Molecular Psychiatry (2012) 17, 1056–1076 & 2012 Macmillan Publishers Limited All rights reserved 1359-4184/12 www.nature.com/mp EXPERT REVIEW The neuropathological profile of mild cognitive impairment (MCI): a systematic review BCM Stephan1, S Hunter1, D Harris2, DJ Llewellyn3, M Siervo4, FE Matthews5 and C Brayne1 1Department of Public Health and Primary Care, Institute of Public Health, Cambridge University, Cambridge, UK; 2NIHR Peninsula Collaboration for Leadership in Applied Health Research and Care (PenCLAHRC), Peninsula College of Medicine and Dentistry, University of Exeter, Exeter, UK; 3Epidemiology and Public Health Group, Peninsula College of Medicine and Dentistry, University of Exeter, Exeter, UK; 4Nutrition Physiology, Department of Neuroscience, University of Naples, Naples, Italy and 5MRC Biostatistics Unit, Cambridge, UK

Whether mild cognitive impairment (MCI) has a distinct neuropathological profile that reflects an intermediate state between no cognitive impairment and dementia is not clear. Identifying which biological events occur at the earliest stage of progressive disease and which are secondary to the neuropathological process is important for understating pathological pathways and for targeted disease prevention. Many studies have now reported on the neurobiology of this intermediate stage. In this systematic review, we synthesize current evidence on the neuropathological profile of MCI. A total of 162 studies were identified with varied definition of MCI, settings ranging from population to specialist clinics and a wide range of objectives. From these studies, it is clear that MCI is neuropathologically complex and cannot be understood within a single framework. Pathological changes identified include plaque and tangle formation, vascular pathologies, neurochemical deficits, cellular injury, inflammation, oxidative stress, mitochondrial changes, changes in genomic activity, synaptic dysfunction, disturbed protein metabolism and disrupted metabolic homeostasis. Determining which factors primarily drive neurodegeneration and dementia and which are secondary features of disease progression still requires further research. Standardization of the definition of MCI and reporting of pathology would greatly assist in building an integrated picture of the clinical and neuropathological profile of MCI. Molecular Psychiatry (2012) 17, 1056–1076; doi:10.1038/mp.2011.147; published online 6 December 2011 Keywords: Alzheimer’s disease (AD); brain; dementia; mild cognitive impairment; neuropathology; vascular dementia

Background state, or is simply early-stage dementia, which itself has multiple possible substrates. In this review, we The concept of mild cognitive impairment (MCI) describe the published literature on the neuropatho- defines an intermediate state between no cognitive logical profile of MCI (Supplementary Table 1). The impairment (NCI) and pathological cognitive ageing methodology for article selection is outlined in with the aim of identifying individuals who are at Supplementary Appendix 1. The focus is on findings high risk of dementia. Numerous different definitions which might establish MCI as a distinct and identifi- of MCI have been proposed with large variations in able state between control and dementia groups. Key the type of impairment that needs to be present (for issues addressed include: (1) the extent to which example, amnestic vs non-amnestic) and the level of 1 Alzheimer’s disease (AD) and non-AD pathology deficit severity that is required. Despite intense associated with clinical dementia are observed in research to characterize the neuropsychological and MCI; (2) the molecular changes that have been found clinical profile of MCI and its progression to demen- and, (3) changes in key processes such as neurotrans- tia, the actual neuropathological substratum has not mitter signaling and cellular responses (including yet been elucidated for any definition of MCI. Defined downregulation and oxidation). as a transitional state, MCI is a difficult biological phenomenon to investigate, especially in terms of testing claims about whether it represents an at-risk Diagnosing MCI

Correspondence: Dr BCM Stephan, Department of Public Health Across articles, there is considerable variability in and Primary Care, University of Cambridge, University Forvie MCI criteria as shown in Supplementary Table 1. Site, Robinson Way, Cambridge CB2 0SR, UK. Although the diagnosis of Mayo Clinic-defined MCI E-mail: [email protected] 2 Received 23 March 2011; revised 19 September 2011; accepted (amnestic MCI (aMCI)) is the most widely applied in 4 October 2011; published online 6 December 2011 clinical and research practice, only 45 articles have Neuropathology of MCI BCM Stephan et al 1057 mapped this definition. The amnestic subtype has memory is impaired (that is, aMCI) than groups in been applied almost exclusively (vs non-amnestic or which memory is intact (that is, non-amnestic MCI multi-domain MCI). In 55 articles, MCI is defined (nMCI)).4 using a Clinical Dementia Rating (CDR) score of 0.5. However, there is still disagreement about whether AD and non-AD pathology such individuals should be included with AD cases.3 In the Religious Orders Study, which of all studies Neuropathological manifestations of AD has produced the most publications (46 articles), the In terms of AD neuropathology, MCI generally reflects MCI case definition was: ‘a neuropsychological an intermediate state that shares to a certain degree impairment (by clinical rating) with no dementia’ the histopathological hallmarks of AD in vulnerable for all but 1 publication. Therefore, reproducibility of cortical regions including memory-related regions, findings in the associations is affected by differences the hippocampus and the visual association in MCI case diagnosis across sources, neuropsycho- cortex.4–50 In some studies, these changes are inter- logical testing batteries used to define cognitive preted to represent one of the earliest pathological impairment and differences in clinical ratings. This substrates of this condition with MCI more similar to is problematic as participants diagnosed with MCI are AD than to control cases.8,24,26,45–47,51 Relative to not directly comparable across all studies. Further- controls, individuals with MCI have been found to more, there may be bias towards specific types of show a significant increase in neuritic plaques (NPs) pathology (for example, AD pathology) depending on in the neocortical regions52 and neurofibrillary the MCI definition used. Here, we give an overview of tangles (NFTs) in the hippocampus, amygdala and all neuropathological findings relating to an immedi- entorhinal