Differential Proteomic Investigations of Normal Appearing Gray Matter in Multiple Sclerosis and Control Post-Mortem Brain Tissue

DIFFERENTIAL PROTEOMIC INVESTIGATIONS OF NORMAL APPEARING GRAY MATTER IN MULTIPLE SCLEROSIS AND CONTROL POST-MORTEM BRAIN TISSUE.

A dissertation submitted to Kent State University in partial fulfillment of the requirements for the degree of Doctor of Philosophy

Laurie A. Broadwater

May, 2013

Dissertation written by

Laurie A. Broadwater

B.S., Kent State University, 2002

M.S., Kent State University, 2003

Ph.D., Kent State University, 2013

Approved by

______, Co-Chair, Doctoral Dissertation Committee Roger Gregory, Ph. D.

______, Co-Chair, Doctoral Dissertation Committee Jennifer McDonough, Ph. D.

______, Member, Doctoral Dissertation Committee Soumitra Basu, Ph. D.

______, Member, Doctoral Dissertation Committee Nicola Brasch, Ph. D.

______, Member, Doctoral Dissertation Committee Edgar Koojiman, Ph. D.

Accepted by

______, Chair, Department of Chemistry & Biochemistry Michael Tubergen, Ph. D.

______, Dean, College of Arts and Sciences Raymond Craig, Ph. D.

TABLE OF CONTENTS

LIST OF FIGURES………………………………………………………………………iv

LIST OF TABLES……………………………………………………………………...viii

ACKNOWLEDGMENTS……………………………………………………...………...ix

INTRODUCTION…..…………………………………………………………………….1

DIFFERENTIAL PROTEOMIC INVESTIGASTIONS………………………………...32

Methods……………………………………………………………….…….32 Results……………………………………………………………….……....43 Discussion………………………………………………………….…….….73

METHOD DEVELOPMENT…..…………………………………………………….….93

Optimization of Matrix Formulation for Mass Spectral Analysis……...…...93 Synthesis of Novel Matrix Molecule…………………………………..…..103 Affinity Capture for the Identification Verification……………….………113

CONCLUSIONS……………………………………………………………....…….....133

BIBLIOGRAPHY……………………………………………………………..……….137

iii

LIST OF FIGURES

Figure 1: Geographical Distribution of Multiple Sclerosis………………………………..2

Figure 2: The role of mitochondria in axonal degeneration………………………………7

Figure 3: Panel A illustrates the MALDI process while Panel B illustrates the ESI process……………………………………………………………………………………10

Figure 4: Mass Analyzers for the Proteomic Laboratory………………………………...14

Figure 5: Types of surfaces available for Proteinchip® interaction studies……………..17

Figure 6: The key to the selectivity of SELDI…………………………………………...18

Figure 7: Schematic of the SELDI system……………………………………………….21

Figure 8: The correlating time of flight with m/z ratios in laser desorption mass spectrometry……………………………………………………………………………...22

Figure 9: Graphical representation of the transformation of the coordinate system during

PCA………………………………………………………………………………………25

Figure 10: A scree plot summarizes the total variability accounted for in the PCA…….28

Figure 11: Scatter plot of scores aid in the visualization of data segregation……….…...30

Figure 12: Hierarchical clustering analysis shows similar objects……………….……...31

Figure 13: Proteomic Work Flow employed in this research……………………………34

Figure 14: Tissue Characterization of representative MS NAGM tissue section using PLP staining…………………………………………………………………………………...44

Figure 15: PLP staining of MS and Control Motor Cortex……………………………...45

Figure 16: Representative western blot demonstrating the relative purity of the cellular fractionation……………………………………………………………………………...46

Figure 17: Scheme outlining the data analysis strategy for identification of differentially expressed proteins………………………………………………………………………..47

Figure 18: Representative SELDI-TOF spectra from a single patient…………………...48

Figure 19: A box and whiskers plot visualizes the distribution of data points…………..51

Figure 20: Multivariate analysis of cohort 1, fraction 3 using hierarchical clustering analysis…………………………………………………………………………………...55

Figure 21: Multivariate analysis of cohort 1, fraction 6 using hierarchical clustering analysis…………………………………………………………………………………...56

Figure 22: Multivariate analysis of cohort 2, fraction 6 using hierarchical clustering analysis…………………………………………………………………………………...57

Figure 23: Principal Component Analysis of Cohort 1 Fraction 3………………………58

Figure 24: Principal Component Analysis of Cohort 1 Fraction 6………………………59

Figure 25: Principal Component Analysis of Cohort 2 Fraction 6………………………60

Figure 26: Decision Tree used in the selection of differentially expressed to submit for peptide fingerprint mapping (PFM)……………………………………………………..63

Figure 27: SELDI-TOF mass spectra of differentially expressed proteins identified

PFM……………………………………………………………………………………...68

Figure 28: SELDI-TOF mass spectra of differentially expressed proteins identified by

MS/MS at 9.7 and 9.8 kDa and 15.9 kDa…………………………………………...... 69

Figure 29: SELDI-TOF mass spectra of differentially expressed proteins identified by

MS/MS at 16.7 and 17.2 kDa.……………………………………………………………70

Figure 30: SELDI-TOF mass spectra of differentially expressed proteins identified by

MS/MS at 22.7 kDa……………………………………………………………………...71

Figure 31: Affinity pull down assay for the confirmation of protein identification…...... 73

Figure 32: Cytochrome c Oxidase, subunits I, II, II, 5a and 5b (pdb 2EIJ file)….….…..82

Figure 33 Creatine Kinase Reactions. ……………………………………………………86

Figure 34: The CK Shuttle- connecting the utilization and production of energy……….90

Figure 35: Comparison of SELDI TOF mass spectra of mitochondrially-enriched protein samples from human brain tissue with CHCA and SPA matrix solutions………………98

Figure 36: Comparison of SELDI TOF mass spectra of mitochondrially-enriched protein samples from human brain tissue obtained with formic acid (FA) and OGP……………99

Figure 37: Representative SELDI TOF mass spectra of mitochondrially-enriched protein samples from human brain tissue obtained with OGP alone…………………………...101

Figure 38: Structure of α-cyanocinnamic acid and chloro-cyanocinnamic acid……….104

Figure 39: 1H and 13C NMR spectra of the purified chloro-cyanocinnamic acid……..107

Figure 40: Representative SELDI spectra acquired using both CHCA and Cl-CCA…..108

Figure 41: Signal to Noise ratio comparison of peaks detected using CHCA and Cl-

CHCA…………………………………………………………………………………..109

Figure 42: SELDI spectra acquired using stored Cl-CHCA matrix……………………111

Figure 43 : 13C and 1H NMR spectra acquired using stored Cl-CHCA matrix……….112

Figure 44: SELDI spectra of nitrated ribonuclease A (1) and hen egg white lysozyme

(2)……………………………………………………………………………………….121

Figure 45: Nitrated Ribonuclease A peak intensity is significantly changed after exposure to anti-nitrotyrosine beads………………………………………………………………122

Figure 46: Reproducibility of normalized ribonuclease A intensity before and after exposure to anti-nitrotyrosine affinity beads…………………………………………...126

Figure 47: Mining nitrated proteins from a complex brain mitochondrially enriched sample…………………………………………………………………………………..128

Figure 48: Identifiy verification of the 10.6 kDa SELDI mass spectral peak with anti-

COX5b affinity beads…………………………………………………………………..130

Figure 49: The verification of COX5b identity by affinity capture…………………….131

vii

LIST OF TABLES

Table 1: Performance Characteristics of Typical Proteomic Mass Analyzers ……..….. 13

Table 2: Donor Demographics……………………………………….....…….…..……...33

Table 3: Differentially Expressed Spectral Peaks…...... 50

Table 4: Differentially Expressed Proteins Identified by PMF………………………….67

Table 5: Confident Candidates for MS/MS Identifications of Differentially Expressed

Peaks……………………………………………………………………………………..67

Table 6: The effect of OGP concentration and sample:matrix dilution factor on protein ionization, desorption and detection……………………………………..……………..102

Table 7: p values generated by Homoscedastic two-tailed Student’s t-tests comparing

CHCA and Cl-CCA……………………………………………….……………………110

Table 8: Mean peak intensities before and after affinity bead exposure……………….125

viii

ACKNOWLEDGEMENTS

I would like to thank Professor Roger Gregory for his guidance, skillful censure and valuable mentorship during my graduate years.

I would like to express my gratitude to Dr. Jennifer McDonough for the most amazing scientific project and the freedom to explore those aspects of this project.

Chapter 1 Introduction

1.1 Multiple Sclerosis and its molecular pathology

Multiple sclerosis (MS) is an inflammatory neurodegenerative disorder of the central nervous system which results in physical and cognitive disability. It is the most common chronic neurological disease among young adults, with a mean onset age of approximately 30 years and a prevalence of 1.3/1000 people in the developed world

(Ziemann, 2011). Geographical distribution (Figure 1, Multiple Sclerosis Research

Center, UK) is not uniform with the greatest incidence in the extreme latitudes and women are two to three times more likely to develop MS than men (Hassan-Smith,

2011). Symptoms include numbness in the arms or legs, pain, loss of vision, muscle weakness or tremors, paralysis, vertigo, fatigue, speech difficulties and depression.

Existing pharmaceutical treatments slow the disease progression by targeting immune components of MS; however, no cure is currently available (National Multiple Sclerosis

Society online publication, 2012).

Figure 1: Geographical Distribution of Multiple Sclerosis

Attempts to characterize the genetic cause of MS have revealed that familial reoccurrence is approximately 20% (Compston, 2008). The human leukocyte antigen

(HLA) on chromosome 6 is a major histocompatibility complex (MHC) and alleles DR15 and Dq6 are shown to be associated with MS (Compston, 2008). Genetic susceptibility studies indicate a limited genetic contribution from HLA-DRB1*1501 and one study concluded that the Cytotoxic T Lymphocyte-associated antigen 4 (CTLA-4 [CD152]) gene was associated with DRB1*15 haplotype in multiple sclerosis (Alizadeh, 2003).

Consequently, these limited genetic differences do not provide adequate distinctions to make conclusions regarding disease pathology, suggesting that MS is polygenic and most probably epigenetic. The gender prevalence of multiple sclerosis, low concordance among homozygous twins and connection between several genes, also suggest an epigenetic component to MS which creates an even more complex paradigm for MS investigators (Casaccia-Bonnefil, 2008). Genetic studies of MS patients have revealed several distinct genes involved in its pathology. Mononuclear blood cell samples taken

from eight pairs of monozygotic Finnish twins discordant for MS confirmed two-fold up regulation of twenty-five genes and down regulation of 15 genes in 25% of the twins

(Särkijärvi, 2006). Genetic microarrays revealed differences between MS and control tissue samples in both white and gray matter (Dutta and Trapp, 2012). In a gene profile study using normal appearing white matter (NAWM) tissue, Graumann (2003) compared nine MS cases with seven matched controls to investigate molecular alterations which precede lesion formation. Hierarchal clustering of the data established the induction of genes involved in response to oxidative stress, increased metabolism and neuroprotection.

Vascular endothelial growth factor (VEGF), VEGF receptor, members of the PI3K/Akt signaling pathway and its targets- hypoxic induction factor 1α (HIF1α), hexokinase-1, brain glucose transporter 3 and CREB were all up regulated, implicating a response to oxidative stress. Neuroprotection mechanisms can be correlated with the presence of various anti-apoptotic genes including 14-3-3, BAG-1, bcl-x, bcl-w, adenosine A1 receptor, GABA A/B receptor, xeroderma pigmentosa. In addition, REF-1, hexokinase 1 and brain glucose transporter 3 (GLUT3) were all up regulated in MS NAWM.

(Graumann, 2003). In a study of normal appearing gray matter (NAGM) from six MS and six control donors, genetic microarray results indicated significant alterations in two ontology-based biological functions: electron transport chain enzymes and GABA synaptic transmission. Twenty six electron transport genes were under-expressed in MS

NAGM suggesting that mitochondrial dysfunction is an important factor in MS pathology

(Dutta, 2006). A study of the coordination of the down regulation of these electron transport genes using nuclear extracts from NAGM samples found a decrease in a transcription factor complex containing nuclear respiratory factor 2 (NRF-2). This

decrease is correlated with decreased expression of electron transport chain subunit genes and increased oxidative damage measured by increased anti-nitrotyrosine immunoreactivity. Researchers concluded that this increase in chronic oxidative damage leads to aberrant regulation of transcription of genes involved in energy metabolism

(Pandit, 2009).

White matter (WM) lesions are indicative of immune pathology in MS and studies involving myelin have provided evidence of variation between MS and control samples.

In a comparative proteomic study, using six MS and two control patients, Han identified tissue factor and protein C inhibitor as differentially expressed within chronic active WM plaque samples. Experimental autoimmune encephalomyelitis (EAE) is a mouse model used to investigate the immune component of MS and when recombinant protein C was administered to EAE mice, symptom severity decreased (Han, 2008). Using formalin fixed WM from three diseased and four control donors, Ly (Ly, 2011) found various myelin proteins were under expressed in chronic lesions of all three MS donors. Due to a decrease in myelin in chronic lesions this result was expected. However, glial fibrillary acidic protein, a major cytoskeletal protein of astrocytes, was up-regulated in chronic lesions, correlating with the astroglial proliferation that characterizes such chronic plaques. Contactin, a cell adhesion molecule involved in NOTCH signaling, was also found to be down regulated and this may account for the loss of astrocytes and oligodendrocytes in WM lesions (Ly, 2011). Another differential study of eight MS and six normal donors found differences of post-translational modifications of isolated myelin basic protein (MBP) tryptic fragments. MS peptides contained an increase in mono and dimethylation of arginine 107, increased deimination in various arginine residues and a

decrease or complete absence of phosphorylation (Kim, 2003). These white matter proteomic results correlate with the demyelination observed in lesioned tissue.

Recent publications indicate that gray matter lesion load in MS is extensive

(Stadelmann, 2008) and damage to both neurons and axons correlate with progression and cognitive impairment (Rudick, 2009, Inglese, 2004). Cortical damage has been suggested as an initiating event which is thought to launch a cascade of alterations leading to demyelination and chronic MS pathology (Desmazieres, 2012). MS proteomics studies utilizing gray matter are extremely limited (Zeimann, 2011; Derfussa, 2009). Our study employing proteomic techniques to mitochondrial fractions isolated from human gray matter is the single investigation published to date (Broadwater, 2011). Lesioned tissue exhibits alterations unique to disease while normal appearing tissue has no such indications and therefore is an ideal candidate for investigating those changes which are specific to disease onset. Through the analysis of normal NAGM from both MS and control individuals, key differentially expressed proteins in NAGM will provide evidence to the molecular events which signal the onset of MS and may lead to the development of superior therapeutics and possible cures for this disorder.

Energy failure in MS is being investigated as a primary cause of neurodegeneration (Paling, 2011; Lassmann and van Hossen, 2011) and mitochondrial dysfunction plays a key role in this phenomenon (Mahad 2008, Stadelmann, 2011). A recent review by van Horssen (van Horssen, J, 2012) highlights the critical role mitochondrial function plays in MS pathology. Van Horssen describes the key features of

MS pathology as axonal injury which is perceived as the main correlate for permanent clinical disability and thought to be mediated by the production of reactive oxygen

species (ROS) from microglia and infiltrated macrophages. Evidence of mitochondrial alterations in both white and gray matter has been reported (Dutta, 2006) and axonal injury results in an increased energy requirement. In normal neuronal function , an action potential travels by depolarization and repolization of the cell membrane at the Nodes of

Ranvier which are characterized by the presence of Na+/K ATPase and voltage-gated

Na+ion channels (Mahad, 2008, Brain). This signal relies on an intact myelin sheath, a source of ATP, voltage-gated Na+ ion channels and Na+/K+ ATPase (Figure 2A). Upon the destruction of the myelin sheath, propagation of the action potential by saltatory conduction is blocked (Figure 2B). In response, voltage-gated Na+ ion channels and

Na+/K+ ATPases are up regulated, restoring depolarization and repolarization, both energy dependent processes (Dowling, 1992). Furthermore, mitochondrial density and

Complex IV activity are enhanced to compensate for the increased energy demand

(Zambonin, 2011), (Figure 2C). The increased density of mitochondria in MS tissue contributes to the production of additional ROS (Mahad, 2008 Neuropathology and

Applied Neurobiology; Pandit, 2009), promoting the further demyelination of the neuron

(Smith, 2011). Evidence of the deleterious effects of ROS production is provided by 1H

MRS studies. The levels of glutathione, a powerful antioxidant, are known to negatively correlate with oxidative stress (Bains, 1997). Indeed, the levels of glutathione were found to be reduced in MS patients when compared to healthy patients (Choi, 2011; Srinivasan,

2010). Lactate, a product of anaerobic respiration, can indicate dysfunctional mitochondria and is found to be elevated in acute MS lesions (Zaaraoui, 2010; Paling,

2011). Additionally, mitochondrial DNA deletions, a product of mitochondrially

generated ROS as a result of impaired respiration, has been reported in NAGM of MS patients (Campbell, 2011).

Figure 2: The role of mitochondria in axonal degeneration. Panel A

indicates the normal function of the axon showing Nodes of Ranvier,

voltage-gated Na+ ion channels, Na+/K+ ATPase and an intact myelin

sheath. Panel B represents the loss of the myelin sheath. The Nodes of

Ranvier are no longer functional and the conduction of the action

potential is blocked. The conduction of the action potential can be

restored by the up-regulation of voltage-gated Na+ ion channels and

Na+/K+ ATPases as shown in Panel C. The density of the mitochondria

has increased as well as the activity of Complex IV (Horssen, 2012).

Comparative proteomic studies focusing on disease etiology are increasing in number and the involvement of mitochondria is recognized as critical in a variety of disease states (Jiang and Wang, 2012). Several studies have identified mitochondrial protein differential expression between normal tissue and MS or EAE. Western blotting of mitochondrially enriched fractions identified NADH dehydrogenase ubiquinone 1 subcomplex 6 (NDUF6) and low molecular mass ubiquinone binding protein (QP-C) as under expressed in MS NAGM (Dutta, 2006). Two-dimensional gel electrophoresis and

MALDI TOF/TOF detection of WM EAE mouse proteins showed NADH dehyrogenase

Fe-S protein 8, cytochrome c oxidase 5a, cytochrome c oxidase 5b, ATP5B, NADH dehydrogenase flavoprotein 2, glutaredoxin 5, estradiol 17 beta-dehydrogenase 8, isocitrate dehydrogenase, glutaredoxin 5, estradiol 17 and beta-dehydrogenase 8 to be down regulated (Fazeli, 2010). In the present study, anion exchange fractionation of mitochondrially enriched proteins followed by SELDI-TOF mass spectrometry detection also demonstrated cytochrome c oxidase 5a and 5b to be down regulated in mitochondrially-enriched fractions of human gray matter (Broadwater, 2011).

Investigations of non-lesioned brain tissue are expected to provide valuable information on the molecular changes which may take place prior to disease onset. Mitochondrial dysfunction has been implicated in the neuropathology of MS (Mahad, 2008) and so this proteomic investigation focuses on mitochondrially enriched fractions from NAGM and will provide evidence of key differentially expressed proteins and peptides which play a critical role in the early molecular events in MS pathology and may lead to the development of superior therapeutics agents and possible cures for this disorder.

1.2 Mass Spectrometry

1.2.1 Introduction Mass spectrometry is an analytical technique used to determine the masses of molecules. This is accomplished by the ionization of a solid, liquid, or gaseous sample.

This technique is useful for determining elemental composition and mass as well as elucidating the structure of peptides and other organic molecules. The data is visualized in a spectrum which plots the intensity of each ion as a function of their mass to charge ratio (m/z) (Aebersold & Mann, 2003).

In order to acquire data, the sample is first ionized into the gas phase. The ions are then separated by their m/z and finally detected using a device which counts charged particles. Each mass spectrometer requires four components: an ionization source, a device engineered to focus ions in space, a mass analyzer and a detector. Soft ionization techniques are required in the analysis of peptides and proteins. These ionizers are capable of generating gas phase ions without fragmenting the molecule and consist of matrix assisted laser desorption ionization (MALDI) and electrospray ionization (ESI).

Both techniques are used to analyze biomolecules and each has advantages and limitations. After co-crystallization of the analyte molecule with a matrix solution,

MALDI then sublimes and ionizes the biomolecule, producing singly charged ions. This technique provides a simple spectrum that is easily interpreted, tolerable to some salts and capable of ionizing molecules in excess of 500 kDa (El-Aneed, 2009). ESI ionizes molecules from the liquid phase and is therefore readily coupled to liquid chromatography systems, providing an orthogonal separation mode for increased

resolution. This hybrid system produces multiply charged ions that are easily suppressed by the salts used to solubilize biomolecules (El-Aneed, 2009). A schematic representation of MALDI and ESI can be seen in Figure 3.

Figure 3: Panel A illustrates the MALDI process while Panel B

illustrates the ESI process. In MALDI, analyte molecules are mixed

in large molar excess with an energy transferring matrix molecule and

upon laser ablation, molecules are sublimed and ionized, generating

charged species. In ESI, the solution sprays out a very small nozzle

creating charged droplets. A sheath gas evaporates the remaining

liquid leaving behind ions (Steen, 2004).

Next an extraction system is required to move ions into a mass analyzer. Since the ions are charged, this is typically accomplished by using an electric field. Ions are pushed out of the ionizing chamber into the mass analyzer. Proteomics research normally

employs four major types of mass analyzers: ion traps (quadrupole ion trap, QIT; linear ion trap, LIT or LTQ), time-of-flight (TOF) mass analyzer, LTQ Orbitrap and Fourier- transform ion cyclotron resonance (FTICR) mass analyzer. All have distinct advantages and limitations and through hybridization, instrument capabilities can be optimized

(Aebersold & Mann, 2003). The performance of mass analyzers is measured by several criteria. Resolution, a parameter which describes the separation of spectral peaks is measured using the following equation:

Equation 1: where M= m/z ratio

Resolving power can be defined as the mass to charge ratio of the observed mass divided by the difference between two mass to charge ratios. Mass accuracy is the ratio of the m/z measurement error to the true m/z and is reported in ppm. Sensitivity indicates the minimal detection amount. Dynamic range describes the range over which values can be detected. For example, some proteins may have a concentration of millimoles but others may have concentrations of micromoles. Can they still be detected? If the technique has a dynamic range greater that 1E+03, then both subsets of proteins can be detected.

