25.1. INTRODUCTION TO METABOLOMICS

There are four biochemical components that control biological systems by serving as building blocks and as information databases: genes, transcripts, proteins, and metabolites. The study of these four components have become entire fields of biological study and have often been referred to collectively as the omics, including genomics, transcriptomics, proteomics, and metabolomics. The ability to study each of these biological components in great detail and to study the relationship between them has led to significant advances in medical discovery and understanding. The goal of medical systems biology is to integrate all biological information to understand mechanistic information about cellular events and functions that may contribute to disease propensity, development, progression, diagnosis, and/or treatment.

Having a systems perspective on human biology is desirable, where details of various system components can be integrated with increasing complexity to better understand properties of the entire system. The systems-oriented approach requires extensive and complex datasets; reliable analytical techniques; thoughtful data integration across platforms; and advanced biostatistical methods. Medical systems biology necessitates an unbiased and comprehensive approach when interpreting experimental results and biological interpretations need to be carefully explained, justified by the data, and tested on larger data sets.

Traumatic brain injury (TBI) patients would benefit from a medical systems biology understanding of the systemic dysregulation and cellular changes that follow an insult to the head. A subspecialty in the critical care environment, neurocritical care, evolved from the acceptance that recovery from the primary injury to the brain tissue is affected by systemic alterations that can result in secondary injuries to the brain. The neurological intensive care unit (ICU) has realized significant improvements in patient outcomes due to protocols to address and prevent secondary injuries and due to neurointensivist-led teamwork, both aided by modern technological advances in multimodality neuromonitoring (Elf et al., 2002, Le Roux et al., 2012, Varelas et al., 2006).

Considering the notable advances achieved through incorporating a systems-level approach to treating head injury and improving outcomes, in this review we discuss metabolomics applied to TBI. First, we will introduce metabolomics for readers not familiar with the field. Second, we summarize research on the metabolic changes following TBI to highlight what information has been translated to the clinic and what treatments exist. Finally, we discuss metabolomics techniques applied to TBI metabolism, reviewing the examples in the literature, and offering the authors’ suggestions for using NMR spectroscopy to study biofluids from head injured patients. As researchers and clinicians report and validate metabolomics findings, building a medical systems biology perspective on post-TBI metabolic dysfunction is likely to aid in informing physicians’ decisions and in integrating treatments into daily practice.

Metabolomics refers to the study of the metabolome, which has been defined as “the quantitative complement of metabolites in a biological system” (Dunn et al., 2011). A metabolome, estimated to contain thousands of compounds, is organism-specific and sample type–specific. The human serum metabolome has been reported to contain 4,229 unique compounds, detection of which involved the use of several analytic techniques, and is still not considered exhaustive (Psychogios et al., 2011). Metabolomics studies aim to discriminate pathological metabolic profiles from that of a normal physiological state and to predict class assignment based on this set of metabolite biomarkers (Baker, 2011; Holmes et al., 2008; Nicholson et al., 2012).

The field of metabolomics research consists of several investigative methods. First, there is a distinction to be made between targeted and exploratory metabolomics studies (Lenz and Wilson, 2007). In the latter, the goal is to generate a metabolomic fingerprint for each case and to use multivariate analysis to probe class-specific patterns. Generally, the focus of such studies is not to identify and quantify metabolites nor to propose mechanistic explanations of the results, but rather to predict class assignment based on the metabolomic fingerprint.

Targeted metabolomics studies aim to identify and quantify specific metabolites. These metabolites may be hypothesized to be biomarkers of disease progression or may be considered an indicator of the severity of a physiological state. Targeted metabolomics studies may use the same multivariate statistical techniques as the metabolome fingerprint-type studies, but also typically include more traditional univariate and multivariate analyses on the metabolite concentrations. Targeted studies can be targeted to a set of endogenous metabolites or can be targeted to study an exogenous substance, including labeled tracer metabolites or a pharmaceutical.

Blood plasma, blood serum, urine, and cerebrospinal fluid (CSF) have been extensively investigated in the metabolomics literature. These biofluids are readily available and are interpreted as an average representation of the surrounding tissue. Researchers working with animal models have access to tissue after sacrifice, which is considerably rarer in human studies. As the field has grown, online metabolite databases containing biological, structural, and experimental information have been developed and are a key tool for metabolomics researchers (Ulrich et al., 2008; Wishart et al., 2007).

The term metabolomics resulted from research in the 1980s and 1990s (Nicholson et al., 1999), yet the concept behind metabolomics was a focus of research for several decades prior. What distinguishes contemporary metabolomics studies from past studies on metabolic changes is the technology available for analyzing such biofluid samples and, therefore, the extent and accuracy of the metabolome quantified. In addition to the larger data set, there have also been computational and statistical advances that make the prospect of drawing meaningful conclusions from thousands of metabolites and the changes that occur between classes possible. With improvements in technology, metabolomics research has reached a level of complexity requiring a multidisciplinary team and has made providing biological rationale for the findings challenging because of data set complexity. The Institute of Medicine of the National Academies published a report on translational omics that issued recommendations for improving the overall quality of the metabolomics research and for translating these findings to the clinical setting (Committee on the Review of Omics-Based Tests for Predicting Patient Outcomes in Clinical Trials, 2012).

