This study included a subgroup of 182 patients from the prospective longitudinal ACLF-I cohort study. The ACLF-I cohort comprises prospectively enrolled adult patients with ACLD who were hospitalized at the University Hospital Frankfurt (November 2020-June 2023). Eligibility was defined by any inpatient admission with established liver cirrhosis, while patients under 18 or over 80 years of age, with hepatocellular carcinoma beyond Milan criteria (Mazzaferro et al., 2011), any other malignant disease, ongoing immunosuppressive therapy, prior organ transplantation, or pregnancy were excluded.

All patients were subjected to screening measures for MDRGN (including NCR-DRGN) on admission and in the course of their hospital stay. These measures ground in nation-wide regulatory demands by the German Infection Protection Law (IfSG) and in-house guidelines outlined in the hygiene plan to prevent transmission of infective agents. Patients remained in the study until death or last follow-up. Study approval was obtained by the local Ethics Committee for Medical Research of the Medical Faculty, University Hospital, Goethe University, Frankfurt am Main, in accordance with the 1975 Declaration of Helsinki prior to research (file number 20-653). Informed consent was obtained upon study inclusion, and the database was pseudo-anonymized. Analysis of bacteria was also approved by the local Ethics Committee (waiver 2024-2145) and conducted within legal requirements given by German Infection Protection Law (IfSG), in particular §§13, 23 IfSG.

All laboratory procedures were conducted in accordance with quality-controlled DIN ISO 15189:2024 standards (certificate number D–ML–13102–01–00). NCR-DRGN bacteria included in this study were defined as Enterobacterales, Pseudomonas aeruginosa and Acinetobacter baumannii-calcoaceticus complex having pheno- or genotypic resistance against at least third generation cephalosporins. Bacterial isolates with a phenotypic carbapenem resistance were excluded from further analysis. Rectal and nasopharyngeal swabs were the primary sample types utilized to screen for NCR-DRGN, with additional samples, including skin swabs, tracheal swabs, and urine samples as required. Amies collection and transport medium (Hain Lifescience, Nehren, Germany) was used for sample collection. For bacterial cultivation, all swabs were directly streaked onto CHROMagar™ ESBL plates (Mast Diagnostica, Paris, France) without prior enrichment culture allowing the growth and selection of ESBL-producing and other Gram-negative bacteria with resistance to penicillins and third generation cephalosporins.¹ Species identification of suspected colonies was performed using matrix-assisted laser desorption ionization-time of flight (MALDI-TOF) analysis or the automated VITEK^® 2 system (bioMérieux, Nürtingen, Germany). Antibiotic susceptibility testing was carried out in accordance with the European Committee on Antimicrobial Susceptibility Testing (EUCAST) guidelines (EUCAST, 2020a,b). Methods included VITEK^® 2 automated systems, antibiotic gradient tests (Liofilchem, Roseto degli Abruzzi, Italy; Etest^®, bioMérieux, Nürtingen, Germany), and agar diffusion assays (Oxoid, Wesel, Germany). Carbapenemase screening using the CARBA-5 immunochromatographic lateral flow assay (NG Biotech, Guipry, France) targeting NDM, VIM, IMP, OXA-48-like, and KPC carbapenemases was conducted based on the criteria outlined by the German National Antibiotic Susceptibility Committee (Hamprecht et al., 2021).

Definitions for colonization and infection were based on established microbiological and clinical standards: Colonization was defined as the detection of NCR-DRGN in nasal, rectal, or pharyngeal swabs. Infection was defined when NCR-DRGN were isolated from wounds or sterile compartments. According to German microbiological-infectiological quality (MiQ) standards (Schubert et al., 2020), urinary tract infection was confirmed in cases where no more than two bacterial pathogens, including NCR-DRGN or potential pathogens, were identified alongside leukocytes. In the absence of leukocytes, infection was defined by the presence of at least 10⁴ bacteria/mL of urine.

