Context
Blood-based analytes may be indicators of pathological processes in Alzheimer disease (AD).
Objective
To identify plasma proteins associated with AD pathology using a combined proteomic and neuroimaging approach.
Design
Discovery-phase proteomics to identify plasma proteins associated with correlates of AD pathology. Confirmation and validation using immunodetection in a replication set and an animal model.
Setting
A multicenter European study (AddNeuroMed) and the Baltimore Longitudinal Study of Aging.
Participants
Patients with AD, subjects with mild cognitive impairment, and healthy controls with standardized clinical assessments and structural neuroimaging.
Main Outcome Measures
Association of plasma proteins with brain atrophy, disease severity, and rate of clinical progression. Extension studies in humans and transgenic mice tested the association between plasma proteins and brain amyloid.
Results
Clusterin/apolipoprotein J was associated with atrophy of the entorhinal cortex, baseline disease severity, and rapid clinical progression in AD. Increased plasma concentration of clusterin was predictive of greater fibrillar amyloid-β burden in the medial temporal lobe. Subjects with AD had increased clusterin messenger RNA in blood, but there was no effect of single-nucleotide polymorphisms in the gene encoding clusterin with gene or protein expression. APP/PS1 transgenic mice showed increased plasma clusterin, age-dependent increase in brain clusterin, as well as amyloid and clusterin colocalization in plaques.
Conclusions
These results demonstrate an important role of clusterin in the pathogenesis of AD and suggest that alterations in amyloid chaperone proteins may be a biologically relevant peripheral signature of AD.
Peripheral compartments including blood and cerebrospinal fluid exhibit signals reflecting neuropathological changes in Alzheimer disease (AD).1,2 In cerebrospinal fluid, these include a decrease in amyloid-β peptide (Aβ) and an increase in total and phosphorylated tau concentrations,3 reflecting amyloid sequestration as plaques and neurofibrillary degeneration, respectively.4,5 Similarly, while numerous articles suggest that plasma concentrations of several metabolites and proteins might represent responses to neuropathological changes in AD,6-11 these findings have not been conclusively replicated.12 A limitation of such studies may be their reliance upon demonstrating changes between affected and unaffected people, a design of study that might identify secondary changes lacking relevance to core disease biology.
Advances in methods such as proteomics present a further challenge in case-control studies, often generating data showing numerous analytes differentially expressed in AD patients. However, validating these results with alternative methods in independent patient populations has been difficult.13,14 These studies also ignore the clinical heterogeneity in disease progression in AD, wherein some patients show rapid cognitive decline, while others remain relatively stable and/or progress slowly.15,16
We applied mass spectrometry–based proteomics to discover plasma proteins associated with disease, using brain atrophy in AD as well as rapid clinical progression, rather than binary distinction between case and control. As a proxy measure of in vivo pathology, we used structural neuroimaging of atrophy in the hippocampus and entorhinal cortex (ERC), 2 components of the medial temporal lobe (MTL) that show early pathological changes in AD.17 For rate of clinical progression, we used both retrospective and prospective measures of cognitive decline. We initially performed 2 independent discovery-phase studies using proteomic analysis of plasma in separate groups of subjects. In the first, we sought proteins that reflect hippocampal atrophy in mild cognitive impairment (MCI) and established AD. In the second, we identified proteins differentially expressed in rapidly progressing AD patients relative to those with a less aggressive disease course. Our aim was to identify plasma proteins common to both paradigms, followed by replication using quantitative immunoassays such as enzyme-linked immunosorbent assay (ELISA) in a large independent cohort of AD, MCI, and control subjects. Figure 1 illustrates the design of this study.
We used samples from 2 studies: the Alzheimer Research Trust–funded cohort at King's College London (KCL-ART)7 and the AddNeuroMed study.18 The KCL-ART study, which began in 2001, includes a cohort of people with AD and MCI19 and healthy elderly individuals. All subjects are white UK citizens with grandparents born in the United Kingdom and are assessed annually. AddNeuroMed is a cross-European cohort; AD cases are assessed at 3-month intervals in the first year and annually thereafter; MCI and control groups are assessed annually. All subjects are white Europeans recruited from 6 centers in the United Kingdom, France, Italy, Finland, Poland, and Greece. Standardized assessments include demographic and medical information; cognitive assessment, including the Mini-Mental State Examination (MMSE) (both studies; all subjects), Alzheimer Disease Assessment Scale–cognitive subscale (ADAS-cog) (AddNeuroMed only), and Consortium to Establish a Registry for Alzheimer's Disease battery; and scales to assess function, behavior, and global levels of severity, including the Clinical Dementia Rating. Cases with probable AD (National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer's Disease and Related Disorders Association [NINCDS-ADRDA] criteria) and amnestic MCI were identified as previously described7 and evaluated with a standardized assessment shown to have high diagnostic validity.20 Cases with amnestic MCI were defined as having subjective memory complaints, Clinical Dementia Rating scores of less than 1, and evidence of objective memory impairment using the Consortium to Establish a Registry for Alzheimer's Disease delayed word list recall (−1.5-SD cutoff). Normal elderly controls, defined as having no evidence of cognitive impairment (MMSE score >28), were recruited systematically from primary care patient lists in the KCL-ART study and from both primary care services and elsewhere in the AddNeuroMed study. Blood samples were collected and stored as previously described.7,18 In total, we studied 95 and 689 subjects in discovery and validation studies, respectively, with an additional 60 subjects from the Baltimore Longitudinal Study of Aging (eTables 1-4).21 Ethical approval was obtained in each of the participating countries.