The use of mass spectrometry (MS)-based and nuclear magnetic resonance (NMR)-based quantification are the most common in the metabolomics literature. Both of these analytical instruments are reliable, accurate, and widely available. There are advantages and disadvantages associated with each, some of which will be briefly mentioned, and the reader is referred to a number of excellent metabolomics review articles (Dunn et al., 2011; Lenz and Wilson, 2007; Nicholson et al., 1999). Because an individual’s metabolome is highly influenced by environment and diet, population studies require a large number of subjects, and the reliability and reproducibility of these analytical techniques is key. The focus of this review is NMR-based metabolomics applied to TBI, but both analytical methods will be described. The reader is referred to extensive review articles focused on the application of MS and/or NMR to metabolomics (Dettmer et al., 2007; Zhang et al., 2010).

MS detects compounds in the picomolar concentration range that become ionized after injection into the mass spectrometer; the readout is the mass-to-charge ratio of the detectable compounds in solution. MS-based metabolomics have used gas chromatography MS and liquid chromatography MS. Preparing samples for MS analysis requires extraction of metabolites and may require derivitization, which can be a labor-intensive process. Metabolite extraction involves a series of experimental steps in which metabolite loss can occur and where additional sample-to-sample variability may be introduced. The high sensitivity of MS-based quantification makes it a powerful tool in targeted metabolomics studies. In metabolome fingerprinting studies, it is challenging to measure all compounds with the same efficiency and accuracy for technical reasons.

NMR spectroscopy is used to identify and quantify compounds in solution containing elements that are magnetic resonance–detectable (i.e., elemental isotopes that will absorb photons when placed in a magnetic field). NMR is considerably less sensitive than MS and is able to detect concentrations in the micromolar concentration range, but does not destroy the sample in the process of measurement. Application of a radiofrequency field at a known frequency and power excites the spin of the magnetic resonance–detectable isotopes. Spin is a fundamental property of elements akin to mass and charge and both the absorption and emission of radiofrequency photons is nondestructive and noninvasive. Each unique chemical structure in a molecule will resonate in the magnetic field at a specific frequency as the spins relax to equilibrium alignment with the magnetic field. The signal collected by the NMR spectrometer is then Fourier transformed into a NMR spectrum with spectral peaks at specific frequencies corresponding to the molecular structure of the compound being measured. The integrated area of the spectral peaks is proportional to the concentration of the compound. All compounds in solution above a certain concentration will be detected, unlike the variable efficiency of MS-based quantification. There is minimal sample preparation required when compared with MS. There are a number of biologically relevant isotopes that can be measured, including ¹H, ¹³C, ³¹P, and ¹⁵N. ¹H is the most abundant isotope of hydrogen (99.99%) and, because biologically relevant molecules contain hydrogen, ¹H NMR is widely used. NMR spectrometers are standard equipment in research environments and increased spectral resolution is possible due to the prevalence of high-field spectrometers with field strengths ≥400 MHz (9.4 T). High-resolution magic angle spinning spectroscopy is able to quantify metabolites in intact tissue using solid-state NMR spectrometers (Beckonert et al., 2010).

Another aspect of modern metabolomics research is application of multivariate statistical approaches. Unsupervised multivariate techniques such as principal component analysis (PCA) reduce the number of variables to a few principal components. Principal components are orthogonal to one another, are linear combinations of the original data, and can reduce hundreds of input variables to three or four. There are many NMR-based metabolomics fingerprint-type studies that use the complete NMR spectrum as the set of variables. Some metabolomics studies are designed to build a prediction model with supervised multivariate techniques, for example partial least squares (PLS) or PLS-discriminant analysis (PLS-DA) among others (Bylesjo et al., 2006). Most metabolomics studies generate a PCA model of the data to test whether the groups can be reasonably separated based on metabolic information. To build a predictive model, validation is vital and the data set is randomly separated into a larger training set and a smaller test set; the model generated from the training set is then tested on the test set.

In reality, metabolomics studies generally quantify fewer than 100 metabolites per sample. Several advances are required to achieve high-throughput quantification of the entire metabolome and to translate metabolomics to the clinical setting. The steps following data collection, including processing and statistical analyses, will be discussed later in this chapter within the context of metabolomics of TBI.

25.2. METABOLIC CHANGES AFTER TBI AND CURRENT TREATMENT RECOMMENDATIONS

TBI results from a strong force applied to the head and, although the primary injury may penetrate and/or cause physical strain on the cerebral tissue requiring surgery, the primary injury also initiates cellular metabolic changes and hemodynamic dysregulation. Experimental models of head trauma have characterized what is known about pathophysiological changes after TBI. The primary injury initiates indiscriminate excitatory neurotransmitter release (primarily glutamate and aspartate) and an increase in extracellular potassium (Katayama et al., 1990). Flux of calcium ions from intracellular stores into the cytoplasm leads to mitochondrial damage and membrane microporation (Verweij et al., 2000). The cellular response to these changes, such as operation of membrane ion pumps to restore the ionic gradient, requires large amounts of adenosine triphosphate (ATP). There is a notable increase in glucose uptake during the acute period without a concomitant increase in oxygen uptake. Increased anaerobic metabolism of glucose, termed hyperglycolysis, is followed by a prolonged period of depressed glucose metabolism with respect to oxygen metabolism. Hyperglycolysis has been interpreted as the cerebral response to the high energy demands required to respond to the metabolic disruption and can lead to increased lactate production (De Salles et al., 1987).

