Genetic Variations in MexAB-OprM Eux Pump Regulators and Their Association with Antibiotic Resistance and Sequence type in Clinical and Epidemiologically High-risk Clones of Pseudomonas Aeruginosa.

Background: Pseudomonas aeruginosa is a major opportunistic pathogen involved in healthcare-associated infections with high mortality rates. This bacterium exhibits elevated resistance to a wide range of antibiotics, resulting in part from the overexpression of eux pumps, among which MexAB-OprM stands out as constitutive. Antibiotic resistance in clinical isolates is associated with mutations in the mexR, nalC, and nalD repressors that modulate the expression of this eux pump. This study identies point mutations in the mexR, nalC, and nalD genes and investigates their associations with antibiotic resistance and sequence type in clinical and epidemiologically high-risk clones of P. aeruginosa. Results: A total of 91 P. aeruginosa strains isolated at a pediatric hospital in Mexico (2007–2015) were classied according to their resistance to antibiotics. The strains were typed by multilocus sequencing of 7 genes. The MexAB-OprM eux pump phenotype was determined using the minimal inhibitory concentration for the reporter antibiotic carbenicillin in the presence/absence of the eux pump inhibitor Phe-Arg-β-naphthylamine. Sequencing of the mexR, nalC, and nalD genes to identify mutations was performed. Genetic relationship among the strains was evaluated by a phylogenetic inference analysis using maximum likelihood to construct a phylogenetic network. The relationship between variables was determined by a principal component analysis. STs revealed six main complexes. Mutations in the mexR, nalC, and nalD genes revealed 27 different haplotypes. Pan-drug and extensive drug resistant proles were associated with specic STs with haplotypes 1 (ST1725, endemic clone), 8, 12 (ST233, epidemiologically high-risk clone), and 5 [related to dead when compared to ST1725 and ST233 (RRR 23.34; p=0.009 and RRR 32.01; p=0.025)], however the resistance in these strains was not mainly attributed to the MexAB-OprM phenotype. Strains with the same haplotype and resistant prole showed different pump behavior. Conclusions: A signicant relationship between ST and resistant proles was observed; (phylogeny), antibiotic susceptibility (MDR, and (PDR)), and the survival outcome of in a The new STs were deposited in the P. MLST data base. Variability parameters were determined as previously described.

Classi cation of the P. aeruginosa strains (origin, haplotype, susceptibility, MexAB-OprM phenotype, and carbapenemase product    74H  12  2560  R  R  R  R  R  R  R  R  I  R  R  R  R  R  XDR  2048   66H   †   12  2559  R  R  R  R  R  R  R  R  S  R  R  R  R  R  XDR  2048   1A  13  2561  S  S  S  S  R  S  R  I  R  S  S  R  R  S  MDR  512   23H  14  1736  S  S  S  S  S  S  S  S  I  I  I  R  R  R  MDR  Carbapenemases are not the main contributor of resistance in P. aeruginosa strains Carbapenemase expression was evaluated in 74 strains con rmed to be meropenem and/or imipenem resistant. The commercial kit β CARBA Test showed invalid results for 15 strains (Table 1), which produced an orange coloration instead of red (positive result) or yellow (negative result). For the remaining 59 strains, carbapenemase typing (serine carbapenemase or metallo-β-lactamase) was conducted ( Table 1) and showed that 8 strains were positive for serine carbapenemases and 11 strains were positive for metallo-β-lactamases.
Multilocus sequence typing veri es the diversity of most P. aeruginosa strains, and reveals the emergence of outstanding Sequence types. Mutations in the regulatory mexR, nalC, and nalD genes show nalC gene as the most diverse It was observed a total of 62 mutations (49 synonymous; 13 non-synonymous) in the mexR, nalC, and nalD genes (    Table 2 Haplotype was de ned as the DNA sequence of the concatenated mexR-nalC-nalD genes. A total of 27 different haplotypes were identi ed in the 91 strains, including 26 haplotypes with substitutions and one haplotype with a 12-bp deletion ( Table 3). The hospital strains showed the largest number of haplotypes (n = 19) ( Table 1), while the environmental strains had the greatest diversity in nucleotides (Pi) and genes (Hd) (Pi, 0.00922; Hd, 0.879), with the greatest diversity observed for the nalC gene (Pi, 0.01184; Hd, 0.771) ( Table 3). We observed a total of 62 mutations, of which 61 were in hospital strains and most were located in the nalC gene (n = 28) ( Table 2). Data including the number of haplotypes, nucleotide diversity, gene diversity, and nucleotide substitutions in the mexR, nalC, and nalD repressor genes are summarized in Table 3. Table 3 Genetic variations (haplotypes) identi ed in the mexR, nalC and nalD repressor genes in P. aeruginosa strains.
