Neurochemical Markers of Traumatic Brain Injury: Relevance to Acute Diagnostics, Disease Monitoring, and Neuropsychiatric Outcome Prediction

Considerable advancements have been made in the quantification of biofluid-based biomarkers for traumatic brain injury (TBI), which provide a clinically accessible window to investigate disease mechanisms and progression. Methods with improved analytical sensitivity compared with standard immunoassays are increasingly used, and blood tests are being used in the diagnosis, monitoring, and outcome prediction of TBI. Most work to date has focused on acute TBI diagnostics, while the literature on biomarkers for long-term sequelae is relatively scarce. In this review, we give an update on the latest developments in biofluid-based biomarker research in TBI and discuss how acute and prolonged biomarker changes can be used to detect and quantify brain injury and predict clinical outcome and neuropsychiatric sequelae.

https://doi.org/10.1016/j.biopsych.2021. 10.010 Traumatic brain injury (TBI) is a major cause of death and disability worldwide (1). Clinically, TBI is classified as mild, moderate, or severe based on loss of consciousness, posttraumatic amnesia, and structural damage visualized on head computed tomography (CT) or magnetic resonance imaging (MRI). Most TBI cases are concussive or mild. The current clinical imaging techniques, although useful for diagnosis of moderate to severe TBI, are not sensitive enough to detect subtle brain injury. Biofluid-based biomarkers may be complementary and/or alternative methods to detect and quantify the amount of injury to different cell types and structures of the brain, as well as tissue reactions to and recovery processes after a TBI. Recently, the United States Food and Drug Administration cleared blood GFAP (glial fibrillary acidic protein) and UCH-L1 (ubiquitin C-terminal hydrolase L1) for prediction of the absence of intracranial injuries on head CT (2)(3)(4). In addition, the Scandinavian Neurotrauma Committee has proposed serum S100B (S100 calcium-binding protein B) for detection of intracranial findings following head trauma (5,6). These blood-based biomarkers have been shown to be useful diagnostic tools and may reduce the use of CT scans in the emergency department setting (3,7).
The outcome following TBI is highly variable (8,9). The traditional view is that most individuals who sustain a concussive or mild TBI recover within days to weeks (10). In about 15% of individuals, postconcussive symptoms persist for more than a year, which is referred to as postconcussive syndrome (PCS) (11,12). The symptoms of PCS generally can be categorized into four domains: physical, cognitive, emotional, and sleep. PCS, albeit an outdated term because of lack of granularity, still gives an indication of the long-term impact of a mild TBI. Biofluid-based biomarkers that correlate with or predict physical, cognitive, emotional, and sleep outcomes after TBI would be very useful, especially in the clinical setting.
In this review, we give an update on the latest developments in biofluid-based biomarker research in TBI and discuss how acute and prolonged biomarker changes can be used to detect and quantify brain injury and predict clinical outcome.

DISEASE MECHANISMS IN TBI OF RELEVANCE TO FLUID BIOMARKERS
TBI is characterized by multifaceted postinjury acute and chronic processes that may contribute to recovery and neurodegeneration. Acute TBI results in axonal injury with release of cytoskeletal proteins and disrupted axonal transport of proteins such as amyloid-b, phosphorylated tau (P-tau), TDP-43 (TAR DNA-binding protein 43), and a-synuclein that may build up in the brain tissue and contribute to neurodegenerative processes, along with neuroinflammatory responses, including microglial and astrocytic activation, as well as injury to oligodendrocytes and cellular and structural components of the neurovascular unit.