Several of the metabolic changes after TBI, including hyperglycemia, high cerebral glutamate, and high cerebral lactate/pyruvate ratio (LPR), have been extensively characterized in the literature (Goodman et al., 1999; Soustiel and Sviri, 2007; Vespa et al., 2003, 2005) because restoring metabolic homeostasis becomes a major focus of clinical treatment after surgical intervention. A metabolic crisis has been defined as a high LPR, high glutamate, and low glucose concentration in the extracellular space of cerebral tissue, as measured with cerebral microdialysis. Stein and colleagues demonstrated that almost three-quarters of patients with brain injury suffer from a metabolic crisis during the acute postinjury period (Stein et al., 2012). The deleterious effects of these metabolic derangements have been correlated with poor outcomes (De Salles et al., 1986; Marcoux et al., 2008; Obrist et al., 1979; Stein et al., 2012; Xu et al., 2010). In this section, we will discuss these changes, focusing on glucose because it is routinely monitored in the clinic. Additionally, we will discuss current treatment recommendations for TBI.

25.3. METABOLOMICS OF HEAD INJURY

¹H NMR-based metabolomics has been used to study diabetes, Alzheimer’s disease, cancer, stroke, and subarachnoid hemorrhage (Fan et al., 2009; Floegel et al., 2013; Jung et al., 2011; Lanza et al., 2010; Srivastava et al., 2009; Tukiainen et al., 2008). In this section, we review examples in the literature of metabolomics applied to head injury not previously mentioned. There is a large body of the head injury animal model literature that uses NMR (¹H primarily, ³¹P to study phosphate containing metabolites such as ATP, ¹³C for tracer studies) to quantify metabolites in cerebral tissue extract. Many of these studies fall into the targeted metabolomics study category previously described, using either endogenous or exogenous metabolites. These metabolomics studies are not reviewed here because our focus is to highlight modern human metabolomics studies with a systems biology-level perspective of TBI.

To the best of the authors’ knowledge, there is one example of ¹H NMR tissue analysis combined with multivariate PCA in the TBI literature. Using a lateral fluid percussion rat model of head injury, Viant et al. compared metabolite levels 1 hour postinjury in the hippocampus and cortex cerebral tissue and in the plasma (Viant et al., 2005). The authors report discernible changes in cerebral tissue metabolites linked to oxidative stress (ascorbate), excitotoxic damage (glutamate), membrane disruption (choline-containing metabolites), and neuronal injury (N-acetylaspartate), but are not able to discern these changes with plasma alone, perhaps because of systemic dilution of cerebral metabolites. There are a growing number of research studies that use in vivo magnetic resonance spectroscopy to measure cerebral metabolites after injury (Harris et al., 2012; Lin et al., 2012).

NMR-based analysis of CSF collected from TBI patients revealed several changes in the injured population when compared with controls and studied these changes over the first 2 weeks of injury (Toczylowska et al., 2006). The authors reported increased lactate and pyruvate compared to controls, with no discernible difference in the CSF LPR between injured and noninjured patients, which persisted for the 2 weeks studied. The changes observed over the 2 weeks are described, but little is offered as a biological interpretation, and these changes are not studied in the context of secondary injuries. As previously mentioned, ¹H NMR analysis of CSF from TBI patients compared with CSF from noninjured controls showed increased propylene glycol and decreased creatinine in the injured cohort (Glenn et al., 2013). Correlation analysis revealed different correlative patterns between metabolites in the two groups. Generalized linear models and multimodel inference revealed a subset of the metabolites measured to be the strongest predictors of outcome and clinical measures of oxygen metabolism and intracranial pressure.

There are two examples in the literature of metabolomics applied to traumatic injuries that exclude head injury. An outcome study on trauma patients, excluding isolated head-injured patients, using blood collected within the first 24 hours of admission was able to discriminate survivors from nonsurvivors with blood lipid biomarkers (Cohen et al., 2010). Study of the metabolomic network altered by a porcine model of trauma, hemorrhagic shock, and resuscitation supports a dual-phase metabolic response to trauma and hemorrhagic shock (Lusczek et al., 2013). It would be interesting to test these findings in a head injured population.

25.4. EXPERIMENTAL PERSPECTIVE ON NMR-BASED METABOLOMICS APPLIED TO HEAD INJURY

In this section, we highlight specific steps in metabolomics protocols where samples from head-injured patients or animals differ from other populations. Additionally, we have a few recommendations from personal experience about metabolomics studies that we hope prove to be useful to researchers.

Samples are stored frozen from the time of collection until the time of NMR sample preparation. There are a variety of techniques to prepare biological samples for NMR analysis that aim to remove proteins and other compounds that reduce the reliability of quantification, while leaving metabolites in solution. There have been concerns about metabolite loss during solvent removal, methods including drying under a flow of nitrogen gas, freeze drying, using a centrifugal concentrator. In our experience, the NMR spectral quality is significantly improved by solvent removal (and replacement with deuterium oxide: ²H₂O or D₂O) and there is no significant metabolite loss. It should be kept in mind that samples in D₂O will, over time, exchange NMR-invisible deuterium with metabolite hydrogen, so it is best to collect NMR spectra soon after sample preparation. The primary advantage of replacing the solvent with D₂O is reducing the large solvent peak before NMR data collection. To detect the low concentration metabolite peaks in solution, the solvent peak must be suppressed, but spectral techniques for water suppression are known to influence nearby peak quantification. For this reason, reducing the solvent peak before NMR data collection improves quantification.

Spectral peak location is significantly influenced by solvent, temperature, ionic strength, and pH. We have found the use of a 0.1 M phosphate-buffered solution in D₂O, in lieu of checking sample pH before NMR measurements, is advantageous to save time and to keep solvent conditions consistent. Quantification with NMR spectroscopy requires an internal standard in solution at a known concentration. We prepare the phosphate-buffered solution containing the NMR internal standard used for quantification, minimizing variability between samples.