2) Complex II: Includes ST1724, ST1726, ST1728 and ST1727, all of nosocomial origin with the ST1724 identi ed as the AT. This complex forms part of the globally CC235. All STs that make up this complex presented haplotype 5. This haplotype is possibly associated with fatal patient outcomes. Complex II was associated with death when compared to the singletons STs (RRR = 40.01; p = 0.006), to complex I (RRR = 23.34; p = 0.009), and to complex 4 (RRR = 32.01; p = 0.025).
3) Complex III: Includes ST1729, ST540, ST2250, and ST2251, all of environmental origin except for one ST1729 strain. If considering a pro le match at n-3 loci, ST2564 clone can be included in this complex. No AT was identi ed within these STs, however they are part of the global CC253. All STs that make up this complex presented haplotype 26.
4) Complex IV: Includes ST2559, ST2560, and ST233 (n = 6), all of nosocomial origin. ST233 (an epidemiologically high-risk clone) is considered the AT of this complex, conforming the CC233 worldwide. All STs that make up this complex presented haplotype 12.
5) Complex V. Includes ST1737 and ST561, both of nosocomial origin and being part of the CC245 worldwide. Both STs that make up this complex presented haplotype 21 and are possibly associated with fatal patient outcomes.
6) Complex VI: Includes ST2710, ST2704, ST2713, and ST2731. If considering a pro le match at n-3 loci, ST2716 clone can be included in this complex. All of these STs were of nosocomial origin without AT neither global CC identi ed. All STs that make up this complex presented haplotype 8, except for a ST2313 strain that presented haplotype 9, being the only difference between these haplotypes a mutation in the nalC gene S209R.
The rest of the STs are considered singletons since more than 3 locus variants are noticed between them and other STs. The remaining 19 mexR-nalC-nalD haplotypes were identi ed in the STs considered singletons.
Close phylogenetic relationship between CCI and CCII is evident, while the CCIV is the most distant complex. However, all CCs appear to be related somehow to the CCI.
The second phylogenetic network is focus in the relationship between the 27 identi ed mexR-nalC-nalD haplotypes, the 48 STs, and the susceptibility pro les of the P. aeruginosa strains ( Furthermore, the genetic relationship between different STs and the mexR-nalC-nalD haplotypes were corroborated with the phylogenetic neighbor-net network (Fig. 3), in which the 6 important clonal complexes previously described were observed highly conserved. The phylogenetic network showed a close relationship between CCI, CC2 and CC3, and a distant relationship between these clonal complexes and CCIV. However, close relationship between CCIV and CCVI was only evident in this network. The presence of rectangular boxes in the network represents the high probability of extensive homologous recombination, which was corroborated with the PHI test that revealed statistically signi cant recombination events (p < 0.05).
Principal component analysis reveals correlation between ST and mexR-nalC-nalD haplotype PCA analysis of variables associated with each strain showed that the variability of component 1 is 47.81% and that of component 2 is 18.34% (Fig. 4); combined, these 2 components have a variability of 66.15%. PCA also revealed a strong relationship between resistance and ST (Fig. 4) inversely proportional to all analyzed variables, although an association was observed between haplotype 5 strains (mexR-nalC-nalD) (Fig. 4, green dots), haplotype 1 strains (blue dots), and fatal outcomes. Fatal outcomes were observed in 12.5% of haplotype 1 strains, 80% of haplotype 5 strains, 11.11% of haplotype 12 strains, and 16.67% of haplotype 26 strains (p = 0.051) ( Fig. 4; Fig. 1, see haplotype colors). However, it should note that patients' underlying conditions were not considered in this study.
