Mass Spectrometry-Based Approaches for Clinical Biomarker Discovery in Traumatic Brain Injury

Abstract

Purpose of Review:

Precision treatments to address the multifaceted pathophysiology of traumatic brain injury (TBI) are desperately needed, which has led to the intense study of fluid-based protein biomarkers in TBI. Mass Spectrometry (MS) is increasingly being applied to biomarker discovery and quantification in neurological disease to explore the proteome, allowing for more flexibility in biomarker discovery than commonly encountered antibody-based assays. In this narrative review, we will provide specific examples of how MS technology has advanced translational research in traumatic brain injury (TBI) focusing on clinical studies, and looking ahead to promising emerging applications of MS to the field of Neurocritical Care.

Recent Findings:

Proteomic biomarker discovery using MS technology in human subjects has included the full range of injury severity in TBI, though critically ill patients can offer more options to biofluids given the need for invasive monitoring. Blood, urine, cerebrospinal fluid, brain specimens, and cerebral extracellular fluid have all been sources for analysis. Emerging evidence suggests there are distinct proteomic profiles in radiographic TBI subtypes, and that biomarkers may be used to distinguish patients sustaining TBI from healthy controls. Metabolomics may offer a window into the perturbations of ongoing cerebral insults in critically ill patients after severe TBI.

Summary:

Emerging MS technologies may offer biomarker discovery and validation opportunities not afforded by conventional means due to its ability to handle the complexities associated with the proteome. While MS techniques are relatively early in development in the neurosciences space, the potential applications to TBI and neurocritical care are likely to accelerate in the coming decade.

Introduction

Traumatic brain injury (TBI) is a common, deadly disease. More than 64,000 TBI-related deaths occurred in the United States in 2020[1], which translates to approximately 176 TBI-related deaths every day. To date, there is no single Food and Drug Administration (FDA) approved therapy for moderate or severe TBI. Identifying treatments for TBI has been challenging, owing in part to its complex, heterogeneous pathophysiology that is produced by head trauma. Clinicians struggle to accurately diagnose the subtypes or endophenotypes of TBI, relying on level of consciousness and motor function to guide treatment rather than specific pathophysiology. To address the multifaceted pathophysiology of TBI with precision treatments, clinicians need ancillary tools to identify endophenotypes in TBI[2]. These gaps have led to the intense study of biomarkers in TBI, especially fluid-based protein biomarkers [3,4]. As of this writing, there is only one FDA approved biomarker for use in TBI, the glial fibrillary acidic protein/ubiquitin carboxyl-terminal esterase L1 panel in blood, which is used to discriminate the presence of lesions on head computed tomography in mild TBI[5], though other markers and indications are likely to follow. Comprehensive reviews on specific fluid-based biomarkers have been published elsewhere[3,6,7]; this review will focus on biomarker measurement techniques used in TBI, especially mass spectrometry (MS).

Fluid-based protein biomarkers many be measured with a number of different techniques, depending on the clinical or research scenario. Many of the assays used in clinical practice and biomedical research rely on antibody binding to proteins of interest for detection (classically, enzyme-linked immunosorbent assays or ELISAs)[8]. Antibody-based assays have previously been favored due to their performance characteristics—they can be highly specific and precisely quantitative without interference from albumin or other high abundance proteins[9]. Antibody array instruments are considerably less expensive than MS platforms and require less extensive training. However, discovery of new targets is limited as the specific protein of interest must be defined, and an antibody specific to the protein or modification of the protein must be developed prior to protein detection, with the appropriate sensitivity and specificity [10].

