Interplay between MexAB-OprM and MexEF-OprN in clinical isolates of Pseudomonas aeruginosa

MexAB-OprM and MexEF-OprN are Pseudomonas aeruginosa efflux pumps involved in the development of antibiotic resistance. Several studies developed with laboratory strains or using a few clinical isolates have reported that the regulation system of MexEF-OprN is involved in the final levels of MexAB-OprM expression. Therefore, this study was aimed to determine the interplay between MexAB-OprM and MexEF-OprN in 90 out of 190 P. aeruginosa clinical isolates with an efflux pump overexpression phenotype. Regarding oprD, 33% (30/90) of isolates displayed relevant modifications (RM) defined as frameshift or premature stop, both related to carbapenem resistance. On the other hand, 33% of the isolates displayed RM in nalC, nalD or mexR, which were significantly associated with multidrug resistance (MDR), non-susceptibility to carbapenems, OprD alterations and strong biofilm production. Meanwhile, the RM in MexS were associated with presence of pigment (p = 0.004). Otherwise, when all the regulators were analysed together, the association between RM in MexAB-OprM regulators and MDR was only significant (p = 0.039) when mexS was the wild type. These data show the modulatory effect of MexEF-OprN on MexAB-OprM in a clinical population of P. aeruginosa. Further studies may contribute to design of novel molecules acting on this interplay to fight against antimicrobial resistance.

MexEF-OprN) exhibit low-level production of MexAB-OprM and virulence factors including a lower ability of biofilm formation 15,16 . This inverse relationship between the expression of MexAB-OprM and the expression of MexEF-OprN could reflect an overlapping of antimicrobial substrates or similar cell-associated extruded products by any of these systems 17 . Nonetheless, this finding may also cause differences in virulence, antimicrobial resistance or specific properties 1 .
Overall, several studies have proposed the interaction between the two efflux systems 2,7,16 , however most of them have mainly been developed in laboratory strains or a few clinical isolates. The purpose of this study was to determine the interplay between MexAB-OprM and MexEF-OprN in antimicrobial resistance, the oprD gene, biofilm formation, swarming motility and pigment in a wide variety of clinical isolates of Pseudomonas aeruginosa from two Peruvian hospitals.
The present data demonstrate an inverse relationship between multidrug resistance (MDR) and EPO (p = 0.0006). In addition, the EPO phenotype was associated with LVX and carbapenem susceptible isolates (p = 0.0093 and 0.0013, respectively) ( Fig. 1). Most of the P. aeruginosa RND-efflux pumps may extrude fluoroquinolones. Therefore, increases in their activity (either that of an efflux pump alone or of two or more concomitantly) may be easily detected using a fluoroquinolone such as LVX together with an efflux pump inhibitor (EPI) 1 . Nonetheless, the present results agree with a limited effect of efflux pumps, by itself and in the absence of other mechanisms, on the change of the clinical strain classification from Susceptible to Intermediate / Resistant 1 . Additionally, these results may reflect the different substrate affinities presented by different efflux pumps 1  Overall, 67% (60/90) of isolates displayed amino acid changes or deletions which did not affect the OprD frameshift. Thus, 7 isolates showed sequences identical to the P. aeruginosa PAO1 strain, 35% (21/60) isolates showed punctual mutations, and 53% (32/60) of the isolates displayed the amino acid deletions S 373 /G 383 . These amino acid deletions were presented in addition to several punctual mutations including V 127 L, E 185 Q, P 186 G, V 189 T, E 202 Q, I 210 A, E 230 K, S 240 T, N 262 T, T 276 A, A 281 G, K 296 Q, Q 301 E, R 310 E/G, G 312 R, A 315 G, K 347 M, V 359 L, S 403 A, Q 424 E as well as a series of changes between amino acid 372 and 383 ( 372 V-DSSSSYAGL-383 ). Eighteen additional isolates presented these mutations and also possessed relevant modifications. Overall, punctual mutations were presented conforming sets (Table 1) has been related to increased susceptibility to meropenem 19,20 ; therefore, these types of alterations were classified as "irrelevant modifications" when presented alone. The presence of a potential association has been suggested between specific amino acid substitutions in OprD and MLST profiles, observing a series of amino acid deletions plus a set of punctual mutations on analysing 12 isolates belonging to the ST111 21 . This pattern of OprD amino acid substitutions was almost