Between 2011 and 2021, 651 patients admitted to the semi- and intensive care units were carriers of KpKPC according to rectal swab analysis (carriage isolates). Infection isolates were those obtained from blood, bronchial lavage, ascitic fluid, abdomen secretion, tracheal secretion, and urine, and out of the 651 carriers, 128 developed a KpKPC infection after 30 days of carriage detection (i.e., positive rectal swab sample for KpKPC). The viable bacterial samples that could be reactivated in culture media and corresponded to the first KpKPC-positive carriage and/or infection sample of non-repeated patients were considered for the study. Considering these criteria, both carriage and infection isolates from 54 patients who had infection following colonization (n = 108 isolates) and 68 carriers (n = 68 isolates) were selected. A total of 176 isolates from 122 different patients were evaluated in this study. All 176 isolates were characterized by PFGE, and 124 were sequenced, corresponding to at least one isolate per patient included in the study. For patients who developed an infection associated with the same KpKPC isolate according to PFGE analyses (≥90% similarity), only the infection isolate was sequenced.

For the relative quantification of the intestinal load, rectal swabs from KpKPC carriers admitted to the hospital between June 2020 and December 2021 were selected. In this period, 165 patients were KpKPC carriers, from which 58 developed an infection associated with the presence of KpKPC within 30 days of the KpKPC-positive rectal swab. Only one sample per patient was included in the study, corresponding to the first KpKPC-positive rectal swab.

This study was approved by the Institutional Review Board (CAAE: 98373618.0.0000.0071).

K. pneumoniae isolates were identified by Matrix Assisted Laser Desorption Ionization Time Of Flight Mass Spectrometry (MALDI-TOF MS) (Bruker Daltonics, Billerica, MA, USA). Antibiotic susceptibility was determined using the Vitek 2 XL System (bioMérieux, Craponne, France) and confirmed by the epsilometric (Etest^®) method for carbapenems (imipenem or meropenem). Antimicrobial susceptibility results were interpreted according to the most recent Brazilian Committee on Antimicrobial Susceptibility Testing/European Committee on Antimicrobial Susceptibility Testing (BrCAST/EUCAST) guidelines (BrCAST, 2020). Carriage isolates were tested for imipenem and meropenem (n = 122). Infection isolates (n = 54) were tested for imipenem and meropenem and also for cephalosporins (ceftazidime and cefepime), aminoglycosides (amikacin and gentamicin), and ciprofloxacin.

Real-time PCR was used to detect the bla _KPC gene in isolates showing reduced susceptibility to carbapenem. For this protocol, DNA extraction was performed using the PrepMan™ kit (Thermo Fisher Scientific, Waltham, Massachusetts, United States). DNA (10 ng) was added to the TaqMan master mix containing specific probes for bla _KPC (FAM fluorophore – 5´ TG ATA ACG CCG CCG CCA ATT TGT 3´) and 16S rRNA (CY5 fluorophore – 5´ CA CGA GCT GAC GAC AR*C CAT GCA 3’), as well as specific primers for bla _KPC (Forward: 5´ GGCCGCCGTGCAATAC 3´ and Reverse: 5’ GCCGCCCAACTCCTTCA 3’) and 16S rRNA (Forward: 5’ TGGAGCATGTGGTTTAATTCGA 3’ and Reverse: 5’ TGCGGGACTTAACCCAACA 3’). The mixture was inserted in a microfluidic cartridge and processed by the BD MAX™ equipment using the PCR Only module. Results were released qualitatively: the presence or absence of the gene.

The genetic relatedness was established by pulsed-field gel electrophoresis (PFGE) as previously described (Han et al., 2013). For the PFGE result interpretation, the Dice similarity coefficient was used and isolates were considered identical when patterns showed ≥ 90% similarity (Koroglu et al., 2015).