Statistical analysis of the relationship between the mutations in the mexR-nalC-nalD genes, the sequence type, the resistance, the MexAB-OprM phenotype, and patient death outcome gave the following results: Sequence type (ST): It was observed that phylogenetically related sequence types presented equal or similar mexR-nalC-nalD haplotypes (p < 0.05) (Fig. 1, see color by haplotype and phylogenetic relationships between STs; Table 3, see similar mexR-nalC-nalD haplotypes).
Patient death outcomes: It was observed that 7 mutations were associated with patient death outcomes (all identi ed in haplotype 5 strains), including one in the mexR gene (V20V) 3 in the nalC gene (R120R, A123A, and E153Q), and 3 in the nalD gene (C92C, I111I, and D180D) (p < 0.05) (Table 3). However, it should note that patients' underlying conditions were not considered in this study.

Discussion
This study identi es point mutations in the regulatory mexR, nalC, and nalD genes and their associations with antibiotic resistance and sequence type in clinical and epidemiologically high-risk clones of P. aeruginosa. STs revealed six complexes. Mutations in the mexR, nalC, and nalD genes revealed 27 different haplotypes. Pan-drug and extensive drug resistant pro les were associated with speci c STs with haplotypes 1 (ST1725, endemic clone), 8, 12 (ST233, epidemiologically high-risk clone), and 5 (related to dead), however the resistance in these strains was not mainly attributed to the MexAB-OprM phenotype.
Strains with the same haplotype and resistant pro le showed different e ux pump behavior. The results suggest a signi cant relationship between mexR-nalC-nalD haplotypes and phylogenetically related ST; and a signi cant relationship between ST and high-drug resistance in P. aeruginosa strains. However, no statistically signi cant relationship between mexR-nalC-nalD haplotypes and positive MexAB-OprM phenotypes (to which the resistance was completely attributed +) was revealed.
In our previous study, we identi ed and characterized P. aeruginosa strains with MDR, XDR, and PDR susceptibility pro les [19]. Of these strains, the ST1725 endemic clone stands out for its frequency and persistence, with characteristics that strongly suggest high adaptation to the nosocomial environment and potential for epidemiological risk. This study also identi ed the ST233 clone, reported in other parts of the world as an epidemiologically high-risk clone [29,30] and reported in a study in Mexico as resistant to colistin and sensitive to aztreonam. However, because aztreonam is not commercially available in Mexico, the ST233 clone presents a risk to patients in this country.
P. aeruginosa has a great capacity to resist adverse environments, as evidenced by its nosocomial survival. Here we observed a high fraction of PDR (33.7%) and XDR (78%) in strains isolated from a hospital setting, while those from environmental settings were only abundant in MDR (78%) ( Table 1). In our work and worldwide, the worrisome observation has been made that colistin resistant strains are present in both nosocomial and other environments. Colistin is the last therapeutic option for MDR and XDR isolates [31,32]. Dößelmann et al. (2017) warn about the rapid acquisition of colistin resistance by 2 P. aeruginosa clinical strains after they observed a 10-fold increase in resistance after 10 days of exposure to this antibiotic and 100-fold increase after 20 days [32].
The presence of antibiotics or heavy metals in the environment induces MDR and even XDR environmental strains. The nding of resistant bacteria in the environment is attributed to the discharge of antibiotics into waste waters [33] and in industrial waste and the misuse of antibiotics as a preventive measure in livestock and sh farms [4]. These factors may explain our observed high resistance to numerous antibiotics in environmental and clinical strains, which may interact in nature.
The resistance of the 91 P. aeruginosa strains to 14 different antibiotics [28] observed in this study (  Fig. 1). ST111 is considered an epidemiologically high-risk clone and was recently reported in 4 patients with cystic brosis [39]. This clone also was identi ed in Croatia and France, with MDR association [40,41]. Occasionally, P. aeruginosa strains migrate from the environment and cause animal or human infections [24,42]. For this reason, this type of ST is considered highly dangerous and should be kept under surveillance.
MDR P. aeruginosa strains likely result from several factors. The involvement of overexpression of e ux pumps has gained recognition, particularly the MexAB-OprM pump for its constitutive expression and attribution of resistance to most antibiotics [14,15,16,17,18].