cortex (EC).33,53 In contrast, in other ate cognitive group between NCI and dementia reports, MCI cases are found to be more similar and regardless of definition. Further research is necessary often indistinguishable from individuals with to elucidate whether different MCI criteria are NCI.20,22,36,44,54–63 associated with different neuropathological profiles, These results raise questions regarding the thresh- and this would require detailed information on the old necessary for the burden to be pathological and operationalization of MCI criteria in specific study the existence of neural tolerance or compensation populations. effects that may be related to the severity and location of the pathology (that is, hippocampus and EC spreading to the neocortex and primary visual system Global diagnostic findings at later, more severe disease stages), the presence of Neuropathological diagnosis of AD disease comorbidity, brain reserve and donor The diagnostic scheme usually used is the NIA-Reagan age.5,14,19,20,54,56,60,64–67 Overall, despite variability in (National Institute on Aging-Reagan Institute criteria) MCI definition, generally there is an increased with MCI brains showing a variable dementia prob- presence of NFTs and NPs, although not always at ability rating ranging from low (as typically seen in the level associated with a neuropathological NCI cases) to high (as typically seen in dementia) as diagnosis of AD. shown in Supplementary Table 1. Similar variability Abnormal tau has been found to increase with and a high probability of a neuropathological diag- increasing dementia severity measured using the CDR nosis of AD is also found when classification is based score. Abnormal phosphorylation and conformational on the Consortium to Establish a Registry for AD changes (MC-1 and TG-3) are observed in mild (CERAD) protocol, with MCI brains diagnosed with (CDR = 1)-to-severe (CDR = 5) dementia stages, with possible, probable or definite AD, or the Khachaturian MCI (CDR = 0.5) and control (CDR = 0) cases simi- criteria with most MCI brains diagnosed as having lar.68,69 The antibody MC-1 recognizes a pathological, Alzheimer-type pathology. However, there is also a compact conformation of tau that is associated with large proportion of MCI cases that do not show increased potential for tau to aggregate into paired significant AD neuropathological changes. The helical filaments (PHF).70,71 The TG-3 antibody recog- schemes used to classify AD pathology were designed nizes an altered tau conformation when phosphory- to distinguish NCI and AD groups, and this could in lated at Thr231.72,73 Both these antibodies recognize part explain their lack of discriminative accuracy in changes in tau structure and/or phosphorylation at categorizing MCI. what are considered to be early stages of disease. Compared with NFT pathology, the severity of Braak NFT Staging plaque pathology has been found to be less discrimi- Mild cognitive impairment cases have variable Braak nating of cognitive status across the NCI-MCI-early staging scores ranging from a complete absence of AD continuum.4,9,14,18,19,25,31,32,63,67,74,75 In some19,20,44,76 pathology (stage 0) to stages V and VI representing reports, the volume of plaques seems to remain stable severe dementia. Similar variability is seen in NCI with disease progression, with significant changes and dementia cases. There is no obvious boundary only observed at late disease stages. These results between control, MCI and AD groups using Braak have been taken to suggest that abnormal accumula- staging. However, it has been suggested that Braak tion of plaques is not critical for the transition from staging is better at identifying MCI groups in which NCI to MCI,56 being necessary only for transitions to

Molecular Psychiatry Neuropathology of MCI BCM Stephan et al 1058 dementia.20 However, the effect of amyloid deposition However, evidence of vascular disease29,87 and vas- on cognitive function is understood to be mediated by cular-related pathology has been found including plaque subtype (greater burden with neuritic vs infarcts,7–9,19,26,37,88 cortical micro-infarcts,19,25,36,89,90 diffuse plaques),20,76,77 co-occurring NFT burden,78 thalamic and basal ganglia lacunes,36,50 periventricu- apolipoprotein E (APOE) e4 gene dosage,18,41,42 lar demyelination,50,89,90 deep white matter demyeli- age42,56 and compensatory mechanisms linked to nation,50,89 subdural hematoma,88 hippocampal education level.79 Plaque–disease associations may sclerosis,7,19,23,26 atherosclerosis,4 cerebral contu- also be missed because of methodological features.80 sion,23 metabolic encephalopathy23 and evidence of By adopting a stereological rather than a semi- morphological alterations in capillaries.91 quantitative approach to amyloid assessment, a In one article, in which the sample of individuals significant relationship between cognitive status was restricted to cases without significant tangle (CDR score) and the total volume occupied by pathology (Braak stage I–II only), the presence amyloid deposits in the subiculum and the EC (CDR of micro-vascular changes including cortical and Mini Mental State Examination (MMSE) score) micro-infarcts, deep white matter demyelination and has been found.44 Furthermore, studies of amyloid periventricular demyelination in the hippocampus, load based on amyloid-beta-protein (Ab) levels have frontal, parietal and temporal areas predicted found that Ab42 (a cleavage product of the amyloid cognitive function defined using the CDR score.89 precursor protein: APP) (and Ab43), and to a lesser Similarly, using the same neuropathology resource,90 extent Ab40 correlate with dysfunction in AD and micro-infarcts and periventricular demyelination also may mediate pathological changes associated with predicted CDR scores in individuals with significant early disease.57,69,77,81 However, although age-adjusted tangle pathology (Braak stage III). In contrast, focal group comparisons of insoluble Ab40 and Ab42 levels cortical and white matter gliosis were not associated in the cortex correlate with neuropathological AD with clinical outcome. status (CERAD and Braak staging),82 they do not correlate with clinical diagnosis at death (NCI, MCI, Interhemispheric distribution of AD and non-AD early AD). MCI and early AD are only distinguishable pathology from controls in terms of the Ab42/Ab40 ratio (both The hemispheric distribution of AD and vascular soluble and insoluble).57 The importance of this pathology seems to be associated with severity of change remains to be elucidated. cognitive impairment (CDR score).6 AD pathology Overall, there seems to be no clear relationship including NFT and