Oral and Rectal Colonization by Antimicrobial-Resistant Gram-Negative Bacteria and Their Association with Death among Residents of Long-Term Care Facilities: A Prospective, Multicenter, Observational, Cohort Study

Antimicrobial resistance (AMR) is a global concern [1]. In particular, the prevalence of third-generation cephalosporin-resistant Escherichia coli, including extended-spectrum β-lactamase (ESBL)-producing Enterobacterales, has been increasing every year [2]. A recent study showed that in 2015, a higher number of deaths in the EU and European Economic Area were attributable to third-generation cephalosporin-resistant E. coli infections than to other antibiotic-resistant bacterium combinations [3].

Studies have reported that the proportion of carriers of ESBL-producing Enterobacterales among healthy people varies geographically; this proportion is lower than 10% in Europe but higher than 20% in Asia [4]. Meanwhile, residents of long-term care facilities (LTCFs) are known to be reservoirs of antimicrobial-resistant bacteria (ARB) [5‒7]. For example, the proportion of carriers in European LTCFs ranges from 2.5% in the Netherlands to 60% in Italy. Alternately, the proportions in LTCFs of the USA, Oceania, and Asia are in the ranges of 2.6–32.5%, 1.7–16.2%, and 21.8–46.9%, respectively [8]. These studies indicate that the proportion of carriers among healthy people and LTCF residents is higher in Asia than in other regions. Moreover, Rodríguez-Villodres et al. [9] reported that Asia had the highest proportion (71%) of ESBL-producing Enterobacterales. Notably, Japan has the highest global population (approximately 30%) of over 65-year-old individuals [10]. Therefore, the distribution of ARB in LTCFs in such countries is of global interest.

Japan has a national AMR surveillance system called Japan Nosocomial Infections Surveillance (JANIS), one of the world’s largest AMR surveillance systems. It comprehensively collects all routine bacteriological testing data from more than 2,000 medical institutions (more than 1/4 of all medical institutions) [11]; however, it does not collect data from LTCFs. Therefore, the prevalence of ARB in Japanese LTCFs remains to be evaluated. Furthermore, although oral pathogenic microorganisms, including ARB, are likely to cause pneumonia [12], the presence of ARB in the human oral cavity had been unexplored until our recent study in Japanese LTCFs [13]. In addition, the effect of ARB colonization on the clinical outcome of LTCF residents remains unreported.

In this study, we determined the colonization of both oral and rectal ARB in LTCF residents based on the genomic sequences of all isolated strains. We also evaluated the relationship between these bacterial strains and the clinical outcomes of the residents.

The elderly care system in Japan provides services with the long-term care insurance system through different facilities, including Geriatric Health Service Facilities and Welfare Facilities for the Elderly Requiring Long-term Care. These facilities have different personnel standards and professional collaborations [14]. In the present prospective, multicenter, observational, cohort study, we determined the carriage of oral and rectal ARB in 178 residents of six LTCFs, including three Welfare Facilities for the Elderly Requiring Long-term Care and three Geriatric Health Service Facilities in Hiroshima, Japan. We evaluated the clinical outcomes of these residents (dead or censored at the end of this study). Oral and rectal samples were collected using swabs from 2019 to 2020, and follow-ups of the study participants were completed by July 2021. The clinical data of the participants were obtained from their medical records and nursing care plans. The data included demographics (age, sex, and unit of residence), Eastern Cooperative Oncology Group (ECOG) performance status, comorbidities, prior antibiotic use within 6 months before the sample collection, enteral nutrition, length of stay in facilities in days (from admission to sample collection), survival time in days (from sample collection to participant discharge or the end of the follow-up period), and mortality.