The QIT is a robust, sensitive workhorse that is limited by its mass accuracy

(Aebersold & Mann, 2003). The low mass accuracy is a function of the limited number of ions which can accumulate in the point-like center of the trap. As the density of ions increases, space-charging distorts their distribution and interferes with the measurement of their mass (Aebersold & Mann, 2003). Figure 4 illustrates the small area for ion accumulation in the QIT. The mass resolution of the QIT is about 1000 with a mass accuracy of 100-1000 ppm. Picomolar sensitivity with an m/z range of 50-4000,

depending upon the sampling mode, provides suitable performance for most proteomic investigations. Coupled with MS/MS capabilities, this mass analyzer is an instrument of choice for many scientists (El-Abeen, 2009). The LTQ mass analyzer is a modification of the trapping space of the QIT. The linear trap captures the ions in cylindrical space, replacing the spherical space thus increasing the area of ion storage. This allows for an increased number of ions to be detected and thereby increase the mass accuracy of the technique. Figure 4 illustrates the difference in trapping space and configuration of the two traps (Aebersold & Mann, 2003). The resolution is double that of the QIT with a mass accuracy of 100-500 ppm. The scan rate is faster than the QIT with a dynamic range of four orders of magnitude while maintaining MS/MS capabilities (Han, X., 2008).

The technical aspects of Time of Flight (TOF) mass analyzers are discussed in detail in the Section 1.2.2. TOF analyzers have a mass resolution of 10,000-20,000 with a mass accuracy of 10-20 ppm. The sensitivity is in the femtomole range and there is no theoretical limit to the m/z range. In fact, TOF detectors are known to routinely characterize samples having m/z in excess of 500 kDa (Han, X., 2008). TOF mass analyzers have a fast duty cycle, a dynamic range of four orders of magnitude and are ideally suited for MALDI ionization because the pulsed laser of ionization provides a t=0 for calculating m/z.

FTICR instruments also employ an ion capture mechanism to measure m/z. They are extremely accurate with a mass accuracy of less than 2 ppm. The mass resolution can reach 750,000 with the same sensitivity and dynamic range as the QIT. However, this instrument is difficult to operate and maintain and has limited fragmentation efficiency

(Abersold & Mann, 2003). Figure 4 Panel C illustrates the capture and separation

mechanism of the FTICR mass analyzer. The Orbitrap captures and separates ions based upon their frequency generation due to harmonic oscillation around a central spindle-like electrode. The frequency detected is a function of the ions m/z. A hybrid LTQ coupled to the Orbitrap combines the robustness, sensitivity and MS/MS capabilities of the LTQ with extremely high mass accuracy and resolution of the Orbitrap. This instrument can attain a mass resolution up to 100,000 with mass accuracy less than 5 ppm and femtomolar sensitivity (Han, X., 2008). The duty cycle and dynamic range exceed that of the FTICR as does the ease of use (Makarov, 2006). Table 1 summarizes the characteristics for each mass analyzer (Han, X., 2008).

Table 1 Performance Characteristics of Typical Proteomic Mass Analyzers

Mass Mass Mass m/z range Scan Dynamic MS/MS Ion Accuracy Sensitivity Analyzer Resolution (Da) Rate Range capability Source (ppm) 50-2000; QIT 1000 100-1000 picomole moderate 1E+03 Yes ESI 200-4000 50-2000; LTQ 2000 100-500 femtomole fast 1E+04 Yes ESI 200-4000 10,000- No practical TOF 10-20 femtomole fast 1E+04 No MALDI 20,000 limit 50,000- 50-2000; ESI; FTICR < 2 femtomole slow 1E+03 Yes 750,000 200-4000 MALDI LTQ- 30,000- 50-2000; moderate ESI; < 5 femtomole 4E+03 Yes Orbitrap 100,000 200-4000 to fast MALDI

Figure 4: Mass Analyzers for the Proteomic Laboratory. Panels A and B illustrate three-dimensional ion traps and their ability to capture ions. The QIT (panel A) and the LTQ (Panel B) capture fragments ions of a particular m/z. The LTQ Orbitrap instrument

(Panel C) captures ions initially using an ion trap but then separates and isolates those ions using frequency generated by harmonic oscillation around a central spindle-like electrode. The frequency detected is a function of the ions m/z. The FT-MS (Panel D) traps ions using strong magnetic fields in combination with the linear ion trap.

All the above instruments are suitable for MS and MS/MS experiments.

1.2.2 Surface Enhanced Laser Desorption Ionization

High throughput mass spectrometry based proteomic techniques have made substantial contributions to systems biology in the past decade (Aebersold & Mann,

2003; Cravatt, 2007; Abu-Farha, 2009; Lu, 2009). Differential proteomic investigations provide a platform to characterize the expression of proteins from a variety of specimens and cohorts simultaneously. DIGE (differential gel electrophoresis), cICAT (cleavable isotope coded affinity tags), and iTRAQ (isobaric tags for relative and absolute quantification) offer relative quantitative differential expression comparisons of up to eight samples (Wu, 2006). Protein co-migration and isoforms complicate quantification of gel based techniques such as DIGE. Gel free methods such as ICAT, quantify and identify cysteine labeled tryptic fragments so those peptide fragments lacking cysteine residues or those containing cysteine residues blocked by posttranslational modifications cannot be quantified and consequently, relative quantification is limited to a subset of proteins. iTRAQ requires separate sample preparation prior to trypsin digestion which increases handling error and allows for only a maximum of eight samples for comparison.

High mass accuracy, high resolution instruments such as those equipped with FT-ICR mass analyzers, are capable of detecting multiple species in a single experiment in a label free quantitation format (Grossmann, 2010). While providing superior resolution, mass accuracy and sensitivity, such instrumentation is limited by the difficulty of use and expense. While these techniques offer absolute or relative quantification, sufficient statistical significance can only be achieved by increasing the sampling population and a cheaper, more efficient, alternative is needed.

Surface enhanced laser desorption ionization time of flight mass spectrometry

(SELDI TOF MS) is a high throughput label free discovery proteomics tool which can analyze large sample populations (Simpson, 2009, Mannello, 2009). The major advantage of the SELDI-TOF system is its selectivity and tolerance for contaminants in the sample preparation process. Salts and detergents, which can reduce or eliminate gas phase ionization, are rinsed away when using the SELDI-MS work flow. Additionally, affinity binding chemistries are used to capture specific types or subsets of proteins. Sample binding is accomplished by exposing the sample solution to a specific chemistry on the surface. These surfaces consist of hydrophilic, hydrophobic, ion exchange, and IMAC.

The chip can also be derivatized with target molecules to be used in the investigation of

DNA, protein or antibody binding interaction studies. Figure 5 shows a few surface chemistries used in the SELDI process.

Figure 5: Types of surfaces available for Proteinchip® interaction

studies. Hydrophilic, hydrophobic, ionic and IMAC surface

chemistries help reduce the complexity of the detected peptides and

proteins by selectively binding a subset of the species in the sample.

The biochemical surface chemistries can be created by chemically

attaching a specific protein or drug to the surface.

The physiochemical properties of the sample interact with the Proteinchip® while those species which do not have affinity for the specific surface can then be rinsed away

(Figure 6).

Figure 6: The key to the selectivity of SELDI. The surface of the

SELDI Proteinchip® is coated with different types of chemistries to

allow for affinity interactions between the surface and the sample.

Panel A represents the surface immediately after application of the

analyte. Many species are present in varying concentrations, making

detection complicated. Panel B represents the spot after washing. The

solution is less complex and therefore spectra are easier to interpret.

Those species which remain bound are then resuspended with a matrix molecule. This matrix molecule, typically an aromatic acid with high absorptivity for the wavelength of

laser radiation employed, is used in large molar excess to transfer sufficient energy from the laser to the analytes, causing desorption and ionization (Dreisewerd, 2003). Exposure to an electric field then moves the positive ions toward a detector which then counts the number of ions. Figure 7 illustrates the desorption process graphically. In order to identify the molecular weight of the species on the Proteinchip®, it is necessary to accelerate the ion using an electric field. The laser is pulsed to desorb and ionize the sample which generates ions. This pulse is t=0. An electric field is then used to accelerate the ions toward a slit. This converts all potential energy to kinetic energy. Passage through this slit into an electric field free drift tube allows the ions to move toward the detector solely as a function of their momentum which is a direct result of their mass and charge. The velocity of the charged particle after acceleration will not change in the field- free time-of-flight tube. The velocity of the particle can be determined in a time-of-flight tube since the length of the path (x) of the flight of the ion is known. Upon striking the detector, the time is recorded and used to calculate the mass to charge ratio (m/z) according to equation 7, derived below.

Kinetic energy (Ek) is a a function of both velocity (ν) and mass (m) as well as charge

(z) and voltage (V). See Equation 2 and Equation 3, below.

2 Equation 2: Ek=1/2 mv

Equation 3: Ek=zV

By combining the two equations and rearranging such that the mass to charge ratio is a function of velocity and voltage, Equation 4 is derived:

Equation 4:

Velocity is also a function of time (t) and distance (x) as seen in Equation 5.

Equation 5:

Substitution of Equation 5 into Equation 4, followed by rearrangement gives Equation 6.

Equation 6:

A specific voltage is applied to all ions which exposes them all to the same electric field in the same field free drift tube making 2V/x2 a constant. Therefore the equation becomes-

Equation 7:

Equation 7 illustrates the physical principles exploited by time of flight mass detection.

By simply measuring the time an ion requires to travel the length of the flight tube, the m/z can be calculated. Figure 8 shows the mathematical relationship between m/z and the flight time of a molecule.

Disadvantages of the SELDI work flow include the inability to identify species through tandem mass spectrometry, low mass accuracy and low resolution. Low mass resolution presents challenges when analyzing complex spectra. Those peaks which are resolved from one another may be post translational modifications (PTM), adducts generated in the ionization process or just proteins of similar molecular weight.

Orthogonal separations and reducing the sample complexity are both techniques used to manage this issue. However, the ability to acquire sufficient data to generate statistically significant differential expression patterns is unique to the gel free work-flow afforded by

SELDI-MS.

Figure 7: Schematic of the SELDI system. The chip is loaded into the instrument. The laser then ablates the sample causing ions to be generated. These ions are accelerated toward the detector using an electric field. Once through a slit (not shown) the electric field is no longer present and the ions move as a function of their momentum until they impede upon the detector.

Figure 8: The correlating time of flight with m/z ratios in laser

desorption mass spectrometry. Energy (E) is a function of mass (m),

velocity (v) as well as voltage (V) and charge (z) which can be used to

derive the equation used to determine the m/z of the detected ions,

where t is time. See text for derivation.

A variety of proteins have been identified using this differential expression workflow including hemoglobin beta chain in cystic fibrosis (Gomes-Alves, 2010) and platelet factor 4 with connective tissue activating peptide III in pediatric acute lymphoblastic leukemia (Shi, 2009). The SELDI IIc platform does not have protein identification capabilities and must be used in conjunction with a tandem mass spectrometer. A peak of interest is isolated from possibly contaminating species by

HPLC or one dimensional gel electrophoresis prior to analysis by tandem mass

spectrometry. Purification methods of the above work consisted of gel electrophoresis and HPLC, respectively, ensuring only a single species was submitted for further analysis. HPLC purification is time consuming and dilutes the species such that additional concentration steps are required. One dimensional gel electrophoresis purification has a low resolving power and can generate more than one confident result from database perturbations.

1.3 Multivariate Analysis

Multivariate statistical techniques are often employed in the analysis of complex biological data patterns (Komori, 2012; Varghese and Ressom, 2011; Wang and

Mizaikoff, 2008). Principal component analysis (PCA) is an unsupervised statistical method that uses orthogonal transformation to change one large data set of possibly correlated variables (in this research- spectral intensities and their m/z ratio) into a linear combination of uncorrelated variables called principal components (Figure 9).

The objective of PCA is to extract the most important information from the data set thereby simplifying the description of that data set. This allows us to analyze the structure of the variables by using a new set of orthogonal variables, the principle components (PC). This new set of representative variables can then be plotted as points in maps to show patterns which would be unrecognizable in the original data set (Abdi &

Williams, 2008)

The number of PCs can never exceed the number of original variables and the first PC accounts for the largest amount of variability in the data set. Each subsequent PC accounts for the next largest amount of variability while also being orthogonal to the

previous PC. Because the first PC accounts for the greatest variability in the data set with each subsequent PC accounting for decreasing amounts of variability, the first several

PCs will reveal which variables contribute most to the differences in the data set.

A Figure 9: Graphical representation of the transformation of the coordinate system during PCA. Panel A shows the data in its original coordinate system as indicated by the blue axes. The spread of the data across both axes is similar and both x and y values appear to contribute equally to the variability of the data set. Panel B shows the data coordinates after the transformation. It is apparent that the new coordinates describe the contribution of variability more completely than the previous coordinate system with axis A accounting for the majority of the variability in the data while axis B accounts for less variability.

Mathematically, PCA is defined as an orthogonal linear transformation that changes the data to a new coordinate system so that the greatest variance by any projection of the data comes to lie on the first coordinate (the first principal component), the second greatest variance on the second coordinate, and so on (Jolliffe, 2002).

Computing the PCA of a data matrix is a multistep process and requires a computer algorithm. MATLAB, SPSS and other statistical packages allow the investigator to select the number of PCs and choose to use either the covariance or correlation matrix.

The data must be organized into a matrix with n rows and m columns. In this research all data treatments were performed by the Ciphergen Express software but the data would have been organized by donor (rows). Each column would represent a time of flight (which would directly correlate with an m/z ratio) and in that column would be the integer of the spectral intensity at that time. Each fraction would have its own data matrix including the control and MS donors. This data matrix would then be mean centered.

Mean centering generates a matrix with values between -1 and 1 with 0 as the mean and allows the data to be analyzed without regard to the magnitude of the measurements. As the magnitude of a measurement increases its contribution to the variability would be artificially increased. In practical terms, a peak at 25 kDa with a fold difference of 1.2 would have a more profound impact on the PCA than a peak at 10 kDa with a 2.5 fold expression difference. By mean centering the data all measurements have the same weight (Jolliffe, 2002).

Next the covariance matrix is calculated. The covariance is calculated between each measurement and arranged into a matrix which is symmetrical about the main diagonal. The covariance between two values is calculated using Equation 8 below:

Equation 8

The covariance matrix is then assembled according to Equation 9:

Equation 9:

cov(x,x) cov(x,y) cov(x,z)

C = cov(y,x) cov(y,y) cov(y,z)

cov(z,x) cov(z,y) cov(z,z)

After the covariance matrix has been calculated, the eigenvalues and eigenvectors are calculated. This requires software, as previously discussed. The columns of the eigenvalues and eigenvectors matrices are arranged, while maintaining the order and pairing of the eigenvalues and eigenvectors. The matrix of eigenvalues is the PC scores.

Several techniques are available to select the appropriate number of PCs to represent the data. If the analysis must account for a specified amount of variance, a simple table can be used to rank the eigenvalue of each PC with the amount of variability it describes. In the table in Figure 10, the sum of the eigenvalues of the first three components is 6.504.

If that is divided by the sum of all the eigenvalues the percent is 92.8%. This means that

over 90% of the variability of the data set is captured in the first three PCs. A scree plot is also, a valuable visualization tool for the selection of the number of PCs (Figure 10). By plotting the eigenvalues according to their rank, a change in slope is apparent and those

PCs whose eigenvalue are before the slope change are retained (Abdi & Williams, 2008).

Figure 10: A scree plot summarizes the total variability accounted

for in the PCA. The principle component data consisting of the

eigenvalues and the percent variation is shown organized into a table.

By totaling the percent variability, the number of PCs to be used in

data representation can be easily visualized. The total variability of

90% is accounted for using the first three PCs. Shown below the table,

a scree plot is a graphical representation of the eigenvalues and the PC.

As the slope decreases, the amount of variability accounted for in that

PC decreases. As the slope approaches 0 additional PCs do not add

value to the data reduction and should not be used.

In spectral interpretation, each PC consists of a linear combination of spectral peak intensities and m/z values and may contain data from a single spectral peak or many. The contribution of a specific PC to a sample's variability is indicated by the PC score. To aid in the visualization of the data reduction process, these scores are plotted in a two or three dimensional scatter plot (Figure 11). Samples with similar scores for a specific PC will segregate together in a scatter plot, allowing the identification of similar parameters of groups of parameters. The loadings of each PCA will then indicate which spectral peak is contributing to the variability in that specific PC. Because PCs are uncorrelated by definition (Jolliffe, 2002), this process can be used to verify those factors which independently influence the data segregation.

Figure 11: Scatter plot of scores aid in the visualization of data

segregation. Thousands of PC scores are plotted in the figure above. A

simple table of this data would not allow the analyst to determine if any of

the samples were related. PC scatter plots easily demonstrate the ability of

PCA to segregate data based on uncorrelated PCs. Source:

http://acgt.cs.tau.ac.il/expander/screenshots.html#3

Hierarchical clustering (HCA), as a data mining technique, was originally applied to genomic differential expression data by Eisen (Eisen, 1998). The application of hierarchical clustering to differential protein expression data identifies elements which may be linked in their occurrence (Broadwater, 2011; Becker, 2012). This unsupervised

classification method can be used to identify and validate those biomarkers responsible for specific patterns and plays an essential role in current disease investigations (Yen,

2011).

HCA clusters data by calculating distance between variables based on a user defined algorithm. The most common HCA method, agglomerative, defines clusters of the most similar measurements and then progressively adds less similar objects until all measurements have been included into a single large cluster. This data structure is visualized in a tree format called a dendogram. The branches of the dendogram link the most common objects and the length of each branch is indirectly proportional to the correlation coefficient calculated between those objects (very short branches indicate highly correlated variables). Figure 12 illustrates the data structure in the HCA output and how the objects are clustered.

Figure 12: Hierarchical clustering analysis shows similar objects.

This figure illustrates the relationships between objects.

Chapter 2 Differential Proteomic Investigation

2.1. Methods

2.1.1 Tissue Preparation and Characterization

Reagents used in the preparation of all buffers were obtained from Sigma-

Aldrich (St. Louis, MO). Brains were obtained from The Rocky Mountain MS

Center Tissue Bank (Englewood, CO), the Brain and Spinal Fluid Resource

Center (UCLA) and The Kathleen Price Bryan Brain Bank (Durham, NC) under

IRB protocol. Tissue was obtained by rapid autopsy and frozen at −70 °C. This research used two cohorts of samples, each consisting of four controls and four secondary progressive multiple sclerosis (SPMS) donors. Tissue from each cohort was homogenized on separate dates using identical protocols. MS and control brains were matched primarily for brain region (parietal, Brodmann areas 1–3, and frontal cortex, Brodmann area 9), and also for age, sex, and post mortem interval (PMI) as closely as possible. Table 2 describes the donor demographics.

Table 2 Age PMI Brain Sample Cohort (years) Gender (hours) Region C1 1 57 M 4.5 Fr C2 1 87 M 6.5 Pa C3 1 72 F 30 Pa C4 1 75 M 19 Pa C5 2 80 M 14 Fr C6 2 65 M 3.5 Pa C7 2 86 F NA Pa C8 2 73 F 12 Fr MS1 1 48 M 4 Fr MS2 1 30 M 5 Pa MS3 1 62 F 6 Pa MS4 1 36 F 3 Pa MS5 2 30 M 5 Fr MS6 2 78 M 15.5 Pa MS7 2 81 F 11.8 Pa MS8 2 52 F 20.6 Fr

C- control donors; MS- MS donors; M- male; F- female; Fr- frontal cortex; Pa- parietal cortex; NA- not available

A diagram of the work flow employed in this research is shown in Figure

13. Briefly, tissue sections were stained for myelin proteolipid protein (PLP), homogenized, separated by charge, and finally detected by mass to charge ratio.

Those peaks that were differentially expressed were purified by gel electrophoresis, bands were excised, trypsin digested and submitted for identification by LC-MS. This work flow is commonly applied to tissue, blood and other bodily fluids in order to investigate proteomic differences between samples.

Figure 13: Proteomic Work Flow employed in this research

Frozen sections, 30 μm thick, from adjacent sides of the tissue blocks were immunostained for PLP to ensure the absence of demyelinated lesions. Each section was fixed in 70% ethanol at room temperature for 30 minutes and then rinsed for 5 minutes in 50 mM phosphate buffered 500mM saline, pH 7.4 (PBS).

Endogenous peroxidases were quenched by incubation in 1% hydrogen peroxide at room temperature for 30 minutes. The sections were then rinsed three times for five minutes each in PBS at room temperature and allowed to air dry. Each sample was encircled by a hydrophobic pen to reduce amount of primary antibody required for incubation. Anti PLP (Chemicon MAB388) was used at a 1/200 dilution with 3% donkey serum in PBS 0.5% Triton X-100 as the diluent. Each section was covered with diluted antibody solution and incubated at 4˚C for seventy two hours. The sections were rinsed in 3% donkey serum, PBS Triton X-

100 three times for five minutes each. A biotinylated anti-mouse secondary antibody (PLP host animal was a mouse) was applied at a 1/500 dilution (3% donkey serum in PBS 0.5% Triton X-100 as diluent) for sixty minutes at room temperature. The sections were then rinsed three times for five minutes each in

3% donkey serum, PBS 0.5% Triton X-100 and then incubated in freshly prepared

Vector Elite ABC solution for thirty minutes. The sections were rinsed in PBS three times for five minutes each and then peroxidase substrate solution, diaminobenzidine (DAB) was applied until the desired staining level was achieved. The sections were rinsed three times in PBS for five minutes each, blotted and allowed to air dry. Each slide was protected from UV radiation using

Permount® under the coverslip and allowed to dry overnight in a hood. Vector

ABC solution contains both avidin and biotinylated horseradish peroxidase enzyme (HRP) solutions. Avidin contains four biotin biding sites and can thus be used to amplify a signal. Additionally, the HRP is now localized to the region of the primary antibody binding and the brown staining created by the reduction of

the peroxide can be used to visualize this location. Counter staining in standard hemotoxylin and eosin for 30 seconds was used to increase image contrast.

Images were acquired using an Olympus FV500 confocal microscope using bright field imaging mode. Images were acquired from each tissue section in a rastering fashion using a 500 μm x 500 μm field of view. The stacks were compiled using

ImageJ. Figure 14 shows representative image compilations from three MS samples (digital photography courtesy of Mike Sulak). The blue circles indicate those areas which were excised for proteomic analysis. Additional features were observed by applying a novel imaging technique to narrow the bandwidth of the excitation wavelength allowing for an enhancement of the DAB signal. Mike

Model (Imaging Core Facility in the Department of Biology) reconstructed the confocal filter block to create a 460-490nm bandwidth filter (Gordon, 1988).

