Introduction

Burn injuries are linked to sustained hypermetabolism and immune dysfunction that may be linked to increased risk of long-term morbidities such as infections, cardiovascular diseases and cancers1,2. One significant long-term impact is increased hospitalisation for mental health conditions for more than 30 years after a paediatric burn injury3. Epidemiological studies have shown that patients who experienced a burn injury when they were children were more likely to be admitted to hospital for mental health conditions. These patients also had a significantly longer length of stay when admitted compared to non-burn injured patients, suggesting an increase in the severity of the conditions experienced3,4. Whilst this increase may be in part due to psychological impacts of paediatric burns5, many of these patients have relatively small injuries with limited scarring3. In addition, burns commonly occur at early ages such that these children are unlikely to have any conscious memory of the original traumatic event6, suggesting other impacts of paediatric burn trauma may be important.

Post burn inflammation can reach the central nervous system as elevated inflammatory markers damage the integrity of the blood-brain barrier, increasing its permeability7. Studies of severe burns have found evidence of metabolic, molecular, cognitive, endocrine and neuropathic sequela in the brain. These include elevated stress hormone levels, changes in glucose metabolism, dysfunction in motor and sensory pathways, and impacted memory7. However, little is known about the impact of non-severe (considered as less than 10% of the total body surface area (TBSA)) burns that account for up to 84.4% of paediatric admissions in developed countries8. Additionally, the majority of studies have focused on the acute period of recovery from the burn, with little yet known about long-term, sustained impacts of burn injury on the central nervous system (CNS). Moreover, it is also not clear whether this impact on the brain is unique to burn trauma or if it is also present after non-burn trauma (NBT), which would impact a much greater number of patients. The aim of this study was to investigate the long-term physiological impact of non-severe burn injury (NSBI) and NBT on the brain to understand how non-severe burn trauma may be linked to sustained physiological changes in the CNS that may consequently increase the risk of admission to hospital for mental health conditions.

Methods

Animals

Nine-week-old female C57BL/6J mice (n = 115 total) were sourced from the Animal Resource Centre and housed at the Pre-clinical Facility of the Animal Care Services in The University of Western Australia’s Crawley campus with a 12-hour light cycle. The mice underwent isoflurane anaesthesia and were assigned to one of three interventions through simple randomization. One group received a full thickness 7–8% TBSA contact burn injury, another group received a NBT which was a full-thickness surgical excision of the same size, and the last control group received no injury (sham) as described by Valvis et al.9. Injury site was on the right dorsal flank for the burn and the medial posterior dorsal side for the excision so as not to restrict movement. All mice (including sham) received 0.1 mg/kg dose of buprenorphine subcutaneously to minimise pain during recovery and paracetamol in their water for five days post-surgery. No confounders were controlled for in the experiments. All experiments were conducted in accordance to the ARRIVE guidelines. All experiments were conducted according to the National Health and Medical Research Council guidelines, with approval by the University of Western Australia Animal Ethics Committee (RA 03/100/1624 and 2022/ET000172). The study is reported in accordance with ARRIVE guidelines.

Behavioural tests

For the behavioural tests, 45 mice (15 per group) underwent four behavioural tests at baseline (1 week before the injury), 1 month post intervention, and at 3 months post intervention. Behavioural tests for anxious behaviour included two tests that investigated exploration of novel and open areas, which were the open field test (OFT)10, and the elevated plus maze (EPM) test11. Two other behavioural tests investigated depression-like behaviour; the sucrose preference test (SPT) which tested reward centres and anhedonic behaviour through a drinking choice of either plain water or sucrose water to provide a rewarding experience for the mice12, and the forced swim test (FST) which tested willingness to escape from stressors13. Full descriptions of the methods of the behavioural tests are in the supplementary material.

Euthanasia and sample collection

Three months after the injury procedure, mice were euthanised through an intraperitoneal injection of 160 mg/kg Pentobarbitone. Before the heart stopped, the plasma was collected through cardiac acupuncture. The brain tissue was dissected, and the hippocampus, the frontal cortex, the cerebellum, and the piriform cortex were isolated for RNA sequencing and High-Resolution Magic-Angle Spinning (HR-MAS) Nuclear Magnetic Resonance (NMR) spectroscopy. The rest of the brain tissue was frozen and stored for subsequent metabolic phenotyping. Different sections of the frontal cortex were used for different experiments, where the prefrontal cortex was used for the RNA sequencing experiments, while as the whole frontal cortex with the anterior olfactory nucleus was used for the HR-MAS NMR experiment as more volume of the brain was required for the experiment.

