The study was retrospective and performed at the Hospital Universitario La Paz (HULP) in Madrid, Spain; a 1,200 bed tertiary care center. The pediatric patients included in this study were all those that had at least one stay at the pediatric Intensive Care Unit (pICU) of the hospital between April, 2018 and December, 2019. These patients were then followed throughout the duration of the study even when they were transferred, admitted, or visited other hospital wards at HULP. The samples received from these patients were originally collected for routine epidemiological surveillance and there wasn’t any sample that was collected specifically for this study. All the samples were re-used after fulfilling their diagnostic purposes and were analysed in bulk. Also, no results were communicated in real time to neither the clinicians nor the patients, and therefore the results obtained from this study did not interfere with the clinical management of the patients involved. Pediatric liver transplant patients were excluded since they received extensively more antibiotics than other patient groups and were included in a previous study performed by our group (Dahdouh et al., 2022).

During the time of the study, rectal swabs were routinely collected and sent to the microbiology department of HULP to determine the presence of ESBL, OXA-48, and VIM producers was implemented due to an ongoing outbreak caused by carbapenem-resistant K. pneumoniae that started in 2010 and has become endemic at the hospital (Pérez-Blanco et al., 2018). The rectal swabs were received from patients whose health condition allow for taking the swab and were initially cultured on MacConkey plates containing 4µg/mL cefotaxime (custom made by Maim^®, Madrid, Spain) for the detection of ESBL producers, and on ChromID CARBA SMART™ agar plates (Biomérieux^®, Marcy l’étoile, France) for the detection of carbapenemase producers. This was part of the routine epidemiological surveillance performed at HULP. After reaching the clinical conclusions, the same swabs were then collected to be reused in this study. Their contents were suspended in 0.5 mL of TE buffer (10mM Tris and 1mM EDTA; pH = 8.0) through vigorous shaking. The suspension was then stored at -20°C until used, and analysed in batches when a sufficient number of samples was reached. In the cases where the isolates obtained from the rectal swabs were stored in the hospital’s bacterial collection, they were recovered. Moreover, all the MDRO isolates that were obtained from clinical samples of the patients included in this study (such as urine, blood, abscesses, etc…), and that were stored in the bacterial collection as part of HULP’s routine practice were recovered to be used in this study. These latter isolates are labelled “extra-intestinal isolates” throughout the rest of the article.

100µL of the rectal swab suspensions were diluted with 900µL of TE buffer in order to avoid inhibition of the qPCR reactions. Total DNA was then extracted by heating this suspension at 95°C for 20 minutes, performing mechanical lysis at 7,000rpm for 70 seconds using MagNA Lyser Green Beads and MagNa Lyser™ (Roche^®, Mannheim, Germany), and extracted using the MagNA Pure Compact Nucleic Acid Isolation Kit I and MagNA Pure Compact™ system (Roche^®, Mannheim, Germany). The DNA was then stored at -20°C.

The qPCR assays designed in one of our previous studies were used to determine the Relative intestinal Loads (RLs) of the beta-lactamase genes that are endemic at HULP (bla _{CTX-M-1-Family}, bla _OXA-1, bla _OXA-48, and bla _VIM) (Dahdouh et al., 2022). Briefly, reaction mixtures containing 10µl 2X PowerUP™ SYBR^® Green Master Mix (Applied Biosystems, Waltham, MA, USA), 0.05µM of each respective primer pair, 6µl H₂O and 2µl DNA were prepared. The CFX Connect™ Real-Time System (BioRad, Madrid, Spain) was used for thermal cycling under the following conditions: 95°C for 3 minutes, then 40 cycles of 95°C for 15 seconds and 60°C for 1 minute; followed by melting curve analyses.

RLs of the tested genes were determined in relation to the C_t values of the 16SrDNA gene and normalized to the RLs of these genes in pure bacterial cultures using the 2^-ΔΔCt method (Livak and Schmittgen, 2001). The RLs were then transformed into an inverse logarithmic scale where 0 represents RLs that are equivalent to pure bacterial cultures, –1 represents 10% of the bacterial population collected in the rectal swab, –2 represents 1%, and so on until –6 which represents 0.0001% of the total bacterial population, and our detection limit. The values are also expressed (where relevant) as percentages (%RL) where the values were converted to percentages using the formula 10^RL × 100.