2.1.2 Tissue Homogenization

Approximately 250 mg tissue from frozen blocks was carefully excised excluding vascular vessels and white matter. Tissue was homogenized using a

Wheaton homogenizer with a Teflon® pestle in whole cell homogenization buffer

(20 mM KCl, 3 mM MgCl2, 10 mM 4-(2-hydroxyethyl)-1- piperazineethanesulfonic acid (HEPES) pH 7.9, 0.5% NP-40, 5% glycerol with protease inhibitors (P2714, Sigma-Aldrich, St. Louis, MO) in forty strokes. The homogenate was centrifuged for ten minutes at 500g at 4°C. The supernatant was removed and centrifuged at 10,000xg for 30 minutes at 4°C. The pellet containing

the mitochondrially enriched fractions was further purified by washing twice in

20 mM phosphate buffered saline (PBS), pH 7.4. The mitochondrial pellet was lysed in mitochondrial lysis buffer (50 mM Tris, 7 M urea, 3% CHAPS with protease inhibitors) by vortexing for 1 minute and then incubating for 20 minutes at room temperature. The mitochondrial lysate was centrifuged for 10 minutes at

10,000xg at 4°C. A modified Lowry assay was used to quantify the supernatant protein concentration. All samples were stored at −80°C until further analysis.

2.1.3 Western Blotting

In order to demonstrate, equal amounts of mitochondrial protein, the mitochondrially enriched tissue lysates were separated by gel electrophoresis using NuPage® 4–12% Tris gel (Invitrogen, Carlsbad, CA) for 35 minutes at 200

V, transferred to nitrocellulose paper for 1 hour at 45 V and blocked using 5% milk fat in PBS 0.5% Tween 20. The blots were then incubated with the mouse anti-cytochrome c oxidase subunit II (COXII) antibody (Mitosciences MS405,

Eugene, OR.) at 1/2000 dilution overnight at 4˚C with gentle rocking. The blots were exposed to the corresponding secondary horseradish peroxidase conjugated secondary IgG (sc2005-mouse, Santa Cruz Biotechnology, Inc. Santa Cruz, CA) and visualized using enhanced chemiluminesence reagents (sc2048, Santa Cruz

Biotechnology, Inc. Santa Cruz, CA). Figure 16 shows levels of COXII in the tissue lysates.

Cytochrome c Oxidase subunit 5b antibodies (ProteinTech Group,

Chicago, IL, PN 11418-2-AP) were used to investigate the differential expression of this protein in mitochondrial tissue lysates.

2.1.4 Ion Exchange Fractionation

The mitochondrially enriched protein extracts were further fractionated using ion exchange chromatography in a spin column format (column-

UFC30HV00; centrifuge tube-UFC3000TB Millipore, Billerica, MA). Samples

(100 μg each) were equilibrated using buffer A (9 M Urea, 2% CHAPS in 50mM

Tris HCl pH 9.0) and incubated for 20 minutes at room temperature. Buffer B

(Buffer A diluted 1 part to 8 parts 50 mM Tris HCl, pH 9.0) was used to dilute all samples to 200 μL which were stored on ice while the remaining samples were prepared (30 minutes maximum). The quaternary ammonium anion exchange beads (Q ceramic HyperD F, Pall Biosepra, NewYork, NY) were equilibrated three times in buffer B. The samples were loaded onto the column/tube and mixed in an end to end fashion for 30 minutes at room temperature.

After this initial incubation, the samples were centrifuged for 1 minute at

1000g. This eluent was the flow through fraction. The column was removed from the centrifuge tube and placed in the next centrifuge tube labeled according to the wash buffer. The pH 9 elution buffer (50 mM Tris HCl, pH9.0, 0.1% octyl glucopyranoside (OGP), 200 μL) was added to the column and mixed in an end to end fashion for 10 minutes, then centrifuged as before. This process was repeated for each buffer at pH 7.0, 5.0, 4.0, and 3.0, and a final organic solvent elution,

yielding six fractions with pI ranges of greater than 9.0 in fraction 1, 7.0–9.0 in fraction 2, 5.0–7.0 in fraction 3, 4.0–5.0 in fraction 4, 3.0–4.0 in fraction 5, and less than 3.0 in fraction 6, respectively. Specific buffer compositions were as follows: pH 7.0 elution buffer; 50 mM HEPES pH 7.0, 0.1% OGP, pH 5.0 elution buffer; 100 mM sodium acetate pH 5.0, 0.1% OGP, pH 4.0 elution buffer; 100 mM sodium acetate pH 4.0, 0.1% OGP, pH 3.0 elution buffer; 50 mM sodium acetate pH 3.0, 0.1% OGP; organic elution buffer; 33.3% isopropyl alcohol (IPA),

16.7% acetonitrile (ACN), 0.1% trifluoroacetic acid (TFA). All protein solutions were stored on dry ice until returned to the −80 °C freezer.

2.1.5. SELDI-TOF-MS acquisition

All samples were mixed with a saturated α-cyano-4-hydroxycinnamic acid

(CHCA) matrix solution consisting of 50% ACN and 0.5% TFA by volume with

0.6% OGP by weight in a 1:5 dilution. Each matrix diluted sample (1 μL) was applied to a spot on an NP20 Proteinchip® (normal phase chromatographic surface, Ciphergen Biosystems, Fremont, CA) chip in a randomized manner.

Mass spectra were acquired with a model PBSIIc SELDI-TOF mass spectrometer manufactured by Ciphergen Biosystems (Fremont, CA). The optimal sensitivity and laser intensity were established on a location not used in transient averaging.

These values were then used to develop a spot protocol. This same protocol was applied to all chips. Data were acquired at a digitizer rate of 250.0 MHz in positive ion mode with a chamber vacuum of less than 5×10−07 Torr. The source voltage was 20 kV and the detector voltage was 2700 V. A total of 65 transients

were averaged for each spectrum. All spectral processing (smoothing and baseline subtraction) was performed with Proteinchip 3.1 Software (Ciphergen

Biosystems, Fremont, CA). Spectra were calibrated externally over the appropriate mass range using at least three of the following standards: recombinant Hirudin BHVK (6,964 Da), bovine cytochrome c (12,230 Da) equine cardiac myoglobin (16,952 Da), bovine RBC carbonic anhydrase (29,024 Da) and

Saccharomyces cerevisae enolase (46,670Da). In order to make appropriate comparisons, all spectra were normalized against the total ion current (Avasarala,

2005).

2.1.6. Data analysis

Manual peak picking, using a signal to noise ratio of greater than 4.0, allowed for unbiased peak selection. Automatic peak detection in Ciphergen

Express often neglects those features which appear as shoulders on larger peaks

(Carlson, 2005) and consequently manual peak selection across all spectra of the same fraction provided a more accurate assessment of the spectral features.

Criterion for the selection of peaks across the entire data set required peak occurrence in 25% of the spectra with a signal to noise ratio greater than 4.0. Peak occurrence percentages were selected based upon the expected absence of some peaks from either the control or the MS group. Peaks selected for further evaluation and validation had a signal to noise ratio of greater than 3.0. All statistical calculations were done with Ciphergen® Express software (Ciphergen

Biosystems, Fremont, CA). Univariate analysis included the non-parametric

Mann–Whitney test. Applicable multivariate techniques included principal component analysis (PCA) and hierarchical clustering (Eisen, 1998). Peaks considered differentially expressed were those altered by at least 1.8 fold and p≤0.05.

2.1.7. Protein purification and peptide fingerprint mapping (PFM)

Initial PFM on individual ion exchange fractions were purified by gel electrophoresis on 16% acrylamide Tris glycine gels at 200 V for 45 minutes

(Laemmli, 1970). The gels were stained with Coomassie blue (Echan, 2002) and gel plugs of interest were removed. The gel plugs were destained (Jimenez, 1998), dehydrated and stored in 1% acetic acid prior to trypsin digestion and peptide fingerprint mapping. Trypsin digestion followed by HPLC-ESI-MS/MS on an

LCQ Deca XP mass spectrometer (Thermo Finnigan, San Jose, CA) was performed at the Genomics and Proteomics Core Laboratories at the University of

Pittsburgh.

A second analysis was performed using NuPAGE 12% Bis Tris gels using a 2-[N-morpholino]ethanesulfonic acid (MES) running buffer system. The gel was stained and destained as previously stated. The gel was plugged at the desired bands and the pieces were rinsed in 1:1 50 mM ammonium bicarbonate and acetonitrile (HPLC grade) for 30 minutes. The gel pieces were dehydrated using acetonitrile (ACN) and stored at -20C until trypsin digestion (Shevchenko,

2006).

Samples were trypsin digested using sequencing grade modified trypsin

(Promega). A single vial of the trypsin solution was diluted in 1.5 mL10 mM ammonium bicarbonate containing 10% acetonitrile. Any remaining ACN was removed from the tubes by pipet and 75 uL of the above trypsin buffer was added to each sample. The samples were incubated at 4˚C for 30 minutes and then checked to insure the trypsin solution would cover the gel plugs after they swelled. After 30 minutes the samples were transferred to the 37˚C oven for overnight incubation. The next day, the solution was transferred to a clean tube, spun to dryness and stored at -20C until analysis. The second set of samples was taken to the Proteomics Center at Case Western Reserve University for analysis on the LTQ Orbitrap Velos (Thermo Scientific).

2.1.8. Protein Identity Confirmation

Antibodies against COX5b (Proteintech Group, Chicago, IL, PN 11418-2-

AP) were derivatized to 3 μm diameter carboxylated polystyrene microspheres

(Polysciences PN 09850) using a modified carbodiimide method (Hermanson,

2008). Briefly, the beads were washed in 50 mM MES (2-(N- morpholino)ethanesulfonic acid, Sigma PN M8250) twice. A 200 mg/mL solution of EDC (N-Ethyl-N′-(3-dimethylaminopropyl) carbodiimide hydrochloride Fluka

PN-03449) in 50 mM MES was prepared. The washed beads (100 μL) were incubated with the EDC solution such that the final concentration was 2 mg/mL for approximately 2 minutes in a vortexer set on low speed. 200 μg of mitochondrially enriched NAGM protein was added to 500 μL of the bead slurry.

Protein and beads were incubated with constant mixing at room temperature for 3 hours. Beads were washed in 10 mM Tris, pH 8.0, 0.5% BSA twice. Beads were stored in 150 μL 10 mM Tris, pH 8.0, 0.5% BSA at 4°C. Bright field microscopy was used to confirm low self-aggregation and antigen-antibody activity of conjugated beads. Mitochondrial protein extracts from specific ion exchange fractions (2.5 μL) were incubated with bead solution (2.0 μL) in 10mM Tris, pH

8.0, 0.5% BSA to a final volume of 10 μL for 1 hour with constant mixing at room temperature. Solutions from both before and after bead incubation were diluted 1/10 in a saturated CHCA (Sigma PN 145505-5g) aqueous solution consisting of 50% ACN (Fisher PN A996-4), 0.6% OGP (Sigma PN O8001-5g) and 0.5% TFA (JT Baker PN9470-00). These solutions were randomly spotted onto an NP-20 Proteinchip® at 0.5 μL, air-dried and repeated.

2.2 Results

2.2.1 Characterization of Postmortem Brain tissue

A range of techniques were used in the characterization of the tissue prior to the proteomics analysis. It was necessary to establish the normal appearing nature of the samples prior to protein extraction. This was performed by immunohistological staining for proteins known to be absent in the lesioned areas.

PLP is present in NAGM and its absence would signal a possible diseased state.

Figure 14 shows both digital and PLP stained representative image compilations from a single MS donor (digital photography courtesy of Mike Sulak). The blue

circle indicates the area which was excised for proteomic analysis. Additional features were observed in the PLP stained images by applying a novel imaging technique to narrow the bandwidth of the excitation wavelength allowing for an enhancement of the DAB signal. By reconstructing the confocal filter block to create a 460-490nm bandwidth filter, image acquisition using DAB as a chromophore was significantly enhanced.

Figure 14: Tissue Characterization of representative MS

NAGM tissue section using PLP staining. PLP was used to

identify those regions which may contain MS lesions. The blue

encircled area marks the region excised for proteomic analysis in

this particular sample.

Figure 15 demonstrates PLP immunostaining of MS motor cortex showing a lesion in the center of the section. The immunostaining of the control motor cortex shows normal gray matter.

Figure 15: PLP staining of MS and Control Motor Cortex:

PLP immunostaining of MS motor cortex shows a lesion in

the center of the section. The immunostaining of the control

motor cortex shows normal gray matter.

Mitochondrial samples from either parietal or frontal cortex were obtained from NAGM as confirmed using PLP staining. Analysis of the mitochondrial fractions was divided into two separate groups or cohorts due to tissue sample availability at study onset. The first cohort of samples consisted of mitochondrial fractions obtained from four controls and four MS brain slices (C1–4 and MS1–

4). As additional brain tissue became available the analysis was extended to incorporate the second group, cohort 2, which also consisted of mitochondrial

fractions derived from four controls and four MS brain slices (C5–8 and MS5–8).

Western blots were performed on cytoplasmic and mitochondrial fractions isolated from MS and control cortex. The protein lysates were separated and blotted with an antibody to the neuron specific protein, neurofilament (NF), and an antibody to the mitochondrial encoded COX2 protein. Arrows denote multiple

NF immunoreactive proteins present in cytoplasmic fractions. A marker for mitochondria, cytochrome c oxidase subunit II (COXII), shows that COXII was found in all samples. Figure 16 illustrates representative Western blotting results and this analysis was performed by Ashish Pandit and Sausan Azzam

(Broadwater, 2011).

Figure 16: Representative western blot demonstrating the

relative purity of the cellular fractionation. Western blots were

performed on cytoplasmic (cyto) and mitochondrial (mito) fractions

isolated from MS and control cortex, run side by side and blotted

with an antibody to the neuron specific protein, neurofilament (NF),

and an antibody to the mitochondrial encoded COX2 protein.

Arrows denote multiple NF immunoreactive proteins present in

cytoplasmic fractions. (Broadwater, 2011).

2.2.2 Mitochondrial protein differential expression in postmortem MS cortex

Mitochondrial tissue lysates were separated, analyzed by SELDI-TOF-MS and the subsequent data analyzed according to the flow chart in Figure 17.

Acquire

Process

Peak Pick

Mann Whitney Mean Comparison p value < 0.05 selected for further investigation

Hierarchal Clustering Analysis samples clustering into disease state are selected for further analysis

Principal Component Analysis Samples with similar PC scores are selected for further investigation

Figure 17: Scheme outlining the data analysis strategy for

identification of differentially expressed proteins.

The mitochondrial tissue lysates were separated using ion-exchange fractionation and analyzed by SELDI-TOF-MS. Mass spectra were acquired from five fractions

with pI values ranging from 9 to 3 and numerous peaks were revealed. Figure 18 illustrates a typical series of SELDI TOF spectra acquired from a single donor.

The spectra are labeled according to their pI range. Orthogonal separations allow for the detection of multiple peptides and proteins which may have the same mass to charge ratio but different pI. Peptide spill over between charge fractions can complicate interpretation of the data downstream but reproducibility of the anion exchange does not contribute additional variability beyond what is inherent to the

SELDI TOF MS technique itself.

Figure 18: Representative SELDI-TOF spectra from a single

patient. NAGM mitochondrial tissue lysate was fractionated using

anion exchange resin generating 5 different fractions. This chemistry

reduces the number of species detected in single spectrum thereby

increasing the total number of species detected in five spectra. The

labels indicate the pI range of the protein species present in each anion

exchange fraction.

The Mann–Whitney test was used to identify those peaks with significant differences in peak intensities. Filtering the data for proteins altered by at least 1.8 fold at the 0.05 significance level we identified nineteen peaks which were differentially expressed (Table 3). Cohort 1 contained twelve differentially expressed peaks, with five of these peaks from fraction 3 (pI 5.0–7.0) and five from fraction 6 (pI<3), and cohort 2 contained seven differentially expressed peaks, three of which originated in fraction 6 (pI <3). These fractions were selected for peptide fingerprint mapping and further investigation. Fractions 4 (pI

4.0–5.0) and 6 from cohort 2 were also selected for further purification and peptide fingerprint mapping. After further calculations, peaks at 5050, 7925 and

7975 m/z were identified as doubly protonated peaks of parent ions at 10090,

15850 and 15940 m/z. The mass to charge ratios are half of the parent ions and the expression levels are also correlated with the suspected parent ions.

Additionally, the gel electrophoresis experiments used to purify the samples for

PFM did not contain a band at these molecular weights. Consequently, they were removed from the ID queue.

Table 3 Differentially Expressed Spectral Peaks

Cohort Fraction p value m/z Fold change ID 1 3 0.047 4570 2.1

1 4 0.047 5050 -1.8 2M+ 2 2 0.021 5322 4.3

2 2 0.021 5402 5.1

1 3 0.028 7925 3 2M+ 1 3 0.028 7975 2 2M+ 1 6 0.028 9663 -1.9 COX6c 1 6 0.047 9789 -1.8 COX5a 1 4 0.047 10090 -1.8

1 3 0.05 10600 -1.9 COX5b 2 6 0.043 11355 -1.8

1 6 0.016 12538 -1.8

2 4 0.021 15850 3.2

1 6 0.047 15887 2.3 Hbb/CaM/NDUFA13 1 3 0.009 15940 2 Hbb 2 4 0.021 16012 2.6 MBP 1 6 0.047 17203 -2.1 CaM/MBP 2 6 0.021 22773 -2.3 GFAP 2 6 0.021 42700 2 CKB

A box and whiskers plot can be used to visualize the distribution of data

points. Each group, MS and control, has its own box and whisker, with the box

representing the quartiles, the line in the box indicates the median and the

“whiskers” show the data range. This tool allows for a quick visual inspection of

the data characteristics mentioned. If there is any ovelap of the data spread, then

the two populations are not significantly different. The box and whiskers plot in

Figure 19 shows the distribution of both control (red) and MS (blue) data points

from fraction 4 cohort 2 from the peak at 15, 850 m/z. A Mann-Whitney test

determined the p value to be 0.021 which is supported by the box and whiskers plot.

Figure 19: A box and whiskers plot visualizes the distribution

of data points. This box and whiskers plot from the 15850 m/z

peak of fraction 4 of cohort 2 displays the segregation of control

and MS peak intensities.

2.2.3 Analysis of proteomic data using multivariate statistics

The application of hierarchical clustering to differential protein expression data identifies elements which may be linked in their occurrence. This unsupervised classification method can be used to identify and validate those biomarkers responsible for specific patterns and plays an essential role in current disease investigations. Since it is unlikely that one specific peak is responsible for the differential expression between control and MS NAGM mitochondrially- enriched proteins, a technique which detects correlated sets of variables is needed to analyze this data. In this research, a modified Pearson product–moment correlation coefficient was employed as a distance measure and the average linkage method was used to perform a hierarchical cluster analysis of protein expression and compute a dendrogram. Nodes of the dendrogram join the most

similar objects and the relative length of the branches between two objects is indirectly proportional to the similarity of those objects. The most significant feature in the hierarchical clustering analysis is the segregation of control and MS donors by disease state (fractions 3 and 6 from cohort 1 and fraction 6 from cohort 2 shown in Figure 20, Figure 21 and Figure 22. Other fractions in this data set have only one or two differentially expressed peaks and therefore are not sufficiently complex to warrant the use of multivariate methods such as hierarchical clustering. Fraction 3 of cohort 1 contains two distinct nodes (Figure

20). The short length of the branches indicates that the peak intensities of the proteins in each node are highly correlated, suggesting a potential relationship between the peaks in each node. The highly correlated peaks of the upper node, containing peaks at 10.7, 10.6, 10.2, 10.1 and 9.9 kDa, have a correlation coefficient greater than 0.85. The lower node, containing peaks at 15.9, 15.8, 8.0, and 7.9 kDa, has a correlation coefficient greater than 0.94. Of the four peaks, three were differentially expressed at the 95% confidence level, with p values of

0.009, 0.076, 0.028 and 0.028, respectively. Fraction 6 of cohort 1 (Figure 21) also shows several highly correlated nodes. The upper middle node containing peaks at 22.9, 22.7, 11.5, 23.7, and 11.4 kDa has a correlation coefficient greater than 0.85, while the correlation coefficient of the lower middle node consisting of peaks at 9.8, 9.7, 12.5, 12.6 and 17.2 kDa is greater than 0.92. The linear correlation of the data combined with significant p values of 0.076, 0.028, 0.028,

0.016, and 0.076, respectively, indicate a group of possible biomarkers. Fraction 6

of cohort 2 shows a distinct clustering by disease as well as several nodes (Figure

22).

Principal component analysis (PCA) is a common technique used in the analysis of large data sets and uses computational methods for pattern recognition

(Lavine, 2010). Mathematically, in this particular analysis, a PC is a linear combination of peak intensities from all samples included in the analysis. The contribution of a specific PC to a sample's variability is indicated by the sample score. Each sample has a single unique score from each PC. The first PC accounts for the greatest variability in the data set with each subsequent PC accounting for decreasing amounts of variability. Therefore, the first several PCs will reveal which variable contributes the most to the differences between the populations and the sample specific scores will relate that variability to each sample. In fractions 1 and 3 from cohort 1 and fraction 6 in cohort 2, PC1 and PC accounted for over 85% of the variability in the data set. Applying PCA to this research enabled the identification of key proteins which may be responsible for the differences between control and disease states. This particular analysis generated two dimensional scatter plots of the principal component (PC) scores to aid in the visualization of the data reduction process. Variables with similar scores for a specific PC will segregate together in a scatter plot, allowing the identification of similar variables. Because PCs are uncorrelated by definition (Jolliffe, 2002), this process can be used to verify those factors which independently influence the data segregation. Figure 23, Figure 24 and Figure 25 display PC scatter plots of PC1 and PC2. Notice the segregation of the data into MS and control sample groups in

each fraction. This technique also confirmed that other parameters such as brain region, gender, age or PMI were not responsible for data segregation. For all controlled variables, PCA verified that disease state was the only donor characteristic responsible for sample segregation.

Figure 20: Multivariate analysis of cohort 1, fraction 3 using hierarchical clustering analysis. HCA shows the clustering of MS and control samples. The peaks indicated by an asterisks have been analyzed by peptide fingerprint mapping and were identified as COX5b (10600 m/z) and MBP (15940 m/z).

Figure 21: Multivariate analysis of cohort 1, fraction 6 using hierarchical clustering analysis. The MS and control samples segregate into different clusters indicating a difference protein expression between disease states.

Figure 22: Multivariate analysis of cohort 2, fraction 6 using hierarchical clustering analysis. HCA shows the clustering of

MS and control samples.

Figure 23: Principal Component Analysis of Cohort 1 Fraction 3. MS samples have similar scores for PC 1 & 2 as is demonstrated by their grouping in the PC scatter plot. Additionally, PC 1 scores for controls are all negative.

Figure 24: Principal Component Analysis of Cohort 1 Fraction 6. In contrast to Fraction 3 in Figure 23, control samples have similar scores for

PC 1 & 2 as is demonstrated by their grouping in the PC scatter plot.

Additionally, PC 1 scores for MS samples are all negative.

Figure 25: Principal Component Analysis of Cohort 2 Fraction 6. MS samples have similar scores for PC 1 & 2 as is demonstrated by their grouping in the PC scatter plot. Additionally, PC 2 scores for controls are all negative.

