High frequency of the exoU+/exoS+ genotype associated with multidrug-resistant “high-risk clones” of Pseudomonas aeruginosa clinical isolates from Peruvian hospitals

The type III secretion system of Pseudomonas aeruginosa is an important virulence factor contributing to the cytotoxicity and the invasion process of this microorganism. The current study aimed to determine the presence of the exoU+/exoS+ genotype in P. aeruginosa clinical isolates. The presence of exoS, exoT, exoU and exoY was determined in 189 P. aeruginosa by PCR, and the presence/absence of exoU was analysed according to source infection, clonal relationships, biofilm formation, motility and antimicrobial susceptibility. The gyrA, parC, oprD, efflux pump regulators and β-lactamases genes were also analysed by PCR/sequencing. The exoS, exoT and exoY genes were found in 100% of the isolates. Meanwhile, exoU was present in 43/189 (22.8%) of the isolates, being significantly associated with multidrug resistance, extensively drug resistance as well as with higher level quinolone resistance. However, the presence of β-lactamases, mutations in gyrA and parC, and relevant modifications in efflux pumps and OprD were not significantly associated with exoU+ isolates. MLST analysis of a subset of 25 isolates showed 8 different STs displaying the exoU+/exoS+ genotype. The MDR basis of the exoU+ isolates remain to be elucidated. Furthermore, the clinical implications and spread of exoU+/exoS+ P. aeruginosa isolates need to be established.

Overall, 34 isolates with mutations in gyrA and/or parC showed resistance to at least one FQ. The most frequent amino acid substitutions were T83I and S87L at GyrA and ParC respectively, which were concomitantly found in 24 isolates, and only GyrA T83I was found in 2 isolates and ParC S87L in another 3 isolates. In addition, one isolate showed the amino acid codon substitution D87N in the gyrA gene, and another presented the double substitution D87N in GyrA and S87L in ParC. Meanwhile two isolates concomitantly presented 2 amino acid codon substitutions in gyrA (T83I/D87N) and one in parC (S87L) showing high MICs to all FQs. Finally, one isolate having the GyrA substitution E153K was detected, showing moderate resistance levels to FQ.

exoU+/exoS+ genotype and β-lactamases.
Overall, 67 isolates suspected of carrying metallo-β-lactamases and/or serine-carbapenemases were phenotypically detected: 22 metallo-β-lactamases, 39 serine-carbapenemases and 6 with both metallo-β-lactamases and serine-carbapenemases. It was of note that 4 of these isolates were susceptible to both imipenem and meropenem. PCR results showed that the bla GIM , bla SIM , bla SPM , bla VIM genes and bla IMI and bla KPC corresponding to metallo-β-lactamases and serine-carbapenemases, respectively were not found in any isolate. All isolates suspected of carrying metallo-β-lactamases presented bla IMP while those which were positive for serine-carbapenemases possessed bla GES . All the exoU+ isolates showed a higher proportion of bla GES 32.5% (14/43) than bla IMP 9.30% (4/43), however no differences were observed (p = 0.246). In contrast, the exoU− isolates showed similar proportions of both genes [bla GES      www.nature.com/scientificreports www.nature.com/scientificreports/ exoU+/exoS+ genotype and multi-locus sequence typing (MLST). MLST analysis showed the presence of 16 different sequence types (STs) among the analysed subset of 25 P. aeruginosa isolates analysed, with 13 exoU+ isolates distributed in 8 different ST patterns. Of these, ST235 and ST357 were the most frequently found. ST235 was detected in 6 isolates from HNCH, with 5 isolates being exoU+ (2 XDR, 2 MDR and 1 MR) and 1 exoU− (MR). ST357 was detected in 5 isolates from HAL, all being XDR and only susceptible to colistin and with two presenting the exoU gene. Furthermore, five new STs were found in the present study (ST3300, ST3301, ST3302, ST3303, and ST3305) two (ST3300 and ST3303) being exoU+ [ Table 6].

Discussion
This study aimed to determine the presence of the exoU+/exoS+ genotype and its association with different phenotypic and genetic characteristics, with special emphasis on MDR levels and the underlying mechanisms and efflux pump regulators in clinical isolates of P. aeruginosa. The exoU gene was present in 22.7% of our isolates, with a trend to be more frequent among HAL isolates, which might be explained because the observed association of exoU genotype with patients attending to burn ward (all from HAL). Nonetheless, the exoU+ P. aeruginosa showed no association with a specific source of infection. Other studies have reported that this gene was present www.nature.com/scientificreports www.nature.com/scientificreports/ in 28-42% of P. aeruginosa isolates causing acute infections, being especially related to pneumonia and respiratory infections 8,11,20,21 .
