https://orcid.org/0009-0000-3461-8405
Figures
Abstract
Background
Pancreatitis-associated sepsis (PAS) caused by carbapenem-resistant bacteria poses significant clinical challenges. The objective of this research was to examine the microbial and metabolic profiles of individuals with carbapenem-resistant Acinetobacter baumannii (CRAB) and Escherichia coli (CREC) infections using integrated microbiomics and metabolomics approaches.
Methods
Peripheral blood samples from 11 PAS patients (8 CRAB, 3 CREC) were analyzed using 16S rDNA gene sequencing and untargeted metabolomics via LC-MS. Microbial diversity, community structure, and differential metabolites were examined between CRAB and CREC groups.
Results
CRAB patients exhibited higher microbial diversity compared to CREC patients. p-Proteobacteria, p-Firmicutes, and p-Cyanobacteria predominated in both patient groups. Significant differences in microbial composition were observed, with p-Proteobacteria more abundant in CRAB and p-Cyanobacteria in CREC samples. g-Enhydrobacter and s-Moraxella osloensis were the biomarkers, significantly higher in CREC patients. Metabolomic analysis revealed 328 differential metabolites between groups, with the majority being downregulated in CRAB. The main categories of identified differential metabolites were amino acids and their derivatives. These differential metabolites were closely related to various metabolic pathways. The most significant metabolic difference between the two patient groups was the level of triglycerides. R-2 Methanandamide and 13-(β-D-glucosyloxy) docosanoic acid showed the highest correlation with g-Enhydrobacter and s-Moraxella osloensis.
Citation: Zhu K, Hu H, Yang Z, Zhu Z (2026) Integrated microbiomics and metabolomics analysis reveals distinct profiles in carbapenem-resistant Acinetobacter baumannii and Escherichia coli infections in Pancreatitis-associated sepsis. PLoS One 21(2): e0340895. https://doi.org/10.1371/journal.pone.0340895
Editor: Rajesh P. Shastry, Yenepoya University, INDIA
Received: July 27, 2025; Accepted: December 29, 2025; Published: February 10, 2026
Copyright: © 2026 Zhu et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data are within the paper and its Supporting Information files.
Funding: The author(s) received no specific funding for this work.
Competing interests: The authors have declared that no competing interests exist.
Abbreviations: PAS, Pancreatitis-associated sepsis; CRAB, carbapenem-resistant Acinetobacter baumannii; CREC, carbapenem-resistant Escherichia coli; LC-MS, liquid chromatography-mass spectrometry; OTU, operational taxonomic unit; NMDS, non-metric multidimensional scaling; PCoA, principal coordinate analysis; LDA, linear discriminant analysis; LEfSe, linear discriminant analysis effect size; UPLC-MS/MS, ultra-performance liquid chromatography-tandem mass spectrometry; GC-MS, gas chromatography-mass spectrometry; KEGG, Kyoto encyclopedia of genes and genomes; HMDB, human metabolome database; TG, triglyceride; VIP, variable importance in projection; OPLS-DA, orthogonal partial least squares discriminant analysis; PCA, principal component analysis; MDRO, multidrug-resistant organism; AP, acute pancreatitis; IDA, information-dependent acquisition
1. Introduction
Pancreatitis represents a complex inflammatory condition affecting the digestive system, marked by acute inflammatory responses and significant systemic complications. Its incidence is increasing globally [1,2]. Due to pancreatic function impairment, patients are prone to developing severe pancreatitis, which may ultimately lead to sepsis, PAS frequently develops under these circumstances [3]. Sepsis, a systemic inflammatory response caused by infection, can result in organ dysfunction and is a leading cause of mortality in intensive care units [4]. PAS significantly increases the mortality rate of pancreatitis patients, particularly when infected with drug-resistant bacteria, posing substantial challenges for clinical management. In recent years, the emergence of antibiotic-resistant bacteria such as CREC and CRAB has further exacerbated the treatment difficulties for PAS patients. These superbugs have developed resistance to multiple antibiotics, posing a severe threat to patients’ lives [5]. A review of the current literature reveals a paucity of studies specifically investigating the comparative roles of CREC and CRAB in the pathogenesis, diagnosis, or management paradigms within the distinct context of established PAS. While patients with MDRO infections in infected pancreatic necrosis are known to have higher mortality [6–8], and CRAB and CREC are independently recognized as significant pathogens imposing substantial clinical burdens [9,10], a direct comparison of their impact within the PAS cohort is lacking. Evidence from high-risk settings, such as ICUs, confirms the emergence of high-risk CREC and CRAB strains, with the COVID-19 pandemic further accelerating CRAB spread [11–13].
Current diagnostics mainly rely on microbial culture and susceptibility testing, which are time-consuming and may delay treatment [14]. International guidelines emphasize early, aggressive management [15], but for resistant strains, traditional empirical antibiotics often fail. This diagnostic delay underscores an urgent clinical need. Extrapolating from existing evidence, the early identification of patients infected with these specific pathogens is critical for facilitating the timely adjustment of therapeutic strategies and improving prognosis in PAS.
To address this challenge, researchers are exploring new diagnostic and therapeutic strategies, including genomic sequencing and metabolomics. Genomic sequencing offers rapid, culture-independent pathogen identification [16,17], though it faces challenges such as contamination from host DNA [18,19]. Metabolomics analysis via LC-MS provides a functional readout of the host’s physiological state, offering new insights into the metabolic perturbations caused by specific infections [20–22]. However, comparative studies focusing specifically on sepsis induced by CREC versus CRAB remain notably scarce [23,24]. Significant knowledge gaps persist regarding differential risk factors, clinical prognoses, and optimal therapeutic strategies for infections caused by these distinct pathogens, particularly after the development of PAS [25,26].