Magnetic Resonance Imaging Data Acquisition
In the KCL-ART study, whole-brain coronal 3-dimensional spoiled-gradient recalled images (repetition time = 14 milliseconds, echo time = 3 milliseconds, 256 × 192 × 124 acquisition matrix, 1.5-mm slices) were obtained on a GE Signa 1.5-T neuro-optimized magnetic resonance system. In the AddNeuroMed study, whole-brain sagittal 3-dimensional magnetization-prepared rapid acquisition gradient echo images (repetition time = 8.6 milliseconds, echo time = 3.8 milliseconds, 256 × 192 acquisition matrix, 180 × 1.2–mm slices) were obtained on a 1.5-T magnetic resonance system at each of the 6 centers. Quality control was undertaken using the ADNI Magphan phantom and 2 volunteers who visited each of the centers, ensuring compatibility across the study. Thickness of the ERC was calculated with Freesurfer using a cortical reconstruction technique.22,23
11C–Pittsburgh Compound B Positron Emission Tomographic Studies
Dynamic 11C–Pittsburgh Compound B (11C-PiB) positron emission tomographic (PET) studies (37 time frames across 90 minutes) were acquired in 3-dimensional mode on a GE Advance scanner immediately after intravenous bolus injection of approximately 5.55 × 108 Bq (15 mCi) of 11C-PiB. Dynamic images were reconstructed using filtered back projection with a ramp filter (image size, 128 × 128; pixel size, 2 × 2 mm; slice thickness, 4.25 mm), yielding a spatial resolution of about 4.5 mm full width at half maximum at the center of the field of view. Parametric images of distribution volume ratios were calculated by simultaneously fitting a reference tissue model using linear regression and spatial constraint with the cerebellum as a reference region.24,25 The SPM5 program (Statistical Parametric Mapping 5; Wellcome Department of Imaging Neuroscience, London, England) was used to investigate the association between clusterin and medial temporal 11C-PiB retention (significance threshold of P ≤ .05, with a spatial extent of 25 voxels). Based on a priori hypotheses in light of our results on the association between ERC atrophy and clusterin concentration in AD, a restricted search of the MTL was performed using the regional definition from the WFU PickAtlas.26
Two-dimensional gel electrophoresis and liquid chromatography coupled to tandem mass spectrometry (LC-MS-MS) were performed as previously described.7 Gels were analyzed using image analysis software (either Melanie 2-D or Progenesis SameSpots version, 3.0, Nonlinear Dynamics). Protein spots of interest were excised, washed, digested in gel with trypsin, and analyzed by LC-MS-MS.7 Mass spectral data were processed into peptide peak lists and searched against the Swiss-Prot Database using Mascot software (Matrix Science, London, England).
For validation experiments, plasma clusterin concentration was assayed by a commercially available ELISA kit (Human Clusterin ELISA, RD194034200R; Biovendor Laboratory Medicine Inc, Modric, Czech Republic). Samples were run in duplicate. Coefficient of variation of the ELISA for all studies overall was 3.5% (baseline data, 3.7%; follow-up data, 3.5%; and BLSA substudy, 3.1%).
Gene Expression of Clusterin
Approximately 2.5 mL of venous blood was collected into a PAXgene tube for each subject at the baseline visit, processed according to the manufacturer's instructions, and stored at −20°C overnight prior to −80°C storage. RNA was extracted using the PAXgene Blood RNA kit according to the manufacturer's instructions. Samples were assessed for yield using a spectrophotometer and quality using the RNA 6000 Pico Chip on the Agilent Bioanalyser. Samples with an RNA integrity number greater than 7.0 were used for polymerase chain reaction (PCR) assays.
Using the Quantitect Reverse Transcription kit (Qiagen), 500 ng of RNA was reverse transcribed to complementary DNA in a 40-μL reaction and subsequently diluted to 200 μL. Reverse transcriptase–PCR reactions were performed in 384-well plates in the 7900HT Fast Real-Time PCR machine (Applied Biosystems, Foster City, California). The geNORM housekeeping selection kit (Primer Design Ltd, Southampton, England) was used to assay 12 housekeeping genes in a subset of the samples. Using NormFinder software, the 2 most stable genes for normalization were determined to be SF3A1 and ATP5B. Samples were assayed in duplicate, and a standard curve of known copy number was run on each plate for clusterin, SF3A1, and ATP5B. Data were nonparametric and were therefore log transformed.
Tagger software (http://www.broad.mit.edu/mpg/tagger/) identified 7 single-nucleotide polymorphisms (SNPs) (rs9331908, rs11136000, rs867231, rs867230, rs9331888, rs9314349, and rs484377) that captured more than 90% of variation in the clusterin gene. Genotypes were determined using a TaqMan allele-specific assay (Applied Biosystems). The PCR amplifications were performed on an ABI PRISMR 7000 Sequence Detection System (Applied Biosystems). A total of 946 individuals (358 AD subjects, 373 controls, and 215 MCI subjects) were genotyped for the 7 SNPs.
TASTPM Transgenic Mouse Model Experiments
Heterozygote transgenic mice overexpressing hAPP695swe (Tas10) and presenilin 1 M146V mutations (Tpm) were generated as previously described.27 Western blot analysis of clusterin was performed in plasma samples at 6 months using an anti–apolipoprotein J mouse polyclonal antibody (Abcam AB349-50; 1:5000). For immunohistochemistry, antigen retrieval was undertaken as described previously.27,28 Primary antibodies were 1E8 (pan-Aβ), 20G10 (Aβ42; GlaxoSmithKline; 1:1000), and anticlusterin (R&D Systems goat polyclonal AF2747; 1:20 000). Images were captured at ×4 magnification on an Axioscope microscope and analyzed by Image J software to generate percentage Aβ or clusterin load. Animal experiments were conducted according to the Council of Europe guidelines.
Discovery-phase proteomic data were analyzed by partial least-squares regression using SIMCA-P, version 8.0. Spot data were scaled to unit variance and log10 transformed where appropriate. Observations with greater than 50% missing values were excluded. Partial least-squares discriminant analysis was used to derive a panel of protein spots that discriminated between rapidly and slowly declining AD groups.
