Tetracyclines are a group of broad spectrum antibiotics that have recently faced a reduction in clinical use due to the increase in prevalence of tetracycline resistance (Chopra and Roberts, 2001; Bishburg and Bishburg, 2009). In the UK tetracyclines are the most sold antibiotic for animal use and represent 10% of prescribed antibiotics for human clinical use. This widespread use of tetracyclines exerts a selection pressure on microorganisms to maintain tetracycline resistance genes (Roberts, 2003; Martinez, 2009; Wu et al., 2010). Resistance is mainly attributed to the production of efflux pumps, ribosomal protection proteins (RPPs) that prevent tetracycline binding to the ribosome, and less often, tetracycline degrading enzymes (Aminov et al., 2002; Connell et al., 2003; Yang et al., 2004; Forsberg et al., 2015).

Although resistance to this group of antibiotics is prevalent they are still used in the treatment of some human infections, including Chlamydia infections and some eye infections such as trachoma (Hu et al., 2010; Dukers-Muijrers et al., 2015). Tigecycline is a novel semi-synthetic derivative of tetracycline and the first of the glycylcyclines. It contains a bulky N,N-dimethylglycylamido side group that allows it to overcome RPP and efflux mechanisms of resistance to earlier generation precursors such as tetracycline (Someya et al., 1995; Olson et al., 2006). Tigecycline is used in the treatment of skin and abdominal infections as well as some cases of community acquired pneumonia (Rubinstein and Vaughan, 2005; Shen et al., 2015; Van Berkel et al., 2016). It has been shown that tetracycline resistance genes can obtain mutations that broaden the activity of their products to new tetracycline derivatives (Linkevicius et al., 2015). It is important that we understand the mechanisms of resistance to our current generation of tetracyclines in order for us to identify environments that may harbor genes that could confer resistance to novel tetracyclines, including those that are still in development.

The microbiota of the human oral cavity constitutes a reservoir of tetracycline resistance genes. RPP genes such as tet(M) are the most abundant tetracycline resistant genes in bacteria of the oral cavity (Villedieu et al., 2003; Seville et al., 2009). Tetracycline efflux genes such as major facilitator superfamily (MFS) exporters including tet(L) and the ATP Binding Cassette (ABC) transporter tetAB(46) have also been detected in bacteria in the oral cavity (Lancaster et al., 2005; Seville et al., 2009; Warburton et al., 2013).

ABC transporters are a functionally and structurally diverse family of proteins. They may comprise a single peptide, four peptides or two half transporters. Bacterial half ABC transporters typically contain a transmembrane domain (TMD) and a highly conserved nucleotide binding domain (NBD). Functional dimeric ABC transporters are composed of two half ABC transporter subunits each contributing a TMD spanning the membrane to form a substrate channel (Biemans-Oldehinkel et al., 2006; Dawson and Locher, 2006). The two NBDs interact to form the ABC of the transporter at the cytoplasmic face of the membrane that can bind two ATP molecules (Jones and George, 1999). It is here that ATP binding and hydrolysis triggers conformational changes in the substrate channel (Hollenstein et al., 2007). Cycles of ATP binding and hydrolysis allow the substrate channel to alternate between being open to the cytoplasm for substrate binding and open to the cells external environment or periplasm for substrate efflux (Higgins, 2001; Hellmich et al., 2012). ABC transporters that confer multidrug resistance (MDR) and biocide resistance to bacteria including human pathogens have been described such as YheH/I, LmrCD, PatAB and EfrAB from Bacillus subtilis, Lactococcus lactis, Streptococcus pneumoniae and Enterococcus faecalis, respectively (Lee et al., 2003; Lubelski et al., 2004; Torres et al., 2009; Baylay et al., 2015).

The aim of this study was to identify the gene(s) conferring resistance in a tetracycline and tigecycline resistant clone that was identified from a human saliva metagenomic library in Escherichia coli. Two genes from the clone, tetA(60) and tetB(60) were found to encode for two half transporter proteins which were shown to be responsible for the observed antibiotic resistance phenotype and reduced fitness.

