Molecular Mechanisms of Resistance to Ceftazidime/Avibactam in Clinical Isolates of Enterobacterales and Pseudomonas aeruginosa in Latin American Hospitals

ABSTRACT Ceftazidime-avibactam (CZA) is the combination of a third-generation cephalosporin and a new non-β-lactam β-lactamase inhibitor capable of inactivating class A, C, and some D β-lactamases. From a collection of 2,727 clinical isolates of Enterobacterales (n = 2,235) and P. aeruginosa (n = 492) that were collected between 2016 and 2017 from five Latin American countries, we investigated the molecular resistance mechanisms to CZA of 127 (18/2,235 [0.8%] Enterobacterales and 109/492 [22.1%] P. aeruginosa). First, by qPCR for the presence of genes encoding KPC, NDM, VIM, IMP, OXA-48-like, and SPM-1 carbapenemases, and second, by whole-genome sequencing (WGS). From the CZA-resistant isolates, MBL-encoding genes were detected in all 18 Enterobacterales and 42/109 P. aeruginosa isolates, explaining their resistant phenotype. Resistant isolates that yielded a negative qPCR result for any of the MBL encoding genes were subjected to WGS. The WGS analysis of the 67 remaining P. aeruginosa isolates showed mutations in genes previously associated with reduced susceptibility to CZA, such as those involved in the MexAB-OprM efflux pump and AmpC (PDC) hyperproduction, PoxB (blaOXA-50-like), FtsI (PBP3), DacB (PBP4), and OprD. The results presented here offer a snapshot of the molecular epidemiological landscape for CZA resistance before the introduction of this antibiotic into the Latin American market. Therefore, these results serve as a valuable comparison tool to trace the evolution of the resistance to CZA in this carbapenemase-endemic geographical region. IMPORTANCE In this manuscript, we determine the molecular mechanisms of ceftazidime-avibactam resistance in Enterobacterales and P. aeruginosa isolates from five Latin American countries. Our results reveal a low rate of resistance to ceftazidime-avibactam among Enterobacterales; in contrast, resistance in P. aeruginosa has proven to be more complex, as it might involve multiple known and possibly unknown resistance mechanisms.

bacteria has radically decreased the effectiveness of last-generation b-lactams, including cephalosporins, carbapenems, and therapeutic combinations with b-lactamase inhibitors. The accumulation of resistance mechanisms to b-lactams and some other antibiotic families significantly hinders the treatment of infections, and obliges the use of less effective and more toxic antibiotics such as colistin and aminoglycosides (1,2).
The most effective resistance mechanism to carbapenems in Gram-negative pathogens is the production of carbapenemases. In Enterobacterales, many class A b-lactamase-encoding genes can yield a carbapenem resistant phenotype. However, bla KPC-2 and bla KPC-3 are the most common transmissible genes circulating worldwide and, notably, are endemic to some geographical areas such Latin America (3,4). In P. aeruginosa, resistance to carbapenems can be achieved either by the hyperproduction of the chromosomal cephalosporinase AmpC or by the production of acquired carbapenemases, particularly of class B metallo-b-lactamases (MBL) such as VIM-2. In addition, nonenzymatic mechanisms such as the modification or inactivation of the porin OprD, or the upregulation of different chromosomally encoded efflux pumps, are also common (5)(6)(7).
In the last few years, novel b-lactams/b-lactamase inhibitor combinations are available for the treatment of infections caused by carbapenem resistant Enterobacterales and carbapenem resistant P. aeruginosa (8). Among them, ceftazidime-avibactam (CZA) is the combination of an extended-spectrum cephalosporin and a diazabicyclooctane (DBO)-based, non-b-lactam b-lactamase inhibitor. Avibactam is capable of inhibiting the majority of KPC enzymes, including the most wide-spread types, KPC-2 and KPC-3, in addition to other class A b-lactamases; class C cephalosporinases; and to a various degree class D b-lactamases, like some members of the OXA-48 family. However, avibactam cannot inhibit any class B MBL (9).
Previously, our group determined the rates of susceptibility to CZA and other relevant antibiotics of clinical Enterobacterales isolates collected prior to the introduction of this antibiotic into the clinical practice in Latin America. The resistance rate found in that study was 4.2% (23). Herein, we reassess the phenotypic resistance to CZA of the 94 CZA-resistant Enterobacterales strains identified in that previous study; describe the phenotypic resistance rates to CZA of 492 P. aeruginosa clinical isolates collected between 2016 and 2017; and explore the molecular mechanisms leading to CZA resistance in these clinical isolates using whole-genome sequencing (WGS).

