Investigation of the Human Brain Metabolome to Identify Potential

Subscriber access provided by FORDHAM UNIVERSITY

Article

An investigation of the human brain metabolome to identify potential markers for early diagnosis and therapeutic targets of Alzheimer’s Disease. Stewart Francis Graham, Olivier P. Chevallier, Dominic Roberts, Christian Holscher, Christopher T. Elliott, and Brian Desmond Green Anal. Chem., Just Accepted Manuscript • DOI: 10.1021/ac303163f • Publication Date (Web): 19 Dec 2012 Downloaded from http://pubs.acs.org on January 2, 2013

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Analytical Chemistry is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 30

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

An investigation of the human brain metabolome to identify potential markers for early diagnosis and therapeutic targets of Alzheimer’s Disease.

Stewart F. Graham1*, Olivier P. Chevallier1, Dominic Roberts2, Christian Hölscher3, Christopher T. Elliott1 and Brian D. Green1. 1

Asset Technology Centre, Institute of Agri-Food and Land Use, Queen’s University Belfast,

Stranmillis Road, Belfast, BT9 5AG, UK. 2

Waters Corporation, Atlas Park, Simonsway, Manchester, M22 5PP, UK.

3

School of Biomedical Sciences, University of Ulster, Coleraine, BT52 1SA, UK.

*Corresponding Author Tel: +44 2890976562; Fax: +44 2890976513; e-mail: [email protected]

1 ACS Paragon Plus Environment


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 2 of 30

Abstract A study combining high resolution mass spectrometry (LC-QTof-MS) and chemometrics for the analysis of post-mortem brain tissue from subjects with Alzheimer’s disease (AD) (n=15) and healthy age-matched controls (n=15) was undertaken.

The huge potential of this

metabolomics approach for distinguishing AD cases is underlined by the correct prediction of disease status in 94%-97% of cases. Predictive power was confirmed in a blind test set of 60 samples, reaching 100% diagnostic accuracy. The approach also indicated compounds significantly altered in concentration following the onset of human AD. Using orthogonal partial least squares discriminant analysis (OPLS-DA) a multivariate model was created for both modes of acquisition explaining the maximum amount of variation between sample groups (Positive Mode-R2=97%; Q2=93 %; RMSEV=13 %; Negative Mode-R2=99 %; Q2=92 %; RMSEV=15%). In brain extracts 1264 and 1457 ions of interest were detected for the different modes of acquisition (positive and negative, respectively). Incorporation of gender into the model increased predictive accuracy and decreased RMSEV values. High resolution LC-QTof-MS has not previously been employed to biochemically profile postmortem brain tissue and the novel methods described and validated herein prove its potential for making new discoveries related to the aetiology, pathophysiology and treatment of degenerative brain disorders.


Page 3 of 30

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Introduction Alzheimer’s disease (AD) is the most common form of dementia, accounting for 70 % of all documented cases.

Worldwide 24 million people suffer from dementia with numbers

forecast to reach 80 million by 2040.1 It is estimated every 69 seconds someone in the US is diagnosed with AD and by 2050 this is forecast to accelerate to one person every 33 seconds. 2

AD is a gradual, debilitating, neurodegenerative disorder that is increasingly imposing

significant economic burdens on society in terms of medical costs, residential and informal care. 3 In the US, AD is estimated to be the third most costly condition after heart disease and cancer with an estimated $202 billion of expenditure in 2011. 2, 3 In the UK the cost of AD is estimated to cost £23 billion a year. AD is a neurodegenerative disorder for which there is no cure and few reliable diagnostic biomarkers.4 The pathology is characterised by the accumulation of tau tangles and of βamyloid plaques,5,6 however the actual biochemical basis for neurodegeneration is poorly understood. Treatment is very limited with only five drugs so far receiving approval by the US Food and Drug Administration. The drugs temporarily slow the progression of symptoms for approximately 6-12 months2 but no treatments are available to slow or stop the deterioration of the brain cells in AD. Current therapies are initiated only after diagnosis; their modest benefit, in part, may be explained by the fact that irreversible brain damage already has occurred by the time dementia is recognized. The development of new therapeutic targets for AD will not only expand our understanding of the aetiology and pathophysiology of the disease but will also assist clinicians to treat the disease in its earliest indicative stages. Current strategies of early diagnosis focus on classic biomarkers in the CSF such as amyloid levels, the amyloid (1-40) / (1-42) ratio, and levels of phosphorylated tau.4 It is also worth point out that CSF taking is not universally given because of its invasive nature. Increasingly, imaging techniques are used to identify plaque load in PET imaging studies.7 3 ACS Paragon Plus Environment


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 4 of 30

Also MRI (particularly volumetric MRI) is undergoing validation for measurement of brain atrophy in a number of studies in the US.8 However, the predictive value of these techniques is still very limited, and may be potentially due to the fact that neither amyloid nor tau may be causal factors of the disease, but only build up once the disease has progressed.