In contrast to antibody-based assays, high-throughput mass spectrometry (MS) is a valuable tool for protein identification due to its ability to capture the complexities associated with the proteome. The study of the proteome, or “proteomics” encompasses the full set of protein forms expressed by an organism or biological system[11]. The proteome represents the original “transcriptome” or the full range of translated mRNA molecules [12]. But proteomic studies have an advantage over genomic studies since not all transcriptional changes have concomitant effects on protein levels. Historically, MS was primarily used in the clinical setting for diagnostic purposes such as toxicology testing, diagnosing metabolic deficiencies, or for determining whether biomarkers or enzymes are present[13]. Now, MS is increasingly being applied to biomarker discovery and biomarker quantification in numerous diseases [14] due to its ability to provide quantitative analysis and sequence specific results allowing for more flexibility than antibody-based assays. Within Neurology, MS-based proteomic techniques (neuroproteomics) have been employed in a wide variety of disease states such as Alzheimer’s[15], ALS[16], and neuroinfectious disease[17], among others. In this narrative review, we will provide specific examples of how MS technology has advanced translational research in traumatic brain injury (TBI) focusing on clinical studies where possible, and looking ahead to promising emerging applications of MS to the field of Neurocritical Care.

Protein Detection by Mass spectrometry (MS)

Mass spectrometry measures the mass-to-charge ratio (m/z) of gas-phase ions by using an ion source (to convert molecules for analysis into their respective gas-phase ions), a mass analyzer (to separate ionized analytes based on m/z ratio), and a detector (to record the number of ions at each m/z value)[14]. Tandem mass spectrometry (or MS/MS) is a key technique for protein or peptide sequencing and post-translational modification (PTM) analysis that is generally used to break down certain selected precursor ions into fragments, or product ions, which then reveal aspects of the chemical structure of the original precursor ion(s). MS-based proteomics can be classified by two different application approaches: global discovery or “shotgun” proteomics versus targeted proteomics. Further, the most common MS workflows can be classified into two different approaches: top-down versus bottom up. (Figure 1) Global proteomics involves digestion of a complex assortment of substances (e.g. proteins, tissue, serum, or even isolated cells), reducing them to a peptide mixture, and then applying various off-line high-performance liquid chromatology (HPLC) separations to reduce the complexity of the initial sample. The subsequent fractionated peptides are then analyzed using on-line liquid chromatography (LC)-MS (either 1D or multidimensional) for peptide precursor mass measurement[18] and/or tandem MS (MS/MS) spectra for accurate detection, identification and quantification of the peptides which can then be extrapolated onto their respective proteins. Peptide sequence identification is based upon searching the MS/MS fragmentation against a relevant protein database generated from genomic sequencing[19]. Quantification can be performed as either label-free peak intensity or with more complex strategies to determine both the peptide and the protein quantities. This “bottom up” approach to proteomics provides an opportunity for proteome analysis without a priori knowledge (“unbiased”) by identifying proteins based on their proteolytic peptide forms[20]. In contrast, “top down” proteomics focuses on analysis of intact proteins, and provides a unique whole protein analysis output for investigation of relevant post-translational modifications or protein isoforms [21]. In contrast to global discovery proteomics, targeted proteomics focuses on quantification of a pre-determined set of peptide sequences from known proteins. Choice of technique depends on the research question under consideration. (Figure 2)

Figure 1:

Outline of the two major analytical mass spectrometry strategies. In bottom-up experiments the proteins are processed to peptides prior to mass spectrometry analysis, whereas in top-down proteomics, intact proteins are analyzed from simplified protein mixtures. Both approaches then use MS/MS fragmentation to obtain detailed protein information, with the potential additional step of enzymatic digestion being utilized for some samples prior to fragmentation in the top-down approach.

Figure 2:

Visual outline of global proteomics versus targeted proteomics. In discovery workflows, samples are fractionated, then constituent peptides are measured and proteins identified from a relevant protein database. Targeted proteomics focuses on quantification of a pre-determined set of peptide sequences from known proteins

Applying MS to Biomarker Discovery in TBI

The majority of investigators using MS techniques for biomarker discovery have applied them to experimental models of TBI, often in rodents[22–25]. However, some investigators have applied mass spectroscopy-focused approaches for biomarker discovery in patients with TBI. These human studies have not been restricted only to blood—tissue and urine have been informative sources of the post-TBI proteome. Subjects studied in these clinical investigations have included the full range of injury severity in TBI, though critically ill patients can offer more options to biofluids given the need for invasive monitoring.

