Post-Traumatic Hypoxia Is Associated with Prolonged Cerebral Cytokine Production, Higher Serum Biomarker Levels, and Poor Outcome in Patients with Severe Traumatic Brain Injury

Abstract

Secondary hypoxia is a known contributor to adverse outcomes in patients with traumatic brain injury (TBI). Based on the evidence that hypoxia and TBI in isolation induce neuroinflammation, we investigated whether TBI combined with hypoxia enhances cerebral cytokine production. We also explored whether increased concentrations of injury biomarkers discriminate between hypoxic (Hx) and normoxic (Nx) patients, correlate to worse outcome, and depend on blood–brain barrier (BBB) dysfunction. Forty-two TBI patients with Glasgow Coma Scale ≤8 were recruited. Cerebrospinal fluid (CSF) and serum were collected over 6 days. Patients were divided into Hx (n=22) and Nx (n=20) groups. Eight cytokines were measured in the CSF; albumin, S100, myelin basic protein (MBP) and neuronal specific enolase (NSE) were quantified in serum. CSF/serum albumin quotient was calculated for BBB function. Glasgow Outcome Scale Extended (GOSE) was assessed at 6 months post-TBI. Production of granulocye macrophage-colony stimulating factor (GM-CSF) was higher, and profiles of GM-CSF, interferon (IFN)-γ and, to a lesser extent, tumor necrosis factor (TNF), were prolonged in the CSF of Hx but not Nx patients at 4–5 days post-TBI. Interleukin (IL)-2, IL-4, IL-6, and IL-10 increased similarly in both Hx and Nx groups. S100, MBP, and NSE were significantly higher in Hx patients with unfavorable outcome. Among these three biomarkers, S100 showed the strongest correlations to GOSE after TBI-Hx. Elevated CSF/serum albumin quotients lasted for 5 days post-TBI and displayed similar profiles in Hx and Nx patients. We demonstrate for the first time that post-TBI hypoxia is associated with prolonged neuroinflammation, amplified extravasation of biomarkers, and poor outcome. S100 and MBP could be implemented to track the occurrence of post-TBI hypoxia, and prompt adequate treatment.

Introduction

Traumatic brain injury (TBI) is a poorly recognized epidemic worldwide.^1,2 Despite progress in neuroimaging and critical care, mortality in severe TBI remains at ∼30%.³ TBI is classified as “focal” by lesion formation, localized edema, and development of pericontusional ischemia, whereas diffuse TBI produces widespread axonal and vascular pathology in the white matter, resulting in global edema.⁴ Hypoxia (Hx) and hypotension are frequent secondary insults aggravating brain damage, and both have been correlated with worse outcome.^5–7 Cerebral Hx has been reported in up to 44% TBI patients with underlying brainstem injury, hypotension, and impaired brain perfusion, as well as extracranial trauma, including airway obstruction, chest injury, and peripheral hemorrhage.^{8, 9}

Following TBI, numerous molecular pathways become activated, prompting the release of their neurotoxic end products. The ensuing process leading to secondary brain damage includes the production of pro- and anti-inflammatory cytokines, oxidating radicals, and excitotoxins; blood–brain barrier (BBB) disruption; cell swelling; and neuronal cell death.^10–15 Cytokines comprise a large group of peptides regulating the inflammatory response.^10,14 Their synthesis is elicited not only in blood leukocytes but also in brain resident glial and neural cells. It is generally believed that pro-inflammatory cytokines exacerbate brain damage after injury, whereas anti-inflammatory cytokines promote brain repair.¹⁴ The main cytokines studied in the context of TBI include the pro-inflammatory interleukin (IL)-1, -2, -6, and -8; granulocye macrophage-colony stimulating factor (GM-CSF); interferon (IFN)γ; tumor necrosis factor (TNF); and the anti-inflammatory cytokines IL-4 and IL-10.

Several groups have undertaken clinical research exploring the role of cytokines in the pathophysiology of TBI, including cytokines' potential diagnostic and prognostic value. These studies report conflicting data on associations between cytokines and clinical parameters, and, to our knowledge, there is no consensus on those that may serve as biomarkers of brain injury. In a recently published review article, we summarized the controversy around the validity of cytokines as correlative biomarkers to initial severity of TBI, raised intracranial pressure (ICP), BBB dysfunction, mortality, and outcome scores on the Glasgow Outcome Scale (GOS)/GOS Extended (GOSE) in a multitude of clinical studies in which cytokines were measured in serum, plasma, cerebrospinal fluid (CSF), or microdialysates.¹⁶

A major hurdle in the management of TBI patients is the lack of reliable surrogate markers that could assist in the diagnosis and prognosis. Biomarkers are intrinsic proteins of the CNS, promptly released into the blood upon injury, which have been successfully correlated with Glasgow Coma Scale (GCS), brain morphological changes on CT scans, and outcome (GOS/GOSE).¹⁷ Some of the most characterized biomarkers include myelin basic protein (MBP), neuronal specific enolase (NSE), and S100.¹⁸ Based on their cellular localization, these proteins are believed to reflect distinct structural damage. In the CNS, 30% of the myelin is composed of MBP. Release of MBP into CSF has been attributed to myelin destruction and white matter damage, and is, therefore, considered as a marker of demyelinating activity and axonal pathology. Studies in pediatric TBI showed that peak levels of MBP as well as S100 and NSE correlate with the GOS.¹⁹ Together with NSE, MBP seems to distinguish uninflicted from inflicted TBI in children, the latter often being associated with hypoxemia and worse outcome.^20,21 NSE is found in the cytoplasm of neurons, and reflects neuronal damage. NSE concentrations were associated with severity of injury, CT scan findings, and outcome.^22,23 S100 is a low affinity Ca⁺⁺ binding protein expressed by astrocytes, and is indicative of glial damage. It remains the most studied biomarker in TBI. Elevation of S100 in blood has been shown to differentiate mild from severe TBI, and to correlate with brain injury severity (GCS scores), size of brain damage on CT scans, and 6 month outcome scores (GOS/GOSE).^18,24,25 In more recent years, these and other biomarkers are being increasingly employed in clinical trials as molecular correlates to outcome scores.^26,27

Although previous TBI studies examined changes of cytokine profiles and the clinical validity of brain injury biomarkers, the impact of secondary Hx on these effects was not thoroughly investigated. Given the ability of cerebral Hx per se to induce immune activation,^28,29 we speculated that TBI combined with a hypoxic insult enhances cerebral cytokine production.