It is

necessary to identify new neurological therapeutic targets and to also to discover potential biomarkers in different sample types (i.e. plasma and CSF) and aid early disease diagnosis/prognosis. Metabolomics is a discipline dedicated to the global study of small molecule metabolites in cells, tissues and biofluids.9 It involves the comprehensive, simultaneous and systematic profiling of multiple metabolite concentrations and their fluctuations in response to disease, drugs, diet and lifestyle.9-11 The technique has been applied to the study of a number of metabolic 12-14 and neurodegenerative diseases,15-17 but few have investigated its use in AD. A study undertaken by Salek et al., (2010) measured 24 metabolite concentrations in mice CRND8 brains using 1H NMR. Transgenic CRND8 mice encode the mutant form of the APP 695 with both the Swedish and Indiana mutations and develop extracellular amyloid betapeptide deposits in the brain.18

In this study they produced a partial least squares

discriminant model with a predictive power of 54 and 60 % (cortex and hippocampus, respectively) when comparing CRND8 mice brains with aged matched controls.18

A

metabolomics study of a Finnish patient cohort identified changes in metabolite profiles which are potentially relevant to the pathogenesis and progression of AD19 with the results provide important clues regarding the features of AD such as the occurrence of hypoxia, oxidative stress, and membrane lipid remodelling. In this study they identified 139 molecular lipids and 544 small polar metabolites. In addition a study undertaken by Han et al., (2011) identified changes in the sphingomyelin and ceramide levels in AD plasma and provide new insight into the AD sphingolipidome and the potential use of metabolite signatures as 4 ACS Paragon Plus Environment

Page 5 of 30

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


potential biomarkers. In this study they identified 800 molecular species of lipids using the non-targeted multi-dimensional mass spectrometry based analysis highlighting 8 species of sphingomyelin that were lower in AD and 2 ceramide species which were significantly higher.16 A recent metabolomics study used liquid chromatography coupled with coulometric array detection to analyse the metabolome of post-mortem ventricular cerebrospinal fluid of both AD and non-demented subjects.20

A promising recent study used a non-targeted

metabolomics approach based on capillary electrophoresis-mass spectrometry (CE-MS) and reported small metabolite changes in CSF from AD and cognitively impaired patients21. In the present study a large proportion of the entire polar metabolome was analysed using two modes of acquisition. A comprehensive review of the scientific literature reveals that these are the only clinical investigations of AD involving metabolomics and there is great potential to improve our understanding of the underlying disease, and offer a means for assessing new disease therapies. The powerful technique of high resolution LC-QTof-MS has not previously been applied for biochemically profiling post-mortem human brain tissue. The purpose of this investigation was to assess the relative analytical power and potential usefulness of high-throughput high resolution LC-QTof-MS for studying the global polar metabolite changes occurring in the brain of human subjects with AD.



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 6 of 30

Experimental Section Tissue Samples Metabolomic studies were conducted on human post-mortem (PM) brain tissue obtained from the Newcastle Brain Trust (Brodmann 7 region, neocortex region, n=15 Controls; n=15 AD). Consent for the use of tissue was obtained by the Newcastle Brain Bank which is licensed by the Human Tissue Authority. Details such as Braak stage, age, gender and post mortem delay can be found in Table S-1. Frozen tissue samples were lyophilised (~5 g), milled to a fine powder and 50 mg (± 0.5 mg) added to 1 mL of 50% methanol:water in a 2 ml sterile Eppendorf tube. The samples were mixed for 10 min, sonicated for 15 min and the protein removed by centrifugation at 16,000 g at 4ºC for 20 min.

LC-QTof-MS Analysis An exhaustive process of optimisation was undertaken which assessed various column chemistries, extraction procedures and solvent gradients. The following protocol was found to be optimal for the analysis of polar brain metabolites.

Sample supernatants were

evaporated to dryness, reconstituted in 300 µl of Ultra-Pure water (Sigma Aldrich, UK) and filtered by centrifugation using 0.22 µm Constar Spin-X® Centrifuge Tube Filter (10,000 g at 4ºC for 5 minutes; Corning Incorporated, Corning, NY 14831, USA). All solvents used (water, acetonitrile, formic acid, ammonia solution 35%) were purchased from Fisher (Loughborough, UK) and were LC-MS grade or equivalent. Chromatography was performed on a Waters Acquity UPLC system (Milford, MA, USA), equipped with column oven, coupled to a Waters Xevo G2 QTof mass spectrometer (Manchester, UK) equipped with an electrospray ionisation source operating in positive or negative mode with lock-spray interface for real time accurate mass correction. The source temperature was 120°C with a cone gas flow of 5 L/h in positive mode and 20 L/h in negative mode, a dessolvation