RESULTS
Molecular characterization of CZA-resistant Enterobacterales. To compare the data previously published for the Enterobacterales collection with the new data on the P. aeruginosa isolates from this study, we checked the susceptibility to CZA of 94 isolates previously identified as CZA-resistant. However, after analyzing together the MIC data with Etest, only 18 isolates were confirmed to be truly CZA-resistant. Therefore, the updated CZA resistance rate of this collection of Enterobacterales is 0.8% (18/2,235). Of interest, all 18 CZA-resistant isolates were collected in Colombia, at different times, from nine medical centers located in nine cities. For Enterobacterales, we expanded the battery of tests performed before, adding the RAPIDEC Carba-NP assay to detect carbapenemase activity, and qPCR to confirm the presence of at least one MBLencoding gene in these isolates (Table 1). Furthermore, three isolates of K. pneumoniae and one Enterobacter cloacae complex co-harboring bla NDM and bla KPC were detected. Since resistance to CZA is explained by the presence of at least one MBL-encoding gene in CZA-resistant Enterobacterales, WGS was not performed on any of these isolates.
Antibiotic susceptibility and molecular characterization of CZA-resistant P. aeruginosa. The distribution of the 492 isolates of P. aeruginosa per country is shown in Table 2. Overall, 22.1% (109/492, MIC 50 4/4 mg/L, MIC 90 64/4 mg/L) of the isolates were resistant to CZA. In addition, complete MIC data are presented in Table S1.
All CZA-resistant P. aeruginosa isolates were then subjected to RAPIDEC Carba-NP test and qPCR. These tests found 42 isolates (38.5%) with MBL (three for bla IMP ; 31 for bla VIM ; one for bla SPM-1 ; and seven carrying a combination of bla KPC and bla VIM ), and eight positives for bla KPC . However, 59 did not carry any carbapenemases (Table 3). Notably, the only isolate harboring bla SPM-1 yielded a negative result in the RAPIDEC Carba-NP assay.
WGS analysis of P. aeruginosa isolates resistant to CZA and associated resistance genes. A total of 67 P. aeruginosa genomes were sequenced. This number corresponds to the 59 isolates that yielded negative results for the multiplex qPCR and eight additional isolates that tested positive for the presence of bla KPC . Due to unexpected low sequence coverage (,30Â), we excluded six samples from the subsequent analysis (five isolates negative for any carbapenemase gene and one positive for bla KPC ). The remaining 61 samples showed quality values over 90%. We obtained between 110 to 137 contigs per isolate sequenced, with a length of the assemblies between 6.3 to 7.2 Kb and, a GC content ranging from 65.8% to 66.5%. Sequencing quality data are presented in Table S2.
WGS analysis revealed 23 known sequence types (STs), and five new STs, as shown in Fig. 1. Relevant STs found included ST111 (n = 1) and ST308 (n = 1) from Colombia; ST357 (n = 1) from Chile; and ST309 (n = 6) were found in four isolates from Mexico, one from Colombia, and one from Chile. Clonal dissemination was observed among some isolates: ST575 (n = 9) was only reported in Mexico; ST235 (n = 16) in Colombia, Mexico, Brazil, and Argentina; and ST244 mainly in Argentina.  Confirming their species identity, sequence analysis of the P. aeruginosa genomes showed that all of them carried bla AmpC and bla OXA-50-like . From the 61 genomes analyzed, 17 (27.9%) harbored bla OXA-2 : 11 isolates belonging to the ST235 from Mexico and Colombia, four isolates with ST309 from Mexico, and two belonging to the ST308 and ST261 isolated from Colombia ( Fig. 1). However, none of the evaluated isolates harbored mutations in bla OXA-2 , including duplication in the bla OXA-2 , which encodes for OXA-539.
Three isolates from Mexico belonging to the ST235 and one belonging to the ST30 harbored bla GES-19 . Interestingly, one isolate belonging to the ST309 from Mexico harbored bla GES- 19 and bla GES-20 in tandem. Also, one isolate from Argentina and one from Chile, were found to harbor bla PER-1 and bla PER-3 , respectively. All sequenced isolates harboring bla KPC-2 (n = 7) were isolated in Colombia and belonged to the high-risk clone ST235. To note, none of these isolates showed mutations in bla KPC-2 .
