The Role of Uniform Meropenem Usage in Acinetobacter baumannii Clone Replacement

The dominant carbapenem resistant Acinetobacter baumannii harboring blaOXA-23-like carbapenemase was replaced by blaOXA-40-like carriers in a Hungarian tertiary-care center with high meropenem but relatively low imipenem use. We hypothesized that alterations in antibiotic consumption may have contributed to this switch. Our workgroup previous study examined the relation between resistance spiral and the antibiotic consumption, and the results suggest that the antibiotic usage provoked the increasing resistance in case of A. baumannii. We aimed at measuring the activity of imipenem and meropenem to compare the selection pressure exerted by the different carbapenems in time-kill assays. Strain replacement was confirmed by whole genome sequencing, core-genome multilocus sequence typing (cgMLST), and resistome analysis. Based on results of the time-kill assays, we found a significant difference between two different sequence-types (STs) in case of meropenem, but not in case of imipenem susceptibility. The newly emerged ST636 and ST492 had increased resistance level against meropenem compared to the previously dominant ST2 and ST49. On the other hand, the imipenem and colistin resistance profiles were similar. These results suggest, that the uniform meropenem usage may have contributed to A. baumannii strain replacement in our setting.

Our working group first investigated the molecular epidemiology of carbapenem resistant A. baumannii (CRAb) and the effect of carbapenem consumption on its distribution published in a previous study, where six clones were identified of which five were carbapenem resistant due to bla OXA-23-like genes (78.1% of all tested isolates), while the bla OXA-24/40-like gene was found in two sporadic isolates (1.2%) [17]. The present study investigates the changes in the occurrence of carbapenem resistance genes between 2012 and 2017 in relation to carbapenem usage patterns and characterizes representative isolates from the former (2010) study and from this period to uncover potential mechanisms in the background of strain dynamics.
Our working group first investigated the molecular epidemiology of carbapenem resistant A. baumannii (CRAb) and the effect of carbapenem consumption on its distribution published in a previous study, where six clones were identified of which five were carbapenem resistant due to blaOXA-23-like genes (78.1% of all tested isolates), while the blaOXA-24/40-like gene was found in two sporadic isolates (1.2%) [17]. The present study investigates the changes in the occurrence of carbapenem resistance genes between 2012 and 2017 in relation to carbapenem usage patterns and characterizes representative isolates from the former (2010) study and from this period to uncover potential mechanisms in the background of strain dynamics.

Antibiotic Consumption
Carbapenem consumption increased more or less steadily between 2005 and 2018 (Mann-Kendall tau = 0.684, p < 0.0001; Figure 1), meropenem usage increased four times between the two endpoints; mean usage was 0.64 defined daily dose per 100 occupied bed-days (DDD/100 OBDs) in 2007 while 2.69 DDD/100 OBDs in 2017. Imipenem consumption was similar in 2007 (mean usage 0.48 DDD/100 OBDs), but it did not change dramatically during the study period (mean usage 0.80 DDD/100 OBDs in 2017), as shown in Figure 1.

Resistance Phenotype
Antibiotic resistance was over 75% to all tested drugs except colistin ( Figure S1). Carbapenem resistance steadily increased during the study (Mann-Kendall tau = 0.610, p < 0.0001). Over 90% of the isolates have been CRAb. Resistance to ciprofloxacin and gentamicin remained very high throughout the study, while amikacin and tobramycin resistance decreased in 2016-17 significantly (p < 0.0001). Colistin resistance was sporadic throughout.

Resistance Phenotype
Antibiotic resistance was over 75% to all tested drugs except colistin ( Figure S1). Carbapenem resistance steadily increased during the study (Mann-Kendall tau = 0.610, p < 0.0001). Over 90% of the isolates have been CRAb. Resistance to ciprofloxacin and gentamicin remained very high throughout the study, while amikacin and tobramycin resistance decreased in 2016-17 significantly (p < 0.0001). Colistin resistance wassporadic throughout.