A library of 27,000 clones was constructed from the pooled human saliva of 11 individuals. Screening of this metagenomic library for tetracycline resistant clones resulted in the isolation of two of clones capable of growing on tetracycline (5 μg/ml), including PS9.

Analysis of the 7,765 bp PS9 insert revealed it to have nucleotide similarity along its entire length with Streptococcus sp. 263_SSPC (accession: GCA_001071995.1, 98% cover and 90% identity) and Granulicatella adiacens ATCC 49175 (accession: NZ_ACKZ00000000, 94% cover and 92% identity). The alignments also identified an inversion in the PS9 insert between 1,600 bp and 1,789 bp (Supplementary Figures S1A,B). BlastX analysis of the insert predicted it to contain five putative ORFs (Figure 1A). The hypothetical products of the five ORFs had a homolog with >90% amino acid identity from Streptococcus sp. 263_SSPC and G. adiacens ATCC 49175. Of the ORFs identified, three encoded a UDP-galactose mutase, sulfurtransferase and amidohydrolase. The remaining ORFs were predicted to encode two half ABC transporters that were named TetA(60) and TetB(60) as their putative amino acid sequences had less than 80% similarity to any other tetracycline resistance protein amino acid sequence and therefore fulfilled the criteria for a new tetracycline resistance gene. Interestingly, in the Streptococcus sp. 263_SSPC genome sequence two transposase genes were found 634 bp and 6,196 bp upstream of the UDP-galactose mutase. No such genes were identified in the G. adiacens ATCC 49175 genome. Clustal Omega alignments of the putative amino acid sequences of TetA(60) and TetB(60) to characterized antibiotic resistance heterodimeric ABC transporters showed that they were more closely related to TetA(46) and TetB(46) (39.27% and 42.28% identity, respectively) and YheH and YheI (40.93% and 46.61%, respectively). TetAB(60) were less related to the MDR ABC transporters EfrAB, PatAB and LmrCD of E. faecalis, S. pneumoniae and L. lactis, respectively (≤34.46%), Table 3.

Each putative ABC half transporter peptide was predicted to be 579 amino acids and both contained a predicted NBD and a TMD, which are hallmarks of ABC transporters. Additionally, there is a 4 bp overlap of the genes, with the start codon of tetB(60) being contained within tetA(60), although the genes are in different reading frames.

To determine if these two genes were responsible for the observed tetracycline resistance phenotype of PS9, tetA(60) and tetB(60) were individually and jointly subcloned into E. coli EPI300 using the pHSG396 cloning vector. Only E. coli::pHSG396tetAB(60) grew on 5 μg/ml of tetracycline, showing that both tetA(60) and tetB(60) were required for the tetracycline resistance.

In order to ascertain whether the gene products function as a heterodimeric ABC transporter that confers resistance to tetracycline, a 69 and 57 base pair deletion was made to remove the Walker A motif of the NBD from either tetA(60) or tetB(60), respectively (Figure 1B). Both mutants, E. coli::pHSG396 tetB(60)ΔtetA(60) and E. coli::pHSG396tetA(60)ΔtetB(60) were susceptible to tetracycline. This confirmed that the ABC transporter activity of these gene products is responsible for the tetracycline resistance in PS9 and the E. coli::pHSG396tetAB(60), Table 4.

Using the broth dilution method, the MIC of tetracycline for E. coli EPI300, E. coli::pHSG396, E. coli::pHSG396tetAB(60), E. coli::pHSG396tetB(60)ΔtetA(60) and E. coli::pHSG396 tetA(60)ΔtetB(60) was determined (Table 4). The MIC of tetracycline for E. coli::pHSG396tetAB(60) was found to be 32 μg/ml. The MICs for the mutants and the controls strains were 16-fold lower at 2 μg/ml. To determine if tetAB(60) was able to confer resistance to later generation tetracycline derivatives, MIC assays were conducted using minocycline and tigecycline. The MIC of minocycline for all strains and clones was 1 μg/ml. The MIC of tigecycline for E. coli::pHSG396tetAB(60) was 16-fold higher than the control and mutant strains at 8 μg/ml, which was above the clinical break point for Enterobacteriaceae (0.5 μg/ml).

