Azithromycin resistance levels and mechanisms in Escherichia coli

Despite azithromycin being used in some countries to treat infections caused by Gram-negative pathogens, no resistance breakpoint for Escherichia coli exists. The aim of this study was to analyse the levels and mechanisms of azithromycin resistance in E. coli. The presence of chromosomal (rplD, rplV and 23S rRNA) mutations, 10 macrolide resistance genes (MRGs) and efflux pump overexpression was determined in 343 E. coli isolates. Overall, 89 (25.9%) isolates had MICs ≥ 32 mg/L to azithromycin, decreasing to 42 (12.2%) when assayed in the presence of Phe-Arg-β-Napthylamide, with 35 of these 42 possessing at least one MRG. Efflux pumps played a role in azithromycin resistance affecting the Minimal Inhibitory Concentration (MIC) levels of 91.2% isolates whereas chromosomal alterations seem to have a minimal role. At least one MRG was found in 22.7% of the isolates with mph(A) being the most commonly found gene. The mph(A) gene plays the main role in the development of azithromycin resistance and 93% of the mph(A)-carrying isolates showed a MIC of 32 mg/L. In the absence of a specific resistance breakpoint our results suggest a MIC of 32 mg/L to be considered in order to detect isolates carrying mechanisms able to confer azithromycin resistance.


Results
Antibiotic susceptibility levels. The MICs of azithromycin ranged between 0.06 mg/L and >256 mg/L, with a MIC 50 of 8 mg/L and MIC 90 of 128 mg/L (Table 1).
In the presence of PAβN all groups showed decreased levels of resistance, which was significant (P = 0.0080) amongst EAEC isolates ( Table 1).

Effect of PAβN.
In all cases the isolates were able to grow in the presence of PAβN. As mentioned above the addition of PAβN affected the azithromycin susceptibility levels (Tables 1, 2 Table 2). In 47 out of 89 (52.8%) azithromycin-resistant isolates, the addition of PAβN resulted in a MIC within the range of susceptibility (Table 1, Fig. 1). On the other hand, 35 out of these 47 isolates (74.5%) possessed at least 1 MRG (unidentified in one casesee conjugation results below).
Two commensal and 4 diarrhoeagenic isolates presented a MIC I > 256 mg/L and a MIC PAβN ≥ 256 mg/L, thereby not allowing the effect of PAβN to be accurately established. www.nature.com/scientificreports www.nature.com/scientificreports/ As a general rule the MIC I /MIC PAβN quotient ranged from 4 to 16 (267 isolates, 77.8% of total isolates). The MIC I /MIC PAβN mode was 8 (overall, and among commensal and diarrhoeagenic groups), while the mean effect was 12 (Table 2). When the diarrhoeagenic group was subdivided into pathotypes, only DAEC and EAEC showed slight differences (Tables 1 and 2).
Analysing the effect of PAβN in 255 diarrhoeagenic and 82 commensal isolates, a non-significant trend of a higher number of affected commensal isolates was observed (P = 0.0810). Thus, the effect of PAβN was not observed in 8.6% and 2.4% diarrheogenic and commensal isolates respectively. Despite the significant effect of PAβN on the MIC of EAEC isolates, 11 (14.7%) were not affected by PAβN. Interestingly, 10 out of these 11 isolates presented MIC I of 64-32 mg/L and MIC PAβN of 32-16 mg/L, with MRGs being detected in only 2 cases. In  ; wt: wild type. Any MIC category with ≥5% of the isolates is highlighted in dark grey. If a strain had a L4 and/or L22 mutation(s) and a MRG, then the isolates are included in either the mph(A) or MRG category. 1 One isolate (isolate 3491) in which an unidentified MRG was detected by conjugation.
www.nature.com/scientificreports www.nature.com/scientificreports/ addition, 3 DAEC isolates (15%) were also not affected by PAβN presenting borderline significant differences with commensal isolates. target mutations. Only 17 out of 263 isolates analysed (6.5%) presented mutations in the rplD or rplV genes. Thus, 6 isolates had mutations in the rplD gene and 7 in the rplV gene, while 4 isolates presented amino   www.nature.com/scientificreports www.nature.com/scientificreports/ acid codon alterations concomitantly in both genes. Thirteen of these had a MIC I ≥ 32 mg/L (including 3 presenting mutations in both of the targets analysed), but only one (isolate 3491), in which an unidentified MRG was detected by conjugation (see below), remained resistant when the MIC PAβN was established. In 4 cases were detected concomitant MRGs ( Table 4). None of the isolates analysed had mutations in the 23S rRNA gene.
Macrolide resistance genes. Seventy-eight isolates (22.7%) possessed at least one MRG ( Table 5). The MRG most frequently found was mph(A), which was present in 53 isolates (67.9% of isolates possessing MRG) belonging to all the groups analysed. In 43 cases no other MRG was detected, while in the remaining 10 cases mph(A) was detected together with the erm(A) gene in 4 cases, the erm(B) gene in 3 cases and the mef(A) and ere(A) gene in 2 and 1 cases, respectively. When more than one MRG was identified within the same isolate the mph(A) gene was always present.
MRG were significantly more frequent among EAEC and DAEC isolates than among the remaining groups analysed, except when EAEC were compared with commensals. In addition, significant differences were also observed in the presence of MRGs among commensal and EPEC isolates (P = 0.0195) ( Table 5).   www.nature.com/scientificreports www.nature.com/scientificreports/ The presence of the mph(A) gene was correlated with higher MIC levels ( Table 3, (Figs 1 and 2). The cumulative MIC curves of wt isolates and those presenting a MRG other than mph(A) were similar. The cumulative MIC curves of the isolates possessing target mutations, mph(A) and mph(A) plus other MRG were sequentially displaced towards higher MIC levels. When the cumulative MICs were established in presence of PAβN the results showed that those belonging to wt isolates, and those presenting MRG or L4/L22 amino acid substitutions were close similar, with only a spurious displacement towards high MIC levels of those non-wt, while isolates possessing mph(A) and mph(A) plus other MRG were sequentially displaced towards higher MIC levels in a clear manner (Fig. 3).

