also revealed such polymorphism, it is possible that environmental variants of AAC(6′)-Ib might be related to their functionality. Among the seven major types, the IbL (control), RD, RG, WY and cr variants were selected for functional characterization because these variants are all identical except for the polymorphic residues 102 and 179. Corresponding recombinant proteins constructed by site-directed mutagenesis were purified, and their N-acetylation activity against ciprofloxacin was analyzed. The cr variant displayed the highest N-acetylation activity against ciprofloxacin (Figure 4a), and other variants such as RD, RG and WY also showed activities higher than IbL, which is known to exhibit a basal level of activity (Robicsek et al., 2006). The ciprofloxacin susceptibility of E. coli cells harboring these variant genes was tested through the disk-diffusion assay and the broth dilution method. As previously reported (Robicsek et al., 2006), the cr variant showed prominent resistance to ciprofloxacin compared with IbL, which was tested as the wild type (Figure 4a). Additionally, the RD, RG and WY variants displayed increased resistance to ciprofloxacin, although this resistance was low compared with that of the cr variant (Figure 4a), suggesting that the increased enzyme activity of these variants conferred the resistant phenotype (Table 1). The D179Y mutation in the WY variant was proposed to be important for fluoroquinolone N-acetylation activity, whereas the W102R mutation in the RD variant was not considered critical (Maurice et al., 2008; Vetting et al., 2008). Interestingly, when the RD and WY variants acquired ciprofloxacin resistance, they displayed less reduction in kanamycin N-acetylation activity and resistance than the cr variant, which exhibited a significant reduction as a trade-off (Figure 4b).

Figure 4
figure 4

N-Acetylation activities of the AAC(6′)-Ib variants and the related resistance against (a) ciprofloxacin, (b) kanamycin and (c) gemifloxacin. The N-acetylation activities of the purified variant enzymes are indicated with white bars. Statistical results (t-test) were provided in Supplementary Table 6. The altered resistance of E. coli DH5α harboring the variant genes is shown based on minimal inhibitory concentration values (black bars) and images of the disk-diffusion assay results.

Full size image

To further evaluate the extended resistance spectra of the AAC(6′)-Ib variants, various antibiotics containing primary or secondary amine moieties were also tested for N-acetylation activity. The 4th-generation fluoroquinolone gemifloxacin, against which the cr variant was previously known not to be resistant (Robicsek et al., 2006), was found to be N-acetylated by the WY variant (Figure 4c; Supplementary Tables 7 and 8). Although the gemifloxacin N-acetylation activity of the WY variant was much lower than the ciprofloxacin N-acetylation activity of the cr variant, the activity was sufficient to confer increased resistance to gemifloxacin compared with the other variants (Figure 4c; Table 1). Inactivation of gemifloxacin through N-acetylation was demonstrated using the disk-diffusion assay (Figure 5). Other fluoroquinolones that were recently developed, including tosufloxacin, clinafloxacin, moxifloxacin and zabofloxacin, were also tested with the WY variant; only zabofloxacin was proven to be a substrate of the WY variant (data not shown), suggesting that the aminoalkyl and methoxyimino groups at the pryrrolidine ring are important for binding at the catalytic site. Homology modeling based on protein-ligand binding suggested that the D179Y mutation was critical for recruiting and stacking the quinolone ring into the binding pocket via π-interactions as previously demonstrated (Maurice et al., 2008; Vetting et al., 2008), but the W102R mutation in the cr variant could result in a potential steric clash with the methoxy group of gemifloxacin (data not shown).

Figure 5
figure 5

N-Acetylation of gemifloxacin by the WY variant resulting in the loss of antimicrobial activity. The antimicrobial activities of gemifloxacin and N-acetylgemifloxacin were tested via disk-diffusion assays using authentic gemifloxacin and a reaction mixture of gemifloxacin and the WY variant enzyme against E. coli DH5α. The complete conversion of gemifloxacin was confirmed by HPLC analysis.