We screened for ARB carriers by obtaining oropharyngeal and rectal swabs from all study participants and spreading them directly onto the CHROMagar ESBL medium plates (Kanto Chemical, Japan). This screening medium is used to rapidly and presumptively identify ESBL-producing Enterobacterales [15]. It has also been used to isolate Acinetobacter and Pseudomonas spp. [13]. We incubated the plates at 37°C for 24 h. A single bacterial colony from each positive sample was further grown on new CHROMagar ESBL medium plates and used for subsequent testing. Colonies were identified by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (Bruker Daltonics, Bremen, Germany). Eventually, bacterial susceptibility to antimicrobials was determined using a MicroScan WalkAway (Beckman Coulter, Brea, CA, USA). Biofilm formation assays were performed as described in our previous study [13].

Bacterial cells were lysed with 0.5 mg/mL lysozyme and 2% SDS, and their genomic DNA was purified from the lysate using AMPure XP (Beckman Coulter). Subsequently, we prepared DNA libraries as described previously [16] and performed paired-end sequencing (2 × 150 bp) on the Illumina HiSeq X Five platform (Macrogen Japan Corporation, Tokyo, Japan). Genome reads for P. aeruginosa and E. coli strains isolated and analyzed in this study (online suppl. Tables S1 and S2; for all online suppl. material, see www.karger.com/doi/10.1159/000525759) were deposited in the DNA Data Bank of Japan (DDBJ) and National Center for Biotechnology Information (NCBI) under BioProject accession number PRJDB12072.

The read data of each isolate were used for de novo genome assembly using Shovill [17]. We examined the presence or absence of AMR genes in the genome of each isolate using an in-house script that conducted a BLASTn search against the NCBI Bacterial AMR Reference Gene Database (https://www.ncbi.nlm.nih.gov/bioproject/PRJNA313047). We determined that a BLAST match of ≥90% and identity of over ≥90% of the locus length indicated the presence of a bla_CTX-M gene in ESBL-producing Enterobacterales.

We conducted a phylogenetic analysis using the genomic data of 43 P. aeruginosa strains isolated from Hiroshima, Japan, and sequenced using Illumina MiSeq in previous studies. These strains included strain 71E isolated from an LTCF [13], strain ST235 carrying the bla_IMP-1 integron [18], and 41 strains isolated from several hospitals. These strains did not include a rectal sample-derived P. aeruginosa strain lacking genome sequence data. Genomic read data for these strains (online suppl. Table S1) were deposited in the DDBJ and NCBI under BioProject accession number PRJDB12075. The reads for each isolate were used for de novo genome assembly, as described in the preceding section.

In addition, we downloaded all 49 publicly available assembled genome sequences of P. aeruginosa strains that were isolated in Japan and annotated as good-quality sequences from the PATRIC database [19]. Similarly, we downloaded 1,651 genome sequences for E. coli strains. We then selected a subset of 324 strains based on multilocus sequence typing [20] and phylotypes [21], such that each selected strain represented a combination of multilocus sequence typing and phylotype.

We annotated each assembled genome sequence using Prokka [22] and conducted a concatenated core-genome alignment using the Panaroo pipeline [23]. Subsequently, we separately generated a maximum-likelihood tree using RAxML version 8.2.10 [24] for P. aeruginosa and E. coli and illustrated the trees and metadata using iTOL [25]. The P. aeruginosa and E. coli strains used in this analysis are listed in online supplementary Tables S1 and S2.

Univariate analysis of the association between the clinical characteristics of participants and detection of ARB was conducted using the χ² test for discrete characteristics and Wilcoxon’s rank-sum test for continuous characteristics. For each clinical characteristic identified to be significant according to the univariate analysis, the χ² test and Wilcoxon’s rank-sum test were used to identify confounding factors that are significantly associated with it and detection of ARB (p < 0.05). Subsequently, multiple logistic regression was conducted to obtain the p value after adjusting for covariates. In addition, univariate analysis of the association between participant survival and the detection of ARB was performed using the Kaplan-Meier method, log-rank test, and Cox regression analysis. Confounding factors were similarly explored, and multiple Cox regression analysis was performed to obtain the p value and hazard ratio as the exponential estimate of regression coefficient after adjusting for covariates. All statistical analyses were performed using the R software (version 4.0.5) and JMP Pro version 13 (SAS Institute, Cary, NC, USA).

