Noninvasive Urinary Biomarkers for Early Diagnosis of Mild Traumatic Brain Injury

Abstract

The lack of an easily measured biomarker hampers mild traumatic brain injury (mTBI) diagnosis and limits the ability to monitor an objective feature in patients through recovery. Due to its ease of collection and mirroring of metabolism, urine is a useful source for biomarkers. Urine samples collected from mice at 1, 12, 24, 48, and 96 h following mild lateral fluid percussion injury were subjected to headspace gas chromatography coupled with mass spectrometry (GC–MS). Of the 68 urinary volatiles detected, three were both stable in the sham controls and significantly increased in the injured group. The potential biomarkers identified in the mouse model included a saturated ketone and two unsaturated ketones. Results with mice prompted a targeted search for ketones in concussed youth (aged 7–19 years). Leveraging bio-banked human urine samples, headspace GC–MS analyses were conducted with human urine collected 4.5–71 h following injury from 18 patients with mTBI and eight healthy controls. Of the three α,β-unsaturated ketones identified in human urine, one was found to be significantly increased following mTBI. Taken together, elevation of urinary ketones in mice and humans suggests acute perturbation of fatty acid metabolism immediately following injury. Thus, ketones could form the basis for an easily accessible human biomarker of mTBI and greatly improve diagnosis and monitoring of these injuries and potentially direct efforts for treatment of mTBI.

Introduction

Mild traumatic brain injury (mTBI), or concussion, is exceedingly common (particularly in adolescents and young adults), but the contemporary state of diagnosis remains primarily clinical. The current practice of relying on subjective reporting of symptoms presents significant challenges in terms of diagnosis, prognosis, and improving outcomes. ¹ In addition, broad TBI classification schemes, such as the ordinal Glasgow Coma Scale score (where mTBI is defined as a score of 13–15), provide limited utility in identifying the state of neurofunction and need for targeted treatment in injured patients. Recently, an updated classification scheme was proposed via an NIH-NINDS initiative that comprises four pillars: clinical, biomarkers, imaging, and modifiers that together characterize TBI from a multidimensional perspective. ² This CBI-M approach recognizes that there is no single parameter that can describe concussive neuro-dysfunction, but rather requires complementing the clinical assessment with imaging and biomarker metrics, as well as incorporating modifying parameters with known influence on clinical outcomes.^3,4

A recent review of TBI biomarker discovery demonstrates the wide variety of targets detectable in blood. ⁵ These targets include components of the proteome, lipidome, metabolome, and exosome that provide evidence of damage at the cellular and tissue level. Several large-scale studies have identified promising biomarkers and demonstrated their diagnostic accuracy and recovery outcome prediction by exploiting both protein^6–8 and metabolomic^9,10 biomarkers. These discoveries led to FDA approval for ubiquitin C-terminal hydrolase L1, an enzyme specific to neurons, and glial fibrillary acidic protein, a protein found in astroglia and released after injury, to assist in the decision to obtain neuroimaging in adults following TBI.^3,7,11–13 However, the invasive nature of blood collection and the highly technical nature of sample analyses (e.g., nuclear magnetic resonance, high-resolution tandem mass spectrometry [MS], chemiluminescent microparticle immunoassay, single molecule array immunoassay) render such tools impractical for rapid testing, particularly in nonclinical settings.

In contrast to blood, urine requires no specialized training to collect nor complex processing while awaiting analysis. In previous work, proteomic analysis of urine identified several biomarkers that were downregulated in concussed college athletes (samples collected <7 days of injury) compared with healthy controls. ¹⁴ Due to its role in metabolism and elimination, urine is an excellent source for small metabolite (i.e., metabolomic) studies. In one such study, the urinary metabolome of male youth athletes (collected 24–72 h from injury) was analyzed via ¹H-NMR spectroscopy, identifying 2-hydroxybutyrate and lactose as promising metabolites for the diagnosis of sport-related mTBI. ¹⁵ Another adult human study of urinary metabolites identified down-regulation of certain amines in response to brain injury. ¹⁶

We previously measured the volatile metabolome in urine collected from mice 4 to 28 days after receiving mild lateral fluid percussion injury (mLFPI), a well-characterized model of mTBI that mirrors human clinical pathology. Volatile metabolomic approaches have previously been employed for human studies of autoimmune diseases (Kang et al. 2015), ¹⁷ cancers,^18–20 tuberculosis ,^21,22 gastrointestinal disorders (Probert et al. 2009), ²³ and diabetes. ²⁴ Chemometric evaluations of these conditions can reveal the presence of novel volatile metabolites in patient samples,^25,26 or a pattern of changes in previously established metabolites.^27,28