We also assumed that blood levels of MBP, NSE, and S100 increased proportionally with more severe injury arising from post-traumatic Hx, possibly aggravating changes in CSF/serum albumin ratio and brain protein extravasation. Ultimately, we expected biomarker concentrations to discriminate the occurrence of Hx and correlate with worse outcome scores. This hypothesis relies on our compelling evidence in TBI rats showing that post-traumatic Hx enhanced expression of brain cytokines,³⁰ heightened macrophage infiltration, and aggravated axonal pathology and neurological impairment.^30,31

Methods

Patient population and sample collection

Following approval by the Alfred Hospital Human Ethics Committee, 42 patients were recruited at the Alfred Hospital, Melbourne. Delayed informed consent was obtained from the next of kin. Inclusion criteria were: severe TBI with a post-resuscitation GCS ≤8 (except for three patients with initial GCS>9 who deteriorated at the scene requiring intubation) and the need for an extraventricular drain (EVD) for ICP monitoring and therapeutic drainage of CSF (see Table 1 for demographics). Computer tomogram (CT) scans were performed within 4 h from TBI, followed by surgical evacuation of mass lesions if required and implantation of an EVD. Patients were managed as per our institution's protocol, which included: 1) sedation with morphine, midazolam, and propofol; 2) maintenance of cerebral perfusion pressure >60 mm Hg and ICP <20 mm Hg; 3) maintenance of normothermia; and 4) anti-seizure prophylaxis with phenytoin for 7 days. ICP >20 mm Hg was managed using the following step-up protocol: head elevation (up to 30 degrees), CSF drainage when ICP >20 mm Hg, osmotherapy with mannitol or hypertonic saline, paralysis using cisatracurium, and, if necessary, barbiturate coma using thiopentone infusion to achieve burst suppression. CSF was collected over 24 h and kept at 4°C beginning from admission (day 0) up to day 5 after injury. Parallel blood samples were obtained daily. Exclusion criteria comprised pregnancy, neurodegenerative diseases, HIV and other chronic infection/inflammatory diseases, or history of TBI.

Table 1.

Demographic Data of TBI Patients

Variables	Values
Age, years, median (range)	29 (16–63)
Gender, n (%)
Males	32 (76.2)
Females	10 (23.8)
Mechanism of injury, n (%)
Motor vehicle	20 (47.6)
Motor bicycle	7 (16.7)
Pedestrian	5 (11.9)
Fall	5 (11.9)
Other	5 (11.9)
GCS, median (range)	5 (3–10)
GCS ≤8, n (%)	39 (92.8)
Hypoxia, n (%)	20 (47.6)
Normoxia, n (%)	22 (52.3)
Focal, n (%)	15 (35.7)
Diffuse, n (%)	23 (54.8)
Focal & diffuse, n (%)	4 (9.5)
ISS, median (IQR)	36 (27–43)
ISS Hx, median (IQR)	41 (35–45)
ISS Nx, median (IQR)	30 (23–43)
GOSE, median (range)	3 (1–8)
Unfavorable outcome, n (%)	30 (71.4)
Favorable outcome, n (%)	12 (28.6)
GOSE Hx, median (IQR)	3 (1–4)
GOSE Nx, median (IQR)	4 (2.75–6)

Forty-two patients with severe traumatic brain injury (TBI) were recruited in the study. The following clinical parameters were recorded: Glasgow Coma Scale (GCS) at the scene corresponding to severe ≤8, moderate 9–12, and mild ≥13 TBI, respectively. Injury Severity Score (ISS): 0=no injury, 75=maximal untreatable injury. Hypoxia (Hx) was defined when patients had an SaO₂ <92% at the scene; normoxia (Nx) was defined as SaO₂ >92%. Type of brain injury was assessed using the Marshall classification as explained in detail in legend to Table 2. Diffuse brain injury was recorded in 23 patients, focal TBI was recorded in 19 patients, and a combination of both was recorded in 4 patients. The median ISS score was higher in Hx patients (ISS=41) than in the Nx patients, who had a median ISS=30 (p<0.05). Glasgow Outcome Scale Extended (GOSE) was assessed at 6 months post-injury whereby score 1=dead, 2=vegetative state, 3=lower severe disability, 4=upper severe disability, 5=lower moderate disability, 6=upper moderate disability, 7=lower good recovery, 8=upper good recovery. The median GOSE score of the entire population was 3. Patient dichotomization into groups with unfavorable outcome (GOSE=1–4, n=30) and favorable outcome (GOSE=5–8, n=12), revealed that most patients had a poor outcome. The Hx group had lower GOSE score (median=3) than did the Nx group (median=4, p=0.03).

Clinical parameters and patient grouping

GCS, hypotension (systolic blood pressure <90 mm Hg) and occurrence of Hx were recorded at the accident scene by paramedics and medical staff. Patients were allocated into the Hx cohort (n=20) if they had SaO₂ <92% or were apneic or cyanotic at the field, consistent with Chi et al.³² Patients with normal SaO₂ formed the normoxic (Nx) cohort (n=22). Type of TBI was characterized using the Marshall CT classification;³³ focal brain injury was defined by the presence of an evacuated lesion or an unevacuated high- or mixed-density mass lesion >25 mL (n=15). Grade I–IV diffuse brain injury depended upon the degree of compression of the cisterns and a midline shift (n=23). GOSE was assessed at 6 months post-TBI as described previously.³⁴

Control patients and sample collection

CSF samples from hydrocephalus patients (without previous TBI) undergoing implantation of ventriculoperitoneal shunts (five males, five females, between 30 and 74 years of age) were used as controls. Exclusion criteria were similar to those for the TBI cohort. Control serum samples were obtained from 12 female and 8 male healthy volunteers, between the ages of 21 and 55 years.

