Skip to main content
NCBI home page
Critical Care logoLink to Critical Care
. 2025 Dec 9;30:23. doi: 10.1186/s13054-025-05789-7

The molecular architecture of severe pediatric traumatic brain injury: integrated omics reveal therapeutic pathways

Elora Hussain 1,2,13,✉,#, Jeremy W Prokop 1,3,#, Emily Nonnemacher 1,2, Nadia Ashrafi 4,5, Ali Yilmaz 4,5,6, Romana Ashrafi Mimi 4,5, Abdullah Khalid 4, Karolis Krinickis 7, Vilija Lomeikaite 7, Lena Sanfilippo 2,8, Kylie Maxton 3, Jacob Charron 1, Charitha Subrahmanya 3, Austin Goodyke 3, Annie Needs 9,10, Daniel R Woldring 9,10, Caleb P Bupp 1,11, Nicholas Hartog 1,12, Juozas Gordevicius 7, Stewart F Graham 4,5,6, Surender Rajasekaran 1,2
PMCID: PMC12801565  PMID: 41361486

Abstract

Introduction

Traumatic brain injury (TBI) remains a leading cause of death and long-term disability in children worldwide. Despite its impact, current clinical management is limited to supportive care, with no FDA-approved therapies to reduce mortality or mitigate lasting neurological consequences. This study presents, to our knowledge, the first integrated multi-omics analysis combining transcriptomic and metabolomic data from pediatric patients with severe TBI spanning both acute and subacute phases offering novel insights into the molecular pathways underlying injury and recovery.

Methods

In this prospective, observational cohort study seventeen severe pediatric TBI patients (median age of 13.1 years, median Glasgow Coma Scale (GCS) of 3, and median Injury Severity Score (ISS) of 29) with no pre-existing neurological comorbidities or non-accidental trauma, and corresponding sex and age-matched controls were enrolled between May 2022 and November 2023. The longitudinal bulk transcriptomic analysis of whole blood and metabolomic profiling of serum were performed at three distinct timepoints. The resulting multi-omics datasets were subsequently integrated with validated clinical severity scoring systems to assess changes over a nine-day period of care in the pediatric intensive care unit (PICU).

Results

We showed that despite the heterogeneity of mechanism and presentation, there was overlap in the transcriptomic and metabolic signatures at each timepoint. There were immediate signs of inflammatory and immune activation, metabolic dysregulation, disturbance of the gut-brain axis in the acute phase. Early markers of T-cell infiltration, such as TRAV35 and ANXA2R, are highly correlated with GCS, and lysophosphatidylcholine 18:0 is highly correlated with NK-cell activation. Multiple gut metabolites, such as indole-3-propionic acid (IPA), and RNA signatures of gut flora are elevated in blood early after TBI. Putrescine elevation at time point one highly correlates with Day 9 red blood cell stimulation. At Day9, multiple lipid species in the metabolome are associated with length of stay and Glasgow Outcome Scale-Extended (GOS-E Peds). By Day 9, both the metabolome and transcriptome show incomplete recovery, marked by highly specific TBI IGH, IGK, and IGL clonal expansion.

Conclusions

Despite the heterogeneity in injury mechanisms and clinical presentations, our findings reveal a convergent host response, characterized by shared transcriptomic and metabolic signatures across all time points. This convergence highlights potentially targetable biological pathways and opens the door to the development of novel therapeutic strategies for severe pediatric TBI.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13054-025-05789-7.

Key points

Question

Can integrated omics characterize the host response to severe pediatric TBI and identify novel therapeutic targets?

Findings

A prospective observational cohort study of three timepoints over nine days showed uniformity in secondary response after the primary injury. Circulatory evidence of initial innate immune activation is associated with later TBI-specific antibody production, cellular and metabolic stress, lipid and amino acid alterations, gut dysfunction, inform changes occurring systemically and within the brain.

Meaning

Despite differences between individual patients and mechanisms of injury, the secondary response in pediatric TBI appeared uniform, and multi-omics may provide novel therapeutic options not previously considered.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13054-025-05789-7.

Introduction

Traumatic Brain Injury (TBI) affects up to 280/100,000 children, resulting in approximately 60,000 hospitalizations annually in the USA [1, 2]. Long-term consequences on health and quality of life are profound, with severe injuries often leading to chronic physical, behavioral, and cognitive disabilities. The economic and public health impact is estimated at $2.6 billion annually in the USA [3]. Scientific advances over the last few decades have improved our understanding of the pathobiological processes associated with severe TBI. We now recognize that severe TBI is much more than an acute event from the primary injury and involves ongoing, progressive pathology consisting of interdependent cellular and molecular cascades that contribute to secondary brain injury in the hours to days and even years in some patients following the acute injury. This period of secondary brain injury may offer a therapeutic window for potential interventions. Unfortunately, despite promising experimental treatments in pre-clinical animal models of TBI over the last few decades, no therapeutic interventions have demonstrated efficacy in large clinical trials [4].

Recent advances in our understanding of the cellular disruptions following TBI have led to the identification of several protein-based biomarkers, such as S100 calcium-binding protein B (S100B), neuron-specific enolase, myelin basic protein, glial fibrillary acidic protein, and ubiquitin carboxy-terminal hydrolase-L1, which show diagnostic and prognostic potential in TBI [5]. However, there remains a critical need to identify underlying mechanistic pathways that may be targeted for therapeutic interventions.

Here, to our knowledge, we present the first pediatric study to integrate clinical data with multi-omics technologies, specifically, whole blood transcriptomics and serum metabolomics, to better characterize the host response to severe pediatric TBI. Integrated omics technologies can improve our understanding of mechanisms through which various phenotypes are generated, and combining metabolomics and transcriptomics allows for a greater holistic understanding of dynamic interactions between gene expression changes and metabolites than either technology alone [6, 7]. We hypothesized that this integrated approach, conducted over three timepoints spanning the acute and subacute post-injury periods, would reveal novel molecular signatures of secondary brain injury.