Clonality analyses were done using Random Amplified Polymorphic DNA (RAPD) (Hsueh et al., 1998). The primers used were OPA-2 (5’-TGCCGAGCTG-3’), OPA-12 (5’-TCGGCGATAG-3’), and OPA-18 (5’-AGGTGACCGT-3’) and the thermal cycling conditions were 94°C for 5 minutes, followed by 45 cycles of 94°C for 1 minute, 37°C for 1 minute, and 72°C for 2 minutes, and a final step at 72°C for 2 minutes. PCR products were then visualized on 2% agarose gels and the resulting profiles obtained were used to determine the clonal relatedness of the isolates.

Whole-genomes were sequenced for representative isolates from the most frequent clones (as determined by RAPD). The isolates were cultured overnight on blood agar plates at 37°C and DNA was extracted from 10 single colonies as described previously (section 2.2). The NEBNext^® Fast DNA Fragmentation & Library Prep Set for Ion Torrent™ (New England Biolabs, Ipswich MA, USA) was used to prepare the libraries according to the manufacturer’s instructions. Mag-Bind^® TotalPure NGS beads (Omega Bio-Tek, Norcross GA, USA) was used for the purifications step, and the Ion Chef^® and S5^® Gene Studio (ThermoFisher Scientic, Waltham MA, USA) were used for sequencing. The reads were assembled using Geneious (v.10.1.3, Biomatter Ltd., Auckland, New Zealand). They were also tested for the presence of antibiotic resistance determinants and plasmids, and were allocated a specific Sequence Type (ST) using the tools available on the Center for Genomic Epidemiology website (https://cge.cbs.dtu.dk ).

16SrDNA metagenomic sequencing was performed to rectal swab suspensions that had varying RLs towards the tested genes, and the samples were labelled S1 through S40. Samples S1 to S29 were positive to at least one of the genes tested for, and were ordered in decreasing values of RLs while samples S30 to S40 were negative to all 4 tested genes. The Ion 16S™ Metagenomics Kit and the Ion Plus™ Fragment Library Kit with the Ion Xpress Barcode Adapters (Thermo Fisher, USA) kits were used to prepare the libraries according to the manufacturer’s instructions. Sequencing was performed using the same instruments as above and the sequences were analyzed using ION Reporter™ (Thermo Fisher, USA). All the sequences obtained from this study were deposited in Genbank under the Bioproject with the accession number PRJNA911601.

Patient data (age, gender, colonization status, extra-intestinal infections 14 days before or after rectal swab collection, colonisation status upon sample collection, and antibiotics received) were retrieved from the hospital’s information system database. The patients were given study-specific codes in such a way that no patient can be identified. Approval of HULP’s local ethics committee was obtained for the study (PI-3428).

Normality of the data was tested for using the Kolmogorov-Smirnov (if n>50) and Shapiro-Wilk (if n<50) tests. The RLs of the four tested genes were grouped based on whether or not they correspond to an extra-intestinal infection, whether or not the patients received antibiotics in the 30 days prior to sample collection, and whether or not they correspond to detection of MDROs using the routine screening cultures performed at HULP. Two-sided student T-tests and ANOVA were used for normally distributed quantitative data, and Kruskal-Wallis and Mann-Whitney U tests were used for non-normally distributed quantitative data. Chi-squared was used for the comparison of categorical variables such as antibiotic consumption in relation to having one or more β-lactamase genes detected in the rectal swab. Receiver Operating Characteristic (ROC) analyses were performed using the easyROC online tool (http://biosoft.erciyes.edu.tr/app/easyROC/) in order to determine cutoff values after which there is an increase rate of detection of extra-intestinal MDROs harbouring the same gene as that detected in the rectal swabs. The Youden method (i.e. determining the J value through the formula [sensitivity + specificity – 1] for all the points of the curve) was used in order to determine the cutoff values. Statistical significance was determined when p values were less than 0.05 and all the statistical tests were performed using SPSS (version 24.0; IBM, Armonk, NY, USA).