Cytokine measurement

IL-2, IL-4, IL-6, IL-8, IL-10, GM-CSF, IFN-γ, and TNF were measured in the CSF of TBI patients and control individuals by luminescence multiplex assay system according to the manufacturer's instruction (Bio-Plex Human Cytokine Array, BioRad Laboratories, Hercules, CA) and our published work.³⁰

Measurement of biomarkers

MBP, NSE, and S100 were quantified in serum of TBI patients and control individuals by ELISA (MBP: Diagnostic Systems Laboratories, Webster, USA; NSE: CanAg Diagnostics, Gothenburg, Sweden; S-100: Diasorin Inc., Stillwater, USA) as described previously.²³

Assessment of CSF/serum albumin quotient

Albumin concentration in paired CSF and serum was measured by ELISA using a commercial kit (Bethyl Laboratories Inc., TX). To characterize the profile of the BBB, the CSF/serum albumin ratio (albumin quotient; Q_A) was calculated daily as described.³⁵ The function of the BBB was evaluated as follows: Q_A values between 0.007 and 0.01 indicated a mild disturbance, values between 0.01 and 0.02 indicated a moderate one, and values >0.02 indicated a severe BBB dysfunction.³⁶

Statistical analysis

Data were analysed using the SAS software version 9.2 (SAS Institute, Cary, NC) with the assistance of a biostatistician (E.P.). Cytokine and biomarker measurements were assessed for normality, and logarithmic transformation was applied where appropriate. To account for repeat measures, data were analyzed using the PROC MIXED procedure in SAS with each patient treated as a random effect. Models were fitted using main effects for group (Hx vs. Nx) and time (0–5 days) and an interaction between groups and time to ascertain if the groups behaved differently over time. Because of the predominance of diffuse over focal TBI in our patient population (23 vs. 15 patients, respectively), to exclude the interaction of injury type in our analyses, separate models were fitted for group effect (Hx vs. Nx) by including injury type as a main effect in the model. This approach proved that type of TBI had no influence on the results. Post-hoc comparisons were performed using Tukey's test to account for multiple comparisons. Results from the mixed-effects models are presented as means (95% confidence intervals [CIs]) or geometric means (95% CIs) when logarithmic transformation was applied.

Correlation of Injury Severity Scores (ISS) with cytokines and biomarkers was performed as follows. The ISS of each patient was correlated with the correspondent CSF cytokine concentration (eight cytokines) or serum biomarker value (three biomarkers) using linear regression analysis. The analyses were conducted in all TBI patients and subgroup patients with Hx using GraphPad Prism 6 (GraphPad Software Inc.).

Comparison of GOSE between groups (Hx vs. Nx) was performed using the Mann–Whitney U test, with results reported as medians and interquartile ranges (IQRs). Spearman rank correlation (ρ) was used to assess the relationship between serum biomarkers and GOSE scores. CSF-serum albumin ratios are reported as medians and IQRs. Statistical significance was set at a two sided p value of 0.05. Graphics were created using GraphPad Prism (6.0.1) with data generated in SAS for better visual qualities.

Results

Demographic data of the patients

Forty-two TBI patients ranging between 16 and 63 years of age (median age of 29 years), with the majority being males (n=32, 76.2%) were recruited (Table 1). The median age in control individuals and TBI patients was 35 years with a range of 21–69 years. Most patients (n=32) had been in road traffic accidents (76.2%) and the remaining patients sustained: penetrating injury (n=5); fall/jump (n=6); blunt assault (n=7), or other mechanisms (n=8). GCS at the scene was ≤8 for all patients with the exception of three patients who had initial GCS between 9 and 10 and rapidly deteriorated, requiring intubation prior to admission to hospital (Tables 2 and 3).

Table 2.

Detailed Clinical Scores and Extracranial Injuries Potentially Leading to Hypoxia and Hypotension after TBI (Hx Patients)

Patient No.	GCS	ISS	Marshall Scores	GOSE	SaO2	Extracranial injury hypoxia	Extracranial injury hypotension
1.	5	41	D-II	5	Und	Fractured L ribs, L hemothorax, L lung collapse	Splenic maceration
2.	7	43	D-II	5	82%	Extensive facial fractures (bilateral LeFort III)	Nil
3.	3	58	D-II	3	Und	Extensive bilateral consolidation secondary to aspiration, L pneumothorax, bilateral multiple rib fractures	Nil
4.	3	27	D-II	5	Und	Bilateral pneumothoraces with consolidation of R upper and bilateral lower lobes. R and L lung contusions	Nil
5.	5	48	D-II	2	90%	Extensive facial fractures, L pneumothorax, bibasal contusions	Nil
6.	5	41	D-II	4	88%	L pneumothorax and extensive bilateral lung contusions	Fracture L inferior pubic ramus, diastasis of R sacroiliac joint
7.	3	35	D-II	3	82%	R upper lobe lung collapse, bilateral pneumothoraces	Nil
8.	3	36	D-II	3	Und	L tension pneumothorax, bilateral lung contusion, bilateral rib fractures	Nil
9.	6	17	EML	4	Und	Found by paramedics with obstructed airway 35 minutes after call-out, with central cyanosis	Nil
10.	7	43	EML	3	89%	R lung contusion, bilateral lung LL collapse, extensive facial fractures	Nil
11.	3	35	EML	1	86%	R pulmonary contusion	Nil
12.	6	34	EML	3	88%	Fractured R mandible and maxilla, R upper lobe and lower lobe contusion	Nil
13.	4	38	EML	1	Und	Nil	Nil
14.	3	45	EML	3	Und	Bilateral fractured ribs with bilateral hemothoraces	Nil
15.	3	35	UEML	1	Und	Nil. Unknown period of “downtime”	Bilateral pubic rami fractures
16.	9*	35	UEML	3	90%	L pneumothorax and L lung contusion	Nil
17.	5	45	UEML	1	Und	Multiple L rib fractures, L pneumothorax, bilateral pulmonary contusions	Nil
18.	3	57	UEML / D-III	1	Und	Bilateral pneumothoraces, bilateral hemothoraces, bilateral rib fractures	Jejunal and ileal tear, mesenteric contusions
19.	3	50	UEML / D-III	1	Und	LeFort 2 facial fractures, bilateral pulmonary contusions	Extensive pelvic fractures
20.	3	43	UEML / D-III	4	90%	R pulmonary contusions, extensive facial fractures	Nil