In our previous work, trained biosensor mice could discriminate between urine from injured and sham mice based on volatile urinary metabolites. ²⁹ Chemical analyses further identified that five specific urinary volatiles (exo-brevicomin, α-farnesene, o-toluidine, dimethyl sulfone, and formanilide) were altered by injury. In a more recent study, 10 differentially expressed metabolites were observed in urine collected from rats receiving irradiation injury compared with sham controls. ³⁰ Among these were saturated ketones 3-methyl-2-hexanone and 2-hexanone.

Extending our previous evaluation (and those of others) to an early acute time period (<48 h) is critical to establishing a viable urine-based biomarker(s) of mTBI. To achieve this goal, we first conducted an animal model study using mLFPI focused on early time points post injury. Headspace GC–MS was used to identify several early biomarkers of injury. In an exploratory manner, we then extended these findings to a pilot study of humans by analyzing urine collected from adolescent patients with mTBI and comparing them to that of uninjured controls. The unique and valuable combination of both preclinical and clinical approaches provides the opportunity to advance the new CBI-M nomenclature for TBI diagnosis.

Methods—Mouse Models

Mouse lateral fluid percussion brain injury

Seven- to ten-week-old, male C57BL/6J mice (The Jackson Laboratory, Bar Harbor, ME, USA) were housed in groups of two to five in the animal facility with standard day–night cycle (lights off: 18:00 h–6:00 h; lights on: 6:00 h–18:00 h) and free access to food and water. One hundred mice were randomly assigned to one of the following experimental groups: TBI or Sham. Experimenters were blinded to group assignments. All procedures were conducted in accordance with the guidelines and regulations of the Children’s Hospital of Philadelphia (CHOP) Institutional Animal Care and Use Committee (#IAC 25-000694). Surgery and injury were performed as previously described.^31,32 Sham animals received all the following, except the fluid pressure pulse (see below). Briefly, C57BL/6 mice were anesthetized with a combination of ketamine (2.6 mg/kg) and xylazine (0.16 mg/kg), and without breaching the dura, a 3-mm craniectomy was conducted over the right parietal area between bregma and lambda in the anterior–posterior direction and the sagittal suture and lateral cranial ridge in the medial-lateral direction. A Luer-loc needle hub was secured to the skull over the opening with cyanoacrylate and dental acrylic, the hub was capped, the scalp was sutured closed, and the animal was allowed to recover.

The following day, the animal was anesthetized with isoflurane (2% oxygen at 500 mL/min). The needle hub was filled with isotonic sterile saline and connected to the LFPI device via saline-filled tubing. The device consisted of a large fluid-filled cylinder with a piston and a pendulum that generated the fluid pressure pulse when it struck the piston. The animal was placed on its left side and the pendulum was raised to a predetermined height and released to deliver a pressure pulse of 1.4–1.6 atm to produce both focal and diffuse injury. Previous studies employing this pressure range demonstrated little to no cell death, cavitation, or tissue loss, supporting the mild nature of the injury.^32–35 The fluid line was disconnected, and the mouse righting time was recorded.^32,33,35 Righting time has previously been shown to correlate with injury severity, ³⁶ and righting times of 250–350 sec are considered to represent mild injury. ³² The animal was reanesthetized with isoflurane for scalp closure.

Mouse urine collection

Care was taken to stratify surgeries among replicate mice to ensure that treatment (mLFPI or sham) and urine collection time (1, 12, 24, 48, and 96 h) were evenly distributed over the course of sample collection dates. Urine was collected directly from the bladder of the mice at predetermined times ranging from 1 to 96 h post-injury (Table 1). Mice were maintained under isoflurane anesthesia, and a vertical incision was made on the lower abdomen (taking care not to damage the intestines). The bladder (normally found just above the penis) was identified by its translucent mass, slightly darker than the surrounding tissue. Curved forceps were used to slightly lift the bladder out of the body cavity, being careful not to put pressure on the bladder itself. As the bladder rested on the forceps, it was carefully punctured with a sterile 1-mL TB syringe (27 G × ½) to draw the urine into the syringe for delivery to sterile glass vials. Following urine collection, mice were euthanized by decerebration. Urine samples were stored at −80°C prior to being transferred to the Monell Chemical Senses Center Laboratories for headspace analyses (where they were stored at −40°C).

