Introduction
Klebsiella pneumoniae is an opportunistic pathogen that causes a variety of infections in hospitalized patients, including ventilator-associated pneumonia, septicemia and urinary tract infections. It can readily acquire resistance to a wide range of antimicrobials due to the production of extended-spectrum β-lactamases (ESBL), AmpC-β-lactamases and carbapenemases encoded on mobile genetic elements. Colistin is an old antibiotic with significant activity against the Gram-negative bacteria including
K. pneumoniae, which is now being administered as a salvage therapy in patients in whom none of the other antibiotics are active against such isolates. The most common mechanism of colistin resistance in
K. pneumoniae is the inactivation of the
mgrB gene, encoding a negative feedback regulator of the
PhoQ-
PhoP signaling system which activates the
pmr system, responsible for the modification of the lipopolysaccharide polymyxin target.[
1,
2,
3] Inactivation of the
mgrB gene can be mediated by sequence IS1F or IS5-like, disrupting the gene.[
3] Moreover, high-level resistance to colistin was found to be mediated by various mutations in the
crrB gene among carbapenemase producing
K. pneumoniae from Europe, Turkey, South America and South Africa.[
4] Plasmid-mediated colistin resistance is associated with acquisition of
mcr-1 or
mcr-2 genes and is the dominant mechanism of colistin resistance in
Escherichia coli, particularly among animal isolates.[
5,
6] Recently, plasmid-mediated colistin resistance was also reported in
K. pneumoniae, but it is still very rare.[
7] Colistin resistance in
K. pneumoniae is of great concern and is usually associated with carbapenemase and ESBL production.[
6,
7] OXA-48, VIM-1, or KPC are most frequently linked with colistin resistance. KPC positive and colistin resistant isolates were shown to belong predominantly to the widespread clone ST258 with a few isolates belonging to ST15 and ST273 reported in USA, Greece and Italy.[
8,
9,
10] The majority of the isolates harbored KPC-3 allelic variant and co-produced TEM-1 and OXA-9 contributing to resistance to beta-lactam inhibitor combinations. Colistin resistance was found to be linked to increased virulence in
Galleria mellonella, model. The first colistin resistant
K. pneumoniae isolates in Croatia were recently reported in the frames of the study on the dissemination of OXA-48 carbapenemases.[
11,
12]
Analysis of colistin resistance mechanisms of
K. pneumoniae and
A. baumannii from Croatia was recently published.[
13] Whole genome sequence analysis revealed several mutations in the
mgrB and
rpo genes to be responsible for colistin resistance in
K. pneumoniae.After the first sporadic cases, alarming diffusion of this important resistance determinant was observed in 2019 in several hospitals located in different geographic regions of Croatia and also in the outpatient setting. This prompted us to analyze the molecular epidemiology and routes of spread of this important hospital pathogen.
Methods
Bacterial isolates
In total 46 colistin resistant non-duplicate (one per patient) K. pneumoniae isolates were collected in seven centers located in different geographic regions of Croatia in order to cover the whole country and to get an insight on dissemination of this important resistance determinant. The participating centers were as follows: General Hospital Karlovac (23 isolates), University Hospital Center Sestre Milosrdnice in Zagreb (2 isolates), University Hospital Center Rijeka (4 isolates), Andrija Štampar Public Health Institute in Zagreb (6 isolates), Public Health Institute Sisak (4 isolates), University Hospital Center Split (4 isolates) and General Hospital Dubrovnik (3 isolates). The isolates originated from various clinical specimens including invasive, clinically relevant (n=33: blood cultures n=2, urine n=28, wound swabs n=2, intraoperative specimen n=1) and surveillance cultures (n=13: tracheal aspirates n=5, bronchoalveolar lavage n=3, rectal swabs n=3, stool n=2) during 2019 except for five outpatient strains which originated from 2016 to 2018. The isolates were identified in the participating centers by conventional biochemical testing, Vitek 2 or MALDI-TOF MS (matrix assisted laser desorption ionization-time of flight mass spectrometry). The isolates that were stored in the participating centers were sent for molecular analysis to University Hospital Centre Zagreb in the frames of the retrospective study.