FLUID BIOMARKERS FOR AXONAL INJURY
One of the most well-studied axonal protein is tau, a microtubule-associated protein predominantly expressed in short cortical unmyelinated axons (Table 1) (13). Increased concentrations of cerebrospinal fluid (CSF) total tau (T-tau) have been found in acute samples from patients with moderate to severe TBI, where the initial levels correlated with 1-year functional outcome (14). In a study of Olympic boxers who underwent repeated lumbar punctures, CSF T-tau increased 7-10 days after a bout and normalized after 3 months of rest (15). With the advances in immunoassay technology, T-tau could also be quantified in blood with high analytical sensitivity (16). Plasma T-tau was measured in blood samples from professional ice hockey players within hours after a sports-related concussion (SRC), and the levels were increased compared with the preseason baseline (17,18). In the context of chronic SRC, a recent study found no difference in plasma T-tau in National Football League players with a history of repetitive head trauma compared with control subjects (19).
Another axonal protein that has garnered a lot of attention is neurofilament light (NfL). NfL is a component of the axonal cytoskeleton and is primarily expressed in large-caliber myelinated axons (Table 1) (20). CSF NfL is a sensitive biomarker of neuroaxonal damage (21) and has been validated in several neurodegenerative disorders (22)(23)(24)(25)(26)(27)(28). In 2016, the ultrasensitive assay for quantification of NfL in serum or plasma was first tested in a TBI context (29). In patients with acute moderate to severe TBI, serum NfL could distinguish these patients from healthy control subjects with area under the receiver operating characteristic curve (AUROC) of 0.98-1.0 (29). Furthermore, serum NfL showed moderate to strong correlations with both CSF and ventricular CSF from the same individuals (29). In subsequent studies, serum NfL measured within 48 hours of injury could also distinguish patients with CT findings from those with normal CT with high accuracy (2,30,31). In a recent study conducted in clinic-based patients with a history of mild, moderate, or severe TBI who received follow-up serial blood sampling from 30 days up to 5 years, serum NfL could distinguish patients with mild, moderate, or severe TBI from each other as well as control subjects (32). Serum NfL concentration at 1 year correlated with global brain volumes measured at the same time point as well as with diffusion tensor imaging measures of white matter integrity (32). NfL has also been measured in professional ice hockey players with acute concussion, where higher levels in serum were seen in players with prolonged return to play (18). Furthermore, serum NfL performed better than plasma T-tau in distinguishing concussed athletes from control subjects (18). In the context of subacute and chronic repetitive head impact, NfL measured 7-10 days after a bout was elevated compared with control subjects, and the levels decreased after 3 months of rest (33). Furthermore, serum NfL correlated with the corresponding CSF values (r = 0.86) as well as with number of hits received to the head (33). In a recent study of professional athletes with a history of repetitive SRC who underwent lumbar puncture and blood assessment months after the most recent SRC, serum NfL correlated with CSF (r = 0.76), and serum NfL could distinguish concussed athletes from control subjects with high accuracy (32).
T-tau has been shown to be more sensitive to body trauma, while serum NfL is not affected by body trauma or strenuous exercise (18).
With regard to the temporal profile of T-tau and NfL quantified in blood, T-tau seems to be an acute biomarker, especially in concussion, while NfL in serum peaks 7-10 days after a head trauma and may be detectable months to years after injury (18,32,34) (Table 1).