Genomic DNA was extracted using DNeasy UltraClean Microbial Kit (Qiagen, Venlo, Netherlands) and sequenced on the Illumina NovaSeq X Plus platform (Illumina Inc., San Diego, USA) generating paired-end 150 bp reads (PE150). Raw reads were quality-trimmed, and Illumina-specific adapters were removed using Trimmomatic v0.39 (Bolger et al., 2014). Taxonomic classification of reads was performed as an additional quality control step using Kraken v2.1.3 (Wood et al., 2019). High-quality reads were assembled de novo into draft genomes using Unicycler v0.5.1 (Wick et al., 2017), and assembly quality was evaluated using QUAST v5.3.0 (Gurevich et al., 2013). Contigs shorter than 200 bp were discarded to ensure high assembly accuracy. Multi-locus sequence typing (MLST) was performed using mlst v2.23.0 (Seeman, 2019), and taxonomic classification of isolates was reconfirmed using Kraken v2.1.3 (Wood et al., 2019). Antibiotic resistance and virulence genes were identified by screening with ABRicate v1.0.1 (Seemann, 2016) and BLASTn searches against the ResFinder database (Zankari et al., 2012). For core genome based multilocus sequence typing (cgMLST) and minimum spanning tree construction of E. coli isolates, Ridom SeqSphere+ v10.5.0 was utilized applying the EnteroBase core genome MLST scheme.² Genome annotation was performed using Prokka v1.14.6 (Seemann, 2014), and the resulting GFF files were used for pangenome analysis with Roary v3.13.0 (Page et al., 2015). The pangenome data were visualized using Roary2svg.pl.³ Prediction of localization of antimicrobial resistance (AMR) genes was based on RFPlasmid predictions.⁴ A tree based on the Roary alignment was calculated with IQTree2 version 2.2.5 using ModelFinder (Kalyaanamoorthy et al., 2017) to determine the best-fit model. Branch support was accessed by calculating 100,000 ultrafast bootstraps.

To identify potential epidemiological links in patients colonized or infected with NCR-DRGN bacteria, a review of patient data records was conducted. Clinical and microbiological data, including demographics, hospitalization history, and antimicrobial treatments, were analyzed to evaluate transmission dynamics and resistance patterns. The analysis focused on overlaps in time and location within healthcare facilities including (i) patient admissions and movements across general wards, intermediate care and intensive care units, (ii) close contacts located in the same room, (iii) visits to diagnostic and therapeutic units (e.g., endoscopy, ultrasonography, radiology, operating theaters) and (iv) visits to the outpatient clinics.

Patient characteristics were collected through a prospectively maintained database, updated at every patient’s visit in UHF. Baseline characteristics of the proportion of the ACLF-I cohort described in this paper reflect an advanced cirrhosis population, with a median age of 58 years, and a model-of-end-stage-liver-disease (MELD) score of 17 demonstrating the expected gradients in hepatic dysfunction and organ failure burden. The most common etiology was alcoholism (54.5%). Two patients underwent liver transplantation prior to NCR-DRGN detection. Decompensation at admission presented itself with ascites (45.5%) and spontaneous bacterial peritonitis (4.5%), hepatic encephalopathy (13.6%), or gastrointestinal bleeding (9.1%). All patients were screened on admission as they belong to a defined risk group to be colonized with NCR-DRGN. Characteristics of NCR-DRGN-positive patients are depicted in Table 1.

Over a 32-months period, 22 out of 182 (12.1%) screened patients tested positive for at least one NCR-DRGN organism. A total of 44 distinct NCR-DRGN isolates were identified, including Escherichia coli (n = 40), Klebsiella pneumoniae (n = 2), Klebsiella oxytoca (n = 1), and Morganella morganii (n = 1) (see Table 2). Three E. coli isolates and one K. pneumoniae isolate were excluded from sequencing analysis because they could not be regrown from the in-house isolate collection. All isolates showed phenotypic resistance against third-generation cephalosporins but not against carbapenems. One particular E. coli isolate (isolate 33) was initially classified as NCR-DRGN having an ESBL phenotype, while WGS performed later in this study revealed the presence of bla_CTX–M–15 and bla_OXA–244 gene (Figure 1 and Table 3).