Ab deposition, is found to between the classic neuropathological markers of AD populate the brain symmetrically even in NCI cases and progression from NCI to MCI. Indeed, even in (CDR = 0), with a left predominance of vascular non-dementia normal groups, the level of pathology pathology independent of CDR score. However, can be significant enough for a neuropathological interhemispheric differences, namely an increase in diagnosis of dementia.83,84 Whether the neuropatho- AD pathology staging and vascular pathology in the logical features are directly involved in disease right hemisphere, are found to be associated with an pathways or are markers for other processes, some increase in CDR score. These results are based on a of which may be protective, has not yet been single study and need to be replicated. conclusively shown.85 Argyrophilic grain disease Ab oligomers Argyrophilic grain disease has been found in some In MCI and mild/moderate AD (CDR = 1 or CDR = 2), cases of MCI,10,19,23,25 but has not been reported there is a progressive accumulation of Ab monomers, routinely, and so it is difficult to discuss particular dimers and higher-order oligomers in the frontal patterns. cortex compared with NCI cases.86 Levels of oligo- mers are found to correlate with severity of cognitive Cerebral amyloid angiopathy impairment (MMSE and Blessed Information Memory Cerebral amyloid angiopathy has been observed in Concentration score) and AD neurodegeneration some cases of MCI,16,25,64 but has also not been (Braak staging). Furthermore, oligomers appear to investigated systematically. In one study, cerebral interrupt synaptic function as shown by an associa- amyloid angiopathy was not found to correlate with tion with loss of synaptic markers including CDR score.52 synaptic vesicle protein vesicle-associated membrane protein-2 and postsynaptic density-95 (a marker of LB pathology postsynaptic density). Therefore, accumulation of Ab In rare cases, Lewy body (LB) pathology is found in monomers and oligomers seems to be involved in the individuals with MCI7,8,19,25,29,43,64,88 and in some is of initial stages of disease resulting in synaptic dysfunc- sufficient severity for a neuropathological diagnosis tion and impaired cognition. of LB dementia10,23,26,43,87 or mixed AD and LB dementia.7,43 In other cases, no evidence of LB Vascular-related pathology pathology is found.24,30,37 No variable currently exists Very few MCI cases receive a primary neuropatholo- that distinguishes MCI with and without LB pathol- gical diagnosis of vascular dementia.7,23,26 ogy in life. However, using cognitive and psychiatric

Molecular Psychiatry Neuropathology of MCI BCM Stephan et al 1059 data, one hospital-based autopsy series distinguished correlate with cognitive performance including a neuropathologically confirmed AD-risk state de- MMSE score.101 Synaptic loss is found to provide a fined using criteria for aMCI, from a pre-clinical LB better indicator for cognitive impairment than dementia state (MCI-LB).73 LB pathology was found b-amyloid plaque burden.102 For a narrative review only in MCI-LB cases and more AD pathology (higher of synaptic degeneration in MCI and AD see the study Braak, CERAD and NIA-Reagan scores) was observed by Arendt.103 in the aMCI group, with the exception of diffuse plaques in which the density was comparable. Tauopathy in the cholinergic nucleus basalis (NB) area Alzheimer’s disease-related neuronal changes have Atrophy been found in the cholinergic system including an Relative to NCI cases, MCI and dementia cases show increase in tauopathy (Alz-50 and AT8 immunoreac- significant loss of neurons in the EC (primarily layers tivity). Alz-50 recognizes a conformation-dependent, II and IV).92,93 This supports in vivo imaging studies of non-contiguous, N-terminal epitope of tau and has a significant degeneration of the EC even before the high affinity for tau in PHF.104 AT8 recognizes an onset of dementia, in individuals with MCI and in epitope doubly phosphorylated at serine 202 and at those with subjective memory complaint but no threonine 205 with some cross-reaction with doubly objective memory deficit.94,95 In one study, using phosphorylated tau at either serines 199 and 202 or postmortem magnetic resonance imaging, reduced threonine 205 and serine 208.105,106 Tauopathy con- hippocampal volume was found to accurately predict taining NB neurons are present in controls and neuropathological fulfillment of AD criteria in become significantly more prominent with disease individuals with and without memory impairment.96 progression including MCI and early AD stages.39 This supports findings from in vivo neuroimaging Such changes also correlate with memory perfor- studies linking hippocampal atrophy to cognitive mance and are suggested to contribute to memory impairment and MCI.97 In contrast, only at advanced impairment observed in the NCI-MCI-AD conti- disease stages (CDR = 3) is significant neuronal loss nuum.39 Although expression levels of tau isoforms, observed in the superior temporal sulcus.31,32 b-tubulin, microtubule-associated proteins and neurofilament subunits was not different across NCI, Dendritic spine loss (Sp-ir) MCI and early/mild AD, single gene profiling Dendritic spines are protrusions from dendritic shafts revealed a significant reduction in the expression that are essential for excitatory synaptic transmission ratio of three-repeat tau to four-repeat tau mRNAs in in the central nervous system. Dendritic spine loss cholinergic basal forebrain (CBF) neurons in MCI and measured using the total number of spinophilin AD that does not occur in NCI.107 This is considered immunoreactivity (Sp-ir) puncta in the hippocampal to be linked to aberrant regulation in normal tau CA1 field has been found to be associated with function contributing to NFT formation. Therefore, the cognitive function (CDR score), but the association NB appears to be an early target for NFT changes of AD. was lost after Braak stage adjustment.98 This suggests that the cognitive consequences of dendritic spine The locus coeruleus (LC) noradrenergic system loss in the CA1 are mediated by global NFT burden. In Increasing numbers of NFT are seen in noradrenergic area 9, the association between CDR scores and Sp-ir neurons projecting from the locus coeruleus (LC) in puncta numbers remained significant even after Braak MCI and early AD (AT8-positive labeling).108 Changes adjustment, suggesting that neocortical dendritic in the LC correlate with global cognitive function spine loss may have an independent impact on (MMSE score). Although the role of the LC in disease cognition. In contrast, degradation of the neuronal progression is