Prevalence of the Resistance Genes
Prevalence of the aminoglycoside resistance genes decreased during the examination period ( Figure S2). The dominant aph(3 )-VIa phosphotransferase gene prevalence decreased to 6.2% from 90.6% (p < 0.0001). The aac(6 )-Ib acetyl-transferase and the armA methyl-transferase show a similar pattern (p < 0.0001). All of the isolates contained the bla OXA-51-like genes proving that all investigated isolates were A. baumannii ( Figure S3). The dominant carbapenemase gene family between 2010 and 2014 was the bla OXA-23-like ( Figure S3), but occurrence of the gene decreased significantly in 2016 and 2017 (p < 0.001). The prevalence of bla OXA-24/40-like carbapenemases showed an opposite trend, while in the beginning of the examination period the gene family was almost totally absent, its prevalence increased significantly in 2016 (p < 0.001), and it was even higher in 2017 (p < 0.001).

Whole Genome Sequencing (WGS)
In 2010, bla OXA-23-like carbapenemases proved to be bla OXA-23 in all sequenced strains representing all major pulsotypes, which belonged to ST1, ST2, and ST45. Only two isolates of the same minor pulsotype carried the bla OXA-24/40-like gene [17]; one of which proved to be ST79 and a bla OXA-24 carrier by WGS. All representative isolates from 2017 harboring only bla OXA-23-like acquired carbapenemase gene belonged to ST2. Representative isolates of the more abundant bla OXA-24/40-like carriers (75.0% in 2016 and 92.3% in 2017 of the total isolates), in turn, belonged to ST636 and ST492, all harboring bla OXA-72 or, in case of some ST636 isolates, bla OXA-72 together with bla OXA-23 ( Table 1). The genetic environment of the bla OXA-23-like gene in these latter isolates was identical to that found in ST45 strains from 2010, also including the gene aph(3 )-Via [17]. No other types of carbapenemase genes were found in the sequenced genomes.   (6 ) The WGS confirmed acquisition of the bla OXA-23 carbapenemase at least two times in 2010, in case of ST1 and ST45. ST636 isolates were genetically highly uniform, distance based on allele presence is ≤4 alleles, while ST2 is more diverse, three of four sequenced isolates belonged to a subgroup distinct from the 2010 ST2 isolates by 20 alleles (Figure 2).

Change-Point Analysis
CRAb incidence exhibited nine changepoints (Figure 4). Out of the antibiotic con-  The WGS confirmed acquisition of the blaOXA-23 carbapenemase at least two times in 2010, in case of ST1 and ST45. ST636 isolates were genetically highly uniform, distance based on allele presence is ≤4 alleles, while ST2 is more diverse, three of four sequenced isolates belonged to a subgroup distinct from the 2010 ST2 isolates by 20 alleles (Figure 2).

Time-Kill Assays
The examined isolates were resistant against meropenem and imipenem in clinical aspects, the minimum inhibitor concentrations (MICs) were >32 mg/L, except the only blaOXA-51-like (blaOXA-69 or the blaOXA-66) harboring isolates (MIC = 2 mg/L in both cases). Against

Discussion
The role of the selective pressure exerted by antibiotics as a provoker of resistance is well known. However, the response to selection pressure may be versatile and evolves over time. Hospital ecology of multiresistant bacteria is characterized by emergence and receding of different bacterial strains, is shaped by virulence, survival ability, and transmissibility of these strains and is fundamentally influenced by the selection pressure exerted by antibiotic use; shown to provoke spread and persistence of resistant strains by a number of studies. For example, the increasing amikacin usage provoked appearance of amikacin resistant but tobramycin sensitive Acinetobacter, while reducing amikacin consumption decreased the number of the resistant isolates [18]. Similar findings were reported between the intravenous fluoroquinolone usage and the fluoroquinolone-resistant A. baumannii [19].
In this study, the patterns of carbapenemase gene carriage and results of the WGS strongly suggest a strain switch, i.e., the newly appeared ST636 and ST492 CRAb strains carrying blaOXA-72 carbapenemase gene replaced the ST2 and ST45 blaOXA-23 carriers during the period 2014-2015. Similar emergence and spread of blaOXA-24/40-like carrier strains have been reported from several countries [20][21][22][23][24][25][26][27]. The strain replacement reported hereby is in line with the well-documented evolution of carbapenem resistance in A. baumannii from overexpressed blaOXA-51-like enzymes in sporadic isolates through blaOXA-23-like and blaOXA-58-like enzymes with potent carbapenem hydrolyzing activity carried by pandemic clones to emerging clones carrying blaOXA-24/40-like enzymes with even more efficient hydrolytic activity. This evolution towards more efficient hydrolysis of carbapenems is linked to increasing use of carbapenems in many countries, as documented earlier in this setting [17] and elsewhere [28][29][30][31]. This relationship is the inherent part of the Gram-negative resistance spiral, as carbapenem usage increases in response to the spread of extended spectrum β-