To explore in detail the molecular mechanisms previously associated with resistance to CZA in these P. aeruginosa clinical isolates, we analyzed a variety of genes for any mutation that could lead to overexpression or repression of a particular gene, or to amino acid substitutions that could change the activity of the protein. These genes include b-lactamase encoding genes (e.g., bla PDC ) and their regulator genes (bla AmpD , bla AmpR , bla AmpG ); genes encoding the multidrug efflux MexA-B, and its regulators (MexR, NalC and NalD); (ftsI, and dacB encoding PBP3 and PBP4, respectively); creD, which encodes a predicted inner membrane protein part of the conserved two-component regulatory system CreBC (24); and genes involved in pathogenesis like DnaJ, DnaK, and ATP-dependent Clp protease proteins (13,(25)(26)(27).
Specifically, predicted substitutions in AmpG, DnaJ, DnaK, and ATP-dependent Clp protease proteins were not found. The proteins that had substitutions in most isolates were PDC, PoxB/OXA-50-like, NalC, and CreD. Most of the proteins had multiple substitutions, except peptidases S41, PBP3/FtsI and NalD, which had only one substitution in some isolates (Table S3). Substitutions in MexAB-OprM regulator proteins, most frequently a G71E change in NalC (77%) and a V126E substitution in MexR (47.5%) were observed. Mutations leading to substitutions in PBP3, PoxB, and the PDC/AmpC system were detected in 9.8%, 95.1%, and 82% of the P. aeruginosa CZA-resistant isolates, respectively. Only six isolates had the substitution N117S in PBP3, all of them belonging to the ST309 from Mexico (four), Colombia (one), and Chile (one) ( Table S3).

DISCUSSION
In a previous study, we evaluated the in vitro activity of CZA against a set of 2,252 clinical isolates of Enterobacterales in Latin America, finding that 4.2% were resistant (23). However, combined phenotypic tests performed in this study confirmed the CZAresistant phenotype of only 18/94 isolates. Therefore, the updated resistance rate to CZA of this group of Enterobacterales is 0.8%. Additionally, we analyzed the susceptibility to CZA of a set of 492 clinical isolates of P. aeruginosa collected during the same time period (2016 to 2017) in the same five Latin American countries. Finally, we determined the molecular mechanisms leading to CZA resistance in these isolates by WGS.
Several molecular mechanisms leading to decrease susceptibility to CZA have been described in P. aeruginosa. Among them, specific amino acid substitutions in some b-lactamases, including KPC and SHV have been associated with resistance to CZA (11). In particular, the D179Y substitution in the X-loop of KPC-3, and in other KPC variants, confer resistance to CZA. Of note, this mechanism was reported in a P. aeruginosa isolate from Chile before this antibiotic was clinically available in this country (28). Interestingly, all sequenced P. aeruginosa isolates that carried bla KPC-2 retrieved in Colombia belonged to ST235. This ST has been associated with the disseminations of bla KPC-2 in Colombia (2). As we did not evidence any mutations in bla KPC-2 , CZA-resistance is most probably caused by other mechanisms. All of these seven strains (58PAE to 63PAE and 65PAE in Table S3) have multiple mutations in several genes, including in ampR leading to the substitutions G283E, M288R in AmpR, and mutated ampG, producing the variant A583T. The association of these mutations with CZA resistance is yet to be determined. Moreover, six out of seven isolates showed mutations in nalD (coding for the MexAB-oprM regulator), which could lead to decreased susceptibility to CZA as previously reported (29,30) (Table S3).
Regarding the molecular epidemiology, WGS analysis revealed that some of the CZA-resistant P. aeruginosa isolates belonged to ST235 (n = 16), ST244 (n = 6), and ST111 (n = 1). These STs have been considered as high-risk clones (31,32). Furthermore, ST235 and ST111 are multidrug resistant (MDR) clones disseminated worldwide and linked to the expression of VIM-2 (2). Sixteen of the sequenced isolates belonging to ST235 did not harbor any bla VIM gene but all harbored bla KPC . A surveillance study of P. aeruginosa performed in Colombia found that ST111 is a common host of bla VIM-2 , whereas ST235 is associated with bla KPC-2 , as aforementioned (33). Additionally, an isolate that carried bla SPM-1 belonged to ST277, which is a ST commonly associated with the dissemination of bla SPM-1 in Brazil (12).