Disk diffusion assays were used to discern the spectrum of resistance for this transporter. E. coli::pHSG396tetAB(60) was less sensitive to tetracycline but equally sensitive to cefotaxime, ceftazidime, metronidazole, neomycin, ciprofloxacin, nalidixic acid, gentamicin, amikacin, amoxicillin/clavulanate and trimetoprim/sulfametoxazole as E. coli EPI300 and E. coli::pHSG396. E. coli EPI300 was intrinsically resistant to erythromycin which has been described previously and attributed to AcrAB-TolC mediated efflux and membrane impermeability (Vaara, 1993; Chollet et al., 2004).

We observed, when it was first isolated, that PS9 grew slower and formed smaller colonies than E. coli::pCC1BAC even in the absence of tetracycline in the growth media; this phenotype was also observed for the E. coli::pHSG396tetAB(60) subclone. Furthermore, it was noted that the E. coli::pHSG396tetB(60)ΔtetA(60) and E. coli::pHSG396tetA(60)ΔtetB(60) mutants did not have such a large growth defect. Growth curves revealed that although there were significant differences between E. coli::pHSG396 and E. coli::pHSG396tetA(60)ΔtetB(60) maximum growth rates (0.993 ± 0.05 and 0.917 ± 0.05, respectively; p = 0.005) there was no significant difference in their OD₆₀₀ of cultures at 7 h (1.65 ± 0.08 and 1.565 ± 0.16; p = 0.1807) when grown in the absence of tetracycline (Figure 2).

Compared to E. coli::pHSG396, E. coli::pHSG396tetB(60)ΔtetA(60) and E. coli::pHSG396tetAB(60) reached lower OD₆₀₀ at 7 h when they were grown in the absence of tetracycline (0.983 ± 0.03, 0.733 ± 0.01, respectively, p < 0.0001). Additionally, when grown in the presence of tetracycline, E. coli::pHSG396tetAB(60) reached an even lower OD₆₀₀ at 7 h (0.543 ± 0.04, p < 0.0001). Whilst the maximum growth rates of E. coli::pHSG396tetB(60)ΔtetA(60) grown without tetracycline and E. coli::pHSG396tetAB(60) grown without or with tetracycline were not significantly different from each other (0.702 ± 0.06, 0.68 ± 0.03 and 0.692 ± 0.09, respectively; p = 0.435 to 0.879), they were 1.14-1.46 fold lower than the maximum growth rate of E. coli::pHSG396 grown without tetracycline (p < 0.0001). We therefore suggest that there is a fitness cost due to the activity of TetAB(60) rather than the carriage of the plasmid itself and that TetB(60) contributes more to this fitness cost than TetA(60).

There has been a resurgence in the use of tetracyclines in human therapy due to the recent development of a number of semisynthetic derivatives of the antibiotic that are efficacious against antibiotic resistant pathogens (Rubinstein and Vaughan, 2005; Anstead et al., 2014; Shen et al., 2015; Lin et al., 2016; Van Berkel et al., 2016). Tigecycline is the first of this new generation of tetracyclines to enter clinical use, being effective against MDR pathogens including carbapenem and colistin resistant microorganisms and those expressing specific tetracycline efflux systems and RPPs (Fluit et al., 2005; Cai et al., 2011). However, it is worrisome that resistance to tigecycline has already been described and associated with multidrug efflux systems and ribosomal mutations (Villa et al., 2014; Zhong et al., 2014; Lupien et al., 2015).