concordant with the present pattern C, having also strong similarities with pattern A. Therefore, despite the absence of specific MLST determinations, the present results suggest the relevant presence of this high risk P. aeruginosa clone in the area. In a previous study 22 , the clonal relationships among these isolates was established, observing a high diversity (72 different clonal patterns). Nonetheless, 72.2% (13/18) of the isolates classified within pattern C, suggestive of belonging to ST111 were from HCNH while the remaining 27.7% (5/18) were from HAL, accounting for 26% and 12.5% of the isolates analysed from each hospital. This finding is in accordance with Kim et al. 21 who described differences in the prevalence of ST111 between different hospitals from the same area. In addition, another common set of amino acid substitutions (pattern B) very similar to those reported by Kim et al. 21 for P. aeruginosa ST298 and ST308 was also detected in 19 isolates. On the other hand, alterations affecting porin functionality (lack of gene, premature STOPs or frameshifts) were classified as "relevant modifications". Overall, 33% (30/90) of the isolates showed relevant modifications; 40% (12/30) presenting frameshifts by base pair insertions and 47% (14/30) possessing premature stops in amino acid codons 65 and 49, and in four isolates no PCR amplification was obtained. Overall, the relevant modifications were strongly associated with carbapenem non-susceptible isolates (p < 0.0001) ( Table 1). This finding correlates with other studies showing that carbapenem resistance is mainly associated with inactivation of the oprD gene 21 . In the present study, 33% of our isolates showed functional alterations containing mainly frameshifts and premature stops in the gene leading to truncated proteins, similar to the results reported by Kim et al. 21 . In addition, although PCR impairment due to DNA polymorphisms cannot be ruled out, the lack of amplification of oprD in four isolates could be explained by the presence of an insertion sequence (IS). In this sense the presence of disruption of the oprD gene by different ISs including the ISPa27, ISPa45, ISPa46, ISPa47, ISPa133, ISPa1328, ISPa1635, ISPst12 or ISPpu21 has previously been shown 23, 24 . remaining 10% (9/90) of isolates repeatedly did not amplify by PCR assay. Difficulties in PCR amplification may have been due to the presence of polymorphisms in the primers annealing regions, or the presence of internal DNA sequence modifications resulting in specific DNA conformation which impaired PCR amplification. However, the non amplification of the mexR gene was probably associated with the presence of a disrupting IS, such as IS21, which has previously been described as breaking mexR and leading to an increased transcription of the mexAB-oprM operon 25 .   26 . The slanted line (/) separates different patterns of modifications. The symbol Δ nt means nucleotide deletion being noted the first and last nucleotides deleted. The amino acid changes located after a frameshift are numbered following the sequence of the wild type strain without considering the presence of this frameshift, and therefore do not represent the protein produced and are only reported for facilitating epidemiological interpretations. In parenthesis, the number of each specific alteration or combined alterations described. In all cases if a relevant modification was found the sequences are listed in this section, irrespectively of the remaining modifications detected. a In isolates in which no mutation was observed, the MexAB-OprM regulator sequences were identical to those of P. aeruginosa PAO1 (GenBank: AE004091.2). b In all isolates in which PCR amplification was obtained, the mexS and mexT genes were identical to those of P. aeruginosa PA14 (GenBank: CP000438). In relation to the nalC gene, 87% (77/90) of the isolates showed punctual mutations, being G 71 E (76/77) and S 209 R (67/77) the most frequent. In addition, one isolate showed a 10 base pair deletion from C 234 to G 243 . Regarding nalD, 71% (64/90) of the isolates did not show modifications, and 20% (18/90) showed relevant modifications, being two base pair deletions (Δ nt397-398 ) the most frequent in 39% (7/18) of the isolates (Table 2). Similar to our results, the presence of deletions in nalC has been previously shown in P. aeruginosa isolates 26 . In addition, Haenni et al. 27 described different alterations in the nalD gene, including a gene disruption mediated by ISAs2 in isolates of P. aeruginosa. This finding may have occurred in two of our isolates that did not amplify this gene.