On the other hand, in the current study the presence of the exoS, exoT and exoY genes was found in 100% of the isolates. Similar results were reported for other studies in which the prevalence of these genes varied from 58-72% for exoS, for 89% of exoY, for 92-100% of exoT 11,22 . Interestingly, previous studies have shown the mutual exclusion of the exoU and exoS genes 5,7,10,11 . However, few studies have reported the concomitant presence of both genes in association with acute infection, being for instance detected in 40 out of 60 (67%) isolates of P. aeruginosa from bacteremia, belonging to 42 different pulse-field gel electrophoresis patterns 16 . The clonal relationships among the current analysed isolates were previously determined by Horna et al. 23 , with the 189 P. aeruginosa distributed in 72 different BOX-patterns; of these, 27 BOX-patterns were represented by a single isolate and the remaining 45 BOX-patterns including up to 14 isolates 23 . The exoU+/exoS+ genotypes detected in our study were distributed within 25 out of these 72 different BOX-patterns, therefore, as in the study of Morales-Espinosa 16 , the current results do not represent the spread of a successful local clone. In addition, the presence of 16 BOX-patterns containing both exoU+ and exoU− genotypes suggests genetic events of acquisition/loss of the exoU encoding genomic islands and of intraspecies diversity due to the dynamic nature of the accessory genome of this microorganism [24][25][26] . Analysis of MLST patterns in a subset of 25 isolates, also showed high clonal heterogeneity, even in the exoU+ isolates. In addition, 2 of these ST (ST235 and ST357) had for both exoU+ and exoU− isolates. Although unusual, the presence of the exoU−/exoS+ genotype in isolates belonging to ST235 has also been previously described 27 . This finding agrees with the proposed events of acquisition/loss of the exoU gene. Furthermore, the present results support the proposed P. aeruginosa non-clonal epidemic population structure 28 , both highlighting the presence of several high-risk clones (such as ST235 and ST357) with a worldwide distribution 12,25,29 , and also showing the presence of a number of undescribed P. aeruginosa ST patterns in under studied geographical areas.
It has been proposed that P. aeruginosa possessing swarming motility are more prone to presenting T3SS 30 , and some authors have related the presence of swarming and swimming as well as that of exoU to higher virulence 31 . Nonetheless, in our study, no specific association was observed between the presence of the exoU+ genotype and motility. Regarding biofilm formation, Azimi et al. showed that only 2.5% of the isolates presenting the exoU and exoS genes were biofilm producers and that all non-biofilm producer isolates presented the exoU and/or exoS genes 32 . In agreement with this finding, although only 1/43 exoU+/exoS+ genotype isolates were unable to form biofilm, the present results did not show an association between the presence of exoU and SBP.
Some studies have reported that the exoU+/exoS− genotype was found to be significantly associated with MDR compared to the exoU−/exoS+ genotype 1,2,4,33 . This relationship was not observed in our study, since the exoS gene was found in all the isolates. However, the exoU+ isolates were significantly associated with MDR and XDR when compared to exoU− isolates. This association between the exoU genotype and the MDR/XDR phenotypes may be due to the presence of transferable antibiotic-resistant determinants such as integrons carrying mobile gene cassettes within the accessory genome of exoU+ P. aeruginosa 25,34 .
In agreement with other studies, here, the presence of the exoU genotype was associated with increased levels of FQ resistance as well as with P. aeruginosa isolates displaying high MIC levels to this antimicrobial class 1,2,4,22 . Similar to Agnello et al. 35 , the present results agree that the development of high level of resistance to FQ has a lower fitness cost on exoU+ compared to exoU− P. aeruginosa isolates. Of note, the emergence of the exoU+ ST235, established around 1984 29 , coincides with the beginning of the use of FQ 36 , suggesting that the worldwide dissemination of this (and other) exoU+ ST has been favoured by this lack of deleterious effect on fitness of selected QRDR mutations 29 .