Extended-spectrum b-lactamases (ESBL) such as PER and GES have also been associated with resistance to CZA via biochemical mechanism conferring a weaker inhibitory potency of avibactam to these enzymes (34). This kinetic feature, possibly combined with the lower permeability of P. aeruginosa, can effectively decreased the susceptibility to CZA (9,13,34,35). In our study, one P. aeruginosa isolate from Argentina (ST179) and one isolate from Chile (ST309) harbored bla PER-1 and bla PER-3 , respectively. In addition, five isolates from Mexico carried bla GES-19 , three of them were ST235 and the other two were ST309. In Mexico, a high prevalence of the ESBL GES-19 and the carbapenemase GES-20 have been reported as the most prevalent in P. aeruginosa (36). Moreover, it has been reported that the presence of the ESBL-encoding genes bla GES-19 and bla GES-26 in tandem is associated with resistance to all b-lactams, including CZA (21). Importantly, in the present study one of the P. aeruginosa isolates belonging to ST309 showed a similar feature, where bla GES- 19 and bla GES-20 were found in tandem, which might explain the resistance to CZA. Dissemination of P. aeruginosa isolates harboring either bla PER or bla GES genes is worrisome, as production of these enzymes compromise the efficacy of the latest anti-pseudomonal drugs, CZA and ceftolozane-tazobactam (14,37).
A recent study by Fraile-Ribot et al. found that the duplication of the residue D149 in OXA-2 led to resistance to CZA in vivo (8). This new variant of OXA-2, called OXA-539 was reported for the first time in a P. aeruginosa isolate resistant to CZA belonging to ST235, from a patient with a susceptible isolate who was previously treated with CZA (8). In our analysis, 17 P. aeruginosa isolates carried OXA-2, 11 of them belonging to ST235 but none of them had the D149 duplication. Worth noting, all P. aeruginosa resistant to CZA and harboring bla OXA-2 were exclusively recovered from Mexico and Colombia.
Several enzymes of class D, including PoxB (OXA-50-like), which is encoded in the chromosome of all P. aeruginosa strains, are not efficiently inhibited by DBOs (38). Compared to the PoxB encoded in the reference strain P. aeruginosa PAO1, multiple substitutions in PoxB were found in our isolates. However, there is no evidence that these mutations can lead to resistance to CZA. On the contrary, Castanheira et al. described substitutions in PoxB in both susceptible and resistant isolates, suggesting that these changes are not directly leading to CZA-resistance (25).
Although CZA shows potent inhibitory activity against PDC (AmpC) of P. aeruginosa, mutations in bla PDC conferring resistance to CZA have been reported (39). Here, we found 14 different PDC variants, being PDC-3, PDC-35, and PDC-1 the most frequent (Table S3). However, these variants have not been associated with a particular antimicrobial resistance pattern in previous studies. Moreover, previous investigations have suggested that amino acid substitutions in the PDC enzyme are unlikely to be the main mechanism conferring resistance to CZA, because a correlation between the PDC enzyme variations and the MIC has not been detected (40). However, the recent emergence of P. aeruginosa clinical isolates overexpressing variants of PDC is worrisome and may compromise the efficacy of CZA (40). Indeed, the E247K, G183D, T96I, and DG229 to E247 substitutions and deletions appear to perform a 2-fold effect on the catalytic cycle of PDC, allowing to evade avibactam inhibition, while hydrolyzing ceftazidime with enhanced efficiency (40). More biochemical studies are needed to elucidate the relation between the PDC variants identified in this study and CZA-resistance in P. aeruginosa.
As previously mentioned, changes in PBPs can lead to CZA resistance. For instance, FtsI (PBP3) of P. aeruginosa, is the PBP to which many b-lactams, including monobactams and some cephalosporins, have the highest affinity for. FtsI is the primary target of ceftazidime, however, avibactam is also known to covalently bind to the PBPs of P. aeruginosa (1). The FtsI variants R504C and P527S have been strongly associated with reduced susceptibility to different types of b-lactams, including ceftazidime (5). We did not find these mutations in our isolates. However, six sequenced P. aeruginosa isolates showed the same FtsI variant, N117S, which, has not been associated to CZA resistance, and given its location within the protein, an effect on CZA-resistance is unlikely. Interestingly, all the strains harboring the N17S variant of FtsI belonged to the ST309, which has been described in serious infections involving MDR and XDR P. aeruginosa strains. Furthermore, all six isolates were recovered from different geographical locations Mexico, Colombia, and Chile, suggesting that the geographic distribution of ST309 is widespread (21).