A common internal NMR standard in metabolomics is sodium 3-trimethylsilyltetradeuteriopropionate (TSP) (Pohl and Eckle, 1969). TSP and other NMR internal standards can bind to proteins and/or protein-like macromolecules (Bell et al., 1989; Nowick et al., 2003) that lead to line broadening because of the relatively slow rotational dynamics of macromolecules in solution when compared with the small hydrocarbons of interest. A broad internal standard peak leads to inaccuracies in setting the chemical shift and in metabolite quantification. 4,4-dimethyl-4-silapentane-1-amomonium trifluoroacetate (DSA) was recently introduced, with a claim that this molecule does not interact with peptides in solution as much as other common NMR internal standards do (Nowick et al., 2003), but results in our laboratory do not support this claim.

We recommend ¹³C-labeled formate as an internal standard for aqueous solutions. Although formate is an endogenous metabolite with a sharp, single spectral peak at 8.44 ppm, ¹³C-labeled formate does not overlap with endogenous formate because the universal ¹³C-labeling splits the signal in ¹H NMR spectra (¹³C satellite peaks located at 8.25 and 8.64 ppm). Unpublished research from our laboratory shows formate does not suffer from line broadening, even in very protein-rich biofluids such as plasma.

Blood serum and blood plasma contain larger quantities of proteins than urine and CSF. NMR spectra of protein-rich biofluids are characterized by broad spectral peaks from protein and lipid macromolecules. Overlap between the sharp metabolite resonances of interest and the broad macromolecule resonances makes quantification difficult and inaccurate, may hide some low concentration peaks under the macromolecule peak envelope, and makes true baseline correction impossible.

Common practice in metabolomics is to change the method of NMR acquisition for protein-rich biofluids; instead of using a simple excitation-acquisition pulse sequence (e.g., zgpr on Bruker spectrometers), researchers use the Carr-Purcell-Meiboom-Gill (CPMG) pulse sequence, which includes a relaxation delay between excitation and acquisition (cpmgpr1d on Bruker spectrometers). The idea behind the relaxation filter is to remove the broad protein peaks without significantly affecting metabolite spectral peaks; the slow rotational dynamics of large macromolecules make their NMR signal decay to zero at a faster rate than small, low-molecular-weight molecules.

Modifying the pulse sequence is an extremely straightforward solution to this problem, but we would caution our readers to two potential problems. First, the combination of TSP or DSA internal standards with CPMG spectral acquisition of protein-rich biofluids leads to extremely inaccurate quantification. The broad TSP or DSA peak in these solutions is a sign that the relaxation rate of the internal standard is increased above normal. CPMG spectra acquired under these conditions are characterized by a TSP or DSA peak that is nearly zero and will lead to vast overestimation of metabolite concentrations in solution. Figure 25.1 plots three spectra acquired on a plasma NMR sample and, although there is a noticeable decrease in the DSA peak at 0 ppm in the CPMG spectra, there is very little change in the ¹³C-labeled formate peaks at 8.25 and 8.64 ppm. Additionally, the broad macromolecule peaks overlapping metabolite peaks between 1.5 and 0.8 ppm are reduced as the CPMG relaxation delay increases. The ¹³C-labeled formate spectral peaks are not changed significantly by acquisition with the CPMG pulse sequence because the relaxation rate of formate is closer to the relaxation rates of metabolites in solution. Second, CSF from TBI patients can contain a relatively high level of blood from cerebral hemorrhage, from disruption of the blood–CSF barrier, from clearance of blood from the brain, and/or from contamination during sample collection through the ventricular drain. Metabolomics of CSF from TBI patients potentially requires some of the techniques typically applied to blood plasma or serum.

FIGURE 25.1

NMR spectra acquired on a sample prepared from 230 µL blood plasma in 0.1 M phosphate-buffered D₂O solution containing 0.2 mM DSA and 0.1 mM ¹³C-labeled formate. The three spectra were acquired with a simple excitation-acquisition pulse sequence (more...)

An important aspect of the multivariate methods applied in metabolomics, including PCA and/or PLS-DA, is that the input variables have a normal distribution and are homoscedastic. Data transformation and scaling are aspects of multivariate analysis that should be justified before application. In the authors’ experience, metabolite variable distributions are highly non-normal as determined by the Shapiro P test of normality, but discussion in the literature of how researchers address data’s non-normality is rare.

Because of the complexity of human biology and of experimental designs, there is a strong danger of bias in metabolomics studies. It is human nature to seek a pattern when there is none and to pay close attention to aspects of the experiments that directly relate to hypotheses. Even though researchers are trained take steps to avoid bias and false discoveries, it may still occur unknowingly. There was a metabolomics study published in 2002 that reported ¹H NMR metabolomics analysis of human serum accurately diagnosed coronary heart disease (Brindle et al., 2002). A PLS-DA model generated from 80% of the data (the training set) predicted, with a sensitivity of 92% and a specificity of 93%, the presence of coronary heart disease in the test set; the spectral regions that led to class separation correspond to lipids. Subsequent analysis of blood serum identified confounding factors, such as gender and drug treatments, that were not considered in the original publication and that significantly reduced the predictive value of ¹H NMR-metabolomics analysis of blood serum in diagnosing coronary heart disease (Kirschenlohr et al., 2006). Avoiding bias in scientific experiments is important, as all scientists know, but because of the interdisciplinary nature and complexity of the omics fields, is worth repeating. There is an excellent opinion piece on the threat of bias to biomarker discovery in cancer research that is applicable to all metabolomics and head injury research (Ransohoff, 2005). We also highly recommend a review article on metabolomics statistical strategies to avoid false discovery (Broadhurst and Kell, 2006).