Validation-phase protein data were examined using SPSS, version 17. Covariates were chosen in cases in which such variables were significantly different between the groups of interest or in which they were likely to influence the dependent variable. To test associations between plasma clusterin concentration and ERC thickness, partial correlation analysis was performed with age and sex as covariates. In analyzing associations between MMSE score and plasma clusterin concentration, partial correlation was performed with age as a covariate. While testing differences in clusterin concentration between rapidly and nonrapidly declining AD patients, age and sex were not included as covariates because they were not significantly different between the 2 groups. However, duration of disease was significantly different between these groups (retrospective analysis) and was therefore included as a covariate in an analysis of covariance (ANCOVA) model. In the prospective analysis, there was no significant difference in disease duration between rapid and nonrapid decliners, and clusterin concentration between these groups was therefore compared using an independent samples t test. Linear regression adjusting for disease status, age, sex, and APOE ε4 status was performed to investigate the association between CLU SNPs and clusterin plasma levels and to examine the relationship between CLU messenger RNA (mRNA) and disease. Image analysis is described in the relevant sections. All other statistical analyses were performed using SPSS, version 17, and are described in the text.
Proteomic identification of plasma proteins associated with hippocampal atrophy and rapid clinical progression in ad
To identify plasma proteins associated with disease as reflected by cerebral atrophy, we first performed a discovery-phase proteomics experiment using 2-dimensional gel electrophoresis and LC-MS-MS, with hippocampal atrophy as the independent variable. We analyzed samples from 44 subjects from the KCL-ART cohort, representing a continuum of disease (27 individuals with mild to moderate AD and 17 with MCI; eTable 1). Bivariate correlation of integrated optical densities of spots detected by 2-dimensional gel electrophoresis revealed 13 spots that were significantly associated with hippocampal volume (r ≥ ±0.35, P < .05). Subsequently, using partial least-squares regression,29 a method suited to analysis of proteomic data in which colinearity among predictor variables is common, a model with 2 components was fitted to the hippocampal volume data. This was constituted by 8 of the 13 spots which, together, explained 34% of the variance (R2Y, ie, explained variance in the outcome variable) in hippocampal volume. Using LC-MS-MS, we identified these 8 spots as complement C3, γ-fibrinogen, serum albumin, complement factor I, clusterin (in 2 spots), α1-macroglobulin, and serum amyloid P (Figure 2). We then performed a second discovery-phase experiment in an independent set of samples in 51 carefully matched (for age, sex, severity at the time of blood sampling, and cholinesterase inhibitor treatment [all were taking the drug]) AD subjects from the AddNeuroMed cohort who we could divide into fast (n = 22) or slow (n = 29) progressors based on their annualized rate of cognitive decline (eTable 1). We defined a priori fast decline as a fall of 2 or more points on the ADAS-cog scale during 6 months. A partial least-squares discriminant analysis model distinguishing the rapidly from the slowly progressing AD groups was constituted by the integrated optical densities of 27 silver-stained 2-dimensional gel electrophoresis spots. Of these, 8 were well defined, discrete, and present in all 51 gels and were identified by LC-MS-MS. These spots contained complement component C4 (in 3 spots), complement C8, clusterin, apolipoprotein A1 (in 2 spots), and transthyretin (Figure 2).
Clusterin and atrophy of the erc, severity of cognitive impairment, and speed of progression in ad
Only 1 protein was common to both discovery-phase studies: clusterin. We therefore sought to confirm this finding in a large cohort of 689 subjects, including 344 from the AddNeuroMed study (119 with AD, 115 with MCI, and 110 controls) and 345 (all with AD) from the KCL-ART cohort (eTable 2). We used atrophy in the ERC as an alternative measure of disease pathology (Figure 1). The 689 validation-phase subjects included the 95 subjects in the discovery phase albeit with entirely different analytical measures in the 2 studies.
Confirming the discovery-phase study, we observed a trend toward association between clusterin concentration and ERC atrophy in the combined AD and MCI cohort (n = 219; R = −0.12, P = .06) after covarying for age and sex. This relationship was driven primarily by a highly significant association between ERC atrophy and clusterin concentration in AD patients (n = 113; R = −0.30, P = .001). We also correlated plasma clusterin concentration with MMSE score—a measure of cognition available in 576 subjects with MCI and AD—and again found a highly significant negative correlation (r = −0.22; P < .001, age as a covariate).
We then compared clusterin levels in rapidly declining AD patients relative to slow decliners using both retrospective and prospective measures of decline relative to the time of blood sampling (Figure 1 and eTable 2). Retrospective decline was estimated from the duration of disease and the MMSE score at the point of blood sampling, allowing the annualized fall in MMSE score to be calculated. We used MMSE score, as the ADAS-cog score was not available in all subjects, and defined fast decline as a fall of 2 points or more during a 1-year period relative to the time of blood sampling. Prospective decline was directly measured as the fall in MMSE score 1 year after blood sampling. We observed a significant increase in clusterin concentration in AD patients with accelerated cognitive decline prior to blood sampling (n = 344; ANCOVA, t341 = 3.40, P< .001, duration of disease as covariate) (Figure 3A) and an increase in clusterin concentration in AD patients with faster cognitive decline subsequent to blood sampling (n = 237; independent samples t test, P = .01) (Figure 3B). Cox proportional regression analysis showed that higher plasma clusterin concentration was associated with a greater risk of rapid cognitive decline 1 year after blood sampling (Figure 3C). We then performed an analysis of variance (age and sex as covariates) between AD, MCI, and control groups in the entire sample to test for differences in plasma clusterin concentration. There were no significant differences: AD, 82.4 ng/mL (SD, 25.6 ng/mL; n = 336); MCI, 77.6 ng/mL (SD, 22.5 ng/mL; n = 222); and control subjects, 82.2 ng/mL (SD, 23.8 ng/mL; n = 385). Finally, we compared differences in plasma clusterin concentration between APOE ε4 carriers and noncarriers (independent samples t test) in the combined cohort of AD, MCI, and control subjects and did not find any significant difference.