Bioinformatics analysis of genome assembly yielded an average N50 value of 200 kb, with most genomes comprising fewer than 150 contigs. The majority of isolates (37 isolates obtained from 16 patients) were classified as E. coli. Distinct sequence types (STs) were observed among patients, with ST131 present in multiple patients (P2, P4, P12, P13, and P24). In addition, ST69 was identified in three (P13, P15, P18) and ST38 in two patients (P6, P22). Furthermore, three non-E. coli species were identified: (i) M. morganii (not typeable, P7), (ii) K. oxytoca (ST271, P14), and (iii) K. pneumoniae (ST307; P11) (Figure 1 and Table 2). The majority (n = 7/10) of E. coli STs were represented by only one particular isolate or multiple isolates (“copy strains”) from the same patient (Figure 1). Among the remaining STs, ST131 was most frequently detected in five patients, and, when including ST13730 (belonging to the clonal complex CC131), even in six patients. This variability in STs suggests multiple independent sources of bacterial introduction (except for P6 and P22) rather than inter-patient transmission.

A number of 15 different β-lactam-resistance genes were detected. One isolate from one patient (isolate 33, P23) carried a bla_OXA–244 gene, indicating impaired carbapenem susceptibility. All remaining isolates lacked the presence of carbapenemase genes. For all other isolates, no differences between the phenotypic or genotypic resistance profiles were detected (see Table 3).

In total, 12 ESBL-genes from the bla_CTX–M-, bla_OXA-, bla_OXY-, bla_SHV- and bla_TEM-type, two AmpC genes (bla_CMY–4, bla_DHA–13) and one carbapenemase gene (bla_OXA–244) were found, mostly predicted as plasmid-borne, indicating the possibility of intra- or interspecies transmission. Additionally, the following resistance genes were found spread among all isolates: 11 different genes for aminoglycoside-resistance (aac, ant, aad, aph), one for rifamycine resistance (arr), three for chloramphenicol resistance (cat, flo), four for diaminopyrimidine resistance (dfr), three for fosfomycin resistance (fos), one for lincosamide resistance (Inu), three for tetracycline resistance (tet), two for macrolides resistance (mph), two for fluoroquinolone resistance (qnr), three for sulphonamide resistance (sul) and the rest represented by other efflux pumps (mef, oqx) with one (mdf) coding for a multidrug resistance protein conferring resistance against rifampicin, tetracyclines, chloramphenicol, erythromycin, aminoglycosides and fluoroquinolones as well as benzalkonium, a quaternary ammonium compound disinfectant. Figure 1B provides an overview of the detected resistance genes associated with the isolates and their genomic localization (plasmid or chromosomal).

An allele difference of ≤ 10 was defined as clustering threshold indicating closely related isolates. The observed allele differences for isolates within clusters range between 0 and 2, while the closest intercluster allele difference was 55. Clusters were further supported by pangenome analysis (Figure 2). The cgMLST analysis highlighted the presence of multiple STs among the E. coli isolates (Figure 1). While certain STs were unique to individual patients, others were shared among multiple patients. Notably, although MLST identified ST131 in 14 isolates originating from five patients (Table 1), cgMLST analysis demonstrated significant genetic diversity within these isolates, indicating multiple independent introductions rather than direct transmission (Figure 1). Again, ST38 was observed in both P6 and P22, forming a distinct cluster in the minimum spanning tree (MST) which suggests transmission of NCR-DRGN bacteria between these two ACLD patients (no allele differences nor core gene single nucleotide variants in isolates 39–42). These findings underscore the genetic variability among isolates and the potential transmission dynamics within the study population.