Clinical data of 20 hypoxic (Hx) patients, including Glasgow Coma Scale (GCS) score at the scene, Injury Severity Score (ISS), Marshall classification of traumatic brain injury (TBI), Glasgow Outcome Scale Extended (GOSE) scores at 6 months post-injury, and extracranial injuries more often associated as causes leading to Hx or hypotension. GCS was ≤8, with the exception of one Hx patient (No 16) and two normoxic (Nx) patients (see Table 3, Nos. 19 and 22) who had initial GCS 9 and 10 at the scene but rapidly deteriorated requiring intubation prior to hospital admission as indicated with^*. The Marshall classification comprises: D-I=diffuse axonal injury grade 1; D-II=diffuse axonal injury grade 2; D-III=diffuse axonal injury grade 3; D-IV=diffuse axonal injury grade 4; EML=evacuated mass lesion; UEML=unevacuated mass lesion. Twenty-three patients had diffuse brain injury, and 19 had focal TBI, including 9 with EML and 6 with UEML. Of those with focal TBIs, three patients had UEML combined with D-III, and one had EML combined with D-II. After TBI, Hx (SaO₂<92%) was detected in 20 patients likely caused by chest injury (rib fractures, mono-/bilateral pneumothorax, lung contusions) (n=12), airway obstruction (facial fractures) (n=3), a combination of both (n=3), and central Hx of unknown causes (n=2). In Nx patients minor chest injuries did not lead to decreased SaO₂. Hypotension was only ascertained in one Hx patient (18) and in none in the Nx patient group (not shown). Higher ISS scores in Hx patients are consistent with higher frequency of severe chest trauma as well as extracranial injuries including pelvis/ fracture, and spleen and liver rupture as recorded in five Hx patients.

Table 3.

Detailed Clinical Scores and Extracranial Injuries Potentially Leading to Hypoxia and Hypotension after TBI (Nx Patients)

Patient No.	GCS	ISS	Marshall Scores	GOSE	SaO2	Extracranial injury hypoxia	Extracranial injury hypotension
1.	3	43	D-I	3	94%	Complex facial fractures, bibasal lung collapse	Nil
2.	3	14	D-I	6	94%	Nil	Nil
3.	7	13	D-II	6	97%	Nil	Nil
4.	7	34	D-II	3	100%	Small R pneumothorax	Nil
5.	5	43	D-II	1	98%	Fracture R ribs	Nil
6.	8	26	D-II	4	97%	Nil	Nil
7.	5	25	D-II	3	96%	R tension pneumothorax, fractured R ribs	Nil
8.	3	43	D-II	3	100%	Nil	Hepatic laceration
9.	8	30	D-II	5	94%	Nil	Nil
10.	7	26	D-II	8	100%	R pulmonary contusions, fractured bilateral ribs, small R pneumothorax	Nil
11.	6	43	D-II	1	98%	Small bilateral pneumothoraces	Extensive L pelvis fractures
12.	3	50	D-II	2	96%	Nil	Nil
13.	7	21	D-II	3	99%	Small R pneumothorax	Nil
14.	3	45	D-III	2	100	Nil	Hepatic contusion
15.	3	50	D-III	1	98%	Nil	Hepatic laceration
16.	7	24	EML	4	100%	Nil	Nil
17.	7	17	EML	6	100%	Nil	Nil
18.	7	30	EML	5	96%	Nil	Nil
19.	10*	30	EML / D-II	7	99%	Nil	Nil
20.	6	41	UEML	4	99%	Nil	Duodenal tear
21.	5	35	UEML	6	96%	Extensive facial fractures	Nil
22.	10*	20	UEML	8	99%	Complex facial fractures, bibasal lung collapse	Nil

Clinical data of 22 normoxic (Nx) patients, including Glasgow Coma Scale score (GCS) at the scene, Injury Severity Score (ISS), Marshall classification of traumatic brain injury (TBI), Glasgow Outcome Scale Extended (GOSE) scores at 6 months post-injury, and extracranial injuries more often associated as causes leading to hypoxia (Hx) or hypotension. GCS was ≤8 with the exception of 1 Hx patient (Table 2 No. 16) and two Nx patients (Nos. 19 and 22) who had initial GCS 9 and 10 at the scene but rapidly deteriorated requiring intubation prior to hospital admission as indicated with^*. The Marshall classification comprises: D-I=diffuse axonal injury grade 1; D-II=diffuse axonal injury grade 2; D-III=diffuse axonal injury grade 3; D-IV=diffuse axonal injury grade 4; EML=evacuated mass lesion; UEML=unevacuated mass lesion. Twenty-three patients had diffuse brain injury, and 19 had focal TBI, including 9 with EML and 6 with UEML. Of these focal TBIs, three patients had UEML combined with D-III, and one had EML combined with D-II. After TBI, Hx (SaO₂ <92%) was detected in 20 patients likely caused by chest injury (ribs fractures, mono-/bilateral pneumothorax, lung contusions) (n=12), airway obstruction (facial fractures) (n=3), a combination of both (n=3), and central Hx of unknown causes (n=2). In Nx patients, minor chest injuries did not lead to decreased SaO₂. Hypotension was only ascertained in one Hx patient (Table 2, No. 18) and in none in the Nx patient group (not shown). Higher ISS scores in Hx patients are consistent with higher frequency of severe chest trauma as well as extracranial injuries including pelvis/ fracture, and spleen and liver rupture as recorded in five Hx patients.

Post-TBI Hx (SaO₂ <92%) was detected by paramedics at the scene in 20 patients, mostly because of chest injury (ribs fractures, mono-/bilateral pneumothorax, lung contusions) (n=12), airway obstruction (facial fractures) (n=3), a combination of both (n=3), or peripheral hypoxia of unknown causes (n=2) (Table 2). The occurrence of minor chest/upper respiratory tract injuries in 9 of 22 Nx patients did not lead to decreased SaO₂ (Table 3). Hypotension was only ascertained in one Hx patient (No. 18, Table 2) and in none in the Nx patient group (not shown).

