Biomarker discovery in human cerebrospinal fluid: the need for integrative metabolome and proteome databases

The number of metabolites identified in human cerebrospinal fluid (CSF) has steadily increased over the past 5 years, and in this issue of Genome Medicine David Wishart and colleagues provide a comprehensive update that brings the number of metabolites listed in the CSF metabolome database to 476 compounds. There is now a need for an integrative metabolome-proteome CSF database to maximize the impact of this achievement in biomedical research. Only by such efforts can we hope to unravel the complexity of molecular pathophysiological processes.


Technological advances expand the spectrum of CSF metabolites
A new study from David Wishart and coworkers, published in Genome Medicine, represents an important work in the field of biomarker discovery [2]. The authors describe how continuing advances in analytical tech nologies have led to the discovery of many new CSF metabolites, allowing expansion of the CSF metabolome database. They applied five different metabolomic plat forms to characterize multiple CSF samples, and this resulted in identification of new constituent metabolites and an increase in size of the database by approximately 50%. The authors now give an updated CSF metabolome database that contains detailed information of 476 compounds. Furthermore, the authors have carried out an extensive literature review for additional information on these compounds, including their concentrations and disease associations. The newly identified molecules consist of a number of metabolites (6 acylcarnitines, 13 amino acids, hexose, 42 phosphatidylcholines, 2 lyso phosphatidylcholines and 14 sphingolipids) that should aid neurological studies involving changes in energy metabolism. In addition, Wishart et al. also identified 37 metal ions, which could be useful in studies of neuro degenerative diseases such as Alzheimer's disease, since alterations in metal ions are known to occur in this condition [3].

Integrating metabolome and proteome profiles of CSF
Despite the comprehensiveness of this updated CSF meta bolome database, there is still a need for continued developments, including integration with a CSF proteo mics database. The most extensive CSF proteome charac terized thus far contains 2,630 proteins [4]. However, integration of these databases will require the application of sophisticated bioinformatic approaches. Also, studies involving changes in CSF metabolites and proteins may require analyses using a single platform to rule out artifacts found as a result of crossplatform comparisons. This is important considering that proteins, metabolites

Abstract
The number of metabolites identified in human cerebrospinal fluid (CSF) has steadily increased over the past 5 years, and in this issue of Genome Medicine David Wishart and colleagues provide a comprehensive update that brings the number of metabolites listed in the CSF metabolome database to 476 compounds. There is now a need for an integrative metabolomeproteome CSF database to maximize the impact of this achievement in biomedical research. Only by such efforts can we hope to unravel the complexity of molecular pathophysiological processes.

© 2010 BioMed Central Ltd
Biomarker discovery in human cerebrospinal fluid: the need for integrative metabolome and proteome databases Emanuel Schwarz 1 , E Fuller Torrey 2 , Paul C Guest 1 * and Sabine Bahn 1,3 *

R E S E A R C H H I G H L I G H T
*Correspondence: pg110@cam.ac.uk; sb209@cam.ac.uk 1 Institute of Biotechnology, University of Cambridge, Cambridge CB2 1QT, UK Full list of author information is available at the end of the article and other molecules are interactive as components of the same biological networks and this is manifested at multiple levels in a systems biology manner. For example, protein hormones, including insulin and VGF, regulate the levels of metabolites such as glucose, adreno corticotropic hormone regulates production and release of the steroid hormone cortisol, and growth factor proteins such as plateletderived growth factor can affect the action of dopaminergic and serotonergic neuro trans mitters. As biomedical research evolves from traditional clinical and biological investigations to incorporate multiomic technologies, integration of the resulting data has emerged as a critical next stage.
In line with this, several groups, including our own, have investigated both metabolic and proteomic profiles of CSF samples in normal and disease states. For example, Blanchet and colleagues [5] have shown that a fusion of proteome and metabolome data leads to higher predictive accuracy in a rat model of autoimmune encephalomyelitis. Interestingly, the molecular signature comprises transporter proteins such as hemopexin, serum albumin and serotransferrin, along with metabo lites such as succinate, pantothenate and glutamine, which are involved in multiple complex biological path ways. An alternative example comes from our own group, by identification of a metabolome profile comprising glucose, acetate, alanine and glutamine that could distin guish patients with schizophrenia from control patients, with high accuracy [6]. As a complement to this meta bolome study, we also profiled proteins and peptides in the same CSF samples [7]. The key alterations identified were increased levels of a VGFderived peptide and decreased levels of the transport protein transthyretin. The two studies together provide a more complete picture of the changes seen in patients with schizophrenia as VGF is known to affect the levels of energyrelated metabolites such as glucose, and transthyretin directly interacts with and is responsible for the transport of metabolites such as the thyroid hormone thyroxine (T 4 ) and the vitaminArelated molecule retinol.
As an extension, we are now profiling CSF from patients with schizophrenia and controls using multiplex immunoassay profiling to simultaneously measure multiple proteins and metabolites (Figure 1). This multi plex platform will allow researchers to target a broader combination of metabolites. As all of these proteins interact with small molecules to exert their functions, it is clear that integration of 'omic' platforms is required for a better understanding of disease processes.

Future perspectives in CSF metabolomics and beyond
Wishart and colleagues have made a significant contri bution to our ability to unravel disease processes through their use of multiple platforms to increase the size of the metabolome database [2]. However, future profiling studies will require incorporation of assays for proteins and metabolites into single platforms to identify compo nents that are altered in disease. This is because cross platform comparisons are useful for cataloguing compo nents but not for reliable quantitative studies. It is clear from the metabolome and proteome studies described above that small molecules and proteins are highly inter active in bringing about physiological effects in complex biological systems. Therefore, a more complete under standing of diseases and other biological effects requires a massive integration of technologies and statistical methods. However, there is now reason for optimism  that further technological and interdisciplinary advance ments will overcome current limitations in the field to help usher biomarkers fully into the 21st century. This can be achieved either by integrating the analyses using a single platform, such as the multiplex immunoassay method described above, or by using sophisticated bio informatic and biostatistical methods to integrate the metabolite and proteomic data acquired using different platforms.