not clear, there is evidence in the cell body, nucleus and nucleolus in the hippocampus wider literature linking changes in the LC to cognitive has been observed only at advanced disease stages.99 impairment.109 Therefore, LC pathology may mediate the onset of cognitive impairment in the NCI-MCI- Synaptic loss dementia continuum. Significant reduction in the synaptic number in the inferior temporal gyrus and volume of the lamina 3 Components of the cell cycle has been observed in AD and aMCI compared with NCI cases.100 Synaptic counts in the lamina 3 were The mechanisms leading to neuronal death and also found to correlate with MMSE scores and total neurodegeneration with disease progression remain synaptic number with word recognition and animal unclear. One mechanism is believed to be linked to naming, but not APOE genotype, plaque density or regionally specific deregulation of various compo- Braak stage.100 An electron microscopy study of nents of the cell cycle.110 Aberrant expression of cell- synaptic loss in the outer molecular layer neurons of cycle proteins and unscheduled cell-cycle events are the hippocampal dentate gyrus found significant loss deleterious to neurons and can result in apoptosis and in individuals with early-onset AD compared with neurodegeneration.111 In MCI and AD, a significant the MCI and NCI groups.101 Although there was a increase in neurons immunopositive for factors reduction in the MCI relative to the NCI group, including proliferating cell nuclear antigen involved the results were not significant. Synapse loss did in control of DNA replication, and the cyclins, cyclin D

Molecular Psychiatry Neuropathology of MCI BCM Stephan et al 1060 and cyclin B, involved in regulation of cyclin- 12–40 years), and inhibition of Csp-6 may therefore dependent kinases, has been observed predominately delay or even prevent disease progression.122 Caspase- in the hippocampus, basal nucleus and EC.34 The cleaved tau is detected early in the progression of highest level of reactivity has been found in regions AD being observed in MCI brains, but not in NCI.126 with the highest tangle pathology (that is, phosphory- However, only at more advanced disease stages lated tau expression levels). Expressions of cyclin- including AD and the transitional state between dependent kinase-2, cyclin-dependent kinase-5 and MCI and AD, does DTau become more insoluble. cyclin G1 have also been found to be increased in Therefore, Csp inhibitors could be a potential ther- MCI and AD in the hippocampus and inferior parietal apeutic target for AD or for the prevention/delay of lobe (IPL), suggesting re-entry of cells into the cell MCI progression to dementia. division cycle in early disease phases.27 However, In MCI and early AD, changes to cellular mechan- these results could be indicative of aberrant signaling isms associated with neuronal protection against involving intracellular kinase cascades at early stages oxidative and other central nervous system stressors of disease progression.103,112 Perturbations in cell- have been observed. The expressions of both the pro- cycle proteins, with cyclin-dependent kinase-5 being apoptotic protein Bax (Bcl-2-associated X protein) the best studied, could reflect aberrant signaling and the anti-apoptotic protein Bcl-2 (B-cell lympho- systems involved in multiple synaptic processes such ma 2) are increased. Bax is increased in the IPL after as axonal transport,113 organization of the cytoskele- aminophospholipid (PtdSer) asymmetry in MCI and ton,114 tau phosphorylation,115 synapse formation and AD (vs NCI)123 and in the hippocampus.128 Erythro- dendritic organization,116 synaptic vesicle endo- poietin is a glycoprotein hormone secreted mainly by cytosis,117,118 regulation of gene expression119 and the kidney. In the brain, it acts as a cytokine and expression of neurotransmitter receptors at the cell regulates inflammatory and apoptotic pathways. It surface120,121 rather than aberrant entry into the cell acts with Bcl-xL (antiapoptotic member of the Bcl-2 cycle per se. family) to inhibit apoptosis.129 There is an increased expression of the erythropoietin receptor in MCI and Factors associated with apoptosis early AD in the hippocampus and temporal cortex The caspases (Csp) are a family of cysteine endopep- (layers I–IV), with the highest level in the MCI group tidases involved in the regulation of inflammation observed in the temporal cortex.37 and apoptosis. The executioner Csps, namely Csp-3 Heat-shock proteins (HSPs) promote cell survival in and Csp-6, promote apoptosis and both are increased response to environmental stressors, including oxida- in MCI and AD.122,123 Caspase gene expression is also tive stress.130 The molecular chaperones HSP 27 and altered with an increase in disease severity.124 In the HSP 70 are increased in areas affected by AD EC, the messenger RNA expression of Csp-1 and Csp- pathology (hippocampus and IPL, but not cerebel- 7 is elevated in MCI (CDR = 0.5), and coincides with lum), whereas induction of other HSPs including increased poly (ADP-ribose) polymerase cleavage but HSPs 90, 60 and 32 are low.131 Such changes may not apoptotic cell injury.124 At later disease stages have a neuroprotective effect early in the pathogen- (CDR = 5), a greater elevation is observed in Csp-1, esis of AD in the MCI stage, reaching peak density in Csp-2L, Csp-3, Csp-5, Csp-6, Csp-7, Csp-8, Csp-9 and preclinical stages and then diminishing with disease terminal deoxynucleotidyl transferase dUTP nick-end progression. Maintenance of these responses during labeling-positive cells. However, conflicting evidence disease progression may prevent conversion from suggests that expression of Csp-3, Csp-6 (and Csp-7) is MCI to dementia in which oxidative stress is the not increased by increased signaling earlier in the Csp underlying dysfunction. pathway and instead apoptosis is aborted, promoting neuronal survival.125 Peptidyl-prolyl cis/trans-isomerase (Pin1) Cleavage of tau at Asp421 (DTau) by Csps is found Pin1 is a nuclear protein involved in the regulation of to occur early in AD tangle evolution.126,127 Caspase mitosis and is a negative regulator of the transforming cleavage of tau has been shown to enhance filament growth factor beta pathway. Pin1 activity and phos- formation, and Csp activation is considered to phorylated tau levels are found to be increased in the be induced by the accumulation of Ab. In AD, frontal cortex in MCI and AD.35 Levels of each are Csp-cleaved tau colocalizes with Ab1–42 and activated positively correlated with each other in MCI (trend) Csp-3.126 Csp-6 is activated in neurons on apoptotic and AD (robust relationship), suggesting that insult and the active Csp-6 and tau cleaved by Cps-6 increased Pin1 response in early disease stages may (TauDCsp-6) is found in NP, neuropil threads and reflect a compensatory mechanism to counteract NFTs in the hippocampus in all stages of AD (that is, increases in AD pathology such as neurodegeneration mild, moderate and severe) independent of disease and neuronal apoptosis. Pin1 function has also been stage and severity.122 Such changes have also been found to be abnormal in MCI and AD due to oxidative found in the hippocampus and EC of individuals with modification.132 Outcomes from this have been MCI and the oldest-old who do not manifest cognitive hypothesized to include NFT formation, neuronal impairment, being inversely related to global cogni- apoptosis, loss of protector protein transported and tive scores in the latter group.122 Csp-6 activation is downregulation of other key proteins, in addition to absent in cognitively normal young individuals (aged oxidized and damaged DNA.132