Page 7 of 30

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


temperature of 450°C in positive mode and 400°C in negative mode, and a dessolvation gas flow of 800 L/h (both mode). The capillary voltage was set at 0.8 kV with a cone voltage of 20 V for positive ion mode, and 1.5 kV and 25 V respectively for negative ion mode. A lock-mass solution of Leucine Enkephalin (2 ng.µL-1) in acetonitrile/water containing 0.1% formic acid (50:50, v/v) was continuously infused into the MS via the lock-spray at a flow rate of 8 µl.min-1. Mass spectra data were acquired in centroid mode using MSE function (low energy: 4eV; high energy: ramp from 20 to 35 eV) over the range m/z 50-1200 range with a scan time of 0.1s. A 1.5 µL aliquot of extracted tissue sample was injected onto an Acquity UPLC BEH HILIC column (2.1 x 100 mm, 1.7 µm, Waters, Milford, MA, USA). The main principle of HILIC (Hydrophilic interaction chromatography) separation is based on a compounds polarity and degree of solvation. The more polar compounds are separated by their stronger interaction with the stationary aqueous layer than the less polar compounds, therefore resulting in a stronger retention on the analytical column. The column oven was set at 45°C, and the sample manager temperature was 6°C. The gradient elution buffers were A (20 mM ammonium formate with 0.2% formic acid) and B (acetonitrile containing 0.025% formic acid), and the flow rate was 0.6 mL.min-1. The elution gradient (A:B, v/v) was as follows for positive ion mode: an isocratic period of 1 min at 10:90 followed by a linear gradient from 10:90 to 40:60 over 7 min then an isocratic period at 40:60 for 0.5 min followed by a linear change from 40:60 to 90:10 over 1.5 min and then returned to initial condition (10:90) over 0.1 min. The initial composition was then kept for a further 2.90 min before the next injection. For negative ion mode, chromatographic elution started by an isocratic period at 5:95 for 1 min, followed by a linear gradient from 5:95 to 40:60 over 7 min. These conditions (40:60) were kept for 0.5 min and were followed by a linear change to 10:90 over 0.5 min.



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 8 of 30

After a 1 min period at 10:90, a linear gradient was applied over 0.1 min to return to the initial composition 5:95 which was held for 2.9 min before the next injection. Prior to the analysis of both ESI+ and ESI- 10 pooled conditioning samples were injected. To determine the chromatographic reproducibility of retention times and peak intensities, pooled samples were injected after every 10 samples throughout the experiment22. Data Analysis The raw data from the spectral analysis of tissue extracts was processed using MarkerLynx XS (Waters Corporation, Milford, MA) software and the parameters for the analysis are available in Table S-2. The analysed spectral data was exported to Simca 13 (Umetrics, Umea, Sweden) for multivariate analysis. Prior to any in-depth data analysis, data quality was assessed in terms of reproducibility by an approach adopted by other leading metabolomics researchers22. Clustering of the pooled samples was assessed using principal component analysis (PCA) to reveal if platform stability had been achieved. For both ESI+ and ESI- acquisition modes the pooled samples were tightly clustered which indicated good reproducibility of the data22. Following this data were mean centered and Pareto scaled and grouped into AD and Controls prior to analysis using orthogonal projection to latent structures-discriminant analysis (OPLS-DA). Pareto scaling was used since it augments the representation of the low concentration metabolites by dividing each variable by the square root of the standard deviation of the variable, without increasing the noise contribution to the model.17

R2 (cumulative), Q2 (cumulative) and Root Mean Squared Error of validation

(RMSEV) were used to determine the validity of the model. R2 (cum) indicates the variation described by all components in the model and Q2 is a measure of how accurately the model can predict class membership. It does this by leaving out 1/7th of the data from the model and then predicting their class membership.19 Additional predictive models (cross validation) were built using 2/3 of the original data (training set; n=120) and used to blindly predict the


Page 9 of 30

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


remaining 1/3 (test set; n=60). In each of the models all of the test set observations were correctly assigned to control or AD groups 100% of the time for both ESI+ and ESI(Available in Figures S1 and S2). The ions of interest (highlighted by the S-plot) to be at different levels between the two sample groups were analysed using a two-tailed homoscedastic student t-test (assuming that the two sample groups were of equal variance) using Microsoft Excel (2010). Graphical representations of the data were produced using Prism (Version 5.0). Results Figure 1A is a typical chromatogram acquired in positive mode of an extract of post-mortem brain tissue. Subsequent data processing with MarkerLynx XS (used to process complex multivariate data from LC-MS) detected 1264 ions. Orthogonal partial least squares discriminant analysis (OPLS-DA) was used to build a model differentiating control (red) and AD (blue) post-mortem tissue (n=6 for each sample). The scores plot outlined in Figure 1B displays the results of the model building. For this model one component and twelve orthogonal projections were calculated to explain the maximum amount of variation between the two sample groups (R2 (cum) = 99 %; Q2 (cum) = 95 %; Root mean squared error of validation (RMSEV) = 11 %). Figure 1C is the loadings plot corresponding to the scores plot in Figure 1B; it identifies the compounds that are responsible for the variation between the two sample groups.