RNA sequencing

The RNeasy mini kit (Cat No./ID: 74104, QIAGEN, Germany) was used to extract RNA in the brain sections according to manufacturer’s instructions. Aliquots of 10 µg of the obtained processed samples were sent for sequencing at the Australian Genome Research Facility, where 1 µg of total RNA was used for sequencing via the Illumina Nova-Seq Control Software (v1.6.0.). Real Time Analysis (v3.4.4) for real-time base calling was used, and it produced a 100 bp single end run. The Illumina bcl2fastq (v2.20.0.422) pipeline was used to generate the primary sequence data.

HR-MAS NMR

Approximately 12 mg of the hippocampus, prefrontal cortex, and cerebellum frozen biopsies were homogenised with 10 µL of buffer (1.5 M KH2PO4, 2mM NaN3, 0.1% trimethylsilyl propionate-[2,2,3,3- 2H4] (TSP), pH 7.4) and rapidly introduced into a HR-MAS disposable insert. This was sealed and introduced into a 4 mm rotor and immediately transferred into the HR-MAS NMR spectrometer. The acquisition experiment was started 5 min after equilibrium in the spectrometer. Spectra were recorded on a Bruker 600 MHz spectrometer fitted with a 4 mm dual band (1 H and 13 C) probe). Samples were spun at 4500 Hz at 4˚C. Two experiments were acquired, the standard 1D experiment with solvent pre-saturation and a spin-echo CPMG experiment (65 K data points, and a spectral width of 21 ppm). The number of scans for both experiments was varied according to the brain mass, for 12 mg 88 scans was used. When mass was reduced an increase in scan number to achieve the same signal to noise ratio was implemented.

Lipid analysis

Leftover brain tissue and plasma were thawed, and the brains were homogenised with a volume of 10 nM of ammonium acetate equal to 1:1 weight: volume ratio of the brains’ weight. 20 µl of the brain homogenate were added to 180 µl of 2-isopropanol (IPA) and put on a thermomixer (Eppendorf® ThermoMixer® C, Eppendorf, Germany) at room temperature. Samples were shaken at 2000 g for 15 min. Lipid extraction and analysis was then performed in accordance with a previously published method14. In brief, 10 µl of the homogenate solution was added to 90 µl of IPA mixed with lipid internal standards (Lipidyzer™ Internal Standards Kit from Sciex (Framingham, MA, USA), SPLASH™ LIPIDOMIX™, Lyso PI 17:1, Lyso PG 17:1, and Lyso PS 17:1 (Avanti Polar Lipids, Alabaster, AL, USA)). This procedure was replicated with plasma (10 µl of plasma with 90 µl of IPA extractant) with all the samples placed in a 96-well plate and sealed using a foil film for a Ultra-high-performance liquid chromatography–tandem mass spectrometry (UHPLC–MS/MS) analysis through a targeted approach using predefined MRM transitions. Pooled quality controls (QC) made from a mix of a few samples from all intervention groups across each sample type were used as a surrogate replicate quality control to demonstrate analytical and signal reproducibility. Additionally, commercially available pooled human plasma (BioIVT, Westbury, NY, USA) was analysed alongside the pooled QCs to act as a long-term reference sample, and both QC types were and they were analysed at frequent intervals throughout the run.

Tryptophan pathway analysis

Brain and plasma samples were thawed, and tryptophan, along with 17 other catabolites within its metabolic pathway were quantified from 50 µl of the brain homogenate which included the ammonium acetate and internal standards, and from 50 µl of the plasma, using a method described by Whiley et al.15. Catabolite concentrations were determined using UHPLC–MS/MS, with analyses conducted on a Waters Acquity UPLC® system (Waters Corp., Milford, MA, USA) coupled to a Waters Xevo TQ-XS mass spectrometer (Waters Corp., Wilmslow, UK). Sample extraction was automated with a Biomek i5 system. Before protein precipitation, 20 µL of stable isotope-labelled internal standards were added to each sample, followed by 250 µL of methanol containing 2 mM ammonium formate. After thorough mixing, samples were transferred to Phenomenex PHREE phospholipid removal solid-phase extraction plates (pre-wetted with methanol) placed above 700 µL Waters 96-well collection plates. The plates were centrifuged at 4000 rpm for 10 min, washed with an additional 150 µL of methanol containing 2 mM ammonium formate, and centrifuged again for 5 min at 4000 rpm. The collected eluents were dried using a SpeedVac vacuum concentrator (Thermo Fisher, MA), then reconstituted in 100 µL of water containing 0.1% formic acid for UHPLC–MS/MS analysis. In total, 29 brain samples (10 Burn, 9 Excision, and 10 Sham) and 44 plasma samples (15 Burn, 14 Excision, and 15 Sham) were analysed for tryptophan pathway catabolites.

Analysis pipelines

Behavioural tests analysis

Statistical analysis was completed using the GraphPad Prism software (GraphPad Software, v8, USA). The Robust regression and Outlier removal (ROUT) test with a false discovery rate of 1%, was used to identify and remove outliers16, and a repeated measure two-way ANOVA with Geisser-Greenhouse correction was performed with Tukey’s test as a post hoc.