One group of investigators examined urine samples from a small number of subacute, severe TBI patients upon entering rehabilitation compared with samples from non-traumatized matched controls to attempt to characterize a TBI-specific urinary biomarker signatures[26]. The authors chose urine as the biofluid to assay as they asserted that the renal barrier would remove small, protolyzed peptides and metabolites from circulation – decreasing the need for high abundance proteins depletion which is often necessary for blood plasma/serum samples[27]. They found 834 reproducible markers 3-fold more abundant in TBI subjects than controls. Additionally, a TBI molecular profile subset correlated with measures of TBI severity indices and neurocognitive outcomes. Despite several limitations (e.g. cohort size, individual factors such as pharmacology, therapeutic interventions, renal function etc.) this feasibility study demonstrated clinical translation of mass spectrometry to human TBI biomarker discovery [26]. The same team later expanded on their prior work in isolating a TBI urinary peptidome by correlating the trajectory of these biomarkers to functional performance outcomes in post-TBI inpatient rehabilitation[28].

Given the complexity of biomarker identification in biofluids, Abu Hamedeh and colleagues took a novel approach to global biomarker discovery through utilization of brain biopsies of structurally normal-appearing cortex, obtained while placing ICP monitors in 16 severe TBI subjects versus similarly obtained biopsies from normal pressure hydrocephalus (NPH) patients undergoing shunt placement[29]. Using LC-MS/MS analysis, they identified 316 unique proteins expressed in all tissue samples and found that the expression of 45 proteins was quantitatively different in TBI patients compared to NPH controls. Then, using high mass accuracy label free LC-MS analysis, they compared patients with diffuse axonal injury (DAI) to patients with focal injury, and found variation in expression of 20 proteins in DAI compared to both focal TBI and NPH samples. This observation suggests that DAI may produce a greater degree of alteration in cortical protein expression than focal TBI. Additionally, they determined that numerous cytoskeletal proteins had altered expression in DAI compared to focal TBI, including proteins such as glial fibrillary acidic protein (GFAP), fascin, γ-adducin. Other potential TBI biomarkers with altered expression included neuron-specific enolase (NSE), neurogranin (NRGN), cathepsin D (CTSD), and fatty acid-binding protein (FABP3) and Tau. The authors noted that in this study the elevations of pro-inflammatory cytokines and chemokines were relatively modest compared with many prior studies, which they speculated was caused by biopsy samples only in normal-appearing cortical tissues[30]. This more well-preserved tissue might have been spared from the intense inflammation of the traumatized parenchyma. Though there were a limited number of subjects studied, the authors demonstrated the feasibility and breadth of large-scale proteomic studies in humans to detect perturbations in multiple cellular pathways in brain regions radiographically distant from TBI foci and demonstrated distinct proteomic profiles in TBI subtypes[29].

Serum or plasma has also been investigated as a source of biomarker discovery. Haqqani and colleagues analyzed a small cohort of pediatric patients with severe TBI (Glasgow Coma Scale score ≤ 8), evidence of TBI on CT scan, and supported by mechanical ventilation [31]. Serum samples obtained <8 hours from injury were analyzed via the coupling of isotope-coded affinity tag (ICAT) and LC/MS-MS to evaluate protein patterns. Specifically, they utilized 2% of the serum sample from each subject in the LC-MS analysis to quantify the heavy/light ICAT ratios, then re-injected 2% of the sample into the MS analysis to further sequence the peptides. These samples were compared against controls obtained from healthy adult subjects. The researchers were able to identify upregulation of S100β in the trauma subjects, as well as 95 unique proteins that were significantly up- or down-regulated in comparison to controls. Despite notable limitations including the difficulties in comparing adult and pediatric samples, the authors concluded that the use of LC-MS and comparative ICAT protein display could be used for biomarker discovery pediatric TBI patients.