Regulatory genes studies in
It has been described that genetic events such as frameshifts, disruptions or premature stops, which lead to loss of functionality of nalC, nalD or mexR are expected to up-regulate the mexAB-oprM operon [25][26][27][28] , and therefore were considered as relevant modifications. In the present study, 33% (30/90) of the isolates displayed relevant modifications in the mexR, nalC or nalD genes which were significantly associated with MDR (p < 0.0001), carbapenem non-susceptible isolates (p < 0.0001) and relevant modifications of the oprD gene (p < 0.0001). In addition, these relevant modifications were significantly associated with strong biofilm producer (SBP) isolates (p = 0.006) [Fig. 2a].
Meanwhile, several of the amino acid changes detected, including some of the most frequently found, such as V 126 E detected in mexR in 39 isolates or G 71 E, S 209 R or G 71 E, A 145 V, S 209 R detected in nalC in 54 isolates and 5 isolates, respectively, have previously been described in isolates not displaying MexAB-OprM overexpression 26,29 . Thus, mexR, nalC and nalD amino acid changes were classified as "irrelevant modifications".
To fully determine the role of the modifications detected in the final expression levels of the mexA gene, 20 isolates carrying different modifications in efflux-pump regulator genes were selected. The results showed that 8 out of 11 isolates (1084, 1085, 1086, 1089, 1090, 1093, 1094, 1096) carrying relevant modifications in mexR, nalC or nalD presented relative mexA expression levels of 1.61 to 5.10 compared to PAO1. Meanwhile, only 2 out of 9 isolates (1082 and 1092) carrying irrelevant modifications presented expression levels higher than PAO1 (1.51 to 3.58) ( Table 3). Therefore, on analysing the selected isolates together it was observed that relevant modifications were associated with higher mexA expression levels (p = 0.02). Previous studies have reported that the MexAB-OprM efflux system contributes to the intrinsic resistance of P. aeruginosa to several antimicrobials such as quinolones, chloramphenicol and most β-lactams and its overexpression contributes to MDR phenotypes 1,2,30 . Similarly, in the present study, isolates with relevant modifications in any of the MexAB-OprM regulators analysed were significantly associated with MDR. This finding is in accordance with the above mentioned different effect on final resistance levels of the overexpression of different efflux pumps, contributing to explain the observed lack of association between MDR (or LVX) and overall EPO, highlighting the role of MexAB-OprM in the development of antibiotic resistance.
In addition, we observed that the presence of relevant modifications in the MexAB-OprM regulators efflux system was significantly associated with reduced susceptibility to carbapenems (p < 0.0001) and relevant modifications in the oprD gene (p < 0.0001) ( Table 4). This is in accordance with the role of the loss or low expression of OprD combined with the overexpression of this efflux system in carbapenem resistance mechanisms in P. aeruginosa isolates 18,[30][31][32][33][34] . Interestingly, the presence of relevant modifications in MexAB-OprM regulators and OprD lack of functionality would seem to be independent phenomena. Nonetheless, the association observed may reflect an external pressure (e.g.: antibiotic consumption) affecting both systems. Furthermore, in agreement with previous studies in which overproduction of MexAB-OprM was correlated with biofilm production 35 , our results showed that the presence of relevant modifications in MexAB-OprM regulators was significantly associated with strong biofilm producer isolates.     V 73 A (2) and G 224 S (1). All were considered as fully functional and able to inhibit the expression of MexEF-OprN (Table 2) 15 . The remaining 41% (37/90) of the isolates did not amplify the mexS gene (Table 2). Of these 37 isolates, it was possible to amplify the mexS gene N-and C-terminal regions in 11 isolates, while in 23 isolates only the N-terminal region was amplified. In 3 isolates neither the N-nor C-terminal regions was amplified (Fig. 3).