RNA sequencing pipeline

The RNA sequencing data was processed on a remote server. Poor quality reads were checked using FastQC (v0.11.3) where the first 15 base pairs were trimmed using the fastp software17. Raw RNA sequences were aligned with the reference genome Mus musculus GRCm39.104 from Ensembl18(http://asia.ensembl.org/Mus_musculus/Info/Index). Using the package DESeq2 (v1.24.0)19, differential gene analysis was performed to quantitate changes in expression levels between the different intervention groups of the samples. P-values of differentially expressed genes (DEGs) were adjusted by the Benjamini-Hochberg false discovery rate correction (p.adj) for multiple comparisons. Significant DEGs were selected based on an absolute log2 fold-change (log2FC) ≥ 1 and p.adj ≤ 0.05. DEG results were visualised using volcano plots and heatmaps using the ggplot2 (v3.5.2), ggrepel (v0.9.6) and pheatmap (v1.0.13) packages in R (v4.4.1, R foundation, Vienna, Austria) within R studio (v1.4.1, R Studio, Boston, MA, USA). For analysis of functions and pathways, overrepresentation analysis of the DEGS list on gene ontology (GO) terms and Kyoto Encyclopaedia of Genes and Genomes (KEGG)20,21,22 based pathway analysis was run using ClusterProfiler (v3.12.0)23. GO pathway significance was set to an adjusted p-value < 0.05, and it included three types of pathways; biological functions (BP), molecular functions (MF), and cell component (CC) pathways.

HR-MAS NMR analysis

For the HR-MAS NMR data, each spectrum was referenced manually to TSP (0 ppm) using Topspin 3.6.2. All subsequent NMR data processing was completed in R using in-house open-source packages nmr-parser (v3.0.3) and nmr-spectra processing (v19.0.1)24,25. The spectra were baseline-corrected using an asymmetric least squares routine; spectral regions corresponding to the residual water resonance signal (δ 4.95 − 5.30) or predominantly noise (δ < 0.5 and δ > 9.5) were excluded from analyses. Spectra were normalised via a probabilistic quotient method using the median spectrum as reference. The integration of the metabolite regions was achieved by summation of a fixed spectral region after preprocessing to get relative concentrations.

Data were mean-centered and scaled to unit-variance prior to multivariate modelling. Principal component analysis (PCA) was used to assess the main sources of structured variation within the dataset. Data was acquired in two batches, which required orthogonal projection to latent structures discriminant analysis (OPLS-DA) models to be constructed to compare the interventions using batch 1. Statistical tests were performed using GraphPad Prism software (GraphPad Software, v8, USA). Univariate t tests were performed for pairwise comparisons of the analyte concentrations between the intervention groups.

Lipid and tryptophan pathway analysis

For lipid analysis, Skyline v23.1 open-source software was used to pre-process the raw lipid spectra (MacCross Labs, Seattle, USA26, and data cleaning was performed in R studio. For data cleaning, lipid species were filtered by < 30% relative standard deviation (RSD) in the replicate long-term reference QC samples, where if samples showed > 50% missing values, they were removed in the analysis. The remaining missing values were treated as a lower limit of detection imputed by using the minimum concentration divided by two. Multivariate statistical analysis of the data was performed on SIMCA® v16 (Sartorious, Göttingen, Germany) to produce the OPLS-DA plots, and Graphpad Prism software was used to perform Mann-Whitney U pairwise comparisons and create dot plots for significantly different lipid subclasses. Lipid subclass data incorporated the sum of all lipids specied within the lipid subclass for each sample.

For the tryptophan analysis, raw data were pre-processed for peak integrations and the calculation of catabolite concentrations using the TargetLynx package in MassLynx v4.2 (Waters Corp. Cambridge MA, USA). Catabolite quantification data were scaled to account for the 1:1 dilution of sample with water and converted to appropriate scales across both assays. Mann-Whitney U tests were performed on each of the metabolites tested for pairwise comparisons of the catabolite concentrations between the intervention groups within the GraphPad Prism software.

Results

Behavioural tests

In the SPT, burn injured animals showed a significant increase in sucrose consumption at the 3-month timepoint compared to baseline (Fig. 1A), with a trend for increased consumption observed at the 1-month timepoint. No significant change in sucrose consumption was observed in either the excision or sham groups over time (Fig. 1B-C). No significant differences in sucrose consumption were observed between groups at any timepoint.