Gao and colleagues evaluated serum amyloid A (SAA), an acute phase reactant sensitive to inflammation, utilizing 2D-gel electrophoresis combined with mass spectrometry in serum from pediatric patients with suspected acute head trauma [32]. The authors had three sample pediatric populations with acute head trauma: group one with a GCS of 13–15 from which serum studies were analyzed with 2-D gel electrophoresis and Western blots; another with the GCS was 3–15 which was used to validate the western blot findings in group one; and the last group with GCS 3–15 utilizing ELISA to measure SAA. In this paper, MS was combined with the abovementioned gel electrophoresis to compare the proteins with age-matched controls, and the results were confirmed by western blot. They concluded that SAA was elevated in pediatric patients with acute head trauma, although the presence of SAA was not influenced by the severity of head trauma. Another group utilized time-of-flight mass spectrometry (TOF MS) analysis to quantify SAA in serum from a prospective, observational cohort of 120 patients with TBI within 24 hours of injury [33]. Mass spectrometry was performed on the serum of randomly chosen patients to identify protein levels that were subsequently verified using ELISA. Here, the authors were able to show that SAA1 levels correlated with clinical severity, although it could not be said that SAA1 was a marker specific to TBI alone.

Applying MS to Biomarker Quantitation in TBI

Targeted Proteomics

Targeted MS aims to quantify specific proteins of interest in biospecimens, often within blood (plasma or serum). This approach can be especially useful to quantify proteins that do not have a readily available affinity antibody for conventional immunoassays, involve particular modifications (i.e. phosphorylation), or encompass specific sequence variants that are challenging for affinity reagent development. Targeted MS employs an approach titled Selective Reaction Monitoring (SRM) in which a MS instrument platform, often a triple quadrupole, is utilized that can provide successive isolation of peptide fragments to reduce background signal and increase detection and quantitative accuracy[34]. Such an approach can be highly multiplexed[35,36], resulting in the descriptive term multiple reaction monitoring (MRM). Requirements include the addition of heavy labeled peptide standards of the sequence of interest which then provides an internal standard for the accurate quantification. As the level of sensitivity of SRM is determined by the initial LC separations and any upfront sample de-complexing approaches such as high abundant protein depletion, it is beneficial to know at what relative abundance targets are present in the plasma/serum so appropriate SRM approaches can be utilized for their detection[36,37]. Additionally, targeted MS can be combined with discovery proteomic techniques in a complementary fashion; they are not always mutually exclusive.

In an exploratory analysis of an observational TBI cohort, our group analyzed plasma samples with the aim of determining the feasibility of identifying known biomarkers by use of LC-MS in critically ill brain injured patients[38]. We performed MS analysis on plasma samples obtained < 8 hours from trauma, with (SRM) to identify angiopoietin 2, matrix metalloproteinase 9, plasminogen activator inhibitor-1, vascular adhesion molecule 1, and intercellular adhesion molecular 1. Noted limitations of this study included a limited range of detection for the current SRM platform employed which constrained the range of detection of prospective targets. Additionally, SRM is not meant to compete directly with a well-developed conventional immunoaffinity type analysis for a particular protein target, as such assays have been optimized for maximum throughput and sensitivity. MS and immunoaffinity techniques can be complementary. For instance, using the Collaborative European NeuroTrauma Effectiveness Research in Traumatic Brain Injury (CENTER-TBI) prospective observational data, Whitehouse and colleagues used traditional electrochemiluminecesence immunoassays to measure six established TBI serum biomarkers (GFAP, NFL, NSE, S100B, t-tau and UCH-L1) in 2869 patients. They aimed to determine if there was a biomarker pattern that would correlate with TBI imaging phenotypes (diffuse axonal injury, extradural hematoma, intraventricular hemorrhage intraparenchymal hemorrhage, traumatic subarachnoid, and subdural hemorrhage). The authors found that traditional biomarkers concentrations correlated to injury severity, rather than pathoanatomical entities[39]. This study highlights both the feasibility high throughput studies using conventual techniques, but also demonstrates the need for discovery using alternative methods.

From Proteins to Metabolites

Beyond analyzing the specific proteomic profile associated with TBI, it is also possible to conduct an assessment of the specific metabolic profile (e.g. metabolomics) which develops following TBI. Metabolomics offers an independent view as compared to proteomics, and includes a somewhat more comprehensive array of molecules. Whereas proteomics is concerned with the protein expression profile, metabolomics includes carbohydrates, amino acids, lipids, products of catabolisis, etc. The metabolome therefore represents an array of small molecules, any of which may potentially have functional significance [40]. Metabolomic analysis appears to be especially useful for discriminating injured patients from controls, or between diseases, as well as offering a window into the dynamics of cerebral metabolism in critically ill patients.