In this study we characterized a clone, PS9, isolated from a human oral saliva metagenomic library that exhibited high levels of resistance to tetracycline and tigecycline. BlastN alignments of the clone with Streptococcus sp. 263_SSPC and G. adiacens ATCC 49175 revealed the PS9 insert to have 90% and 92% nucleotide similarity to these species, respectively, indicating a probable Gram-positive origin for the insert. Streptococcus spp. are predominant in the oral cavity although to the best of our knowledge Streptococcus sp. 263_SSPC has not been identified (Segata et al., 2012). Granulicatella spp. including G. adiacens are also abundant in the oral cavity, typically inhabiting the mucosa (Aas et al., 2005). Tetracycline resistance mediated by Tn916 encoded tet(M) has been described for oral Streptococcus and Granulicatella spp. (Lancaster et al., 2005). Additionally, MFS and ABC transporter genes conferring resistance to tetracyclines have also been identified in oral Streptococcus spp. including tet(L) and tetAB(46), respectively (Chen et al., 2013; Warburton et al., 2013). Although tetracycline resistance has been observed in Granulicatella spp., minimal characterisation studies have been conducted (Zheng et al., 2004; De Luca et al., 2013). Although Streptococcus and Granulicatella spp. are abundant in the oral cavity, it is not known how prevalent tetAB(60) is in the oral cavity and further work beyond the scope of our characterisation study is required to address this.

Two transposase genes were located up stream of the UDP-galactose mutase gene in Streptococcus sp. 263_SSPC, however, it is unknown if these transposases are found in the host genome of the PS9 sequence or if they are capable of transposition of tetAB(60). The alignments also identified an inverted region in the PS9 insert when compared to the Streptococcus sp. 263_SSPC and G. adiacens ATCC 49175 genomes which may have resulted from a DNA breakage followed by repair or a transposition event.

Analysis of the insert revealed it contained five ORFs, predicted to encode a putative UDP-galactose mutase, a sulfurtransferase, an amidohydrolase and two ABC half transporters. Each predicted protein had amino acid sequences with high similarity (>90% identity) to proteins from Streptococcus sp. 263_SSPC and G. adiacens. Numerous heterodimeric ABC transporters capable of conferring antibiotic resistance have been described, including the MDR transporters EfrAB of E. faecalis (Lee et al., 2003), PatAB from S. pneumoniae (Baylay et al., 2015), LmrCD from L. lactis (Lubelski et al., 2004) and the recently characterized EfrCD of E. faecalis (Hurlimann et al., 2016). These transporters have been shown to confer resistance to fluoroquinolones, tetracyclines and biocides among other antimcrobials. Alignment of the putative amino acid sequences of TetAB(60) to other antibiotic resistance heterodimeric ABC transporters showed that they were most closely related to TetAB(46) and YheH/I and less so to the MDR ABC transporters EfrAB, PatAB and LmrCD. As TetAB(46) has been shown to be most closely related to YheH/I this suggested that TetAB(60) was also tetracycline specific (Warburton et al., 2013).

We showed that both tetA(60) and tetB(60) were required to confer tetracycline resistance in E. coli EPI300, suggesting that the product of these genes formed a heterodimeric ABC transporter with each gene product containing a TMD and NBD as revealed by BlastX (Dawson and Locher, 2006). Previous studies have used E. coli as a host to characterize Gram-positive antimicrobial ABC transporters including EfrAB from E. faecalis and LmrA from L. lactis (Lee et al., 2003; Achard-Joris et al., 2005). As E. coli::pHSG396tetAB(60) did not contain the inverted sequence it is likely that this sequence does not affect expression of tetAB(60).

Walker A motifs are found in many ATP utilizing enzymes including ABC transporters and are required for binding and stabilizing ATP (Ramakrishnan et al., 2002). Deletion of these motifs from ATP transporters has been shown to result in a loss of function (Warburton et al., 2013). Individual in-frame deletions of these motifs from either tetA(60) or tetB(60) led to a loss of the tetracycline and tigecycline resistance phenotype providing further evidence that the products of these genes form a heterodimeric ABC transporter.