The general characteristics of the patients and the samples included, as well as the qualitative results obtained by routine cultures and qPCRs are presented in Table 1 . Of the 90 patients included in this study, 42 were chronic. Of these chronic patients, 19 were transplant patients (9 kidney, 4 multi-organ, 3 hematopoietic, and 3 cardiac transplants). Of the 382 rectal swabs obtained from the 90 patients included in the study, 317 (82.98%) were collected from chronic patients and 65 (17.02%) were collected from acute patients (2 or less swabs from each acute patient that had average hospital stays of less than 1 week). Sixty-seven patients from which 340 rectal swabs were collected had at least one swab that was positive for at least one of the genes tested by PCR. Thirty-eight of these patients had both ESBLs and carbapenemases, 18 had only carbapenemases, and 11 had only ESBLs. Just 3 of the 23 patients that were negative for these genes where chronic patients (15 (4.73%) samples), and the rest were acute patients (40 (61.54%) samples). By routine cultures, 28 patients had at least one sample positive for ESBLs, 24 for carbapenems, and 10 for both carbapenems and ESBLs.

K. pneumoniae was the most frequent species isolated from the rectal swabs ( Table 1 ). To study the relationship between the different isolates, clonality analyses through RAPD was performed for 43 isolates that could be recovered from 21 patients. The most frequent clone comprised 20 isolates originating from 9 different patients, and was found (through whole-genome sequencing) to belong to sequence type (ST) ST39. Another clone was found to belong to ST37 and was detected in 7 isolates from a single patient. A third clone was detected in 2 isolates obtained from a single patient and belonged to ST3155, and a fourth clone that belonged to ST1928 was also detected in 2 isolates obtained from another patient. The rest of the isolates (twelve isolates from nine patients) included in the clonality analyses did not have a profile that was similar to any other isolate.

In total (out of the 382 rectal swabs), 136 (35.6%) were positive for bla _{CTX-M-1-Family}, 109 (28.5%) for bla _OXA-1, 71 (18.6%) for bla _OXA-48, and 134 (35.1%) for bla _VIM by PCR. Compared to the PCR, the sensitivity of the routine cultures for the detection of OXA-48 producers was 51.52% and its specificity 98.39%. For VIM producers, the sensitivity was 38.1% and the specificity 98.98% ( Supplementary Table 1 ). ESBL-producers were not included in this analysis since ESBL-harboring E. coli were reported as negative in the hospital’s information system database, according to the activated protocols at the time of the study.

The RLs of the four β-lactamase genes endemic at HULP (bla _{CTX-M-1-Family}, bla _OXA-1, bla _OXA-48, and bla _VIM) were determined through normalizing the Ct values of these genes to that of the 16SrDNA gene. The RLs of the four genes measured in all the swabs fell within a six log range, from almost pure cultures (0; 100% of bacterial population) to -6 (0.0001% of the total bacterial population). The median RLs of bla _{CTX-M-1-Family} and bla _OXA-1 were significantly higher than those of bla _OXA-48 and bla _VIM (-1.7 (1.98%) and -1.79 (1.64%) versus -2.49 (0.32%) and -2.41 (0.4%), respectively; P < 0.05; Figure 1 ). The RLs determined for bla _OXA-48 and bla _VIM were more evenly spread out across the detection ranges, but showed a bi-modal distribution pattern where they clustered in either high RLs (over -2 (i.e. 1% of the total bacterial population)) or low RLs (under -4 (i.e. 0.001% of the total bacterial population)) ( Figure 1 ). Samples that were PCR-positive for bla _{CTX-M-1-Family} and bla _OXA-1 often showed RLs that were higher than -2 (1% of the total bacterial population). There was no difference in RLs between male and female patients, nor was there any difference based on the age of the patient, the type of disease for which the patient was receiving treatment, the wards in which the patients were admitted to at the time of sampling, and whether or not the patients were critically ill.