2.2.4 Identification and confirmation of differentially expressed proteins in MS cortex

The decision tree used to determine which peaks would be submitted for identification by PFM or LC MS/MS is shown in Figure 26.

. The Mann-Whitney p values less than 0.05 with fold changes exceeding

1.8 were considered for further investigation. Identification confidence was determined by the Mascot score. The Mascot score is a probability based value which compares the experimentally acquired mass with a database of theoretically determined masses. The score is directly proportional to the probability that the match of the acquired mass to the theoretical mass is a random event. Because scores may have a large range, probabilities are reported as a logarithmic value according to the Equation 10.

Equation 10: Mascot Score= -10 log P, where P is the probability the m/z is a random

event.

In a peptide fingerprint mapping experiment, a score of approximately 90 is considered an identity for many instruments. In MS/MS experiments, the results are more complicated and the output contains a score for every peptide detected. This score is probability based as in Equation 10 but is a function of the fragmentation of a single peptide. The number of fragments and their matched m/z are then used in a probability algorithm to assign a peptide score. The confidence of the peptide score is largely a function of the mass accuracy of the instrument acquiring the data. A higher the score provides a more confident result.

The peptides detected from a specific protein are grouped together and the resulting protein score is a ranking of how well the detected peptides match the protein stated. This protein score is based on the number of peptides sequences detected, how many times each sequence was detected, and how accurately the expected peptide mass matches the empirical mass. (Matrix Science) This score is not probability based and the user must gather other information such as tissue specificity, physical properties of the excised band and other available biochemical information in order to decide which match best fits all the available data. If there is any ambiguity, the ID must be confirmed by other methods such as western blotting or the antibody pulled down assay used in this research and described in detail in Chapter 3.

Figure 26: Decision tree used in the selection of

differentially expressed to submit for identification. A

Mann Whitney p value of 0.05 and a fold change of 1.8

were set as the minimum requirements for PFM

submission.

The unambiguous identification of specific differentially expressed peaks in cohort 1 by peptide fingerprint mapping was complicated by the low resolution of one dimensional gel electrophoresis and the complex mixture of analytes in the

10 kDa and 16 kDa regions. Each band contained at least two protein species as evidenced by the SELDI-TOF mass spectra. Regardless, the entire band of each sample was excised and submitted for peptide fingerprint mapping on a

ThermoFinnigan LCQ Deca XP mass spectrometer. Table 4 lists the most probable identifications with Mascot scores, number of peptides, sequence coverage, probabilities and protein characteristics. The list of matches in the

database for the 10 kDa protein fragment included cytochrome c oxidase subunit

5b (COX5b) (UniProt accession number P10606), an electron transport chain subunit of cytochrome c oxidase (Complex IV) with a Mascot score of 20.2 from two peptides detected and 12.24 % sequence coverage. The 15.9 kDa peak in fraction 3 was identified as myelin basic protein (MBP) (UniProt accession number P02686) with a Mascot score of 80.21 from 10 peptides with a sequence coverage range of 25.7% for isoform 1 to over 49% for isoform 6 (see Table 4).

This fraction certainly represents a proteolytic fragment of one of the MBP isoforms. In cohort 2, proteins from fraction 4 and 6 were purified by gel electrophoresis and bands were removed at 16 kDa and 42 kDa, respectively. The protein at 42 kDa was identified as creatine kinase type B (UniProt accession number P12277) with a Mascot score of 70.26 from seven peptides and sequence coverage of 27.11%. The peak at 16 kDa was identified as hemoglobin β-chain

(UniProt accession number P68871) with a Mascot score of 60.25 from six peptides and a sequence coverage of 51.37%. The SELDI-TOF mass spectra for these four proteins are shown in Figure 27, panels A–D, and illustrates the overexpression of hemoglobin β, MBP, and creatine kinase B, and decreased expression of COX5b in MS relative to controls.

In a second identification experiment, bands were excised and submitted for LC MS/MS on a Thermo LTQ Orbitrap mass spectrometer. Table 5 lists the most probable identifications with the number of sequences and peptides, sequence coverage as well as the number of peptides with Mascot peptide scores greater than 60. The list of matches in the database for the 9.7/9.8 kDa peptides

included cytochrome c oxidase subunit 6c (COX6c) (UniProt accession number

P10606) and cytochrome c oxidase subunit 5a. Both are subunits of the electron transport chain complex cytochrome c oxidase (Complex IV). The protein Mascot score for COX6c was 988 with a sequence coverage of 72%. COX5a had a

Mascot protein score of 146 with a sequence coverage of 22%. Figure 28 Panel A shows SELDI TOF spectra of four control and four MS patients in this mass region.

The 15.9 kDa peak from fraction 6 contained three confident biologically reasonable identifications and included hemoglobin β chain, Calmodulin (UniProt accession number P62158) and NADH dehydrogenase [ubiquinone] 1 alpha subcomplex (accession number Q9P0J0). The protein Mascot scores were 890, 557 and 548, respectively with sequence coverages of 69.2%, 29.7% and 48.6%, respectively. All three identification contained three peptides with Mascot scores exceeding 60 which is considered identity.

Myelin basic protein (MBP) (UniProt accession number P02686), had a protein Mascot score of 411 from 6 peptides and a sequence coverage of 32.2% for isoform 5 (see Table 5). This hit is probably a proteolytic fragment as this spectral peak has an m/z of 17.3 and the full length species has a molecular weight of 18.5 kDa. Calmodulin is also a confident hit at this band with a protein

Mascot score of 972 from 6 different sequences with a coverage of 48.6%.

Additionally, the SELDI TOF spectra show a distinct band at 16.7 kDa which is the molecular weight of active Calmodulin. Figure 29 shows this area of the spectra.

Finally, the MS/MS analysis of the differentially expressed peak at 22.7 kDa returned an extremely confident identification of Glial Fibrillary Acidic

Protein (GFAP) (UniProt Acccession number P14136) with a protein Mascot score of 1757 from 12 different sequences for a sequence coverage of 30%. A total of twenty-five peptides were detected and five peptides returned Mascot scores of 60 or greater. The spectra are illustrated in Figure 30.

Table 4 Differentially Expressed Proteins Identified by PMF

pI range Mann- Fold UniProt Mascot Sequence of Whitney Change Entry Accession Protein Peptides Coverage fraction p value wrt MS Name Number Score Detected (%) 5-7 0.05 -1.9 COX5b P10606 20.2 2 12.24 5-7 0.009 2 MBP1a P02686-3 80.2 10 39.59 5-7 0.009 2 MBP2a P02686-4 80.2 10 41.94 5-7 0.009 2 MBP3a P02686-5 80.2 10 45.61 5-7 0.009 2 MBP4a P02686-6 80.2 10 48.75 3-4 0.021 2.6 HBB P68871 60.2 6 51.37 < 3 0.021 2 CKB P12277 70.3 7 27.11

a- denotes possible isoforms of MBP from the peptides detected.

Table 5: Confident Candidates for MS/MS Identifications of Differentially Expressed Peaks

m/z of # of Mann- Fold UniProt Protein # with Sequence Differential Entry Sequences/ Whitney Change Accession Mascot peptide Coverage expressed Name Peptides p value wrt MS Number Score score >60 (%) peak Detected

9.7 0.028 -1.9 COX6c P09669 988 Sep-32 2 72 9.8 0.028 -1.9 COX5a P20674 146 3-Feb 2 22 HBB P68871 890 20-Sep 3 69.2 15.9 0.047 2.3 CaM P62158 557 8-May 3 29.7 NDUFA13 Q9P0J0 548 11-Jun 3 48.6 17.2 0.047 -2.1 CaM P62158 972 13-Jun 3 44.6 17.2 0.047 -2.1 MBP3 P02686-5 411 12-Jun 2 32.2 22.3 0.021 -2.3 GFAP P14136 1757 25-Dec 5 30

Figure 27: SELDI-TOF mass spectra of differentially expressed proteins identified PFM. Panel A: The 16 kDa peak was identified as hemoglobin β chain, panel B: The 42 kDa peak was identified as creatine kinase type B, panel C: The 10.6 kDa peak was identified as

COX5b, panel D: The 16 kDa peak was identified as MBP. Mass spectra of mitochondrial samples outlined in red are control samples and those outlined in blue are MS samples.

Figure 28: SELDI-TOF mass spectra of differentially expressed proteins identified by MS/MS. Panel A: The peaks at 9.7 and 9.8 kDa were identified as COX6c and COX5a, respectively. Panel B: The

15.9 kDa peak contained three confident identifications.

Figure 29: SELDI-TOF mass spectra of differentially expressed proteins identified by MS/MS. Panel A: The peaks at 16.7 and 17.2 kDa were identified as CaM and MBP3, respectively. Panel B: This expanded view of the spectral peak 17.2 kDa highlights the complexity of the spectra in this region.

Figure 30: SELDI-TOF mass spectra of differentially expressed proteins identified by MS/MS. The differentially expressed peak at 22.7 kDa was identified as GFAP, an intermediate filament protein found in astrocytes.

Confirmation of the identification of the peak at 10.6 kDa as COX5b was obtained by employing COX5b antibody labeled beads. Chapter 3 discusses the specific details of this technique. Briefly, SELDI TOF-MS spectra acquired both before and after protein solutions were exposed to this COX5b antibody confirmed that the differentially expressed peak at 10.6 kDa was COX5b as shown in Figure 31 . Relative peak intensity ratios were used to determine which peak resulted from COX5b. The 10.1 kDa peak was verified as unchanged in mass spectra acquired before and after treatment with anti-COX5b antibodies by one way ANOVA and was therefore used as the normalization peak to determine which peak intensities changed as a result of exposure to the COX5b IgG beads.

Representative spectra from one of three trials are shown in Figure 31. Only the peak at 10.6 kDa demonstrated a change in relative peak height. This molecular weight corresponds to the mass of COX5b after signal peptide cleavage.

Figure 31: Affinity pull down assay for the confirmation of protein

identification. A control donor tissue sample was incubated with anti-

COX5b affinity beads at varying concentrations and demonstrates a

reduction in protein peak intensity at varying concentrations of beads.

2.3 Discussion

2.3.1 Hemoglobin β (Hbb; P68871)

The differential protein expression analysis of mitochondrially enriched fractions from MS and control gray matter demonstrated that the differentially expressed proteins in cohort 1 fraction 4 and fraction 6, with a mass of 16 kDa, were consistent with hemoglobin-β. Notably, both hemoglobin-α2 and hemoglobin-β genes are upregulated in mononuclear blood cells from MS patients

as reported recently in a study comparing monozygous twins discordant for multiple sclerosis (Särkijärvi, 2006). This phenomenon appears to be extended to the brain, since the data show increased expression of hemoglobin-β in MS cortex. Additional investigations in our research group localized hemoglobin expression to pyramidal neuronal cell bodies by immunofluorescent staining with antibodies to both hemoglobin and neurofilament. The discovery of hemoglobin expression in neurons, until recently considered specific to red blood cells, is novel. Ohyagi, (Ohyagi, 1994) demonstrated an over expression of both hemoglobin alpha and beta chain mRNA expression in developing neurons when compared to normal adult neurons and then confirmed the over expression by western blot. Additionally, research by the Ohyagi group indicates a potential toxic effect of extrinsic hemoglobin. Mixed neuronal and glial neocortical cell cultures exposed to hemoglobin for 24-48 hours experienced concentration dependent cell death. Intrinsic hemoglobin may mediate iron homeostasis in neurons which is relevant due to the effects of iron accumulation in MS pathogensis (Williams, 2012).

The unexpected expression of hemoglobin in rat, mouse and human neurons has been demonstrated by several research groups (Broadwater, 2011;

Schelshorn, 2009; Biagoli, 2009; Richter, 2009; Ohyagi, 1994), contradicting the traditional philosophy that hemoglobin functions solely as an oxygen transporter in erythrocytes. Recent studies support our finding, including studies by Biagoli.

(Biagoli, 2009) and Richter. (Richter, 2009) in which the expression of both α-

and β-globin transcripts was shown in dopaminergic neurons from mouse, rat, and human brains.

Richter detected both α and β globin transcripts in pyramidal, mesencephalic, dopaminergenic (DA) and striatal GABAergenic rat neurons without evidence of erythoid transcripts. Colocalization of the globin transcripts with neurons was confirmed using immunohistochemistry of neuronal nuclear antigen marker and in situ hybridizations of the globin transcripts. Following the administration of the mitochondrial complex I inhibitor, rotenone, down regulation of both α and β globin transcripts was observed while other brain globins (neuroglobin and cytoglobin) were unaffected. This result was confirmed by quantitative RT-PCR. Similarly, Biagoli found both alpha and beta hemoglobin transcripts in A9 dopaminergic neurons and confirmed this expression using western blots. Then using gene microarray analysis and dopaminergic cell lines over expressing α and β globin chains, 4617 genes were found to have a fold change of greater than 1.2. Ingenuity software determined the two major biological pathways involved were oxygen homeostasis and oxidative phosphorylation, implicating hemoglobin expression in mitochondrial function.

Specifically, 36 of the 78 nuclear encoded genes for the subunits of the electron transport chain complexes were up-regulated in hemoglobin over expressing A9

DA cell lines, further implicating hemoglobin in the function of mitochondria.

Taken with the Richter results, mitochondria can influence hemoglobin gene regulation and hemoglobin can influence mitochondrial gene regulation. These

complicated and perhaps, contradictory results indicate that there is still much to learn regarding the expression and function of hemoglobin in neurons.

In other neurodegenerative disorders, hemoglobin is known to interact with amyloid β in Alzheimer’s disease, mediating the binding of the amyloid β and heme bound iron (Chuang, 2012). Hypoxic-like tissue damage, known to be associated with MS (Lassmann, 2003), induces the expression of erythropoietin

(EPO) and its receptor in the brain (Marti, 2004). Using microarray and qRT-PCR it has been demonstrated that brain mRNA of an EPO transgenic mouse line, when compared to their wild-type littermates, had an elevated level of α-globin mRNA. When a single dose of human recombinant EPO was administered, hemoglobin expression was induced in the brains of wild-type mice after 24 hours at both the RNA and protein level. Compared with control animals, a strong time- dependent induction of α and β-chain mRNA (1.6- and 1.3-fold at 6 hours; 13.7- and 10.8-fold after 24 hours, respectively) was shown (Schelshorn, 2009).

In addition to oxygen transport in erythroid cells, a growing body of evidence supports additional roles for hemoglobin. In endothelial cells, nitric oxide (NO) signaling regulates blood pressure, blood flow and oxygen delivery.

Nitric oxide synthase (NOS) which produces NO in response to different stimuli is regulated by hemoglobin α (Straub, 2012). Hemoglobin α is expressed in both human and mouse endothelial cells and enriched at myoendothelial junctions where it regulates nitric oxide synthase in endothelial cells. Hemoglobin may also aid in protecting neurons from oxidative damage, since the ability of hemoglobin

α2β2 to scavenge peroxynitrite (ONOO−) in a variety of O2 bound states has been suggested (Herold, 2005).

A new class of bioactive peptides has been linked to hemoglobin derived peptides. These peptides have been shown to interact with both opoid and cannabinoid receptors to induce hypotension and reduce pain sensation through both opoid and non-opoid pathways (Gomes, 2010). These peptides are derived from both α and β- globins through possible interactions with cathepsin and other proteases (Gomes, 2010). Additionally, in separate PFM and MS/MS experiments, a peptide with an m/z matching the Hbb derived peptide,

LLVVYPWTQR, has been detected on three separate occasions. The biological mechanism required to generate this peptide is unknown but due to a Mascot peptide score of 44 in the most recent MS/MS, trypsin may generate these biologically active peptides. This peptide is known to interact with opoids, angiotensin IV, bombesin 3 and angiotensin converting enzyme which effect analgesic pain sensitivity, blood pressure, learning and memory as well as the potentiation of cholinergic transmission. Indeed, Kooi (Kooi, 2011) found the cholinergic transmission to be imbalanced in MS suggesting that perhaps the detected peptides may be purposefully affecting the MS brain and are not just a byproduct of excessive protease activity.

2.3.2 Myelin Basic Protein (MBP; P02686)

In the initial peptide fingerprint mapping experiment, the spectral peak at

16 kDa in cohort 1 fraction 3 was identified as MBP, a component of myelin. The

fold change was 2 with a p value of 0.009, a Mascot score of 80 with 10 peptides detected and 48.8 % sequence coverage constituting a confident identification.

This was an unexpected result and one that was consistent in mitochondrial fractions from both controls and MS donors. MBP is a very basic protein which is normally associated with membranes and it is possible that MBP released from myelin during tissue homogenization was bound adventitiously to mitochondrial membranes and remained in the mitochondrial fraction even after the washing steps. Additional work from our lab suggests that this adventious binding may also be a result of freezing tissue. When both fresh and frozen mouse brains were compared by Western blot analysis, the frozen samples were found to contain

MBP fragments, while the fresh tissue did not (Azzam, 2013, in press).

Regardless, samples were consistently found to have increased levels of MBP associated with MS fractions and its presence is an important distinguishing factor separating controls from MS as determined in our principal component analysis.

Proteomic studies of mouse brain have also shown mitochondrial proteins, including several involved in the electron transport chain, as well as hemoglobin, and creatine kinase B in myelin preparations Taylor, 2004; Vanrobaeys, 2005;

Ishii, 2009; Mugnaini, 1977; Ravera, 2009). However, this is the first time that myelin basic protein has been observed associated with mitochondrial fractions.

Western blot analysis also indicated an increase in the amount of MBP associated with mitochondrial fractions in MS donors relative to controls by Western blot and have also shown that the expression of MBP isoforms is altered in the EAE mouse model of MS. These data suggest that MBP can be localized and

associated with mitochondria both in normal and diseased brain, but the significance of this for MS disease pathology is not clear.

In a second experiment, MS/MS proteomic analysis identified a differentially expressed protein to be MBP at the 17.2 kDa spectral peak in cohort

1 fraction 6. The Mascot score was 411 with 12 peptides matched and a sequence coverage of 32.2%. The specific peptides matched indicate that isoform 5,

(MBP3) not 4 or 6, was the identified protein. The full length MBP3 protein has a molecular weight of 18.5 kDa but is known to undergo proteolytic cleavage in MS

(Medvecky, 2006). Generation of a 17.2 kDa fragment from the 18.5 kDa isoform

5 might have resulted from an S10-K11 cleavage known to be catalyzed by matrix metalloproteinase 3 (Bacheva, 2011). MMP-2 is also known to cleave MBP at site

15, as well as many others (Shiryaev, 2009). Cathepsin D, a brain specific proteolytic enzyme, is implicated in lesion production in MS (Cuzner, 1973) and also known to cleave MBP at regions near the N terminus in a variety of human

MBP isomers (Bacheva, 2011). The preference for proteolytic cleavage near the

N terminus as opposed to others sites may be related to conformational changes.

(Cuzner, 1973). CD spectroscopy and molecular dynamic simulations have shown that the deimination of MBP disrupts its tertiary structure (Harauz, 2007).

Differential citrullination coupled with protease activity may provide an explanation for the generation of the 17.2 kDa peptide from the 18.5 kDa full length species.

2.3.3 Calmodulin (CaM; P62158)

Careful inspection of the spectra reveals the presence of more than one species (Figure 29) in the 17.2 kDa differentially expressed spectral peak. A second strong candidate for this spectral peak is calmodulin (CaM), a known binding partner for MBP (Majava, 2010, Libich, 2003). Calmodulin has a full length molecular weight of 16.7 kDa with four known phosphorylation sites and two acetylation sites (UniProtKB- P62158). In this identification, 13 peptides were detected for a total sequence coverage of 44.6%, generating a Mascot score of 972, indicating high confidence. CaM-MBP binding has been implicated in MS pathology (Harauz, 2009) through the interactions of MBP with actin. Calmodulin binding to the deiminated C terminus of MBP causes MBP to dissociate from actin and the plasma membrane. Consequently, the myelin structure held in place by MBP is disrupted by the interactions of Calmodulin and MBP.

2.3.4 Cytochrome c Oxidase Subunits 5a & 5b (COX5a; P20674 & COX5b; P10606)

Peptide fingerprint mapping identified the protein isolated from fraction 3 of cohort 1 in the 10 kDa region as COX5b, a component of Complex IV of the electron transport chain. The gene expression of several nuclear encoded proteins of the cytochrome c oxidase complex have been found to be down regulated in

MS (Dutta and McDonough, 2006), correlating with the present result of statistically significant under expression of subunit 5b in MS samples. The -1.9 protein fold change in the present research supports the -1.41mRNA fold change found by Dutta. Most nuclear encoded mitochondrial proteins, including subunit

5b, are transcribed with a positively charged N terminal signal sequence which is cleaved upon import into the appropriate mitochondrial location. The full length

COX5b has a molecular weight of 13,696 Da and a pI of 9.07. However, upon removal of the 31 residue transit sequence, the molecular weight is reduced to

10,613 Da and the pI decreases to 6.33. Both physical properties are in strong agreement with the differentially expressed species in fraction 3 having a molecular weight of 10.6 kDa and a pI between 5.0–7.0.

While the mitochondrially encoded subunits of Complex IV are catalytically active, the nuclear encoded subunits function to maintain Complex

IV by providing structural and assembly support (Galati, 2009). Indeed, in null yeast mutants of subunit 4, 5a, 5b, 6c and 7a, the cytochrome c oxidase complex failed to assemble (Capaldi, 1990). Additionally, Complex IV dysfunction may impact the production of free radicals, since it has been shown that various cytochrome c oxidase subunits provide protection from highly reactive peroxynitrite (Fontanesi, 2006). The positioning of subunits 5a and 5b between the catalytic subunits and the mitochondrial matrix further suggest a role in ROS protection. Specifically, COX5b essential tyrosine residues 31 and 89, located on the periphery of the matrix side of the protein, suggest a role in protecting the catalytic complex from ROS. Figure 32 generated from pdb file 2EIJ shows the crystal structure of COX in its native dimeric form. COX subunits I, II, III, 5a and

5b are represented in yellow, orange, green, purple and blue, respectively. The essential tyrosine residues at position 31 and 89 are in red. The location of the 5b

subunit with respect to the catalytic components and the mitochondrial matrix support a ROS protective function of both subunit 5a and 5b.

Figure 32: Cytochrome c Oxidase, subunits I, II, II, 5a and 5b

(pdb 2EIJ file). COX subunits I, II, III, 5a and 5b are shown in

yellow, orange, green, purple and blue, respectively. The essential

tyrosine residues at position 31 and 89 are in red. The location of

the both 5b subunits with respect to the catalytic components and

the mitochondrial matrix support a ROS protective function of

both subunit 5a and 5b. IMS- intermembrane space; IMM- inner

mitochondrial membrane.