Highlighted in red are ions found to be at significantly higher

concentrations in control than in AD (n=15). Highlighted in blue are ions found to be at significantly higher concentrations in AD than in control (n=24). The S-plot in Figure 1D indicates the relative importance of each molecule in differentiating sample groups. Peak height and colour indicate relative importance of the candidate molecules. Table 1 lists the retention times, masses of the ions, their percentage increase and their significance values.



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 10 of 30

Figure 2 displays the relative abundance of all the ions of interest between the two sample groups including the standard error of the mean indicated by the error bars. Figure 3A is a typical chromatogram acquired in negative mode of an extract of post-mortem brain tissue. Data were analysed as previously described for positive mode acquisition and 1457 ions were detected by MarkerLynx. The OPLS-DA scores model (Figure 3B) required one component and six orthogonal projections to explain the maximum amount of variation between the two sample groups (R2 (cum) = 99%; Q2 (cum) = 92 %; RMSEV = 15 %). The loadings plot depicted in Figure 3C indicates ions of interest higher in a particular group (Red = Controls (n=18); Blue = AD (n=19)) and ones which are similar in both sample sets. Significance and relative variable importance is indicated by the S-plot in Figure 3D. Table 2 lists the retention times, masses of the candidate molecules, percentage increases between sample groups and their significance values acquired in both positive and negative mode. The numbered ions of interest and their relative abundances between sample groups are displayed in Figure 4. Data from positive mode acquisition were segregated by gender (Figure 5A: male; Figure 5B: female). The OPLS-DA scores model for the male sample data (Figure 5A) required one component and three projections to explain the maximum variation between the groups (R2 = 99 %; Q2 = 96%; RMSEV = 9 %).

The model for the female data (Figure 5B) was based on

one component and five projections (R2 = 99 %; Q2 = 97 %; RMSEV = 8 %). Figures 5C (male) and 5D (female) indicate the relative importance of individual ions explaining the variation between control and AD (also see Table S-3). Red indicates molecules at significantly higher concentrations in controls (male, n=15; female, n=11) and blue indicates compounds at significantly higher concentrations in AD (male, n=16; female, n=24).


Page 11 of 30

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Data from negative mode acquisition were segregated by gender (Figure 6A: male; Figure 6B: female). The OPLS-DA scores model for the male sample data (Figure 6A) required one component and four projections to explain the maximum variation between the groups (R2 = 99 %; Q2 = 91%; RMSEV = 13 %).

The model for the female data (Figure 6B) was based

on one component and four projections (R2 = 99 %; Q2 = 93 %; RMSEV = 13 %). Figures 6C (male) and 6D (female) indicate the relative importance of individual ions explaining the variation between control and AD (also see Table S-4). Red indicates ions at significantly higher concentrations in controls (male, n=17; female, n=19) and blue indicates ions at significantly higher concentrations in AD (male, n=19; female, n=12). Discussion To our knowledge this is the first retrospective study using high resolution LC-QTof-MS to biochemically profile the polar metabolome of post-mortem AD brain tissue and elderly controls. By acquiring MS data in positive and negative modes two separate metabolomic models were developed which clearly and unambiguously distinguish AD subjects from controls. The predictive power of these metabolomics models was profound with predictive accuracy ranging from 91 to 97% (RMSEV 8-15 %). The closest published metabolomics studies in brain tissue predicted Batten’s disease with 41 % accuracy15 and predicted AD with 54-60 % accuracy.18 Strikingly, data acquired in positive mode assigned a total of 1264 polar compounds, whilst data acquired in negative mode assigned a total of 1457 polar compounds. This number of candidate molecules is 2-3 times greater than in the closest related study, where 544 polar metabolites were examined in the plasma of AD subjects.19 Furthermore, previous metabolomics investigations in brain tissue have typically employed low sensitivity and low resolution techniques such as NMR and coulometry. A major disadvantage of these methods is that they only measure 25-30 metabolites.15,18,20 Alternatively CE-MS has been used to profile ~160 metabolites in the CSF of AD patients, cognitively impaired subjects and 11 ACS Paragon Plus Environment


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 12 of 30

age-matched controls21. This study indicates the potential of an MS-based approach for discovering metabolite biomarkers for AD. A total of 14 identified metabolites appeared to differ in the three groups of subjects but only 5 were significantly altered in concentration (p

Investigation of the Human Brain Metabolome to Identify Potential

Recommend Documents