Regulatory gene studies in
Only two combinations of the mexS and mexT genes were further analysed: i) the mexS and mexT genes identical to those of PA14 (47 isolates), and ii) the non amplified mexS gene and mexT gene identical to that of PA14 (34 isolates). The remaining scenarios were not further analysed due to the small numbers of isolates presenting the required characteristics (9 isolates). The second combination was significantly associated with the presence of pigment (p = 0.004) (Fig. 2b).
As mentioned above, amino acid changes both in mexS and mexT were classified as irrelevant modifications. Nonetheless, a direct effect of the amino acid substitutions detected on the functionality of these regulators cannot be ruled out 15  The mexE gene expression analyses showed that no amplification of mexS only correlated with mexE overexpression in 2 isolates (1087 and 1095). In isolate 1095, in which the expression levels of mexE were of 8.92, both N-and C-terminal regions were amplified while in isolate 1087 (mexE expression levels of 1.98) a PCR product was only obtained by amplifying the N-terminal region. In both cases, this may have occurred due to the presence of an IS disrupting the mexS gene 27 . Meanwhile, in other 6 isolates (1084, 1085, 1088, 1094, 1097, 1101) no deregulation of mexE was observed (mexE expression levels ranging from 0.53 to 1.16), despite the presence of a fully functional mexT, thereby suggesting the presence of polymorphisms in mexS primer annealing regions and/ or specific internal DNA conformation hampering PCR amplification, although a possible impairment of MexT activity in isolate 1088 related to specific amino acid changes cannot be ruled out. Nonetheless, it should be mentioned that a similar scenario of non mexE overexpression in the presence of fully functional MexT and inactive MexS has previously been described 12 . In the remaining isolate analysed (1091) with relevant alterations in mexS, the mexE expression levels of 0.37 were concordant with the absence of mexT amplification, which as mentioned above might be related to the presence of a disrupting internally inserted sequence (Table 3) (Table 3). This result shows the role of other regulators in the final expression levels of mexEF-oprN. In this line, modifications of the mvtA gene have been related to mexEF-oprN overexpression 36 .
Different from what was observed on analysing the MexAB-OprM regulators, no association was found between relevant modifications in MexEF-OprN regulators and MDR or the presence of an oprD gene frameshift. The only association observed was present among isolates with MexEF-OprN regulator relevant modifications and the presence of pigment (Fig. 2b). This finding disagrees with the reduced production of virulence factors such as biofilm formation, pyocyanin or rhamnolipids among others, in isolates overexpressing MexEF-OprN 34,37 . Nonetheless, it should be taken into account that despite the above commented impairment in the production of pyocyanin in isolates overexpressing MexEF-OprN, a role of MexEF-OprN in the excretion of intermediates of pyocyanin biosynthesis has been proposed 38 .

Interplay of the MexAB-OprM and MexEF-OprN. Previous studies have reported that C4-HSL induces
the expression of the mexAB-oprM operon directly by binding at the MexR-MexAB-OprM operator-promoter region 7 . It has been reported that the nfxC mutant isolates overexpress MexEF-OprN, decreasing the production of C4-HSL 7 , and subsequently those of MexAB-OprM, thereby having a negative effect on MexAB-OprM exported products and homoserine lactone-dependent virulence factors 7 . Likewise, the association between relevant modifications in MexAB-OprM regulators and MDR was only significant (p = 0.039) when mexS was wild type, and therefore able to exert a negative regulation effect on the expression levels of MexEF-OprN. Furthermore, in 2 out of 3 isolates (isolates 1095 and 1100) in which the presence of relevant modifications in the mexAB-OprM regulators did not result in mexA overexpression (expression levels of 1.04 and 0.8 respectively), the expression levels of mexE were of 4.81 and 8.92 (Table 3).