25.5. CONCLUSION

Prevention of secondary brain injury after trauma is the single most important task of the treating physician. Clinical management of acute, aberrant metabolic changes most certainly represents the cornerstone of what metabolomics has to offer TBI research. Therefore, this endeavor is likely to become one of the most important and relevant forms of post-TBI metabolism investigation.

Metabolic crisis has been associated with poor outcome, yet few treatment options exist that address this significant problem. As a result, exhaustive metabolic analyses, followed by exhaustive validation, will need to be conducted to further delineate the complex metabolic composition of the injured brain and to deepen our understanding of the biochemical pathways involved. It is our belief that only under these conditions will treatment options and interventions emerge. Ultimately, measuring and monitoring postinjury metabolic profiles could assist not only in the treatment of metabolic crisis, but also in predicting an individual’s clinical parameters that may deteriorate in the days after injury and in determining long-term prognosis. Metabolomics represents one of the most important tools in the arsenal of the modern-day TBI scientist.

REFERENCES

1. Ackermann R, Lear J. Glycolysis-induced discordance between glucose metabolic rates measured with radiolabeled fluorodeoxyglucose and glucose. J. Cerebr. Blood Flow Metab. 1989;9:774–785. [PubMed: 2584274]
2. Andersen B, Marmarou A. Isolated stimulation of glycolysis following traumatic brain injury. In: Hoff J.T, Betz A.L, editors. Intracranial Pressure. VII. Springer-Verlag; Berlin, Germany: 1989. pp. 575–580. In.
3. Baker M. Metabolomics: From small molecules to big ideas. Nat. Methods. 2011;8:117–121.
4. Bartnik B, Hovda D, Lee P. Glucose metabolism after traumatic brain injury: Estimation of pyruvate carboxylase and pyruvate dehydrogenase flux by mass isotopomer analysis. J. Neurotrauma. 2007;24:181–194. [PubMed: 17263682]
5. Bartnik B, Sutton R, Fukushima M, Harris N, Hovda D, Lee S. Upregulation of pentose phosphate pathway and preservation of tricarboxylic acid cycle flux after experimental brain injury. J. Neurotrauma. 2005;22:1052–1065. [PubMed: 16238483]
6. Beckonert O, Coen M, Keun H, Wang Y, Ebbels T, Holmes E. et al. High-resolution magic-angle-spinning NMR spectroscopy for metabolic profiling of intact tissues. Nat. Protoc. 2010;5:1019–1032. [PubMed: 20539278]
7. Bell J, Brown J, Sadler P. NMR studies of body fluids. NMR Biomed. 1989;2:245–256. [PubMed: 2701808]
8. Bergsneider M, Hovda D, McArthur D, Etchepare M, Huang S.-C, Sehati N. et al. Metabolic recovery following human traumatic brain injury based on FDG-PET: Time course and relationship to neurological disability. J. Head Trauma Rehabil. 2001;16:135–148. [PubMed: 11275575]
9. Bergsneider M, Hovda D, Shalmon E, Kelly D, Vespa P, Martin N. et al. Cerebral hyperglycolysis following severe traumatic brain injury in humans: A positron emission tomography study. J. Neurosurg. 1997;86:241–251. [PubMed: 9010426]
10. Bor-Seng-Shu E, de Lima Oliveira M, Teixeira M. Traumatic brain injury and metabolism. J. Neurosurg. 2010;112:1351–1353. [PubMed: 20415524]
11. Brindle J, Antti H, Holmes E, Tranter G, Nicholson J, Bethell H. et al. Rapid and noninvasive diagnosis of the presence and severity of coronary heart disease using ¹H-NMR-based metabonomics. Nat. Med. 2002;8:1439–1444. [PubMed: 12447357]
12. Broadhurst D, Kell D. Statistical strategies for avoiding false discoveries in metabolomics and related experiments. Metabolomics. 2006;2:171–196.
13. Brunkhorst F, Engel C, Bloos F, Meier-Hellmann A, Ragaller M, Weiler N. et al. Intensive insulin therapy and pentastarch resuscitation in severe sepsis. N. Engl. J. Med. 2008;358:125–139. [PubMed: 18184958]
14. Bylesjo M, Rantalaineen M, Cloarec O, Nicholson J, Holmes E, Trygg J. OPLS discriminant analysis: Combining the strength of PLS-DA and SIMCA classification. J. Chemometrics. 2006;20:341–351.
15. Carre E, Ogier M, Boret H, Montcriol A, Bourdon L, Risso J.-J. Metabolic crisis in severely head-injured patients: Is ischemia just the tip of the iceberg? ¹⁴⁶Front. Neurol. 2013;11 [PMC free article: PMC3795329] [PubMed: 24130548]
16. Cochran A, Scaife E, Hansen K, Downey E. Hyperglycemia and outcomes from pediatric traumatic brain injury. J Trauma. 2003;55:1035–1038. [PubMed: 14676647]
