In this prospective study, we identified candidate prognostic biomarkers using proteomics and developed prognostic models at 6 months after injury in a longitudinal cohort of 42 patients with mTBI. First, proteins significantly associated with the 6-month outcome were selected among the proteins quantified by the highly quantitative multiple reaction monitoring-mass spectrometry (MRM-MS) assays (Fig. 1). Next, we identified sequential associations among clinical assessments affecting the 6-month outcome of mTBI (Fig. 2). Then, we identified candidate prognostic biomarker proteins that indirectly affected the 6-month outcome via neuropsychological symptoms (BDI-II or K-MoCA and FAB paths) (Fig. 3). Using the candidate biomarker proteins, the prognostic models for each phase that can predict the 6-month outcome of mTBI were developed (Table 2).
By analyzing the biological terms of these prognostic candidate biomarker proteins, we could determine implications for the pathophysiology of mTBI. The sequence from the events that occurred within 72 hours after injury to the prognostic outcomes at 6 months is summarized as follows (Fig. 4): (i) acute events caused inflammation during the first week, followed by depressive symptoms, which worsened the RPCQ and GOSE scores at 6 months; (ii) at 1 week, the secretion of proteins from axons and neurofilament to plasma was associated with a poor RPCQ score via impaired cognitive function tested by the K-MoCA, whereas the mitochondrial proteins related to inflammation were associated with a poor GOSE score via cognitive decline tested by the FAB; (iii) the effect of depression on the RPCQ score varied depending on the myelin sheath degradation at 1 month and alterations in the hormone levels and glucose homeostasis at 3 months; and (iv) the effect of cognitive decline (FAB score) on the GOSE score varied depending on neuronal damage at 1 month and alterations in lipoproteins and LPS-IFNγ proteins at 3 months.
At 1 week, the main biological term associated with the outcome was inflammation. Inflammation has an important role in the pathophysiology of secondary brain injury after TBI21. Since post-traumatic inflammatory response plays an important role in neurodegeneration, corticosteroids were used after TBI until the early 21th century, when the CRASH trial was published22. Although the CRASH trial showed that corticosteroids should not be used routinely after TBI, this does not negate the role of inflammation in TBI. The neuroinflammation following mTBI has been hypothesized to be associated with post-concussive symptoms23,24. An interesting study reported that initially elevated C-reactive protein levels might be an independent predictor of persistent PCS in patients with mTBI25. Similar results that inflammation may influence the outcome, mediating depressive symptoms and cognitive function, were found in this study. However, further investigation of the role of inflammation in the sequelae following mTBI is required.
Axon and neurofilament proteins in the plasma at 1 week showed a correlation with RPCQ at 6 months via cognitive decline at 3 months in this study. The subsequent axonal damage is one of the most common pathological mechanisms of long-term dysfunction in mTBI21. A few studies have reported that cognitive dysfunction can correlate with the extent of microstructural white matter damage21,23. Thus, axon-related proteins, such as neurofilament light (NFL), spectrin N-terminal fragment (SNTF), and A-tau, have been extensively studied as biomarkers of TBI26. In this study, candidate biomarkers discovered at 1 week were neurofilament heavy and light polypeptides, which is consistent with other studies. Activated leukocyte cell adhesion molecule (ALCAM) was also discovered at 1 week. ALCAM is known to play a role in cell adhesion, axon growth, axonal pathfinding, and neuronal migration and differentiation27. There is a paucity of studies on ALCAM as a TBI biomarker; however, it may have potential as a candidate biomarker in TBI. Mitochondrion was also identified in the subacute phase (1 week) as a biological term that can influence the outcome of mTBI via cognitive decline. Mitochondrial dysfunction has been reported in animal models and humans and is considered to be an acute response after TBI28,29. The primary function of mitochondria is the production of adenosine triphosphate (ATP) and reactive oxygen species30. Thus, mitochondria dysfunction can exacerbate the cellular energy crisis and aggravate cell damage28,31. This energy crisis can impact post-concussive symptoms and increase the vulnerability to second impact syndrome in mTBI patients31,32.
At 1 month, the effect of depression on the 6-month RPCQ score varied depending on the degradation of the myelin sheath. Myelin surrounds and protects axons in the central nervous system (CNS). After TBI, myelin damage can result from axon injury, neuronal cell death, or secondary damage that causes oligodendrocyte loss with subsequent demyelination of intact axons33. Although the role of myelin injury in the pathophysiology of mTBI remains poorly understood, it is known that myelin damage contributes to white matter injury along with axonal injury34,35; thus, myelin can be associated with cognitive function36. However, the present study differs in that myelin showed an association with depressive symptoms. Although the association between depressive symptoms and decreased white matter integrity in multiple brain areas after mTBI was reported in a previous study37, it needs further investigation.
Moreover, microglial activation terms, such as neuronal damage, were associated with 6-month GOSE via cognitive decline at 1 month. Microglia are the immune cells of the CNS that have roles in brain inflammation through their interaction with white matter38,39. Microglial activation has recently been of increasing interest because they have been recognized as a key regulator of degeneration and regeneration of the white matter39. As white matter injury is known to be a major contributor to the impairments associated with mTBI, microglial activation has been studied as a biomarker and a potential target of therapeutics in mTBI39,40. Although the results of the present study might be supportive of this, there is still much to be revealed.