Methods

Participants

After Institutional Review Board approval (2021-096), male and female patients aged one to 21 years with severe TBI (defined by GCS ≤ 8) requiring mechanical ventilation were recruited from a pediatric tertiary intensive care unit. Patients’ healthcare proxies provided consent for biofluid collection. Non-injured age- and gender-matched control patients undergoing routine sedated procedures were recruited from the Pediatric Sedation Unit. Exclusion criteria included significant comorbidities, pregnancy, suspected or confirmed non-accidental trauma, or unclear mechanism of injury.

Sample collection

Blood samples (RNA PAXgene tube and serum separated) were collected from TBI patients at three time points (Fig. 1). First samples were collected within 12–24 h of PICU admission (hospital day 0), second samples within 36–60 h of PICU admission (hospital day 2), and third samples within 204–228 h of PICU admission (hospital day 9). PICU admission for all patients occurred within 24 h of injury. The first two time points were considered the acute phase of injury. The third time point was considered the subacute phase of injury. A single blood sample was collected for each control. Demographics, severity scores, comorbidities, injury characteristics, including neuroimaging and clinical data, were abstracted from the electronic medical record and entered into a REDCap (Research Electronic Data Capture) database.

Fig. 1.

Fig. 1

Open in a new tab

Patient demographics and study design. TBI, Traumatic Brain Injury; TP1, timepoint 1; TP2, timepoint 2; TP3, timepoint 3; DI/LC-MS/MS, Direct Injection/Liquid Chromatography-Mass Spectrometry; 1H NMR, proton magnetic resonance; GCS, Glasgow Coma Scale; LOS, length of stay; ISS, Injury Severity Score; GOS-E Peds, Glasgow Outcome Scale-Extended Pediatric Revision

RNASequencing

PAXgene tube RNA was sequenced by Azenta Life Sciences in paired-end 150-basepair reads on Illumina NovaSeq in a single flow lane. Reads were processed for quality control with FastQC and MultiQC (eFigure S1), aligned to the human transcriptome using Salmon relative to Gencode v45, screened for multiple species using KRAKEN2 relative to the PlusPFP index, and assessed for immune repertoire using MiXCR and ImmunArch [8–10]. Significant differential expression of transcripts was assessed using Bonferroni corrected p-value < 0.05 and log2 fold change > 2 or <−2 for pairwise comparisons. Enriched pathways associated with significant genes were identified using STRING. Deconvolution was performed using CIBERSORTx with the LM22 signature matrix [11]. Mapping data and statistics are available for RNAseq at 10.6084/m9.figshare.28303880.v1.

Metabolomics

Serum samples were analyzed with Direct Injection/Liquid Chromatography-Mass Spectrometry (DI/LC-MS/MS) and proton magnetic resonance (1H NMR) Spectroscopy. In short, DI/LC-MS/MS samples were analyzed using the biocrates Q500 XL kit on a Waters I-class Ultra Performance Liquid Chromatography (UPLC) unit coupled with a Sciex 7500 mass spectrometer. DI/LC-MS/MS data were extracted using WebIDQ and MetaboINDICATOR to calculate metabolite sums and ratios, respectively. For 1H NMR, samples (200 µL) were analyzed in a Bruker 3 mm thin wall NMR tube, and 1H detection was performed at 600.13 MHz on a Bruker ASCEND 600 NMR spectrometer equipped with a cryogenically cooled 3 mm TCI probe. Bayesil was used to identify and quantify all 1D 1H-NMR data [12]. Further description of these methods is available as supplementary information.

Metabolites were filtered for missing values, log-transformed, and imputed using k-nearest neighbors algorithm with 5 nearest neighbors via the impute R package (v1.68.0). Metabolite intensities were normalized using R (v4.3.1) scale function and were used to fit a linear model using limma R package (v3.56.2). A mixed-effect linear model was used, including a random effect for patient ID to account for repeated measures within individuals. The relationships between metabolite levels and clinical characteristics were assessed using similar individual mixed-effect models, with each clinical measure included as a covariate. The model was further adjusted for sex, age, time point, and condition. To ensure that biomarker discovery was driven by the most conservative and biologically robust approach, we conducted all analyses using a cross-sectional design. This framework allowed each participant to contribute a single, independent measurement at the relevant timepoint, thereby eliminating within-subject repetition and preventing correlation structures arising from longitudinal sampling. As a result, all comparisons across timepoints and controls reflect independent observations, strengthening the validity of the statistical inferences. Metabolite-set enrichment was carried out using the GSEA function of clusterProfiler R package (v4.10.0) with the fgsea method with 1,000 permutations [13]. For linear and enrichment analyses, p-values were adjusted using the Benjamin-Hochberg [14] method for multiple testing correction, and adjusted p-values (q-values) < 0.05 were considered significant [15].

Integrated statistics

Clinical data, transcripts, and metabolites were integrated using Pearson correlation for each time point relative to other time points, and the results were plotted using Matplotlib. Clinical scoring included PICU length-of-stay (PICU_LOS), hospital length-of-stay (Hospital_LOS), Pediatric Risk of Mortality (PRISM_Score), Glasgow Coma Scale (Initial_GCS), Injury Severity Score (ISS), Glasgow Outcome Scale Extended Disability Checklist (GOSE_DC), Glasgow Outcome Scale Extended (GOSE_3mo), Functional Status Scale at Discharge (fss_dc), and Functional Status Scale assessed at 3 months post-discharge (fss_3m). To build a cross-day analysis of transcripts and metabolites, we utilized correlations of transcript panels, where the Z-score was added for all transcripts of a given panel of genes, relative to other significant panels or metabolites using the eight TBI samples that had all three timepoints available.