Molecular Psychiatry Neuropathology of MCI BCM Stephan et al 1061 Regulation of the cell cycle, its role in the brain and synapsin isoform. Synapsin is involved in neuro- regulation of apoptosis are currently not well under- transmitter release with the a- and b-type functionally stood. The contributions of various components of distinct. Type-a has a role in synaptic plasticity and cell-cycle signaling pathways and their relationships the kinetics of neurotransmitter release. As the MCI to disease progression require further investigation. and NCI groups had a similar level of AD pathology As described above, both neurotoxic and neuropro- (NP and NFTs), the results suggest that decreased tective factors are upregulated in MCI and AD, expression of the a-type synapsin was not due to suggesting that environmental stressors and cellular neuronal loss, but a selective alteration in protein responses to stress have an important role early in expression. A later study by the same group used disease progression. Therefore, maintenance of these protein array analysis to screen 750 proteins of which neuroprotective responses during disease progression in MCI (CDR = 0.5), 50 were found to be altered may prevent progression from MCI to dementia. > 2-fold (typically downregulation) compared with However, this requires testing. the NCI group (CDR = 0). The differentially regulated proteins were categorized into five functional groups Synaptic protein expression including proteins related to synaptic function and neurotransmitters, cytoskeleton and cell adhesion, Synaptic protein expression the cell cycle, apoptosis and transcription and Specific synaptic proteins have been found to have translation.136,137 discordant expression across groups depending on the protein studied and the neocortical region a-Synuclein levels investigated. In MCI, drebrin (DRB: an actin-binding a-Synuclein is normally a soluble protein, localized protein involved in dendritic plasticity and cell primarily at the pre-synaptic region of axons. It may motility) levels are found to be significantly down- regulate dopamine release, induce fibrillization of tau regulated in the superior temporal cortex (vs NCI) and and protect neurons from various apoptotic stimuli. upregulated in the superior frontal cortex (SFC) (vs Insoluble a-synuclein levels have not been found to NCI and AD).133 In contrast, cortical synaptotagmin133 correlate with neuropathological criteria for AD (a group of transmembrane proteins involved in (CERAD or Braak stage) nor clinical diagnosis.82 membrane trafficking)134 and postsynaptic density- However, decreases in soluble a-synuclein in the 95 (frontal cortex)135 levels do not differ across the frontal cortex have been found to correlate with NCI-MCI-AD continuum. In a later study, postsynap- CERAD diagnosis and global neuropsychological tic density-95 level was found to be decreased in the scores, and are also found to be significantly lower hippocampus in aMCI cases relative to controls.128 In in early AD vs NCI and MCI.28 No difference is addition, N-methyl-D-aspartate receptor subunit 2A observed between the NCI and MCI groups. Therefore, and the low-density lipoprotein receptor-1 levels both a-synuclein pathology levels may be a reliable marker of which are associated with postsynaptic density-95 for detecting synaptic changes at early disease stages, were also decreased in the MCI hippocampus. Using but not MCI. the MCI case definition of Cognitive Impairment no Dementia in an oldest-old sample, synaptophysin levels in the frontal cortex were found to be increased Metabolic factors relative to individuals with dementia (mixed subtype). Oldest-old individuals with dementia Metabolic changes showed a decrease relative to the non-dementia group Peroxisome proliferator-activated receptor g-coactiva- (NCI and Cognitive Impairment no Dementia com- tor-1alpha is a co-activator of the peroxisome prolif- bined).135 These results were replicated using the CDR erator-activated receptor-g and also interacts with score to classify participants with no difference in sirtuin 3. These molecules are involved in the synaptophysin expression in the EC in the CDR = 0 regulation of fatty-acid metabolism, reactive oxygen and CDR = 0.5 groups.62 Disparity in synaptic species generation and glucose homeostasis.138 In protein expression in early disease stages is believed the hippocampus, peroxisome proliferator-activated to be linked to changes in dendritic plasticity within receptor g-coactivator-1alpha immunoreactivity regions most vulnerable to incipient AD pathology. decreases with AD progression (CDR score) and is For example, the former response is believed to negatively associated with plaque pathology and Ab reflect increased vulnerability of the temporal (Ab1À42 and Ab1À40).139 Decreased peroxisome prolif- cortex to synaptic dysfunction, and the latter to a erator-activated receptor g-coactivator-1alpha is plasticity response to compensate for neuropathologi- believed to promote AD neuropathology by Fox03a- cal damage. mediated responses. However, such changes are only Using a cDNA microarray, genes related to synaptic significant at advance AD stages (CDRX2), with function, namely synapsin splice variants I±III of the levels in MCI (CDR = 0.5) similar to controls. Promo- a-type isoform, were found to be decreased in MCI tion of peroxisome proliferator-activated receptor (CDR = 0.5) in the EC (but not in the visual cortex) g-coactivator-1alpha expression, thereby maintaining compared with NCI cases.62 No difference was found glucose homeostasis, may be a promising therapeutic between MCI and NCI in variant II of the b-type strategy for late-stage AD, but not MCI.