At 3 months, alterations in lipoproteins and LPS-IFNγ were associated with the 6-month outcome via cognitive decline. Among lipoproteins, lipoprotein receptor protein 1 (LRP1) was a candidate biomarker identified in this study. LRP1, a regulator of blood-brain barrier (BBB) integrity, may have roles in dementia progression41,42. TBI-induced BBB dysfunction has been reported in previous studies42. However, there is a paucity of studies examining the role of LRP1 in TBI. Meanwhile, IFN-γ may play a significant role in the pathophysiology of TBI as a neuroinflammatory mediator43. A study reported that IFN-γ level in mTBI patient group was significantly increased compared with that in the community control group at 1 year after injury, suggesting an activation of the innate immune system up to 1 year post-injury44. Although our study also showed similar results, further research is needed.
In addition, glucose homeostasis showed a correlation with RPCQ at 6 months, mediating depressive symptoms at 3 months. In this study, the candidate biomarker protein related to glucose homeostasis was IGFBP5, which can exert both stimulatory and inhibitory effects on IGF-1 signaling45. Although IGF-1 exhibits both neuroprotective and neuroregenerative effects after TBI46, further studies are needed on the potential of IGFBP5 as a mTBI biomarker.
Because of the heterogeneity of mTBI, it is difficult to predict the prognosis of the patients. Until recently, several studies have attempted to predict the prognosis using multiple factors7,47,48. These models were mainly based on clinical factors, such as admission characteristics, post-injury symptoms, and psychological symptoms. However, in an external validation study using the Collaborative European NeuroTrauma Effectiveness Research in Traumatic Brain Injury study data, none of these predictive models had both good calibration and discrimination in patients with mTBI7.
Blood-based biomarkers have been proposed as novel predictors of mTBI prognosis. Commonly studied mTBI prognostic biomarkers, such as GFAP, S100B, neuron-specific enolase, and UCH-L1, have shown poor discriminating ability18. Ghrelin, NFL, SNTF, and A-tau have been studied as potential biomarkers for the prognosis of mTBI26,49–51. For example, a recent study that was conducted to predict recovery after sports-related concussion in professional flat-track jockeys using blood biomarkers, such as GFAP, NFL, and tau, suggested that cognitive testing and blood biomarkers may be potential objective measures to assist in the monitoring of concussion recovery52. However, the relation of ghrelin, NFL, SNTF, and A-tau is limited to axonal injury, which is one of the most common pathological mechanisms of long-term dysfunction in mTBI21. The pathophysiology of mTBI is highly complex; therefore, it is necessary to discover various candidate biomarkers that can comprehensively cover the pathophysiology.
Unlike previous studies of mTBI biomarkers that used a hypothesis-driven approach, we utilized proteomics, a data-driven approach, to discover novel prognostic candidate biomarkers. Although the hypothesis-driven approach is less likely to yield false-positive results, it is limited by current understanding and the time it requires18. In data-driven methods such as proteomics, a large number of candidate biomarkers can be discovered quickly because candidate biomarkers can be screened without regard to the pathophysiology, although it is more likely to yield false-positive results53. To reduce false-positive results of proteomics, serial mediation and moderation analyses were performed in this study. Although cross-sectional approaches to mediation are possible to generate biased estimates of longitudinal parameters in the special case of complete mediation54, serial mediation analysis is powerful and widely used to find mediators between variables. We investigated a causal chain linking the mediators between clinical symptoms and time progression and then selected the proteins that mediated these causal linking chains and influenced the 6-month outcome. Thus, the selected proteins might be more reliable as mTBI prognostic biomarkers.
This study identified candidate mTBI prognostic biomarkers at multiple time points from the early to the chronic phase after injury, whereas previous studies have mostly identified biomarkers only at a single time point. A biomarker that is available in the acute phase of injury is not appropriate in the chronic phase. For example, the level of GFAP in peripheral blood reaches a peak at 20 hours after trauma but is detected at a very low concentration after 72 hours17. Biomarkers such as GFAP can be used for diagnosis or prediction of the outcome in the acute phase of injury, but it is difficult to use them after 3 days of injury. Because patients with mTBI may not visit the hospital in the acute stage of the injury alone, the candidate prognostic biomarkers at various time points identified in this study are expected to be clinically useful.
The present study used both the RPCQ and GOSE as outcomes for developing prognostic models of mTBI. Although there is an overlap between both measures, there is also a difference in that RPCQ reflects patients’ discomfort, while GOSE reflects the overall function. Therefore, in previous studies, post-concussive symptoms measured by RPCQ or GOSE at 6 months after injury have been used as the outcome of mTBI7. However, both measures have limitations as outcome variables. Although RPCQ is employed most often in measuring post-concussive symptoms55, there is no guidance for assessing PCS using RPCQ. GOSE is widely employed as a primary outcome measure in TBI studies, but its utility as an outcome measure in patients with mTBI is controversial because GOSE is not sensitive enough to find different health problems despite good functioning7. Therefore, to complement these limitations, we used both outcomes of mTBI for developing prognostic models in this study.
This study has some limitations. Measurements at some time points were missed in a few participants because of the time of their enrollment. In particular, we could not find the independent variables in the logistic regression due to a lower number of blood samples within 72 hours after trauma and could find only single variables. To enhance patient care in the emergency department (ED), biomarkers that can be applied within 72 hours should be identified in further studies. In addition, there may be a selection bias toward more severe patients because this study was conducted at a tertiary hospital. A future multicenter study would ensure the representativeness of the sample group and a larger sample size.
In conclusion, we identified prognostic candidate biomarkers using proteomics in light of clinical course after mTBI. Although our results should be externally validated for generalization, we suggest that these protein biomarkers established in the consideration of the course after mTBI are expected to have a wide clinical application.