Results

Transcriptomic profiles indicate sequential changes over time post-trauma

Principal component analysis (PCA) plots using Gencode v45 aligned protein-coding reads produce clustering of datapoints from timepoints 1 and 2 (Day0 and Day2) in the TBI cohort, and unique clustering in timepoint 3 (Day9). Data points from the control cohort also showed unique clustering. (Fig. 2A). We identified 345 significant transcripts for pairwise comparison groups (eFigure S2). While there was sample dropout for later collections, the Day0 samples did not show any significant differences between those with or without a Day9 collection (Fig. 2B). Twenty-two transcripts were identified as potential Day 0 biomarkers and 28 transcripts as potential Day 9 biomarkers using multiple Log2 fold change comparisons. (eFigure S3). Figure 2C shows the per-sample heatmap for the major insights discussed below, enabling both per sample assessment of significant factors and visualization of non clustering data statistics (age, sex, sequencing depth, % GC content, and % of reads mapping to the human reference).

Fig. 2.

Fig. 2

Open in a new tab

RNA sequencing data for TBI cohort (A) Principal component analysis (PCA) of all protein-coding transcripts in control and TBI time points. Different colored lines represent connected samples within individuals. A 3D interactive PCA HTML file can be downloaded at 10.6084/m9.figshare.30379273.v1. The top 50 loading components for each axis can be found at 10.6084/m9.figshare.30379144.v1. (B) and (C) Box and whisker plots displaying gene ontology pathways over time in TBI patients relative to controls. An interactive 3D plot for significant transcripts is available at 10.6084/m9.figshare.30379285.v1. (D) RNA sequencing heat mapping data for controls and Day 0, Day 2, or Day 9 TBI samples. Demographics at the top include age groups (blue = ages 3–10 years, white = ages 10–14 years, red = 14 + years) and sex (blue = female, red = male), which show no clustering for control or day groups. For remainder of values, blue represents the lowest value of each row; red represents the highest value of the row. For bottom two immune chain plots, white = zero values; red = top mapped values

Significant day0 transcripts

Peripheral blood transcriptomics identified genes of early leukocyte activation and T-cell receptor mobilization. M0 macrophages and neutrophil-related transcripts, including those related to degranulation, are abundant compared to controls. As the innate immune system is triggered, higher NK cell gene expression is also observed. There were 56 TBI Day0 higher transcripts from 40 genes with five isoforms of BCL6, 4 of ADAM9, and 3 of SLC2A3/WDFY3. The 40 genes were enriched for S100/CaBP-9k-type calcium binding subdomain (CL:19236), the S100A8/A9 complex (GOCC:1990662), extracellular exosome (GO:0070062), neutrophil degranulation (HSA-6798695), and ficolin-1-rich granule membrane (GO:0101003). There were 250 transcripts from 214 genes lower in TBI Day0, including six isoforms of LINC00299 and five isoforms of CD247/SPON2/EPHX2. The 214 genes were enriched for neural plate pattern specification, (GO:0060896), alpha-beta T cell receptor complex (GO:0042105), immune system process (GO:0002376), leukocyte activation (GO:0045321), natural killer cell-mediated leukemia response, and T-cell response in multiple sclerosis [16, 17]. Neutrophil aggregation and plasma cell differentiation transcripts are altered early after TBI and return to control levels, while other pathways remain altered on Day9 (eFigure S4). There was one transcript from Day2 samples relative to Day0 significant (RN7SL663P-201, ENST00000578592.2, a misc_RNA biotype).

Significant day9 transcripts

Day9 TBI vs. control identified 37 higher transcripts from 31 genes including three isoforms for ALAS2/SLC14A1. The Day9 TBI biomarkers are enriched for erythrocyte development (GO:0048821), erythropoietin treatment, erythrocyte maturation (KW-0265), and mean corpuscular hemoglobin concentration (EFO:0004528) [18]. The hemoglobin and erythrocyte development transcripts are only altered on Day9.

Taxonomic transcript classification

RNA sequencing reads primarily map to the human transcriptome (49.3 ± 0.3%) based on KRAKEN2, followed by unclassified (0.72 ± 0.01%), virus (0.43 ± 0.05%), bacteria (0.34 ± 0.03%), and plants (0.32 ± 0.03%). A total of 51 species were identified, including a cluster of three samples of Day2 TBI that contained bacterial RNA in their blood from several gut flora species (Escherichia coli, Clostridium, Akkermansia muciniphila, eFigure S5). A cluster of six patients at Day9 TBI had RNA of species that included Staphylococcus aureus and Klebsiella pneumoniae. (eFgure S5)

Immune repertoire analysis: T and B cell changes

Immune cell deconvoluted revealed significantly lower CD8 T-cell and CD4 memory T-cell biomarkers in TBI (eFigure S6). Within the immune repertoire analysis, there is a significantly lower level of T-cell receptor alpha/delta (TRAD) and beta (TRB) unique sequences in Day0 TBI (p-value 2.81e-11 and 8.37e-10) and Day2 TBI (p-value 5.50e-10 and 5.27e-08) relative to controls (eFigure S7). The number of unique antibody sequences is higher in Day9 TBI samples for immunoglobulin heavy chain (IGH) (p-value 4.20e-06), immunoglobulin kappa light chain (IGK) (p-value 4.33e-06), and immunoglobulin lambda light chain (IGL) (p-value 1.19e-05). Several IGK (CQQYGNSPRTF, CQQYHSTPYTF, CQQSYTTPTF, CQQYYSPPWTF, CQQRGSWPLTF, CQQYGSAPYTF) and IGL (CCSYAGSYTHWVF, CQSYDTSLSGWVF) sequences were found to overlap Day9 TBI samples. Of the top 500 IGH chains, only nine were observed (at low levels) in other critical care patients, with few of these antibodies detected in the comparison of 16,243 additional blood PAXgene tube RNAseq datasets in NCBI SRA (eFigure S8).

Serum metabolomics indicates widespread Temporal perturbations

Similar to transcriptomics, the metabolite PCA plot shows clustering of Day0 and Day2 TBI samples with separate clusters of Day9 TBI and control cohorts (Fig. 3A).

Fig. 3.