On the other hand, the final expression levels of MexAB-OprM and MexEF-OprN with isolate 1086 showed increased expression levels of both mexA (expression levels of 5.10) and mexE (expression levels of 7.92) ( Table 3), which agree with the concomitant overexpression of both efflux systems previously described in several P. aeruginosa clinical isolates by different authors 12,13 .
Overall, the present data showed a relevant role of modifications leading to the loss of MexR, NalC and NalD functionality in the clinical isolates analysed, which were associated with higher levels of antibiotic resistance and different bacterial virulence including biofilm formation. The effect of these modifications on multidrug primer 2. (d) No amplification of Mex S gene and no amplification of N-and C-terminal regions. Two scenarios are considered. Scenario 1: the mexS gene has been deleted or is absent. Scenario 2: polymorphisms are present in both annealing position of primers 1 and 2. If this late option was right, the most probable is the presence of additional differences in the sequence.

Methods
Bacterial strains. We studied a total of 190 isolates of P. aeruginosa from clinical samples of patients attended at the HAL (78 isolates) and the HNCH (112 isolates) in Lima (Peru), from December 2012 to June 2013. In all cases only non-duplicated isolates from different patients were included in the study. The isolates were stored at −70 °C in skim milk medium (Oxoid, Hampshire, UK) until use. The clonal relationships, carbapenem susceptibility and MDR levels, biofilm formation, swarming motility and pigment presence were determined in a previous study 22 . In all cases antibiotic susceptibility was classified, according CLSI breakpoints 39 . MDR was defined as resistance to three or more unrelated families of antibiotics (aminoglycosides, β-lactams, fluoroquinolones and polymyxin). The isolates intermediate or resistant to both imipenem and meropenem were classified as "carbapenem resistant". Throughout the text the term "resistance" englobes resistant and intermediate isolates.
Efflux pump inhibition test. EPO  oprD gene amplification. Amplification of the oprD gene was performed by PCR (Table 4) 41 . Negative PCRs were performed twice in order to avoid false negative results. In all cases the PCR products were recovered and fully sequenced and compared to the reference strain PAO1 (GenBank: AE004091.2).
Efflux pumps gene regulators. The primers and PCR amplification conditions of the efflux regulator-encoding genes mexR, nalC, nalD, mexT and mexS were designed by Solé et al. 9 with slight modifications of the annealing conditions (Table 4). All PCR products were sequenced as above. When the PCR product did not amplify, the assay was performed twice to avoid false negative results. After that, negative PCR were considered to as genes with "relevant modifications". The mexR, nalC and nalD genes were compared with those of P. aeruginosa strain PAO1 (GenBank: AE004091.2). However the mexT and mexS genes were analysed according to the full functional MexS and MexT of P. aeruginosa PA14 (GenBank: CP000438), because PAO1 lacks the functionality of those genes related to the presence To determine the presence of undetected insertions within to the mexS gene, a PCR strategy was designed. Briefly, in those isolates in which PCR product was repeatedly not obtained, two new PCR reactions were designed in order to amplify the N-and C-terminal regions, respectively (Table 4).
Efflux Pump expression. The expression levels of mexA and mexE were determined in 20 P. aeruginosa isolates representative of the different alterations encountered in the regulator genes. mRNA extraction and qRT-PCR were performed following the primers and methodology previously described (Table 4) 13,42 . In all cases, gene expression was normalised versus rpsL housekeeping gene and expression levels were indicated as a ratio to the expression level in strain PA01 (mexA) or PA14 (mexE).
Statistical analysis. The χ 2 (Chi square test) was used to determine the presence of significant differences which were considered with a p value of ≤0.05. R studio version 3.4.0.was used for all statistical analyses. Resistant and intermediate isolates were classified together as "non-susceptible" for statistical analyses.
Compliance with ethical standards. The study was approved by the Ethical Committee of the Universidad Peruana Cayetano Heredia (Lima, Peru) and by the Ethical Committee of Hospital Clinic (Barcelona, Spain), and all experiments were performed in accordance with relevant guidelines. All samples were obtained within routine clinical practice; no personal data was requested or available to researchers.