Mild TBI can be challenging to diagnose as there is no single objective marker of mild TBI. Fiandaca and colleagues aimed to differentiate subjects with mild TBI from age-matched controls by analyzing 4 serial (≤6 hours, 2 days, 3 days, and 7 days post-injury) plasma samples[41]. Using an initial discovery-cohort of college athletes, six-metabolite panel was developed. Specifically, 38 athletes who sustained a mild TBI were age-matched with 24 controls, and provided samples which underwent analysis via LC-MS, and subsequent confirmatory testing utilizing tandem MS/MS [41]. An initial total of 2811 prospective biomarkers were narrowed to 294 annotated metabolite species via a proprietary web-based application (MSF Metabolomics). Six of these metabolites were found to provide significant discrimination between TBI and non-TBI samples. This panel was then validated in an external cohort composed of trauma patients from multiple hospital emergency rooms with a mixed severity of injury (mild-severe). Despite major differences between the external validation cohort and experimental cohort, the biomarker panel was able to again objectively discriminate acute mild TBI cases from controls within 6 hours of injury [41].

The same team subsequently employed a similar LC-MS metabolomic platform to compare 75 subacute mild TBI subjects to 60 Parkinson’s disease (PD) subjects. Their aim was to define novel metabolic biomarkers that could distinguish mild TBI from controls and, subsequently, identify similarities that could implicate TBI in the pathogenesis of PD. They utilized global metabolomic profiling of plasma specimens using ultra-high performance LC-MS/MS to annotate and score each metabolite by abundance. They also performed targeted metabolic analysis using MS/MS to isolate and quantify a 144-lipid panel using MRM. Through both untargeted and targeted methods, they discovered and validated several plasma metabolic biomarker panels. They applied similar methods in the PD cohort and discovered that there was a reciprocal relationship in serum glutamic acid levels in TBI and PD groups; glutamic acid levels increased in both the subacute TBI and PD cohorts compared to controls[42]. While limited in sample size and temporal comparison of their cohort’s metabolome, this study highlighted feasibility of using a global discovery approach with MS in annotating new serum biomarker panels and locating potential links between TBI and other neurodegenerative diseases [42].

Targeted MS metabolomics has also yielded intriguing results in human TBI studies, particularly in critically ill subjects that offer access to biofluids in addition to blood. Orešič and colleagues used gas chromatography, time of flight MS (GC, TOF-MS) to examine at 465 metabolites (including a range of amino acids, sugar derivatives, fatty acids, sterols etc.) in 144 TBI patients[43]. They compared serum and brain micro-dialysate samples and found two medium-chain fatty acid (decanoic and octanoic acids) and sugar derivatives were strongly associated with the severity of TBI and predictive of clinical outcome. While significant limitations in study validation and generalization to polytrauma patients existed in their study, it also demonstrated the feasibility of uncovering the human “TBI metabo-type” using MS. The first and largest of these limitations was the large range of possible metabolic permutations and combinations, and the large range of possible confounding molecules present in samples. Although the authors performed data filtering for propofol and its metabolites, as well as for ibuprofen, the need to do so typifies the complexity of this method of analysis. Given the large range of potential contributions to the metabolic profile not all potential confounders could be controlled for – age, alcohol use, and diet were not assessed, for instance. Despite these limitations however, the study was able to identify multiple molecules (including fatty acids, amino acids, and carbohydrates) which were associated with the occurrence of TBI. Importantly, the magnitude of difference these metabolites displayed from baseline control samples reflected the severity of TBI. The authors therefore propose that TBI might be characterized by a specific metabolomic profile (which they refer to as a “metabotype”) and that this “metabotype” correlates with both injury severity and patient outcomes.

