Polyclonal Endemicity of Carbapenemase-Producing Klebsiella pneumoniae in ICUs of a Greek Tertiary Care Hospital

Carbapenemase-producing Klebsiella pneumoniae (CPKP) emerged in Greece in 2002 and became endemic thereafter. Driven by a notable variability in the phenotypic testing results for carbapenemase production in K. pneumoniae isolates from the intensive care units (ICUs) of our hospital, we performed a study to assess the molecular epidemiology of CPKP isolated between 2016 and 2019 using pulse-field gel electrophoresis (PFGE) including isolates recovered from 165 single patients. We investigated the molecular relatedness among strains recovered from rectal surveillance cultures and from respective subsequent infections due to CPKP in the same individual (48/165 cases). For the optimal interpretation of our findings, we carried out a systematic review regarding the clonality of CPKP isolated from clinical samples in ICUs in Europe. In our study, we identified 128 distinguishable pulsotypes and 17 clusters that indicated extended dissemination of CPKP within the hospital ICU setting throughout the study period. Among the clinical isolates, 122 harbored KPC genes (74%), 2 harbored KPC+NDM (1.2%), 38 harbored NDM (23%), 1 harbored NDM+OXA-48 (0.6%), 1 harbored NDM+VIM (0.6%) and 1 harbored the VIM (0.6%) gene. Multiple CPKP strains in our hospital have achieved sustained transmission. The polyclonal endemicity of CPKP presents a further threat for the selection of pathogens resistant to last-resort antimicrobial agents.


Introduction
In recent years, hospital-acquired infections caused by carbapenem-resistant Gram negative bacteria, especially carbapenem-resistant Klebsiella pneumoniae (CRKP), have been observed worldwide causing important public health problems and posing serious infection control issues. CRKP are opportunistic pathogens that cause infections with high morbidity and mortality mainly in hospitalized patients [1,2]. In Europe, the burden of CRKP predominantly affects the south and the east. According to the annual report of the European Centre for Disease Prevention and Control on antimicrobial resistance in Europe,
At a similarity level of 80% or above, the majority of CPKP isolates (95.6%, 158/165) were assigned into 17 clusters (A-Q), demonstrating multiclonal dissemination. The remaining seven genomes resulted to be unrelated and were consequently classified as sporadic isolates. KPC as well as NDM genetic determinants demonstrated polyclonal dissemination being present in 14 and 11 distinct clones, respectively.
In more detail, four predominant clusters E, G, K, and M consisting of 31, 19, 16 and 41 CRKP isolates, respectively, were identified: isolates of cluster G were almost exclusively obtained from ICU 1 (18 of 19), while clusters E, K, M consisted of clinical isolates from all ICUs under study. Interestingly, within the above clusters, indistinguishable pulsotypes shared by isolates from different patients and different ICUs were identified. Furthermore, looking at indistinguishable pulsotypes, we could identify common pulsotypes among isolates obtained from different patients during different time periods of the study (P5, P27, P48 and P106).
We also used PFGE analysis for revealing the genetic association among CRKP strains from rectal and clinical samples of 48 representative patients. According to PFGE, in the majority of the cases (81.3%, 39 of 48), the clinical and rectal strains from the same patients were identical ( Figure 2). Different pulsotypes were observed for pairs PAT_1422a/b, PAT_1386a/b, PAT_854a/b, PAT_476a/b, PAT_326a/b, PAT_1529a/b, PAT_735a/b, PAT_191a/b and PAT_1216a/b.

Non-Susceptibility Rates of CRKP Isolates in the Hospital
The non-susceptibility rates of CRKP isolates in our institution's ICUs for amikacin, aztreonam, colistin, fosfomycin, gentamicin, piperacillin/tazobactam and tigecycline in isolates during the study period are shown in Table 1 and Figure 3. CRKP isolates pre-sented with high level of resistance to both meropenem and imipenem (MIC 50 ≥ 16 mg/L) throughout the study years. On the other hand, a significant increase in resistance was observed for gentamycin ranging from 15% in the first semester of 2016 to 66.7% in the first semester of 2019. In regard to tigecycline, an increase was also observed (23.5% in the 2016a semester to 61.1% in the 2019a semester). As for colistin, an increase was observed (15.8% in the 2016a semester to 27.8% in the 2019a semester).