Fig. 3

Open in a new tab

Metabolomic data for TBI cohort (A) A score plot of Principal Component Analysis (PCA) created using metabolomics data from control and TBI timepoints across the first three principal components (PC’s). An interactive PCA is available at 10.6084/m9.figshare.30380188.v1. The top 20 loading components for each axis can be found at 10.6084/m9.figshare.30379144.v1. (B) Hierarchical clustering of metabolites with significant differential abundance in TBI patients versus controls

Significant metabolites by timepoint

Day0 TBI metabolites relative to controls were involved in multiple lipid pathways (sphingolipids, phosphosphingolipids), glutamate metabolism, and amino acid metabolism (eFigure S10). A total of 12 metabolites or metabolism indicator ratios produced a q-value < 0.05 (eFigure S9) in TBI Day 0 vs. Control. Higher on Day0: LDH (lactate dehydrogenase) activity, phosphatidylcholines (sum.of.PUFA.PC.O.s, sum.of.PUFA.PCs), L-lactic acid, aconitic acid, and indole derivatives (sum.of.indoles). Higher in controls: sphingomyelins (SM,34:1, SM 36.1), and amino acids (L-ornithine, lysine, and isoleucine). Of note, at Day0 there is a marked decrease in ornithine and elevation of putrescine reflective of ornithine decarboxylase enzyme activation (eFigure S11), returning to control levels in most patients by Day2. Only two significant metabolites/metabolic indicators were seen TBI Day 0 vs. TBI Day 2. Higher on Day2: glycosphingolipid (Hex2Cer.d.18:1/16.0), and sphingomyelin (SM.34:1), Day9 relative to Day0 or control metabolites were enriched for sphingolipids and glycosphingolipids (eFigure S10). Six metabolites/metabolic indicators were significant in TBI Day 0 vs. TBI Day 9 (higher Day0: triacylglycerol (TG.18:1_36:2), LDH.Activity; higher Day9: sphingomyelins (0- SM.34:1, SM.36:1, SM.42:2), ceramide (Cer.d18:1/24:1), and 21 metabolites/metabolic indicators in TBI Day9 vs. control. Higher Day9 sphingomyelins (SM.36:1, SM.42:2, SM.35:1, SM.34:1, SM 44:2, sum.of.LCFA.SMs), ceramides (Cer.d18:1/20:0-OH, Cer.d18:1/18:0, Cer.d18:1/24:1), glycosylceramide (Hex2Cer.d18:1/20:0), triacylglycerols (TG.22:5_32:0, TG.18.2_32.0), and acylcarnitine (MA.NBS). Higher in controls: triacylglycerols (TG.18:1_36:1, TG.18:1_36:2), amino acids (CPS.deficiency.NBS), indole derivatives (3-IPA), phosphatidylserines (sum.of.MUFA.PSs). Integration of significant metabolites produced several clusters of metabolite profiles over the samples (Fig. 3B).

Metabolite correlations with clinical parameters

Multiple comparisons included known metabolites involved in Alzheimer’s disease, including IPA (eFigure S12), which showed trends correlating with multiple clinical parameters at Day9. Several metabolites, including creatinine and lactate, correlated with the Pediatric Risk of Mortality (PRISM) III score and initial Glasgow Coma Score (GCS) (eFigure S13).

Integrated clinical, transcriptomic, and metabolomic analysis

The clinical data, as assessed through eight different metrics, showed strong correlations with transcripts and metabolites (Fig. 4, top). Of the top 20 transcript or metabolite correlations, the PRISM_Score had 14 transcripts (enriched for endoplasmic reticulum targeting) with values > 0.7 or <−0.7, while the Hospital_LOS had 13 metabolites. GOSE_DC had 8 transcripts (enriched for mitochondrial translation elongation) correlated at Day2, while the PRISM_Score had 10 metabolites correlated at Day2. ISS had 16 Day9 transcripts correlated while GOSE_DC had all 20 top Day9 metabolites correlated. Initial_GCS had the highest correlation of all metrics with Day0 TPGS2-201 (0.92), a member of the neuronal polyglutamylase complex. PICU_LOS had a high correlation with Day2 CNN2-209 (0.93), Initial_GCS with Day9 UCP2-208 (0.98), and GOSE_DC with Day9 TG 20:1_31:0 (−0.99).

Fig. 4.

Fig. 4

Open in a new tab

Top Correlations of clinical metrics, transcriptomics, and metabolomics The top three panels display the correlations of various clinical metrics relative to day 0, day 2, or day 9, with the top 20 metabolites or transcripts. The bottom panel shows the top transcriptomic-based panels relative to all day TBI significant metabolites. For all charts, Pearson correlations above 0.7 are shown with red lines representing positive and blue negative correlations scaled with darker colors closer to 1 and lighter closer to 0.7. Below each panel is a heatmap of all trait correlations with one minus Pearson correlation clustering

Cross-day, cross-platform correlations revealed multiple novel insights from immune cell regulation, various biological pathways, and with the immune repertoire dynamics (Fig. 5, bottom). The Day0_T cells CD8 transcript levels were correlated with the Day9_Xanthine (−0.94), Day0_Macrophages M0 with Day0_PC O-42:5 (0.91) and Day9_TrpBetaine (0.89), Day0 Transcript Biomarkers with Day0_AconAcid (0.83), and Day9 Transcript Biomarkers with Day9_LPS 20:0 (0.88) and Day0_PC O-36:0 (−0.92). Day0_GO:0045321 (Leukocyte activation, 28 genes) correlated with Day0_SM 41:2 (0.92), Day0_PI 18:2_20:4 (0.88), and Day0_Hex2Cer d18:1/20:0 (0.87). Day0_GO:0002376 (Immune system process, 59 genes) correlated with Day2_SM 34:1 (0.87), Day0_PI 18:2_20:4 (0.87), and Day9_PI 16:1_18:2 (0.87). Day9_EFO:0004528 (Mean corpuscular hemoglobin concentration, 11 genes) correlated with Day0_PC O-36:0 (−0.93), Day0_PG 18:2_20:4 (−0.92), Day0_PG 18:2_20:3 (−0.90), and Day9_LPS 20:0 (0.85). Day0_TRAD Vol correlated with Day0_His (−0.83) and Day9_Xanthine (−0.82) while the Day9_IGH Clones/Vol correlated with Day0_Cer d18:1/20:0 (−0.94), Day0_PC 36:4 (−0.94), and Day9_PG 16:0_18:3 (0.85).