Molecular Psychiatry Neuropathology of MCI BCM Stephan et al 1062 Oxidative/nitrosative damage to the neurodegenerative damage observed in AD and Evidence of oxidative stress including damage to MCI. protein, lipids, as well as nucleic, mitochondrial and ribosomal acids (that is, nuclear DNA, mitochondrial Endosomal–lysosomal system dysfunction DNA, mRNA and rRNA) and its by-products (for The Rab family of GTPases is involved in intracellular example, carbonyls) are found in both MCI and AD in membrane trafficking and signaling, and has been multiple brain regions, typically those affected by AD linked to pathways associated with AD including Ab, (for example, IPL, hippocampus and frontal cortex, NFT formation, APP, b-protein precursor, amyloid-b but not the cerebellum) when compared with con- peptide and apolipoprotein E. Altered expression trols.30,33,38,123,131,132,140–154 Supplementary Table 2 levels and mutations of Rab GTPases have been gives a detailed description of oxidative/nitrosative associated with AD, and have recently been found changes observed in MCI, and for detailed reviews see to be observed as early as MCI. In MCI and AD, a the studies by Sultana et al.,155 Mangialasche et al.,156 significant increase in the expression levels of genes Butterfield et al.,157 Lovell and Markesbery,158,159 that regulate early (GTPase rab5) and late (GTPase Bonda et al.,160 Sultana and Butterfield161 and rab7) endosomes is observed in CA1 hippocampal Markesbery and Lovell.162 Such changes are believed pyramidal neurons.169 Changes in the hippocampal to occur in a stepwise manner, with a threshold of CA1 region are found to be associated with impaired pathogenesis associated with major changes in pro- neurotrophin receptor signaling linked to neuronal tein synthesis and oxidative metabolism consumption apoptosis including downregulation of genes encod- occurring during the MCI stage of disease.163 ing TrkB and TrkC in MCI and AD.169 Increase in the In AD and MCI, oxidative damage has been gene expression levels of rab4 and rab24 and implicated in cellular dysfunction, metabolism dys- decreased expression of rab3 are also observed in regulation and alteration to antioxidant defense the CA1 neurons but, only at late disease stages (AD mechanisms. An increase in reactive oxygen species vs MCI and NCI).169 Relative to NCI cases, protein and reactive nitrogen species generation may be expression levels of rab5 and rab7 are also found to be associated with impaired signaling and enzymatic increased in MCI and AD in regions vulnerable to AD functions, ion fluxes and pH regulation, mitochon- pathology including the hippocampus and frontal drial functions and gene expression, which may have cortex, but not the cerebellum or striatum. In AD, a a causal role in the onset of dementia through the significant upregulation is also observed in the basal contribution of several mechanisms associated with forebrain.170 cellular neurodegeneration (for example, upregula- tion of the enzymes that process APP to generate Ab Proteomic analysis of insoluble proteins and contributing to NFT and plaque formation), Using proteomics, B125 proteins have been found to cell-cycle alteration, neuroplasticity, immune exhibit altered detergent-insoluble levels in the response, apoptosis and neuronal cell death.163–165 temporal cortex in late-onset AD including Ab and Oxidative stress may be induced by impaired tau and those linked to dendritic spines or synapses, protein synthesis. For example, with disease progres- cytoskeletal proteins, mitochondrial enzymes, sion, ribosome dysfunction occurs, resulting in a chaperonins, ubiquitination, oxidative stress- decreased capacity for protein synthesis in the response enzymes and extracellular matrix proteins.69 ribosome, decreased rRNA and transfer RNA levels, By western blot, levels of Ab, APOE, tubulin, 14-3-3, and excessive RNA oxidation in memory-related glyceraldehyde 3-phosphate dehydrogenase, cyto- cortical areas including the superior middle temporal chrome c, glial fibrillary acidic protein and myelin gyri and the IPL.166,167 Such changes have been linked basic protein were also found to be increased in to impaired neuronal viability and function and prodromal AD (defined as no dementia and a oxidative stress.142,168 However, the presence of Cognitive Abilities Screening Instrument score < 90) oxidative damage does not always predict clinical relative to controls. In contrast, tau protein was only diagnosis or AD severity.168 This raises the question of found to be elevated at late AD stages, being indis- the extent to which a protein must be oxidized to tinguishable in control and prodromal groups. Over- impact cognition.165 all, these results have been taken to suggest that the pathological process leading to abnormal proteins is not exclusively linked to AD-related pathology, such Protease activity as Ab and tau. Proteasomes are cellular organelles present in the nucleus and in the cytoplasm the function of which is Antioxidant enzymes the clearance of unneeded or modified proteins, Individuals with MCI show alterations in the anti- necessary for maintaining the internal cellular milieu. oxidant system, designed to counteract the poten- Protease activity has been found to be impaired in the tially hazardous reactions that are initiated by brain of individuals with AD and MCI.143 Thus, this oxidative stress (for example, reactive oxygen reduction in enzymatic activity could result in an species).152 In the MCI hippocampus, superoxide accumulation of damaged proteins that can form dismutase activity, glutathione-S-transferase activity protein aggregates (aggresome) and possibly be linked and the glutathione (involved in metabolic regulation