Oral and rectal samples were collected from 178 participants, of which 146 were followed up (Fig. 1). The clinical characteristics of all participants, stratified by the two kinds of LTCFs they lived in, are shown in Table 1. The average age of the participants was 87.0 years (range, 44–104 years), and 36 (20.2%) of the participants were males. Of all participants, 90.0% had an ECOG performance status of 4. Long-term enteral nutrition therapy was administered to 23 participants (12.9%), all of whom resided in “Welfare Facilities for the Elderly Requiring Long-term Care.” Percutaneous endoscopic gastrostomy was performed in 15 of the 23 participants (65.2%). ESBL-producing Enterobacterales and P. aeruginosa were detected in 42.7% (n = 76) and 2.8% (n = 5) of the rectal swabs and 5.6% (n = 10) and 3.4% (n = 6) of the oral swabs, respectively. Additionally, Citrobacter spp. were detected only in the rectal swabs.

We identified 41 and 127 ARB strains from the oral and rectal swabs, respectively (Table 2). These isolates belonged to the two bacterial genera Acinetobacter and Pseudomonas and the order Enterobacterales. The most common oral bacterial species were Acinetobacter spp. (15 isolates, 36.6%), P. aeruginosa (7 isolates, 17.1%), and E. coli (6 isolates, 14.6%). In contrast, the most common rectal bacterial species were E. coli (77 isolates, 60.6%), E. cloacae complex (9 isolates, 7.1%), and Citrobacter spp. (8 isolates, 6.3%). P. aeruginosa was the most detected in the oral cavities of the residents of “Welfare Facilities for the Elderly Requiring Long-term Care.”

Univariate analysis of the association between the clinical characteristics (Table 1) and ARB detection is summarized in Table 3, which revealed that the detection of both ESBL-producing Enterobacterales and P. aeruginosa was associated with enteral nutrition in participants. Remarkably, participants subjected to enteral nutrition had a significantly higher proportion of ESBL-producing Enterobacterales (26 vs. 7%, p = 0.002, χ² test). Similarly, participants on enteral nutrition were more frequently positive for P. aeruginosa than those without enteral nutrition (64 vs. 12%, p < 0.001). Moreover, there were no confounding factors regarding the clinical background of each participant that were related to both enteral nutrition and the presence of ESBL-producing Enterobacterales or P. aeruginosa (p < 0.05). Further, multiple regression analysis after adjusting for age, gender, and facility difference confirmed that there was almost no change in p value compared with that obtained using univariate analysis (Table 3). Another significant association was found between the detection of ESBL-producing Enterobacterales and prior use of antibiotics within 6 months (64 vs. 39%, p = 0.003, χ² test). Similarly, there were no confounding factors regarding the clinical background of each participant, and multiple regression analysis adjusting for age, gender, and facility difference confirmed that there was almost no change in p value (Table 3).

Next, we evaluated the relationship between participant survival time and their clinical background, including the detection of ARB (Table 4), and found a clear difference in terms of P. aeruginosa (Fig. 2). However, there was no significant difference between the two types of facilities (p = 0.43, log-rank test). Before testing the statistical significance, the comparison of data in Tables 3and4 confirmed no confounding factors associated with participant survival and P. aeruginosa detection (at p < 0.05). However, for the subsequent multiple regression analysis, we considered enteral nutrition as a covariate because it was possibly associated with participant survival (p = 0.169, Table 4, Cox regression) and P. aeruginosa detection (p < 0.0001, Table 3, χ² test). Age was also a covariate, as it was possibly associated with both participant survival and P. aeruginosa detection (p = 0.056, Table 4, Cox regression, and p = 0.147, Table 3, Wilcoxon’s rank-sum test). Although 86% of the deaths occurred in the “Welfare Facilities for the Elderly Requiring Long-term Care,” the differences in participant survival curves were insignificant between both types of LTCF (p = 0.43, log-rank test).