Peptide mass fingerprinting following an anti-nitrotyrosine immunoprecipitation detected the peptide containing tyr31 as nitrated in unpublished data from our lab. Additionally, an anti-nitrotyrosine antibody pull down assay detected this peak at 10.6 kDa as being nitrated. Further details are provided in section 3.3.3. Furthermore, Pandit (2009) found cytosolic proteins in

MS samples to contain an increased level of nitrotyrosine when compared with controls. This result implies that MS tissue may be under increased oxidative stress and therefore has a higher turnover of ROS protecting species, like COX subunits 5a and 5b. COX5a was detected in a differentially expressed peak in cohort 1 fraction 6. MS/MS analysis returned two hits for the differentially expressed peak at 9.7/9.8kDa. One hit was for COX5a with two sequences and three peptides with scores of 65 and 64, yielding a protein hit score of 146.

Inspection of the SELDI mass spectra shows a minor species as a shoulder at 9.8 kDa and a major species at 9.7 kDa. The relatively low protein hit score and low number of peptides suggests that COX5a is the minor species in the shoulder. The active protein has a mass of 12.5 kDa and the peptides detected included those from the internal sequence of the protein indicating that this could be a proteolytic fragment of COX5a. The under expression of this species in MS tissue supports a possible role in protection from reactive oxygen species as discussed earlier.

Currently, this is the only proteomic study indicating a possible role of COX5a in neurodegeneration.

2.3.5 Cytochrome c Oxidase Subunit 6c (COX6c; P09669)

Another COX subunit, COX6c was identified in the mass spectral peak at 9.7/9.8 kDa of cohort 1, fraction 6. Both differentially expressed peaks are under- expressed in MS and it is apparent from the spectra (Figure 28) that this peak contains multiple species. The MS/MS analysis returned a confident hit for

COX6c with a protein score of 988 and sequence coverage of 72%. Inspection of the spectra indicates that the identity of this peak is most likely the predominant

9.7 kDa peak. The number of peptides detected (32) suggests that this is the major species in this band and not the differentially expressed shoulder at 9.8 kDa. The mass of the full length protein is 8.8 kDa and peptides from both the N and C terminus were detected indicating that a full length protein is indeed present.

Additionally, there is no known signal peptide for this nuclear encoded COX subunit so the active protein is the full length species. Post-translational modifications (PTM) may be responsible for the mass difference and further investigation is required to both confirm the protein identity and characterize any

PTM present. COX6c has been implicated as a biomarker for patient survival in breast cancer (Emerson, 2009) but its involvement in neurodegeneration is presently uncharacterized.

2.3.6 Glial fibrillary acidic protein (GFAP; P14136)

Glial fibrillary acidic protein (GFAP) has been detected as the major species in the differentially expressed spectral peak at 22.5 kDa. This peak is under- expressed in MS tissue with a fold change of 2.3 and a p value of 0.021. Figure 30 shows there are more than one species at the m/z but the major component is most

likely responsible for the differential expression. There were 25 peptides detected from 12 different sequences and 5 of those peptide sequences had Mascot scores greater than 60. Taken together these data provide strong evidence of a confident identification. The molecular weight of the intact protein, 49.8 kDa, does not match that of the differentially expressed band. GFAP is an intermediate filament protein common to astrocytes (Anderton, 1981). GFAP fragments in the 22.5 molecular weight range have been detected in rat brain protein lysates (Nicholas,

2003), indicating a possible neurochemical mechanism which generates GFAP fragments. Rat and human GFAP have a 91 percent homology with the human isoform containing two additional amino acid residues with respect to the rat.

GFAP has been implicated in MS pathology and studies revealing differences in the full length protein in CSF have been published (Maddeddu, 2013). The CSF of fifty-one patients with MS and eighteen with other neurological diseases

(OND) were analyzed and the results demonstrated that GFAP was higher in MS than in OND with a p-value of 0.001 (Chi-square Moods Median test for differences between medians). CSF from patients diagnosed with clinically isolated syndrome (n =46), relapsing–remitting MS (n=67) or primary-progressive

MS (n=22) were compared to controls diagnosed with other non-inflammatory neurological disease (NIND) (n=22). Western blot analyses of GFAP showed a statistically significant difference between NIND and RRMS patients with a p- value of 0.001 calculated using one-way ANOVA (Avsar, 2012).

2.3.7 Creatine Kinase B (CKB; P12277)

Creatine kinase was identified as over expressed in MS at the 42 kDa spectral peak with a fold change of 2 and p value of 0.021 in cohort 2, fraction 6. The full length protein molecular weight of 42.5 kDa aligns with the differentially expressed spectral peak at 42.7 kDa. Seven detected peptides with a sequence coverage of 27 % and a Mascot score of 70.3 is evidence of a confident identification.

Creatine kinase (CK) is a key enzyme in cellular bioenergetics and plays a crucial role in supplying ATP to tissues which have large, fluctuating energy requirements such as muscle and brain. CK catalyzes the addition of a high energy phosphate from phosphocreatine to ADP generating ATP and creatine.

The reverse reaction, phosphorylation of creatine is also catalyzed by CK. Figure

33 illustrates the transfer details of this reaction.

Figure 33 Creatine Kinase Reactions: CK catalyzes both the

addition and removal of highly energetic phosphates involving

phosphocreatine.

There are several different isotypes of CK which include brain (CKB), muscle (CKM), and three different isotypes in mitochondria (CKMT1A,

CKMT1B, CKMT2) (UniProt KP, 2013). The primary function of mitochondrial

CK is generation of phosphocreatine. Conversely, the major role of the cytosolic isozymes is generation of ATP and creatine. The names imply a difference in cellular location and indeed their compartmentalization is paramount for efficient enzymatic function (Wallimann, 1992). The mitochondrial and cytosolic (muscle or brain) isoforms function in concert to provide energy for a variety of cellular processes in excitable cells. This cycle has been referred to as the phosphocreatine circuit by Wallimann and five different roles have been suggested.

The primary function of the phosphocreatine circuit is to provide a temporal energy buffer for excitable cells such as brain cells, including oligodendrocytes, neurons (Manos, 1991), and astrocytes (Thompson, 1980). CK has high activity in the brain (Norwood, 1983) and its ability to buffer energy supplies is crucial for ideal function. Simply increasing the concentration of ATP is an insufficient mechanism for meeting the energy needs of excitable cells. ATP is a key regulator in multiple metabolic pathways and shifting the ratios of ATP and its metabolites such as ADP and AMP would contribute to multiple downstream complications. By instituting a complementary but separate mechanism, the deleterious effects of manipulating ATP concentrations are avoided. The basal cellular concentration of ATP in excitable cells is approximately 2-5 mM (Wallimann) which is sufficient to contact muscle for only several seconds. Phosphocreatine, which is metabolically inactive, has a concentration of 5-10 mM in brain tissue. This allows the CK system to replenish

ATP efficiently and continuously. In fact, the rate of ATP synthesis by CK is ten

times that of oxidative phosphorylation in rat cardiac muscle (Bittl and Ingwall,

1985).

Another major role of the CK circuit is to act as a spatial energy buffer.

The transport of energy equivalents across the mitochondrial membrane to the location of utilization is essential for tissue with high energy requirements such as brain and muscle. This function is often referred to as the CK shuttle (Bessman &

Geiger, 1981). Figure 34 (Andres, 2008) illustrates the transport of energy equivalents across the outer mitochondrial membrane utilizing the CK shuttle.

Creatine transports across the outer mitochondrial membrane via porin in a cation selective confirmation. Creatine is then transphosphorylated by mitochondrial CK to produce PCr and ADP. PCr is then transported out of the mitochondria via porin in its anion selective state where it is free to diffuse to the plasma membrane and supply ATP to ATPases. The diffusion rate of both creatine and phosphocreatine are significantly faster than ADP and ATP, respectively, which makes them much better suited to transport phosphate equivalents to the cytosol and plasma membrane for use in a variety of ATP dependent processes

(Selivanov, 2004; Vendelin, 2004 ; Kay, 2004). Figure 34 also demonstrates the compartmentalization and coupling of CK isozymes to ATP producing and utilizing enzymes. This mechanism also allows the CK circuit to function as an

ADP sensor. CK has a low Km for ADP and by maintaining the ADP in the vicinity of ATP utilization, cytosolic CK (CKB in the brain) effectively increases the efficiency of ATP hydrolysis (Kammermeier, 1987). When energy consumption exceeds production, the accumulating free ADP can have negative

metabolic consequences. CK/ PCr system controls the free ADP concentration such that mitochondrial respiration is regulated, ATPases remain active and cellular adenine nucleotides are not lost (Wallimann, 1992). Since cytosolic CK is coupled to ATP consuming enzymes, the ratio of ATP/APD is maintained. This same principle applies at the site of ATP generation, as well.

Figure 34: The CK Shuttle- connecting the utilization and production of energy. Mitochondrial CK is sequestered in the intermitochondrial space and generates PCr at the expense of ATP produced as a result of oxidative phosphorylation (1). Cytosolic

CK is functionally bound to the glycolic process and uses ATP synthesized during glycolysis to produce PCr. This pool of PCr buffers ATP/ADP ratios in the cytosol and supplies ATP for local consumption (2). Cytsolic CK is also coupled to membrane bound

ATPases, ATP gated ion channels and ATP-regulated receptors.

This shuttle provides a source of energy remote from the

production site and provides a mechanism to delivery ATP to cell

types with rapidly changing energy needs.

Creatine has been linked to several functions unrelated to energy production. Of particular interest is the anti-apoptotic function of the CK system.

In the presence of creatine, Mt CK activity was found to regulate mitochondrial permeability transition pore opening and modulate apoptosis (Dolder, 2003).

Creatine was also found to be a direct scavenger of mitochondrially generated

ROS (Sestili, 2006) and has been shown to reduce ROS through the tight coupling of creatine kinase and the oxidative phosphorylation process (Kay, 2000).

CK presence and activity in the central nervous system (CNS) has been investigated and found to be significant in a variety of neurodegenerative diseases including Alzheimer’s Disease (AD), Amyotrophic Lateral Sclerosis (ALS),

Charcot Marie Tooth Disease (CMT), Huntington’s Disease, (HD) and

Parkinson’s Disease (PD) (Aksenov , 2000; Wendt, 2002; Smith, 2006; Ryu,

2005; Hass, 2007). The association of CK with MS has also been investigated and results vary based on type and disease state of the tissue analyzed, attesting to the complexity of the molecular pathology of MS. Serum CKB activity detected by immunoassay was found to be decreased in the serum of MS patients with respect to control patients (Vassilopoulos, 1987). In a proteomic analysis of cerebral spinal fluid (CSF) from MS and non-MS patients, CKB was detected in the CSF of Relaxing-Remitting MS patients but not the non-MS control group CSF. These results support our finding of up regulation of CKB in control MS NAGM

(Nobem, 2005). CKB activity and concentrations were measured in secondary

progressive post mortem NAGM and found to be significantly decreased with respect to controls (p=0.0001 and 0.0007, respectively). The same study included

31P MRS measurements taken from NAWM of individuals with progressive MS and showed increased ratios of PCr/ATP when compared to control patients.

However, benign individuals were not significantly different than controls (Steen,

2010). Using two dimensional gel electrophoresis to identify differentially expressed proteins and identification by trypsin digestion followed by LC-

MS/MS, mitochondrial CK was found to be decreased in NAWM and NAGM of human specimens as well as brain and spinal cord samples from EAE mice

(Jastorff, 2009). This result links the previous discussed mitochondrial dysfunction and energy failure of MS to the creatine kinase circuit.

MS is characterized by dysfunctional mitochondria which generate increased ROS and possibly initiate apoptosis. The various functions of CKB in brain chemistry provide the means to counteract the excessive ROS generated by dysfunctional mitochondria. The two fold over expression of CKB in NAGM offers evidence of a compensatory role in MS.

Chapter 3 Method development

Section 1: Optimization of matrix formulation for the mass spectral analysis of human brain mitochondrial proteins

3.1.1 Introduction:

Matrix assisted laser desorption ionization mass spectrometry (MALDI-

MS) and Surface Enhanced Laser Desorption/Ionization Mass Spectrometry

(SELDI-MS), a modified MALDI-MS technique combining selective chromatographic surfaces with sensitive MALDI-TOF detection (Hillencamp and

Karas, 1988; Tang., 1993; Mann and Talbo, 1996; Poon, 2007; Liebler, 2002) are rapid and powerful techniques for the characterization of complex protein and peptide mixtures. Matrix-assisted laser ablation experiments require the protein sample to be mixed with an energy-transferring matrix. This matrix molecule, typically an aromatic acid with high absorptivity for the wavelength of laser radiation employed, is used in large molar excess to transfer sufficient energy to the analytes, causing their desorption and ionization (Dreisewerd, 2003). The mechanisms of protein desorption and ionization in laser ablation experiments are poorly understood but efficient energy transfer is known to require co- crystallization of the matrix and protein (Bae, 2012). Protein and matrix co- crystallization is enabled by the co-solubility of the protein and energy-

transferring matrix (Bird, 2002). The quality of MALDI mass spectra is highly dependent on the sample–matrix preparation procedure (Cohen and Chait, 1996).

Variations in the sample–matrix formulation can have dramatic effects on the populations of proteins that are detected by MALDI-MS as protein and polypeptide components partition between the matrix and analyte solutions. For example, matrix formulations with and without formic acid gave strikingly different MALDI mass spectra of E. coli proteins, with fewer than ten proteins common to both preparations (Ogorzalek-Loo and Loo, 2007). Efficient desorption, ionization and detection of proteins therefore requires careful optimization of the protein solubilization conditions, additives and matrices employed.

Membrane proteins can be particularly difficult to analyze by MALDI-MS and SELDI-MS due to membrane protein insolubility and the suppression of mass spectral peaks in the presence of certain detergents (Cadene and Chait, 2000).

Formic acid has been employed to enhance MALDI mass spectra of hydrophobic proteins, although its use can complicate spectral analysis by the formylation of some species and acid-catalyzed hydrolysis of peptide bonds (Ogorzalek_Loo,

2007; Cadene, 2000). Certain detergent additives also improve MALDI-MS detection. Ionic detergents are incompatible with positive ion mass spectrometry of proteins, but several non-ionic detergents, including OGP (Octyl β-D- glucopyranoside), NP-40, Triton X-100, and Tween 80 as well as zwitterionic detergents such as N,N-dimethyldodecylamine-N-oxide (LDMS) and Zwittergent

3-12(ZG) have proved useful in enhancing MALDI mass spectra (Cohen and

Chait, 1996; Brinkwort and Bourne, 2007; Gharahdaghi., 1996; Meetani and

Voorhees, 2005). Initially, SELDI analysis of mitochondrial proteins from postmortem human brain tissue showed no protein present in the sample.

Optimization of the sample-matrix formulation was therefore required before any conclusions could be made regarding the differential expression of proteins in NAGM and control donors.

3.1.2 Methods:

Brain tissue from parietal cortex was obtained from the Rocky Mountain

MS Center. Tissue of a known weight was homogenized in a buffer containing

20mM KCl, 3mM MgCl2, 10mM Hepes pH 7.9, 0.5% NP-40, 5% glycerol, and protease inhibitor cocktail with a Wheaton homogenizer with Teflon® pestle.

After 10-minute incubation, tissue was centrifuged for 10 minutes at 500xg in

4ºC. The pellet was stored at -80ºC. The supernatant was spun at 10,000xg for 30 minutes in 4ºC and the supernatant was stored at -80ºC. The pellet, the mitochondrially enriched fraction, was further purified by washing in 20mM PBS pH 7.4. The pellet was lysed in a buffer containing 50mM Tris, 7M urea, 3%

CHAPS and protease inhibitors for 20 minutes at 20ºC. The mitochondrial lysate was centrifuged for 10 minutes at 10,000xg at 4ºC. A bicinchoninic acid (BCA) assay was used to quantify the supernatant protein concentration. Samples were diluted with 50mM Trizma base to a standard concentration and stored at -80ºC.

The purity of the mitochondrial fraction was assessed by Western blotting with

antibodies specific for subunits of the mitochondrial electron transport chain

(Figure 16).

The mitochondrial proteins were fractionated according to the previously stated protocol in Chapter 2 Section 1.4. Proteins were separated using anion exchange chromatography in a spin column (Millipore PN UFC3OHV00) bead format using Q ceramic HyperD® F beads (Pall Life Sciences PN 20066-031).

All samples were stored on ice and transferred to -80ºC until mass spectral analysis. Proteins with pIs between pH 5 and 7 were chosen to optimize the matrix solution due to the number of mitochondrial proteins in this range [http:/ bioinfo.nist.gov/hmpd]. The matrix solution consisted of various formulations of acetonitrile, isopropanol, formic acid, trifluoroacetic acid, Octyl β-D- glucopyranoside (OGP), NP-40, Sinapinic Acid (SPA), and α-cyano-4-hydroxycinnamic acid (CHCA).

Mass spectra were acquired using a SELDI-TOF MS model PBSIIc

(Ciphergen Biosystems, Fremont, CA). Each fraction was applied to a spot on a gold chip (Ciphergen Biosystems, Fremont, CA) in a randomized manner. The optimal sensitivity and laser intensity were established on a location not used in transient averaging. These values were then used to develop a spot protocol which was applied to all chips. Data were acquired at a digitizer rate of 250.0 megahertz in positive ion mode with a chamber vacuum of less than 5 x 10 -07 torr. The source voltage was 20 kilovolts and the detector voltage was 2,700 volts. A total of 65 transients were averaged for each spectrum.

Proteins were purified by gel electrophoresis using Tris-glycine gels with

16% acrylamide and run at 125V for 75 minutes. In gel trypsin digestion and protein identification on a Thermo Finnegan LCQ Deca XP Mass spectrometer were performed at the University of Pittsburgh Genomics and Proteomics Core

Laboratory.

3.1.3 Results and Discussion:

SPA is considered the matrix of choice for analysis of large molecular weight proteins. CHCA is usually reserved for the analysis of peptides less than

10,000 Da (Mann and Talbo, 1996). A SELDI-MS acquisition protocol developed to detect higher molecular weight proteins was used to analyze the effectiveness of saturated and 50% solutions of SPA as well as saturated solutions of CHCA in generating mass spectra with the mitochondrial protein samples. A modified

“dried droplet” method of sample application was employed. Matrix solution

(0.5L) was spotted onto a gold chip at 40ºC and allowed to air dry creating an area of matrix crystals to enhance subsequent co-crystallization of the matrix and protein. The matrix solutions were mixed with the protein sample at a 1:10

(sample to matrix) dilution and 0.5µL was applied twice to each spot with air- drying between applications. The ability of SPA and CHCA to desorb and ionize human mitochondrial brain proteins is seen in Figure 35. A low molecular weight acquisition protocol yielded similar results. The saturated CHCA matrix produced a significant improvement in the detection of brain mitochondrial proteins over

SPA matrices.

10000 20000 30000 10

7.5

5 HMW sat SPA 1% OGP MS 17-2Saturated frac 3 1:10 SPAdil 2.5

0 10 10000 20000 30000 7.5 5 50%HMW 50% Saturated sat SPA 1% OGP SPA MS 17-2 frac 3 1:10 dil 2.5

10 relative intensity relative relative intensity relative 10000 20000 30000 7.5

5 HMW sat CHCA 1%OGP MS17-2Saturated frac 3 1:10 CHCA dil 2.5

1000010000 2000020000 3000030000

m/z Figure 35: Comparison of SELDI TOF mass spectra of

mitochondrially-enriched protein samples from human brain

tissue with CHCA and SPA matrix solutions. Each solution

contained 1% OGP and 0.1% TFA with a 1:9 (sample:matrix)

dilution.

Formic acid is also used as a formulation enhancer. Figure 36 shows the results in detecting mitochondrial proteins using formic acid as a solubilizer. The formic acid matrix solution consisted of 50% acetonitrile, 20% formic acid, 20% isopropanol, and 1% OGP. The OGP-only matrix solution contained 50% acetonitrile, 0.1% TFA acid and 1% OGP. Sample application used the modified dried droplet method. The peak at approximately 1000 Da in the spectra obtained with the formic acid matrix solution may represent acid catalyzed protein hydrolysis products. Regardless, in this study, formic acid solubilization offers no additional protein detection advantage over non-ionic detergent solubilization.

10000 20000 30000 10

7.5

5 Formic20% Acid FA 2.5

0 10 10000 20000 30000 7.5

1%OGP relative intensity relative relative intensity relative 1% OGP 2.5

1000010000 2000020000 3000030000 m/z

Figure 36: Comparison of SELDI TOF mass spectra of

mitochondrially-enriched protein samples from human brain

tissue obtained with formic acid (FA) and OGP as solubilizing

agents. The formic acid matrix solution contained 50%

acetonitrile, 1% OGP, 20% formic acid and 20% isopropanol. The

OGP solution contained 50% acetonitrile, 0.1% TFA and 1% OGP.

With an effective matrix and solubilizing agent chosen, the detergent concentrations that provided optimum protein detection could be defined. The ideal acquisition conditions detect a maximum number of peaks at the highest resolution. These two indicators were used to select the optimum conditions for detection of brain mitochondrial proteins. Auto detection in Ciphergen

ProteinChip® software with a sensitivity of S/N >4.4 and manual spectral inspection was used to select peaks in each mass spectrum. Initial trials using 1:3 sample to matrix dilutions did not yield a signal (data not shown). Additional sample to matrix dilutions of 1:5, 1:10 and 1:15 were examined to establish the

100

ideal concentration of the protein sample. Table 6 summarizes the effects of protein sample dilution and OGP concentration on the number of detected peaks and their resolution while Figure 37 displays representative SELDI mass spectra.

While the 1.0% OGP containing matrix produced the maximal mean resolution

(835.8) and the greatest number of detected peaks (115), the dependence of these performance measures on protein sample dilution was pronounced. The 0.6%

OGP containing matrix was more consistent across all dilutions, giving resolutions that varied from 660 to 697 and detected between 91 and 100 protein peaks, performance measures that are close to the maximum mean resolution and maximum number of peaks detected.

101

10000 20000 30000 10

7.5

5 0.0%HMW 0% OGPOGP, 1:5 dil 1:5 2.5

0 10 10000 20000 30000 7.5

5 HMW0.6% 0.6% OGP OGP, 1:5 dil 1:5 2.5

0 10 10000 20000 30000 7.5

5 HMW sat CHCA 1%OGP relative intensity relative relative intensity relative MS17-21.0% frac OGP 3 1:5 dil 1:5 2.5

0 10 10000 20000 30000 7.5

5 HMW1.5% 1.5% OGP OGP dil 1:5 1:5 2.5

1000010000 2000020000 3000030000 m/z

Figure 37: Representative SELDI TOF mass spectra of mitochondrially-enriched protein samples from human brain tissue obtained with 0%, 0.6%, 1.0%, 1.5% OGP. The enhancement seen in the 1.0% OGP spectrum is varied and 0.6%

OGP offered the most consistent improvement in SNR and peaks detected.