The ISS reflect the combined injuries sustained, including brain and peripheral trauma. The median ISS detected in Hx patients (median [IQR]=41 [35–45]) was significantly higher than that calculated in Nx patients (median [IQR]=30 [23–43]; p<0.05). This finding is consistent with the higher frequency of severe chest trauma, airway obstruction, and respiratory arrest as well as extracranial injuries including pelvis fracture, and spleen and liver rupture in the Hx cohort. According to the CT Marshall score, out of 20 patients with hypoxia, 8 had diffuse axonal injury (DAI)-II, and 3 had a DAI-III combined with an unevacuated mass lesion (UEML) (Table 2). In the Nx counterpart, 15 patients had DAI-I–III, 6 had UEML/EML, and 1 had a mixture of EML/DAI-II (Table 3).

Outcome at 6 months was generally poor, with a median GOSE of 3. Of the seven patients who died, four died in the 1st week, and three died between 9 and 20 days post-injury. Five patients died in the Hx cohort and two died in the Nx cohort. Following dichotomization, 30 patients had an unfavorable outcome with GOSE 1–4 (71.4%), whereas 12 patients had a favorable outcome with GOSE 5–8 (28.6%). Patients with post-TBI hypoxia presented a slightly worse outcome (median [IQR] GOSE=3 [1–4]) as compared with Nx patients (median [IQR] GOSE=4 [2.75–6]; p=0.03).

Increased concentration of cytokines in CSF after TBI

Eight cytokines that have been more frequently characterised in human TBI were measured in CSF over 6 days. They included the pro-inflammatory mediators IL-2, IL6, IL-8, GM-CSF, IFN-γ, and TNF and the anti-inflammatory cytokines IL-4 and IL-10.^{14,16,35,37–40} With the exception of GM-CSF, the concentration of all cytokines increased significantly in CSF of TBI patients (n=42) over 6 days relative to controls (p<0.05, Fig. 1). When the total concentration of cytokines in all patients was considered, IL-6 and IL-8 presented elevated geometric means (95% CI) of 1756 (1232–2502) pg/mL and 653.5 (483.7–882.8) pg/mL, respectively, as compared with the corresponding control values being 6.23 (2.98–13.01) pg/mL and 8.12 (4.76–13.84) pg/mL. IFN-γ and TNF, although at lower concentrations, were also significantly elevated after TBI with geometric means (95% CI) of 20.05 (16.83–23.89) pg/mL and 10.24 (8.81–11.90) pg/mL, respectively. IL-4 and IL-10 were produced at much lower levels with geometric means (95% CI) of 0.3 (0.25–0.35) pg/mL and 3.06 (2.06–5.17) pg/mL, respectively. This large imbalance in the production of cytokines indicates an overwhelming pro-inflammatory response acutely post-TBI.

FIG. 1.

Profiles of cytokines in cerebrospinal fluid (CSF) of the entire traumatic brain injury (TBI) population. CSF samples were collected from day 0 (admission) until day 5 following TBI. The data represents the profiles of each cytokine measured in CSF of all 42 patients over 6 days. The graphs show respectively: (A) interleukin (IL)-2, (B) IL-4, (C) IL-6, (D) IL-8, (E) IL-10, (F) granulocyte macrophage-colony stimulating factor (GM-CSF), (G) interferon-γ (IFN-γ), and (H) tumor necrosis factor (TNF). The number of samples for each time point is indicated at the bottom of x-axis of each graph. There were missing patient samples over the course of study period because of early removal of extraventricular drain (EVD), no CSF drained on the day, or patient death. *Indicates significant differences of cytokine geometric means at the specific time points between TBI and control CSF obtained from non-TBI patients (n=10) (p<0.05).

Analysis of cytokine profiles detected either a significant increase or a trend toward higher concentrations during the first 24–48 h immediately post-TBI (Fig 1). Subsequently, IL-6, IL-8, and IL-10 gradually decreased reaching concentrations similar to controls (Fig. 1 C–E). IL-2, IL-4, and IFN-γ peaked on day 0, with levels remaining statistically elevated until day 5 (Fig. 1 A, B, G). In contrast, TNF was found to be increased for the entire study period after TBI, but it was statistically different from controls only on day 3 (Fig. 1 H).

GM-CSF and IFN-γ are prolonged following post-traumatic hypoxia

We next explored whether hypoxia enhances cerebral inflammation elicited by TBI by dividing patients into Hx and Nx groups. All cytokines, except GM-CSF, increased significantly to a similar degree in both cohorts relative to controls (p<0.05) (Fig. 2). Although IL-6, IFN-γ, and IL-10 had higher geometric means in Hx patients, there was no statistical difference when compared with the Nx group. Conversely, GM-CSF was significantly elevated only in Hx patients over controls (Fig. 2 F). Cytokine profiles over 6 days post-injury showed significant differences between Hx and Nx groups for GM-CSF and IFN-γ (Fig. 2 f and g) with post-hoc analysis revealing higher cytokines in the Hx group at 4–5 days (Fig. 2 f and g). TNF showed a trend toward higher concentrations in Hx patients on these days (Fig. 2 h).

FIG. 2.

Comparison of cytokine concentrations and profiles in hypoxic (Hx) and normoxic (Nx) traumatic brain injury (TBI) patients. This figure shows cytokine data on TBI patients grouped into HX (n=20) and Nx (n=22) cohorts and uninjured controls (n=10). The graphs indicate: (A, a) interleukin (IL)-2, (B, b) IL-4, (C, c) IL-6, (D, d) IL-8, (E, e) IL-10, (F, f ) granulocyte macrophage-colony stimulating factor (GM-CSF), (G, g) interferon-γ (IFN-γ), and (H, h) tumor necrosis factor (TNF), respectively. The upper letter graphs represent the average cytokine concentrations over 6 days after TBI. A significant increase in the geometric means of IL-2, IL-4, IL-6, IL-10, GM-CSF, IFN, and TNF, but not for IL-8, was detected in the Hx group relative to control. Significant elevation was also observed between the Nx cohort and control for IL-2, IL-4, IL-6, IL-10, and TNF, but not for IL-8, GM-CSF, and IFN-γ. No differences were found between the Hx and Nx cohorts for any cytokines. *Indicates significant differences between the groups (p<0.05). The graphs in small letters depict the profile of the daily geometric means of each cytokine over 6 days post-injury in Hx and Nx patients relative to controls. Although the pro-inflammatory cytokines IL-2 (a) and IFN-γ (g) increased more markedly on day 0 in the Nx group, in general no differences were observed for most cytokines between the TBI cohorts with the exception of GM-CSF (f ) and IFN-γ (g), which were significantly prolonged in the Hx group over the last 2 days of the study (4 and 5), suggesting sustained cerebral inflammation. A trend toward cytokine increase on these days was also detected for TNF (h). The number of samples for each time point is indicated at the bottom of x-axis of each graph. There were missing patient samples over the course of study period because of early removal of the extraventricular drain (EVD), no cerebrospinal fluid (CSF) drained on the day, or patient death.