Fig. 1
Fig. 1The alternative text for this image may have been generated using AI.
Full size image

Mean sucrose consumption and standard error of mean per mouse in the SPT across intervention groups. Sucrose consumption was tested at three timepoints; baseline (1 week before intervention), 1 month and 3 months post intervention. The three intervention groups were: burn (A), excision (B), and sham (C). Sucrose consumption was significantly increased in the burn group between baseline and the 3-months timepoint. Average sucrose consumption is calculated from the amount consumed per cage (n = 27 per intervention group, with n per cage between 3–5). * p < 0.05.

In the EPM, FST, and OFT behavioural tests, significant changes were evident across timepoints in all three groups including the sham as shown in the supplementary material where many timepoint comparisons are showing p < 0.0001 significance. However, no significant differences were found between groups, indicating that this change is likely due to habituation from multiple exposures to the testing environment rather than a response to burn or non-burn trauma (Supplementary Fig, 1–5, Supplementary Tables 1–6).

RNA sequencing

Analysis of hippocampal samples revealed 59 significantly DEGs between the burn and sham group (n burn = 10, n sham = 9), and 14 DEGs between the burn and excision group (n burn = 10, n excision = 10). For the prefrontal cortex, 31 DEGs were identified between the burn and sham group (n burn = 10, n sham 8), and 18 DEGs were found between the burn and excision group (n burn = 10, n excision = 10). There were no DEGs between the excision and sham groups in either of these regions. For the piriform cortex, no DEGs were found between the burn and sham groups. However, 55 DEGs were found between the burn and excision groups (n burn = 5, n excision = 4), and one DEG was found between the excision and sham groups (n excision = 4, n sham = 5). The genes from these comparisons are listed in Supplementary Table 7. There were 55 downregulated DEGs and 4 upregulated DEGs in the hippocampus in the burn vs. sham group comparison, while there were 12 downregulated and 19 upregulated in the prefrontal cortex, and 28 downregulated and 27 upregulated in the piriform cortex in the burn vs. sham comparison. Heatmaps and volcano plots of these DEGs are shown in Supplementary Fig. 6.

Analysis of gene pathways using the GO enrichment analysis on the intervention groups revealed 55 significant Biological pathways (BP) between the burn and sham groups in the hippocampus and one KEGG pathway that was significant, while there were 65 significant BP and no significant KEGG pathways between the burn and excision groups in the hippocampus. In the prefrontal cortex, there were 115 significant BP between the burn and sham groups and 10 KEGG pathways, while there were 11 BP and one KEGG pathway that were significant between the burn and excision groups. Interestingly, the highest number of BP pathways between groups was between the burn and excision groups in the piriform cortex, with 372 significant BP and 27 KEGG pathways. Many of the BP were related to regulatory control of RNA, stress responses, extracellular matrix and cognition/memory, suggesting broad changes across pathways and potentially a sustained change in these regions of the brain after burn trauma (Fig. 2). The full list of pathways including the KEGG and the BP, MF and CC GO pathways is shown in Supplementary Table 8.

Fig. 2
Fig. 2The alternative text for this image may have been generated using AI.Fig. 2The alternative text for this image may have been generated using AI.Fig. 2The alternative text for this image may have been generated using AI.
Full size image

Dot plots and concept network plots for significant gene pathways using the BP GO enrichment analysis. Dot plots are showing the top 10 gene pathways based on the count of DEGs on each pathway, while the network plots are showing 5 of the gene pathways based on the number of DEGs in each pathway. Plots were created for the intervention group combinations with more than one DEG including: burn vs. sham hippocampus (A, B), burn vs. excision hippocampus (C, D), burn vs. sham prefrontal cortex (E, F), burn vs. excision prefrontal cortex (G, H), burn vs. excision piriform cortex (I, J).

Lipid analysis

After removing missing values and outliers, the datasets consisted of 884 reproducible lipid species in the brain (whole brain with frontal cortex, piriform cortex, hippocampus and cerebellum removed for other experiments) and 834 lipid species in the plasma from 19 lipid subclasses. OPLS-DA modelling showed significant separation in both the brain and plasma samples from the burn injured group compared to both sham and excision injury groups (Fig. 3), where only plots were significant differences in intervention groups were generated.

Fig. 3
Fig. 3The alternative text for this image may have been generated using AI.
Full size image

OPLS-DA plots of brain and plasma lipid analyses showing the pairwise comparisons with significant differences between intervention groups. Comparisons included: Burn (red, n = 9) vs. Sham (blue, n = 10) in the brain (A), Burn (red, n = 9) vs. Excision (yellow, n = 9) in the brain (B), Burn (red, n = 10) vs. Sham (blue, n = 10) in the plasma (C), and Excision (yellow, n = 9) vs. Sham (blue, n = 10) in the plasma (D).