CLUSTERIN AND FIBRILLAR Aβ BURDEN IN THE ERC IN NONDEMENTED OLDER INDIVIDUALS
Because high clusterin levels are associated with brain atrophy and a more rapid rate of cognitive decline in AD patients, we hypothesized that increased clusterin concentration might be an antecedent marker of pathology in otherwise normal older individuals. We tested this hypothesis in participants of the Baltimore Longitudinal Study of Aging who had stored samples of plasma and underwent PET imaging of fibrillar Aβ burden with 11C-PiB (n = 60; eTable 3). Although all participants were nondemented at the time of the PiB-PET study, a range of in vivo amyloid burden is observed in cognitively normal individuals30 and increased amyloid deposition may represent the earliest phase of AD pathology in these subjects. Measuring plasma clusterin concentration from samples collected 10 years before the PiB-PET studies, we investigated associations between clusterin concentration and subsequent development of in vivo fibrillar amyloid burden.
We conducted a directed search of significant associations between clusterin and MTL PiB values using the MTL region defined by the WFU PickAtlas26 and the SPM5 multiple regression module, adjusting for age and sex. These results indicated that higher antecedent clusterin concentrations were associated with greater PiB retention in bilateral ERC; it was higher on the right (right ERC, P = .009; and left ERC, P = .03) (Figure 4A). This suggests that increased plasma concentration of clusterin, even in nondemented older individuals, predicts a greater extent of fibrillar amyloid burden in the ERC, the same region where we have also demonstrated robust association with atrophy in subjects with MCI and AD.
Gene expression of clusterin is altered in ad
To investigate the mechanisms underlying the associations between plasma concentration of clusterin and both imaging measures of atrophy and accelerated clinical progression, we measured clusterin mRNA levels in blood cells from AD patients (n = 182), MCI subjects (n = 179), and controls (n = 207) (eTable 4). Diagnosis had a significant effect on clusterin gene expression (ANCOVA, df = 2, P < .001, age as a covariate). Pairwise comparisons between the 3 groups showed significantly higher clusterin gene expression in AD patients than in MCI and control subjects (P = .008 and P < .001, respectively, Bonferroni adjustment for multiple comparisons) (Figure 4B). Sex and presence of the APOE ε4 allele did not have a significant effect on clusterin mRNA levels. We did not observe a significant association between clusterin mRNA in blood cells and plasma concentration of clusterin protein nor did we find a correlation between plasma mRNA levels and either MMSE score or rate of decline in MMSE score within groups or with atrophy on neuroimaging.
Lack of effect of variation in the clusterin gene on peripheral clusterin expression
We did not observe significant effects of the 7 clusterin gene SNPs on either clusterin mRNA expression in blood cells or plasma concentration of clusterin (eTable 5 and eTable 6). The SNPs analyzed included those reported on in the recent large Genome-Wide Association Studies to be associated with risk of sporadic AD.31,32
Plasma concentration of clusterin in transgenic mice with plaque pathology
To extend our findings on the association of clusterin with brain amyloid deposition, we examined its plasma concentration in a transgenic mouse model of AD. TASTPM mice overexpress the hAPP695swe and presenilin 1 M146V mutations, resulting in overproduction of human amyloid precursor protein,27 and mimic various hallmarks of AD including amyloid plaques as well as cognitive and behavioral deficits.27,28 In light of our magnetic resonance imaging data in AD patients and PiB-PET results in nondemented older individuals, we hypothesized that plasma clusterin concentration in transgenic TASTPM mice would be higher than wild-type controls. As predicted, we observed a significantly greater plasma concentration of clusterin (P = .02, independent samples t test) in 6-month-old transgenic TASTPM mice (n = 10) relative to wild-type litter mates (n = 10) (Figure 4C). Previous studies have established both marked cerebral Aβ deposits as well as cognitive deficits in TASTPM mice at this age relative to wild-type litter mates.27,28
Brain clusterin and amyloid in a transgenic mouse model of ad
Using double-labeling immunohistochemistry, we demonstrated that cortical plaques in TASTPM mice contained both Aβ and clusterin (Figure 4D). Finally, we established the close association between Aβ and clusterin by showing that both cortical Aβ burden and clusterin deposition increase with age in TASTPM mice (n = 9-11) (Figure 4E) and that there is a highly significant correlation (F1,37 = 107.57, P < .001, adjusted R2 = 0.737) between cortical Aβ and clusterin load (Figure 4F).
We have combined a novel proteomic and neuroimaging approach to establish that plasma concentration of clusterin is associated with in vivo pathology, disease severity, and clinical progression in patients with AD. The primary outcomes in our discovery-phase studies were association with both atrophy of the MTL and the rate of progression of cognitive decline. In the discovery phase, we used hippocampal atrophy derived from manual tracing of the hippocampal formation from magnetic resonance imaging, and in the much larger validation phase, from automated regional analysis of the ERC, an adjacent region of the MTL and the site of earliest pathology in AD.
Hippocampal atrophy is an early event in the pathogenesis of AD, is associated with an increased risk of conversion from MCI to AD, and may even precede the development of cognitive decline.33,34 Cerebrospinal fluid levels of phosphorylated tau correlate with hippocampal volume, indicating that this measure reflects an integral feature of AD pathology.35 Moreover, decreased hippocampal volume in AD patients is associated with neuronal loss, confirming its validity as a marker of neurodegeneration.35 A second independent outcome variable in the discovery-phase studies was rate of cognitive decline, derived as a measure of decrease in the ADAS-cog scores during a 6-month interval in AD patients. Using this measure, we dichotomized AD patients as fast and slow decliners, an approach previously shown to predict long-term prognosis in AD.36
Only clusterin was associated both with hippocampal atrophy in AD and MCI subjects and with fast progressing, or more aggressive, AD. Evidence from human cerebrospinal fluid, postmortem brain, and transgenic animal models suggests a plausible link between clusterin and AD pathology.37-40 We therefore sought to confirm the association of clusterin with AD pathology, severity, and progression in a much larger validation-phase study.