Systematic Review Results
Our systematic review search strategy yielded 290 results. After implementation of the exclusion criteria, 39 studies remained. Fourteen studies reported monoclonal dissemination. Data extracted from the remaining 25 studies are reported in Table 2.

Discussion
The present study evaluated the type of carbapenemases and the molecular epidemiology of K. pneumoniae strains circulating in the ICUs of a tertiary hospital in Thessaloniki, Greece between 2016 and 2019. During the study years, KPC was the predominant carbapenemase; NDM were also present, and a few double-carbapenemaseproducers were isolated. On the basis of PFGE, a total of 17 different CRKP transmission clusters were identified.
CRKP isolates have been introduced in our hospital since 2004. At that time, no phenotypic or molecular testing was performed to reveal the type of carbapenemase up to 2010. From 2010 to 2014, phenotypic testing among CRKP isolates revealed that the majority carried KPC enzymes (60%), 25% produced MBL, and 7% co-produced KPC and MBL enzymes. In the time period of 2011-2014, we also observed the first OXA-48 producers in our hospital at a rate of 1.7%. Molecular testing showed that KPC positive strains harbored the bla KPC-2 , while MBL-positive strains harbored bla VIM-1 .
During 2013-2015, an oligoclonal outbreak caused by 45 CRKP occurred in our hospital [40]. All the patients were hospitalized in the three intensive care units of the hospital, and 17 (68%) of them developed bloodstream infections; the overall mortality of the patients involved in the outbreak was 48% (12/25). Molecular testing verified that all 45 K. pneumoniae isolates co-harbored bla KPC-2 and bla VIM-1 and were associated with OmpK35 deficiency and OmpK36 porin loss. PFGE clustered all isolates into a single clonal type, and multi-locus sequence typing (MLST) assigned them to the emerging high-risk ST147 clonal lineage.
In our study, PFGE analysis identified 128 distinguishable pulsotypes and 17 clusters indicating an extended dissemination of CRKP within the hospital setting. Moreover, the dissemination took place over a long-time frame since we included in our study isolates recovered during a 3.5-year period. The presence of identical isolates in all three ICUs highlights their successful dissemination through different hospital wards. More worryingly, the persistence of certain strains throughout the whole study period, despite the various infection control measures that were applied, reflects the difficulties that undermine the efforts for their eradication once they are well-established in a certain geographical area. Indeed, such strains may have persisted in the hospital or/and may have been re-introduced by carrier admissions.
In fact, most of the CRKP carriers that later presented a CRKP infection had identical pulsotypes between their rectal and clinical isolates. The most probable explanation for this finding is that gut colonization preceded infection. In some cases, however, the pulsotypes of rectal and clinical isolates were different indicating that the infection was caused by another nosocomial K. pneumoniae.
There were also two clinical-rectal pairs (PAT_1436a/b and PAT_1569a/b) that harbored different carbapenemase-encoding genes according to PCR results but had identical PFGE pulsotypes. This could be explained by the mobilization of mobile genetic elements, most likely by the loss and acquisition of plasmids [43] even though an infection by a different isolate of the same pulsotype harboring different resistance determinants could not be excluded. PFGE studies in our hospital performed in several CRKP strains during the period of 2011 to 2013 revealed that KPC strains prevailed and that the majority of them belonged to two distinct clones (unpublished data).
A similar pattern of carbapenemases was observed in Hippokration General Hospital of Thessaloniki, Greece where KPC carbapenemases have prevailed among CRKP since 2009 outnumbering the VIM-type carbapenemases that predominated previously [44].
In the same report, KPC-producers belonged to two distinct clones, the predominant of which correspond to the hyperepidemic Greek clone. In a multicenter nationwide surveillance study conducted in several Greek hospitals for CRKP from 2014 to 2016, NDMproducing isolates belonged mainly to one clone, whereas KPC, VIM, OXA-48 and double carbapenemase-producers were mainly categorized in three clones [45]. On the contrary, in our study, both KPC and NDM CRKP isolates showed extremely multiclonal profiles.
In our study, we also observed a rise in the resistance rates of tigecycline, gentamicin and colistin. This is in accordance with other studies reporting elevated resistance rates to last resort antibiotics driven, among other factors, by a vicious cycle of increased last resort antibiotic consumption and subsequent resistance [46].
Our systematic review results showed that non-monoclonal dissemination of K. pneumoniae strains in ICU settings has been described before in countries participating in the EARS-Net (Table 2). Of note, such observations have been reported almost exclusively by Mediterranean countries, mainly from Greece and Italy, and this is in accordance with the epidemiological situation of the region regarding carbapenem-resistance determinants. However, and despite the heterogeneity of settings and methods used, most studies reported rather oligoclonal transmission with further identification of sporadic cases.
In our study, multiple clones circulating simultaneously achieved sustainable dissemination and, according to our knowledge and our systematic review results, this is the first time that such polyclonal dissemination has been observed in Europe. This multi-clonal PFGE observation highlights a possible additional reason for their endemic persistence in our hospital even though infection control measures, including hand hygiene, surveillance for colonization among high-risk patients and contact precautions have been established.
In this context, active surveillance with rectal swab cultures is of outmost importance to control the spread of these pathogens by isolation or cohorting of the colonized patients [29]. However, the spread of CRKP in an endemic environmental niche is a dynamic and multifaceted phenomenon that involves many variables. In a similar situation, more than one CRKP clone may be simultaneously present in the hospital; whereas new admissions may be CRKP carriers most likely by previous hospitalizations in the same or other hospitals. Consequently, a multi-clonal spread is very likely to occur and, even when a previously colonized subject presents a CRKP infection, it is not certain that this infection is directly related to the strain that colonized the patient upon admission.
Our study has several limitations. A multi-centric study would be able to evaluate whether the epidemiological pattern that we observed in our single center study was an isolated phenomenon or more widespread. Including non-ICU along with ICU K. pneumoniae strains would yield a more complete picture for their dissemination. In our analysis, we did not include clinical patient level information, and this limits our ability to draw conclusions with regard to precipitating factors. Finally, we were not able to employ sequencing-based methods to better characterize the molecular epidemiology of the strains included in our study.