In agreement with other studies, the most frequent mutations were found in the amino acid codon 83 (T83I) and/or 87 (D87N) of gyrA and 87 (S87L) of parC 1 . In addition, a higher proportion of multiple TSM in gyrA and parC was found in exoU+ than in exoU− isolates. However, these were not significantly different, as also previously reported by Takata et al. 4 . Three resistant isolates had a single mutation in parC showing that a previous gyrA mutation is not a strict requisite for the acquisition of mutations at other target genes leading to resistance 1 1123  T83I  S87L -G71E,S209R  NA  -H18Y  --64  128  32 Y49* + 4 R R Table 5. Modifications in target genes of the QRDR, efflux pump regulators, the oprD gene and antimicrobial susceptibility to fluoroquinolones and carbapenems in exoU+ and exoU− isolates. The symbol "−" represent wild type isolates; NA: No amplification; the symbol ∆ nt represents nucleotide deletion being noted the first and last nucleotides deleted; the symbol ∆ AA represents amino acid deletions being noted the first and last amino acid deleted; ins nt : nucleotide insertion; The symbol "*" represents codon STOP. FQ-MIC: MIC to fluoroquinolones (LVX: Levofloxacin; OFX: Ofloxacin; CIP: Ciprofloxacin); IMI/MER are isolates showing susceptibility (S); resistance or intermediate susceptibility (R) to imipenem or meropenem performed by the disk diffusion assay. Alterations at nalC, nalD, mexR, mexS and mexT as well as the patterns of OprD were previously described 17 . The patterns of OprD are named accordingly to Horna et al. 17 . 1. Amino acid www.nature.com/scientificreports www.nature.com/scientificreports/ an uncommonly reported substitution in GyrA (E153K) having moderate resistance to FQ (MIC LVX = 32 mg/L, OFX = 64 µg/mL and CIP = 8 mg/L). This mutation has been previously identified in a P. aeruginosa isolate having a MIC of CIP of 8 mg/L and a concomitant amino acid change S87L in parC, and in two FQ-resistant unrelated Escherichia coli isolates, but neither data on MIC levels nor information of concomitant TSM was provided 37,38 .
P. aeruginosa has several RND-type efflux pumps, being MexAB-OprM, MexCD-OprJ, MexEF-OprN and MexXY-OprM efflux pumps well investigated 39 . In our study the exoU+ isolates showed higher proportions of irrelevant modifications in the regulators of MexAB-OprM and MexEF-OprN, and therefore these efflux pumps could presumably show normal basal expression levels. In agreement, in a previous study analysing mexA expression in a subset of isolates included in this study, those isolates having irrelevant modifications in MexAB-OprM regulators showed mexA expression levels equivalent to PAO1 and significant lowers (p = 0.02) than those reported in isolates having relevant modifications in these regulator genes 17 . Furthermore, this finding agrees with other studies showing that the isolates overexpressing mexB were less likely to be found among exoU+ isolates, and the overexpression of mexF and mexD was not correlated with the exoU+ genotype 4 . Although, in our study, almost all isolates showed the mexZ gene with irrelevant modifications (and as above may be considered as fully functional), it has been reported that the isolates overexpressing mexY were significantly associated with the exoU+ genotype 4 .
The 5 isolates possessing resistance to any of the FQ tested in the absence of a TSM were exoU−, further showing relevant alterations in at least one efflux pump regulator gene. Furthermore, data on 1094 isolate MexA and MexE expression levels were recorded in a previous study, showing increased mexA gene expression 17 . This finding suggests that in exoU− isolates FQ resistance will be more prone to be developed by mechanisms different to TSM supporting the proposed higher fitness cost of TSM in P. aeruginosa isolates presenting the exoU− genotype 35 .
Previous studies have reported that oprD mutations alone is the source of non-susceptibility to imipenem, and the mechanisms leading to meropenem resistance are thought to be multifactorial 40 . Although in a previous study an association between oprD defective mutations and the presence of exoU was observed 4 , this scenario was not found in the current study, also in agreement with the lack of association between exoU and imipenem non-susceptibility.
Regarding the presence of β-lactamases, non-conclusive associations were found, despite a higher proportion of bla GES being found in exoU+ isolates and bla IMP in the exoU− isolates. Likewise, Takata et al. did not find differences in the prevalence of bla IMP between exoU+ and exoU− isolates 4 . These data together suggest that the prevalence of specific transferable genes may be more related to the specific prevalence of the gene in the area analysed than with to specific exoenzyme genetic background. Furthermore, GES-type β-lactamases and IMP-metallo-carbapenemases have been widely reported in South America, including Peru 23,41-43 . Overall, the association between the exoU+ genotype and MDR/XDR was shown, despite the presence of β-lactamase, mutations in gyrA and parC, relevant modification in efflux pumps and OprD not being significantly associated with exoU+ isolates. Thus, the MDR/XDR phenotypic basis of the exoU+ genotype remains to be elucidated. One limitation of this study was that not all the mechanisms of resistance were determined, and therefore, other mechanisms might be correlated to MDR and exoU+ genotype.