17. Cohen M, Serkova N, Wiener-Kronish J, Pittet J.-F, Niemann C. ¹H-NMR-based metabolic signatures of clinical outcomes in trauma patients—beyond lactate and base deficit. J. Trauma. 2010;69:31–40. [PubMed: 20622576]
18. Committee on the Review of Omics-Based Tests for Predicting Patient Outcomes in Clinical Trials. The National Academies Press; Washington, DC: Evolution of Translational Omics: Lessons Learned and the Path Forward. 2012 [PubMed: 24872966]
19. De Salles A, Kontos H, Becker D, Yang M, Ward J, Moulton R. et al. Prognostic significance of ventricular CSF lactic acidosis in severe head injury. J. Neurosurg. 1986;65:615–624. [PubMed: 3772448]
20. De Salles A, Muizelaar J, Young H. Hyperglycemia, cerebrospinal fluid lactic acidosis, and cerebral blood flow in severely head-injured patients. Neurosurgery. 1987;21:45–50. [PubMed: 3614603]
21. Dettmer K, Aronov P, Hammock B. Mass spectrometry-based metabolomics. Mass Spectrom. Rev. 2007;26:51–78. [PMC free article: PMC1904337] [PubMed: 16921475]
22. Dunn W, Broadhurst D, Atherton H, Goodacre R, Griffin J. Systems level studies of mammalian metabolomes: The roles of mass spectrometry and nuclear magnetic resonance spectroscopy. Chem. Soc. Rev. 2011;40:387–426. [PubMed: 20717559]
23. Dusick J, Glenn T, Lee W, Vespa P, Kelly D, Lee S. et al. Increased pentose phosphate pathway flux after clinical traumatic brain injury: A [1,2-¹³C₂]glucose labeling study in humans. J. Cerebr. Blood Flow Metab. 2007;27:1593–1602. [PubMed: 17293841]
24. Fan T, Lane A, Higashi R, Farag M, Gao H, Bousamra M. et al. Altered regulation of metabolic pathways in human lung cancer discerned by ¹³C stable isotope-resolved metabolomics (SIRM). Mol. Cancer. 2009;8:41. [PMC free article: PMC2717907] [PubMed: 19558692]
25. Finfer S, Chittock D, Su S, Blair D, Foster D, Dhingra V. et al. Intensive versus conventional glucose control in critically ill patients. N. Engl. J. Med. 2009;360:1283–1297. [PubMed: 19318384]
26. Floegel A, Stefan N, Yu Z, Muhlenbruch K, Drogan D, Joost H.-G. et al. Identification of serum metabolites associated with risk of type 2 diabetes using a targeted metabolomic approach. Diabetes. 2013;62:639–648. [PMC free article: PMC3554384] [PubMed: 23043162]
27. Gallagher C, Carpenter K, Grice P, Howe D, Mason A, Timofeev I. et al. The human brain utilizes lactate via the tricarboxylic acid cycle: A ¹³C-labelled microdialysis and high-resolution nuclear magnetic resonance study. Brain. 2009;132:2839–2849. [PubMed: 19700417]
28. Glenn T, Hirt D, Mendez G, McArthur D, Sturtevant R, Wolahan S. et al. Metabolomic analysis of cerebral spinal fluid from patients with severe brain injury. Acta Neurochir. Suppl. 2013;118:115–119. [PubMed: 23564115]
29. Glenn T, Kelly D, Boscardin W, McArthur D, Vespa P, Oertel M. et al. Energy dysfunction as a predictor of outcome after moderate or severe head injury: Indices of oxygen, glucose, and lactate metabolism. J. Cerebr. Blood Flow Metab. 2003;23:1239–1250. [PubMed: 14526234]
30. Goodman J, Valadka A, Gopinath S, Uzura M, Robertson C. Extracellular lactate and glucose alterations in the brain after head injury measured by microdialysis. Crit. Care Med. 1999;27:1965–1973. [PubMed: 10507626]
31. Green D, O’Phelan K, Bassin S, Chang C, Stern T, Asai S. Intensive versus conventional insulin therapy in critically ill neurologic patients. Neurocrit. Care. 2010;13:299–306. [PubMed: 20697836]
32. Harris J, Yeh H.-W, Choi I.-Y, Lee P, Lee P, Berman N. et al. Altered neurochemical profile after traumatic brain injury: ¹H-MRS biomarkers of pathological mechanisms. J. Cerebr. Blood Flow Metab. 2012;32:2122–2134. [PMC free article: PMC3519407] [PubMed: 22892723]
33. Hillered L, Enblad P. Nonischemic energy metabolic crisis in acute brain injury. Crit. Care Med. 2008;36:2952–2953. [PubMed: 18812809]
34. Hillered L, Persson L, Nilsson P, Ronne-Engstrom E, Enblad P. Continuous monitoring of cerebral metabolism in traumatic brain injury: A focus on cerebral microdialysis. Curr. Opin. Crit. Care. 2006;12:112–118. [PubMed: 16543785]
35. Holmes E, Wilson I, Nicholson J. Metabolic phenotyping in health and disease. Cell. 2008;134:714–717. [PubMed: 18775301]
36. Jung J, Lee H.-S, Kang D.-G, Kim N, Cha M, Bang O.-S. et al. ¹H-NMR-based metabolomics study of cerebral infarction. Stroke. 2011;42:1282–1288. [PubMed: 21474802]
37. Katayama Y, Becker D, Tamura T, Hovda D. Massive increases in extracellular potassium and the indiscriminate release of glutamate following concussive brain injury. J. Neurosurg. 1990;73:889–900. [PubMed: 1977896]