Conjugation
Colistin, ertapenem and cefotaxime resistance transfer experiments were carried in mating assays employing
E. coli J65 recipient strain resistant to sodium-azide.[
19] Equal volume of cultures of donor and recipient strains were mixed in Brain-Heart infusion broth and incubated for 18h at 35°C. The transconjugants were selected on MacConkey agar supplemented with either colistin (1 mg/L), ertapenem (1 mg/L) or cefotaxime (2 mg/L) to suppress the recipient strain and sodium azide (100 mg/L) to inhibit the growth of the donor strains.
Molecular detection of resistance genes
A fresh bacterial colony was suspended in 100 µL of sterile, distilled water and boiled at 100°C for 15 min. After centrifugation the supernatant was used as DNA template. Genes conferring resistance to β-lactams, including broad spectrum and extended-spectrum β-lactamases (
blaSHV,
blaTEM,
blaCTX-M, and
blaPER-1), plasmid-mediated AmpC β-lactamases (pAmpC), class A (
blaKPC), class B carbapenemases (
blaVIM, blaIMP and
blaNDM), carbapenem hydrolyzing oxacillinases (
blaOXA-48-like) and fluoroquinolone resistance genes (
qnrA,
qnrB,
qnrS) were amplified by PCR using protocols and conditions as described previously.[
20,
21,
22,
23] For each PCR reaction 25 µL of emerald master mix, 20 µL of ultrapure water and 1 µL of each primer were mixed with 3 µL of DNA template to obtain the total volume of 50 µL. Genetic context of
blaCTX-M genes was determined by PCR mapping with forward primer for IS
Ecp1 and IS
26 combined with primer MA-3 (universal reverse primer for
blaCTX-M genes). Flanking regions of
blaOXA-48 genes were analysed by PCR using forward primer for IS
1999 and reverse primer for
blaOXA-48. PCR was used to analyze the inactivation of
mgrB genes and to detect plasmid encoded colistin resistance genes
mcr-1 and
mcr-2.[
6]
PCR products were separated in 1% agarose gel and visualized under UV light after staining with ethidium bromide.
Characterization of plasmids
Plasmids were extracted from donor strains and their respective transconjugants with Qiagen (QIAGEN Hamburg, Germany) mini kit according to the manufacturer’s instructions. After staining with ethidium bromide, the DNA was visualized by ultraviolet light. The size of the plasmid bands was determined by comparison with those of
E. coli NTCC 50192 yielding four bands of know sizes of 148, 64, 36 and 7 kb. PCR-based replicon typing (PBRT) was applied to determine the plasmid content of the tested strains.[
24] Plasmid extractions obtained from transconjugant strains were subjected to PCR for the detection of KPC and ESBLs in order to determine the resistance gene content of the transconjugants. PBRT was also applied on transconjugants to identify incompatibility groups such as in their respective donors. Positive control strains for PBRT were kindly provided by dr. A. Carattoli (Instituto Superiore di Sanita, Rome, Italy).
Pulsed-field gel electrophoresis (PFGE)
In total 34 isolates, which were available for testing, were subjected to PFGE. Bacterial DNA was prepared and
XbaI- was used as restriction endonuclease. Plugs were loaded onto agarose gel and DNA separation was performed with CHEF-DRIII system (Bio-Rad, USA); as described elsewhere.[
25] The DNA macrorestriction patterns were compared and analysed using the Gel-Compar software to determine the level of similarity. The dendrogram was computed after band intensity correlation using global alignment with 1.5% optimization and 1% tolerance and unweighted pair-group method using arithmetical averages (UPGMA) clustering. PFGE cluster analysis was carried out with Bionumerics 7.6 (Applied Maths, Belgium) using Dice similarity coefficient and clustering by the UPGMA. Bands smaller than ~50 kb were not included in analysis. Pulsotypes were defined at 80% similarity between macrorestriction patterns. If the strains had two to four different fragments in their
Xba patterns they were designed as subtypes. One isolate KC 5852 was subjected to multilocus sequence typing according to the instructions on the Pasteur website.
Infection control measures
Screening methods used for surveillance of nosocomial colonization according to the local ICU procedures include nasal, pharyngeal and rectal swabs as well as urine and stool cultures taken at admission and during ICU stay. Patients are screened for methicillin-resistant Staphylococcus aureus (MRSA), vancomycin-resistant Enterococcus faecalis (VRE), ESBL producing Enterobacteriaceae and carbapenem and colistin-resistant Enterobacteriaceae, Pseudomonas aeruginosa and Acinetobacter baumannii. In case of suspected or confirmed infection, microbiological examination depends on the suspected site of infection (blood culture, tracheal aspirate, bronchoalveolar lavage fluid or urine culture).