As shown in Table 5, multiple Cox regression analysis with age and enteral nutrition as covariates revealed a statistically significant difference in survival between the participants positive for P. aeruginosa and the other participants (p = 0.002, hazard ratio 5.7 [95% confidence interval, CI: 1.9–17.3]). On the contrary, a similar multiple Cox regression analysis on the detection of ESBL-producing Enterobacterales did not reveal statistical significance (p = 0.05, hazard ratio 0.48 [95 CI: 0.23–1.00]).

Genome sequencing of all the isolated strains enabled the construction of core-genome phylogeny (including other strains isolated in Japan to provide context to our findings) using metadata from the strains. The core-genome phylogeny of P. aeruginosa (Fig. 3) revealed that the isolated strains were widely distributed across the phylogeny. The most frequently detected ST was ST235 (4 out of 11 isolated strains). We found that the ST235 strains harbored the β-lactamase encoding the bla_OXA-488, bla_PDC-35, and bla_PAO gene variants. The cluster of ST235 strains (Fig. 3) included a 71E strain isolated from another LTCF in Hiroshima, Japan [13], suggesting clonal evolution of these strains in LTCFs of that geographical region. This strain was previously shown to be a high biofilm producer [13]. The four ST235 strains isolated in this study that were phylogenetic neighbors to the 71E strain also had significantly higher biofilm-producing abilities than the other seven strains (median OD_{590 nm} is 2.3 vs. 0.1, p < 0.028, Wilcoxon’s rank-sum test). The OD_{590 nm} values of the strains were measured using Luria-Bertani medium in the absence or presence of 1% glucose (Spearman’s correlation coefficient 0.93 between the two conditions) and are shown in online supplementary Table S1.

In contrast, the core-genome phylogeny of E. coli (Fig. 4) revealed a clearly localized phylogenetic distribution of the strains isolated in this study (red in Fig. 4). Most of these strains were ST131 belonging to phylotype B2. The phylogenetic distribution of bla_CTX-M genes directly responsible for the ESBL phenotype is shown in the outermost part of Figure 4: bla_CTX-M-27, bla_CTX-M-15, bla_CTX-M-55, and bla_CTX-M-14 had 48.8%, 28.8%, 17.1%, and 6.1% frequencies, respectively.

Our study showed that 54.5% (97/178) and 42.7% (76/178) of the participants were carriers of ARB and ESBL-producing Enterobacterales, respectively. The proportion of ESBL-producing Enterobacterales carriers in this study (42.7%) was higher than that reported in previous studies conducted in Western countries and 20% higher than the global average [8] (see online suppl. Table S3 for more information from previous studies). Similarly, a Chinese study in 2014 reported a high proportion (46.9%) of ESBL-producing E. coli in stool samples [26], suggesting that East Asia harbored a consistently high proportion of ESBL carriers.

A Japanese study in 2013 reported that 19.6% of stool samples obtained from three LTCFs in Osaka harbored ESBL-producing Enterobacterales [27], which is lower than that identified in the present study. A review of 134 studies on LTCFs published from 1987 to 2020 revealed that the increase in the percentage of carriers of ESBL-producing Enterobacterales started in 2015 [9]. Nevertheless, the molecular mechanisms underlying this increase remain unknown.

The present study is based on our previous study on oral colonization by antimicrobial-resistant gram-negative bacteria among LTCF residents in Hiroshima, Japan [13]. Compared to our previous study, this study examined both oral and rectal samples from six LTCFs and attempted to determine how colonization by ARB affects the clinical outcomes of participants. Moreover, for the first time, the inclusion of both oral and rectal samples enabled comparative analyses that we discuss in the following paragraphs. Both studies are consistent with respect to the highest proportion of Acinetobacter spp. isolated from oral swabs (Table 2), which is most likely a characteristic reflecting the oral environment of LTCF residents. However, the proportion of P. aeruginosa detected in this study (17.1%) was higher than that detected in the previous study (7%). In contrast, the proportion of any isolate detected from oral swabs of LTCF residents was lower (17.4%) in this study than in the previous study (38%). Such discrepancy might reflect the difference in the extent of oral care or the practice of oral administration of antibiotics among LTCFs. However, further research is required to ascertain the confidence intervals of these statistics and explore such potential differences by including more LTCFs.