38. Kirschenlohr H, Griffin J, Clarke S, Rhydwen R, Grace A, Schofield P. et al. Proton NMR analysis of plasma is a weak predictor of coronary artery disease. Nat. Med. 2006;12:705–710. [PubMed: 16732278]
39. Lam A, Winn H, Cullen B, Sundling N. Hyperglycemia and neurological outcome in patients with head injury. J. Neurosurg. 1991;75:545–551. [PubMed: 1885972]
40. Lanza I, Zhang S, Ward L, Karakelides H, Raftery D, Nair K. Quantitative metabolomics by 1H-NMR and LC-MS/MS confirms altered metabolic pathways in diabetes. PLoS ONE. 2010;5 e10538. [PMC free article: PMC2866659] [PubMed: 20479934]
41. Lenz E, Wilson I. Analytical strategies in metabonomics. J. Proteome Res. 2007;6:443–458. [PubMed: 17269702]
42. Lin A, Liao H, Merugumala S, Prabhu S, Meehan W, Ross B. Metabolic imaging of mild traumatic brain injury. Brain Imaging Behav. 2012;6:208–223. [PubMed: 22684770]
43. Lindquist J, LeRoy G. Studies of cerebral oxygen consumption following experimental head injury. Surg. Gynecol. Obstet. 1942;75:28–33.
44. Lusczek E, Lexcen D, Witowski N, Mulier K, Beilman G. Urinary metabolic network analysis in trauma, hemorrhagic shock, and resuscitation. Metabolomics. 2013;9:223–235.
45. Marcoux J, McArthur D, Miller C, Glenn T, Villablanca P, Martin N. et al. Persistent metabolic crisis as measured by elevated cerebral microdialysis lactate-pyruvate ratio predicts chronic frontal lobe brain atrophy after traumatic brain injury. Crit. Care Med. 2008;36:2871–2877. [PubMed: 18766106]
46. Meier R, Bechir S, ad Ludwig M, Sommerfeld J, Keel M, Steiger P, Stocker R. et al. Differential temporal profile of lowered blood glucose levels (3.5 to 6.5 mmol/l versus 5 to 8 mmol/l) in patients with severe traumatic brain injury. Crit. Care. 2008 doi:10.1186/cc6974. [PMC free article: PMC2575586] [PubMed: 18680584]
47. Meierhans R, Bechir M, Ludwig S, Sommerfeld J, Brandi G, Haberthur C. et al. Brain metabolism is significantly impaired at blood glucose below 6 mM and brain glucose below 1 mM in patients with severe traumatic brain injury. Crit. Care. 2010 doi:10.1186/cc8869. [PMC free article: PMC2875528] [PubMed: 20141631]
48. Nicholson J, Holmes E, Kinross J, Darzi A, Takats Z, Lindon J. Metabolic phenotyping in clinical and surgical environments. Nature. 2012;491:384–392. [PubMed: 23151581]
49. Nicholson J, Lindon J, Holmes E. ‘Metabonomics’: Understanding the metabolic responses of living systems to pathophysiological stimuli via multivariate statistical analysis of biological NMR spectroscopic data. Xenobiotica. 1999;29:1181–1189. [PubMed: 10598751]
50. Nowick J, Khakshoor O, Hashemzadeh M, Brower J. DSA: A new internal standard for NMR studies in aqueous solution. Org. Lett. 2003;5:3511–3513. [PubMed: 12967312]
51. Obrist W, Gennarelli T, Segawa H, Dolinskas C, Langfitt T. Relation of cerebral blood flow to neurological status and outcome in head-injured patients. J. Neurosurg. 1979;51:292–300. [PubMed: 469577]
52. Pohl L, Eckle M. Sodium 3-trimethylsilyltetradeuteriopropionate, a new water-soluble standard for ¹H-NMR. Angew. Chem. Int. Ed. 1969;5:381.
53. Preiser J, Devos P, Ruiz-Santana S, Melot C, Annane D, Groeneveld J. et al. A prospective randomised multi-centre controlled trial on tight glucose control by intensive insulin therapy in adult intensive care units: The Glucontrol study. Intensive Care Med. 2009;35:1738–1748. [PubMed: 19636533]
54. Psychogios N, Hau D, Guo A, Mandal R, Bouatra S, Sinelnikov I. et al. The human serum metabolome. PLoS ONE. 2011;6 e16957. [PMC free article: PMC3040193] [PubMed: 21359215]
55. Ransohoff D. Bias as a threat to the validity of cancer molecular-marker research. Nat. Rev. 2005;5:142–149. [PubMed: 15685197]
56. Rovlias A, Kotsou S. The influence of hyperglycemia on neurological outcome in patients with severe head injury. Neurosurgery. 2000;46:335–342. [PubMed: 10690722]
57. Soustiel J, Larisch S. Mitochondrial damage: A target for new therapeutic horizons. Neurotherapeutics. 2010;7:13–21. [PMC free article: PMC5084108] [PubMed: 20129493]
58. Soustiel J, Sviri G. Monitoring of cerebral metabolism: Non-ischemic impairment of oxidative metabolism following severe traumatic brain injury. Neuro. Res. 2007;29:654–660. [PubMed: 18173902]
59. Srivastava N, Pradhan S, Gowda G, Kumar R. In vitro, high-resolution ¹H and ³¹P NMR based analysis of the lipid components in the tissue, serum, and CSF of the patients with primary brain tumors: One possible diagnostic view. NMR Biomed. 2009;23:113–122. [PubMed: 19774696]