To address this critical knowledge gap, we initiated this investigation by collecting peripheral blood samples from patients diagnosed with PAS and infected with either CRAB or CREC. We employed an integrated multi-omics strategy, hypothesizing that this approach could reveal distinct host-pathogen interaction profiles and provide novel biomarkers. We utilized 16S rDNA gene sequencing to characterize and compare the microbial community structures, identifying differential taxa and potential biomarkers using LEfSe analysis. Concurrently, untargeted LC-MS-based metabolomics was used to profile metabolic alterations, identifying differential metabolites and perturbed pathways. Finally, we performed correlation analyses to investigate the interrelationships between microbial biomarkers and differential metabolites. This study aims to move beyond single-level descriptions to explore the complex interplay between the host microbiome and metabolome in response to CRAB versus CREC infection during PAS. We hypothesize that this integrative analysis may reveal pathogen-specific signatures that could serve as a reference point for rapid diagnostics, aiding clinicians in differentiating these critical infections hours or days before culture results become available. Such timely identification is paramount for optimizing antimicrobial therapy, advancing precision medicine for PAS, and potentially illuminating novel therapeutic targets related to host metabolic responses.
2. Materials and methods
2.1. Patients
This single-centre observational investigation enrolled 11 patients with PAS (CRAB n = 8; CREC n = 3). The research received approval from the Clinical Research Ethics Committee, Renmin Hospital of Wuhan University (Ethics approval number: 2024K-K304(C01)). Sample collection for this study was conducted from [Dec 25, 2024] to [Mar 20, 2025]. Recruitment lasted for 146 days. The recruitment method involved collecting blood samples from patients with severe pancreatitis complicated by intra-abdominal infection. Peripheral blood reflects systemic microbial and metabolic changes associated with PAS and is readily accessible clinically. Therefore, peripheral blood samples were collected from all patients for 16S rDNA sequencing and untargeted metabolomics analysis. Baseline demographic and clinical characteristics of the patients are provided in S1 Table. All patients received standardized treatment for severe pancreatitis, which included the STEP-UP (Step-up Approach) strategy for peripancreatic necrotic tissue debridement. All patients underwent varying degrees of abdominal paracentesis and video-assisted pancreatic necrosectomy, as well as combined drug therapy with specialized antibiotics such as cefoperazone-tazobactam, tigecycline, and polymyxin. The Minimum Inhibitory Concentration (MIC) values for all patients are presented in S2 Table. MIC values were used to determine the resistance of 17 antibiotics. This research followed the guidelines set forth in the Declaration of Helsinki, and all participants provided their informed consent before enrollment. Inclusion Criteria: (1) Diagnostic Criteria: Patients meeting the Atlanta criteria for severe acute pancreatitis and with confirmed intra-abdominal infection via paracentesis drainage. (2) Infecting Pathogen: Blood or pus culture results indicating infection with CREC or CRAB. (3) Informed Consent: Patients or their legal representatives willing to participate in the study and having signed an informed consent form. Exclusion Criteria: (1) Non-severe Pancreatitis: Patients not meeting the Atlanta criteria for severe acute pancreatitis. (2) No Intra-abdominal Infection: Patients without confirmed intra-abdominal infection via paracentesis drainage. (3) Non-CREC/CRAB Infection: Patients whose blood or pus culture results do not indicate CREC or CRAB infection. (4) Immunosuppressed Individuals: Patients receiving immunosuppressive therapy, organ transplant recipients, or those currently undergoing chemotherapy, as this may affect the assessment of infection characteristics. (5) Pregnant or Breastfeeding Women: Pregnant or breastfeeding women are excluded to ensure the safety of the study participants. (6) Severe Comorbidities: Patients with severe liver or kidney dysfunction, terminal illness, or other conditions that may confound the study results. (7) Inability to Provide Valid Informed Consent: Patients unable to provide informed consent due to mental disorders or other reasons.
2.2. 16S rDNA sequencing
Genomic DNA was isolated from peripheral blood samples utilizing the QIAamp Fast DNA Stool Mini Kit (Cat. No. 51604, Qiagen, USA) following the guidelines provided by the manufacturer [27]. The negative control and samples were assessed for contamination by agarose gel electrophoresis during the amplification stage (S1 Fig). Amplicons targeting the bacterial V4 region were amplified with primers 515F and 806R purchased from Thermo Fisher Scientific. PCR utilized Phusion® High-Fidelity PCR Master Mix with GC Buffer (New England Biolabs) under the cycling conditions recommended by Earth Microbiome Project protocols. Amplicons were pooled, purified with AMPure XP beads, and sequenced on an Illumina NovaSeq 6000 platform (2 × 250 bp). Representative sequences were generated through read assembly, filtering, clustering, or denoising methods. OTU abundance statistics were calculated using filtered tags. Sequence information was compared against databases for species annotation and abundance analysis.
The top 10 organisms with the greatest prevalence at each taxonomic rank (Phylum, Class, Order, Family, Genus, Species) were selected for each specimen or group to create stacked bar charts depicting relative organism abundance. The top 35 microbial taxa, ranked by the total of quantitative measurements across all samples, were chosen. Heatmaps were constructed by organizing both species and sample levels according to their quantitative data in each sample. Analyses of alpha and beta diversity were utilized to clarify differences in species composition inside samples and community structure among samples, respectively. NMDS and weighted UniFrac PCoA were utilized to reflect inter- and intra-group differences among samples. LEfSe analysis was conducted to identify differential biomarkers between CRAB and CREC patients.