Overall, these data suggest that exoU+ genotype might be genetically favoured in environments with high antibiotic pressure, such as ICUs 10 . In fact, it has been observed its adaptation to FQ-rich environment 1 . As has been commented above, the exoU+ isolates were more prone to be associated with ICU and burn wards, and subsequently with the most fragile patients of hospital environment. However, no data on background and final outcome of the patients was recorded, therefore lacking data about patient risk factors facilitating exoU+ P. aeruginosa infections, and information about patient mortality. Analysed together, these data agree with previous studies showing that exoU+ isolates were significantly found in man-made environmental sites while the exoS+ isolates were found in natural environmental sites 13,44 .
Regarding the ST detected, ST179, ST235, ST308, ST348, ST357 and ST699 are reported in the Pseudomonas aeruginosa MLST Database as having been previously described on different continents, being therefore widely disseminated 45 . ST235 and ST357 are among the most widespread high-risk clones. The results of the subset of isolates analysed agree with this distribution, being the two most frequently detected STs. Furthermore, the present data suggest a different hospital distribution of both STs. ST235 has been associated with a poor clinical outcome in part due to its high level of antibiotic resistance and the presence of the exoU+ gene 12,29 . Regarding antibiotic resistance, ST235 and ST357 are usually resistant to FQ, aminoglycosides and β-lactams 12,25,46,47 . Likewise, all the isolates belonging to these STs were at least MR, with all ST357 isolates being XDR and displaying resistance to all antibacterial agents tested except colistin. Regarding antibiotic resistance mechanisms, 3 out of 6 isolates belonging to ST235 (isolates 1077, 1078 and 1079) displayed the presence of bla GES and an additional isolate (isolate 1080) possessed bla IMP (Table 6). Similarly, the presence of bla GES and bla IMP was detected in 2 (isolates 1093 and 1105) and 1 (isolate 1089) respectively, of the isolates belonging to the ST357. In addition, all ST235 and ST357 isolates identified showed the presence of QRDR mutations. www.nature.com/scientificreports www.nature.com/scientificreports/ The remaining exoU+ isolates belonged to ST308, ST348, ST759 and ST2726, as well as to the newly identified ST3300 and ST3303. Among these, the presence of exoU has been largely described on the high-risk clone ST308, which usually presents an MDR/XDR phenotype, related to a variety of molecular mechanisms including carbapenemases such as NDM 48,49 . On the other hand, ST641 has been previously reported in Korea in association with antimicrobial resistance and the exoU−/exoS+ genotype 50 . Data of the remaining STs are scarce, with ST179 being the most well characterised. In the present study, ST179 was found in an MDR isolate from a bronchial secretion. Accordingly, this clone has been previously associated with MDR P. aeruginosa, causing chronic respiratory infections in Spanish hospitals 51,52 .
In conclusion, an unusual high number of unrelated clinical isolates of P. aeruginosa showing the exoU and exoS genes concomitantly were found, suggesting the presence of specific pressures which facilitate the stable presence of both genes and highlight their concomitant dissemination in the area studied. The exoU+/exoS+ were associated with MDR and XDR. Furthermore, these isolates showed an enhanced ability to acquire higher levels of FQ resistance, which might be related to lower fitness cost. Rapid diagnostic determination of virulence genotypes and antibiotic resistant profiles as well as continuous surveillance are needed to monitor these high-risk P. aeruginosa isolates.
Material and Methods study area. The Hospital Nacional Cayetano Heredia (HCNH) is a level III-1 hospital with 452 beds (including a total of 24 in the ICUs, of these 6 beds belonging to a Neonatal ICU), which receives patients from around all the country 53,54 . The direct reference area is composed by the districts in the north of the city of Lima, the largest urban area of the capital, in which the population is heterogeneous: urban, rural and marginal urban. In 2011 the HCNH receives 147,642 outpatients, with 17,558 hospital admissions 53 . The Hospital Arzobispo Loayza (HAL) is also considered as a level III-1 hospital accounting for 806 beds (Of these 26 presents in ICUs) 55 . In addition, HAL acts as reference center for burn patients 56 . Its referral area comprises districts in the center of Lima, attending also population of other Lima districts and Peruvian areas, with a heterogeneous population: urban and urban marginal. In 2011 the HAL receives 218,123 outpatient visits (10.3% from outside of Lima), with 29,158 hospital admissions 57 . In 2014 the population of Lima was reportedly 9,752,000 inhabitants, 2,475,432 and 1,796,112 living in the north and center districts respectively 58