Characterization of Proteome of Human Cerebrospinal Fluid

https://doi.org/10.1016/S0074-7742(06)73002-1Get rights and content

Publisher Summary

Human cerebrospinal fluid (CSF) is an ideal source for identifying biomarkers for neurodegenerative diseases such as Alzheimer's disease (AD), Parkinson's disease (PD), and dementia with Lewy bodies (DLB). Proteomics has been used to analyze CSF in order to discover disease-associated proteins and elucidate the basic molecular mechanisms that either cause, or result from, central nervous system disorders. To identify as many CSF proteins in well-characterized healthy young subjects as possible, sodium dodecyl sulfate- polyacrylamide gel electrophoresis (SDS-PAGE) was used to prefractionate the CSF proteins before further separation by multidimensional liquid chromatography and analyzed with LCQ or LTQ-FT mass spectrometry (MS). LCQ-MS/ MS identified 466 proteins and LTQ-FTMS/MS identified 608 proteins, which was 30% over those identified by LCQ-MS/MS. Issues related to sample preparation, proteomic instrumentation, and database search are discussed further in the context of characterization of human CSF proteome.

Introduction

Human cerebrospinal fluid (CSF) circulates within the ventricles of the brain and surrounds the brain in the subarachnoid space (Blennow 1993a, Blennow 1993b, Blennow 1993c). Secretion and absorption of CSF is closely regulated with an average circulating volume of 125–150 ml in an adult. Several reasons make human CSF an ideal source for identifying biomarkers for neurodegenerative diseases such as Alzheimer's disease (AD), Parkinson's disease (PD), and dementia with Lewy bodies (DLB). These include CSF's close proximity to the site of pathology, its high availability, and the advantage of minimal ambiguities that are commonly encountered in experimental models. However, translational research using CSF to identify biomarkers in these increasingly common diseases have not been successful to date, largely due to, in our opinion, two major grounds: (1) the underlying pathogenic events are too complex to be accurately reflected in a single molecule or small group of molecules, and (2) the lack of a complete understanding of the composition of human CSF in normal persons.

Recently, we utilized discovery based proteomics (Link et al., 1999)—the identification of proteins in a given sample—to address both difficulties. The three fundamental steps in discovery‐based proteomics are: (1) protein harvest or isolation; (2) protein identification by mass spectrometry (MS), typically via tandem MS (MS/MS); and (3) data mining using bioinformatics tools and databases. In the first step, proteins are often separated by three different methods: two‐dimensional gel electrophoresis (2‐D gel) (Hochstrasser et al., 2002), liquid chromatography (LC) (Aebersold 2001, Washburn 2001), and more recently, “protein chips” or activated surfaces that bind proteins based on chemical characteristics (Yip and Lomas, 2002). For the second step, MS or MS/MS identification of proteins also involves three basic steps: ionization, ion separation, and detection (Gross, 2004). Most 2‐D gel and protein chip proteomic protocols are coupled with matrix‐assisted laser desorption ionization (MALDI). One type of protein chip‐coupled MALDI is called surface enhanced laser desorption/ionization (SELDI). In contrast, the LC‐based platforms are usually interfaced to MS with electronspray ionization (ESI) (Chait and Kent, 1992). ESI‐MS has emerged as one of the premier methods for examining biological molecules in solution, as it permits the direct analysis of nonvolatile compounds, such as peptides, proteins, glycoproteins, phospholipids, glycolipids, and complex carbohydrates in liquid solutions, as intact molecules without derivatization or digestion. Furthermore, ESI can be interfaced with LC readily while maintaining high sensitivity (subfemtomole). Our study utilized two popular, commercially available ESI‐ion trap MS systems, the LCQ and LTQ proteomic stations made by ThermoElectron.

The general method we used is referred to as shotgun proteomics, which includes methods like Multidimensional Protein Identification Technology (MudPIT) and Isotope Coded Affinity Tag (ICAT), both of which use multidimensional LC and MS/MS to separate and fragment peptides for protein identification (Link et al., 1999). The main advantage of shotgun proteomics over 2‐D gel electrophoresis followed by MS is higher throughput. The advantages over the SELDI method (Forde and McCutchen‐Maloney, 2002) include better coverage of proteins with high molecular weight and the ability to identify proteins directly (SELDI typically identifies unique peaks only, i.e., pattern recognition; unique protein can be identified at a later stage with extensive off‐chip workup). With MudPIT and an LCQ system, we were able to identify more than 300 proteins in a previous study, focused on aging related changes in human CSF (Zhang et al., 2005a). Over the last year or so, however, it has become increasingly clear that the LCQ system is associated with several limitations with one of the major ones being slow scanning speed (Yi 2002, Zhang 2005b), which translates into lower reproducibility of samples when analyzed multiple times and less total proteins identified as compared to MS with faster scanning speeds, for example, an LTQ‐ Fourier transform (FT) ion trap. Thus, in the current work with a goal of identifying as many CSF proteins in well‐characterized healthy young subjects as possible, we tested the LTQ‐FT system as well as the LCQ system with a better sample preparation procedure to circumvent the limitation associated with LCQ.