Discussion
Diarrhoea-related deaths in children remain among the most relevant health challenges worldwide, being of special concern in low-and middle-income countries 1,2 . In these countries, antibiotic therapy when needed may be crucial to achieve a successful outcome 21,22 . However, antibiotic resistance to commonly used antibacterial agents is dramatically increasing requiring new alternatives.
Regarding the feasibility to considered azithromycin as an alternative to treat diarrhoeagenic E. coli in the studied areas, the present study showed moderate azithromycin resistance levels highlighting some concerns about its usefulness as treatment in the absence of antibiotic susceptibility data, especially when EAEC or DAEC isolates are present.
In accordance with what has been previously described 20,23 , the relevant role of PAβN-inhibitible efflux pumps in azithromycin resistance has been demonstrated once more. However, differences related to the specific bacteria groups were observed. The presence of a series of EAEC isolates in which no PAβN-effect was observed opens the door to different options, including the presence of alterations in the outer membrane composition which results in a possible azithromycin impaired permeability leading to an increase in the basal azithromycin resistance levels, combined with lesser efflux pump activity, at least in regard to PAβN-inhibitible efflux pumps. Another possibility is the presence of different patterns of overexpressed efflux pumps. In this line, selecting azithromycin resistant mutants in the presence of PAβN a similar scenario was observed (MIC of 32-16 mg/L with no further PAβN effect). In all these mutants the presence of an overexpressed OmpW was observed 24 . In fact, OmpW has been associated with EmrE, an efflux pump belonging to the small multidrug resistance (SMR) family 9,25 . Furthermore, the overexpression of EmrE has been related to E. coli grown in the presence of erythromycin 26 .
In agreement with the presence of up to 7 gene copies and the subsequent need for multiple mutated alleles to visualize an effect on macrolide resistance 9 , in the present study no mutations in the 23S rRNA gene were www.nature.com/scientificreports www.nature.com/scientificreports/ observed in the 66 isolates analysed. Regarding L4 and L22, the alterations detected seem to have a minor role in the development of azithromycin resistance, and most might be gene polymorphisms without antibiotic resistance relevance. Regarding the alterations at L4 and L22 observed, to our knowledge only the alterations at amino acid codon K82, D94 and K98 of L22 have previously been described in in vitro obtained E. coli macrolide-resistant mutants but always concomitantly with other L22 amino acid alterations 27 . The L22 alteration L46Q was present in 3 cases, all having a MIC ≥ 32 mg/L. Although in one case the addition of PAβN resulted in a MIC of 2 mg/L, and another was concomitantly present with the mph(A) gene, a possible slight effect of this alteration on macrolide susceptibility cannot be ruled out.