E ux pumps are part of the intrinsic protective mechanism used by bacteria to avoid stress and therefore work in response to natural signals and against antibiotics. Resistance-nodulation-cell division pumps in P. aeruginosa have overlapping but non-identical substrates. While MexB transports (among other antibiotics) a broad spectrum of β-lactams, MexD has a narrower spectrum, excluding some of the MexB substrates such as carbenicillin. MexY has the narrowest spectrum, further excluding other antibiotics [43]. Studies suggest that MexB accepts negatively-charged substrates that other pumps do not recognize, as in the case of carbenicillin, which has a high negative charge (-2) compared to other β-lactams [43]. The MIC of a speci c substrate in the presence of the e ux pump inhibitor PaβN is one of the most widely used assessments for evaluating e ux pumps. PaβN both inhibits e ux pumps and permeabilizes the outer membrane of gram-negative bacteria in a dose-dependent manner, such that low doses inhibit e ux pumps and higher doses destabilize membranes [43,44].
Our results corroborate the relationship between the MDR of a given P. aeruginosa strain and its MexAB-OprM e ux pump activity, with 66.23% of nosocomial strains demonstrating activity, principally in PDR strains (73%). E ux pump activity was also higher in strains ST1725 (76.5%) and ST233 (50%), an epidemiologically high-risk clone, compared with other STs (p < 0.0001) ( Table 1) MexAB-OprM expression and multidrug resistance in 50 P. aeruginosa clinical strains [46]. However, in our study, of the P. aeruginosa strains that showed MexAB-OprM activity (phenotypic positive-strains), only in 52.9% the MexAB-OprM e ux pump was the most likely cause of resistance (+), in the remaining 47% of the strains the MexAB-OprM e ux pump was contributing to the resistance (*), which demonstrated that although the MexAB-OprM pump shows positive activity and contributes to resistance, this is not the only mechanism of resistance in these strains [13]. While recent studies have identi ed point mutations in the mexR, nalC, and nalD repressor genes [13,25], few have investigated the relationship between these mutations and the occurrence of multidrug resistance and ST. Here, we observed 27 different haplotypes in the 3 repressor genes. Regarding the mexR gene 3 non-synonymous substitutions were identi ed in 71.42% of the strains, with the V126E amino acid variation being the most frequent, agreeing with previous studies [25]. The 268C→T nonsense mutation, the only mutation encoding a stop codon (Q90*), was observed in 2 strains (haplotype 18), however, until now we do not know its effect on the nal structure of the protein.  [25]. In addition, a 12-bp deletion was identi ed in one strain (Tables 2 and 3). Another study also identi es the nalC gene as the main site of point mutations target, reporting relevant point mutations (non-synonymous substitutions) in 87% of nosocomial isolates (n = 77/90) as well as some deletions [25]. Most of the substitutions reported in that study differ from those we observed (Table 3).
It has been reported that genetic variations as stop codons, frameshifts or deletions lead to loss of functionality of the repressor genes and may contribute to the MexAB-OprM over expression [47]; the deletion identi ed in the nalC gene, was possibly the cause of the MexAB-OprM positive-phenotype that was associated to a PDR phenotype in the strain (haplotype 2), however, in the case of the stop codon identi ed in the mexR gene, no activity was detected from the MexAB-OprM e ux pump that contributes to the XDR shown by these strains, which is possibly partly attributed to the production of metallo-.βlactamases.
The contribution of the MexAB-OprM e ux pump to the resistance of P. aeruginosa strains was evident, as 17 out of 19 statistically signi cant mutations associated with MexAB-OprM e ux pump activity also correlated with resistance of the studied strains (p < 0.05). Some authors suggest that mutations in the mexR, nalC, and nalD genes can impair their function, favoring MexAB-OprM e ux pump overexpression with a consequent increase in bacterial resistance [6,12,13], however, in our study, same haplotypes (mutations) lead to different MexAB-OprM e ux pump phenotypic behavior and resistance.