All significant intergroup differences in lipid subclasses (12 out of the 19 lipid subclasses) were found between the burn and sham groups in the brain with the exception of the FFA class which was also significantly different between excision and sham (Supp Table 9). Between the burn and sham groups, all lipid subclasses had higher concentrations in the burn compared to the sham group. The significantly elevated lipid subclasses were ceramides (CER), lactosylceramides (LCER) lysophosphatidylethanolamines (LPEs), free fatty acids (FFAs), lysophosphatidylinositols (LPIs), phosphatidylcholines (PCs), lysophosphatidylcholines (LPCs), lysophosphatidylglycerols (LPGs), phosphatidylglycerols (PGs), cholesteryl esters (CEs), phosphatidylserines (PSs), and phosphatidylinositols (PIs) (Fig. 4).

Fig. 4
Fig. 4The alternative text for this image may have been generated using AI.
Full size image

Dot plots with lipid subclasses which had significantly different concentrations in brain samples between intervention groups using two-way ANOVA Tukey’s multiple comparisons test. Differences between lipid subclasses were only found between the burn and sham groups in the brain samples. (Burn, n = 9; Excision, n = 9; Sham, n = 10).**p < 0.01, *p < 0.05.

In plasma samples, no significant differences between the burn and sham groups were observed in lipid subclasses. However, three lipid subclasses were significantly different between the excision and sham groups, including hexosylceramides (HCERs), phosphatidylethanolamines (PEs), and PIs which were also significantly different between the burn and excision groups (Fig. 5).

Fig. 5
Fig. 5The alternative text for this image may have been generated using AI.
Full size image

Dot plots with lipid subclasses which had significantly different concentrations in plasma samples between intervention groups using two-way ANOVA Tukey’s multiple comparisons test. Differences between lipid subclasses were only found between the burn and sham groups in the brain samples. (Burn, n = 9; Excision, n = 9; Sham, n = 10) **p < 0.01, * p < 0.05.

Tryptophan pathway analysis

The concentrations of 18 catabolites were analysed to investigate changes in the tryptophan pathway. This pathway was analysed as burn injuries have previously been found to cause dysregulation in metabolism and immune responses27, and as this pathway is responsible for the production, and synthesis of many neurotransmitters and regulators in the brain28. In the brain tissue samples, seven catabolites measured had significantly different concentrations between the burn and sham groups including 2-aminophenol, 3-hydroxyanthranilic acid, 5-hydroxyindole acetic acid, indole 3 acetic acid, kynurenine, nicotinic acid, and picolinic acid, while three had significantly different concentrations between excision and sham including 2-aminophenol, picolinic acid and tyramine (Fig. 6). No significant differences in concentrations were found between the burn and excision groups in the brain.

In the plasma, four catabolites had significantly different concentrations between the burn and sham groups including 3-hydroxyanthranilic acid, 5-hydroxy-tryptophan, dopamine, and quinolinic acid Dopamine was also significantly different between the excision and sham groups (Fig. 6). A full list of concentrations and corresponding p values can be found in Supplementary Table 12.

Fig. 6
Fig. 6The alternative text for this image may have been generated using AI.
Full size image

Dot plots of the catabolites with significant differences between intervention groups and their location on the tryptophan pathway. Dot plots with significantly different concentrations between groups in the brain (A-H), and in the plasma (I-L), where the black lines show the median in each group. A visualisation of the tryptophan pathway (Created in BioRender. Allahham, A. (2025) https://BioRender.com/qjumwwk) with the catabolites which had significantly different concentrations in the burn and excision groups relative to sham shows the relationship between these significantly different catabolites (M). (Brain: burn, n = 10 burn; excision, n = 9; sham, n = 10. Plasma: burn, n = 15 ; excision, n = 14; sham, n = 15). ****p < 0.0001, **p < 0.01, *p < 0.05.

HR-MAS NMR results

Brain regions were investigated using HRMAS NMR to investigate changes in small molecular weight metabolites and the metabolomic profile in the brain after injury. 15 cerebellum, 26 frontal cortex, 29 hippocampus and 10 piriform cortex samples were analysed through HR-MAS NMR and the spectral assignments completed (Supplementary Fig. 7). The PCA scores plot showed clustering of each brain region (Supplementary Fig. 8) and while no significant differences in individual metabolite concentrations were found between groups, most likely due to the small sample size, significant OPLS-DA models could be produced across intervention groups for the NMR spectral data for the prefrontal cortex and cerebellum (Fig. 7). In the prefrontal cortex a model between the burn and sham groups was achieved with an AUROC of 1.00 where a decrease in 4-aminobutyrate, N-acetylasparatate, choline, myo-inositol and glutamine were found in the burn group compared to sham, whilst lactate and acetate were increased. Within the cerebellum, comparing sham vs. burn produced a model with an AUROC of 0.77, with lactate and aspartate increased in the burn group while choline, phosphocholine, glycerophosphocholine, taurine and myo-inositol were reduced. In the cerebellum samples, a comparison of burn and excision groups generated an AUROC of 0.64, with glycine and N-acetylaspartate increased in the burn group and alanine, 4-aminobutyrate, glutamate, creatine, phosphocholine, glycerophosphocholine and taurine decreased compared to excision. Full univariate comparisons can be found in the supplementary materials (Supplementary Table 13).