We confirmed highly significant associations of plasma clusterin concentration with atrophy of the ERC (P = .001), MMSE score (P < .001), and rate of progression in AD (P < .001). We also demonstrated a significantly greater risk of subsequent accelerated cognitive decline associated with increased concentration of clusterin in patients with AD and, in normal individuals, with subsequent deposition of fibrillar Aβ in the ERC. Our finding of raised plasma clusterin concentration 10 years before fibrillar Aβ deposition in the brain in normal elderly individuals suggests that clusterin is raised very early, possibly as an etiopathological event, and is not simply a reaction to other pathology in AD. The observation that clusterin mRNA is significantly increased in blood cells in AD suggests that the observed changes in protein levels reflect changes in expression in disease and not, for example, altered turnover. However, the increase in clusterin mRNA in AD patients does not correlate directly with plasma clusterin concentration, suggesting that the primary sources of plasma clusterin that we find predictive of more aggressive disease are organs other than blood cells such as the liver or possibly even the brain. In the course of this study, 2 groups, including one in which we participated, reported from genome-wide studies that polymorphic variation in CLU, which encodes clusterin, was associated with AD.31,32 One possible mechanism for this association would be for the SNPs associated with disease to be modifiers of gene expression. To investigate this, we determined the effect of variations in the clusterin gene on both peripheral mRNA levels and plasma concentration of clusterin protein, including the principal variant associated with disease and 6 other SNPs determined to cover most of the variation in the gene. We did not find significant effects of these SNPs on either peripheral mRNA levels or plasma clusterin concentration, suggesting that our observed association of clusterin protein and mRNA with AD-related pathological processes is independent of genetic variation in the clusterin gene. Our findings raise the possibility of 2 possibly linked mechanisms whereby both altered expression and some other factor in the gene linked to the disease-associated SNPs are active in moderating disease pathology. However, we cannot exclude an effect of genetic variation not examined in this study on clusterin expression or a small effect of CLU variation, below the power of our study to detect, on expression. Nonetheless, the finding of association with both genetic variants and, as we now report, gene and protein expression adds considerable weight to the importance of clusterin to AD pathogenesis. It is interesting that we observe clusterin in 2 closely related but distinct spots in the discovery-phase 2-dimensional gel electrophoresis studies. Proteins are components of multiple spots on 2-dimensional gel electrophoresis because of changes in posttranslational modification, complex formation, and splicing changes resulting in different isoforms. It is possible that some of these variations might be associated with disease processes in addition to the overall amount of protein as measured in the validation-phase study. Finally, we confirmed a previous report of significantly higher plasma concentration of clusterin in TASTPM mice overexpressing APP/PS1 mutations,41 and we also show that clusterin is closely associated with cortical amyloid plaques, showing an age-dependent concomitant increase with brain amyloid burden.
Previous studies suggest that clusterin belongs to a family of extracellular chaperones that regulate amyloid formation and clearance.42 In vitro experiments show that clusterin regulates amyloid formation in a biphasic manner with low clusterin to substrate ratios enhancing and higher ratios inhibiting amyloid formation, respectively.43 In mice, in vivo binding of Aβ to clusterin enhances its clearance and efflux through the blood-brain barrier.44 However, previous studies reporting differences in cerebrospinal fluid clusterin concentration between AD patients and controls have been inconclusive.39,40 Our findings may have implications for the discovery and characterization of other amyloid chaperone proteins in blood linked to AD pathogenesis. In this context, α2-macroglobulin has recently been characterized as an amyloid chaperone that inhibits fibril formation.45,46 In a previous proteomic analysis of plasma, we reported the differential expression of α2-macroglobulin in AD patients and have also found associations between the plasma concentration of α2-macroglobulin and hippocampal metabolite abnormalities in AD.7,47 In this previous study,7 in addition to α2-macroglobulin, we also identified components of the complement pathway associated with AD. In the discovery phase of the current study, we note many of the same proteins and also that clusterin may itself play a role in complement activation, suggesting that further examination of this pathway may be useful to identify markers associated with AD.7
In summary, we have used a novel proteomic-neuroimaging discovery paradigm in which the primary end points were well-established measures of pathology in the MTL and rate of disease progression. We identified clusterin as a plasma protein associated with disease pathology, severity, and progression in AD. Although these findings do not support the clinical utility of plasma clusterin concentration as a stand-alone biomarker for AD, they reveal a robust peripheral signature of this amyloid chaperone protein that is responsive to key features of disease pathology. Our findings clearly implicate clusterin, but there may well be other proteins in plasma related to the disease process, and indeed our previous studies and those of others suggest this is the case. These results may have wider implications for the identification of other amyloid chaperone proteins in plasma, both as putative AD biomarkers as well as drug targets of disease-modifying treatments.
Correspondence: Simon Lovestone, MRC Psych, PhD, Department of Old Age Psychiatry, National Institute for Health Research Biomedical Research Centre for Mental Health, Department of Psychological Medicine, King's College London, Institute of Psychiatry, De Crespigny Park, London SE5 8AF, England (simon.lovestone@kcl.ac.uk).
Submitted for Publication: September 13, 2009; final revision received December 17, 2009; accepted December 17, 2009.
Author Contributions: Dr Lovestone had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Funding/Support: The AddNeuroMed study was funded by the European Union as part of the FP6 InnoMed program. The authors also received funding from the Alzheimer's Research Trust, the National Institute for Health Research (NIHR) Biomedical Research Centre for Mental Health at the South London and Maudsley National Health Service (NHS) Foundation Trust and King's College London, the Bupa Foundation, and the Alzheimer's Society (fellowship to Dr Thambisetty).