Full size image

Discussion

The use of targeted gene sequencing approach has been emphasized to reveal the vast diversity of antibiotic resistance genes in the environment (Schmieder and Edwards, 2012). The results of our study showed that the polymorphic diversity of AAC(6′)-Ib was a natural phenomenon in the environment and that certain dominant polymorphisms were ecological niche-specific. The prevalence of the cr type as a dominant polymorphism in soil where the antibiotic selection pressure was absent suggested that the clinically relevant variant was a naturally-occurring polymorphic form. Additionally, the soil and intestine samples showed close relationships in their polymorphic diversity and the compositions of their major types (Figures 2c and 3). These results coincide with previous findings regarding interconnections between the resistomes of the human gastrointestinal tract, soil and clinical pathogens (Benveniste and Davies, 1973; Forsberg et al., 2012; Woolhouse et al., 2015). Comparative genomics analysis also revealed an inter-link of antibiotic resistance between humans and livestock (Mather et al., 2013; Ward et al., 2014). Therefore, the soil, animal and human microbiomes are considered major reservoirs of antibiotic resistance (D'Costa et al., 2007; O'Toole, 2014; Woolhouse et al., 2015). The cr variant was also found in hospital wastewater (SW) and Han river samples (SR) downstream of metropolitan Seoul but not in other water samples from much less populated regions (Figures 1 and 2c; Supplementary Figure 1). These results suggest that anthropogenic activity may influence the emergence of certain antibiotic resistance genes. In previous studies, it was proposed that steady increase in the clinical use of ciprofloxacin during the 1990s has generated selection pressure for the cr variant (Park et al., 2006; Robicsek et al., 2006; Jacoby et al., 2009). Moreover, several reports have shown that human activity is closely associated with the antibiotic resistome (Perry and Wright, 2014; Stalder et al., 2014; Li et al., 2015).

While the cr variant was previously reported to inactive ciprofloxacin and aminoglycosides (Robicsek et al., 2006), other types of polymorphic variants were functionally characterized in this study. A novel N-acetylation activity of the WY variant against gemifloxacin was discovered and the resulting resistance was demonstrated. According to previous reports (Baquero, 2001; Andersson and Hughes, 2014; Martinez et al., 2015), this low-level resistance in the environment should not be overlooked because it could be ‘a gateway to clinical resistance’, as previously shown for the cr variant (Robicsek et al., 2006; Hawkey and Jones, 2009). Furthermore, the altered resistance spectra of environmental variants may confer advantages under certain environmental conditions in various ecological niches. Indeed, we observed different enzyme activities of AAC(6′)-Ib variants under different pH levels and magnesium ion concentrations (Supplementary Figure 6).

Since these variants were discovered only in metagenomic sequences, attempts to isolate environmental bacteria harboring these variant sequences were conducted using four mountain soil samples tested for targeted gene sequencing. As a result, the cr variant was found to be a dominant type in soil isolates harboring the aac(6′)-Ib gene; other major variants, such as RD, RG and WY, were also detected (data not shown). Resistance to such a novel antibiotic in environmental bacteria may lead to the emergence of novel resistance in the clinic. Indeed, some major variants, such as the RD and WY variants, have recently been reported in clinical and animal isolates and have been found to be associated with mobile genetic elements (Supplementary Table 9; Moura et al., 2012; Deng et al., 2014; Vredenburg et al., 2014; Vaz-Moreira et al., 2016), suggesting that they may have already been disseminated to the clinic as a potential source of resistance. Since 460 variant sequences identified in this study were mostly not reported in the database, other novel polymorphic variants related to resistance could still be present. For example, the substitution G180S in the cr variant was prevalent in the intestine samples (8.4% of the total sequence reads of the intestine samples; Figures 2b and 3).

In addition to the previous discovery in polymorphic mutations conferring resistance to two different classes of antibiotics in the clinical isolate of the cr variant (Robicsek et al., 2006), our study demonstrated the altered resistance spectra of other polymorphic variants, including the novel gemifloxacin resistance of the WY variant. More extensive analysis of other environmental variants to determine their functionality will lead to the identification of a wider spectrum of multi-drug resistance and novel resistance mechanisms. In conclusion, our results suggested that polymorphisms occurring in nature could alter the substrate specificity of antibiotic-inactivating enzymes and therefore determine resistance spectra, ultimately expanding the antibiotic resistome throughout the environment.