Discussion
The role of the selective pressure exerted by antibiotics as a provoker of resistance is well known. However, the response to selection pressure may be versatile and evolves over time. Hospital ecology of multiresistant bacteria is characterized by emergence and receding of different bacterial strains, is shaped by virulence, survival ability, and transmissibility of these strains and is fundamentally influenced by the selection pressure exerted by antibiotic use; shown to provoke spread and persistence of resistant strains by a number of studies. For example, the increasing amikacin usage provoked appearance of amikacin resistant but tobramycin sensitive Acinetobacter, while reducing amikacin consumption decreased the number of the resistant isolates [18]. Similar findings were reported between the intravenous fluoroquinolone usage and the fluoroquinolone-resistant A. baumannii [19].
In this study, the patterns of carbapenemase gene carriage and results of the WGS strongly suggest a strain switch, i.e., the newly appeared ST636 and ST492 CRAb strains carrying bla OXA-72 carbapenemase gene replaced the ST2 and ST45 bla OXA-23 carriers during the period 2014-2015. Similar emergence and spread of bla OXA-24/40-like carrier strains have been reported from several countries [20][21][22][23][24][25][26][27]. The strain replacement reported hereby is in line with the well-documented evolution of carbapenem resistance in A. baumannii from overexpressed bla OXA-51-like enzymes in sporadic isolates through bla OXA-23-like and bla OXA-58-like enzymes with potent carbapenem hydrolyzing activity carried by pandemic clones to emerging clones carrying bla OXA-24/40-like enzymes with even more efficient hydrolytic activity. This evolution towards more efficient hydrolysis of carbapenems is linked to increasing use of carbapenems in many countries, as documented earlier in this setting [17] and elsewhere [28][29][30][31]. This relationship is the inherent part of the Gram-negative resistance spiral, as carbapenem usage increases in response to the spread of extended spectrum β-lactamase (ESBL) producing Enterobacterales in our setting [32]. The link between the strain switch and carbapenem use is further supported by the coincident changepoints.
Sequenced ST636 isolates were highly similar, corresponding to clonal spread of these isolates within this setting, which may be an advent of the emergence of a new high-risk clone. Some isolates within this clone acquired another carbapenemase (bla OXA-23 ) most probably from ST45, one of the earlier resident clones, suggesting that ST636 is currently evolving in the setting.
Investigating the pattern of antibiotic consumption as a potential driver of this change, notable differences were found between killing kinetics of imipenem and meropenem. While meropenem killed the novel strains ST492 and especially ST636 markedly less efficiently at higher concentrations, the effect of imipenem was comparable in case of all examined strains. As monthly meropenem use exceeds imipenem use 5-10-fold, the difference in meropenem resistance levels may have provided the selective advantage fueling the emergence of bla OXA-72 producing strains, especially ST636 with exceedingly high meropenem resistance. This assumption may similarly be applied in other settings; in Colombia Villalobos et al. reported a high prevalence of CRAb together with strong preference for meropenem over imipenem in Colombia [33]. This is in line with the well-documented process that the more resistant strain is expected to outcompete the less resistant provided that the selection pressure is high enough to exceed the fitness cost of resistance [34], i.e., if it has a fitness advantage in a situation with high antibiotic pressure. This is elegantly demonstrated in case of trimethoprim and Escherichia coli by Baym et al. in an ingenious experimental setting [35]. It was also shown that transmission between patients is also strongly linked to drug use [36,37].
This principal precept is also applicable to small differences in fitness in the long run. Though the less efficient killing found in the present case may provide only a small advantage at clinically attainable carbapenem concentrations, such small competitive advantages may be important in selection between resistant strains as well as in shaping competitive interactions. For instance, in case of subclones of E. coli ST131, the success of pandemic H30Rx clones with very high fluoroquinolone resistance over other resistant subclones was linked to fluoroquinolone use [38] or methicillin-resistant Staphylococcus aureus (MRSA) strain dynamics was driven by drug use patterns [39]. Thus, it may be assumed that the problem leading to and maintaining the problem of CRAb lies not only in overuse, but also in imbalanced consumption of carbapenems, i.e., the preference of meropenem over imipenem as empirical as well as targeted therapy. This is further supported by the acquisition of another potent carbapenemase by the already highly resistant ST636 and the extremely high level of meropenem resistance (but not of imipenem resistance) in these isolates, pointing out that strain evolution in this setting is driven towards even higher level of carbapenem resistance. Thus, even strains with clinically relevant high-level resistance may be outcompeted by more resistant novel clones under high selection pressure and thus attaining higher resistance levels may be evolutionarily advantageous even for highly resistant strains with significant clinical resistance.
Antibiotic stewardship efforts generally aim at reducing consumption of a drug family, considering individual drugs with the same spectra within this family equivalent. The present study draws attention to the fact that related compounds may not be equal, and one may favor spread of certain resistant strains and/or fuel evolution of higher resistance levels in already resistant strains. The other way around, versatile use of individual compounds within a drug group may oppose spread of resistance by decreasing uniform selection pressure. When total drug restriction is not feasible, such an approach may be advantageous.  [17]. Susceptibility to amikacin, tobramycin, gentamicin, colistin, imipenem, meropenem, and ciprofloxacin were determined using the European Committee on Antimicrobial Susceptibility Testing (EUCAST) guidelines. Changes in susceptibility were analyzed by means of chi-square test.