Determination of virulence factors
String test was attempted by stretching a mucoviscous string from the colony using a standard bacteriologic loop. If a viscous string >5 mm was formed, isolate was defined as hypermucoviscous.
The production of β hemolysin was tested on human blood agar plate and was considered positive when bacteria were stabbed with a sterile straight wire into 5% human blood agar, and after 18 to 24 h of incubation at 37°C, a clearing zone was observed.
For serum sensitivity testing blood was obtained by venipuncture from three healthy volunteers and was allowed to clot at room temperature for 30 min and overnight at 4°C. After centrifugation at 1,000 g for 15 min at 4°C, serum was removed and pooled. A portion of the pooled serum was decomplemented by heating at 56°C for 30 min and used as test controls. Bacterial susceptibility to serum killing was measured by assessing regrowth after incubation in normal human serum according to Schiller and Hatch method.
Discussion
The study demonstrated a sustained outbreak of colistin resistant K. pneumoniae in five hospitals, located in different geographic regions of Croatia and in the outpatients setting. mcr genes responsible for transferable colistin resistance were not found, indicating that resistance was probably due to mutations in mgrB genes, but experimental data supporting this speculation are required. The clarification of the resistance mechanism was beyond this study.
Colistin resistance was associated in the majority of cases with OXA-48 which is now the most widespread type of carbapenemase in Croatia.[
11] A few isolates from southern regions of Croatia (Dubrovnik and Split) were found to co-harbor KPC carbapenemases. The isolates positive for OXA-48 exhibited very low MICs of meropenem and imipenem with the majority of isolates being in the susceptible range. All except seven isolates co-harbored additional group 1 CTX-M β-lactamase associated with high level resistance to ESC and gentamicin. OXA-48 was probably carried by IncL plasmid responsible for horizontal spread of this important resistance determinant. Difficulties in transfer of plasmid from
K. pneumoniae donor strains are likely due to the high molecular mass expected for the IncL plasmid identified in these strains. Because of this technical limitation, it was not possible to determine the location of
blaOXA-48 genes. Four KPC producing organisms showed high level of resistance to all antibiotics except for sulfamethoxazole/trimethoprim.
Croatia has a national surveillance system set in place and specific guidelines for the management of carbapenemase-producing Enterobacteriaceae, along with the obligation to report them to the health authorities. Despite these measures we have seen an increase of OXA-48 producing organisms, similar to the situation in other European countries. However, there is no national surveillance programme for monitoring colistin resistance.
The production of ESBL, predominantly belonging to the CTX-M family, was associated with resistance to ESC. In this study the majority of OXA-48 positive organisms harbored ESBLs and attributable ESC resistance, which is in contrast with the previous study which analysed OXA-48 colistin susceptible
K. pneumoniae.[
11,
12] ESBL positive isolates exhibited resistance to gentamicin and fluoroquinolones due to the additional
aac and
qnr genes usually located at the same plasmid. All isolates showed a high level of resistance to amoxicillin/clavulanate and piperacillin/tazobactam, which is consistent with both OXA-48 β-lactamase resistant to inhibitors and KPC which also efficiently hydrolyze penicillins. Almost half of the isolates were XDR, limiting the therapeutic options.
Based on resistance gene content, three clones were identified: a large cluster with OXA-48 positive isolates containing 39 isolates and two small clusters, one with KPC (4 isolates) and one with CTX-M-group 1 as the sole β-lactam resistance mechanism (two isolates), respectively. One single isolate harbored NDM-1 carbapenemase. OXA-48 exhibited variable MICs of carbapenems ranging from susceptibility to frank resistance. All except two OXA-48 positive isolates demonstrated resistance to ESC due to additional CTX-M β-lactamase and were classified as MDR. Four KPC and one NDM-1 producing organisms were XDR. Interestingly, colistin resistance in northern or middle regions of Croatia (Zagreb, Sisak and Karlovac) was associated with OXA-48 whereas in southern areas of Croatia (Split and Dubrovnik) with KPC. In other European countries colistin resistance was in the majority of cases linked to KPC carbapenemases, in contrast to our study where OXA-48 was the dominant carbapenem resistance determinant.[
9,
10] Colistin resistance in some centers such as Karlovac and Split reached an alarming rate of 7.9% and 9.8%, respectively.