The haplotype networks highlight the important relationship between ST and mexR-nalC-nalD haplotype in P. aeruginosa strains. In the rst haplotype network (Fig. 1) it was observed 6 important clonal complexes made up of phylogenetically related ST and its relationship with the pump haplotype, and in the second network (Fig. 2) the relationship between ST/haplotype and resistant pro les in the P. aeruginosa strains is evident. These relationships are further supported by principal components analysis (Fig. 4), which shows a clear correlation between ST and resistance and between ST and mexR-nalC-nalD haplotype, but an inversely proportional relationship to the outcome variable (patient death) with all variables analyzed.
Haplotype 1 (mexR-nalC-nalD) was identi ed with high frequency (n = 40) (Fig. 1). All ST1725 strains presented this haplotype except one isolate classi ed as haplotype 2 because of a deletion (△105-116). ST111, an epidemiologically high-risk clone also presented this haplotype (although it appears to be phylogenetically distant from CC1); both haplotypes (1 and 2) were associated with XDR and PDR (Fig. 2). Haplotype 1 was also identi ed in other strains of different ST (ST1723, ST1730, ST2243, ST2245, ST2247, and ST2244) (Fig. 1, CC1), suggesting that these substitutions are speci c to phylogenetically related ST. Likewise, the neighbor-net graph based on the MLST genotyping of the P. aeruginosa strains showed a strong relationship between the ST mentioned above, which suggests that these ST are highly maintained (Fig. 3). However, high-drug resistance shown by haplotype 1 strains was not signi cantly related to the MexAB-OprM activity, since only in 47.5% of the strains the resistance was completely attributed to this e ux pump, in 22.5% the e ux pump contribute to resistance, and in 27.5% of the strains other resistance mechanisms are suggested. Association between haplotypes and high-drug resistance may be due to potential relationship between the ST and mexR-nalC-nalD haplotypes, as resistance is hightly related to speci c STs.
Haplotype 5 (mexR-nalC-nalD) was identi ed in phylogenetically related STs, which were highly associated with XDR (Fig. 2). The high-drug resistance exhibited by these strains could be in part attributed to the MexAB-OprM e ux pump activity (Table 1), however no signi cant relationship between resistance due to the pump was observed, since different behavior of the pump was noticed within haplotype 5 strains. Finally, the relationship between this haplotype and fatal patient outcomes was remarkable.
Haplotypes 8 and 12 (mexR-nalC-nalD) correlate closely with XDR and PDR pro les in the analyzed strains (Fig. 2), with few strains exhibiting contribution of the MexAB-OprM e ux pump to resistance (20% and 33.33%, respectively) ( Table 1), suggesting the relationship between haplotypes and resistance is mainly due to the potential association between haplotypes and phylogenetically related STs, and the latter with resistance (Fig. 2). In these cases, resistance could be explained by the presence of carbapenemases, other pumps, among other mechanisms, as explained by Correa et al (2015) [21]. Interestingly, the carbapenemase types differ between strains of haplotype 8 (serine carbapenemases) and those of haplotype 12 (metallo β-lactamases) [5].
In addition, the 6 clonal complexes previously identi ed in the phylogenetic networks stand out as groups of highly maintained STs in the neighbor-net graph ( Fig. 3): CCI with haplotype 1 strains (mainly ST1725) associated with XDR and PDR. CCII with haplotype 5 strains (ST1724, ST1726, ST1727, and ST1728) were associated with XDR and fatal outcomes (Fig. 2, CC II). CCIII with haplotype 26 strains (ST540, ST2250, ST2251, and ST2564) were identi ed in environmental sites and in a nosocomial strain associated with a fatal outcome (Fig. 2, CC III) suggesting that nosocomial strains might originate from environmental strains, and once in hospitals, can persist for extended periods due to their highly adaptive features and that haplotypes are highly maintained within phylogenetically related STs no matter what environment they came from. CCIV with haplotype 12 strains, characterized by non-synonymous substitutions in the nalC repressor gene (Table 3), correlates closely with the XDR phenotype and ST233 (epidemiologically high-risk clone), ST2559, and ST2560 (Fig. 1, CCIV). CCV with haplotype 21 strains (ST1737 and ST561) both MDR. CCVI with haplotype 8 strains (ST2704, ST2710, ST2716, and ST2731) strongly associated with XDR and PDR in nosocomial strains (CCVI, Fig. 1).