Antibiotic Consumption and Changepoint Analysis
We collected monthly antibiotic consumption data between 2005 and 2017 in defined daily dose (DDD) per 100 occupied bed days (OBDs). Trend analysis was performed using a Mann-Kendall test in R statistical environment (Kendall package, MannKendall test). Timeseries of CRAb incidence density per 100 bed-days and of consumption of different antibiotic groups were subjected to changepoint analysis in R statistical environment (changepoint package; changepoints in mean). To identify drug groups possibly involved, the changepoints of the CRAb incidence density was matched to changepoints of the different drug use series (coamoxiclav, third generation cephalosporins, carbapenems, fluoroquinolones, aminoglycosides, and colistin) to find temporally close changepoint patterns.

Whole Genome Sequencing
For whole genome sequencing, isolates representing all pulsotypes (A1; A2; B; C1; C2; D and those representing unique isolates) of the 2010/11 study (n = 9) were chosen. In this study, the isolates from 2017 (n = 14) were selected based on their differences in the resistance phenotype (i.e., susceptible to all investigated antibiotics, amikacin resistant but tobramycin susceptible; susceptible only to gentamicin and colistin, susceptible only to ciprofloxacin and colistin, and resistant against all investigated antibiotics except colistin).

Time-Kill Assay
In time-kill assays the effect of meropenem, imipenem, and colistin were investigated against 12 carbapenem resistant A. baumannii isolates representing all major STs (ST2; ST636, ST1, ST492). The starting inoculum was 10 6 CFU/mL, the drug concentrations were 16, 128-1024 mg/L in case of carbapenems, and 2-32 mg/L in case of colistin in Mueller-Hinton broth medium (Lab M Limited, Heywood, UK). In a smaller subset of strains (n = 4) the meropenem and imipenem time-kill assays were also performed at lower concentrations (16 and 32 mg/L) as well. All tubes were plated at 0, 2, 4, 6, 8, 10, 12, and 24 h to solid Mueller-Hinton media (Lab M Limited, Heywood, UK). The colonies were counted after 24 h of incubation at 37 • C. Killing rates (k) were determined: N t = N 0 × e −kt , where N t is the number of viable bacteria colonies at time t, N 0 is the number of viable bacteria colonies in the initial inoculum, k is the killing rate, and t is the incubation time. Negative and positive k values indicate growth and killing, respectively. Bactericidal activity was defined as a 99.9% reduction of the starting inoculum. All tests were performed at least twice.
Supplementary Materials: The following are available online at https://www.mdpi.com/2079-638 2/10/2/127/s1, Figure S1: Results of the Kirby-Bauer disk diffusion tests, based on EUCAST recommendation. Figure S2: The prevalence of aminoglycoside resistance genes among the A. baumannii isolates by year. Figure S3: The prevalence of carbapenemases among the A. baumannii by years.