TBI alters CSF and serum protein levels and upregulates neuroinflammatory pathways
Using t-SNE, data grouped along compartment (serum and CSF) and disease status (TBI and control) (Fig. 2A). In general, t-SNE 1 corresponded to compartment (CSF/serum), and t-SNE 2 to individual patient characteristics. Interestingly, BBB integrity seemed to be related to t-SNE 2, particularly in CSF (Fig. 2B). This indicates that the CSF and serum proteomes are distinctly different following TBI, and that certain injury characteristics may be reflected in protein alterations.
Next, we examined which specific protein levels that changed following TBI. Among control subjects, CNS-originating proteins (e.g. GAP43, log2 fold change [FC] 3.41, p < 0.001) were enriched in CSF compared with serum, while for example complement proteins (e.g. C1QB, log2 FC -2.38, p < 0.001) were enriched in serum (Figure S8). Following TBI, n = 124 (unique) proteins were altered in either CSF or serum compared with controls (Fig. 2C-D, Table S3). This allowed us to assess currently used TBI biomarkers, comprising the astrocytic proteins S100B and glial fibrillary acidic protein (GFAP), as well as the neuronal proteins neuron-specific enolase (NSE, also referred to as ENO2), neurofilament-light (NFL), and ubiquitin carboxy-terminal hydrolase-L1 (UCH-L1) (38). We could confirm previous findings of upregulation of S100B, GFAP, NSE (ENO2), and NFL post-TBI (Table S3).
Next, we characterized all proteins altered following a severe TBI. As expected, far more proteins were altered in CSF (n = 109) than in serum (n = 35) following TBI. In CSF, n = 81 (74%) of all altered proteins were CNS enriched, whereas n = 11 (10%) were immune system function related. Proteins notably enriched in CSF following TBI were among else MBP (ΔMFI = 3655, p < 0.001), and AQP4 (ΔMFI = 2208, p = 0.002). In contrast, there were n = 7 (20%) altered proteins with immune system function in serum following TBI. Similarly to CSF, the majority of altered proteins were predominant CNS enriched (n = 23, 66%). The proteins in serum that exhibited the highest ΔMFI were the complement proteins CFB (ΔMFI = 2131, p < 0.001) and C9 (ΔMFI = 2000, p < 0.001). Pathway analysis of these revealed that top-altered pathways in CSF included the lectin-induced complement pathway, erythropoietin-mediated neuroprotection through Nuclear Factor Kappa-Light-Chain-Enhancer of Activated B cells (NF-κB), synaptic proteins at the synaptic junction, and Role of Tob in T-cell activation (Fig. 2E). This was partially mimicked in serum with regard to the neuroinflammatory pathways, entailing both signal transduction through IL1R, cytokines and inflammatory response, as well as the complement system (Fig. 2F).
Surprisingly, merely n = 19 proteins were altered concurrently in both CSF and serum following TBI. Among these, n = 12 proteins (63%) were CNS enriched and n = 4 (21%) had an immune system related function. Among immune system proteins, notably all but one (CXCL1) were complement system proteins (CFI, FCN1, MASP2).
BBB disruption following severe TBI yields a protein signature in CSF and is predictive of outcome
Median QA was 0.004 (0.002–0.011) (Fig. 3A) and BBB disruption was present among 23 TBI patients (32%). The QA reference interval is defined for lumbar albumin (36), and due to the CNS rostro-caudal gradient, the amount of ventricular albumin comprises ~ 40% that of lumbar albumin under homeostasis (66). As one would expect the rostro-caudal gradient to be inverted following a supratentorial trauma with ventricular albumin consequently higher than the lumbar ditto, we did not attempt any rostro-caudal correction for the QA reference interval, in line with our previous work (12, 15). A few control subjects exhibited pathological QA values (Table 1), in accordance with previous work where ~ 15% of healthy subjects exhibit a pathological QA in the absence of neurological disorder (67).
QA was an independent significant predictor of GOS (p = 0.044, ΔNagelkerke’s pseudo-R2 = 8.89%). This finding is novel and highlight BBB disruption as a prognostic marker for severe TBI. This finding could not be attributed to APOE4 carriership, as APOE4 variant was not associated with QA adjusted for age and sex (p = 0.494), or if injury severity was added to the model (p = 0.634).
In total, 114 unique proteins had a CSF/serum ratio significantly correlated with QA, conferring a median correlation coefficient τ 0.33 (0.29–0.40) (Table S4). The ten proteins with highest correlation coefficient τ between CSF/serum ratio and QA were complement proteins, except VCAM1 (Table 2). In fact, the majority of proteins that correlated with QA were either nervous system or immune system proteins, where the latter entailed (aside from the complement system proteins) for example the cytokines IL-1α, IL-1β, and IL-6 (Fig. 3B-C, Table S4). APOE4 was not a predictor of the QA associated protein levels in either CSF or serum.