Fig. 5.

Fig. 5

Open in a new tab

Summary of discoveries via integrated transcriptomic (blue) and metabolomic (red) analysis over three time points of TBI recovery. ODC1, ornithine decarboxylase 1; PUFA, polyunsaturated fatty acid

Discussion

TBI pathology is initiated by a singular event, aggravated by a series of secondary insults evolving over days, months, and sometimes years [19]. In this original study, our integrated multi-omic approach provided high-resolution insights into the host response and uncovered potential convergent therapeutic windows, establishing a framework for the development of targeted interventions in pediatric severe TBI. This study focuses on the secondary response to find associations between interdependent processes: the whole blood transcriptome (an immuno-cellular signal) and the serum metabolome (a reflection of the circulatory milieu) in a clinical setting. The secondary injury consisted of adaptive and maladaptive changes, which were primarily inflammatory, immunological and metabolic perturbations, as shown before [20]. Despite individual differences in ages and mechanisms of injury, omics data indicated that the post-traumatic response was relatively uniform in this cohort. The first two timepoints reflect inflammatory mechanisms as well as immune activation and metabolic stress, with a unique response in the third timepoint demonstrating incomplete metabolic recovery and a well-established acquired immune response.

We found a significant increase in neutrophil aggregation transcripts across acute and subacute timepoints. The capacity for neutrophils to increase blood-brain barrier permeability in TBI has been previously shown, with migrating neutrophils across central nervous system (CNS) barriers releasing inflammatory and cytotoxic factors, promoting endothelial injury and vascular permeability [21–23]. Our data also show a significant decrease in a number of immune-related transcripts in the TBI population as compared to controls across all timepoints, including leukocyte activation, NK cell response, and T cell-related transcripts (eFigures S4 and S6). This is notable given a growing body of literature describing peripheral immune suppression following severe TBI, with secondary infections being a common complication and leading cause of morbidity and mortality in this population [24, 25].

In our study, gene expression changes involving S100 proteins are also notable. While S100β is a well-established biomarker of pediatric TBI, S100A8/A9 is emerging as a mechanistic driver of neuroinflammation, promoting neutrophil recruitment and reactive oxygen species production [26, 27]. While blood sample limitations in this study did not allow for correlation of RNA changes to protein, the S100A8/A9 pathway is often found highly correlated between RNA and protein changes in most immune activation pathologies, and others have shown significant changes at the protein level in TBI [28–31]. S100A9 has also been implicated in the amyloid-neuroinflammatory cascade and chronic neurodegeneration post-TBI [31]. These proinflammatory pathways could in the future be a therapeutic target. Notably, S100A8/A9 inhibition in preclinical models reduces neuroinflammation and neuronal death, though its therapeutic potential in pediatric and adult TBI remains largely unexplored.

Metabolomic analysis is similarly consistent with the acute inflammatory response and mitochondrial stress, with an increase in glutamate, lactate, and LDH levels, as well as specific lipid changes across timepoints. Higher amounts of phosphatidylcholine are seen in the early phase of injury in this cohort. Lipids make up approximately 60% of the brain’s dry weight [32]. Choline phospholipids are essential membrane lipids that can be transported across the BBB, and have been inversely associated with TBI severity in adults, with higher levels seen in patients with more favorable outcomes, implicating neuroprotective effects [33].

One unique approach in pediatric patients to develop pathway-level insights and identify potential therapeutic pathways is to integrate correlations between indicators of severity, such as GCS, ISS, PRISM, and GOS-E scores, and the transcriptome and metabolome. Severity scores such as Day0 PRISM III scores and Day9 GOS-E Peds scores correlate negatively with phosphatidylcholine levels. Others have reported early increases in pro-inflammatory lipids following TBI. While phosphatidylcholine levels rise acutely, our data reveal sustained elevations in triacylglycerols, phosphatidylglycerols, and phosphoethanolamine through Day9 post-injury [34]. This lipid surge in the TBI cohort coincides with transcriptomic evidence of heightened neutrophil activity and T cell migration. Excess lipid accumulation may contribute to mitochondrial dysfunction, potentially reflected in the early observed lactate elevation and change in PUFAs namely arachidonic, docosahexaenoic, and linoleic acid associated with redox imbalance [35, 36]. Notably, at least one adult study has linked acute-phase metabolomic profiles with long-term functional outcomes, including GOS-E scores at 3 and 12 months [37]. In our cohort, GOS-E Peds scores were available at 3 months. Notably, some patients remained on hyperalimentation at Day9, which may confound lipid findings. Importantly, neither the TBI nor control groups received propofol at the time of sampling, a common confounder in adult TBI studies due to its lipid emulsion formulation.

The integrated model of our study noted disruptions in the polyamine pathway with increased ornithine decarboxylase (ODC) gene expression connected with elevated plasma putrescine levels. (eFigure S11) These changes have been linked to increased inflammation, neuronal dysfunction and vasogenic edema in rodent models of TBI [38]. Disruption of polyamine homeostasis is hypothesized to lead to a neurotoxic environment contributing to secondary brain injury, and approaches that counteract polyamine catabolism may be a promising therapeutic approach [39].

Immunologic changes as evidenced by early increases in T-cell marker gene expression and an unexpected early clonal B-cell expansion suggest sequestered epitopes triggering an acquired immune response. Adult studies indicate post-TBI adaptive immune responses, both T-cell dependent and independent, have significant functional effects [40, 41]. Although autoantibody production following TBI has been previously reported, the specificity and functional significance of these autoantibodies in children remain poorly defined. Whether they contribute to repair or exacerbate injury warrants urgent investigation [42]. Our study suggests that antibody production begins early enough to have therapeutic implications, being initiated in the subacute phase of injury. This is consistent with other studies highlighting B-cell behavior [43–45]. Given the antibody sequence resolution, we demonstrate the specificity of these antibodies to TBI, relative to the immune profiles of thousands of other pathologies. This study highlights the power of transcriptomics to identify clonal activity and antibody production.