Fig. 7
Fig. 7The alternative text for this image may have been generated using AI.
Full size image

OPLS-DA and spectra figures of the HR-MAS NMR of significant overall differences between intervention groups in the prefrontal cortex and the cerebellum. OPLS-DA scores plot of 1D 1H NMR profiles sham (blue) vs. burn (orange) in the prefrontal cortex (R2X = 0.22, CV-AUROC = 1.00) (A), 1H NMR backscaled coefficients plots from the OPLS-DA model showing significant metabolite peaks (B). OPLS-DA scores plot of 1D 1H NMR profiles sham (blue) vs. excision (red) in the prefrontal cortex (R2X = 0.22, CV-AUROC = 0.90) (C), 1H NMR backscaled coefficients plots from the OPLS-DA model showing significant metabolite peaks (D). Key metabolites in (B) and (D): (1) Lactate; (2) Alanine; (3) 4-aminobutyrate (GABA); (4) Acetate; (5) N-acetylaspartate; (6) Glutamine; (7) Aspartate; (8) Creatine; (9) Choline; (10) Myo-inositol. OPLS-DA scores plot of 1D 1H NMR profiles of sham (blue) vs. burn (orange) in the cerebellum (R2X = 0.10, CV-AUROC = 0.77) (E), 1H NMR backscaled coefficients plots from the OPLS-DA model showing significant metabolite peaks (F). OPLS-DA scores plot of 1D 1H NMR profiles of sham (blue) vs. excision (red) in the cerebellum (R2X = 0.14, CV-AUROC = 0.87) (G), 1H NMR backscaled coefficients plots from the OPLS-DA model showing significant metabolite peaks (H). OPLS-DA scores plot of 1D 1H NMR profiles of burn (orange) vs. excision (red) in the cerebellum (R2X = 0.13, CV-AUROC = 0.64) (I), 1H NMR backscaled coefficients plots from the OPLS-DA model showing significant metabolite peaks (J). Key metabolites in (F, H, and J): (1) Lactate; (2) Alanine; (3) 4-aminobutyrate (GABA); (4) Acetate; (5) N-acetylaspartate; (6) Glutamate; (7) Glutamine; (8) Aspartate; (9) Creatine; (10) Choline; 11. Phosphocholine; 12. Glycerophosphocholine; 13. Taurine; 14. Myo-inositol; 15. Glycine.

Discussion

This study examined the impact of non-severe burn trauma and non-burn trauma on the brain in mice after recovery (3-months post injury) through behavioural tests, RNA sequencing and metabolite analyses. Overall results indicate that non-severe burn trauma leads to significant, sustained behavioural, inflammatory, transcriptional, and lipidomic changes in several regions in the brain, whilst NBT caused some changes but to a lesser extent and in many cases not significant level. These changes did not appear to be sustained in the plasma samples, suggesting the observed changes in the brain are not solely the result of sustained systemic metabolic changes in the circulation. Whilst most behavioural tests showed changes likely associated with conditioning, an increase in sucrose consumption from baseline to the 3-month timepoint was only observed in the burn injured group. In humans, stress is associated with consumption of food high in sugar and fat29. Although the relationship between stress and food consumption is less clear with mice, there is evidence that links an increase in sucrose consumption in the SPT with chronic variable and moderate to severe stress30. This stress is chronic and is induced by repeated environmental stressors, including cage tilt, restraint, space reduction, forced swimming, and flashing light, applied for variable durations each day over several weeks30. However, stress from the burn is likely more severe compared to the chronic mild stress from the environment. This is inconsistent with studies with other types of stress such as chronic unpredictable mild stress, where mice are exposed to more natural stressors such as exposure to different lighting and temperature, water and food deprivation and changes in housing conditions which results in a decrease in sucrose consumption31. This indicates that the changes in sucrose consumption observed after NSBI more closely mirror chronic moderate to severe stress.

Both the hippocampus and the prefrontal cortex were found to have transcriptomic changes in the burn group compared to sham, while there were significant differences between the burn and excision group in the piriform cortex. The hippocampus and frontal cortex are regions associated with mental health conditions such as depression and anxiety32,33,34. Previous studies analysing the hippocampus post burn injury in mice found an elevation in inflammatory chemokines and cytokines, and increased astrocyte activation in the hippocampus, including in the dentate gyrus that is involved with storing memories in the acute phase 24 h after the burn35. Our results support these findings and suggest sustained transcriptomic changes that may be a result of an inflammatory milieu post burn or contribute to it.