2.2.1. The criteria for filtering and clustering reads.
Raw sequencing data were demultiplexed according to barcode sequences and PCR amplification primers to assign reads to their respective samples. Barcode and primer sequences were subsequently trimmed. Raw reads were processed using fastp (v0.22.0, https://github.com/OpenGene/fastp) to obtain high-quality reads by applying the following filtering criteria: automatic detection and removal of adapter sequences; removal of reads containing one or more N bases; removal of reads where over 40% of bases had a quality score < 15; trimming based on a 4-base sliding window where reads were truncated if the average quality within the window dropped below 20; removal of terminal polyG sequences; and discarding reads shorter than 150 bp. The resulting high-quality paired-end reads were merged using FLASH (v1.2.11, http://ccb.jhu.edu/software/FLASH/) to generate high-quality tags (Clean Tags). These Clean Tags were then aligned against a species annotation database using vsearch (v2.22.1, https://github.com/torognes/vsearch/) to detect chimeric sequences. Identified chimeric sequences were removed, yielding the final dataset of effective tags.
Effective Tags from all samples were clustered into OTUs using the Uparse algorithm (USEARCH v7, http://www.drive5.com/uparse/) with a default 97% sequence identity threshold. Based on the algorithm’s principles, the most abundant sequence within each cluster was selected as the representative sequence for that OTU. Single OTU in all samples tags sum’s lowest value, defined as all OTUs after flattening tags total number’s 0.005% [28]. Taxonomic assignment for OTU representative sequences was performed using the Mothur software (v1.48) against the SILVA SSUrRNA database (v138.1, http://www.arb-silva.de/) with a confidence threshold range of 0.8–1.0 [29]. This yielded taxonomic information, and the community composition for each sample was summarized at various taxonomic levels. Multiple sequence alignment of all representative OTU sequences was performed using MAFFT (v7.520, https://mafft.cbrc.jp/alignment/software/) to establish phylogenetic relationships. Finally, the OTU abundance data were normalized by rarefying each sample to the lowest sequencing depth present across the dataset. Subsequent diversity analyses were based on these rarefied data.
2.3. Untargeted metabolomics [30–32]
Serum aliquots stored at −80 °C were thawed on ice, vortexed for 10 s, and combined with 300 μL extraction solvent (acetonitrile: methanol 1:4, v/v) containing isotopically labelled internal standards (SCIEX). After vortexing for 3 min, samples were centrifuged at 12,000 × g (4 °C, 10 min); 200 μL supernatant was precipitated at −20 °C (30 min) and recentrifuged, and 180 μL was transferred to autosampler vials. Pooled QC samples, generated by combining equal aliquots from every extract, were injected after system conditioning and then after every tenth analytical sample; solvent blanks bracketed each batch to monitor carryover. Chromatography employed a Waters ACQUITY Premier HSS T3 column (1.8 μm, 2.1 × 100 mm) held at 40 °C with the gradient 0–2 min 5–20% B, 2–5 min 20–60% B, 5–6 min 60–99% B, 6–7.5 min hold 99% B, 7.5–7.6 min return to 5% B, 7.6–10 min re-equilibrate (flow 0.4 mL·min ⁻ ¹, injection 4 μL). Mass spectrometry was conducted on an AB SCIEX TripleTOF 6600 + instrument operated in positive and negative electrospray ionisation modes under information-dependent acquisition (IDA) using Analyst TF 1.7.1. Source parameters were: ion spray voltage +5000 V (positive) and −4000 V (negative); source gas 1 = 50 psi; source gas 2 = 60 psi; curtain gas = 35 psi; source temperature 550 °C (positive) and 450 °C (negative); declustering potential ±60 V. TOF MS scans covered m/z 50–1000 (200 ms accumulation), while product ion scans covered m/z 25–1000 (40 ms accumulation) with collision energy ±30 V and collision energy spread 15 V.
Raw data were converted to mzXML via ProteoWizard. Peak detection, alignment, and retention-time correction were performed in XCMS with centWave parameters. Features with >50% missing values per group were removed; intensity-dependent signal drift was corrected using QC-based support vector regression (SVR). Remaining missing values were imputed using k-nearest neighbour (k = 5). Features detected in solvent blanks were removed unless the sample-to-blank ratio exceeded 5:1. Positive and negative mode data were merged by retaining the highest MSI identification level and lowest QC coefficient of variation per compound. Intensities were log₂ transformed and unit-variance scaled prior to PCA and OPLS-DA. In order to avoid overfitting, a permutation test (200 permutations) was performed (S2 Fig).
Metabolite identification relied on an in-house spectral library augmented with KEGG, HMDB, METLIN, and metDNA resources. MSI levels were assigned as: Level 1a (reference standard matching for retention time, MS1, and MS/MS), Level 1b (reference standard matching with minor MS/MS deviations), Level 2 (library MS/MS match with accurate mass and retention time), Level 3 (putative class annotation), and Level 4 (accurate mass only; excluded from differential analysis). Differential metabolites were defined by VIP > 1 and Student’s t-test P < 0.05, with MSI Levels 1a–4 annotated.
2.4. Spearman statistical analysis
Unsupervised PCA was performed by statistics function prcomp within R (www.r-project.org, Version: 4.1.2). Correlation data were filtered according to a correlation coefficient |r| ≥ 0.9 and a significance test P < 0.05. All plots were generated using the ggplot2 package in R.