Table 1.

Number of Mice at Each Postinjury Time Point for Which Sufficient Sample Volumes Enabled Headspace Analysis (Total n = 91)

Treatment	1 h	12 h	24 h	48 h	96 h
Sham surgery	9	9 (1)	10	9 (1)	10
mLFPI	9 (2)	9 (1)	8	9 (1)	9

Six outliers (in parentheses) were not included in statistical analyses (final total n = 85).

mLFPI, mild lateral fluid percussion injury.

Mouse urine headspace

Fifty microliters of urine was placed in a 20-mL headspace vial and fortified with 10 µL of an internal standard consisting of 70 µg/mL l-carvone in water (such that 700 ng was delivered to each sample). Multiple quality control (QC) samples (to monitor chromatographic system suitability), consisting of empty vials or vials containing only l-carvone solution were distributed throughout each chromatographic run. Quality control samples were also used to assist in the exclusion of chromatographic peaks not related to urine. Samples were analyzed using an HT3 dynamic headspace analyzer (Teledyne Tekmar) with a Supelco Trap K Vocarb 3000 thermal desorb trap (Sigma-Aldrich) attached to a Trace Ultra gas chromatograph (Thermo Scientific) equipped with a single quadrupole mass spectrometer (ISQ, Thermo Scientific). Sample vials were incubated at 40°C and swept with helium for 10 min at 75 mL/min as volatile metabolites were collected on the thermal trap. Following collection, the trap was heated to 260°C and volatiles desorbed directly onto the gas chromatograph equipped with 30 m × 0.25 µm Stabilwax-DA (Restek) capillary column. Split injections (5:1) were made with a column flow of 2.0 mL/min and a split vent flow of 10.0 mL/min. The starting GC oven temperature was 40°C (3 min) and ramped to 260°C at a rate of 7°C/min. The mass spectrometer was operated in scan mode from 33 to 400 m/z following a 5-min solvent delay necessitated by the presence of ethanol co-solvent in the internal standard solution. Tentative chromatographic peak identifications were based on the NIST Standard Reference Database. ³⁷

Mouse urine chemometric analyses

Chromatographic data were exported to MetAlign® software for peak alignment and noise elimination. ³⁸ Resulting data were processed by MSClust for mass spectral extraction and assignment of a single peak response value based on the extracted selected ion monitoring trace. ³⁹ All chromatographic peak responses were normalized to the l-carvone response from each sample. Chromatographic peaks that were present in urine samples and not present in quality control samples were considered to be of urine origin. Carvone-normalized data were subjected to principal component analysis (PCA) using Unscrambler® (CAMO Software; Oslo, Norway) to visually identify outliers exhibiting undue influence or leverage in residual plots.

To assess which volatiles remained at constant concentrations over 96 h in sham mice, slopes of carvone-normalized peak responses from mice in the sham treatment were examined by analysis of covariance (ANCOVA) using proc GLM in SAS. ⁴⁰ A decision criterion (α = 0.1) was chosen to determine which volatiles did not increase or decrease in sham mice over the 96-h study period. Normalized peak responses of those volatiles remaining constant in sham animals for 96 h were subjected to repeated measures analysis of variance (RM-ANOVA) using the repeated statement with proc GLM in SAS. Treatment (LFPI or sham control), time (1, 12, 24, 48, and 96 h), and the interaction were fixed effects. A false discovery rate controlling procedure was employed to account for multiple tests. ⁴¹ Residuals were plotted to examine their distribution and heteroscedasticity.

Methods—Human Participants

Human urine collection

Urine samples employed in this study were scavenged from two unrelated studies. Only those samples for which further use was permitted by consent or assent were transferred to the Monell Chemical Senses Center for analysis. In one of these studies, samples from concussed youth (age 7–19) were obtained upon the patient’s initial presentation to the CHOP Emergency Department as part of a study from July 2021 to December 2023. Inclusion criteria included 5–23-year-olds presenting for concussion care within 72 h of injury. Exclusion criteria were a prior mTBI within 30 days of enrollment, current ocular trauma, lower extremity injury precluding normal gait, or underlying disorders suggesting visio-vestibular dysfunction. Among those samples, 18 were collected within 3 days of injury (Table 2), which matched the timeframe of the mouse study and aligned with the typical timeframe of presentation to acute care, primary care, and specialists for pediatric mTBI.^42,43

Table 2.