Gut barrier dysfunction and impaired brain-gut homeostasis are known to occur post-TBI [20, 46]. RNA and metabolomic data indicate early gut barrier loss, evidenced by bacterial transcripts found in the blood from gut flora, and secondary bile acids, indoles, glutamine, and putrescine in circulation (eFigures S5 and S11). While microbiome analysis wasn’t performed, early gut dysbiosis, with reduced commensals and increased pathogens, is known to drive inflammation [47–50]. This study demonstrates the novel finding of decreased levels of a neuroprotective gut-derived metabolite, indole-3-propionic acid (IPA), after pediatric severe TBI. This is consistent with adult data showing decreased IPA levels in varying severities of TBI [5]. IPA is an indole derivative produced exclusively from dietary tryptophan by gut microbiota such as Lactobacillus, Clostridium, and Peptostreptococcus species. It is capable of crossing the BBB, with animal studies showing that IPA supplementation can decrease neuronal functional loss through regulation of astrocyte and microglial-mediated inflammation, and an increase in neuroprotective neurotrophins, such as brain-derived neurotrophic factor and nerve growth factor [51, 52]. Recent pre-clinical studies are examining IPA as a therapeutic supplement to promote neurologic recovery for neurodegenerative diseases, arterial ischemic stroke, hypoxic-ischemic injury, and peripheral nerve injury [1, 53–55]. Notably, although the lowest IPA levels were seen at later timepoints, higher IPA levels in this cohort were associated with mortality. It is unclear whether this finding relates to an increase in stress response or increase in gut permeability in these patients. While prior studies demonstrate anti-inflammatory and anti-oxidant effects of IPA, there exists evidence that at higher doses IPA could induce mitochondrial dysfunction with suppressed mitochondrial respiration [56, 57]. While we did not study the gut microbiome, the changes in IPA levels, and presence of gut bacteria highlights the influence of the gut brain axis and the a role for nutritional interventions promoting neuronal recovery after TBI.

This study has many limitations, including a small sample size, lack of longitudinal omics data for controls, heterogeneity of patients, the presence of polytrauma, and lack of protein-level confirmation of transcriptomic findings. The inclusion of non-TBI trauma controls would be a consideration for future studies, as well as tracking longer-term outcomes as patients move from the ICU for rehabilitation.

Conclusions

This study provides the first integrated analysis of whole blood transcriptomics and plasma metabolomics alongside clinical data in pediatric severe TBI. Our findings reveal both established and novel metabolic and transcriptomic signatures, notably the lipid burst, the involvement of IPA, and immune responses occurring within a critical therapeutic window to potentially mitigate secondary injury. These insights offer promising avenues for targeted interventions in pediatric TBI.

Supplementary Information

Supplementary Material 1. (1.6MB, docx)
Supplementary Material 2. (2MB, docx)

Acknowledgements

Corresponding Author: Elora Hussain, MD, Corewell Health Helen DeVos Children’s Hospital, Michigan State University, 100 Michigan St NE, MC 117, Grand Rapids, MI 49503 email: (elora.hussain@helendevoschildrens.org) Author Contributions: Dr. Rajasekaran had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Drs. Hussain and Prokop are co-first authors.

Author contributions

Dr. Rajasekaran had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Drs. Hussain and Prokop are co-first authors *Concept and design: * Hussain, Prokop, Graham, Rajasekaran*Acquisition, analysis, or interpretation of data: * All the authors Elora Hussain, Jeremy W Prokop, Emily Nonnemacher, Nadia Ashrafi, Ali Yilmaz, Romana Ashrafi Mimi, Abdullah Khalid, Karolis Krinickis, Vilija Lomeikaite, Lena Sanfilippo, Kylie Maxton, Jacob Charron, Charitha Subrahmanya, Austin Goodyke, Annie Needs, Daniel R Woldring, Caleb P Bupp, Nicholas Hartog, Juozas Gordevicius, Stewart F. Graham, Surender Rajasekaran.*Drafting of the manuscript: * Hussain, Rajasekaran, Graham, Prokop*Critical review of the manuscript for important intellectual content: * Hussain, Rajasekaran, Prokop, Graham, Lomeikaite, Hartog, Gordevicius*Statistical analysis: * Prokop, Gordevicius, Lomeikaitė, Krinickis*Administrative, technical, or material support: * Hussain, Rajasekaran, Prokop, Graham, Marlie Vipond, Marlene Gomez-Vergara*Supervision: * Rajasekaran, Graham, Prokop.

Funding

This study was supported by the Helen DeVos Children’s Hospital Foundation. The funder had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Data availability

Mapping data and statistics are available for RNAseq at [https://doi.org/10.6084/m9.figshare.28303880.v1].

Declarations

Ethics approval and consent to participate

The protocol was approved by the Ethics Committee of Helen DeVos Children Hospital as the corresponding institution (2021-096) or the responsible ethics committee of each respective study center and conducted in accordance with the revised Declaration of Helsinki. Informed written consent was obtained from all patients’ legal surrogates prior to inclusion.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Footnotes

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Elora Hussain and Jeremy W. Prokop are equally contributing first authors.