FLUID BIOMARKERS FOR ASTROCYTIC ACTIVATION
S100B is a protein primarily expressed by astrocytes and was the first biomarker to be proposed for clinical use by the Scandinavian Neurotrauma Committee (5,6). In a metaanalysis of 2466 patients with TBI, S100B had a pooled sensitivity, specificity, and negative predictive value of 97%, 40%, and 99%, respectively, in predicting CT findings (35). These findings have been confirmed in additional studies (6,36). Furthermore, S100B led to a 32% reduction in unnecessary head CTs compared with either the Canadian CT Head Rule or the New Orleans Criteria. In another study, a secondary elevation of S100B following TBI was shown to significantly predict secondary pathological CT/MRI findings (mainly ischemic-like lesions) with high sensitivity and specificity in patients with mild to severe TBI (37). This was superior to common clinical features (pupil response, Glasgow Coma Scale, admission CT findings, intracranial pressure, and hemoglobin levels) used to predict secondary pathological findings (37).
In the context of SRC, serum S100B increased 1 hour after a concussion compared with preseason control results (18). However, when compared with NfL and tau, S100B had lower diagnostic and prognostic value (18). A major limitation of S100B is that it significantly increased after strenuous exercise (18). Similarly, other studies have reported exercise-related elevation in S100B (38)(39)(40). The increase in S100B observed in these studies may be because S100B expression is found in adipocytes, skeletal tissue, and various other organs (41)(42)(43)(44).
Another biomarker of astrocyte reactivity is GFAP, which is an intermediate filament protein (Table 1) (45). Serum GFAP has been shown to distinguish patients with TBI with intracranial findings on head CT from those without with high accuracy (2)(3)(4). Recently, GFAP and UCH-L1 (a protein abundantly found in neurons) were cleared by the United States Food and Drug Administration for detection of intracranial injury on head CT following TBI (4). In the context of acute TBI, serum GFAP and UCH-L could distinguish patients with intracranial lesions on CT from those without with high accuracy. In the largest study of GFAP and UCH-L1 to date, including 1959 patients with mild to moderate TBI, serum GFAP and UCH-L measured within 12 hours of injury had high sensitivity and negative predictive value for the detection of traumatic intracranial injury on head CT (3). Several studies have assessed the combination of GFAP and UCH-L1 for predicting CT findings after acute TBI. The combination of GFAP and UCH-L1 performed better than either biomarker alone in predicting intracranial injuries on head CT after TBI (3,46,47). Recent studies have found that GFAP alone may perform similarly as the GFAP and UCH-L1 combination for predicting CT findings after mild TBI (4,48,49). These studies also found that GFAP or the GFAP and UCH-L1 combination outperformed S100B for predicting CT findings (47,48,50). Several studies have also compared GFAP and UCH-L1 with NfL and T-tau in predicting intracranial pathology on head CT or brain structural MRI. In one of these studies, serum GFAP performed similarly or slightly better than NfL, while T-tau performed worse in detecting MRI findings after a mild TBI, and UCH-L1 had variable levels (2). In another study, serum GFAP, UCH-L1, and NfL had almost similar performance, while T-tau performed worse in detecting CT pathology associated with TBI (51).
In the context of SRC, increased levels of GFAP measured within 48 hours after a concussion have been seen in collegiate athletes (52,53). In a recent study, concussed collegiate athletes had increased concentrations of serum GFAP, NfL, and UCH-L1 measured within 24-48 hours after a concussion compared with preseason baseline, with the highest concentrations in concussed athletes with loss of consciousness or posttraumatic amnesia (54). The levels of GFAP and NfL remained elevated for several days in these types of concussions. In another recent study, GFAP, NfL, and UCH-L1 increased in U.S. cadets who sustained a concussion, as well as in cadets who participated in the same combative training exercise but did not incur a concussion (55). These recent studies provide support for the potential utility of blood biomarkers, especially GFAP and NfL, for SRC or military concussion.
Serum S100B is an acute biomarker that peaks within hours after injury ( Table 1). The utility of S100B beyond the acute phase and in relation to TBI severity is yet to be examined in detail. Serum GFAP has been shown to increase acutely after injury (Table 1). Recently, we found that GFAP is detectable in serum following mild, moderate, or severe TBI even months to years after head trauma (56). A drawback of GFAP as a biomarker of intracranial injury on head CT after TBI is that it seems to perform worse in older patients (57) ( Table 1).

FLUID BIOMARKERS FOR INJURY TO OLIGODENDROCYTES
MBP (myelin basic protein) is a marker of oligodendrocytes (Table 1), which is detectable in blood and indicates potential disruption in myelin (58)(59)(60). In animals exposed to various degrees of blast TBI, MBP was elevated in serum (58). Elevated serum MBP was also seen in patients with severe TBI (58-60). The marker has not been examined in mild TBI.