3. Results
3.1. Microbial diversity analysis between CRAB and CREC groups
High-throughput sequencing revealed distinct microbial community compositions in the CRAB and CREC groups. The CRAB group exhibited significantly higher microbial diversity, with 556 unique OTUs compared to 88 in the CREC group, while 42 OTUs were shared between groups (Fig 1A). Despite observable differences in overall community structure (Fig 1B and 1C), alpha diversity analyses indicated that there were no notable alterations in diversity and richness indices (S3 Fig). At the taxonomic rank of phylum, Proteobacteria, Firmicutes, and Cyanobacteria were the most abundant in both groups, collectively accounting for over 75% of total abundance (Fig 1D). Notably, the CREC group displayed a higher proportion of Cyanobacteria and lower Proteobacteria compared to the CRAB group. Heatmap analysis of the top 35 phyla further illustrated these compositional differences (Fig 1E). LEfSe identified g-Enhydrobacter and s-Moraxella osloensis as biomarkers in the CREC group (Fig 1F, S3 Table). Significant differences at the genus level for Enhydrobacter and species level for Moraxella osloensis were observed in three CRAB and three CREC patients (S4 Fig).
(A) Venn diagram of Amplicon Sequence Variants (ASVs) among the two groups. (B, C) Non-Metric Multi-Dimensional Scaling (NMDS) and weighted Unweighted UniFrac (UniFrac) Principal Coordinate Analysis (PCoA) conducted on the Operational Taxonomic Units (OTUs) of the microbiota from the CREC and CRAB groups. (D) Relative phylum abundance of ASVs among the CREC and CRAB groups. (E) Cluster analysis of the top 35 phylum abundances among the CREC and CRAB groups. (F) Cladogram generated from Linear Discriminant Analysis Effect Size (LEfSe) analysis showing the association between taxa (the levels represent, from inner to outer rings, phylum, class, order, family, and genus). The thresholds were set at a Linear Discriminant Analysis (LDA) score >4 and P < 0.05, with a focus on family, genus, and species level biomarkers. n = 8 in the CRAB group, and n = 3 in the CREC group.
3.2. Differential metabolite analysis between CRAB and CREC groups
Sample quality control was performed by analyzing the total ion current (TIC) chromatograms, the peak profiles of internal standards in blank samples, the Pearson correlation analysis of QC samples, and the distribution of the coefficient of variation (CV) for internal standard responses. The results showed that the instrument was stable, the data were reliable, and the repeatability was good (S5 Fig, S4 Table). PCA and OPLS-DA demonstrated clear separation between CRAB and CREC samples, indicating significant metabolic profile differences (Fig 2A and 2B). The CRAB group showed more dispersed sample distribution, suggesting greater heterogeneity. Differential metabolite analysis identified 283 downregulated and 45 upregulated metabolites in the CRAB group compared to CREC (Fig 2C). The metabolite with the highest VIP score was TG (14:1(9Z)/18:4(6Z,9Z,12Z,15Z)/20:5(5Z,8Z,11Z,14Z,17Z)) (Figs 2D and 3A). The leading 20 metabolites exhibiting the greatest disparity in concentration between the two cohorts are depicted in Fig 3A and Table 1. Enrichment analysis revealed that amino acids and their metabolites were the most abundant type of differential metabolites, accounting for 23.07% of the differences between groups (Fig 3B and 3C).
(A) 2D Principal Component Analysis (PCA) plot illustrating overall differences in metabolite composition between CRAB (red) and CREC (green) samples. (B) Orthogonal Partial Least Squares Discriminant Analysis (OPLS-DA) score plot. (C) Volcano plot of differential metabolites, with the x-axis representing log2 fold change in metabolite expression and the y-axis representing statistical significance (-log10 (p)). (D) Variable Importance in Projection (VIP) score plot displaying metabolites with the highest VIP values in the OPLS-DA model based on screening criteria for differential metabolites identified in group comparisons. n = 8 in the CRAB group, and n = 3 in the CREC group.
(A) Violin plot of the top 20 differential metabolites. (B) Enrichment analysis. (C) Differential metabolite classification. n = 8 in the CRAB group, and n = 3 in the CREC group.
3.3. Differential metabolite KEGG analysis between CRAB and CREC groups
KEGG analysis showed that differential metabolites were primarily associated with metabolism, clustering predominantly in Metabolic pathways (84.42%) and Glycerophospholipid metabolism (50.65%) pathways (Fig 4).
(A) KEGG enrichment analysis. (B) KEGG classification. n = 8 in the CRAB group, and n = 3 in the CREC group..
3.4. Correlation between Metabolomic Signatures and Microbial Community
Correlation analysis revealed that TG (14:1(9Z)/18:4(6Z,9Z,12Z,15Z)/20:5(5Z,8Z,11Z,14Z,17Z)) was positively correlated with microbes within the Flavobacteriaceae family and inversely associated with those within the Alcaligenaceae family (Fig 5A). At the Enhydrobacter genus and Moraxella osloensis species levels, 14 differential metabolites showed correlations (Fig 5B and 5C). Notably, tacalcitol and 4-tetradecanamidobenzylphosphonic acid exhibited positive correlations, while the remaining 12 metabolites showed negative correlations. Among the top 20 differential metabolites, R-2 Methanandamide and 13-(β-D-glucosyloxy) docosanoic acid demonstrated the strongest correlations with microbes at the Enhydrobacter genus and Moraxella osloensis species levels (Tables 2–4). However, no differential metabolites with correlation coefficients |r| ≥ 0.9 were observed at the family level (S5 Table).
(A) Correlation heatmap of top 20 differential metabolites and differential microbiota at the family level. (B) Correlation heatmap of top 20 differential metabolites and differential microbiota at the genus level. (C) Correlation heatmap of top 20 differential metabolites and differential microbiota at the species level. Rows represent microbiota, columns represent metabolites. Red ellipses indicate positive correlations; blue ellipses indicate negative correlations. The thinner the ellipse, the greater the absolute value of the correlation. Blank cells indicate P > 0.05.