The OXA-48 positive isolates exhibited variable MICs of imipenem and meropenem ranging from 0.5 to >128 mg/L making the phenotypic detection difficult and unreliable. All isolates showed reduced susceptibility to ertapenem in disk-diffusion test, which is used as screening method for carbapenemase production. Due to the very variable level of carbapenem resistance, microbiologists rely on the phenotypic tests. An important proportion of OXA-48 positive isolates exhibited resistance only to ertapenem, with MICs of imipenem and meropenem being in the susceptible or intermediate susceptible range. For that reason, laboratory detection of OXA-48 poses a serious challenge to clinical microbiologists. KPC and NDM positive organisms showed high level of resistance to almost all antibiotics and XDR phenotype. False negative Hodge and CIM tests observed in some isolates were linked solely to OXA-48 carbapenemase.
High variability of carbapenem MICs could be attributable to variable expression of blaOXA-48 genes. Surprisingly, the MICs of meropenem were higher than those of imipenem in the majority of isolates in spite of the fact that imipenem is, generally, better hydrolyzed by oxacillinases.
Part of the OXA-48 positive isolates were positive for IS1999 upstream of the blaOXA-48 gene which is responsible for the mobilization of blaOXA-48 genes and enhances the expression of the gene. Analysis of the flanking regions of blaOXA-48 gene revealed IS1R element between IS1999 and the OXA-48 encoding gene. ISEcp which mediates the mobility and expression of the blaCTX-M gene was found in only three strains, which could explain lack of transferability of cefotaxime resistance.
IncFIIs plasmid was dominant among isolates. In the previous studies it was associated with the spread of KPC carbapenemase, but in the present study it was found in isolates harboring OXA-48 and CTX-M β-lactamases. IncL/M was found in the NDM positive organism which is in concordance with the previous results. Similar types of plasmids were obtained in the previous studies on carbapenemases in Croatia.[
11,
12,
13] Unlike in Italy, our KPC producing organisms did not harbor
blaTEM-1 or
blaOXA-9 genes.[
9,
10] The antibiotic susceptibilities were similar with high level of resistance to the majority of antibiotics, but our isolates were uniformly resistant to gentamicin in contrast to the Italian isolates reported. Regarding the virulence factors, production of β-hemolysins was linked to OXA-48 positivity whereas hyperviscosity was found only in one KPC producing organism. Both groups of isolates showed resistance to serum bactericidal activity allowing them to cause severe systemic infections such as bloodstream infections. However, septicemia was diagnosed in only two patients and in general, mortality rate was lower than expected concerning MDR or XDR phenotype. Ceftazidime/avibactam was shown to be associated with favorable therapeutic outcome. There was no association between the severity of infection or outcome and the presence of certain virulence traits.
Spread of carbapenemase producing
K. pneumoniae was polyclonal, similarly as in the previous studies and probably mediated by horizontal spread of plasmids belonging to the same incompatibility group.[
11,
12,
13] Polyclonal spread including various STs and PFGE patterns was also reported in Italy,[
9,
10] while in USA dissemination of colistin resistant
K. pneumoniae was monoclonal.[
8] However, some clonal similarity was observed between isolates from Karlovac and those from Zagreb and Sisak indicating a possible epidemiological link probably due to patient or staff transfer. Infection control measures were implemented and the outbreak in Karlovac was terminated. ST37 was previously reported among isolates from Croatia with different types of carbapenemases, indicating that the same isolate lineages can evolve and acquire different resistance traits. The strength of the study is detailed molecular analysis of additional resistance mechanisms such as ESBLs and carbapenemases and inclusion of multiple centers, but the limitation is that the mechanism of colistin resistance was not clarified and that the genotyping was done by an “old” method such as PFGE whereas MLST as the portable method enabling international comparisons of the isolates, was done for only one isolate. Future studies should employ whole genome sequencing to analyze the complete resistance and also virulence gene content of these successful isolates.