Section snippets

Collection of Human CSF by Lumbar Puncture and Exclusion Criteria

Written informed consent was obtained from all subjects and the Human Subjects Division of the University of Washington approved this study. All subjects were compensated community volunteers consisting of 10 women and 12 men aged 22–36 (median age = 25). All subjects underwent evaluation that consisted of medical history, physical and neurological examinations, laboratory tests, and brief neuropsychological assessment. Laboratory evaluation included a complete blood count and quantitative

Proteins Identified by μLC‐LCQ‐MS/MS

Online MudPIT analysis of 10 SDS‐PAGE fractions with the LCQ‐MS/MS identified a total of 466 proteins (Table I, Table II), 115 of which have the dubious distinction of being so‐called single‐hits (Table II). Single‐hits refer to the fact that a protein is identified from the MS/MS spectrum of a single peptide and, therefore, judged as being a less reliable identification than those proteins identified with multiple peptide tandem mass spectra. We include them here in our totals with reasons

Acknowledgments

Proteomic characterization of human CSF is supported by NIH grants (S10RR17262 and P30ES007033) to DRG and NIH grants (R01AG025327 and R01ES012703) to JZ.

References (44)

  • Y. Zhou et al.

    Analysis of alpha‐synuclein‐associated proteins by quantitative proteomics

    J. Biol. Chem.

    (2004)
  • R. Aebersold et al.

    Mass spectrometry in proteomics

    Chem. Rev.

    (2001)
  • K. Blennow et al.

    Protein analysis in cerebrospinal fluid. II. Reference values derived from healthy individuals 18–88 years of age

    Eur. Neurol.

    (1993)
  • K. Blennow et al.

    Protein analyses in cerebrospinal fluid. I. Influence of concentration gradients for proteins on cerebrospinal fluid/serum albumin ratio

    Eur. Neurol.

    (1993)
  • K. Blennow et al.

    Protein analysis in cerebrospinal fluid. III. Relation to blood‐cerebrospinal fluid barrier function for formulas for quantitative determination of intrathecal IgG production

    Eur. Neurol.

    (1993)
  • T.M. Block et al.

    Use of targeted glycoproteomics to identify serum glycoproteins that correlate with liver cancer in woodchucks and humans

    Proc. Natl. Acad. Sci. USA

    (2005)
  • B.T. Chait et al.

    Weighing naked proteins: Practical, high‐accuracy mass measurement of peptides and proteins

    Science

    (1992)
  • G.L. Corthals et al.

    The dynamic range of protein expression: A challenge for proteomic research

    Electrophoresis

    (2000)
  • C.E. Forde et al.

    Characterization of transcription factors by mass spectrometry and the role of SELDI‐MS

    Mass Spectrom. Rev.

    (2002)
  • J. Gross

    “Mass Spectrometry.”

    (2004)
  • D.F. Hochstrasser et al.

    Proteomics and its trends facing nature's complexity

    Proteomics

    (2002)
  • A.J. Hughes et al.

    What features improve the accuracy of clinical diagnosis in Parkinson's disease: A clinicopathologic study

    Neurology

    (1992)
  • Cited by (30)

    • Cerebrospinal fluid proteomics in multiple sclerosis

      2015, Biochimica et Biophysica Acta - Proteins and Proteomics
      Citation Excerpt :

      To our knowledge there is no large scale study of the proteome changes in CSF due to the circadian rhythm. CSF used for biomarker discovery studies should ideally be withdrawn at the same time of the day [13,33], but since this can be difficult to establish in clinical practice, recoding the time of lumbar puncture (LP) is necessary [36]. The sex related changes of the CSF proteome is not well studied.

    • Proteomics of human cerebrospinal fluid: Discovery and verification of biomarker candidates in neurodegenerative diseases using quantitative proteomics

      2011, Journal of Proteomics
      Citation Excerpt :

      For example, alcohol and cigarette smoking can alter the concentration of different components like hemovannilic acid and serotonin in CSF [72,73]. Heavy cigarette smoking and alcohol consumption other than socially have therefore been used as exclusion criteria in some studies [36]. Different medications have also been shown to cause alterations of CSF proteins [22].

    View all citing articles on Scopus
    View full text