There were 75 (70.1%), 32 (45.1%), and 78 (58.2%) swabs positive by PCR for bla _{CTX-M-1-Family}, bla _OXA-48, and bla _VIM genes, respectively, that were negative in antibiotic-selective cultures. However, according to HULP’s protocols at the time of the study, ESBL-harboring E. coli were reported as negative in the hospital’s information system database, and therefore the total number of positive cultures for ESBL-producing organisms cannot be known. Comparing the RLs of the carbapenemase genes to the results of the routine cultures showed that the median RLs of bla _OXA-48 and bla _VIM were significantly higher in samples that were positive by PCR and culture as compared to those only positive by PCR (-1.45 (3.2%) and -0.82 (15.1%) versus -4.38 (0.004%) and -4.02 (0.01%); respectively; P<0.05; Figure 2 ). Moreover, the bi-modal distribution pattern disappeared where the RLs of the samples that were positive through both methods clustered in the higher RL ranges while those that were only positive by PCR clustered in the lower RL ranges ( Figure 2 ). The bla _OXA-1 gene was not included in this analysis since it was not screened for in the routine cultures.

Antibiotic consumption data was retrieved from the hospital’s information system database for the patients that were included in this study throughout the duration of the study, and 30 days before inclusion. Twenty one (23.33%) patients did not receive any antibiotic throughout the study period, and in 37 (41.11%) patients, all the samples collected from them corresponded to them receiving antibiotics during the previous 30 days. These treatments were not consecutive and were often mixed where different antibiotics were received at different times, but they always received at least one type in the 30 days before sampling. The remaining 32 (35.56%) patients were mixed where some swabs corresponded with antibiotic consumption during the 30 days before sample collection, and others did not. From the first group of 21 patients, 42 (10.99%) rectal swabs were collected, and from the second group of 37 patients 139 (36.39%) swabs were collected. From the third group of mixed patients, 138 (36.12%) rectal swabs corresponded with antibiotic consumption and 63 (16.49%) of them did not correspond with the patient receiving antibiotics in the previous 30 days.

In total, 277 (72.51%) of the rectal swabs corresponded with antibiotic consumption during the 30 days prior to collection, and 105 (27.49%) did not. For 72 rectal swabs, the patients received one type of antibiotic, for 40 rectal swabs the patients received antibiotics from 2 different classes, and for the remaining 165 swabs the patients received 3 or more antibiotics from different classes. Table 2 shows the type of antibiotics that the patients received before sample collection. In the majority of the incidents (82%), the antibiotics were administered intravenously, in 14.5% of the incidents antibiotics were received orally and intravenously, and in 3.5% of the incidents the antibiotics were only administered orally.

The consumption of carbapenems, non-carbapenem β-lactams, and glycopeptides was significantly associated with testing negative for bla _{CTX-M-1-Family} and bla _OXA-1 (P < 0.05). Consumption of aminoglycosides and Trimethoprim/Sulfamethoxazole (TMX) was also significantly associated with testing negative for bla _OXA-48 (P < 0.05). Additionally, the consumption of TMX was significantly associated with lower median RL values for bla _OXA-48 (-1.82 (1.5%) versus -4.45 (0.004%); P < 0.05). No statistically significant associations were determined for the rest of the antibiotic classes and the other antibiotic resistance genes.

The RLs determined for the four genes tested for were compared to incidents where extra-intestinal isolates were collected within 14 days of rectal swab collection. Thirty-three extra-intestinal β-lactam resistant isolates have been registered in the hospital information system from 18 patients included in the study. Fourteen isolates were obtained during the two weeks before the collection of the rectal swab, seven on the same day, and twelve during the two weeks after the collection of the rectal swabs. These incidents were at least 2 weeks apart when collected from the same patient in order to exclude ongoing infections caused by the same isolate. Twenty of the extra-intestinal isolates were related to infections while the remaining thirteen were considered colonisations, as registered by the attending physician at the time ( Table 3 ). In 28 of these episodes (84.85%), the patients received antibiotics up to 30 days before extra-intestinal detection of MDROs. It was possible to compare the extra-intestinal isolates to the intestinal ones in only 21 of these cases (13 of them were infections) since in the remaining cases, no growth was reported through routine culture from the corresponding rectal swab. In 16 of these 21 (76.2%) cases the same organism was detected in both the rectal swabs and the extra-intestinal site. Clonality analyses through RAPD comparing these isolates was only possible for 9 incidents (where the isolates were stored in the hospital’s bacterial collection), and in 8 of them (88.9%) the same clone was detected. In terms of the β-lactam resistance genes, in out 27 of the 33 isolates (81.8%), the same gene was detected in the intestinal (detected through qPCR) and extra-intestinal sites (registered in the hospital’s database).