102

Table 6: The effect of OGP concentration and sample:matrix dilution

factor on protein ionization, desorption and detection.

total number % sample:matrix of mean OGP dilution peaks resolution 0 1:5 8 105.5 1:10 62 718.5 1:15 44 849.4 0.6 1:5 93 697.6 1:10 91 660.2 1:15 100 683.7 1.0 1:5 46 315.2 1:10 115 835.8 1:15 78 754 1.5 1:5 74 755.3 1:10 103 779.1 1:15 73 682.5

In order to confirm that the optimized sample-matrix formulation enabled the detection of mitochondrial proteins, proteins were purified by SDS polyacrylamide gel electrophoresis and bands corresponding to protein peaks observed in the mass spectrum were excised and identified by ESI LTQ MS/MS analysis of their trypsin digests. These included COX5b (COX5b_human) from cytochrome c oxidase, NADH dehydrogenase (NDUA5_human) and ubiquinonol- cytochrome c reductase (QCR7_human). The presence of COX5b in the protein sample was confirmed by western blotting and its detection in the mass spectrum was additionally confirmed by a novel COX5b antibody pull-down analysis (see

Chapter 3).

This optimized sample-matrix formulation increased the number of proteins detected in mitochondrial fractions as compared to earlier reports

(Nyblom, 2006) in which mitochondrial proteins isolated from a single cell type,

103

INS-1E cells, were analyzed using a traditional saturated SPA matrix solution with 50% acetronitrile and 0.5%TFA. This formulation failed to detect any proteins in our mitochondrially-enriched protein samples isolated from human brain tissue which is a considerably more complex and heterogenous tissue than

INS-1E cells. Varying the detergents, matrices and sample dilutions employed in sample-matrix formulations can increase the number of proteins detected and their mass spectral resolution. It is unlikely that a single sample-matrix formulation will allow the detection of all proteins in a complex protein mixture, and instead a strategy of employing different sample-matrix formulations, each of which allows a different set of proteins to be detected, may enable the detection of a larger set of proteins.

Section 2: Synthesis of a novel matrix designed for detection of human brain mitochondrial proteins using SELDI TOF MS

3.2.1 Introduction:

4-Chloro-cyanocinnamic acid is an advanced, rationally designed MALDI matrix (Jaskolla, 2008). Molecules which enhance peptide detection in peptide fingerprint mapping experiments are requisite. Enhanced gas phase ionization increases the detection of peptides and their resolution. The specific mechanisms involved in gas phase ionization are currently under investigation but by decreasing proton affinity of the matrix molecule gas phase ionization may be improved (Jaskolla, 2008). Figure 38 shows the structure of 4-chloro- cyanocinnamic acid and its electron withdrawing potential. The electron

104

withdrawing properties of the chlorine bound to the benzene ring increase the acidity of the molecule possibly allowing it to more readily donate a proton and thereby increase gas phase ionization. To investigate this theory, Jaskolla. synthesized several matrix molecules and tested their performance. Chloro- cyanocinnamic acid when compared with CHCA increased both the peak intensity and signal to noise ratio in BSA peptides generated by tryptic digestion.

Additionally, spectra acquired using the new matrix contained peptides previously unseen in MALDI experiments, thereby increasing the sequence coverage of the protein. Sequence coverage is a parameter used in calculating the probability of identity of a specific protein in a peptide finger print mapping experiment. To explore the effects this matrix may have on brain proteomics, 4-chloro- cyanocinnamic acid was synthesized and its performance evaluated against α-

CHCA.

Figure 38: Structure of α-cyanocinnamic acid and chloro-

cyanocinnamic acid. The structural differences of the two

chemicals are highlighted in red. The electron withdrawing

105

properties of the chlorine substituent may contribute to the proton

donating characteristics of chloro-cyanocinnamic acid.

3.2.2 Methods:

The mechanism of the formation of Cl-CHCA is a Knoevenagel condensation, a modified aldol condensation. This reaction occurs between an aldehyde (or ketone) and an active hydrogen compound with electron withdrawing functional groups bond to the target carbon atom. Nucleophilic addition is catalyzed by a slightly basic amine which is followed by dehydration to generate the conjugated enone. Briefly, 4-chlorobenzaldehyde, cyanoacetic acid and ammonium acetate (Sigma Chemicals) were refluxed in toluene to produce chloro α-cyano-4-hydroxycinnamic acid (Cl-CHCA). One equivalent of

4-chlorobenzaldehyde was combined with one equivalent of cyanoacetic acid and

0.15 equivalents of ammonium acetate. The mixture was refluxed at 220°C in toluene for three hours. The mixture was returned to room temperature and stored at 4°C overnight. Crystals were harvested by filtration from the 4°C liquor. The crystals were washed with ice cold toluene, redissolved in warm methanol:water

(1:1) solution and warm filtered to remove particulates. The filtrate was transferred to a beaker and allowed to recrystallize at room temperature. The crystals were isolated by filtration and washed in ice cold methanol:water. This recrystallization process was repeated twice.

Protein samples consisted of mitochondrial enriched fraction from control brains which had been anion exchanged to contain proteins and peptides with a pI between 5 and 7. Each sample was mixed with saturated matrix (CHCA or Cl-

106

CHCA) solution consisting of 50% ACN and 0.5% TFA by volume with 0.6%

OGP by weight in a 1:5 dilution. 1 μL of this solution was applied to a spot on a gold SELDI chip in a randomized manner and allowed to air dry. Mass spectra were acquired using a model PBSIIc SELDI-TOF mass spectrometer manufactured by Ciphergen Biosystems. The optimal sensitivity and laser intensity was established on a location not used in transient averaging. These parameters were then used to develop a spot protocol. This same protocol was applied to all chips. Data was acquired at an approximate digitizer rate of 250.0

MHz, in positive ion mode with a chamber vacuum of less than 5×10−07 Torr. The source voltage and detector voltage were set at 20 kV and 2700 V, respectively.

Approximately 65 transients spectra were averaged for each spectrum. All spectral processing (smoothing and baseline subtraction) was performed with

Proteinchip 3.1 Software. In order to make appropriate comparisons, all spectra were normalized against the total ion current (Avasarala, 2005). Peaks used in statistical analysis were present in all spectra.

3.2.3 Results:

Both 13C NMR and 1H NMR of the purified Cl-CHCA crystals were acquired on a Bruker NMR using DMSO-d6 as a solvent. The spectra are presented in Figure 39. The 13C NMR spectra indicate a total of seven different carbon species, which agrees with both the structure and the theoretical spectral analysis. The 1H NMR spectra contain 4 species with chemical shifts in alignment with the theoretical spectra and those published by Jaskolla. The 1H and 13C NMR

107

spectra confirm that the intended product was indeed synthesized and further characterization of its ability to enhance the detection of peptides and proteins in a

MALDI experiment was conducted.

Figure 39: 1H and 13C NMR spectra of the purified chloro-

cyanocinnamic acid. Panel A shows the actual 13C NMR spectra

of chloro-cyanocinnamic acid. Panel B is the 1H NMR spectra of

chloro-cyanocinnamic acid. All carbon and hydrogen species are

accounted for in the spectra indicating that the synthesis was

successful.

Representative SELDI TOF mass spectra are shown in Figure 40, below.

The spectra acquired using Cl-CHCA contained enhanced spectral peak intensities between 4 and 6 kDa m/z. Improvement in spectral peak mass resolution continues until approximately 12,000 m/z. However, above 12,000 m/z the performance of the matrices as measured by mass resolution and signal to noise ratio (SNR) was nearly identical. Figure 41 contains a graphical

108

representation of this performance trend showing the mean SNR of peaks common to both populations from four trials of both matrices. Error bars indicate the standard deviation for each peak intensity. Homoscedastic two-tailed

Student’s t-tests were performed on the SNR and mass resolution for each spectral peak. Table 7 summarizes the results which again indicate that for the majority of peaks the most pronounced effects take place at the lower m/z. This trend supports the use of this matrix for peptide fingerprint mapping of tryptic fragments for protein identification but not detection of intact proteins.

Figure 40: Representative SELDI spectra acquired using both

CHCA and Cl-CCA. Panel A shows a significant enhancement in

the 4000-6000 m/z range. Panel B also shows a significant

improvement in spectral intensity over the 6000-12000 m/z range.

Cl-CHCA shows no improvement when compared to CHCA in

ranges of m/z greater than 12,000 as indicated in Panel C.

109

Figure 41: Signal to Noise ratio comparison of peaks detected using CHCA and Cl-CHCA. The enhancement of Cl-CHCA is most pronounced in the low molecular weight range less than 5000 m/z.

110

Table 7

p values generated by Homoscedastic two-tailed Student’s t-

tests comparing CHCA and Cl-CCA. At the peaks listed, there

was a statistical difference between CHCA and Cl-CCA if the p

value listed was less than 0.05. (N=3)

p value (α=0.05) m/z SNR resolution 4782 0.003 0.007 5029 0.005 2.76E-05 5079 0.026 0.019 5347 0.032 0.246 7987 0.526 0.0006 10043 0.006 0.025 10138 0.248 0.014 10649 0.038 0.083 12446 0.600 0.340 13491 0.846 0.314 16013 0.567 0.040 16916 0.566 0.775 18199 0.460 0.106 18480 0.095 0.087 20911 0.058 0.032

The shelf life of this matrix was also investigated. Results with fresh Cl-

CCA matrix showed a clear enhancement of signal below 12,000 m/z. However, spectra acquired approximately ninety days after synthesis do not demonstrate such results. Figure 42 displays the SELDI mass spectra acquired using Cl-CHCA and CHCA matrices which were stored at room temperature, protected from light for ninety days. Clearly, the improvements seen previously are missing. New

NMR spectra were acquired (Figure 43), in order to demonstrate the absence of

111

new species due to oxidation or degradation. Indeed, when compared to Figure

39, the spectra are nearly identical in relative abundance, coupling and chemical shift. Different storage conditions could improve the shelf life of the Cl-CHCA but were not investigated in this research due to the inability of Cl-CHCA to enhance the detection of proteins and peptides in the target range (10-50kDa).

Figure 42: SELDI spectra acquired using stored Cl-CHCA

matrix. In a second experiment, using a similar protein sample,

CHCA was compared to Cl-CHCA and the enhanced SNR and

resolution were not observed.

112

Figure 43 : 13C and 1H NMR spectra acquired using stored Cl-

CHCA matrix. Spectra acquired after a ninety day storage

indicate no chemical changes have taken place due to oxidation or

degradation in the Cl-CHCA.

Cl-CHCA has been used in a variety of MALDI TOF experiments to enhance the detection of glycosylated peptides (Carulli, 2011), lipids (Fuchs,

2010) and tryptic peptides (Meyer, 2011). The increase in identification certainty attributed to Cl-CHCA has been cited (Papasotiriou, 2010) attesting to its impact on the detection of molecules analyzed by UV laser MALDI TOF. However, all citing authors used freshly prepared matrix, further verifying its tendencies toward decomposition. The improvements offered by this matrix are profound and efforts to characterize and increase its stability are warranted.

113

Section 3: Affinity Capture for the Verification of PFM Identifications

3.3.1 Introduction:

Surface enhanced laser desorption ionization time of flight mass spectrometry (SELDI TOF MS) is a high throughput label free discovery proteomics tool which can analyze larger sample sizes (Simpson, 2009). The

SELDI platform does not have protein identification capabilities and must be used in conjunction with a tandem mass spectrometer. A peak of interest can be purified prior to analysis by tandem mass spectrometry. Purification methods of the above work consist of gel electrophoresis, ion exchange or reverse phase chromatography. HPLC purification is time consuming and one dimensional gel electrophoresis has a low resolving powers which often generate more than one confident result from database perturbations. A high throughput, highly specific method capable of verifying protein identification in complex samples would provide important information for differential proteomic investigations.

Antibodies are known to be precise in protein target detection and have been used extensively to enhance specificity (Whiteaker, 2007). Whiteaker achieved a signal enhancement of three orders of magnitude by employing an antibody enrichment step prior to mass spectrometry. Coupling this affinity with mass spectrometry allows for accurate molecular weight determinations of multiple protein forms. Sparbier has reviewed the applications of antibodies to mass spectrometry, (Sparbier, 2009) including a discussion on the relevance of immune-MALDI-TOF MS to clinical proteomics. One limitation of the typical

114

mass spectrometry system is sample processing. The initial sample preparation phase employing capillary electrophoresis or liquid chromatography has low throughput and consequently alternative methods need to be developed to optimize mass spectrometry detection in differential proteomic applications.

Derivatized affinity surfaces on beads (Robijn, 2007) and MALDI target plates

(Dunn, 2008) have provided a platform for high throughput analysis while still utilizing increased specificity and mass spectrometry sensitivity.

Initial studies published by Meuwis (Meuwis, 2008) indicate the applicability of SELDI to high throughput detection of specific molecules. An antibody to platelet aggregation factor 4 (PF4) was used to deplete serum which was then analyzed by SELDI TOF MS. The resulting spectra showed a decrease or absence of the PF4 peak. This indirect detection method is intended to quickly confirm the presence of a specific protein in numerous complex serum mixtures.

Further applications of this technique would include identification of truncated or covalently modified forms of a specific protein.

This research extends the serum depletion work of Meuwis by applying the method to tissue samples and post translational modifications, specifically nitrotyrosine. The assay is further characterized by investigating sensitivity, reproducibility and bead binding capacity. A detailed analytical assessment of this antibody-bead depletion method using mass spectrometry detection provide additional tools for the quick, high throughput confirmation of protein identity, regardless of modification.

115

3.3.2 Methods

3.3.2.1 Tissue Preparation:

Human brain tissue of a known weight was homogenized in buffer A (20mM

KCl, 3mM MgCl2, 10mM Hepes pH 7.9, 0.5% NP-40, 5% glycerol, protease inhibitor cocktail) using a Wheaton homogenizer with Teflon® pestle. After a 10 minute incubation, tissue was centrifuged for 10 minutes at 500xg in 4ºC. The pellet was stored at -80ºC. The supernatant was spun at 10,000xg for 30 minutes in 4ºC. The supernatant was stored at -80ºC. This pellet, the mitochondrially enriched fraction, was further purified by washing in 20mM PBS pH 7.4 twice.

The pellet was lysed in buffer B (50mM Tris, 7M urea, 3% CHAPS with protease inhibitors) for 20 minutes at 20ºC. The mitochondrial lysate was centrifuged for

10 minutes at 10,000xg at 4ºC. BioRad DC protein assay (PN 500-0111) was used to quantify the supernatant protein concentration. Samples were diluted with

50mM Tris buffer pH 8.0 to a standard concentration and store at -80ºC.

The mitochondrial proteins were fractionated using anion exchange chromatography in a spin column (Millipore PN UFC3OHV00) bead format using

Q ceramic HyperD® F beads (Pall Life Sciences PN 20066-031). Protein separation yielded six fractions having isoelectric point ranges of greater than 9,

7-9, 5-7, 4-5, 3-4 and below 3. All samples were stored on ice immediately after centrifuging and transferred to -80ºC until further analysis.

116

3.3.2.2 Standard Protein Solution Preparation: Nitrated ribonuclease A was prepared using ribonuclease A from bovine pancreas (Sigma R-5125) according to Riordan (Riordan, 1972) with slight modifications. Tetranitromethane (Fisher Scientific, NC9690825) (TMN) in one hundred molar excess was used to nitrate the six tyrosine residues of ribonuclease

A. Ribonuclease A was solubilized in 50mM Tris pH 8.0, while TMN in 95% ethanol was added dropwise. The reaction proceeded at room temperature for one hour. A PD10 column, charged with 50mM Tris at pH 8.0, was used to separate unreacted TMN. The extent of modification was measured using

-1 -1 spectrophotometry at 428nm (at pH 8.0 ε nitrotyrosine = 4100M cm ). A standard mixture of native proteins, serving as the negative control, was mixed at 1.1mM cytochrome c (Sigma, C-7752), 6.3mM hen egg white (HEW) lysozyme (Sigma,

L-6876) and 3.5mM whale sperm myoglobin (Miles Laboratories, 95-061-13), and prepared in 10mM phosphate buffered with 150mM NaCl and 0.5% Triton X-

100 (Sigma, X-100). Varying concentrations of the nitrated ribonuclease A positive control standard were added to this mixture. Of this standard control solution, 1 µL was spotted randomly onto a Proteinchip® and the remaining solution was incubated with nitrotyrosine affinity beads. After centrifugation, 1µL of the resulting supernatant was spotted randomly onto an NP20 Proteinchip®.

3.3.2.3 Bead Derivatization:

117

Antibodies against nitrotyrosine (Upstate PN-05-233) and cytochrome c oxidase subunit 5b (COX5b Proteintech Group PN 11418-2-AP) were derivatized to 3m diameter carboxylated polystyrene microspheres (Polysciences PN 09850) using a modified carbodiimide method (Hermanson, 2008). Briefly, the beads were washed in 50mM MES (2-(N-morpholino)ethanesulfonic acid, Sigma PN

M8250) twice. A 200mg/mL solution of EDC (N-Ethyl-N′-(3- dimethylaminopropyl) carbodiimide hydrochloride Fluka PN-03449) in 50mM

MES was prepared. The washed beads (100 μL) were incubated with 2 L of

EDC solution (to a final concentration of 4mg/mL) for approximately 2 minutes in a vortexer set on low speed. Due to low volumes end over end mixing was not possible. A volume of protein was added such that the micrograms of protein were equivalent to the 0.4volume of bead slurry (200g of protein to 500l bead slurry). Protein and beads were incubated with constant mixing at room temperature for 3 hours. Beads were washed in 10mM Tris pH 8.0 0.5%BSA twice. Beads were stored in 150L 10mM Tris pH 8.0 0.5%BSA at 4°C.

Brightfield microscopy on a Leica microscope using a 40x brightfield objective was used to confirm low self-aggregation and antigen-antibody activity of conjugated beads.

3.3.2.4 Bead-Sample Incubation:

Prior to incubations, aliquots of each sample were reserved for SELDI

TOF MS analysis. Nitrated standard solutions and mitochondrial protein extracts

118

(2.5L) were incubated with varying volumes of beads in 10mM Tris pH 8.0

0.5%BSA to a final volume of 10L for 1 hour with constant mixing at room temperature. The supernatants were removed for further analysis.

3.3.2.5 SELDI-TOF-MS:

standard solution, nitrated ribonuclease A and negative control proteins, were spotted (0.5µL) randomly onto a Proteinchip®. After incubation, 0.5µL of supernatant was spotted, in a random fashion, onto a Proteinchip®. After sample application, matrix consisting of 50% saturated sinapinic acid in 50% acetonitrile and 0.5% trifluoroacetic acid was applied in two separate aliquots with air drying between. Mitochondrial protein extracts were diluted 1/10 in a saturated -cyano-

4-hydroxycinnamic acid (CHCA; Sigma PN 145505-5g) aqueous solution consisting of 50% acetonitrile (Fisher PN A996-4), 1% octyl -D- glucopyranoside (Sigma PN O8001-5g) and 0.5% trifluoroacetic acid (JT Baker

PN9470-00). This analyte solution was then randomly spotted onto an NP-20

Proteinchip® (BioRad Laboratories PN C57-30043) at 0.5L, air-dried and repeated. Mass spectra were acquired using a SELDI-TOF MS, model PBSIIc, manufactured by Ciphergen Biosystems (Fremont, CA). The optimal sensitivity and laser intensity were established on a location not used in transient averaging.

These values were then used to develop a spot protocol designed to optimize detection at 10-20kDa. This same protocol was applied to all chips. Data were acquired at a digitizer rate of 250 megahertz in positive ion mode with a chamber

119

vacuum of less than 5.0 E-07 torr. The source voltage was 20 KV and the detector voltage was 2,700 volts. A total of 65 transients were averaged for each spectrum.

3.3.2.6 Spectral Processing:

Spectral normalization by total ion current, smoothing and baseline subtraction was performed using Proteinchip 3.1 Software. Calibration equations were developed using hirudin BHVK (6,964 Da), bovine cytochrome c (12,230

Da) and equine cardiac myoglobin (16,952 Da) mass standards for all brain tissue experiments. The negative control proteins, cytochrome c from equine heart

(12384Da), HEW lysozyme (14313Da) and whale sperm myoglobin (17200Da) were used as internal calibrants for the standard nitration protein mixture investigations. Peaks were picked manually with a signal to noise ratio greater than 4.0.

3.3.2.7 Statistical Analysis:

Microsoft® Excel was used to calculate linear regression, Pearson’s product-moment correlation coefficient, one-tailed paired Students t-test and single factor ANOVA. All tests were performed at an alpha level of 0.05. The coefficient of variance (%CV) was used to compare reproducibility.

3.3.3 Results:

120

Data acquisition was set up in a paired fashion such that a specific aliquot taken before affinity bead incubation was matched to the aliquot removed after incubation. The nitrated standard protein experiments were designed to test the reproducibility of the assay with a known protein composition. Figure 44 illustrates several aspects of the initial phases of this study. Firstly, the dependence of spectral intensity on concentration is demonstrated for the nitrated ribonuclease A standard protein. The signal of the nitrated ribonuclease A prior to affinity bead exposure over a concentration range of 35ng/µL to 426/µL shows a direct correlation to the amount of nitrated protein in solution with a linear response of y=0.023x + 1.89 and a Pearson’s product-moment correlation coefficient of 0.99. The chart in Figure 45 illustrates this point. Additionally, the intensity of the nitrated ribonuclease A protein appears decreased after exposure to the affinity beads, while the intensity of the remaining species does not change, confirming the nitrated ribonuclease A does indeed possess an affinity for the anti-nitrotyrosine derivatized beads while the negative control species do not. A student’s single tailed paired t-test yielded a p value of 0.009, validating the hypothesis that the intensity after exposure to the affinity beads was indeed different than before exposure. This experiment served as a proof of concept and a more in-depth characterization of reproducibility and response followed.

121

Figure 44: SELDI spectra of nitrated ribonuclease A (1) and hen egg white lysozyme (2): The signal intensity of nitrated ribonuclease A increases as the concentration of the enzyme increases while the intensity of the lysozyme is relatively stable.

The intensity of the nitrated protein decreases after exposure to the antinitrotyrosine affinity beads indicating removal of the nitrated species from the solution. The incomplete removal of the nitrated

122

protein from the solution seen in the last spectra demonstrates the

capacity of the beads.

Figure 45: Nitrated Ribonuclease A peak intensity is

significantly changed after exposure to anti-nitrotyrosine

beads. The signal of the nitrated ribonuclease A is markedly

reduced after exposure to an affinity anti- Nitrotyosine beads. .