Indeed, metabolomics has the power to reflect the complex and dynamic perturbations of intracerebral metabolism of acute TBI patients in the intensive care setting. Another group of investigators collected paired arterial and jugular venous samples from 26 ICU patients with non-penetrating acute, severe TBI with GCS score ≤8 compared to 6 healthy volunteers, over several intervals[44]. Using LC-MS, they identified a total of 156 metabolites of interest which they further isolated to 69 unique metabolites that were present consistently across subjects and batches. They performed multiple comparison corrected univariate and multivariate analysis which demonstrated, primarily, that most arterial and jugular metabolites were reduced to ≤60%, compared to the concentration of controls, with urea cycle metabolites and proteinogenic amino acids most highly affected. They also noted that three of the metabolites differed significantly between arterial and jugular venous blood: net cerebral uptake of glucose-6-phosphate and net cerebral release of xanthine and choline. However, there was no clear relationship between metabolite levels and TBI severity or outcome. Sedatives such as propofol or midazolam did not seem to impact metabolite levels, but glycerol was significantly increased (74%) in subjects receiving pentobarbital. Pentobarbital reduces intracranial pressure through decreasing cerebral metabolic demand with resultant decrease cerebral blood flow, which may explain this association. While this study was limited by employing only targeted metabolite selection and limited sample size, the authors provided a reasonable proof-of-concept of the potential clinical impact of real-time biomarker monitoring of cerebral metabolism and the impact of interventions in the management of severe TBI patients [44].

In contrast, Thomas and colleagues completed a comprehensive, well-powered prospective metabolomics study with the goal of describing the serum metabolome and lipidome associated with TBI within 24 hours of injury, its relationship to injury severity and patient functional outcome. Their cohort included 716 individuals with varying degree of TBI from the CENTER-TBI project and they employed two mass spectrometry-based analytical methods: gas chromatography coupled to quadrupole time-of-flight MS (GC-QTOFMS) for polar metabolomic detection and liquid chromatography (LC)-QTOFMS for lipidomic detection. They identified a total of 459 metabolites (147 polar metabolites and 312 lipid metabolites), observing that lysophosphatidylcholines, were significantly associated with TBI severity while higher levels of choline-containing phospholipids portended better outcomes. The authors hypothesized that increased choline-containing phospholipids is representative of the compensative transcytosis of essential membrane lipids across the blood brain barrier, which fails in cases of severe TBI. This may have clinical implications for TBI, specifically treatment with choline phospholipids. They also observed several sugar derivatives such as myoinositol were elevated TBI patients, both proportional to severity of injury and associated with unfavorable outcomes. This investigation provides addition evidence for the utility of mass-spectroscopy based metabolomic and lipidomics in clinical TBI subjects for identifying promising diagnostic and prognostic biomarkers of TBI. These investigations may provide avenues to therapeutic interventions after sufficient validation in future studies[45].

Conclusion/Future Directions

Though early in their application to TBI, emerging MS technologies offer biomarker discovery and validation opportunities not afforded by conventional immunoassays. Combining multi-omics data (genomics, proteomics, metabolomics, lipidomics, and glycomics) provides a great potential for breakthroughs in human health and disease, especially with the development of comprehensive mapping and functional annotation of normal human proteomes within specific biofluids[46]. The proteome of a number of biofluids (including blood, urine, and cerebral tissue) have proven an informative source for the discovery of potential biomarkers. Targeted proteomics can complement global discovery methods by offering a means to quantitate specific proteins, especially when antibody-specific methods are not readily available. Metabolomic analysis appears to be especially useful in critically ill patients, offering a window into the dynamics of cerebral metabolism.

Future directions in the MS analytical space include driving towards single cell proteomic analysis. Such approaches will provide a greater resolution of molecular signals within differential cell types reflecting the pathophysiology of interest. Two of these innovations for near single cell analysis include coupling laser capture microdissection[47,48] with nanoscale-type isolations like nanoPOTS[49,50]. Advanced ion manipulation approaches such as structures for lossless ion manipulations[51] also have the potential to dramatically increase the isolation of ion/molecular populations for higher resolution separation of molecular species and peptide isoforms. Neurons and glial cells could potentially be specifically interrogated with these technologies in the future, adding several orders of magnitude of granularity to proteomic analysis. While MS techniques are relatively early in development in the neurosciences space, the potential applications to TBI and neurocritical care are immense and likely to accelerate in the coming decade.