Table 2
Complement Proteins Exhibited Highest Correlations with QA
Protein, Antibody | Specific function | 𝛕 | adjusted p-value |
C1QB HPA052116 | innate immunity/complement system | 0.67 | < 0.001 |
CFB HPA001817 | innate immunity/complement system | 0.66 | < 0.001 |
C9 HPA029577 | innate immunity/complement system | 0.65 | < 0.001 |
C9 HPA070709 | innate immunity/complement system | 0.65 | < 0.001 |
C1QA HPA002350 | innate immunity/complement system | 0.64 | < 0.001 |
MASP2 HPA029314 | innate immunity/complement system | 0.58 | < 0.001 |
VCAM1 HPA069867 | cell cell communication | 0.54 | < 0.001 |
FCN3 HPA071173 | innate immunity/complement system | 0.54 | < 0.001 |
MASP2 HPA029313 | innate immunity/complement system | 0.52 | < 0.001 |
C5 HPA075945 | innate immunity/complement system | 0.52 | < 0.001 |
Top 10 QA correlated proteins as deemed by correlation coefficient Kendall 𝛕. Correlations were calculated between protein CSF/serum ratio and QA. Abbreviations: CNS, central nervous system; CSF, cerebrospinal fluid; QA, albumin quotient. Full protein names are detailed in Table S1. |
Cluster analysis of QA correlated proteins revealed that in CSF, but not in serum, protein levels paralleled QA (Fig. 3D-E). Of note, the protein levels did not exhibit any association with APOE4 (Fig. 3D-E). Among proteins significantly different between the CSF clusters, pathway analysis exhibited that structural (synaptic proteins at the synaptic junction) and inflammatory pathways (complement pathway, lectin-induced complement pathway, IL5-signaling, Role of Tob in T-cell activation, signal transduction through IL1R, TGF-β signaling pathway) were upregulated (Fig. 3F). Finally, we examined whether any proteins were significantly different between TBI patients dependent on intact or disrupted BBB. In total, merely n = 7 of all our QA associated proteins were significantly altered dependent on QA status in both CSF and serum (Fig. 3G-H). In CSF, the majority were inflammatory (CFB, C9, IL6, FCN1), whereas in serum the only significant protein was the structural protein OLIG1.
Proteins associated with BBB disruption comprise outcome predictors following severe TBI
There was an overlap between proteins that were significantly altered (in either CSF or serum) following TBI and that were altered in the CSF cluster analysis among QA associated proteins (Fig. 4A-B). For these, and also for the proteins that were significantly different between patients with intact and disrupted BBB, we performed outcome analyses (Table S5-S6). Among protein intersects between cluster/bicompartmental analyses, n = 40 proteins comprised independent outcome predictors (the representative examples CASKIN1, MMP9, and complement C5 are highlighted in Fig. 4C-E). The proteins with highest ΔNagelkerke’s pseudo-R2 from both analyses are summarized in Table 3. Of these, the majority were CNS enriched (n = 23, 58%) as compared to (neuro)inflammatory (n = 9, 23%) proteins.
Table 3
BBB correlated proteins improved outcome prediction independently following severe TBI
Protein, Antibody | Compartment | Highest Tissue Enrichment | Coefficient | 𝚫R2 | adjusted p-value | QA subgroup analysis |
STMN4 HPA078407 | CSF | cns | -0.00505 | 0.121 | 0.04548 | no |
C5 HPA075945 | CSF | liver/gallbladder | -0.00095 | 0.106 | 0.04548 | no |
GPR26 HPA062736 | CSF | cns | -0.00684 | 0.099 | 0.04548 | no |
CFB HPA001817 | Serum | liver/gallbladder | 0.00098 | 0.092 | 0.04548 | yes |
FCN1 HPA001295 | Serum | blood | 0.00303 | 0.082 | 0.04548 | yes |
C9 HPA070709 | CSF | liver/gallbladder | -0.00123 | 0.074 | 0.04548 | yes |
IL6 HPA064428 | Serum | adipose/soft tissue | 0.00185 | 0.071 | 0.04548 | yes |
All proteins that comprised the intersect between CSF-altered proteins and CSF cluster-derived proteins or serum-altered proteins and CSF cluster derived proteins were used for outcome analysis. Outcome prediction was conducted by univariable followed by multivariable proportional odds regression analysis where GOS was used as dependent variable and the protein level as independent variable. The IMPACT variables were used as covariates. Here we show the n = 3 proteins that conferred the highest 𝚫Nagelkerke’s pseudo- R2 (decimal number) in CSF (row 1–3), in serum (row 4, 5, 7), and upon specific outcome analysis for proteins significantly different between patients with intact and disrupted BBB (row 4–6). Proteins that were significantly different between disrupted and intact BBB (CFB, FCN1, C9, IL-6) were subjected to a sub-group analysis (“QA subgroup analysis” column), for which adjusted p-values are described in Table S6. Abbreviations: BBB, blood-brain barrier injury; CNS, central nervous system; Coeff., Regression Coefficient; CSF, cerebrospinal fluid; GOS, Glasgow Outcome Score; IMPACT, International Mission for Prognosis and Analysis of Clinical Trials in TBI; TBI, Traumatic Brain Injury; QA, Albumin Quotient. All full protein names are listed in Table S1. |
Among proteins that had significantly altered levels if the TBI patient had a BBB injury we also found independent outcome predictors (Table S6, Table 3). In order to see which of these proteins that were particularly important, we made a step-down analysis, which comprised all proteins that were significant within the specific compartment upon multivariable analysis followed by sequential deletion until merely significant proteins were retained in the model. In CSF, C9 (Fig. 4F, p = 0.0143, ΔR2 = 7.4%) was the only protein retained. In serum, CFB (p = 0.0031, ΔR2 = 9.2%) was the only protein retained.