Molecular Psychiatry Neuropathology of MCI BCM Stephan et al 1063 and essential for cellular detoxification)/oxidized (Brodmann areas 22 and 26) has been found to glutathione disulfide ratio is decreased. In contrast, correlate positively with CDR score, with only small there is a significant increase in the level of decrements in the groups with none/mild impairment superoxide dismutase activity and glucose-6-phos- (CDR = 0 or CDR = 0.5).179 Furthermore, although phate dehydrogenase activity compared with age- expression levels of the muscarinic AChRs (mAChRs) matched controls. Overall, the results suggest that subtype mRNAs among NB neurons is unchanged in antioxidant imbalance may have a role in dementia NCI, MCI and mild/moderate AD cases, increased progression. expression of the a7nAChR subunit mRNA within NB neurons has been found in mild/moderate AD (vs NCI or MCI),180 but not in the SFC.61 Neurotransmitter systems Increases in a7 may reflect a compensatory Cholinergic system and its markers neuroprotective response to maintain basal cortical Selective degeneration of CBF neurons, including a cholinergic activity in the presence of pathological loss of choline acetyltransferase (ChAT) activity changes associated with AD. The contributions of the within the NB, has been implicated in the develop- various cholinergic receptor subtypes to regulation ment of neurological disorders including AD. Loss of of cholinergic signal transmission both pre- and cortical and hippocampal ChAT activity has been post-synaptically, are not well understood; and found to correlate with AD severity and disease hence, how changes in the cholinergic receptor duration171–173 (for a review see Schaeffer and subtype may contribute to cognitive impairment and Gattaz174). However, such changes have only been disease progression is not clear. found to occur in late disease stages.171 In NCI, MCI and early/mild AD cases, ChAT activity is found to be Galanin stable in the inferior parietal, superior temporal and The signaling peptide galanin is closely associated anterior cingulate cortices,172 as well as in neurons with cholinergic function in the human brain and is within the basal forebrain175,176 and primary visual co-expressed in a subset of cholinergic neurons cortex.177 within the NB.181,182 Galanin hypertrophy within the Relative to controls and individuals with mild AD, anterior NB subfield only occurs at late AD stages.183 individuals with MCI have been found to show an It is currently unclear whether galanin contributes to elevation in ChAT activity in the hippocampal15,172 dementia progression by inhibiting cholinergic and neocortical regions.21 Activation in the SFC is signaling182 or is protective by promoting neuronal also significantly elevated compared with controls function and survival.184 Evidence that noradrenergic but not significantly different from AD.172 Upregula- neurons co-expressing galanin that project from the tion of cortical ChAT activity is believed to be linked LC survive for longer, supports the protective hypoth- to compensatory mechanisms early in the disease esis.185 Further investigation is required to clarify the process.15 The mechanism by which this is achieved contribution of galanin to disease progression. is suggested to be linked to the altered metabolic status of cholinergic neurons that project to the SFC, Neurotrophins rather than an increase in fiber projection. Indeed, a Although ChAT activity is not found to be signifi- significant loss of ChAT-like immunoreactive (ChAT-ir) cantly altered in MCI, recent observations suggest that fiber and axon varicosity in the SFC has been found in regulation of CBF neurons may be abnormal. The early AD with comparable densities in the NCI and high-affinity nerve growth factor (NGF)-specific re- MCI groups.15 A similar metabolic change is observed ceptor tyrosine kinase TrkA (and TrkB and TrkC) in the NB as reflected by an increase in the size of the density has been found to be significantly reduced in Golgi apparatus in MCI.178 Increased NB metabolic MCI and AD (mild and severe) in cholinergic neurons activity may reflect a compensatory response to of the NB compared with controls.176,186,187 In con- incipient pathology, reflecting the upregulation of trast, TrkA levels have been found to remain stable in cortical ChAT activity. Such metabolic changes may the cortex in MCI with significant declines only in be a marker of disease transition from early to late AD mild/moderate and severe AD cases.188 Furthermore, stages. Expression levels of the vesicular acetylcho- NGF levels in CBF cortical target sites including the line transporter protein do not vary significantly anterior cingulate, superior frontal, inferior parietal, across clinical groups.175 superior temporal, middle temporal regions and hippocampus also remain stable with disease pro- Nicotinic and muscarinic cholinergic receptor binding gression (that is, NCI, MCI, mild and severe AD).189 Although disruption to nAChR (nicotinic acetylcho- There is also a marked reduction in the number of p75 line receptor) binding has been observed in the mid- neurotrophin (p75NRT)-immunopositive neurons in frontal cortex in patients with AD (vs NCI), nAChR the NB in MCI and mild AD.75 In contrast, p75NRT receptor binding is unchanged in MCI (vs NCI).22 levels are stable in the cortex in NCI, MCI and mild/ Disruption to nAChR binding is believed to be a moderate/severe AD groups.188 This change is not due secondary outcome after development of AD pathol- to downregulation of p75NTR gene expression in CBF ogy. Indeed, loss of high-affinity nicotinic agonist neurons, which is found to be stable across clinical binding (epibatidine binding) in the temporal cortex groups.176 Brain-derived neurotrophic factor (both