Demographics, Sampling, and Clinical Characteristics of the Exploratory Human Mild Traumatic Brain Injury Cohort

Characteristic	Healthy controls (n = 8)	Injured group (n = 18)
Mean (median, IQR) time of urine collection after injury (hours)	NA	16.5 (8.6, 19.2)
Male	4 (50%)	11 (61%)
Female	4 (50%)	7 (39%)
Age (years) (median, IQR)	12.8 (13.5, 3.8)	12.9 (12.5, 4.8)
Sports and recreation-related injury	NA	9 (50%)
Nonsports-related injury	NA	9 (50%)

Samples from uninjured controls were scavenged from a separate study of patients presenting with musculoskeletal complaints (without bony fracture or need for immobilization) to CHOP Sports Medicine clinics; or presenting to the CHOP Minds Matter Concussion Specialty Care Program for concussion baseline testing from March 2021 to May 2024. Eight urine samples were selected from this larger population such that an equal number of males and females were represented, and the age distribution of the controls was similar to those of the injured participants (Table 2). Exclusion criteria for controls included a prior mTBI within 28 days of enrollment or any current ongoing symptoms from a previous concussion.

In both studies, urine samples (>25 mL) were collected from participants using the “clean catch” method and stored on dry ice for a few hours, then stored at −80°C until transfer for analysis. In addition to samples, the following data were abstracted from injured participants’ medical record: age, race, ethnicity, sex, time since injury for sample collection, clinical symptoms, and mechanism of injury. Any imaging and related abnormalities were also abstracted. Age and sex were self-reported by control participants. Both studies were approved by the CHOP Institutional Review Board (17-013875 and 20-018393).

Human urine headspace

Headspace collection of urinary volatiles was achieved with the TriPlus RSH autosampler (Thermo Scientific, Waltham, MA) for solid-phase microextraction (SPME) with the novel “Arrow” 1.10-mm DVB/C-WR/PDMS fiber. One milliliter of urine was transferred to a 20-mL glass headspace vial and fortified with 10 µL of an aqueous internal standard solution of acetophenone-phenyl-d5 (3 µg/mL). The vial was sealed immediately with a magnetic crimp cap and incubated for 10 min at 37°C at 500 rpm prior to 10 min SPME extraction at 1,000 rpm. The fiber was inserted into the GC injection port and thermally desorbed in splitless mode at 230°C for 2 min. The ISQ single-quadrupole gas chromatograph–mass spectrometer (Thermo Scientific, Waltham, MA) was equipped with a 30 m × 0.25 mm id Stabiliwax®-DA fused-silica capillary column (Restek, Bellefonte, PA). The oven temperature was held for 2 min at 40°C and then programmed at 5.0°C/min to 230°C with a 2 min hold at this final temperature. Helium carrier gas was used at a constant flow of 1.1 mL/min. The mass spectrometer was operated with a 3.3-Hz scan rate over a scan range of m/z 33–400 and an ion source temperature of 260°C.

Human urine chemometric analyses

Based on the results of the mouse experiment, a targeted approach for identifying ketones was performed by monitoring specific mass spectrometric ions (m/z = 57 and m/z = 69). Suspected ketone identities were confirmed by analyses of known standards. Ketone responses were normalized using the internal standard added to each sample.

The upper bounds of expected values in urine samples from healthy males and females were calculated separately for each identified ketone using means and standard deviations of responses observed in healthy subjects. Each upper bound was defined as the mean normalized response plus three times the standard deviation (3 SD), which is a common definition for an upper control limit. Difference scores were determined by subtracting the upper bound (defined for each ketone and subject sex) from each observation. Nonparametric tests of difference scores were conducted in SAS using the NPAR1WAY procedure (specifying the Wilcoxon Rank Sum Exact Test) to determine if difference scores varied between control and injured groups. The number of observations exceeding an upper bound was also counted. The probability that the actual number of observations exceeding an upper bound was greater than the number expected from the spread of a normal distribution was determined using the FREQ procedure in SAS and specifying the Exact Binomial Test. By definition, the spread defined by 3SD of a normal distribution encompasses 99.87% of the observations.