According to the neighbor-net graph, CCII (Haplotype 5) and CCIII (Haplotype 26) appear to be closely associated with the CC1 (Haplotype 1), and distant from CCIV (Haplotype 12) and CCVI (Haplotype 8) (related to production of carbapenemases) which are closely related, indicating that despite the highly recombinant nature of P. aeruginosa, some substitutions are highly maintained among the strains (CCs) (Fig. 3, Table 3). This close relationship between speci c ST that differs little in sequence (CCs) is troubling, as it raises the possibility of the eventual selection of an e cient epidemiologically high-risk clone with high dissemination capacity.

Conclusions
In this study, a signi cant relationship between ST and resistant pro les was observed. The mexR-nalC-nalD haplotypes were not related to the MexAB-OprM e ux pump phenotypic behavior, however activity of this pump was most evident in XDR and PDR strains. In addition, there was a signi cant relationship between mexR-nalC-nalD haplotypes and phylogenetically related ST, suggesting mutations in these repressors are highly maintained within these STs.
The haplotypes 1, 5, 8, and 12 stand out for their frequency, strong relationship with resistance and their association with outstanding STs, and should be under supervision, specially haplotype 5 which was closely related to the death outcome of the patients, however it should be noted that patients' underlying conditions were not considered in this study.
The MexAB-OprM e ux pump is one of the most important mechanisms, to whom P. aeruginosa resistance is attributed; however not in all strains analyzed in this study, the resistance is entirely caused by this pump, in many cases the MexAB-OprM pump contributes to resistance, but other resistance mechanisms must be taken into consideration in the P. aeruginosa strains high-drug resistance.

Methods
This study aims to identify point mutations in the regulatory mexR, nalC, and nalD genes of the MexAB-OprM e ux pump and their associations with antibiotic resistance and sequence type in clinical and epidemiologically high-risk clones of P. aeruginosa.
Previously, 58 MDR, XDR, and PDR strains of P. aeruginosa were collected from patients at the Hospital Infantil de México Federico Gomez, a level 3 health care institute. Responsible for a high mortality rate in patients during 2007-2013, these strains were biochemically and molecularly characterized by Aguilar-Rodea and colleagues (2017) [19]. These 58 previously analyzed strains plus 33 clinical and environmental isolated strains from 2014 to 2015 were analyzed in this study.
In total, 91 P. aeruginosa strains were analyzed: 77 strains of nosocomial origin, isolated at the Central Laboratory of the Hospital Infantil de México Federico Gomez during 2007-2015, and 14 environmental strains isolated from soil, water, and plants by the Aerobiology Laboratory at the Centro de Ciencias de la Atmósfera, UNAM during 2014. Strains were taxonomically identi ed using the automated system MALDI-TOF (Biomerieux Marcy l'Etoile, France). All P. aeruginosa strains were cultured on blood agar plates at 37ºC for 24h.
The following assessments of P. aeruginosa phenotypes were conducted: observation of macroscopic and microscopic morphology, oxidase and catalase tests, hemolysis test in blood agar, pigment production in Mueller Hinton agar (pyocyanin and pyoverdine), odor, Kliger biochemical test, and growth in cetrimide agar at 42ºC [48].

Susceptibility pro les
The susceptibility pro les of 58 P. aeruginosa strains were reported by Aguilar-Rodea et al., 2017 [19]. For the remaining 33strains, the susceptibility pro les were determined according to the minimal inhibitory concentration (MIC) for 9 different antibiotic categories [28] using the agar dilution method described by Identi cation of carbapenemase-producing P. aeruginosa Carbapenemase-producing P. aeruginosa were screened using the phenotypic technique βCARBA Test (BIO-RAD, France). Rapid detection of carbapenemaseproducing strains is a qualitative colorimetric test used to detect strains with decreased susceptibility to carbapenems due to carbapenemase production. This assessment is based on the color change of a pH indicator following hydrolysis of the β-lactam ring in carbapenem.