Elevation of MBP and S100 is associated with post-traumatic hypoxia

We subsequently investigated the ability of MBP, NSE, and S100 serum concentrations to discriminate between Hx and Nx. When all TBI data were pooled disregarding the time, all three biomarkers were significantly heightened over control levels. MBP increased by >5-fold displaying a geometric mean of 1.17 (0.87–1.57) ng/mL over controls 0.22 (0.21–0.23) ng/mL; NSE was >12-fold higher (mean=0.57 [0.27–1.22] ng/mL) than controls (mean=0.046 [0.002–0.866] ng/mL), and S100 increased by nearly 4-fold, with a mean of 0.30 (0.24–0.38) ng/mL against control 0.08 (0.05–0.12) ng/mL. Over 6 days after TBI, NSE and S100 concentrations peaked on day 0 (Fig. 3 B, C) and remained significantly higher than control until day 4 post-TBI. The increase of serum MBP occurred with a delay between days 0 and 4, and reached statistical significance only on day 5 (Fig. 3 A).

FIG. 3.

Increased concentration of brain injury biomarkers myelin basic protein (MBP), neuronal specific enolase (NSE), and S100 following traumatic brain injury (TBI). The graphics on the upper row show the time course of (A) MBP, (B) NSE, and (C) S100 in all TBI patients over the study period. NSE (B) and S100 (C) were significantly increased at day 0 after injury and slowly decreased over time, whereas MBP (A) showed a constant elevation between days 0 and 4 before a significant difference to control value was observed on day 5. The number of samples for each time point is indicated at the bottom of x-axis of each graph. There were missing patient samples over the course of study period because of early removal of extraventricular drain (EVD), no cerebrospinal fluid (CSF) drained on the day, or patient death. Graphs D, E, and F show the protein levels in hypoxic (Hx) and normoxic (Nx) patients. The average biomarker concentrations over 6 days after TBI were used for each patient to calculate the geometric mean (ng/mL) with 95% confidence intervals in each group. *Indicates a significant difference between the time points and controls (p<0.05).

Although all biomarkers were markedly elevated after TBI, MBP and S100 were superior in differentiating Hx and Nx patients (Fig. 3 D, F), and were significantly increased by 5.4- and 7.7-fold respectively in the Hx cohort when compared with control. Conversely, NSE concentrations were not statistically different between the three cohorts. Comparison of MBP and S100 levels between Hx and Nx groups showed a nonsignificant elevation by 2.7- and 2.0- fold in Hx patients. A trend increase in serum NSE levels was also observed in Hx over Nx patients, but these differences were not significant (Fig. 3 D–F).

MBP, NSE and S100 correlate with adverse outcome in hypoxic patients

When analyzing the entire patient population, we only found a correlation between maximal S100 and the GOSE (ρ=−0.446, p=0.01, n=33), and lack of correlations for MBP (ρ=−0.360, p=0.071, n=26) and NSE (ρ=−0.179, p=0.305, n=34). Interestingly, following the separation of cohorts, the correlation between S100 and GOSE became stronger (ρ=−0.486, p=0.048, n=17) in the TBI-Hx compared with the Nx group (ρ=0.358, p=0.097, n=16). In Hx patients, MBP approached a significant correlation to the GOSE (ρ=0.550, p=0.051, n=13).

Furthermore, dichotomization of all TBI patients into those with favorable (n=9) and unfavorable (n=26) outcome, revealed an association between a higher concentration of all three biomarkers and unfavorable outcome (GOSE 1–4) at 6 month post-TBI (1.367 [0.746–2.507], 8.405 [5.618–12.574], 0.423 [0.304–0.588] ng/mL geometric mean [confidence intervals] for MBP, NSE, and S100, respectively) as compared with control values (0.217 [0.482–0.098], 0.974 [2.455–0.387], 0.081 [0.148–0.045] ng/mL geometric mean [confidence intervals] for MBP, NSE and S100, respectively). In contrast, there was no difference in any of the biomarkers' geometric means between patients with favorable outcomes (GOSE 5–8) and controls. After subdividing patients with unfavorable outcome according to their Hx (n=15) and Nx (n=11) status, all three biomarkers showed significantly higher concentrations in the Hx cohort than in controls, with no difference between Nx patients and controls (Fig. 4 A–C).

FIG. 4.

Analysis of brain injury biomarkers in relation to neurological outcome. The graphs show respectively: (A) myelin basic protein (MBP), (B) neuronal specific enolase (NSE), and (C) S100. Patients with unfavourable outcome (unfav; Glasgow Outcome Scale Extended [GOSE] 1–4) were subdivided into the hypoxic (Hx) or normoxic (Nx) cohorts. A significant elevation in MBP, NSE, and S100 was observed in the Hx group, whereas NSE and S100 increased also in Nx groups relative to control. *Indicates a significant difference between the groups (p<0.05). The average biomarker concentrations of each patient over 6 days after TBI were used to calculate the geometric mean (ng/mL) with 95% confidence intervals in each group.

CSF/serum albumin quotients do not differ in hypoxic and normoxic patients

To determine whether a greater diffusion of biomarkers into blood of Hx patients may underlie a more severe disturbance of the BBB, we compared the profiles of CSF albumin in both the Hx and Nx groups. All TBI patients had a marked increase of CSF albumin concentration, rising from a normal level <0.34 mg/mL to a median concentration of 1.21±0.3 mg/mL at day 0, reflecting an increased neurovascular permeability and/or possible blood contamination caused by hemorrhages. Over time, CSF albumin gradually declined to 0.42±0.24 mg/mL at day 5 post-TBI.