Genomic DNA isolation for whole-genome sequencing (WGS) was performed as previously described (Salvà Serra et al., 2018). The concentration and purity of genomic DNA was assessed with a NanoDrop™ One Spectrophotometer (Thermo Fisher Scientific). DNA fragmentation for library construction was performed using the Ion Shear Plus reagents kit, and libraries were constructed using the Ion Plus Fragment Library Kit (Thermo Fisher Scientific). Barcoded libraries were quantified using the Bioanalyzer 2100 and High Sensitivity DNA kit (Agilent, Santa Clara, California, United States). Clonal amplification of the libraries was carried out using Ion PI™ Hi-Q™ Chef Kit (Thermo Fisher Scientific) and sequenced using the Ion PI™ Chip Kit v3 and Ion PI™ Hi-Q™ Sequencing 200 Kit (Thermo Fisher Scientific) in the Ion Proton Sequencer. Quality filtering was done with cutadapt v3.4 (Martin, 2011). Assembly was performed with SPAdes genome assembler software (v3.15.0) using the "iontorrent" and "careful" options (Prjibelski et al., 2020).

MLST, the detection of virulence and antibiotic resistance genes, and lipopolysaccharide (O) and capsular polysaccharide (K) characterization were based on the analysis of WGS data. Sequence types (STs) were determined using the multilocus sequencing typing (MLST) script available at https://github.com/tseemann/mlst implemented at Institut Pasteur (Diancourt et al., 2005). Resistance and virulence genes were identified using the ABricate tool available at https://github.com/tseemann/abricate. ABricate allows the screening of contigs against multiple databases, including the AMRFinderPlus resistance gene database (Feldgarden et al., 2019) and the Virulence Factor Database (VFDB) (Chen et al., 2005).

Rectal swabs were eluted in 0.5 mL of the TE 1X buffer (Tris-EDTA; 10 mM Tris, 1mM EDTA, pH 8.0) and used for serial 10-fold dilution in 0.9% saline as previously described (Lerner et al., 2013). The dilutions were plated on tryptic soy plates supplemented with 5% sheep blood (BiobioMérieux, Craponne, France) and CHROMagar ™ msupercarba™ plates (CHROMagar, Paris, France), to obtain total viable aerobic bacteria (TAB) and viable CRE, respectively. Bacterial counts were determined after 18 h of growth at 37°C, and the ratio of CRE to TAB [colony-forming unit (CFU/ml)] was determined and expressed as log CRE/TAB. The rectal swab suspensions were stored at -20°C for further molecular analysis.

Total DNA was extracted from rectal swab suspensions. Briefly, 250 µl of bacterial suspensions were heated at 100°C for 10 min and centrifuged at 12,000g for 5 min, and the lysate was transferred to a new tube. Quantitative real-time PCR was performed in multiplexed reactions using primers for the 16S rRNA gene (see above) and following conditions previously described (Centers for Disease Control and Prevention (CDC), 2019). DNA from a K. pneumoniae isolate harboring the bla _KPC gene, Kp378, previously sequenced by our group (Migliorini et al., 2021) was used as positive control. The bla _KPC copy number of the positive control (Kp378) was determined from short-read assemblies by dividing the coverage of the contig containing bla _KPC by the average coverage for the assembly (weighted by contig length) as previously described (Stoesser et al., 2020). The 2^-ΔΔCt method was used to compare bla _KPC copy numbers between the rectal swab and the reference samples (Ramos-Ramos et al., 2020) ΔΔCt was defined as the difference between the ΔCt of a sample (rectal swab) and ΔCt of the reference sample (Kp378). Relative loads (RLs) were defined as RL= log(2^-ΔΔCt), in which values close to zero meant that the sample contained a similar copy number of bla _KPC when compared to the reference sample, and values lower or higher than zero were considered as low or high intestinal loads of bacteria carrying bla _KPC gene, respectively.

Electronic clinical records were searched retrospectively. The variables registered were sex, age, comorbidities, hospitalization days within last 2 months, medical ward admission, antibiotic treatments in the previous 30 days, invasive procedures, and infections. The statistical significance of results was tested by non-parametric Mann–Whitney U test and two-way ANOVA for continuous variables and Fisher’s exact test for categorical variables (Graph Pad Software 9.0 Inc). Bivariate analysis to determine the impact of covariables on infection following colonization was carried out by binary logistic regression, adjusting for confounders, using jamovi (Version 1.6).

Out of 122 KpKPC carrier patients from which isolates were selected for this study, 54 had a subsequent infection episode; from these 54 cases, both carriage and infection KpKPC isolates were analyzed by PFGE (n = 108). In 46 of these cases (85%) the same KpKPC strain was detected as carriage and in the infection site (≥90% similarity in the PFGE pattern), suggesting that KpKPC intestinal colonization was the primary source for infection.