FLUID BIOMARKERS FOR MICROGLIAL ACTIVATION
Microglia are found throughout the central nervous system, where their main function is to clear damaged cells and synapses, as well as infectious agents (61). After TBI, microglia can clear cell debris and orchestrate neurorestorative processes that are beneficial to neurological recovery. Microglia can also produce proinflammatory and cytotoxic mediators that hinder central nervous system repair and further contribute to neuronal dysfunction and cell death. The shift between these two opposite functions is not well understood. TREM2 (triggering receptor expressed on myeloid cells 2) is a receptor mainly expressed on the surface of the microglia (62). Recently, TREM2 has been found to play a role in Alzheimer's Neurochemical Markers of Traumatic Brain Injury Biological Psychiatry March 1, 2022; 91:405-412 www.sobp.org/journal disease (AD). CSF-soluble TREM2 has been found to be increased in patients with AD than in control subjects (63). The availability of the CSF assay for CSF-soluble TREM2 also opens a window of opportunity for assessing the potential role of microglia in human TBI.

FLUID BIOMARKERS FOR DISRUPTION OF THE NEUROVASCULAR UNIT/BLOOD-BRAIN BARRIER
TBI causes disruption of the blood-brain barrier (BBB) integrity (64). Clinically, the ratio of CSF albumin to serum albumin is commonly used as a surrogate marker of BBB integrity (65). In the context of TBI, the CSF:serum albumin ratio was measured in a study of professional athletes with a history of repetitive head trauma (66), where the levels of the CSF:serum albumin ratio was unaltered. A plausible explanation could be that the CSF:serum albumin ratio may not be a sensitive enough measure of BBB integrity or that BBB integrity may be disrupted in the acute phase of the injury but not in the chronic phase as this study was performed.
Another biomarker of BBB leakage is sPDGFRb (soluble platelet-derived growth factor receptor), a protein highly expressed in pericytes of the vasculature (67,68). Increased CSF sPDGFRb has been reported in patients with AD compared with control subjects, where the levels of CSF sPDGFRb correlated with CSF:serum albumin ratio (69). In the context of TBI, sPDGFRb is yet to be examined.

FLUID BIOMARKERS FOR TBI-RELATED PROTEINOPATHIES
TBI may also cause tangle pathology, which consists predominantly of P-tau (70) ( Table 1). Recently, P-tau (using antibody that specifically recognizes phosphothreonine-231) and T-tau were measured in plasma samples from 217 patients with TBI, where P-tau and ratio of P-tau to T-tau demonstrated perfect discrimination of mild TBI from control subjects (AUROC of 1.0) (71). The ratio of P-tau to T-tau also showed strong ability to predict positive CT findings (AUROCs 0.921 and 0.923, respectively) (71). In another study, P-tau and GFAP together performed significantly better for predicting CT findings than either biomarker individually (AUROC 0.96) (57). A recent meta-analysis found that the most promising biomarkers for predicting CT findings in TBI were GFAP in combination with UCH-L1, although P-tau was comparable, and S100B was significantly lower (AUROC 0.98, 0.92, 0.72, respectively) (72). In the context of SRC, P-tau was measured in CSF of 16 professional athletes with a history of repetitive concussion and 15 healthy control subjects, and there was no significant difference in the levels of CSF P-tau between the groups (73), suggesting that the marker may not detect longterm P-tau changes, although more studies are needed.
Experimental and postmortem studies suggest that athletes who have had repetitive head trauma may develop brain amyloid deposition (seen in 43% of cases) (74)(75)(76). The amyloid deposition or plaques seen in TBI are predominantly composed of 42 amino acid long and aggregation-prone amyloid-b (Ab 42 ) (Table 1), which are also seen in AD (77,78). In a study, Ab 40 and Ab 42 were measured in CSF from professional athletes with a history of repetitive concussions, and both CSF Ab 40 and Ab 42 were decreased, with the highest effect size seen for Ab 42 (73), suggestive of potential brain amyloid pathology. Altered Ab 42 has also been observed in CSF and plasma of patients with acute severe TBI. In a study, a decreased CSF Ab 42 concentration was seen in 12 patients with severe TBI compared with 20 control subjects when measured acutely (79). In another study, Ab 42 was measured in plasma collected at 24 hours, 30 days, and 90 days after TBI from 34 patients with TBI and 69 healthy volunteers, where the levels of Ab 42 were significantly increased at all measured time points (80).
In addition to the classic pathologies of tangles and amyloid plaques observed in some individuals with TBI, especially those with chronic traumatic encephalopathy, TBI is also associated with TDP-43 inclusions and less commonly with asynuclein inclusions (81). Currently, there are no reliable fluid assays to quantify TDP-43 or a-synuclein inclusions in individuals with TBI. With advances in the detection of misfolded seeds of a-synuclein in biofluids using real-time quakinginduced conversion or protein misfolding cyclic amplification (similar technologies to qualitatively detect trace amounts of diffusible misfolded a-synuclein, through its ability to induce aggregation of added recombinant a-synuclein in CSF over time, using thioflavin T fluorescence), brain a-synuclein pathology can be reliably detected in lumbar CSF from patients with Parkinson's disease and other synucleinopathies (82). While so-called real-time-induced conversion has been used to quantify TDP-43 in CSF of patients with amyotrophic lateral sclerosis and frontotemporal dementia (83), this technique is yet to be used for quantification of TDP-43 or a-synuclein inclusions in individuals with TBI.