A study from Castanheira et al. showed that MexAB-OprM efflux system overexpression was significantly associated with CZA resistance, alone or in combination with alterations or disruptions in other genes (25). Furthermore, it has been shown that disruption of MexR, a negative regulator of MexAB-OprM, leads to high expression of the MexAB-OprM efflux pump slightly raising the MIC of CZA (41). In our analysis, nine isolates belonging to the ST575 from Mexico showed altered versions of MexR. Additionally, 18 isolates (5 from Argentina [ST244 and ST179], 11 from Colombia [ST235 and ST3963], 1 from Chile [ST357], and 1 from Brazil [ST235]) had mutations, framework-shifts, or alterations in the NalD, a repressor of MexAB. Mutations in NalD have been associated with hyperexpression of MexAB and therefore, resistance of all b-lactams (30).
Regarding the Enterobacterales, we determined that the presence of at least one MBL-encoding gene in all evaluated isolates could be the underlying molecular mechanism leading to CZA-resistance. The presence of MBL-encoding genes in CZA-resistant Enterobacterales has been frequently reported in the United States, countries of the Asian-Pacific region, and Europe (9,42,43).
Interestingly, all CZA-resistant Enterobacterales were isolated in Colombia, where KPC-enzymes are considered endemic (44). Although specific amino acid substitutions in the X-loop of KPC leading to CZA-resistance in Enterobacterales have been reported in several countries, we did not find isolates harboring bla KPC without an MBL-encoding gene. Conversely, the prevalence of Enterobacterales carrying MBL, especially NDM, either alone or in combination with a serine carbapenemase has increased in recent years in this country (45). Exemplary for this observation, we found four isolates from Colombia harboring both bla KPC and bla NDM (1,10,42).
Conclusions. By the time of the collection of these isolates, a low rate of resistance to CZA was found among Enterobacterales in the Latin American countries that participated in this study. In this analysis, we demonstrated that the most common mechanism of resistance in Enterobacterales was the production of MBLs. In contrast, resistance to CZA in P. aeruginosa has proven to be more complex, as it might involve multiple known and possibly unknown resistance mechanisms.
Our study has many limitations. Due to budget restrictions, we could only sequence some of the CZA-resistant isolates and none of the CZA-susceptible ones. This impeded us to have the complete molecular snapshot of all Enterobacterales and P. aeruginosa isolates. Consequently, we are only reporting known mechanisms of reduced susceptibility to CZA in these isolates. More studies are needed to investigate emerging mechanisms of resistance to CZA. Nevertheless, as these isolates were collected before the clinical use of CZA in Latin America, the results presented here offer a valuable tool for upcoming comparisons with isolates of Enterobacterales and P. aeruginosa recovered after its introduction in this region. These studies will delineate the evolutionary path of the CZA-resistance and how its use in the clinical practice affects the epidemiology of these MDR pathogens. The knowledge of the evolution of resistance to last-resort antibiotics such as CZA in clinical isolates will help to understand the role of selective pressure in different scenarios.
Ethical approval. The protocol was approved by the ethics committee of Universidad El Bosque, under act #018-2020. Collection of the microbiological isolates was part of the regular diagnostic process, as established by each of the participating health care institutions.

MATERIALS AND METHODS
Susceptibility testing and detection of carbapenemases. Resistance to CZA was confirmed by MICs determined by broth microdilution method using customized Sensititer plates (Trek Diagnostic Systems, Thermo Fisher Scientific, UK) following the manufacturer's recommendations and, Etest (bioMérieux, Marcy l'Etoile, France). Results were interpreted according to the current guidelines of the Clinical and Laboratory Standards Institute (CLSI) (46). Presence of carbapenemases in CZA-resistant Enterobacterales and P. aeruginosa isolates was initially screened by RAPIDEC Carba-NP Assay (bioMérieux, Marcy-l' Etoile, France) (47), followed by qPCR targeted to the bla KPC , bla NDM , bla VIM , bla IMP , bla oxa-48-like , and bla SPM-1 genes. The reference strains Escherichia coli ATCC 25922, K. pneumoniae ATCC 700603 and P. aeruginosa ATCC 27853 were used as the quality control strain, as per CLSI recommendations (46).

Ceftazidime/Avibactam Molecular Resistance Mechanisms mSphere
Data availability. The genome sequencing data are publicly available at NCBI GenBank under the BioProject accession number PRJNA729968.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.