Results

Mouse model experiment

Fifty mice received mLFPI, and 50 mice received sham surgery. Righting times for mice receiving mLFPI ranged from 250 to 350 sec, demonstrating the mild nature of the injury. Sufficient urine volume for reliable analysis was obtained from 91 of the 100 mice (Table 1). Sixty-eight volatile metabolites were attributed to the mouse urine samples. Examination of all normalized responses obtained from 91 samples revealed that five of them (three injured, two sham) exhibited large residual variance and/or made a disproportionate contribution to the PCA results and were excluded as outliers. In addition, examination of RM-ANOVA model residuals indicated one outlier (mLFPI mouse at 1 h) more than six standard deviations from the mean and was excluded. Thus, data from 85 mice representing collections at 1, 12, 24, 48, and 96 h post-injury or sham surgery formed the analytical data set (40 sham and 45 injured; Table 1).

ANCOVA revealed eight volatile metabolites with constant concentrations over 96 h (i.e., slopes statistically equal to zero) in sham mice (Table 3). Among these eight, RM-ANOVA yielded a significant between-subjects treatment × time effect (F_4,75 = 4.03; p < 0.0051); the main effects (treatment and time) were not significant. For within-subjects effects, all effects were significant. Importantly, the volatile × treatment × time effect (F_28,525 = 2.16; p = 0.0006) indicated that at least one volatile differed by treatment at one or more time points. Univariate analyses revealed significant treatment × time effects for three volatile metabolites (Table 3): 6-methyl-3-heptanone (F_4,75 = 4.22; p < 0.0039), 3-methyl-4-hexen-2-one (F_4,75 = 3.63; p < 0.0092) and 4-hepten-2-one (F_4,75 = 3.76; p < 0.0076). For each of these ketones, a substantial concentration increase was observed in urine of injured mice at 1 h post-injury (Fig. 1).

Table 3.

Eight Volatile Metabolites with Constant Mouse Urine Concentrations (i.e., Slopes Statistically Equal to Zero) in Sham Controls over 96 h

Urine metabolite	ANOVA
6-Methyl-3-heptanone	p = 0.0039
3-Methyl-4-hexen-2-one	p = 0.0092
4-Hepten-2-one	p = 0.0076
Dimethyl trisulfide	p = 0.1736
Nonanal	p = 0.2385
Unknown	p = 0.4533
Unknown	p = 0.4992
Dimethylsulfone	p = 0.3035

Among these, analyses of variance (ANOVA) revealed significant time × treatment interactions for three (bold).

FIG. 1.

Potential biomarkers found to increase in mouse urine 1 h post injury using the LFPI model of mTBI. Concentrations are semiquantitative estimates based on normalization to l-carvone internal standard responses. Error bars are the standard error of the mean. mTBI, mild traumatic brain injury.

Human urine evaluation

Urine samples from 18 concussed youth and eight healthy controls were analyzed. The targeted approach to detecting ketones revealed three aliphatic ketones in human urine samples: 3-octen-2-one, 3-nonen-2-one, and 3-decen-2-one. Chromatographic comparison to known ketone standards revealed that all three ketones were present as trans isomers. Wilcoxon signed-rank tests of difference scores indicated that urine concentrations of 3-octen-2-one (Χ²_1,25 = 6.250; p = 0.0124) differed from concentrations of the healthy patients (Fig. 2). However, 3-nonen-2-one (Χ²_1,25 = 3.781; p = 0.0518) and 3-decen-2-one (Χ²_1,25 = 1.235; p = 0.2665) urine concentrations of injured patients did not differ from the concentrations of the healthy patients.

FIG. 2.

Human urine ketone concentrations (semiquantitative estimates normalized to the acetophenone-d5 internal standard response) in samples collected from concussed patients 4.5–71 h post-injury. The dotted lines represent the upper bounds of ketones measured in healthy males (red) and females (blue).

In the absence of external factors (e.g., head injury), only 13 of every 10,000 urine ketone observations (0.13%) should be observed above the upper bound (as defined by a spread of 3SD). Of 18 observations made from injured patients, ten 3-octen-2-one values exceeded the upper bound (Fig. 2). The probability of observing 10 of 18 concentrations in excess of the 3SD upper bound is p = 1.1 × 10⁻¹⁵.