4. Discussion
This study elucidates the distinctive characteristics of CRAB and CREC infections in patients with PAS through an integrated microbiomics and metabolomics approach. Our findings offer new understanding of the infection processes employed by such resistant strains and how they affect host metabolism, while also suggesting potential avenues for diagnostic biomarkers and therapeutic targets. We observed notable variations in the structure of microbial communities in CRAB and CREC-infected patients. The CRAB group exhibited a reduced proportion of Cyanobacteria and an increased proportion of Proteobacteria in relation to the CREC group. This finding aligns with Ceccarani et al., who reported Proteobacteria overgrowth in the gut microbiome of patients with glycogen storage disease [33]. These compositional differences may reflect the specific impacts of different resistant bacteria on the host microecological environment, potentially explaining why no metabolites with high correlation coefficients (|r| ≥ 0.9) were observed at the genus level in our study.
Notably, we identified significant differences in the genera Enhydrobacter and Moraxella osloensis between the two patient groups, with Moraxella osloensis emerging as a potential biomarker. Moraxella osloensis, a Gram-negative bacterium, has been implicated in various infections. Lee et al. reported a fatal case of bacteremia caused by this organism in an immunocompromised patient [34]. Koleri et al. documented multiple bacteremia cases, further emphasizing its pathogenic potential [35]. In the realm of central nervous system infections, Li et al. described a rare case of acute meningitis attributed to Moraxella osloensis [36]. Richards et al. highlighted a case of simultaneous bacteremia and pneumonia caused by Moraxella osloensis in a 17-year-old patient [37]. These findings underscore the clinical significance of Moraxella osloensis in various infectious contexts.
Our metabolomics analysis revealed substantial metabolic differences between CRAB and CREC-infected patients, with most differential metabolites showing downregulation in the CRAB group. This suggests that CRAB infection may exert a broader inhibitory effect on host metabolism. Specifically, we found that amino acids and their metabolites were the primary differential metabolites, consistent with previous research on metabolic changes in sepsis patients. Yang et al. demonstrated that specific metabolite-receptor binding patterns might influence inflammatory responses and immune regulation [38]. Intriguingly, we discovered that R-2 Methanandamide and 13-(β-D-glucosyloxy) docosanoic acid exhibited high correlations with the genus Enhydrobacter and species Moraxella osloensis. This finding opens new avenues for in-depth research on the interactions between specific metabolites and microbes in PAS. Luz-Martínez et al. indicated that Methanandamide affects physiological functions through the TRPV1 channel [39]. These metabolites may play crucial roles in CRAB and CREC infection processes, offering potential targets for future therapeutic strategies.
R-2 Methanandamide, an endogenous cannabinoid receptor agonist, has not been explicitly studied in the context of PAS, CRAB, or CREC infections. However, research has shown that activation of the cannabinoid receptor CB2 can modulate inflammation, infection, and immune responses, potentially playing a significant role in pancreatitis and sepsis treatment [40]. Additionally, the cannabinoid receptor VR1 is upregulated in pancreatic cancer, with its expression level associated with prognosis, suggesting potential involvement in the pathogenesis of pancreatitis and sepsis [41]. 13-(β-D-glucosyloxy) docosanoic acid, a fatty acyl compound [42], has received limited attention in research. However, it has garnered some interest in plant physiology and pharmacology. The strong correlation of both R-2 Methanandamide and 13-(β-D-glucosyloxy) docosanoic acid with Enhydrobacter and Moraxella osloensis provides further microbiological and metabolomic insights into the microbial differences between CRAB and CREC infections.
The prevalence of CRAB and CREC among patients with PAS presents substantial hurdles for effective treatment and control. The urgent unmet clinical need lies in the rapid and accurate differentiation between these infections, as delays or inappropriate empirical therapy stemming from the days-long turnaround of traditional culture methods can significantly worsen patient prognosis and fuel antimicrobial resistance. Our study holds considerable promise in addressing this challenge. We identified Moraxella osloensis as significantly enriched in CREC infections, which, particularly when considered alongside correlated metabolites such as R-2 Methanandamide and 13-(β-D-glucosyloxy) docosanoic acid, suggests the potential for constructing a multi-marker signature. Such a signature could pave the way for the development of point-of-care diagnostic tools. Critically, these tools aim to provide pathogen-leaning information within hours, not days, enabling clinicians to initiate timely, targeted antimicrobial therapy or make informed adjustments to empirical regimens much sooner. Beyond their diagnostic potential, these findings hold significant implications for future therapeutic strategies and provide a foundation for precision medicine in PAS. The strong correlation observed between specific microbial taxa, such as Moraxella osloensis, and differential metabolites, like R-2 Methanandamide, highlights the endocannabinoid system and associated pathways as potential novel avenues for non-antibiotic, host-directed interventions [43,44]. The development of such strategies is critically important to circumvent established antibiotic resistance mechanisms. Furthermore, the capacity to rapidly differentiate between CRAB and CREC using the proposed biomarker signatures can directly inform clinical decision-making. For instance, early biomarker evidence suggestive of CRAB infection might prompt clinicians to consider earlier escalation to last-resort antibiotics and to institute rigorous monitoring for complications potentially linked to its unique metabolic disruption profile. Conversely, a signature indicative of CREC could guide alternative empirical therapeutic choices or different monitoring approaches. By dissecting these pathogen-specific host responses, this research moves beyond simple pathogen identification towards enabling truly personalized management strategies. Such strategies aim to target both the specific etiological agent and the host’s unique physiological response to it. Crucially, however, these preliminary findings require extensive validation before they can be considered for clinical implementation.