mph(A) mph(B) erm(A) erm(B) erm(C) ere(A) mef(A) mef(B) msr(A) msr(D)
Regarding MRGs, in our series the relevant role of Mph(A) is undoubtable. This finding is in accordance with what has been previously described in E. coli and other Enterobacteriaceae 9,28-31 . Those isolates with the mph(A) gene presented the highest percentages of azithromycin resistance both in the presence and the absence of PAβN. Nonetheless, relevant differences were observed in the MIC levels among isolates carrying the mph(A) gene. Thus, while 2 mph(A)-carrying isolates had a MIC I of 8 mg/L which decreased to MIC PAβN of 0.25 and 1 mg/L, another 11 isolates in which no other MRG was detected had a MIC I > 256 mg/L which in no case decreased below the breakpoint considered in the presence of PAβN. This heterogeneity may be observed on analysing together different studies performed either in E. coli or other closely related Enterobacteriaceae 9,28-30 . Different explanations may be proposed, including differences related to expression levels which may be due to the number of copies of the gene related to its genetic environment (e.g.: plasmids with different sizes and copy numbers), with alterations at the promotor sequence or with the presence of other undetected MRGs.
The remaining MRGs, seemed to have a marginal role in azithromycin resistance. In fact, the cumulative MIC curve of these isolates was close to that of wt microorganisms. Nonetheless, those isolates presenting the mph(A) together with another MRG ranked among those most resistant and less affected by the addition of PAβN, suggesting a slight contribution of other MRGs to final MIC levels when mph(A) gene is present. This finding was also showed when cumulative MICs were established.
Of these MRGs, among Enterobacteriaceae, the Msr(A) has only been described in E. coli and Enterobacter spp. 20,32 . In the present study, the msr(A) gene was detected in isolates having MIC I of 8 mg/L, supporting the loss of activity of this gene when cloned in E. coli 33 . The other ATP binding transporter studied, Msr(D), it was detected independently of the presence of Mef(A). Moreover, in no case the mef(A) and the msr(D) genes were detected together. To our knowledge this is the first description of the msr(D) gene alone, since it has always been described concomitantly with mef(A) 9 . Nevertheless, the presence of polymorphisms in the mef(A) primers annealing region cannot be ruled out. While the effect of Msr(D) on the final MIC levels was within the range of those previously described, this dissociation might result in impaired Mef(A) 34 . Contrary to what was observed in the present study, Mef(A) has been described to be frequent in Enterobacteriaceae 31 . This difference may be related to the geographical origin of the samples. This is the first description of Erm(A) in Enterobacteriaceae 9,35 . While no data on erm(A) functionality in Enterobacteriaceae has been found, previous studies have described an impairment in the expression levels of erm(C) 36 , which, if combined with a limited gene copy number, might result in a marginal influence on azithromycin MIC levels such as those detected in present study. Regarding Erm(B), the concomitant presence with mph(A) detected here in 3 isolates, has also been previously described 30 .
Also Ere(A) had a minimal role in the resistance to azithromycin in the present isolates. This finding is in accordance with the proposed lack of activity of Ere(A) in azithromycin 37 .
There is controversy about the ability of Mph(B) to hydrolyse azithromycin. Thus, while Chesneau and col 38 . have described its inability to confer azithromycin resistance, other authors have established a similar activity on hydrolysing erythromycin and azithromycin 39 . The only isolate of our study that possessed the mph(B) gene exhibited an azithromycin MIC of 16 mg/L in the absence of PAβN.
Despite this marginal role of most MRGs in the final azithromycin MIC, the detection of 6 out of 10 MRGs among commensal E. coli is noteworthy because of their role as a gene-reservoir 40,41 . Conjugation studies showed that only the mph(A) or erm(B) genes were transferred alone or together. Additionally, in one case in which no MRG was previously detected, transconjugants were obtained showing the presence of an undetermined MRG. In fact other MRGs have been described in E. coli 9,35 . However, it should be noted that the conjugation assay was designed to detect the transference of high levels of azithromycin resistance (>32 mg/L), and thus, if the resistance levels associated with transferable MRGs was lower, the transference of these elements would probably remain undetected.
Although the presence of non-sought mechanisms of azithromycin resistance, similar to observed in the isolate 3491, may not be discharged, and their presence may influences final MIC as observed when mph(A) was present concomitantly with other MRG. The fact that the cumulative MIC curves of those isolates presenting target mutations or MRG other than mph(A) were only slightly higher than those belonging to wt isolates (on special when role of efflux pumps was discounted with the use of PAβN) confirms the spurious or merely complementary role of these mechanisms as primary azithromycin-resistance cause in E. coli and highlight the relevant role of mph(A).
Thus, the present data showed that the mph(A) gene, is by far, the most effective mechanism of azithromycin resistance present, leading to MIC values higher than 32 mg/L in 93% of the cases, while 88.9% of isolates without mechanisms of resistance remained with MIC levels <32 mg/L. Therefore the use of 32 mg/L seems adequate to suspect the presence of mph(A) and in general of non-wt E. coli isolates. Nonetheless, the presence of sporadic E. coli isolates possessing Mph(A) with MIC values of 8-16 mg/L was also showed. Therefore studies are needed to determine the possible need for more conservative breakpoint.
In summary, the present data demonstrate the presence of azithromycin resistance among intestinal, either pathogenic or not, E. coli from the area of Lima, highlighting the need for susceptibility data to adequately use this antimicrobial agent. Moreover, the relevant and hidden role of efflux pumps in the intrinsic levels of azithromycin www.nature.com/scientificreports www.nature.com/scientificreports/ resistance is highlighted, showing the potential clinical utility of efflux pumps inhibitors. The present data indicate that the majority of isolates harbouring mph(A) will have MICs ≥ 32 mg/L. These data, combined with other epidemiological data will be useful to establish an E. coli ECOFF value. Clinical data will be needed to establish breakpoints for azithromycin in E. coli.