NOVEL CANDIDATE FLUID BIOMARKERS
A recent TBI biomarker avenue of research has been quantifying central nervous system-derived proteins contained in extracellular vesicles (EVs). There are several potential advantages to quantifying proteins in EVs: 1) EVs protect their content from degradation by endogenous proteases that are common in blood (84,85), 2) EVs can easily cross the BBB (86), and 3) EVs are found to be biologically more active than proteins found within circulating blood (87). In a study of veterans with a history of remote trauma, elevated EV NfL was seen in those with a history of multiple mild TBIs and elevated chronic neurobehavioral symptoms (88). Similarly, significantly increased EV tau and EV P-tau were found in veterans with a history of multiple mild TBIs compared with control subjects (87). In a recent study of civilians with a history of TBI, EV NfL and EV GFAP measured at 1 year after injury were elevated in patients with moderate to severe TBIs compared with control subjects, with EV GFAP performing better than EV NfL in distinguishing patients with moderate to severe TBIs from control subjects (89).
EVs may also contain microRNAs (miRNAs) released from injured neurons (90). miRNAs are found throughout the body and are essential to neuronal injury and repair (91,92). Similar to the other established proteins measured in EVs, miRNAs can transverse the BBB easily owing to their small size, which makes them attractive as potential biomarkers of TBI. Furthermore, miRNAs have been implicated in both the primary (90) and secondary (93) damage responses to TBI.
Several studies have investigated the role of miRNAs as biomarkers for TBI. A study compared EV RNA in the CSF of 11 patients with severe TBI and 17 control subjects and found that most of the RNA packaged in CSF microparticles was noncoding RNA and that two of these noncoding RNAs (miR-9 and miR-451) were differentially expressed in patients with severe TBI (94).