In the present study, the two types of Japanese LTCFs exhibited substantial differences in several aspects. For example, rectal samples from 51.4% of the residents of the “Welfare Facilities for the Elderly Requiring Long-term Care” tested positive for ESBL-producing Enterobacterales, a figure that is markedly higher than the 28.4% in “Geriatric Health Service Facilities” (Table 1, p = 0.004, χ² test). In addition, P. aeruginosa was detected in the oral swabs of residents of the “Welfare Facilities for the Elderly Requiring Long-term Care” alone and accounted for approximately 30% of the isolated oral strains (Table 2, p = 0.03, Fisher’s exact test). Furthermore, our study demonstrated that enteral nutrition only conducted in the “Welfare Facilities for the Elderly Requiring Long-term Care” (Table 1) was associated with the detection of ESBL-producing Enterobacterales and P. aeruginosa (Table 3). These data suggest that enteral nutrition is a major factor underlying the difference in the distribution of ARB carriers among the two LTCF types.

Despite its prevalence, the presence of ESBL-producing Enterobacterales was not associated with participant deaths. Reportedly, carriage of ESBL-producing Enterobacterales in stool is associated with the onset of infectious diseases, especially urinary tract infections [28]. However, the present study suggests that carriage of ESBL-producing Enterobacterales does not necessarily contribute to the onset of infectious diseases.

Oral care is an important factor associated with the outcomes of LTCF residents. A recent systematic review [12] demonstrated that E. coli carriage in the human oral cavity is associated with oral hygiene and appropriate oral care. In addition, ensuring a clean oral cavity reduces the chances of developing aspiration pneumonia. Concordantly, it could be inferred that the risk of aspiration pneumonia was lower in the LTCFs investigated in this study because oral care was conducted weekly in the facilities.

Separate analyses for association of oral and rectal P. aeruginosa with the death of participants revealed that 50% (2 [black in Fig. 3] out of 4 [brown in Fig. 3] strains) and 75% (6 [black in Fig. 3] out of 8 [cyan in Fig. 3] strains) of P. aeruginosa strains isolated from rectal and oral samples were associated with participant deaths, respectively. After taking the survival time of each participant into account, the survival analysis revealed a significant difference in survival between participants positive for P. aeruginosa and other participants (Table 5). Furthermore, separate analyses using multiple Cox regression revealed that the estimated hazard ratio was 8.4 (95% CI: 2.2–32.1) and 3.1 (95% CI: 0.82–11.6) for rectally and orally isolated P. aeruginosa, respectively (Table 5). These data suggest that P. aeruginosa had a significant association with the death of participants. Reportedly, P. aeruginosa is associated with patients using nasal gastric tubes [29]. Using a percutaneous endoscopic gastrostomy tube is also one of the risk factors for the existence of AMR bacteria in the oral cavity [13], consistent with the present study, wherein enteral nutrition was performed in 7 out of 11 participants harboring P. aeruginosa. Collectively, these studies indicate that the devices used for enteral nutrition can be colonized by P. aeruginosa, which may be challenging to nullify by oral and skin care. Furthermore, enteral nutrition is performed in patients who face difficulty in oral ingestion and pharyngeal function and is thus a likely cause of aspiration pneumonia [30]. Therefore, participants positive for P. aeruginosa were more likely to have a significantly lower survival rate than others (Fig. 2).

A limitation of the present study is that death due to various causes, including infectious diseases, was defined as the primary outcome in the facilities. Moreover, in Japanese LTCFs, residents at their end stage of life often opt for palliative care rather than receiving further treatment post-hospitalization, and the cause of death is often not determined. In addition, as all the LTCFs included in our study were confined to the same geographical region, our study contains some geographical bias.

Despite these limitations, to our knowledge, this is the first multicenter prospective observational cohort study that seamlessly investigates carriage of both oral and rectal ARB, genomic relatedness and determinants of AMR in isolated strains, and clinical outcomes of LTCF residents. Our study reveals the high carriage of ARB in Japanese LTCFs and the association of P. aeruginosa carriage with participant deaths. Our study provides the first direct evidence of the burden of AMR in LTCFs.