We then calculated the daily CSF/serum albumin ratios (albumin quotient, Q_A). The Q_A profile delineated a pattern similar to the fluctuations of CSF albumin, showing a higher value at day 0 (median=0.0259, range: 0.0178–0.0399; severe disturbance), which gradually decreased at day 5 (median=0.0017, range: 0.0007–0.0336; mild disturbance). When patients were grouped according to Hx or Nx, we found no statistical differences in the median Q_A values, although a few Hx patients still maintained elevated ratios on days 2 and 3, with Q_A values >0.01 (Fig. 5).

FIG. 5.

Profiles of CSF/serum albumin quotients in hypoxic (Hx) and normoxic (Nx) traumatic brain injury (TBI) patients. The curves show the daily median values with 25–75 percentile of the albumin quotient (Q_A) in Hx and Nx patients over 6 days post-TBI. Q_A values 0.02, indicative of severe blood–brain barrier (BBB) disruption, occurred on day 0, and gradually declined over time. Although no significant differences were detected between the groups, the Hx cohort displayed elevated Q_A values on days 2–4 post-TBI.

Discussion

In this study, we report for the first time changes in cerebral inflammation, biomarkers of brain damage and CSF/serum albumin quotients in TBI patients with and without Hx. We demonstrate that secondary Hx is associated with prolonged cytokine production and greater elevation of serum biomarkers, and that increased biomarker levels correlate with lower GOSE scores.

Although pre-hospital Hx remains a major cause of mortality in TBI patients, doubling the risk of adverse outcome, its impact on the pathophysiology of secondary brain damage remains elusive. Surprisingly, Hx is not often taken into consideration as a co-variable when reporting epidemiological data on outcomes.^5,41 The detrimental effect of Hx is supported in our study, whereby Hx patients generally had a worse outcome than Nx patients. Five of seven patients who died after TBI had Hx, compared with the absence of Hx in patients with favorable outcome (GOSE 6–8). The causes of Hx likely depend on the larger incidence of extracranial trauma in this patient cohort. In fact, most Hx patients (18/20) presented severe chest injuries and higher ISS, comprising collateral pathologies potentially leading to post-TBI Hx and poorer outcome or death. Conversely, only three Nx patients had significant chest/lung injuries that did not lead to Hx. Hypotension was only reported in one Hx patient, suggesting that this was not a factor determinant of Hx in our study. Overall, the percentage of Hx patients in our population was relatively higher than in other clinical studies on post-traumatic Hx. The reasons underlying the difference in proportion of Hx versus Nx patients may be because older automobiles lacked preventive safety features (airbags) and there were longer periods of patient transport from the accident site to the hospital as is often encountered in Australia. This was discussed in a comparative epidemiological study with European and United States centers, whereby worse outcomes and incidence of secondary insults in the Australian population may reflect the greater retrieval distances in Australia than in Europe.⁴²

Neuroinflammation after TBI

Cerebral inflammation is one of the prominent secondary cascades elicited in response to TBI, and is thought to exacerbate secondary brain damage, but also promote repair.^14,16,43 Recently, we validated the increased production of cytokines measured in CSF by demonstrating their upregulation at both mRNA and protein levels in postmortem brains within the first minutes of TBI.⁴⁴ The data shown here support previous literature indicating that cytokines increase in human CSF, delineating an early elevation followed by a gradual decrease.^{35,38–40,45} The uniqueness of our study is the stratification of TBI patients in those with and without Hx, showing evidence of higher concentrations of GM-CSF in the Hx group. A prolonged release of GM-CSF and IFN-γ, and a trend for TNF, was also demonstrated in Hx compared with Nx patients.

The cytokine of particular relevance in this study was GM-CSF, whose production was enhanced and sustained in patients with post-traumatic Hx, possibly indicating a more profound pro-inflammatory response proportional to a greater injury severity. This corroborates the potent role of GM-CSF as a stimulator of macrophage migration, which is consistent with marked upregulation of GM-CSF at protein and mRNA levels in human brain regions coinciding with accumulation of CD-68-positive macrophages.⁴⁴ The ability of Hx to exacerbate GM-CSF may amplify macrophage infiltration in the injured brain, as shown in a combined TBI-Hx rat model.³¹ Conversely, beneficial functions of GM-CSF include its ability to promote neuronal survival, inhibit apoptosis, sustain neovascularization, and increase the synthesis of specific neurotrophic factors mediating neurogenesis.^46,47

The association of IFN-γ with post-traumatic Hx is an interesting one. In human brains, IFN-γ was overexpressed within the few minutes from TBI, subsequently reaching a >10-fold increase in patients dying hours later.⁴⁴ IFN-γ is secreted by glial cells and infiltrating monocytes, displaying a potential role in neurogenesis and repair.⁴⁸ Most importantly, in vitro studies reported IFN-γ to be an Hx-specific mediator induced by T-cells.⁴⁹

Although it is surprising that no additional differences in cytokine levels were found between the Hx and Nx patient groups, it is important to emphasize that all the patients enrolled in this study had severe TBI. It is conceivable that cytokine production had already reached a maximal response, thus preventing a further increase in CSF. For example, the average IL-6 level in CSF of our TBI population exceeded 13,000 pg/mL, in comparison to only 400–800 pg/mL observed in the CSF of other neurological patients.⁵⁰

When we performed a number of correlative analyses between maximal levels of all eight cytokines versus the GOSE scores, no significant correlations were detected, whether taking the entire TBI population or the Hx and Nx subgroups (data not shown). We also tested for possible correlations between CSF cytokines and the ISS, using a similar approach, and again, no significant correlations were observed with the entire patient population or after dichotomization into Hx or Nx groups. These data are in line with the conflicting reports by others and by us on the meaningful value of cytokines as biomarkers of brain injury.¹⁶

Serum biomarkers

A reliable serum biomarker is an ideal diagnostic tool to identify type and severity of injury to track the incidence of secondary insults and predict long-term outcomes. Despite their limitations, MBP, NSE, and S100 remain the most utilized biomarkers in observational studies and clinical trials.^26,27