Study Design
This was a retrospective study that was carried out at AHEPA University Hospital, a 700-bed institution with three ICUs, a central surgical and medical ICU (8 beds, ICU 1), a surgical ICU (4 beds, ICU 2) and a cardiosurgical ICU (5 beds, ICU 3) as well as surgical and internal medicine departments. The study was approved by the institutional medical scientific board. Sample related patient data were retrieved from the laboratory database.
CRKP clinical isolates, recovered in the aforementioned ICUs between January 2016 and June 2019 from 165 single patients, were included in the study. In 48 cases, a rectal isolate (isolated upon admission in ICU for infection control purposes) and a subsequent clinical isolate (isolated by an infection that occurred during ICU stay) were considered, thus forming 48 pairs of surveillance-clinical isolates. Isolates taken from the remaining 117 patients were all recovered from clinical specimens only.
Rectal swabs taken from ICU patients upon admission were inoculated on MacConkey agar plates supplemented with meropenem and ceftazidime discs. All Gram-negative colonies that grew after 24 h of incubation near the discs were further identified and K. pneumoniae isolates were tested for carbapenemase production with phenotypic and molecular techniques. For clinical specimens, standard laboratory procedures were followed depending on each specimen source.
Bacterial identification and antimicrobial susceptibility testing were performed with the Vitek2 automated system (Biomerieux, Marcy-l'Étoile, France). Furthermore, the minimum inhibitory concentration of tigecycline was determined by E-test (Liofilchem, Roseto degli Abruzzi, Italy) and for colistin using the broth microdilution method (Liofilchem, Roseto degli Abruzzi, Italy). The results of all antimicrobial testing were interpreted in accordance with the CLSI criteria. For tigecycline, the breakpoints recommended by the United States Food and Drug Administration were used (susceptible: MIC ≤ 2 mg/L; resistant: MIC ≥ 8 mg/L).