Materials and Methods
Bacterial strains. Three hundred forty-three diarrhoeagenic (259 isolates, including 78 EAEC, 41 ETEC, 20 DAEC and 120 EPEC) or commensal (84 isolates) E. coli isolates from faeces samples collected in previous studies from children under 5 years of age in periurban areas of Lima (Peru) were recovered from frozen stocks to be included in the study. The uidA gene of all grown isolates was amplified as previously described by Walk and colleagues as a quality control 42 .
In all cases the previous studies in which were collected the E. coli isolates were approved by the Ethical Committee of the Universidad Peruana Cayetano Heredia, faeces were sampled after informed consent was obtained from parents and/or children legal guardians and all experiments were performed in accordance with relevant guidelines and regulations.
Antimicrobial susceptibility testing. The MIC of azithromycin was determined by the agar dilution method in accordance with the CLSI guidelines 17 (Table 6). In all cases negative and positive controls (microorganisms carrying the MRGs included in the study) were used to validate the results. Additionally random selected positive PCRs were sequenced.

Conjugation assays.
A total of 66 isolates with a MIC ≥ 32 mg/L were selected to determine the transferability of the MRGs. The conjugation was carried out in Luria-Bertani broth (Conda, Madrid, Spain) with azide-resistant E. coli J53 as a recipient strain. Transconjugants were selected in plates containing 150 mg/L of sodium azide and 32 mg/L of azithromycin. In order to avoid considering possible contaminations the relationship of transconjugants and the respective recipient strain was established by REP-PCR 23 . The amplification of the MRGs present in the donor and derived transconjugant strains was performed by PCR as mentioned previously.  www.nature.com/scientificreports www.nature.com/scientificreports/ statistical analysis. The Fisher exact test was used for statistical analysis. P values ≤ 0.05 were considered significant. A microorganism was considered "wt" when no sought mechanism of resistance other than PAβN inhibitable efflux pumps was identified.