Financial Disclosure: Intellectual property has been registered on the use of plasma proteins, including clusterin, for use as biomarkers for AD by King’s College London and Proteome Sciences, with Drs Lovestone and Thambisetty and others named as inventors. Dr Wong has grant support and/or contracts from the National Institutes of Health, Acadia, Avid, Bristol-Myers Squibb, GE, Intracellular, Johnson & Johnson, Eli Lilly, Merck, Orexigen, Otsuka, Roche, and Sanofi-Aventis.
Additional Contributions: A large number of research workers and colleagues made essential contributions to the collection of samples and data on the AddNeuroMed study. Nicola Dunlop and Nicola Archer (London); Tomasz Sobow, Radoslaw Magierski, Iwona Makowska, Marcin Flirski, and Marcin Wojtera (Lodz); Emanuela Costanzi and Roberta Cecchetti (Perugia); Merja Hallikainen, Teemu Paajanen, Ritva Vanninen, and Mervi Kononen (Kuopio); Emma Reynish (Toulouse); and Penelope Mavridaki and Eleni Kantoglou (Thessaloniki) all assisted in the clinical assessments and sampling of research participants. Rikke Lewinsky (Roskilde) contributed to the data from TASTPM mouse experiments. They received no compensation for this work beyond normal salary payments.
1.Aluise
CDSowell
RAButterfield
DA Peptides and proteins in plasma and cerebrospinal fluid as biomarkers for the prediction, diagnosis, and monitoring of therapeutic efficacy of Alzheimer's disease.
Biochim Biophys Acta 2008;1782
(10)
549- 558
PubMedGoogle Scholar 2.Hampel
HBurger
KTeipel
SJBokde
ALZetterberg
HBlennow
K Core candidate neurochemical and imaging biomarkers of Alzheimer's disease.
Alzheimers Dement 2008;4
(1)
38- 48
PubMedGoogle Scholar 3.Blennow
KVanMechelen
EHampel
H CSF total tau, Aβ42 and phosphorylated tau protein as biomarkers for Alzheimer's disease.
Mol Neurobiol 2001;24
(1-3)
87- 97
PubMedGoogle Scholar 4.Buerger
KEwers
MPirttila
TZinkowski
RAlafuzoff
ITeipel
SJDeBernardis
JKerkman
DMcCulloch
CSoininen
HHampel
H CSF phosphorylated tau protein correlates with neocortical neurofibrillary pathology in Alzheimer's disease.
Brain 2006;129
(pt 11)
3035- 3041
PubMedGoogle Scholar 5.Fagan
AMMintun
MAMach
RHLee
SYDence
CSShah
ARLaRossa
GNSpinner
MLKlunk
WEMathis
CADeKosky
STMorris
JCHoltzman
DM Inverse relation between in vivo amyloid imaging load and cerebrospinal fluid Aβ42 in humans.
Ann Neurol 2006;59
(3)
512- 519
PubMedGoogle Scholar 6.Britschgi
MWyss-Coray
T Blood protein signature for the early diagnosis of Alzheimer disease.
Arch Neurol 2009;66
(2)
161- 165
PubMedGoogle Scholar 7.Hye
ALynham
SThambisetty
MCausevic
MCampbell
JByers
HLHooper
CRijsdijk
FTabrizi
SJBanner
SShaw
CEFoy
CPoppe
MArcher
NHamilton
GPowell
JBrown
RGSham
PWard
MLovestone
S Proteome-based plasma biomarkers for Alzheimer's disease.
Brain 2006;129
(pt 11)
3042- 3050
PubMedGoogle Scholar 9.Lönneborg
A Biomarkers for Alzheimer disease in cerebrospinal fluid, urine, and blood.
Mol Diagn Ther 2008;12
(5)
307- 320
PubMedGoogle Scholar 10.Mrak
REGriffin
WS Potential inflammatory biomarkers in Alzheimer's disease.
J Alzheimers Dis 2005;8
(4)
369- 375
PubMedGoogle Scholar 11.Ray
SBritschgi
MHerbert
CTakeda-Uchimura
YBoxer
ABlennow
KFriedman
LFGalasko
DRJutel
MKarydas
AKaye
JALeszek
JMiller
BLMinthon
LQuinn
JFRabinovici
GDRobinson
WHSabbagh
MNSo
YTSparks
DLTabaton
MTinklenberg
JYesavage
JATibshirani
RWyss-Coray
T Classification and prediction of clinical Alzheimer's diagnosis based on plasma signaling proteins.
Nat Med 2007;13
(11)
1359- 1362
PubMedGoogle Scholar 12.Marksteiner
JKemmler
GWeiss
EM
et al. Five out of 16 plasma signaling proteins are enhanced in plasma of patients with mild cognitive impairment and Alzheimer's disease [published online ahead of print April 22, 2009].
Neurobiol Aging doi:10.1016/j.neurobiolaging.2009.03.011
Google Scholar 13.Shi
MCaudle
WMZhang
J Biomarker discovery in neurodegenerative diseases: a proteomic approach.
Neurobiol Dis 2009;35
(2)
157- 164
PubMedGoogle Scholar 14.Zhang
JGoodlett
DRPeskind
ERQuinn
JFZhou
YWang
QPan
CYi
EEng
JAebersold
RHMontine
TJ Quantitative proteomic analysis of age-related changes in human cerebrospinal fluid.
Neurobiol Aging 2005;26
(2)
207- 227
PubMedGoogle Scholar 15.Brooks
JO
IIIYesavage
JA Identification of fast and slow decliners in Alzheimer disease: a different approach.
Alzheimer Dis Assoc Disord 1995;9(suppl 1)S19- S25
PubMedGoogle Scholar 16.Kraemer
HCTinklenberg
JYesavage
JA ‘How far’ vs ‘how fast’ in Alzheimer's disease: the question revisited.