First, although this study provides valuable preliminary insights, the sample size was limited, particularly for the CREC subgroup, which restricts the statistical power and generalizability of our findings. As indicated by the permutation test results, while the very high R2Y value indicates strong explanatory power for the observed class separation, the moderate Q2 suggests limited predictive power. The strong correlation between Moraxella osloensis and key metabolites is noteworthy but requires validation in a larger cohort to rule out chance associations. Second, our analysis was limited to peripheral blood samples. While circulating biomarkers reflect systemic responses, they may underestimate local pathogenic processes in organs such as the pancreas or gut. Future work will involve collecting target organ tissues or specific bodily fluids to uncover molecular events more proximal to the site of pathology. Third, the cross-sectional nature of our sampling provides only a single-time-point snapshot, precluding the tracking of microbiome and metabolome dynamics throughout the infection process. We have planned a prospective longitudinal follow-up study involving multi-timepoint sampling to characterize temporal heterogeneity and validate the reproducibility of the identified signatures. Fourth, metabolite identification was performed by searching an in-house database, integrated public and predictive libraries, and using the metDNA approach, but these identifications lack quantitative validation with chemical standards. Therefore, based on the findings from the correlation and integrative analyses, targeted quantification of specific key metabolites will be performed in future mechanistic studies. Finally, as the serum samples were de-identified upon collection, we were unable to obtain complete clinical covariates for integrated analysis. Therefore, we are establishing a prospective multicenter cohort to concurrently collect standardized clinical phenotypic and metabolomic data. We will collaborate with a biostatistics team to perform multivariate modeling to assess whether these metabolites can predict the need for organ support, length of hospital stay, and treatment response. Concurrently, mechanistic studies focusing on R-2-methylarachidonylamine and 13-(β-D-glucopyranosyloxy) docosanoic acid will be pursued to advance their functional validation and translational potential.
5. Conclusion
In conclusion, this study reveals unique characteristics of CRAB and CREC infections in PAS patients through integrated microbiomics and metabolomic analyses. We identified Moraxella osloensis as a potential biomarker for differentiating CRAB and CREC infections, correlating with R-2 Methanandamide and 13-(β-D-glucosyloxy) docosanoic acid. These findings not only deepen our understanding of the infection mechanisms of these resistant strains but also provide new directions for developing future diagnostic and therapeutic strategies in the management of PAS. Future longitudinal studies will incorporate larger cohorts with pancreatic and/or gut microbiome samples collected over extended time periods to further investigate the underlying biomarker mechanisms.
Supporting information
S1 Fig. The agarose gel experiment monitored the negative control of 11 samples, and after two amplifications, the negative control did not show amplification bands, and the amplification was free of contamination.
https://doi.org/10.1371/journal.pone.0340895.s001
(TIF)
S2 Fig. OPLS-DA model validation, permutation test (200 permutations).
https://doi.org/10.1371/journal.pone.0340895.s002
(TIF)
S3 Fig. Differences in α-diversity and β-diversity between Carbapenem-resistant Acinetobacter baumannii (CRAB) and Carbapenem-resistant Escherichia coli (CREC) groups.
(A-D) Analysis of α-diversity between two groups by Abundance-based Coverage Estimator (ACE) index, Chao1 index, Shannon index, and Simpson index. (E-H) Analysis of β-diversity between two groups by Non-Metric Multidimensional Scaling (NMDS), Principal Component Analysis (PCA), Principal Coordinate Analysis (PCoA), and Bray-Curtis analysis. n = 8 in the CRAB group, and n = 3 in the CREC group.
https://doi.org/10.1371/journal.pone.0340895.s003
(TIF)
S4 Fig. Differential microbiota analysis by Linear Discriminant Analysis Effect Size (LEfSe) analysis.
(A) The family level. (B) The genus level. (C) The species level.
https://doi.org/10.1371/journal.pone.0340895.s004
(TIF)
S5 Fig. Sample quality control analysis.
In positive ion mode, (A) Total ion current (TIC) chromatogram. (B) Extracted ion chromatogram (EIC) of the blank sample. (C) Pearson correlation analysis. (D) The distribution of Coefficients of Variation (CV). In negative ion mode, (E) TIC chromatogram. (F) EIC of the blank sample. (G) Pearson correlation analysis. (H) The distribution of CV.
https://doi.org/10.1371/journal.pone.0340895.s005
(TIF)
S1 Table. Baseline demographic and clinical characteristics.
https://doi.org/10.1371/journal.pone.0340895.s006
(DOCX)
S2 Table. Minimum Inhibitory Concentration/Antibiotic Susceptibility Result.
https://doi.org/10.1371/journal.pone.0340895.s007
(DOCX)
S5 Table. Correlation analysis between microorganisms and metabolites.
https://doi.org/10.1371/journal.pone.0340895.s010
(DOCX)
References
- 1. Tian Y, Huang Q, Ren Y-T, Jiang X, Jiang B. Visceral adipose tissue predicts severity and prognosis of acute pancreatitis in obese patients. Hepatobiliary Pancreat Dis Int. 2024;23(5):458–62. pmid:37648552
- 2. Iannuzzi JP, King JA, Leong JH, Quan J, Windsor JW, Tanyingoh D, et al. Global Incidence of Acute Pancreatitis Is Increasing Over Time: A Systematic Review and Meta-Analysis. Gastroenterology. 2022;162(1):122–34. pmid:34571026
- 3. Shankar-Hari M, Wyncoll DJCC. Activated protein C in severe acute pancreatitis without sepsis? Not just yet. Critical Care. 2010;14:1–2.