60. Stein N, McArthur D, Etchepare M, Vespa P. Early cerebral metabolic crisis after TBI influences outcome despite adequate hemodynamic resuscitation. Neurocrit. Care. 2012;17:49–57. [PubMed: 22528283]
61. Toczylowska B, Chalimoniuk M, Wodowska M, Mayzner-Zawadzka E. Changes in concentration of cerebrospinal fluid components in patients with traumatic brain injury. Brain Res. 2006;1104:183–189. [PubMed: 16793028]
62. Tukiainen T, Tynkkynen T, Makinen V.-P, Jylanki P, Kangas A, Hokkanen J. et al. A multi-metabolite analysis of serum by ¹H NMR spectroscopy: Early systemic signs of Alzheimer’s disease. Biochem. Biophys. Res. Commun. 2008;375:356–361. [PubMed: 18700135]
63. Ulrich E, Akutsu H, Doreleijers J, Harano Y, Ioannidis Y, Lin J. et al. BioMagResBank. Nucleic Acids Res. 2008;36:D402–D408. [PMC free article: PMC2238925] [PubMed: 17984079]
64. Unterberg A, Stover J, Kress B, Kiening K. Edema and brain trauma. Neuroscience. 2004;129:1021–1029. [PubMed: 15561417]
65. Van den Berghe G, Schoonheydt K, Becx P, Bruyninckx F, Wouters P. Insulin therapy protects the central and peripheral nervous system of intensive care patients. Neurology. 2005;64:1348–1353. [PubMed: 15851721]
66. Van den Berghe G, Wilmer A, Hermans G, Meersseman W, Wouters P, Milants I. et al. Intensive insulin therapy in the medical ICU. N. Engl. J. Med. 2006;354:449–461. [PubMed: 16452557]
67. Van den Berghe G, Wouters P, Weekers F, Verwaest C, Bruyninckx F. et al. Intensive insulin therapy in critically ill patients. N. Engl. J. Med. 2001;345:1359–1367. [PubMed: 11794168]
68. Verweij B, Muizelaar J, Vinas F, Peterson P, Xiong Y, Lee C. Impaired cerebral mitochondrial function after traumatic brain injury in humans. J. Neurosurg. 2000;93:815–820. [PubMed: 11059663]
69. Vespa P, Bergsneider M, Hattori N, Wu H.-M, Huang S.-C, Martin N. et al. Metabolic crisis without brain ischemia is common after traumatic brain injury: A combined microdialysis and positron emission tomography study. J. Cerebr. Blood Flow Metab. 2005;25:763–774. [PMC free article: PMC4347944] [PubMed: 15716852]
70. Vespa P, Boonyaputthikul R, McArthur D, Miller C, Etchepare M, Bergsneider M. et al. Intensive insulin therapy reduces microdialysis glucose values without altering glucose utilization or improving the lactate/pyruvate ratio after traumatic brain injury. Crit. Care Med. 2006;34:850–856. [PubMed: 16505665]
71. Vespa P, McArthur D, O’Phelan K, Glenn T, Etchepare M, Kelly D. et al. Persistently low extracellular glucose correlates with poor outcome 6 months after human traumatic brain injury despite a lack of increased lactate: A microdialysis study. J. Cerebr. Blood Flow Metab. 2003;23:865–877. [PubMed: 12843790]
72. Vespa P, McArthur D, Huang N, Stein S.-C, Shao W, Filippou M. et al. Tight glycemic control increases metabolic distress in traumatic brain injury: A randomized controlled within-subjects trial. Crit. Care Med. 2012;40:1923–1929. [PubMed: 22610193]
73. Vespa P, Miller C, McArthur D, Eliseo M, Etchepare M, Hirt D. et al. Nonconvulsive electrographic seizures after traumatic brain injury result in a delayed, prolonged increase in intracranial pressure and metabolic crisis. Crit. Care Med. 2007;35:2830–2836. [PMC free article: PMC4347945] [PubMed: 18074483]
74. Vespa P, O’Phelan K, McArthur D, Miller C, Eliseo M, Hirt D. et al. Pericontusional brain tissue exhibits persistent elevation of lactate/pyruvate ratio independent of cerebral perfusion pressure. Crit. Care Med. 2007;35:1153–1160. [PubMed: 17334254]
75. Viant M, Lyeth B, Miller M, Berman R. An NMR metabolomic investigation of early metabolic disturbances following TBI in a mammalian model. NMR Biomed. 2005;18:507–516. [PubMed: 16177961]
76. Wishart D, Tzur D, Knox C, Eisner R, Guo A, Young N. et al. HMDB: The Human Metabolome Database. Nucleic Acids Res. 2007;35:D521–D526. [PMC free article: PMC1899095] [PubMed: 17202168]
77. Xu Y, McArthur D, Alger J, Etchepare M, Hovda D, Glenn T, Huang S, Dinov I, Vespa P. Early nonischemic oxidative metabolic dysfunction leads to chronic brain atrophy in traumatic brain injury. J. Cerebr. Blood Flow Metab. 2010;30:883–894. [PMC free article: PMC2949156] [PubMed: 20029449]
78. Young B, Ott L, Dempsey R, Haack D, Tibbs P. Relationship between admission hyperglycemia and neurologic outcome of severely brain-injured patients. Ann. Surg. 1989;210:466–472. [PMC free article: PMC1357925] [PubMed: 2679455]
79. Zhang S, Gowda G, Ye T, Raftery D. Advances in NMR-based biofluid analysis and metabolite profiling. Analyst. 2010;135:1490–1498. [PMC free article: PMC4720135] [PubMed: 20379603]