WHICH OF THESE BIOMARKERS PREDICT NEUROPSYCHIATRIC SEQUELAE?
As mentioned earlier, TBI (even a mild one) may cause longterm neuropsychiatric symptoms, including cognitive and emotional symptoms and sleep disturbances (11,12). In civilian patients hospitalized for an orthopedic injury, presence of comorbid mild TBI was associated with an increased risk of posttraumatic stress disorder (PTSD) and depression 3-6 months after injury (95). In another study of 91 patients with TBI and 27 patients with multiple traumas but without evidence of brain damage, major depressive disorder was significantly more frequent among patients with TBI than among the control subjects during the first year after sustaining a TBI (96). In military TBI, there is an increased risk of posttraumatic stress and depressive symptoms that may worsen over time (97). In the context of sports-related TBI, several studies suggest that symptoms of depression, anxiety, and emotional lability are higher in concussed athletes, especially those with a history of repetitive head trauma (98,99). Emerging studies indicate that chronic symptoms of PTSD, depression, and neurobehavioral symptoms following mild TBI are associated with increased concentrations of neuronal injury proteins in peripheral blood. For example, increased PTSD symptoms in service members have been associated with increased plasma tau (100). In another study, it was found that PTSD, depression, and neurobehavioral symptoms following TBI were associated with increased tau and NfL but not Ab 40 or Ab 42 (101). In a recent study of veterans and service members with a remote history of repetitive mild TBI, higher concentrations of serum and EV NfL correlated with increased neurobehavior, PTSD, and depressive symptoms (88). In another study, we found that increased concentrations of serum NfL correlated with poor sleep and lower executive function scores after a remote mild TBI (102). In the context of SRC, we found that higher concentrations of NfL correlated with Rivermead Post-Concussion Symptoms Questionnaire scores both in athletes who have had an acute concussion (18) and in those who developed chronic PCS (32). These studies, despite their caveats, suggest that axonal injury as measured by serum T-tau or NfL may underlie the severity of neuropsychiatric symptoms such as depression and neurobehavioral and PTSD-related symptoms.
Neuropsychiatric symptoms such as anxiety and depression go hand in hand with functional outcome after TBI. In a recent study, depressive and anxiety symptoms correlated strongly with function and disability measures in daily life (103). Currently, there are several candidate biomarkers that have shown promising prognostic utility (72). In a recent study, NfL and GFAP measured within 24 hours predicted an unfavorable outcome (AUROC 0.75 and 0.82, respectively) (31). In a study of professional Swedish ice hockey players, low serum NfL predicted a more favorable functional outcome and a lower risk of PCS (18,33). In the same cohort, the initial level of plasma Ttau predicted return to play but had a lower predictive value than NfL (18,33), while S100B showed no associations with return to play. S100B has shown mixed results in predicting outcome in severe TBI. In a recent study, serum NfL measured at an average of 1 year after injury correlated with functional outcome assessed at the same time, while no relationship with functional outcome was seen for GFAP, T-tau, or UCH-L1 (56). In one study of severe TBI, serum S100B measured within 2 weeks of injury could discriminate patients who would have an unfavorable outcome (defined as severe disability, vegetative state, or death based on the Glasgow Outcome Scale) from those with a favorable outcome (moderate disability, mild disability, or no disability based on the Glasgow Outcome Scale) at 12 months (104). In the same study, S100B was compared with UCH-L1, GFAP, and NfL; NfL and GFAP added the most independent information in predicting functional outcome, while S100B was the least useful (104). In another study, S100B was not associated with outcome at 12 months, while serum NfL measured within 24 hours after injury was associated with outcome at 12 months (29).

SYNTHESIS AND CONCLUSION: WHAT ADDITIONAL RESEARCH IS NEEDED?
Several of the existing large-scale biofluid-based biomarker studies have been focused on distinguishing patients with TBI from control subjects. In the emergency department setting, GFAP, UCH-L1, and S100B have been shown to be useful in distinguishing patients with trauma-related cranial CT findings from those without. Several recent studies have found that a panel of biomarkers may outperform individual biomarkers, especially with regard to diagnostic or predictive value (4,46). For example, a combination of GFAP and UCH-L1 performs better than individual values in predicting the presence of intracranial pathology (4,46). There is a scarcity of literature assessing these biomarkers in the subacute or chronic phase of TBI and longitudinally, which is an important topic for future research.
NfL has been shown to be an excellent biomarker for assessing axonal injury after TBI of varying severity and over months to years after TBI (32). However, serum NfL reflects one aspect of the TBI pathophysiology, and there is a need for assessing other pathologies such as tangles, amyloid deposition, astrogliosis, microglial activation, and BBB disruption. Therefore, we may need a panel of biomarkers for TBI. Blood assays for several of these pathologies are under development or refinement; however, there are few studies that have assessed these assays in TBI.
Finally, recent studies have assessed the relationship between GFAP, T-tau, NfL, and Ab 42 , where higher levels of Ttau, GFAP, and NfL correlate with elevated neuropsychiatric symptoms after TBI (100,101) or worse functional outcome (18,29,31,33,56,104). Although these recent studies show promise for utility of these biomarkers for further understanding of the impact of neuropsychiatric symptoms, larger longitudinal studies are needed to address whether initial levels of these biomarkers would predict neuropsychiatric outcomes.