We gratefully acknowledge the contributions and support of Koji Matsumoto, Koji Sumii, Hirofumi Kobayashi, Tatsuko Okuda, and Masanori Yamamoto at Saiseikai Hiroshima Hospital; Katsushi Yamada, Miwa Igarashi, and Hiromi Kawakami at Hamana-sou; Keiji Kanpachi and Satomi Kado at Takane-sou; Takane-sou Koyaura, Tomomi Yonekawa, and Masaru Yonekawa at Hiroshima Kosei Hospital; Yasunori Hirayama at e-House; Miyako Nishiyama at Hiroshima-hakkeien; Katsunari Miyamoto, Raita Yano, and Hayato Nakano at Hiroshima Memorial Hospital; and Shiro Nakai at Kinenju. The computational calculations using genome data were performed at the Human Genome Center at the Institute of Medical Science (University of Tokyo) and at the National Institute of Genetics. We are also grateful to Editage (www.editage.jp) for English language editing.

All the residents admitted to LTCFs during the study period were eligible for inclusion. Written informed consent was obtained from the participants prior to their enrolment in the study. Additionally, we obtained written informed consent from the families of participants who lacked the mental capacity to consent. The residents were excluded if they or their families refused consent. This study was approved by the Ethical Committees of the Hiroshima University Hospital review board (approval number E-1692) and the National Institute of Infectious Diseases Committee of Ethics (approval number 1017). All study protocols were performed in accordance with the principles of the Declaration of Helsinki.

The authors have no conflicts of interest to declare.

This study was supported by the Ministry of Health, Labour, and Welfare Japan (program Grant No. JPMH19HA1004 and JPMH22HA1002) and the Research Program on Emerging and Re-emerging Infectious Diseases from the Japan Agency for Medical Research and Development under Grant No. 21fk0108604j0001.

Conceptualization, Toshiki Kajihara, Koji Yahara, and Motoyuki Sugai; funding acquisition, Koji Yahara, Motoyuki Sugai, and Hiroki Ohge; investigation, Toshiki Kajihara and Koji Yahara; methodology, Toshiki Kajihara, Koji Yahara, and Motoyuki Sugai; sample collection, Toshiki Kajihara, Mineka Yoshikawa, Azusa Haruta, Mi Nguyen-Tra Le, Maho Takeuchi, Chika Arai, and Kouji Ohta; extracting genomic DNA, Yo Sugawara, Junzo Hisatsune, and Shizuo Kayama; bioinformatics, Koji Yahara; supporting bioinformatics, Norikazu Kitamura; clinical data collection, Toshiki Kajihara; supervision, Miki Kawada-Matsuo, Hitoshi Komatsuzawa, Kazuhiro Tsuga, Hiroki Ohge, and Motoyuki Sugai; validation, Toshiki Kajihara and Koji Yahara; visualization, Toshiki Kajihara and Koji Yahara; writing – original draft, Toshiki Kajihara, Koji Yahara, and Motoyuki Sugai; writing – review and editing, Miki Kawada-Matsuo, Hitoshi Komatsuzawa, Kazuhiro Tsuga, Hiroki Ohge, and Motoyuki Sugai. All authors read and approved the final manuscript.

The genomic data of P. aeruginosa and E. coli strains isolated and analyzed in this study (online suppl. Tables S1 and S2) were deposited in the DNA Data Bank of Japan (DDBJ) and NCBI under BioProject accession number PRJDB12072. The genomic data of 43 P. aeruginosa strains isolated from Hiroshima, Japan, and sequenced using Illumina MiSeq in previous studies (online suppl. Table S1) were deposited in the DDBJ and NCBI under BioProject accession number PRJDB12075. Further inquiries can be directed to the corresponding author.

Toshiki Kajihara, Koji Yahara, and Motoyuki Sugai contributed equally to this work.