At least one isolate per patient was sequenced; general genome data quality, STs, O-, and K-types are reported in Supplementary Table 1 . Most carriage and infection isolates belonged to the most prevalent strains in the period, ST11 or ST437, and no specific ST could be associated with either carriage or infection in this cohort ( Figure 1A ).

Additionally, capsule polysaccharide synthesis and O antigen locus analysis indicated that isolates belonged to 15 distinct K-loci groups, the most recurrent being KL36, corresponding to ST437. We also identified a total of four different O-types, of which O4 was predominant. No differences between carriage and infection samples were observed.

All included isolates were resistant to carbapenems (imipenem or meropenem). Among infection isolates, resistance rates were high for ciprofloxacin (98%), ceftazidime (93%), and cefepime (93%) and lower for amikacin (48%) and gentamicin (44%). The bla _KPC-2 gene was the most frequently found gene encoding a carbapenamase (102/124), followed by bla _KPC-3 (3/124), bla _KPC-30 (2/124), and bla _KPC-33 (1/124). One isolate of K. pneumoniae carried bla _NDM. No other carbapenem resistance mechanisms were found. Additionally, genes associated with resistance to cephalosporins (bla _CTX-M-15, bla _OXA-1, bla _TEM-1, and bla _SHV-158) and to aminoglycoside (aac(6’)-Ib-D181Y) were also detected, as well as genes associated with reduced susceptibility to ciprofloxacin (oqxA and oqxB). All infection-associated isolates resistant to ciprofloxacin, cephalosporins, and/or aminoglycoside carried genes related to resistance to these antibiotic classes, showing a good association between the resistance phenotype and genetic determinants. The number of resistance genes was not significantly altered over time, and no significant differences were observed between carriage isolates and those related to infection ( Figure 1B ). Surprisingly, virulence genes were not mostly found in the isolates involved with infection nor did the number of virulence genes vary ( Figure 1B ). Carriage and infection isolates also carried a very similar set of virulence markers (data not shown), including yersiniobactin (ytb) and colibactin (clb) ( Figure 1C ).

Taken together, these results indicate that the resistance and virulence markers identified in this study, or their combination, possibly contribute but do not explain infections identified following intestinal colonization by KpKPC.

The relative intestinal load of CRE and bla _KPC-carrying bacteria (bacKPC) was estimated by microbiological and molecular quantification methods, respectively. Out of the 165 patients evaluated, 58 presented with an infection subsequent to colonization (infection group) within 30 days, and 107 showed only intestinal colonization by KpKPC (carrier group). Only the stay in intensive/semi-intensive care units was significantly associated with infection in this cohort ( Table 1 ).

The relative intestinal load was determined using two techniques: the microbiological method that measured the load of CRE as related to the total aerobic bacteria present in the sample (results are reported as log CRE/TAB) and the molecular method, comparing bla _KPC gene copy numbers and the total number of bacteria present in the sample as estimated using primers specific to the conserved region of 16S rRNA gene (results are reported as log 2^-ΔΔCt). The relative load of CRE in the infection group was higher (median of -0.54, 95% CI: -0.77 to -0.30) than in the carrier group (median of -1.49, 95% CI: -1.76 to -1.23) (P<0.0001) ( Figure 2A ). Similarly, the infection group showed a higher relative load of bacKPC from swabs (median of -0.28, 95% CI: -0.51 to -0.06), compared to carriers (median -1.36, 95% CI: -1.69 to -1.02) (P<0.0001) ( Figure 2B ).

On the bivariate logistic regression, ICU/semi-intensive care unit admission (OR 2.79, 95% CI 1.375 to 5.687, P=0.005), the relative intestinal load of CRE (OR 3.78, 95% CI 1.658 to 8.640, P=0.002), and relative intestinal load of bacKPC (OR 8.601, 95% CI 2.44 to 30.352, P<0.001), were factors independently associated with infections.

However, when analyzed independently, patients from ICU/semi-intensive care units and those admitted to other wards had higher intestinal loads of CRE and bacKPC when they belonged to the infection group when compared with the carriage group (P<0.05; Figures 3A, B ).

Taken together, these results indicate that patients from the infection group showed higher intestinal loads of CRE and bacKPC when compared with patients who did not present with infection within 30 days of the first positive KpKPC swab.