Transcriptional changes in the brain after a burn also suggest potential functional impacts in the long-term. One of the pathways significantly changed post burn in the prefrontal cortex was associated with learning or memory. Burn patients often have compromised memory following their burns, with one study showing that paediatric burn patients years after their injuries recalled significantly fewer specific memories and more extended versus general memories compared to controls36. Another study on adults ~ 2 years after their burn also showed deficit in cognitive functions including working memory, delayed recall, attention, executive function and language37. In an investigation of prefrontal cortex function after lower extremity burns it was also found that this region of the brain has a higher activation after burns when compared to healthy controls38. As such, the transcriptomic changes observed after burn injuries may be contributing to changes in cognitive function. As no DEGs were found between the excision and sham groups, this suggests the severity of burn trauma is higher than that of the excision and therefore that burn injury may be more damaging than other trauma types.

The significant changes found in the concentrations of different lipid subclasses in the brain between the intervention groups may also have pathological consequences or be indicative of sustained physiological changes in the brain. The main change in the lipid subclasses were in LPIs which were significantly increased in the burn group. LPIs are involved with several cell functions including cell proliferation, migration and tumorigenesis39. Levels of LPIs in different organs can be altered with the activation of macrophages39, with previous studies supporting this interaction in the brain by showing that LPIs can suppress microglial phagocytosis which is suggested to be neuroprotective in otherwise pathological conditions40. PCs, also elevated in the burn group, have been shown to have therapeutic anti-inflammatory effects, whereby reducing inflammation through inhibiting TNF-α and IL-6 pro-inflammatory signalling41. PGs, which were also elevated in the burn group in the brain, have been shown to play an anti-inflammatory role when combined with cardiolipin, where they support mitochondrial metabolism and inhibits inflammation and mRNA expression of COX-2 (an inflammatory marker) by 358-fold42. Therefore, the burn-mediated increase in PG concentrations in the brain could indicate action to counter inflammatory modifications. Although generalised, these changes in lipid concentrations in the brain suggest long-term activation of neuroprotective mechanisms to counter these pathological changes.

Catabolites of the tryptophan pathway are known to be involved with regulating inflammation and have been associated with mental health conditions including depression and schizophrenia43. In our study, analysis of catabolites of the tryptophan pathway suggest that immune and anti-inflammatory mechanisms are activated in the long-term in the brain after burn and excision injury. 3-hydroxyanthranilic acid is often considered to be a neurotoxin due to its free radical–generating and N-methyl-D-aspartic acid (NMDA) agonist activities44. It has also been shown to inhibit mitochondrial respiration, promote oxidative damage and induce apoptosis45. The increase in 3-hydroxyanthranilic acid in the brain following burn injury is consistent with findings from another study on traumatic brain injury, which found that 3-hydroxyanthranilic acid was overexpressed in the white matter astrocytes surrounding the area of brain trauma, suggesting external burn trauma may lead to similar changes in the brain as direct brain injury45. Kynurenine is a key regulator in the immune system, in metabolism, and is involved with neuronal modulation as it regulates neuroplasticity through NMDA receptor signalling and glutamatergic neurotransmission46. It is also activated by immune stress responses where it helps inhibit some T cell functions, activate regulatory T cells, and inhibit natural killer cells, therefore maintaining homeostasis47. Given its involvement with these significant immune and neuronal pathways, dysregulation of kynurenine concentrations can have effects on multiple systems. Elevation of kynurenine is associated with several neuronal pathologies including Alzheimer’s disease, amyotrophic lateral sclerosis, Huntington’s disease, depression and schizophrenia48. In the plasma, quinolinic acid had a lower concentration in both the burn and excision groups, which can be evidence for the activation of anti-inflammatory mechanisms49, however, quinolinic acid has been shown to be elevated in paediatric burn patients at least 3 years after the burns27. This may be due to immune differences between humans and mice50. If the decrease in quinolinic acid is an indicator of anti-inflammatory mechanisms at play, this would suggest that recovery in mice after burn injuries is more effective than humans. The changes observed in the tryptophan pathway metabolites further supports the transcriptomic and lipidomic findings of sustained changes in the brain following burn injuries, strongly suggesting a long-term physiological change in the brain after burn trauma that may be linked to an increased risk of pathology.