References

  • 1.Dewan MC, Mummareddy N, Wellons JC 3rd, Bonfield CM. Epidemiology of global pediatric traumatic brain injury: qualitative review. World Neurosurg. 2016;91:497–509. e491. [DOI] [PubMed] [Google Scholar]
  • 2.Haydel MJ, Weisbrod LJ, Saeed W. Pediatric head trauma. [Updated 2024 Feb 16]. In: StatPearls. Treasure Island (FLL): StatPearls Publishing; 2025 Jan https://www.ncbi.nlm.nih.gov/books/NBK537029 [PubMed]
  • 3.Shi J, Xiang H, Wheeler K, et al. Costs, mortality likelihood and outcomes of hospitalized US children with traumatic brain injuries. Brain Inj. 2009;23(7):602–11. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Cox CS Jr., Juranek J, Bedi S. Clinical trials in traumatic brain injury: cellular therapy and outcome measures. Transfusion. 2019;59(S1):858–68. [DOI] [PubMed] [Google Scholar]
  • 5.Orešič M, Posti JP, Kamstrup-Nielsen MH, et al. Human serum metabolites associate with severity and patient outcomes in traumatic brain injury. EBioMedicine. 2016;12:118–26. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Cavill R, Jennen D, Kleinjans J, Briede JJ. Transcriptomic and metabolomic data integration. Brief Bioinform. 2016;17(5):891–901. [DOI] [PubMed] [Google Scholar]
  • 7.Woodward A, Pandele A, Abdelrazig S, et al. Integrated metabolomics and transcriptomics using an optimised dual extraction process to study human brain cancer cells and tissues. Metabolites. 2021 Apr 14;11(4):240 10.3390/metabo11040240. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Patro R, Duggal G, Love MI, Irizarry RA, Kingsford C. Salmon provides fast and bias-aware quantification of transcript expression. Nat Methods. 2017;14(4):417–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Bolotin DA, Poslavsky S, Mitrophanov I, et al. MiXCR: software for comprehensive adaptive immunity profiling. Nat Methods. 2015;12(5):380–1. [DOI] [PubMed] [Google Scholar]
  • 10.Wood DE, Lu J, Langmead B. Improved metagenomic analysis with kraken 2. Genome Biol. 2019;20(1):257. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Steen CB, Liu CL, Alizadeh AA, Newman AM. Profiling cell type abundance and expression in bulk tissues with CIBERSORTx. Methods Mol Biol. 2020;2117:135–57. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Ravanbakhsh S, Liu P, Bjorndahl TC, et al. Accurate, fully-automated NMR spectral profiling for metabolomics. PLoS ONE. 2015;10(5):e0124219. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Wu T, Hu E, Xu S, et al. ClusterProfiler 4.0: a universal enrichment tool for interpreting omics data. Innovation. 2021;2(3):100141. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Rothhammer V, Mascanfroni ID, Bunse L, et al. Type I interferons and microbial metabolites of tryptophan modulate astrocyte activity and central nervous system inflammation via the aryl hydrocarbon receptor. Nat Med. 2016;22(6):586–97. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Ritchie ME, Phipson B, Wu D, et al. Limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res. 2015;43(7):e47. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Allison M, Mathews J, Gilliland T, Mathew SO. Natural killer cell-mediated immunotherapy for leukemia. Cancers (Basel). 2022. Feb 8;14(3):843. 10.3390/cancers14030843. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Cruciani C, Puthenparampil M, Tomas-Ojer P, et al. T-cell specificity influences disease heterogeneity in multiple sclerosis. Neurol Neuroimmunol Neuroinflamm. 202117;8(6):e1075. 10.1212/NXI.0000000000001075. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Wang G, Durussel J, Shurlock J, et al. Validation of whole-blood transcriptome signature during microdose recombinant human erythropoietin (rHuEpo) administration. BMC Genomics. 2017;18(Suppl 8):817. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Masel BE, DeWitt DS. Traumatic brain injury: a disease process, not an event. J Neurotrauma. 2010;27(8):1529–40. [DOI] [PubMed] [Google Scholar]
  • 20.Corps KN, Roth TL, McGavern DB. Inflammation and neuroprotection in traumatic brain injury. JAMA Neurol. 2015;72(3):355–62. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Hubbard WB, Dong JF, Cruz MA, Rumbaut RE. Links between thrombosis and inflammation in traumatic brain injury. Thromb Res. 2021;198:62–71. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Makinde HM, Cuda CM, Just TB, Perlman HR, Schwulst SJ. Nonclassical monocytes mediate secondary injury, neurocognitive outcome, and neutrophil infiltration after traumatic brain injury. J Immunol. 2017;199(10):3583–91. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Santos-Lima B, Pietronigro EC, Terrabuio E, Zenaro E, Constantin G. The role of neutrophils in the dysfunction of central nervous system barriers. Front Aging Neurosci. 2022;14:965169. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Meisel C, Schwab JM, Prass K, Meisel A, Dirnagl U. Central nervous system injury-induced immune deficiency syndrome. Nat Rev Neurosci. 2005;6(10):775–86. [DOI] [PubMed] [Google Scholar]
  • 25.Hazeldine J, Lord JM, Belli A. Traumatic brain injury and peripheral immune suppression: primer and prospectus. Front Neurol. 2015;6:235. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Friess SH, Kilbaugh TJ, Huh JW. Advanced neuromonitoring and imaging in pediatric traumatic brain injury. Crit Care Res Pract. 2012;2012:361310. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Shi G, Cao Y, Xu J, et al. Inhibition of S100A8/A9 ameliorates neuroinflammation by blocking NET formation following traumatic brain injury. Redox Biol. 2025;81:103532. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Swindell WR, Remmer HA, Sarkar MK, et al. Proteogenomic analysis of psoriasis reveals discordant and concordant changes in mRNA and protein abundance. Genome Med. 2015;7(1):86. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Xu Z, Cheng C, Kong R, et al. S100A8 and S100A9, both transcriptionally regulated by PU.1, promote epithelial-mesenchymal transformation (EMT) and invasive growth of dermal keratinocytes during scar formation post burn. Aging. 2021;13(11):15523–37. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Yang Y, Shen L, Xu M, Chen L, Lu W, Wang W. Serum calprotectin as a prognostic predictor in severe traumatic brain injury. Clin Chim Acta. 2021;520:101–7. [DOI] [PubMed] [Google Scholar]