In our patient population, the elevation and time profile of these biomarkers were similar to previous studies.^23,51–53 The most relevant findings are: 1) MBP and S100 are more robustly increased in patients with Hx; 2) all three biomarkers are more elevated in Hx patients with unfavorable outcome (GOSE 1–4); 3) S100 strongly correlates with GOSE in both, the whole TBI population, and the TBI-Hx group; and 4) the peak of all biomarkers mostly occurred on days 0 and 1, coinciding with maximal CSF/serum albumin ratios potentially indicative of BBB dysfunction. Subsequently, vasculature permeability might be restored and prevent further extravasation of brain proteins. Surprisingly, no differences in CSF/serum albumin ratios were found between Hx and Nx groups, possibly because of high variations in albumin measurements in the patient cohort, the small population size, and the limitations of the Q_A reliability. In fact, we cannot ascertain whether increased albumin in CSF is the result of protein diffusion from damaged vessels at the blood–brain/CSF barrier, or blood contamination caused by brain or ventricular hemorrhages frequently occurring after TBI. However, there is sufficient evidence in human and experimental TBI by means of neuroimaging and histological evaluation that the dysfunction of the BBB occurs in the early hours and days following trauma. This is shown by perivascular edema and albumin/immunoglobulin G (IgG) accumulation in the lesioned parenchyma.^54,55 Although we are aware of these limitations, the Q_A remains the only molecular approach to measure BBB dysfunction in live TBI patients.

This study is consistent with previous work showing an increase of S100 in serum of TBI patients²³ and correlations with lower cerebral perfusion pressure and higher ICP.^22,56 Although defined as a biomarker of brain damage, elevation of S100 caused by peripheral trauma cannot be excluded.⁵⁷ It has been reported that S100 does not cross the intact BBB.⁵⁸ This corroborates earlier observations of S100 and Q_A profiles being similar and closely reflecting vascular damage.^23,59 Elevated MBP in CSF has been observed in patients with various neurological diseases; however its use in the context of TBI is not as frequent as NSE and S100. The unique profile of MBP was interesting; unlike NSE and S100, it showed a steady elevation rather than an early peak followed by a rapid decline. This unique temporal behavior would be worth pursuing by measurement of MBP beyond 6 days post-TBI, possibly allowing detection of additional differences. MBP was more abundantly increased in serum of Hx patients and most importantly, MBP differentiated patients with unfavorable or favorable outcome. These results strengthen the future use of MBP as a prognostic biomarker for TBI patients, possibly suggestive of myelin and axonal damage. NSE showed the poorest correlations with Hx and outcome, thus leading one to question its significance as a reliable biomarker of brain damage. Generally, the correlations of biomarkers with Hx are consistent with a recent report showing that elevation of serum S100, GFAP, and NSE in TBI patients precedes the clinical symptoms of Hx. In this study, Hx was more precisely monitored by intraparenchymal PbO₂ measurements, thus implying these to be reliable predictors of cerebral Hx to guide clinicians to early intervention.⁶⁰

Interestingly, no correlations were detected between a biomarker's levels and ISS scores, therefore strengthening the hypothesis that enhanced serum biomarkers in blood may be a true consequence of brain damage rather than the contribution of brain and peripheral injuries, as biomarker levels were not affected by additional trauma.

Limitations

There are a number of limitations in this study. Despite the fact that the population size was relatively small to detect correlations of higher significance, the hypoxic patient cohort had a high frequency of peripheral injuries of the respiratory tract that likely caused a prevalence of Hx in >47% of the population, being higher than the incidence reported in the literature, which ranged from 6 to 44%.^61,62 The complexity of multitrauma in Hx patients made it difficult for us to definitively discern the impact of systemic Hx on outcome. The occurrence of extracranial injuries reflected by the higher ISS scores in Hx patients may ultimately affect long-term outcomes. It has to be emphasized, however, that all three biomarkers were more elevated in Hx patients with unfavorable outcome, making it improbable that this consistent increase was caused by peripheral injuries.

The lack of measurement of parenchymal O₂ did not permit establishment of the relationship between peripheral and cerebral Hx. According to Stein et al.,⁶⁰ cerebral Hx (using Lycox catheters) was detected in TBI patients treated in the intensive care unit (ICU), although peripheral SaO₂ was normal. This critical monitoring of parenchymal O₂ may assist in identifying the actual presence and duration of early cerebral Hx, as well as the onset of delayed brain Hx arising from secondary ischemic insults.

A direct causality of elevated cytokines and injury biomarkers with post-traumatic Hx cannot be substantiated with this study, and, therefore, remains a correlative association. Although our previous data have validated a direct relationship between enhanced parenchymal cytokine and poorer sensorimotor function in rats exposed to a 30 min Hx period following diffuse TBI, in the clinical studies, confounding factors including the heterogeneity of the patient population, make it difficult to prove the true impact of Hx on molecular and behavioral changes beyond doubt. These associations only suggest a potential role for hypoxia in prolonging neuroinflammation, enhancing blood levels of biomarkers, and worsening outcomes.

Conclusion

In patients with severe TBI, we have shown critical differences in cerebral inflammation and blood injury biomarker levels, which were found exacerbated in patients with secondary Hx. These molecular changes in hypoxic patients seemed to be independent from the CSF/serum albumin ratios. These data add to the current knowledge on the impact of post-traumatic Hx in the TBI clinical setting, and emphasize the association of complex pathological insults, which may amplify brain damage leading to adverse outcome. However, further work is warranted to fully elucidate the causality of Hx on these secondary events.

Acknowledgments

We thank Ann Sutherland from Victorian State Trauma Outcomes Registry and Department of Epidemiology, Monash University, and Shirley Vallance and Lynne Murray from the Intensive Care Unit, Alfred Health, for providing the GOSE scores, and David O'Reilly for partial albumin measurements. We thank Sarah Hellewell for her intellectual input on this project, and for manuscript preparation. This study was funded by Victorian Neurotrauma Initiative/Traffic Accident Commission project grant D009 and Research Fellowships to Drs. Morganti-Kossmann, Yan, and Bye; the National Health Medical Research Council Project Grant #436815; The Henry O'Hara Surgical Research Trust to Dr. Satgunaseelan, and the L.E.W. Carty Charitable Fund.

Author Disclosure Statement

No competing financial interests exist.