The intestinal RLs of the antibiotic resistance genes collected within 2 weeks (average = 2 days; median = same day) from the extra-intestinal episode had median RLs of -1.36 (4.4%) for bla _{CTX-M-1-Family}, -1.48 (3.5%) for bla _OXA-1, -1.05 (8.9%) for bla _OXA-48, and -1.97 (1.1%) for bla _VIM. The RLs of bla _{CTX-M-1-Family} and bla _OXA-48 were significantly higher in the samples close to an episode of extra-intestinal spread than in samples that were not associated to extra-intestinal episodes (-1.36 (4.4%) and -1.05 (8.9%) versus -1.85 (1.4%) and -3.25 (0.06%), respectively; P < 0.05; Figure 3 ). No statistical significance was detected for bla _VIM. There were 11 extra-intestinal samples that corresponded with reporting of a culture-negative rectal swab. Seven of these samples corresponded with infections (bacteremia, peritonitis, and lower respiratory tract infections), and in all of them except for one, the same gene was detected in the rectal swab through PCR compared to the gene reported in the extra-intestinal isolate. In the seventh sample, no antibiotic resistant gene was detected in the rectal swab and an ESBL-producing organism was reported to cause ventilator associated pneumonia. Also among these 11 samples, 7 corresponded with the patient receiving antibiotics and the rest did not correspond with antibiotic consumption during the 30 days before sample collection. In the 4 remaining episodes, nothing was detected by PCR in the rectal swabs and all 4 episodes were registered as asymptomatic bacteruria.

ROC analysis (Youden method) of the capacity of the RLs to identify samples close to an extra-intestinal spread episode determined a cut-off value of 6.5% for organisms harbouring bla _OXA-48 (area under the curve = 0.89, sensitivity = 0.89, specificity = 0.87; P < 0.05; Figure 4C ). In none of the other three genes the intestinal RL could identify samples close to extra-intestinal spread episodes ( Figure 4A, B, D ).

Metagenomic analyses (16SrDNA) were done with the same rectal swab suspensions used for qPCR in a subset of forty selected samples (S1 to S40) ( Figure 5 ). Figure 5B shows the same samples but where only Klebsiella spp. (red) and Pseudomonas spp. (green) are highlighted. These two organisms were chosen since, according to the local epidemiology at HULP, were the most likely organisms to harbour the genes tested for in our study. The observed number of genera, α-diversity (calculated according to the Shannon method), and percent relative loads (%RLs) of the four genes tested for in these samples are shown in Supplementary Table 2 .

Comparing the positive samples to the negative ones, a higher diversity among the negative samples was observed (median of 31 genera versus 15 genera; P>0.05). Moreover, looking at the percentage of Klebsiella spp. and Pseudomonas spp. ( Figure 5B ), it is evident that all but 4 samples that were positive for at least one of the tested genes were dominated by one or both of these organisms, while in all but 2 samples that were negative for these genes these organisms were almost absent. Of the 4 positive samples that showed an exception, S1 was dominated by Enterococcus spp., S7 and S24 were relatively diverse, and S21 showed dominance by Acinetobacter spp. It is also interesting to note that the negative samples S38 and S39 were dominated by Enterococcus spp. and S40 was dominated by Acinetobacter spp. For samples S21 and S24, ESBL- and OXA-48-producing K. pneumoniae were registered in the hospital’s information system database, respectively. For samples S1, S38, S39, and S40, no record of a β-lactam resistant organism in the rectal swabs was registered in the hospital’s database system.