The immunoprecipitated solution differs from the untreated solution in composition and relative amounts of protein. Therefore, the actual spectral property of interest is the ratio of a standard peak to that of a peak of interest not the peak intensity. The absolute intensity of the nitrated ribonuclease A was normalized to the absolute intensity of an internal standard to calculate a ratio with which to compare the relative loss of the nitrated protein. Signal stability is one criterion which can be used to select an internal standard protein. To verify the most stable internal signal, the total ion current normalized intensities of the

123

three negative control proteins were analyzed by ANOVA at 0.05, as a function of three nitrated ribonuclease A concentrations, 71, 142 and 284 nanograms. Each measurement was performed in triplicate for all concentrations as well as before and after treatment. Lysozyme exhibited the most consistent total ion current (TIC) normalized intensity across all nitrated ribonuclease A concentrations with a %CV of 14.1% and p value of 0.08 before and 9.7% CV with a p value of 0.12 after exposure to the affinity beads, indicating that the intensity of the lysozyme is unaffected by bead exposure and there is no indication of non-specific binding. Therefore, lysozyme was used to normalize the nitrated ribonuclease A signal of before and after immunoprecipitation (IP) trials.

The detection limit of the technique is an important attribute of any assay.

As can be seen in Figure 44, top panel, the technique is able to distinguish 35ng of

NO RNase A from instrumental noise. The detection limit of this technique can be further enhanced by decreasing the total protein load of the sample solution through dilution. Decreased total protein loads typically enhance individual protein signals in desorption ionization mass spectrometry, allowing a lower detection limit (Vorderwülbecke1, 2005).

Saturation and bead capacity are other attributes which must be characterized in assays utilizing affinity properties. The last two spectra in Figure

44 demonstrate the capacity of this specific lot of beads. The last spectrum clearly indicates a saturation point by its inability to remove all of the nitrated

124

ribonuclease A protein. Using the TIC normalized intensity of the nitrated ribonuclease A after bead exposure at 426ng, the intensity correlates to approximately 300ng nitrated ribonuclease A bead capacity. However, without the use of isotopes, mass spectrometry is considered to be only semi-quantitative and therefore caution should be exercised when calculating absolute concentrations.

The analytical reproducibility of MALDI-TOF mass spectrometry has been reviewed by Albrethsen (Albrethsen, 2007) and the intraexperimental mean

CV values varied from 4% to 26%. The reproducibility of this specific technique was characterized, in triplicate, using the mean intensity values of the nitrated ribonuclease A standard. The nitrated ribonuclease A intensities were normalized to all protein species in the negative control solution. Figure 46 demonstrates the reproducibility of the assay response when nitrated ribonuclease A intensity was normalized to lysozyme intensity.

Table 8 shows descriptive statistics for the lysozyme normalized results.

The signal intensity ratio prior to affinity bead exposure has a smaller mean coefficient of variance (15%) indicating the signal does not fluctuate with the trial number or with varying concentration of nitrated analyte. The limited usefulness of the CV parameter is apparent in the after affinity bead exposure data with a mean CV of 34%. It is important to remember that as the mean approaches zero, the coefficient of variance becomes increasingly sensitive to minute changes in the mean, limiting its applicability at this extreme value. The hypothesis µbefore >

125

µafter was tested using the student’s t parameter. A one-tailed paired t-test, α=0.05, gave p values of 0.06, 0.001, and 0.006 for 71ng, 142ng, and 284ng, respectively, indicating that thes pectral intensity before the affinity bead exposure was significantly different than the spectral intensity after affinity bead exposure.

Table 8: Mean peak intensities before and after affinity bead exposure

BEFORE Affinity AFTER Affinity Amount of Bead Incubation Bead Incubation Nitrated P value Ribonuclease A Mean Mean (α=0.05) (ng) peak % CV peak % CV intensity intensity 71 0.144 29.2 0.1 12.1 0.06 142 0.462 7.2 0.042 38.3 0.001 284 0.436 8.9 0.134 52.8 0.006

126

Figure 46: Reproducibility of normalized ribonuclease A

intensity before and after exposure to anti-nitrotyrosine

affinity beads. SELDI mass spectra were acquired before and after

appropriate concentrations of proteins were incubated with affinity

beads. The experiment was performed in triplicate resulting in p

values of 0.06. 0.001 and 0.006 for 71, 142, and 284 ng

ribonuclease A, respectively.

The applicability of this assay to mitochondrially-enriched human brain proteins was also investigated. Anti-nitrotyrosine affinity beads were incubated with the mitchondrially enriched human brain protein samples and the resulting spectra can be seen in Figure 47. The reference peak at 10.1kDa was selected because its peak intensity was stable across a wide range of sample concentrations

127

and statistical analysis verified peak intensity stability. This reference peak is labeled with an arrow and was used as the normalization species in this series of experiments. The peaks whose intensity was significantly affected by incubation with the affinity beads are highlighted with asterisks. The identity of the peak at

25kDa is under investigation while the peak at 10.6kDa has been confirmed to be cytochrome c oxidase subunit 5b. The nitration of the COX5b protein in brain tissue is plausible due to the existence of peroxynitrite in neuronal tissue. Nitric oxide generated by mitochondrial nitric oxide synthase (mtNOS) and neuronal nitric oxide synthase (nNOS) is free to diffuse through membranes and is, therefore, available to react with superoxide, generated by complex III, to form peroxynitrite. Peroxynitrite then nitrates tyrosine residues, effecting protein function, sensitivity to proteolysis, and tyrosine phosphorylation (Abello, 2009).

128

Figure 47: Mining nitrated proteins from a complex brain

mitochondrially enriched sample. A donor tissue sample was

incubated with anti-nitrotyrosine affinity beads. The reference peak

at 10.1kDa is marked with an arrow. Peaks at 10.6kDa and 25kDa

are encircled, highlighting the reduction in height with respect to a

reference peak after exposure to affinity beads. Other peaks in

these spectra are not reduced with respect to the reference peak.

Cytochrome c oxidase subunit 5b (COX5b) is a mitochondrial protein known to participate in the dimerization, stability, and modulation of catalytically active cytochrome c oxidase subunits I, II and III (Fontanesi, 2006). This protein has a molecular weight of 13,696 Da. However, upon removal of the 31 amino acid signal peptide, this mass is reduced to 10,613Da. A SELDI spectrum containing a mixture of 10kDa peptides can be seen in Figure 48. Only the peak at

129

10.6kDa, indicated by the blue arrow, is affected by incubation with the anti-

COX5b affinity beads, confirming the identity of the 10.6kDa peak. One dimensional gel electrophoresis cannot fully resolve bands of within molecular mass differences of 500Da, even at high acrylamide concentrations.

Consequently, when bands are excised for LC MS/MS, the identity of several protein species may be present which can unambiguous identification of the protein of interest. A clear advantage of antibody capture coupled to mass spectrometry is the unambiguous identification of a specific peak in a mixture.

Two different volumes of beads were incubated with the same sample to explore the optimal bead concentration. Figure 48 shows results in spectra 2 and 3 which used 2µL and 3 µL of beads, respectively. This experiment was repeated with additional samples yielding similar results. The increase in bead volume did not enhance the performance of this assay.

130

Figure 48: Identify verification of the 10.6 kDa SELDI mass

spectral peak with anti-COX5b affinity beads. A control donor

tissue sample was incubated with anti-COX5b affinity beads at

varying concentrations. The blue arrow indicates the peak whose

intensity was confirmed to be significantly reduced. These spectra

also illustrate bead concentration optimization. The second

spectrum (2) is taken from a sample which was exposed to 2µL

beads as opposed to 3µL of beads in the third spectrum (3).

The anti-COX5b incubation was performed in triplicate using three distinct tissue samples (N=3) and the results can be seen in Figure 49. The p value generated from a one-tailed paired t-test is 0.01, indicating the means of the normalized intensities before affinity bead incubation are statistically significantly

131

different than the mean normalized intensities after bead incubation at the 95% confidence interval.

Figure 49: The verification of COX5b identity by affinity

capture. The COX5b peaks (N=3) are normalized to the 10.1kDa

peak and the above plot demonstrates the statistically significant

difference between the peak intensity before and after incubation

with the anti-COX5b beads.

This fast, flexible, accurate and high-throughput assay allows for the unambiguous identification of intact, as well as truncated proteins and peptides.

Additional applications include the detection of post-translationally modified proteins. Multiplex capabilities, the simultaneous detection of multiple analytes, further enhance the productivity of this assay. The characterization of disease etiology and progression requires novel techniques which improve the detection

132

of all components of the system. This systems biology approach to understanding crucial mechanisms contributes to the growing knowledge of the role of proteins in cellular dysfunction.

Chapter 4 Conclusions

The detection of mitochondrially enriched proteins in human brain initially offered challenges which prompted the development of alternative matrix formulations. Incorporation of this new matrix solution formulation into the optimized analysis protocol provided an opportunity to advance the analysis of differentially expressed proteins in NAGM from MS and control donors. The discovery of an alternative matrix which could further enhance sensitivity was investigated. The synthesis and characterization of this chlorinated derivative

(ClCCA) required four days and only offered improved sensitivity when compared to CHCA in the 5 kDa-10 kDa range. This improvement was only applicable to freshly prepared matrix and experiments conducted using older matrix offered no improvement.

Hemoglobin β was found to be differentially expressed in mitochondrially enriched protein lysates. Both fractions 4 and 6 (pIs 4-5 and < 3) in cohort 1, contained increased Hbb in MS donors with a fold value of 2.6 and 2.3, respectively. Additionally, the p values, calculated by Mann Whitney analysis, were 0.021 and 0,047, respectively. Mascot scores of 60 and 890 provide

133

134

confidence in the identification. Additional investigations by our group further confirmed this unexpected result by localizing Hb to neurons using confocal microscopy (Broadwater, 2011).

Under expression of several COX nuclear encoded subunits provides additional evidence of the energy dysregulation in pre-lesioned MS tissue.

COX5a, COX5b and COX6c are known to provide structural and biochemical support to the catalytic subunits of cytochrome c oxidase (Galati, 2009). Under- expression of these proteins could be evidence of the high turnover experienced by oxidatively damaged proteins (Fontanesi, 2006).

Creatine kinase B, the brain isozyme of the creatine kinase family, was also over expressed in MS tissue by 2 fold with a p value of 0.021. The ATP buffering capacity of creatine kinase in both the mitochondrial and brain isozymes provides a mechanism to supply energy equivalents in response to the fluctuating requirements of neurons (Wallimann, 1992). The over expression of CKB in

NAGM may be a response to the increased energy demands of early demyelinating neurons. The investigation of various isozymes of CK in a multiple sample types from MS patients and donors has generated diverse results, demonstrating the complexity of the molecular pathology of MS (Vassilopoulos,

1987; Noben, 2006).

Myelin basic protein was found to be both over expressed and under expressed in different fractions. The 15.9 kDa spectral peak from fraction 3 with proteins of pI in the 5-7 range contained an over expression of MBP. Using the

135

detected peptides, MBP1-4 could be responsible for the differential expression.

MBP is known to be a target of various proteases in MS (Bacheva, 2011) and the

15.9 kDa protein has been generated in rat brain by trypsin 4, a brain specific isozymes of trypsin (Medveczky, 2006). A 17.2 kDa spectral peak from fraction 6

(pI<3) was under expressed in MS and identified as MBP3. The full length molecular weight of MBP3 is 18.5 kDa and this peak is certainly a proteolytic fragment. The confidence of this identification is high but characterization of the

MBP fragments and the specific mechanisms which generate them would provide useful information regarding MS pathology.

Another under expressed spectral peak with high confidence identification is glial fibril acidic protein (GFAP). A Mascot score of 1757 indicates that regardless of the molecular weight of the full length active protein, this spectral peak is GFAP. A full investigation into the fragments which generate this differential expression would provide vital information into the roles astrocytes play in MS pathology.

The expression of Hbb in mitochondrially enriched fractions is a phenomenon which requires further investigation. The co-localization of Hbb to mitochondria cannot be investigated by light microscopy due to the size of the mitochondria and the resolution of light microscopy (Mobius, 2009).

Immunoelectron microscopy provides superior resolution and would be suitable for investigations of this nature (Fujioka, 2011). Several attempts to acquire images after gold immunostaining ultrathin brain sections (approximately 80nm) yielded no results. Consultation with experts revealed the brain tissue was flash-

136

frozen at autopsy and would require ultrathin cryo-sectioning of the tissue to preserve organelle structure. This equipment was not available and this project was postponed until such time as this equipment would be available.

Bibliography

Abdi. H., & Williams, L.J. (2010). "Principal component analysis.". Wiley

Interdisciplinary Reviews: Computational Statistics, 2: 433–459.

Abello, N.,Kerstjens, H.A.M., Postma, D.S., Bischoff, R. Protein Tyrosine

Nitration: Selectivity, Physicochemical and Biological Consequences,

Denitration, and Proteomics Methods for the Identification of Tyrosine-Nitrated

Proteins. J. Prot. Res. 2009, 8, 3222–3238.

Abu-Farha, M., Elisma, F., Zhou, H., Tian, R., Zhou, H., Asmer, Figey. D.,

Proteomics: From Technology Developments to Biological Applications Anal.

Chem. 2009, 81, 4585–4599

Aebersold & Mann, Mass Spectromety-Based Proteomics, Nature, 2003, 422,

198-207.

137

138

Aksenov, M. Aksenova, M., Butterfield, D.A., Markesbery, W.R., Oxidative modification of creatine kinase BB in Alzheimer’s disease brain, J. Neurochem.

2000, 74, 2520–2527.

Alberthsen, J., Reproducibility in Protein Profiling by MALDI-TOF Mass

Spectrometry. Clin. Chem. 2007, 53, 852-858.

Alizadeh, M., Babron, M.-C., Birebent, B., Matsuda, F., Quelvennec, E., Liblau,

R., Cournu-Rebeix, I., Momigliano-Richiardi, P., Sequeiros, J., Yaouanq, J.

Genin, E. Vasilescu, A., Bougerie, H., Trojano, M., Martins-Silva, B., Maciel, P.,

Clerget-Darpoux, F., Clanet, M., Edan, G., Fontaine, B., Semana, G. Genetic

Interaction of CTLA-4 with HLA-DR15 in Multiple Sclerosis Patients, Ann.

Neurol. 2003, 54, 119–122.

Anderton, B., Intermediate filaments: a family of homologous structures, Journal of Muscle Research and Cell Motility, 1981, 2, 141-166.

Andres, R.H., Ducraya, A.D., Schlattner, U., Wallimann, T., Widmera, H.R.,

Functions and effects of creatine in the central nervous system. Brain Res. Bull.

2008, 76, 329–343.

139

Avasarala, J.R., Wall, M.R., Wolfe, G.M., A distinctive molecular signature of multiple sclerosis derived from MALDI-TOF/MS and serum proteomic pattern analysis, J. Mol. Neurosci. 2005, 25, 119–126.

Bae, Y.J., Shin, Y.S., Moon, J.H., Kim, M.S. Degree of Ionization in MALDI of

Peptides: Thermal Explanation for Gas-Phase Ion Formation. J. Am. Soc. Mass

Spec., 2012, 23, 1326-1335.

Bains JS and Shaw CA. Neurodegenerative disorders in humans: the role of glutathione in oxidative stress-mediated neuronal death. Brain Res Brain Res Rev

1997; 25: 335–358.

Becker L, Liu N-C, Averill MM, Yuan W, Pamir N,. Unique Proteomic

Signatures Distinguish Macrophages and Dendritic Cells. PLoS ONE, 2012, 7, e33297

Bessman, J., Geiger, T., Transport of Energy in Muscle: The Phosphorylcreatine

Shuttle. Science, 1981, 211, 448-452.

Biagioli, M. Pinto, M., Cesselli, D., Zaninello, M., Lazarevic, D., Roncaglia, P.,

Simone, R., Vlachouli, C., Plessy, C., Bertin, N., Beltrami, A., Kobayashi, K.,

Gallo, V., Santoro, C., Ferrer, I., Rivella, S., Beltrami, C.A., Carninci, P., Raviola,

E., Gustincich, S., Unexpected expression of α- and β-globin in mesencephalic

140

dopaminergic neurons and glial cells, Proc. Natl Acad. Sci. USA, 2009, 106,

15454–15459.

Bird GH, Lajmi AR, Shin JA; Manipulation of temperature to improve solubility of hydrophobic proteins and co-crystallization with matrix for analysis by

MALDI-TOF mass spectrometry; Anal. Chem., 2002, 74, 219-225.

Bittl, J.A., Weisfeldt, M.L., Jacobus, W.E., Creatine-Kinase Of Heart-

Mitochondria- The Progressive Loss Of Enzyme-Activity During In vivo

Ischemia And Its Correlation To Depressed Myocardial-Function. J. Biol.

Chem. 1985, 260, 208-214.

Brinkwort CS, Bourne DJ; The effects of Tween 80 on signal intensity of intact proteins by MALDI time of flight mass spectrometry. J. Am. Soc. Mass Spec.

2007, 18, 102-105.

Broadwater, L., Pandit, A., Clements, R., Azzam, S., Vadnal, J., Sulak. M., Wee

Yong, V., Freeman, E.J., Gregory, R.B., McDonough, J., Analysis of the

Mitochondrial Proteome in Multiple Sclerosis Cortex, Biochim. et Biophys.

ACTA, 2011, 1812, 630–641.

141

Cadene, M, Chait, BT; A Robust, Detergent-Friendly Method for Mass

Spectrometric Analysis of Integral Membrane Proteins. Anal.Chem., 2000, 72,

5655-5658.

Campbell GR, Ziabreva I, Reeve AK, Mitochondrial DNA deletions and neurodegeneration in multiple sclerosis. Ann. Neurol. 2011, 69, 481–492.

Capaldi, R.A., Structure and assembly of cytochrome c oxidase, Arch. Biochem.

Biophys., 1990, 280, 252–262.

Carlson, S.M., Najmi, A., Whitin, J.C., Cohen, H.J., Improving feature detection and analysis of surface-enhanced laser desorption/ionization-time of flight mass spectra, Proteomics, 2005, 5, 2778–2788.

Carulli, S., Calvano, C.D., Palmisano, F. Pischetsrieder; M., MALDI-TOF MS

Characterization of Glycation Products of Whey Proteins in a Glucose/Galactose

Model System and Lactose-free Milk. J. Agric. Food Chem., 2011, 59, 1793–

1803.

Casaccia-Bonnefil, P., Pandozy, G., Mastronardi, F., Evaluating Epigenetic

Landmarks in the Brain of Multiple Sclerosis Patients: A Contribution to the

Current Debate on Disease Pathogenesis, Progress in Neurobiology, 2008, 86,

368–378.

142

Choi IY, Lee SP, Denney DR, Lower levels of glutathione in the brains of secondary progressive multiple sclerosis patients measured by 1H magnetic resonance chemical shift imaging at 3 T. Mult Scler 2011, 17, 289–296.

Cohen, SL, Chait, BT; Influence of Matrix Solution Conditions on the MALDI-

MS analysis of peptides and proteins. Anal. Chem. 1996, 68, 31-37.

Cravatt, B.F., Simon, G.M., Yates, J.R., The biological impact of mass- spectrometry-based proteomics. Nature. 2007, 450, 991- 999.

Compston A, Coles A., Multiple Sclerosis. Lancet, 2008, 372, 1502–1517.

Cuzner M.L., Davison A.N., Changes in cerebral lysosomal enzyme activity and lipids in multiple sclerosis. J. Neurol. Sci. 1973, 19, 29–36.

Derfussa,T., Parikhc, K., Velhina, S., Brauna, M., Matheyc, E., Krumbholza, M.,

Kumpfelb, T., Moldenhauerd, A., Radere, C., Sondereggere, P., Pollmannf, W.,

Tiefenthallerg, C., Bauerg, J., Lassmanng, H., Wekerlea, H., Karagogeosh, D.,

Hohlfelda, R., Liningtonc, C., Meinl, E., Contactin-2/TAG-1-Directed

Autoimmunity is Identified in Multiple Sclerosis Patients and Mediates Gray

Matter Pathology in Animals, Proc. Natl Acad. Sci., 2009, 106, 208302–208307.

143

Desmazières, A., Sol-Foulon, N., Lubetzki, C., Changes at the Nodal and

Perinodal Axonal Domains: a Basis for Multiple Sclerosis Pathology? Multiple

Sclerosis Journal 2012, 18, 133–137.

Dolder, M., Walzel, B., Speer, O., Schlattner, U., Wallimann, T., Inhibition of the

Mitochondrial Permeability Transition by Creatine Kinase Substrates:

Requirement For Microcompartmentation. J. Bio. Chem. 2003, 278, 17760–

17766.

Dowling JE. Neurons and Network: An Introduction to Neuroscience. Cambridge,

MA: Harvard Univ. Press, 1992, p.44.

Dreisewerd, K., The Desorption Process in MALDI. Chem. Rev., 2003, 103, 395-

425.

Dunn, J.D., Igrisan,E.A., Palumbo, A.M., Reid, G.E., Bruening, M.L.,

Phosphopeptide Enrichment Using MALDI Plates Modified with High-Capacity

Polymer Brushes. Anal. Chem. 2008, 80, 5727-5736.

Dutta, D., McDonough, J., Yin, X., Peterson, J., Chang, A., Torres, T., Gudz, T.,

Macklin, W., Lewis, D.A., Fox, R.J., Rudick, R., Mirnics, K., Trapp, B.D.,

Mitochondrial Dysfunction as a Cause of Axonal Degeneration in Multiple

Sclerosis Patients, Ann. Neurol. 2006, 59, 478–489.

144

Dutta, R., Trapp, B.D., Gene Expression Profiling in Multiple Sclerosis Brain,

Neurobio. Dis., 2012, 45, 108–114.

Echan, L.A., Speicher, D.W., Protein detection in gels using fixation, Curr.

Protoc. Protein Sci. 2002, 10.5.3–10.5.4.

Eisen, M.B., Spellman, P.T., Brown, P.O., Botstein, D., Cluster Analysis and

Display of Genome-Wide Expression Patterns, Proc. Natl Acad. Sci. USA 1998,

95, 14863–14868.

El-Aneed , Aljandro, Cohen, J., Banoub, J., Mass Spectrometry, Review of the

Basics: Electrospray, MALDI, and Commonly Used Mass Analyzers, Applied

Spectroscopy Reviews, 2009, 44, 210-230.

Emerson, J.E., Marisa Dolled-Filhart, Lyndsay Harris, David L. Rimm and David

P. Tuck. Quantitative Assessment of Tissue Biomarkers and Construction of a

Model to Predict Outcome in Breast Cancer Using Multiple Imputation. Cancer informatics, 2009, 7, 29-40.

Fazeli, A.S., Nasrabadi, D., Sanati, M.H., Pouya, A., Ibrahim, S.M., Baharvand,

H., Salekdeh, G.H., Proteome Analysis of Brain in Murine Experimental

Autoimmune Encephalomyelitis, Proteomics, 2010, 10, 2822–2832.

145

Fontanesi, F., Soto, I.C., Horn, D., Barrientos, A., Assembly of mitochondrial cytochrome c oxidase, a complicated and highly regulated cellular process. Am. J.