- 4. van der Poll T, Shankar-Hari M, Wiersinga WJ. The immunology of sepsis. Immunity. 2021;54(11):2450–64. pmid:34758337
- 5. Jarczak D, Kluge S, Nierhaus A. Sepsis-Pathophysiology and Therapeutic Concepts. Front Med (Lausanne). 2021;8:628302. pmid:34055825
- 6. Jain S, Mahapatra SJ, Gupta S, Shalimar, Garg PK. Infected Pancreatic Necrosis due to Multidrug-Resistant Organisms and Persistent Organ failure Predict Mortality in Acute Pancreatitis. Clin Transl Gastroenterol. 2018;9(10):190. pmid:30287818
- 7. Cen M-E, Wang F, Su Y, Zhang W-J, Sun B, Wang G. Gastrointestinal microecology: a crucial and potential target in acute pancreatitis. Apoptosis. 2018;23(7–8):377–87. pmid:29926313
- 8. Otsuka Y, Kamata K, Minaga K, Watanabe T, Kudo M. Pancreatic colonization of fungi in the development of severe acute pancreatitis. Front Cell Infect Microbiol. 2022;12:940532. pmid:35967861
- 9. Rafique A, Khawar S. Risk Factors Associated with MDR and CR Acinetobacter baumannii Carriage among ICU Patients Hospitalized at Hospitals in Pakistan. Sch Acad J Biosci. 2025;13(03):349–54.
- 10. Durante-Mangoni E, Andini R, Zampino R. Management of carbapenem-resistant Enterobacteriaceae infections. Clin Microbiol Infect. 2019;25(8):943–50. pmid:31004767
- 11. Zhang R, Li Y, Chen J, Liu C, Sun Q, Shu L, et al. Population genomic analysis reveals the emergence of high-risk carbapenem-resistant Escherichia coli among ICU patients in China. J Infect. 2023;86(4):316–28. pmid:36764393
- 12. Ding Y, Zhuang H, Zhou J, Xu L, Yang Y, He J, et al. Epidemiology and Genetic Characteristics of Carbapenem-Resistant Escherichia coli in Chinese Intensive Care Unit Analyzed by Whole-Genome Sequencing: a Prospective Observational Study. Microbiol Spectr. 2023;11(2):e04010-22. pmid:36802220
- 13. Wong SC, Chau PH, So SYC, Lam GKM, Chan VWM, Yuen LLH, et al. Control of healthcare-associated carbapenem-resistant Acinetobacter baumannii by enhancement of infection control measures. JA. 2022;11(8):1076.
- 14. Gauer R, Forbes D, Boyer N. Sepsis: Diagnosis and Management. Am Fam Physician. 2020;101(7):409–18. pmid:32227831
- 15. Evans L, Rhodes A, Alhazzani W, Antonelli M, Coopersmith CM, French C, et al. Surviving Sepsis Campaign: International Guidelines for Management of Sepsis and Septic Shock 2021. Crit Care Med. 2021;49(11):e1063–143. pmid:34605781
- 16. Abellan-Schneyder I, Matchado MS, Reitmeier S, Sommer A, Sewald Z, Baumbach J, et al. Primer, Pipelines, Parameters: Issues in 16S rRNA Gene Sequencing. mSphere. 2021;6(1):e01202-20. pmid:33627512
- 17. Church DL, Cerutti L, Gürtler A, Griener T, Zelazny A, Emler S. Performance and Application of 16S rRNA Gene Cycle Sequencing for Routine Identification of Bacteria in the Clinical Microbiology Laboratory. Clin Microbiol Rev. 2020;33(4):e00053-19. pmid:32907806
- 18. Walker SP, Barrett M, Hogan G, Flores Bueso Y, Claesson MJ, Tangney M. Non-specific amplification of human DNA is a major challenge for 16S rRNA gene sequence analysis. Sci Rep. 2020;10(1):16356. pmid:33004967
- 19. Weinroth MD, Belk AD, Dean C, Noyes N, Dittoe DK, Rothrock MJ, et al. Considerations and best practices in animal science 16S ribosomal RNA gene sequencing microbiome studies. J Anim Sci. 2022;100(2):skab346. pmid:35106579
- 20. Niederman MS, Baron RM, Bouadma L, Calandra T, Daneman N, DeWaele J, et al. Initial antimicrobial management of sepsis. Critical Care. 2021;25:1–11.
- 21. Zhang Y-Y, Ning B-T. Signaling pathways and intervention therapies in sepsis. Signal Transduct Target Ther. 2021;6(1):407. pmid:34824200
- 22. Durazzi F, Sala C, Castellani G, Manfreda G, Remondini D, De Cesare A. Comparison between 16S rRNA and shotgun sequencing data for the taxonomic characterization of the gut microbiota. Sci Rep. 2021;11(1):3030. pmid:33542369
- 23. Kong H, Hu Z, Zhang L, Chen Q, Yang L, Li J, et al. Clinical risk factors and outcomes of carbapenem-resistant Escherichia coli nosocomial infections in a Chinese teaching hospital: a retrospective study from 2013 to 2020. Microbiol Spectr. 2024;12(7):e0422823. pmid:38814065
- 24. Zilberberg MD, Nathanson BH, Sulham K, Fan W, Shorr AF. Carbapenem resistance, inappropriate empiric treatment and outcomes among patients hospitalized with Enterobacteriaceae urinary tract infection, pneumonia and sepsis. J Bid. 2017;17:1–13.