Discussion

The early postinjury period following brain injury is extraordinarily dynamic. Brain metabolism, for instance, shifts from hypermetabolic to hypometabolic, and neuroinflammation may progress from restorative and beneficial to prolonged and deleterious. ⁴⁴ Specifically, soon after injury, microglia contribute to the clearing of debris ⁴⁵ and activate proinflammatory regulatory proteins (e.g., IL-6, TNF-α) for neuron recovery. Later, prolonged activation of microglia can lead to continuous production of proinflammatory molecules, contributing to further neuronal dysfunction.^46,47

In the search for objective biomarkers of mTBI, understanding the time course of presentation of specific biomarkers is critical for clinical application. Our previous work demonstrated that patterns of several mouse urinary volatiles could be used to discriminate mLFPI-injured mice from sham mice over the time period from 2 to 14 days post-injury. ²⁹ However, several of them are pertinent only to rodent biology. The current study was conducted specifically to identify potentially important biomarkers at earlier time points post-injury that may be more relevant for translation to human studies.

Given the diagnostic goal, metabolites that naturally exhibit significant concentration variability over time would not be reliable biomarkers. Thus, we focused on those metabolites that remained at constant concentrations at each time point of the study in sham mice. Among the eight volatile metabolites that remained constant, concentrations of three (3-methyl-4-hexen-2-one, 4-hepten-2-one, and 6-methyl-3-heptanone) were elevated at 1 h post-injury (relative to sham controls) in those mice subjected to mLFPI-induced brain injury (Fig. 1). A possible link between neuronal damage and fatty acid metabolism is strongly suggested by results of the mouse experiment where other factors that may affect urinary ketone concentrations, such as anesthesia, surgery, diet, and age, were highly controlled.

In the human cohort, three ketones (3-octen-2-one, 3-nonen-2-one, and 3-decen-2-one) were observed in the urine samples. Significant evidence indicates that the urinary concentration of 3-octen-2-one is elevated in response to head injury (Fig. 2). It is important to note that increased levels of urinary ketones were observed at 1 h post-injury in mice (Fig. 1). However, no human samples were collected prior to 4 h post-injury from females or 3 h from males. Thus, responses of urinary ketones to head injury in humans may be incomplete in this pilot study.

The ketones observed in human urine are α,β-unsaturated ketones (i.e., ketone group on carbon #2 and conjugation on the adjacent carbon #3), which are formed from oxidized polyunsaturated fatty acids. ⁴⁸ Oxidation of fatty acids occurs in the liver to produce ketone bodies (3-hydroxy butyrate and acetoacetate), which provide energy for thermogenesis and are utilized by extrahepatic organs, particularly the brain. ⁴⁹ Systemic responses are known to be triggered in response to brain inflammation. ⁵⁰ In a mouse model study, rapid changes were observed in the liver within 1 h of brain injury. ⁵¹ Specifically, plasma levels of polyunsaturated fatty acids increased within an hour of injury and returned to preinjury levels after 24 h.

Fatty acid metabolism is not limited to the liver. Circulating lipids and other metabolites are taken up in parts of the brain. ⁵² Specifically, saturated and unsaturated fatty acids are differentially processed in the hypothalamus. Alterations in the lateral hypothalamus have been demonstrated from studies examining sleep impairments following mTBI. Orexin neurons, which reside in the lateral hypothalamus and underlie wakefulness, are altered by mLFPI. Orexin neurons exhibit decreased activity during the dark phase of the light cycle, thus leading to an inability to maintain wakefulness in mLFPI animals.^35,53

Volatile ketones can also be produced through an exchange of metabolites between a host and the host’s supporting bacteria. ⁵⁴ Many α,β-unsaturated ketones are known to be produced by microbial species. ⁵⁵ Further, significant changes in the gut microbiome have been observed within hours of TBI in animal studies.^56–58 Regardless of mechanism, the rapid increase of these particular ketones in humans following brain injury appears to align with known physiological processes that occur with brain injury and emphasizes the importance of further investigating the role of α,β-unsaturated ketones in human urine in the early hours following mTBI.