Arch Neurol 1994;51
(3)
275- 279
PubMedGoogle Scholar 17.Braak
HBraak
E Neuropathological stageing of Alzheimer-related changes.
Acta Neuropathol 1991;82
(4)
239- 259
PubMedGoogle Scholar 18.Lovestone
SFrancis
PStrandgaard
K Biomarkers for disease modification trials–the innovative medicines initiative and AddNeuroMed.
J Nutr Health Aging 2007;11
(4)
359- 361
PubMedGoogle Scholar 19.Petersen
RCSmith
GEWaring
SCIvnik
RJTangalos
EGKokmen
E Mild cognitive impairment: clinical characterization and outcome.
Arch Neurol 1999;56
(3)
303- 308
PubMedGoogle Scholar 20.Foy
CMNicholas
HHollingworth
PBoothby
HWillams
JBrown
RGAl-Sarraj
SLovestone
S Diagnosing Alzheimer's disease: non-clinicians and computerised algorithms together are as accurate as the best clinical practice.
Int J Geriatr Psychiatry 2007;22
(11)
1154- 1163
PubMedGoogle Scholar 21.Resnick
SMGoldszal
AFDavatzikos
CGolski
SKraut
MAMetter
EJBryan
RNZonderman
AB One-year age changes in MRI brain volumes in older adults.
Cereb Cortex 2000;10
(5)
464- 472
PubMedGoogle Scholar 22.Desikan
RSSegonne
FFischl
BQuinn
BTDickerson
BCBlacker
DBuckner
RLDale
AMMaguire
RPHyman
BTAlbert
MSKilliany
RJ An automated labeling system for subdividing the human cerebral cortex on MRI scans into gyral based regions of interest.
Neuroimage 2006;31
(3)
968- 980
PubMedGoogle Scholar 23.Fischl
Bvan der Kouwe
ADestrieux
CHalgren
ESégonne
FSalat
DHBusa
ESeidman
LJGoldstein
JKennedy
DCaviness
VMakris
NRosen
BDale
AM Automatically parcellating the human cerebral cortex.
Cereb Cortex 2004;14
(1)
11- 22
PubMedGoogle Scholar 24.Zhou
YEndres
CJBrasic
JRHuang
SCWong
DF Linear regression with spatial constraint to generate parametric images of ligand-receptor dynamic PET studies with a simplified reference tissue model.
Neuroimage 2003;18
(4)
975- 989
PubMedGoogle Scholar 25.Zhou
YResnick
SMYe
WFan
HHolt
DPKlunk
WEMathis
CADannals
RWong
DF Using a reference tissue model with spatial constraint to quantify [
11C]Pittsburgh compound B PET for early diagnosis of Alzheimer's disease.
Neuroimage 2007;36
(2)
298- 312
PubMedGoogle Scholar 26.Maldjian
JALaurienti
PJKraft
RABurdette
JH An automated method for neuroanatomic and cytoarchitectonic atlas-based interrogation of fMRI data sets.
Neuroimage 2003;19
(3)
1233- 1239
PubMedGoogle Scholar 27.Howlett
DRRichardson
JCAustin
AParsons
AABate
STDavies
DCGonzalez
MI Cognitive correlates of Aβ deposition in male and female mice bearing amyloid precursor protein and presenilin-1 mutant transgenes.
Brain Res 2004;1017
(1-2)
130- 136
PubMedGoogle Scholar 28.Howlett
DRBowler
KSoden
PERiddell
DDavis
JBRichardson
JCBurbidge
SAGonzalez
MIIrving
EALawman
AMiglio
GDawson
ELHowlett
ERHussain
I Aβ deposition and related pathology in an APP × PS1 transgenic mouse model of Alzheimer's disease.
Histol Histopathol 2008;23
(1)
67- 76
PubMedGoogle Scholar 29.Eriksson
LAntti
HGottfries
JHolmes
EJohansson
ELindgren
FLong
ILundstedt
TTrygg
JWold
S Using chemometrics for navigating in the large data sets of genomics, proteomics, and metabonomics.
Anal Bioanal Chem 2004;380
(3)
419- 429
PubMedGoogle Scholar 30.Mintun
MALarossa
GNSheline
YIDence
CSLee
SYMach
RHKlunk
WEMathis
CADeKosky
STMorris
JC [
11C]PIB in a nondemented population: potential antecedent marker of Alzheimer disease.
Neurology 2006;67
(3)
446- 452
PubMedGoogle Scholar 31.Harold
DAbraham
RHollingworth
PSims
RGerrish
AHamshere
MLPahwa
JSMoskvina
VDowzell
KWilliams
AJones
NThomas
CStretton
AMorgan
ARLovestone
SPowell
JProitsi
PLupton
MKBrayne
CRubinsztein
DCGill
MLawlor
BLynch
AMorgan
KBrown
KSPassmore
PACraig
DMcGuinness
BTodd
SHolmes
CMann
DSmith
ADLove
SKehoe
PGHardy
JMead
SFox
NRossor
MCollinge
JMaier
WJessen
FSchürmann
Bvan den Bussche
HHeuser
IKornhuber
JWiltfang
JDichgans
MFrölich
LHampel
HHüll
MRujescu
DGoate
AMKauwe
JSCruchaga
CNowotny
PMorris
JCMayo
KSleegers
KBettens
KEngelborghs
SDe Deyn
PPVan Broeckhoven
CLivingston
GBass
NJGurling
HMcQuillin
AGwilliam
RDeloukas
PAl-Chalabi
AShaw
CETsolaki
MSingleton
ABGuerreiro
RMühleisen
TWNöthen
MMMoebus
SJöckel
KHKlopp
NWichmann
HECarrasquillo
MMPankratz
VSYounkin
SGHolmans
PAO’Donovan
MOwen
MJWilliams
J Genome-wide association study identifies variants at CLU and PICALM associated with Alzheimer's disease.