Results from the multivariate analysis of the HR-MAS NMR data also strongly support sustained metabolite changes in the brain regions investigated, with significant differences in the profile of discrete regions in the brain, specifically the prefrontal cortex and cerebellum. Both showed significant differences between the burn and sham groups, with no significant changes observed in the hippocampus and piriform cortex. Disturbances in several of the same metabolites that were significantly changed in the prefrontal cortex and cerebellum have previously been associated with mental health conditions. Patients with major depression were found to have increased levels of lactate and choline in the anterior cingulate cortex, the lateral ventricles, and the amygdala51. Other studies found lower levels of N-acetylaspartate in the medial prefrontal cortex and anterior cingulate, and lower levels of phosphocholine and glycerophosphocholine in the anterior cingulate in patients with depression compared to controls52. Choline levels have also been reported to be elevated in the anterior cingulate cortex and the lateral ventricles in major depressive disorder, but evidence on this was not consistent in the literature53. Other neuropathologies were also associated with disturbances to some of these metabolites, where N-acetylaspartate was found to be downregulated in traumatic brain injury, resulting in disruptions in energy metabolism in the brain54. Reductions in levels of choline were also found in patients with Alzheimer’s disease, where the neuroinflammatory environment was shown to be contributing to the changes in metabolite levels and the progression of the disease55. In our study, lactate and acetate were increased in the burn group in the cerebellum, while choline, phosphocholine, glycerophosphocholine, taurine and myo-inositol were decreased. In the prefrontal cortex, 4-aminobutyrate, N-acetylaspartate, glutamine and myoinositol were decreased in the burn group. Such changes likely lead to downstream metabolic impacts and have been associated with major depression and an inflammatory environment in the brain56. This suggests a possible relationship between burn injuries and mental health where the overlap of their associated mechanisms may be increasing the risk of developing mental health conditions after NSBI.

The impact of burn injuries on the brain demonstrated by this study may provide insight into the increased risk of burn patients developing brain-related pathologies in the long-term, as previously reported. Cognitive deficit after burn injury is a known but underestimated consequence of burn injuries57, and burn patients often experience cognitive sequelae including memory deficit, delirium, and mental health conditions after the injury7,57. The brain regions chosen for this study including the hippocampus and prefrontal cortex were selected based on their relevance to mental health conditions such as depression and anxiety, which have also been associated with elevated inflammatory markers in the brain58,59,60. The transcriptomic, lipidomic and metabolic impact on these brain regions may therefore increase the risk of burn patients developing such conditions in the long-term through inflammatory associated pathways3. These findings are the first to provide comprehensive evidence for a sustained physiological change in the brain after recovery from a NSBI, and support further studies in patients, potentially using MRI or other non-invasive imaging to identify whether there is sustained physiological change.

Limitations for the behavioural tests experiment included mice having to be group caged to help with their social interaction, which also plays a physiological part in decreasing stress responses. However, this meant that it was not possible to record the individual consumption rate of individual mice in the SPT. As four FSTs were run at the same time for mice in the same cage, there was also a chance that these mice were in a distressed condition due to the movement and noise created to funnel them into the glass beakers, which would have added to the stress from the test itself. For the transcriptomic changes experiment, bulk sequencing was conducted, where a whole brain region had been analysed in a sample rather than single-cell analysis. This means that the data reflected changes in multiple cell types, therefore it was not possible to detect if the changes were driven by a particular type of cell or by changes in the cell profile. Nevertheless, bulk RNA analysis has its advantages with its ability to produce in-depth sequencing on a larger number of cells. There may have also been a risk of RNA degradation, as samples were placed in ice for ~ 1–2 h until enough were accumulated to be bulk analysed following dissection. With regards to the lipidomic experiment, the main obstacle in any current lipidomic study is the fact that most lipid species have not yet been characterized, meaning that we still do not know their biological function. This was an anticipated problem in our study, which is why the focus in our experiment was on general lipid subclasses rather than lipid species. The main limitation of the tryptophan experiment is that it only analysed a subset of metabolites from the pathway. This is both an advantage as the pathway is well characterised, and the changes known to be closely related to mental health and a disadvantage as it is possible that other metabolites in other pathways have been overlooked at this point. Due to the HR-MAS machine undergoing several rounds of cleaning, optimising, and calibrating for the duration of this project, batch effects may have been introduced due to differences in tissue storage time and instrumental drift. However, we attempted to control this by including samples from each group in any given batch, and through post-hoc assessment of batch-to-batch discrepancies.

Conclusion

In this study we have demonstrated sustained changes in the brain after recovery from NSBI (7–8% TBSA) in a mouse model of trauma. Transcriptomic, lipidomic, and inflammatory changes were evident three months after the burn injury. NBT such as excision injuries showed some significant changes at the same timepoint, but these appear to be more limited compared to the burn group. The effects described suggest a sustained physiological change in the brain caused by burn injury and may therefore underpin the observed increase in hospitalisation for mental health conditions observed in burn patients. However, future studies will be needed to explore in more detail the associations between an increased risk of developing neurological conditions after burn injuries, the mechanisms that lead to this development, and possible interventions and preventative measures for burn patients to reduce the risk of developing long-term brain related morbidities after burn trauma.