  • 31.Wang C, Iashchishyn IA, Pansieri J, et al. S100A9-driven amyloid-neuroinflammatory cascade in traumatic brain injury as a precursor state for alzheimer’s disease. Sci Rep. 2018;8(1):12836. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Hamilton JA, Hillard CJ, Spector AA, Watkins PA. Brain uptake and utilization of fatty acids, lipids and lipoproteins: application to neurological disorders. J Mol Neurosci. 2007;33(1):2–11. [DOI] [PubMed] [Google Scholar]
  • 33.Thomas I, Dickens AM, Posti JP, et al. Serum metabolome associated with severity of acute traumatic brain injury. Nat Commun. 2022;13(1):2545. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Hogan SR, Phan JH, Alvarado-Velez M, et al. Discovery of lipidome alterations following traumatic brain injury via high-resolution metabolomics. J Proteome Res. 2018;17(6):2131–43. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Lamade AM, Anthonymuthu TS, Hier ZE, Gao Y, Kagan VE, Bayir H. Mitochondrial damage & lipid signaling in traumatic brain injury. Exp Neurol. 2020;329:113307. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Pomatto LCD, Davies KJA. Adaptive homeostasis and the free radical theory of ageing. Free Radic Biol Med. 2018;124:420–30. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Banoei MM, Lee CH, Hutchison J, et al. Using metabolomics to predict severe traumatic brain injury outcome (GOSE) at 3 and 12 months. Crit Care. 2023;27(1):295. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Henley CM, Muszynski C, Cherian L, Robertson CS. Activation of ornithine decarboxylase and accumulation of Putrescine after traumatic brain injury. J Neurotrauma. 1996;13(9):487–96. [DOI] [PubMed] [Google Scholar]
  • 39.Zahedi K, Huttinger F, Morrison R, Murray-Stewart T, Casero RA, Strauss KI. Polyamine catabolism is enhanced after traumatic brain injury. J Neurotrauma. 2010;27(3):515–25. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Ankeny DP, Guan Z, Popovich PG. B cells produce pathogenic antibodies and impair recovery after spinal cord injury in mice. J Clin Invest. 2009;119(10):2990–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41.Schwartz M, Kipnis J. Protective autoimmunity: regulation and prospects for vaccination after brain and spinal cord injuries. Trends Mol Med. 2001;7(6):252–8. [DOI] [PubMed] [Google Scholar]
  • 42.Ankeny DP, Popovich PG. B cells and autoantibodies: complex roles in CNS injury. Trends Immunol. 2010;31(9):332–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43.Needham EJ, Helmy A, Zanier ER, Jones JL, Coles AJ, Menon DK. The immunological response to traumatic brain injury. J Neuroimmunol. 2019;332:112–25. [DOI] [PubMed] [Google Scholar]
  • 44.Chenouard A, Chesneau M, Braza F, et al. Phenotype and functions of B cells in patients with acute brain injuries. Mol Immunol. 2015;68(2 Pt A):350–6. [DOI] [PubMed] [Google Scholar]
  • 45.Zhang Z, Zoltewicz JS, Mondello S, et al. Human traumatic brain injury induces autoantibody response against glial fibrillary acidic protein and its breakdown products. PLoS One. 2014;9(3):e92698. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 46.El Baassiri MG, Raouf Z, Badin S, Escobosa A, Sodhi CP, Nasr IW. Dysregulated brain-gut axis in the setting of traumatic brain injury: review of mechanisms and anti-inflammatory pharmacotherapies. J Neuroinflammation. 2024;21(1):124. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 47.Foster JA, Lyte M, Meyer E, Cryan JF. Gut microbiota and brain function: an evolving field in neuroscience. Int J Neuropsychopharmacol. 2016. 10.1093/ijnp/pyv114. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 48.Ma Q, Xing C, Long W, Wang HY, Liu Q, Wang RF. Impact of microbiota on central nervous system and neurological diseases: the gut-brain axis. J Neuroinflammation. 2019;16(1):53. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 49.Hanscom M, Loane DJ, Shea-Donohue T. Brain-gut axis dysfunction in the pathogenesis of traumatic brain injury. J Clin Invest. 2021 Jun 15:131(12):e143777. 10.1172/JCI143777. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 50.Hayakawa M, Asahara T, Henzan N, et al. Dramatic changes of the gut flora immediately after severe and sudden insults. Dig Dis Sci. 2011;56(8):2361–5. [DOI] [PubMed] [Google Scholar]
  • 51.Kim CS, Jung S, Hwang GS, Shin DM. Gut microbiota indole-3-propionic acid mediates neuroprotective effect of probiotic consumption in healthy elderly: a randomized, double-blind, placebo-controlled, multicenter trial and in vitro study. Clin Nutr. 2023;42(6):1025–33. [DOI] [PubMed] [Google Scholar]
  • 52.Wang T, Chen B, Luo M, et al. Microbiota-indole 3-propionic acid-brain axis mediates abnormal synaptic pruning of hippocampal microglia and susceptibility to ASD in IUGR offspring. Microbiome. 2023;11(1):245. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 53.Hwang IK, Yoo KY, Li H, et al. Indole-3-propionic acid attenuates neuronal damage and oxidative stress in the ischemic hippocampus. J Neurosci Res. 2009;87(9):2126–37. [DOI] [PubMed] [Google Scholar]
  • 54.Pappolla MA, Perry G, Fang X, Zagorski M, Sambamurti K, Poeggeler B. Indoles as essential mediators in the gut-brain axis. Their role in Alzheimer’s disease. Neurobiol Dis. 2021;156:105403. [DOI] [PubMed] [Google Scholar]
  • 55.Serger E, Luengo-Gutierrez L, Chadwick JS, et al. The gut metabolite indole-3 propionate promotes nerve regeneration and repair. Nature. 2022;607(7919):585–92. [DOI] [PubMed] [Google Scholar]
  • 56.Gesper M, Nonnast ABH, Kumowski N, et al. Gut-derived metabolite indole-3-propionic acid modulates mitochondrial function in cardiomyocytes and alters cardiac function. Front Med. 2021;8:648259. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 57.Chen SJ, Chen CC, Liao HY, et al. Alteration of gut microbial metabolites in the systemic circulation of patients with Parkinson’s disease. J Parkinsons Dis. 2022;12(4):1219–30. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Material 1. (1.6MB, docx)
Supplementary Material 2. (2MB, docx)

Data Availability Statement

Mapping data and statistics are available for RNAseq at [https://doi.org/10.6084/m9.figshare.28303880.v1].

Articles from Critical Care are provided here courtesy of BMC

RESOURCES