Compared to E. coli::pHSG396, E. coli::pHSG396tetAB(60) was 16-fold more resistant to tetracycline (MIC of 32 μg/ml) and tigecycline (MIC of 8 μg/ml). Although E. coli::pHSG396tetAB(60) showed levels of resistance to tetracycline and tigecycline beyond the EUCAST breakpoints, it was as susceptible to minocycline as E. coli::pHSG396, indicating that minocycline is not a substrate for this transporter (Olson et al., 2006; Ramos et al., 2009). Efflux mediated tigecycline resistance has been described previously in Pseudomonas aeruginosa and Klebsiella pneumoniae, being attributed to the activity of an ABC and a resistance nodulation division (RND) transporter, respectively (Dean et al., 2003; He et al., 2015; McDaniel et al., 2016). TetAB(46) was also shown to confer low tigecycline resistance in S. australis (Warburton et al., 2013).

TetAB(60) appeared to be specific for tetracycline and tigecycline as disk diffusion assays demonstrated E. coli::pHSG396 to be as susceptible as E. coli::pHSG396tetAB(60) to cefotaxime, ceftazidime, metronidazole, neomycin, ciprofloxacin, nalidixic acid, gentamicin, amikacin, amoxicillin/clavulanate and trimetoprim/sulfametoxazole and erythromycin, providing further evidence for the tetracycline specificity of the ABC transporter.

The observed fitness cost associated with tetAB(60) was not observed in either mutant as although E. coli::pHSG396tetA(60)ΔtetB(60) had a lower maximum growth rate than the control it had a comparable final OD₆₀₀ to E. coli::pHSG396 andE. coli::pHSG396tetB(60)ΔtetA(60) exhibited faster growth than E. coli::pHSG396tetAB(60). This indicated that the growth defect was a result of TetAB(60) activity rather than from maintenance of the plasmid. Additionally, as E. coli::pHSG396tetB(60)ΔtetA(60) grew less well than E. coli::pHSG396tetA(60)ΔtetB(60) it suggests that TetB(60) produces a greater cost to the E. coli host than TetA(60).

We have identified two novel genes from the human oral cavity that likely produce a heterodimeric ABC transporter, TetAB(60). TetAB(60) specifically exports tetracycline and tigecycline conferring high levels of resistance to these antibiotics in an E. coli host. A limitation of this work is that we do not know the prevalence of these genes in the human oral cavity. Further work should be undertaken to survey its prevalence in various niches, to determine how common these genes are, and their possible clinical relevance for treating bacterial infections with tetracycline derivatives. This work also shows that the human oral cavity harbors unknown tetracycline resistance determinants in the absence of any obvious selection pressure. There is potential for these genes to be acquired by mobile genetic elements and transferred to bacterial pathogens, which is particularly worrying given the recent identification of a carbapenem and colistin resistant strains of E. coli some of which could only be inhibited by doxycycline and tigecycline (Liu et al., 2016; Mediavilla et al., 2016; Yao et al., 2016). However, the associated fitness cost of tetAB(60) observed in E. coli may limit any possible fixation following dissemination of the genes from their native host to E. coli strains in the absence of a tetracycline or tigecycline selective pressure.

LR contributed to the design of the experiments as well as to the acquisition, analysis, interpretation of the data included in this manuscript, wrote initial and revised drafts of the manuscript and approves of the final manuscript being submitted and also agrees to be accountable for the work detailed in the submitted manuscript. AR conceived the project, contributed to the design of experiments conducted throughout it, interpreted results and contributed to the drafting and revising of the manuscript being submitted and also approved the final draft of the manuscript and agrees to be accountable for the work detailed in the submitted manuscript. MA conceived the project, contributed to the design and direction of experiments within it, interpreted results and made revisions to the final manuscript, approved the manuscript being submitted and also agrees to be accountable for the work represented in the submitted manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.