Nat Genet 2009;41
(10)
1088- 1093
PubMedGoogle Scholar 32.Lambert
JCHeath
SEven
GCampion
DSleegers
KHiltunen
MCombarros
OZelenika
DBullido
MJTavernier
BLetenneur
LBettens
KBerr
CPasquier
FFiévet
NBarberger-Gateau
PEngelborghs
SDe Deyn
PMateo
IFranck
AHelisalmi
SPorcellini
EHanon
Ode Pancorbo
MMLendon
CDufouil
CJaillard
CLeveillard
TAlvarez
VBosco
PMancuso
MPanza
FNacmias
BBossù
PPiccardi
PAnnoni
GSeripa
DGalimberti
DHannequin
DLicastro
FSoininen
HRitchie
KBlanché
HDartigues
JFTzourio
CGut
IVan Broeckhoven
CAlpérovitch
ALathrop
MAmouyel
PEuropean Alzheimer's Disease Initiative Investigators, Genome-wide association study identifies variants at CLU and CR1 associated with Alzheimer's disease.
Nat Genet 2009;41
(10)
1094- 1099
PubMedGoogle Scholar 33.Kantarci
KJack
CR
Jr Neuroimaging in Alzheimer disease: an evidence-based review.
Neuroimaging Clin N Am 2003;13
(2)
197- 209
PubMedGoogle Scholar 34.Nestor
PJScheltens
PHodges
JR Advances in the early detection of Alzheimer's disease.
Nat Med 2004;10(suppl)S34- S41
PubMedGoogle Scholar 35.Hampel
HBurger
KPruessner
JCZinkowski
RDeBernardis
JKerkman
DLeinsinger
GEvans
ACDavies
PMöller
HJTeipel
SJ Correlation of cerebrospinal fluid levels of tau protein phosphorylated at threonine 231 with rates of hippocampal atrophy in Alzheimer disease.
Arch Neurol 2005;62
(5)
770- 773
PubMedGoogle Scholar 36.Helmer
CAndrieu
SPeres
KOrgogozo
JMVellas
BDartigues
JFREAL.fr Group, Predictive value of 6-month decline in ADAS-cog for survival without severe Alzheimer's disease.
Dement Geriatr Cogn Disord 2007;23
(3)
168- 174
PubMedGoogle Scholar 37.DeMattos
RBO'Dell
MAParsadanian
MTaylor
JWHarmony
JABales
KRPaul
SMAronow
BJHoltzman
DM Clusterin promotes amyloid plaque formation and is critical for neuritic toxicity in a mouse model of Alzheimer's disease.
Proc Natl Acad Sci U S A 2002;99
(16)
10843- 10848
PubMedGoogle Scholar 38.Lidström
AMBogdanovic
NHesse
CVolkman
IDavidsson
PBlennow
K Clusterin (apolipoprotein J) protein levels are increased in hippocampus and in frontal cortex in Alzheimer's disease.
Exp Neurol 1998;154
(2)
511- 521
PubMedGoogle Scholar 39.Nilselid
AMDavidsson
PNagga
KAndreasen
NFredman
PBlennow
K Clusterin in cerebrospinal fluid: analysis of carbohydrates and quantification of native and glycosylated forms.
Neurochem Int 2006;48
(8)
718- 728
PubMedGoogle Scholar 40.Sihlbom
CDavidsson
PSjogren
MWahlund
LONilsson
CL Structural and quantitative comparison of cerebrospinal fluid glycoproteins in Alzheimer's disease patients and healthy individuals.
Neurochem Res 2008;33
(7)
1332- 1340
PubMedGoogle Scholar 41.Cutler
PAkuffo
ELBodnar
WMBriggs
DMDavis
JBDebouck
CMFox
SMGibson
RAGormley
DAHolbrook
JDHunter
AJKinsey
EEPrinjha
RRichardson
JCRoses
ADSmith
MATsokanas
NWillé
DRWu
WYates
JWGloge
IS Proteomic identification and early validation of complement 1 inhibitor and pigment epithelium-derived factor: two novel biomarkers of Alzheimer's disease in human plasma.
Proteomics Clin Appl 2008;2
(4)
467- 477
Google Scholar 42.Wilson
MRYerbury
JJPoon
S Potential roles of abundant extracellular chaperones in the control of amyloid formation and toxicity.
Mol Biosyst 2008;4
(1)
42- 52
PubMedGoogle Scholar 43.Yerbury
JJPoon
SMeehan
SThompson
BKumita
JRDobson
CMWilson
MR The extracellular chaperone clusterin influences amyloid formation and toxicity by interacting with prefibrillar structures.
FASEB J 2007;21
(10)
2312- 2322
PubMedGoogle Scholar 44.Bell
RDSagare
APFriedman
AEBedi
GSHoltzman
DMDeane
RZlokovic
BV Transport pathways for clearance of human Alzheimer's amyloid β-peptide and apolipoproteins E and J in the mouse central nervous system.
J Cereb Blood Flow Metab 2007;27
(5)
909- 918
PubMedGoogle Scholar 45.French
KYerbury
JJWilson
MR Protease activation of α
2-macroglobulin modulates a chaperone-like action with broad specificity.
Biochemistry 2008;47
(4)
1176- 1185
PubMedGoogle Scholar 46.Yerbury
JJKumita
JRMeehan
SDobson
CMWilson
MR α
2-Macroglobulin and haptoglobin suppress amyloid formation by interacting with prefibrillar protein species.
J Biol Chem 2009;284
(7)
4246- 4254
PubMedGoogle Scholar 47.Thambisetty
MHye
AFoy
CDaly
EGlover
ACooper
ASimmons
AMurphy
DLovestone
S Proteome-based identification of plasma proteins associated with hippocampal metabolism in early Alzheimer's disease.
J Neurol 2008;255
(11)
1712- 1720
PubMedGoogle Scholar