- 25. Lv Y, Jiang M, Ouyang Y, Zheng X, Zhang L, Yao J, et al. Sodium Butyrate-Loaded Microspheres With Enhanced Bioavailability for Targeted Treatment of Intestinal Barrier Injury. Adv Healthc Mater. 2025;14(9):e2402773. pmid:39629540
- 26. Zhao M, Cui M, Fan M, Huang C, Wang J, Zeng Y, et al. Octreotide attenuates experimental severe acute pancreatitis through inhibiting pyroptosis and modulating intestinal homeostasis. Eur J Pharmacol. 2025;994:177314. pmid:39922420
- 27. Ma Y, Zhou Y, Xie G, Chen H, Huangfu Y, Shen L, et al. HEX-1 reduces colitis-driven colorectal cancer via inactivating the prolyl isomerase PIN1 sensitization and remodeling the gut microbiota. Discov Oncol. 2025;16(1):565. pmid:40251462
- 28. Bokulich NA, Subramanian S, Faith JJ, Gevers D, Gordon JI, Knight R, et al. Quality-filtering vastly improves diversity estimates from Illumina amplicon sequencing. Nat Methods. 2013;10(1):57–9. pmid:23202435
- 29. Quast C, Pruesse E, Yilmaz P, Gerken J, Schweer T, Yarza P, et al. The SILVA ribosomal RNA gene database project: improved data processing and web-based tools. Nucleic Acids Res. 2012;41(Database issue):D590–6. pmid:23193283
- 30. Yang Y, Fan C, Zhang Y, Kang T, Jiang J. Untargeted Metabolomics Reveals the Role of Lipocalin-2 in the Pathological Changes of Lens and Retina in Diabetic Mice. Invest Ophthalmol Vis Sci. 2024;65(14):19. pmid:39656472
- 31. Xue Y, Zhang J, Chen M, Fan Q, Huang T, Ke J, et al. Integrated Metabolomics and Proteomics Analysis of the Myocardium in a Mouse Model of Acute Viral Myocarditis. Immun Inflamm Dis. 2025;13(2):e70151. pmid:39912544
- 32. Ray MK, Fu M, Chen Y, Chen T, Nomoto T, Sakai S, et al. Zero-field Hall effect emerging from a non-Fermi liquid in a collinear antiferromagnet V1/3NbS2. Nat Commun. 2025;16(1):3532. pmid:40251162
- 33. Ceccarani C, Bassanini G, Montanari C, Casiraghi MC, Ottaviano E, Morace G, et al. Proteobacteria Overgrowth and Butyrate-Producing Taxa Depletion in the Gut Microbiota of Glycogen Storage Disease Type 1 Patients. Metabolites. 2020;10(4):133. pmid:32235604
- 34. Lee MY, Kim MH, Lee WI, Kang SY. Bacteremia caused by Moraxella osloensis: a fatal case of an immunocompromised patient and literature review. J Clin Lab. 2021;1.
- 35. Koleri J, Petkar HM, Husain AAM, Almaslamani MA, Omrani AS. Moraxella osloensis bacteremia, a case series and review of the literature. IDCases. 2022;27:e01450.
- 36. Li Y, Wang GQ, Ma XL, Li YBJCN. A rare case of acute meningitis caused by Moraxella osloensis. Therapeutics. 2024;30(8).
- 37. Richards HW, Lancioni C, Jones M. Moraxella osloensis Bacteremia and Pneumonia in a Seventeen-Year-Old Patient. Int J Clin Pediatr. 2022;11(2):56–9.
- 38. Yang J-F, Williams AH, Penthala NR, Prather PL, Crooks PA, Zhan C-G. Binding Modes and Selectivity of Cannabinoid 1 (CB1) and Cannabinoid 2 (CB2) Receptor Ligands. ACS Chem Neurosci. 2020;11(20):3455–63. pmid:32997485
- 39. Luz-Martínez BA, Marrero-Morfa D, Luna-Vázquez FJ, Rojas-Molina A, Ibarra-Alvarado C. Affinin, Isolated from Heliopsis longipes, Induces an Antihypertensive Effect That Involves CB1 Cannabinoid Receptors and TRPA1 and TRPV1 Channel Activation. Planta Med. 2024;90(5):380–7. pmid:38219731
- 40. Hashiesh HM, Sharma C, Goyal SN, Jha NK, Ojha S. Pharmacological Properties, Therapeutic Potential and Molecular Mechanisms of JWH133, a CB2 Receptor-Selective Agonist. Front Pharmacol. 2021;12:702675. pmid:34393784
- 41. Miao X, Liu G, Xu X, Xie C, Sun F, Yang Y, et al. High expression of vanilloid receptor-1 is associated with better prognosis of patients with hepatocellular carcinoma. Cancer Genet Cytogenet. 2008;186(1):25–32. pmid:18786439
- 42.
Schomburg D, Stephan D. 13-hydroxydocosanoate 13-beta-glucosyltransferase. In: Schomburg D, Stephan D, editors. Enzyme Handbook 12: Class 232 — 24 Transferases. Berlin, Heidelberg: Springer Berlin Heidelberg; 1996. p. 731–3.
- 43.
Wheeler CE, Coleman SS, Hoyd R, Denko L, Chan CH, Churchman ML, et al. The tumor microbiome as a predictor of outcomes in patients with metastatic melanoma treated with immune checkpoint inhibitors. 2023.
- 44. Ramer R, Weinzierl U, Schwind B, Brune K, Hinz B. Ceramide is involved in r(+)-methanandamide-induced cyclooxygenase-2 expression in human neuroglioma cells. Mol Pharmacol. 2003;64(5):1189–98. pmid:14573769