Physiol. Cell. Physiol. 2006, 291, C1129-C1147.

Fuchs,B., Sus, R., Schiller, J., An update of MALDI-TOF mass spectrometry in lipid research. Progress in Lipid Research, 2010 49, 450–475.

Fujioka, H., Moghaddas, S., Murdock, D.G., Lesnefsky, E.J., Tandler, B., Hoppel,

C.L., Decreased cytochrome c oxidase subunit VIIa in aged rat heart mitochondria: Immunocytochemistry. Ann. Rec. 2011, 294, 1825-1833.

Galati, D., Srinivasan, S, Raza, H., Prabu, S.K., Hardy, M., Chandran, K., Lopez,

M., Kalyanaraman, B., Avadhani, N.G., Role of nuclear-encoded subunit Vb in the assembly and stability of cytochrome c oxidase complex: implications in mitochondrial dysfunction and ROS production, Biochem. J., 2009, 420, 439–

449.

Gharahdaghi, F; Kirchner, M; Fernandez, J; Mische, SM. Peptide-Mass Profiles of Polyvinylidene Difluoride-Bound Proteins by Matrix-Assisted Laser

Desorption/Ionization Time of Flight Mass Spectrometry in the Presence of

Nonionic Detergents. Anal. Biochem. 1996, 233, 94-99.

146

Gordon SR. Use of selected excitation filters for enhancement of diaminobenzidine photomicroscopy. J. Histochem. Cytochem. 1988, 36, 701-4.

Gomes-Alves, P., Imrie, M. Gray, R.D., Nogueira, P., Ciordia, S., Pacheco, P.,

Azevedo, P., Lopes, C., Bugalho de Almeida, A., Guardiano, M., Porteous, D.J.,

Albar, J.P., Boyd, A.C.,

Penque, D. SELDI-TOF biomarker signatures for cystic fibrosis, asthma and chronic obstructive pulmonary disease. Clin. Biochem. 2010, 43, 168–177.

Graumann, U., Reynolds, R., Steck, A.J., Schaeren-Wiemers, N., Molecular

Changes in Normal Appearing White Matter in Multiple Sclerosis are

Characteristic of Neuroprotective Mechanisms Against Hypoxic, Insult Brain

Pathol., 2003, 13, 554-573.

Grossmann, J., Roschitzki, B., Panse, C., Fortes, C., Barkow-Oesterreicher, S.,

Rutishauser, D., Schlapbach, R., Implementation and evaluation of relative and absolute quantification in shotgun proteomics with label-free methods. Journal Of

Proteomics, 2010, 73, 1740-1746.

Han, M.H., Hwang, S.-I., Roy, D.B., Lundgren, D.H. , Price, J.V., Ousman, S.S.,

Fernald, G.H., Gerlitz, B., Robinson, W.H., Baranzini, S.E., Grinnell, B.W.,

Raine, C.S., Sobel, R.A., Han, D.K., Steinman, L., Proteomic Analysis of Active

147

Multiple Sclerosis Lesions Reveals Therapeutic Targets, Nature, 2008, 451, 1076-

1083.

Harauz, G., Musse, A., A Tale of Two Citrullines- Structural and Functional

Aspects of Myelin Basic Protein Deimination in Health and Disease., Neurochem

Res 2007, 32, 137–158.

Harauz, G., Ladizhansky, V., Boggs, J., Structural Polymorphism and

Multifunctionality of Myelin Basic Protein. Biochemistry 2009, 48, 8094–8104.

Hass, C.J., Collins, M.A., Juncos, J.L., Resistance training with creatine monohydrate improves upper-body strength in patients with Parkinson disease: a randomized trial, Neurorehabil. Neural Repair, 2007, 21, 107–115.

Hassan-Smith, G., Douglas, M.R., Epidemiology and Diagnosis of Multiple

Sclerosis, British Journal of Hospital Medicine, 2011, 72, M146-M151.

Hermanson, G.T., Bioconjugate Techniques, second ed., Academic Press, San

Diego, CA, 2008.

Herold, S., Fago, A., Reactions of peroxynitrite with globin proteins and their possible physiological role, Comp. Biochem. Physiol. A Mol. Integr. Physiol.,

2005, 142, 124–129.

148

Hillencamp, F, Karas, M; Laser Desorption Ionization of Proteins with Molecular

Masses Exceeding 10 000 Daltons. Anal. Chem. 1988, 60, 2299-2301.

Inglese, M., Ge, Y., Filippi, M., Falini, A., Grossman, R.I., Gonen, O., Indirect evidence for early widespread gray matter involvement in relapsing-remitting multiple sclerosis, Neuroimage, 2004, 21 1825–1829.

Ion Source: http://www.ionsource.com/tutorial/DeNovo/b_and_y.htm, last accessed March 17, 2013.

Ishii, A., Dutta, R., Wark, G.M., Hwang, S.-I., Han, D.K., Trapp, B.D., Pfeiffer,

S.E., Bansai, R. Human myelin proteome and comparative analysis with mouse myelin, Proc. Natl Acad. Sci. USA, 2009, 106, 14605–14610.

Jaskolla, T.W., Lehmann, W.-D., Karas, M. 4-Chloro-α-cyanocinnamic acid is an advanced, rationally designed MALDI matrix. PNAS, 2008, 105, 12200–12205.

Jastorff, A. M.; Haegler, K., Maccarrone, G. Regulation of proteins mediating neurodegeneration in experimental autoimmune encephalomyelitis and multiple sclerosis., Prot. Clinical App. 2009, 3, 1273-1287.

149

Jiménez, C.R, Huang, L., Qui, Y., Burlingame, A.L., In-gel digestion of proteins for MALDI-MS fingerprint mapping, Curr. Protoc. Protein Sci. 1998 16.4.1–

16.4.3.

Jiang, Y., Wang, X., Comparative mitochondrial proteomics: perspective in human disease. J. Hematol & Oncol. 2012, 5, article 11.

Jolliffe, I.T., Principal Component Analysis, Series: Springer Series in Statistics,

2nd ed. Springer, NY, 2002.

Kammermeier, H., Why Do Cells Need Phosphocreatine and a Phosphocreatine

Shuttle? J. Mol. Cell. Cardiol. 1987, 19, 115-118.

Kay, L., Nicolay, K., Wieringa, B., Saks, V., Wallimann, T., Direct evidence for the control of mitochondrial respiration by mitochondrial creatine kinase in oxidative muscle cells in situ, J. Biol. Chem., 2000, 275 6937–6944.

Kim, J.K., Mastronardi, F.G., Wood, D.D., Lubman, D.M., Moscarello, R.Z.,

Moscarello, M.A., Multiple Sclerosis: An Important Role For Post-translational

Modifications of Myelin Basic Protein in Pathogenesis, Mol. Cell. Prot., 2003,

2:453–462.

150

Komori, M., Matsuyama, Y., Nirasawa, T., Thiele, H., Becker, M., Alexandrov,

T., Saida, T.,Tanaka, M., Matsuo, H., Tomimoto, H., Takahashi, R., Tashiro, K.,

Ikegawa, M., Kondo, T., Proteomic Pattern Analysis Discriminates Among

Multiple Sclerosis–Related Disorders. Ann. Neurol. 2012, 71, 614–623.

Kooi, E-J., Prins, M., Bajic, N., Belie, J.A.M., Gerritsen, W.H., van Horssen, J.,

Aronica, E., van Dam, A-M, Hoozemans, J.J.M., Francis, P.T., van der Valk, P.,

Geurt, J.J.G., Cholinergic imbalance in the multiple sclerosis hippocampus. Acta

Neuropathol., 2011, 122, 313–322.

Laemmli, U.K., Cleavage of structural proteins during assembly of head of bacteriophage-T4, Nature 1970, 227, 680–685.

Lassmann, H., van Horssen, J., The molecular basis of neurodegeneration in multiple sclerosis. FEBS Letters 2011, 585, 3715–3723.

Lavine, B., Workman, J., Chemometrics, Anal. Chem. 2010, 82, 4699–4711.

Liebler, Daniel C., Introduction to Proteomics: Tools for the New Biology;

Humana Press, Totowa, NY, 2002, p.61.

151

Lu, C., Ross D. King, R.D. An investigation into the population abundance distribution of mRNAs, proteins, and metabolites in biological systems.

Bioinformatics, 2009, 25, 2020–2027.

Ly, L., Barnett, M.H., Zheng, Y.Z., Gulati, T., Prineas, J.W., Crossett, B.,

Comprehensive Tissue Processing Strategy for Quantitative Proteomics of

Formalin-fixed Multiple Sclerosis Lesions, J. Proteome Res. 2011, 10, 4855–

4868.

Mahad, D., Lassmann, H., Turnbull, D., Mitochondria and disease progression in multiple sclerosis, Neuropathology and Applied Neurobiology, 2008, 34, 577–

589.

Mahad D, Ziabreva I, Lassmann H, Mitochondrial defects in acute multiple sclerosis lesions. Brain, 2008, 131, 1722–1735.

Makarov, A., Denisov, E., Lange, O., Horning, S., Dynamic Range of Mass

Accuracy in LTQ Orbitrap Hybrid Mass Spectrometer. J Am Soc Mass Spectrom

2006, 17, 977–982.

Mann M, Talbo G; Developments in matrix-assisted laser desorption ionization peptide mass spectrometry. Curr. Opin. Biotech. 1996, 7, 11-19.

152

Mannello, F. Medda, V.Tonti, G.A., Protein profile analysis of the breast microenvironment to differentiate healthy women from breast cancer patients.

Exp. Rev. Prot.. 2009, 6, 43-60.

Manos, P., Bryan, G. K., Edmund, J. J., Neurochem. 1991, 56, 2101-2107.

Matrix Science website- http://www.matrixscience.com last accessed March 17,

2013.

Medveczky P, Antal J, Patthy A, Kékesi K, Juhász G, Szilágyi L, Gráf L: Myelin basic protein, an autoantigen in multiple sclerosis, is selectively processed by human trypsin 4. FEBS Lett. 2006, 580, 545-552.

Meetani, MA, Voorhees, KJ.; MALDI Mass Spectrometry Analysis of High

Molecular Weight Proteins from Whole Bacterial Cells: Pretreatment of Samples with Surfactants. Am.Soc. Mass Spec., 2005, 16, 1422.

Meuwis , M.-A. Fillet, M. Lutteri, L., Marée , R.,Geurts P., de Seny , D. Malaise,

M., Chapelle , J.-P., Wehenkel , L., Belaiche, J., Merville, M.-P., Louis, E.

Proteomics for prediction and characterization of response to infliximab in

Crohn’s disease: A pilot study. Clin. Biochem. 2008, 41, 960-967.

153

Meyer, B., Papasotiriou, D.G., Karas, M., 100% protein sequence coverage: a modern form of surrealism in proteomics. Amino Acids, 2011, 41, 291–310.

Mobius, T., Cryopreparation of biological specimens for immunoelectron microscopy, Ann. Anat. 2009, 191, 231—247.

Mugnaini, E., Osen, K.K., Schnapp, B., Friedrich, V.L., Distribution of Schwann cell cytoplasm and plasmalemmal vesicles (caveolae) in peripheral myelin sheaths. An electron microscopic study with thin sections and freeze-fracturing, J.

Neurocytol. 1977, 6, 647–668.

The MS Disease-Modifying Medications nationalMSsociety.org/DMD., last accessed February, 24, 2013.

Nobem, J.-P., Dumont, D., Kwasnikowska, N., Verhaert, P., Somers, V.

Hupperts, R. Stinissen, P., Robben, J., Lumbar cerebrospinal fluid proteome in multiple sclerosis: characterization by ultrafiltration, liquid chromatography, and mass spectrometry, J. Proteome Res. 2007, 5, 1647–1657.

Norwood, W. I., Ingwall, J. S., Norwood, C. R., Fossel, E. T., Am. J. Physiol.

1983, 244, C205-C210.

154

Nyblom HK, Thorn K, Ahmed M, Bergsten, P; Mitochondrial protein patterns correlating with impaired insulin secretion from INS-1E cells exposed to elevated glucose concentrations.; Proteomics, 2006, 6, 5193-5198.

Ogorzalek-Loo, RR, Loo, JA; Matrix-Assisted Laser Desorption/Ionization- Mass

Spectrometry of Hydrophobic Proteins in Mixtures Using Formic Acid,

Perfluorooctanoic Acid, and Sorbitol. Anal.Chem., 2007, 79, 1115-1125.

Paling, D., Golay, X., Wheeler-Kingshott, C., Kapoor, R., Miller, D., Energy failure in multiple sclerosis and its investigation using MR techniques, J. Neurol.,

2011, 258, 2113–2127.

Pandit, A., Vadnal, J., Houston, S., Freeman, E., McDonough, J., Impaired regulation of electron transport chain subunit genes by nuclear respiratory factor 2 in multiple sclerosis, J. Neurol. Sci. 2009, 279, 14–20.

Papasotiriou, D.G., Jaskolla, T.W., Markoutsa, S., Baeumlisberger, D., Karas, M.,

Meyer, B.; Peptide Mass Fingerprinting after Less Specific In-Gel Proteolysis

Using MALDI-LTQ-Orbitrap and 4-Chloro-r-cyanocinnamic Acid. Journal of

Proteome Research, 2010, 9, 2619–2629.

Poon, TCW. Opportunities and limitations of SELDI-TOF-MS in biomedical research: practical advices. Expert Rev. Prot., 2007, 4, 51-65.

155

Ravera, S., Panfoli, I., Calzia, D., Aluigi, M.G., Bianchini, P., Diaspro, A.,

Mancardi, G., Morelli, A., Evidence of aerobic ATP synthesis in isolated myelin vesicles, Ont. J. Biochem. Cell B, 2009, 41, 1581–1591.

Richter, F., Meurers, B.H., Zhu, C., Medvedeva, V.P., Chesselet, M-F., Neurons express hemoglobin α- and β-chains in rat and human brains, J. Comp. Neurol.,

2009, 515, 538–547.

Riordan, J.F., Vallee, B.L., Nitration with tetranitromethane. Methods in

Enzymology, 1972, 25, 515-521.

Robijn,M. L. M., Koeleman,C.A.M., Hokke, C.H., Deelder, A.M., Schistosoma mansoni eggs excrete specific free oligosaccharides that are detectable in the urine of the human host. Mol. Biochem. Parasit. 2007, 151, 162–172.

Rudick, R.A., Lee, J., Nakamura, K., Fisher, E., Gray matter atrophy correlates with MS disability progression measured with MSFC but not EDSS, J. Neurol.

Sci., 2009, 282, 106–111.

Ryu, H., Rosas, H.D., Hersch, S.M.,. Ferrante, R.J., The therapeutic role of creatine in Huntington’s disease, Pharmacol. Ther., 2005, 108 193–207.

156

Särkijärvi, S., Kuusisto, H., Paalavuo, R., Levula, M., Airla, N., Lehtimäki, T.,

Kaprio, J., Koskenvuo, M., Elovaara, I., Gene expression profiles in Finnish twins with multiple sclerosis, BMC Med. Genet., 2006, 7, 11–20.

Shevchenko, A., Tomas, H., Havlis, J., Olsen, J.V., Matthias Mann. In-gel digestion for mass spectrometric characterization of proteins and proteomes. Nat.

Protocols 2006, 1, 2856-2860.

Shi, L., Zhang, J., Wu, P., Feng, K., Li, J., Xie, Z., Xue, P., Cai, T., Cui, Z., Chen,

X., Hou, J., Zhang, J., Yang, F. Discovery and identification of potential biomarkers of pediatric Acute Lymphoblastic Leukemia. Proteome Science 2009,

7:7.

Selivanov, V.A., Alekseev, A.E., Hodgson, D.M., Dzeja, P.P., Terzic, A.

Nucleotide-gated KATP channels integrated with creatine and adenylate kinases: amplification, tuning and sensing of energetic signals in the compartmentalized cellular environment, Mol. Cell. Biochem. 2004, 256–257, 243–256.

Sestili, P., Martinelli, C., Bravi, G. Piccoli, G., Curci, R., Battistelli, M., Falcieri,

E., Agostini, D., Gioacchini, A.M., Stocchi, A., Creatine supplementation affords cytoprotection in oxidatively injured cultured mammalian cells via direct antioxidant activity, Free Radic. Biol. Med., 2006, 40, 837–849.

157

Simpson, K.L., Whettonb, A.D., Dive, C. Quantitative mass spectrometry-based techniques for clinical use: Biomarker identification and quantification. J. Chr. B,

2009, 877, 1240–1249.

Smith, C.A., Chetlin, R.D., Gutmann, L., Yeater, R.A., Alway, S.E., Effects of exercise and creatine on myosin heavy chain isoform composition in patients with

Charcot-Marie-Tooth disease, Muscle Nerve, 2006, 34, 586–594.

Smith KJ. Newly lesioned tissue in multiple sclerosis—a role for oxidative damage? Brain, 2011, 134 (Pt 7): 1877–1881. Epub 15 June 2011.

Sparbier, K., Wenzel, T., Dihazi, H., Blaschke, S., Müller, G-A, Deelder, A.,

Flad, T.,Kostrzewa, M. Immuno-MALDI-TOF MS: New perspectives for clinical applications of mass spectrometry. Proteomics 2009, 9, 1442–1450.

Srinivasan R, Ratiney H, Hammond-Rosenbluth KE, MR spectroscopic imaging of glutathione in the white and gray matter at 7 T with an application to multiple sclerosis. Magn. Reson. Imaging, 2010, 28, 163–170.

Srinivasan, S., Avadhani, N.G., Cytochrome c oxidase dysfunction in oxidative stress. Free Radical Biology and Medicine. 2012, 53, 1252–1263.

158

Stadelmann, C., Multiple sclerosis as a neurodegenerative disease: pathology, mechanisms and therapeutic implications. Cur. Opin. Neurol., 2011, 24, 224-229.

Steen, C., Wilczak, N., Hoogduin, J.M., Koch, M., De Keyser, J., Reduced

Creatine Kinase B Activity in Multiple Sclerosis, Normal Appearing White

Matter, PLoS, 2010, 5, e10811.

Straub, A.C., Lohman, A.W., Billaud, M., Johnstone, S.R., Dwyer, S.T., Lee,

M.Y., Bortz, P.S., Best, A. K., Columbine, L., Gaston, B., Isakson, B.E., endothelial cell expression of haemoglobin α regulates nitric oxide signaling.

Nature, 2012, 491, 473-477.

Tang, K, Allman, SL, Jones, RB, Chen, CH; Quantitative Analysis of

Biopolymers by Matrix-Assisted Laser Desorption. Anal. Chem. 1993, 65, 2164-

2166.

Taylor, C.M., Marta, C.B., Claycomb, R.J., Pfeiffer, S.E., Proteomic mapping provides powerful insights into functional myelin biology, Proc. Natl Acad. Sci.

USA 2004, 101, 4643–4648.

Thompson, R. J., Kynoch, P. A. M., Sarjant, J., Radioimmunoassay of Brain-Type

Creatine Kinase-B Isoenzyme In Human Tissues And In Serum Of Patients With

Neurological Disorders. Brain Res. 1980, 201, 423-426.

159

Trapp, B.D., Peterson, J., Ransohoff, R.M., Rudick, R., Mörk, S., Bö, L.; Axonal transaction in the lesions of multiple sclerosis, New Eng. J. Med. 1998, 38, 278–

285.

van Horssen, J., Witte, M.E., Ciccarelli, O. The role of mitochondria in axonal degeneration and tissue repair in MS. Multiple Sclerosis Journal, 2012, 18, 1058–

1067.

Vanrobaeys, F., Van Coster, R., Dhondt, G., Devreese, B., Van Beeumen, J.,

Profiling of myelin proteins by 2D gel electrophoresis and multidimensional liquid chromatography coupled to MALDI TOF-TOF mass spectrometry, J. Prot.

Res., 2005, 4, 2283–2293.

Varghese, R.S., Ressom, H.W., LC-MS Data Analysis for Differential Protein

Expression Detection. In Bioinformatics for Comparative Proteomics; Wu, C.H.,

Chen, C., Eds.; Methods in Molecular Biology; Humana Press: New York, 2011;

694, 139-150.

Vendelin, M. Eimre, M., Seppet, E., Peet, N., Andrienko, T., Lemba, M.,

Engelbrecht, J., Seppet, E.K., Saks, V.A., Intracellular diffusion of adenosine phosphates is locally restricted in cardiac muscle, Mol. Cell. Biochem., 2004,

256–257, 229–241.

160

Vassilopoulos, D., Jockers-Wretou, E.; Serum Creatine Kinase B levels in

Diseases of the CNS. Eur. Neurol. 1987, 27, 78-81.

Vorderwülbecke1, S., Cleverley, S., Weinberger, S.R., Wiesner, A. Protein quantification by the SELDI-TOF-MS–based ProteinChip® System. Nature

Methods, 2005, 2, 393-395.

Wang, L., Mizaikoff, B., Application of multivariate data-analysis techniques to biomedical diagnostics based on mid-infrared spectroscopy. Anal. Bioanal.

Chem., 2008, 391; 1641-1654.

Wendt, S. Dedeoglu, A., Speer, O., Wallimann, T., Beal, M.F., Andreassen, O.A.,

Reduced creatine kinase activity in transgenic amyotrophic lateral sclerosis mice,

Free Radic. Biol. Med. 32 (2002) 920–926.

Whiteaker, J.R., Zhao, L., Zhang, H.Y.,Feng, L-C, Piening, B.D., Anderson,

L.,Paulovich, A.G. Antibody-based enrichment of peptides on magnetic beads for mass-spectrometry-based quantification of serum biomarkers. Anal. Biochem.

2007, 362, 44–54.

161

Wu, W.W., Wang, G.H., Baek, S.J., Shen, R.F., Comparative study of three proteomic quantitative methods, DIGE, cICAT, and iTRAQ, using 2D gel- or LC-

MALDI TOF/TOF. J. of Prot. Res., 2006, 5, 651-658.

Yen, T-Y, Macher, B.A., McDonald, C.A., Alleyne-Chin, C., Timpe, L.C.

Glycoprotein Profiles of Human Breast Cells Demonstrate a Clear Clustering of

Normal/Benign versus Malignant Cell Lines and Basal versus Luminal Cell

Lines. J. Proteome Res. 2012, 11, 656–667.

Zaaraoui W, Rico A, Audoin B, Unfolding the long-term pathophysiological processes following an acute inflammatory demyelinating lesion of multiple sclerosis. Magn Reson Imaging 2010; 28: 477–486.

Zambonin JL, Zhao C, Ohno N,. Increased mitochondrial content in remyelinated axons: implications for multiple sclerosis. Brain 2011, 134, 1901–1913.

Ziemann, U., Wahl, M., Hattingen, E., Tumani, H., Development of biomarkers for multiple sclerosis as a neurodegenerative disorder, Prog. Neurobio. 2011, 95,

670–685.