Genotyping via Multilocus Sequence Typing (MLST)
The ST of 58 of the P. aeruginosa strains was previously determined [19]; for the remaining 33 strains, the same genotyping procedure via MLST was performed. Nested PCR for the metabolic genes acsA, aroE, guaA, mutL, nuoD, ppsA, and trpE was carried out using the primers described by Curran et al., 2004 [50]. Sequencing of the PCR products was performed in both senses. The obtained sequences were edited and aligned as previously described. The ST of each strain was obtained by BLAST analysis (nucleotide) of each gene compared with the P. aeruginosa MLST database [51], http://pubmlst.org/paeruginosa/. The new STs were deposited in the P. aeruginosa MLST data base. Variability parameters were determined as previously described.
Identi cation of mutations in the MexAB-OprM e ux pump repressor genes mexR, nalC, and nalD The mexR, nalC, and nalD repressor genes were ampli ed by PCR in a Thermo Hybaid Thermal cycler (PCR Express, California). The following primers were designed using the Primer3 program [52], PCR products were sequenced in both senses using the primers for each gene described above using a Genetic Analyzer 310 sequencer (Applied Biosystems,

Phylogenetic Analysis
The phylogenetic relationship and evolutionary history of the 91 nosocomial and environmental P. aeruginosa strains were evaluated by the construction of a phylogenetic network using maximum likelihood. A minimum-spanning tree was built from the MLST (ST) sequences using the GrapeTree (visualization of genomic relationships) [57] and PhyloViz Online (visualization and phylogenetic inference) [58], https://online.phyloviz.net/index softwares. In addition, groups of related STs (clonal complexes) were identi ed using the BURST analysis. A group of related STs was de ned as a pro le match at n-2 loci to any other member of the group (n= number of loci in the scheme, MLST=7); default settings were used to achieve the most stringent de nition. The GrapeTree, PhyloViz Online and BURST analysis softwares are available at the Pseudomonas aeruginosa PubMLST database [51], http://pubmlst.org/paeruginosa/.
Furthermore, to determine the evolutionary relationships and events of recombination between the STs a phylogenetic network was built from the MLST (ST) of the P. aeruginosa strains using the neighbor-net algorithm (distance-based-method) implemented in SplitsTree ver.4.0. [59]. The robustness of the network was calculated with a bootstrap test after 1000 pseudo replicates and the inference of recombination events during the generation of allelic variation was estimated with the pairwise homoplasy index test (PHI).

Genetic Diversity
For each of the MLST and MexAB-OprM e ux pump repressor genes, the number and frequency of haplotypes was determined, as well as the estimated nucleotide diversity, including the nucleotide diversity per site (Pi) and expected variation per site assuming a neutral evolution (Eta). The number of substitutions (S) for each gene is reported as well. All data were obtained using DnaSP ver 5.10.01 [56]. The DnaSP program allows the analysis of DNA polymorphisms using data from several loci by estimating several measures of DNA sequence variation within and between populations.

Statistical Analysis
The nature of the data determined the type of statistical analysis used. Qualitative variables de ned subgroups of the total cohort; therefore, associations between variables required the construction of contingency tables to identify association patterns from the counts within these values. Statistical signi cance was considered as p <0.05 as determined by Fisher´s exact test using STATA/MP 14.1 [60]. To investigate the effects of explanatory variables on a binary response variable we used logistic regression models, while for a categorical dependent variable with outcomes that have no natural ordering, multinomial logit models were used. All procedures were done with the STATA/MP 14.1 program [60].  Phylogenetic network based on the MLST genotyping of the P. aeruginosa strains (ST/haplotypes/resistance). Relationship between mexR-nalC-nalD haplotypes, STs and susceptibility pro les are shown. P. aeruginosa isolates (n= 91; 48 STs; 27 mexR-nalC-nalD haplotypes). Circles represent sequence types (STs); Circumference of the circle is based on ST frequency; Two or more strains with the same ST are depicted as fractions in each circle (in addition ST1725 (n= 34 strains)); Lines connect locus variants; Numbers indicate the number of locus variants among the connected STs. Clonal complexes (CC) formed are highlighted in rectangles and described as (I, II, III, IV, V and, VI). STs not grouped into a CC are considered singletons (>3 locus variants with other STs).