Carbapenemase Detection
All isolates were phenotypically screened for carbapenemase production with the Modified Hodge Test (MHT), [47] while the type of MBL or KPC was assessed with the Combined Disk Test (CDT) [48]. Following phenotypic identification, PCR assays were performed for carbapenemase-encoding genes using specific primers for bla KPC , bla VIM , bla IMP , bla NDM and bla OXA-48 (Appendix A) [49].

Pulse-Field Gel Electrophoresis
The genetic relationship among the CPKP isolates was determined by PFGE according to standardized protocol [50] with the XbaI endonuclease (New England Biolabs, Beverly, MA, USA) by using a CHEF-DR III apparatus (Bio-Rad Laboratories Inc., Hercules, CA, USA) for the separation of DNA fragments. XbaI-digested DNA from Salmonella enterica serotype Braenderup H9812 was used as a reference size standard, while PFGE patterns were digitally analyzed using the FPQuest (Bio-Rad Laboratories Pty Ltd., Hercules, CA, USA) software package.
PFGE profiles were compared using the Dice correlation coefficient with a maximum position tolerance of 1.5% and an optimization of 1.5%. Similarity clustering analysis was performed by using the Unweighted Pair Group Method using Averages (UPGMA), and a dendrogram was generated. Two PFGE profiles were classified as indistinguishable if the DNA fragment patterns matched each other completely, while clusters were selected using a cutoff at the 80% level of genetic similarity.

Non-Susceptibility Rates of CRKP Isolates
For every semester of the study period, we determined the K. pneumoniae non-susceptibility rates of amikacin, aztreonam, colistin, fosfomycin, gentamicin, piperacillin/tazobactam and tigecycline in imipenem non-susceptible single patient isolates from the ICUs using the CLSI 2020 breakpoints.

Systematic Review
In order to evaluate the extent of non-monoclonal transmission of CRKP strains in the intensive care environment in countries participating in the European Antimicrobial Resistance Surveillance Network (EARS-Net), we undertook a systematic review of the recent literature. We searched MEDLINE via PubMed from 1 January 2000 to 28 April 2021, implementing the search strategy described in Appendix B. Titles and abstracts were screened for studies, which incorporated a molecular epidemiology investigation (PFGE or sequencing methods) of CRKP strains, including samples obtained from ICU patients.
We excluded reviews, case reports, studies restricted to environmental samples, studies that did not explicitly include any ICU clinical samples and studies not conducted in EARS-Net participating countries. No language and patient age restrictions were applied.
Eligibility assessment was conducted in duplicate (D.P. and G.M.) and discrepancies were adjudicated by a third author (E.P.).
Next, studies reporting monoclonal transmission were excluded, and, from the remaining studies, we extracted the following data: study setting and eligible samples, sample type (clinical versus surveillance and infection versus colonization), the number of carbapenem-resistant isolates and mechanism of resistance detected, the method used to investigate molecular epidemiology and the number and size of clusters involving ICU patients. If ICU-level data were not available, hospital-wide data were reported. Data extraction was conducted by A.T., G.M., T.P., D.P. and E.P. All steps were conducted in duplicate.

Conclusions
In our study, we demonstrated that CPKP in our hospital belonged to a great variety of pulsotypes and clusters, thus, indicating their extended dissemination within the hospital ICU settings. Among them, KPC carbapenemase predominated. The presence of multiple clones harboring variable resistance-determinants poses additional challenges. Further studies are required to identify suitable infection control strategies in a setting of polyclonal dissemination within a context of carbapenem resistance endemicity. Our results highlight the need to urgently reinforce infection-control measures along with antimicrobial stewardship and together with a generous increase in the nosocomial budget in order to contain the transmission of antibiotic resistant organisms.  Acknowledgments: We wish to express our gratitude to the laboratory technicians of the Microbiology Laboratory for the performance of the phenotypic and antimicrobial susceptibility tests. We are also grateful to the nurses of the Infection Control Unit of AHEPA Hospital for assistance in collecting the rectal samples.

Conflicts of Interest:
The authors declare no conflict of interest.