Employment of opportunistic sampling (acquisition of biological samples collected for clinical or other purposes) is a particular strength of this study. Opportunistic sampling is becoming increasingly common in pharmacokinetic studies and is endorsed by the Pediatric Trials Network. ⁵⁹ Access to these urine samples scavenged from two unrelated studies allowed us to extend our findings from mice to human injury in an exploratory manner. Studies evaluating human metabolic alterations in the first few hours post-injury are uncommon, in large part due to the difficulty of systematically obtaining access to concussion patients close to the time of injury. Animal models, on the other hand, allow for rigid study design and regularly scheduled sample collection. Despite acceptance of mLFPI as a suitable animal model for the study of brain injuries, some suggest that physiological differences between the human and murine brain can generate distinctly different responses to injury. ⁶⁰ Likewise, diet (intake of polyunsaturated fats in particular) varies among humans and is difficult to scale between mice and humans. The mouse chow provided to the animal models in this study was slightly higher in total fat (9%) as compared with some other common lab diets (4.1–7.2%), but not so high as to be considered a “high fat diet” (>20% fat). Currently, there are no guidelines for standardizing diets between species. ⁶¹

As pertaining to biomarker discovery, the social role of urine scent also varies greatly between mice and humans. In mice, certain ketones are important semiochemicals participating in social communication. For example, 6-hydroxy-6-methyl-3-heptanone accelerates puberty in female mice ⁶² and 4-hepten-2-one is an adrenal-mediated urinary metabolite that suppresses female estrous. ⁶³ Mice also rely on urinary proteins for social communication. Major urinary proteins (or MUPs) are highly polymorphic arbiters of genetic information advertised by males. ⁶⁴ Because these proteins differentially bind odorant molecules, ⁶⁵ observed alterations of urinary volatiles may represent injury-induced changes to the MUP profile.

Additional limitations to this study are related to sample size, subject sex, and timing of postinjury observations. In the animal model, sample collection started at 1 h, with the next sample at 12 h. TBI is incredibly dynamic, and future work should more closely examine the time period between 0.5 and 12 h as a relevant time period for translation to humans. In addition, only male mice were employed in the design (while youth of both sexes were included in human cohort). The exploratory human portion suffers from a small sample size that may not adequately define the upper bound of urinary ketone concentrations in a healthy population, and the important time period immediately following injury (0–3 h post-injury) was not represented among the samples.

Regardless, targeted analysis of ketones in the human cohort uncovered preliminary evidence that certain α,β-unsaturated ketones may serve as early biomarkers of mTBI. Future work should focus on additional human data collection that more precisely accounts for potential sources of systemic or acute metabolic changes unrelated to TBI and explores circulating ketone levels. Linkage of both animal and human experiments increases the importance of this study and should be employed more often to facilitate and accelerate preclinical to clinical translation. This requires partnerships across disciplines at the outset of both human and animal studies rather than the more typical serial attempts at translation where silos can serve as barriers to preclinical impact on human health.

Identification of biomarkers sensitive to different phases of the injury response could have profound implications for studying and diagnosing brain injury in humans. Noninvasive body fluids, such as urine, constitute a rich source of potential biomarkers that can be collected quickly and easily at multiple time points pre- and post-injury. Monitoring for injury-induced alteration of fatty acid metabolism (3-octen-2-one in particular) could provide rapid and early detection of brain injury in nonclinical settings (e.g., playing field, training room, workplace). Such a test could perhaps be as straightforward as a chemical test strip (similar to pH) or, owing to the volatile nature of the ketones, an electronic nose (e-Nose) device.

Authors’ Contributions

A.C.: Conceptualization, methodology, investigation, and writing—original draft; K.B.A.: Conceptualization, writing—original draft, project administration, and funding acquisition; Y.E.L.: Investigation; B.J.: Writing—original draft; O.E.P.: Resources and data curation; C.L.M.: Conceptualization and writing—review and editing; G.K.B.: Conceptualization and writing—review and editing; B.A.K.: Methodology, formal analysis, and writing—original draft.

Abbreviations Used

ANCOVA

Analysis of Covariance

CBI-M

Clinical, Biomarkers, Imaging, and Modifiers (TBI classification framework)

CHOP

Children’s Hospital of Philadelphia

IQR

Interquartile Range

mLFPI

Mild Lateral Fluid Percussion Injury

mTBI

Mild Traumatic Brain Injury

MUPs

Major Urinary Proteins

NPAR1WAY

Nonparametric One-Way Analysis Procedure (SAS)

PCA

Principal Component Analysis

QC

Quality Control

RM-ANOVA

Repeated Measures Analysis of Variance

SPME

Solid-Phase Microextraction

TBI

Traumatic Brain Injury

Cite this article as: Cohen A, Arbogast KB, Lee YE, et al. Noninvasive urinary biomarkers for early diagnosis of mild traumatic brain injury, Neurotrauma Reports, 2026, 7, 2689288X261464654, doi: 10.1177/2689288X261464654.