Molecular Psychiatry Neuropathology of MCI BCM Stephan et al 1064 precursor and mature forms), necessary for CBF Other factors neuronal survival and function is significantly Concanavalin-A-associated brain proteins reduced in the parietal cortex in early AD and MCI relative to controls.190 Furthermore, although brain Relative to controls, altered levels of protein phos- NGF-like immunoreactivity in cortical and hippo- phatase-related protein Sds-22 (Sds22), glial fibrillary campal structures remains stable with disease acidic protein, dihydropyrimidase-2 and glucose- 189 regulated protein 78 are found in the MCI hippocam- progression, the level of its precursor pro-form of 196 NGF (pro-NGF), is significantly increased in the pus. Levels of the latter two proteins are decreased, inferior parietal cortex in MCI and early AD.191 In and the former increased. The only protein with addition, the extracellular collagenase matrix metal- altered levels in the inferior temporal lobe was b-synuclein, with a decrease in levels relative to loproteinase-9 degrades mature NGF. During progres- 196 sion of AD, matrix metalloproteinase-9 (and controls. These proteins support different func- pro-matrix metalloproteinase-9) is unregulated in tions known to be disrupted in AD including the cortex.192 Matrix metalloproteinase-9 activity also metabolism, cellular trafficking, neurotransmission correlates with cognition (Global Cognitive Scores and signal transduction. and the MMSE) and Braak stage. These findings indicate an alteration in the NGF cascade early in Tissue transglutaminase disease progression, at the MCI stage. Tissue transglutaminase is involved in catalyzing the Overall, the results suggest that there are differ- formation of protein cross-links and has been asso- ences in expression changes across different cortical ciated with apoptosis and neurodegeneration.197 projection sites of cholinergic NB neurons depending Tissue transglutaminase immunoreactivity and enzy- on the cholinergic marker and disease severity. Over- matic activity in the frontal gray matter is reported to lap in the neurotrophic molecular pathology in MCI be intermediate in individuals with MCI and highest and AD suggests that such changes occur early in the in individuals with AD. However, cross-group differ- pathogenesis of disease and support MCI as prodro- ences are not significant. Furthermore, isopeptide, the mal AD. The findings support a role for altered catalytic product of tissue transglutaminase, although neurotrophin responsiveness rather than frank higher in AD (vs NCI and MCI), is not statistically 29 neurodegeneration of cholinergic neurons as a possi- different across clinical groups. ble early biomarker for AD.193 The biological consequences of such changes are still unclear. Metal homeostasis Further refinement of the neuropathological changes Zinc (Zn) is an essential co-factor for many proteins in the NB system and their functional significance including the zinc-finger containing transcription remains outstanding. factors and is therefore involved in many cellular processes through regulation of gene expression. Neuropeptides Perturbations in Zn homeostasis have been impli- cated in the pathogenesis of AD and MCI. In both AD Both NCI and MCI cases (CDR = 0.5) are not found to and MCI, regional-specific alterations in the Zn differ on cortical levels of somatostatin-like immu- transporter protein-1 (ZnT-1) have been observed.198 noreactivity or corticotptrophin-releasing factor.173,194 In AD relative to controls, reduced brain ZnT-1 In contrast, significant differences between dementia expression is found in the amygdala, hippocampal/ and NCI cases are found in cortical levels of parahippocampal gyrus and IPL, and an increase in corticotptrophin-releasing factor (reduced at mild expression in the superior middle temporal gyri.198 In and advanced stages; CDR = 1–5) and somatostatin- both early- and late-stage AD, an increase in ZnT-4 like immunoreactivity (reduced only at advanced and ZnT-6 is seen in the hippocampal/parahippo- disease stages; CDRX4).194 campal gyrus.24 In aMCI when compared with age- matched controls, an increase in ZnT-1 levels is Glutamatergic system observed in the superior middle temporal gyri Paradoxical alterations in non-cholinergic synaptic (similar to the results from early-AD cases) but, in integrity have also been observed in MCI relative to contrast to the results from AD groups, MCI cases control and AD groups. Changes include an upregula- show depletion of ZnT-1 in the hippocampal/para- tion of glutamatergic pre-synaptic boutons in the mid- hippocampal gyrus. Levels of ZnT-4 and ZnT-6 frontal gyrus195 and increased DRB levels in the although increased in MCI do not differ from controls frontal cortex.133 These changes are followed by in any regions tested including the hippocampal and successive depletion with increased severity of AD hippocampal/parahippocampal gyrus, superior mid- (mild to severe). Such changes are believed to reflect a dle temporal gyri and the cerebellum.24 Alterations to counter response to pre-existing synaptotoxicity to Zn-transported proteins seem to occur early in the improve transmission probability, thereby allowing transition from MCI to AD in regions most vulnerable maintenance of cognitive capacity early in disease to AD pathology. The exact mechanism(s) by which progression. Alternatively, it has been suggested Zn changes affect cognition remains unknown, but that they may be due to an uncoordinated atypical given its role as a cofactor, multiple processes may be response in early disease stages.195 affected and such changes are believed to contribute

Molecular Psychiatry Neuropathology of MCI BCM Stephan et al 1065 to plaque formation (that is, Ab aggregation) and therapeutic interventions that delay dementia neuronal degeneration.198 progression.202

Cortical sortilin levels Inflammation Sortilin is a sorting receptor involved in protein Neuroinflammatory processes characterized by the transport from the endoplasmic reticulum and the upregulation of complement proteins involved in Golgi apparatus to other cellular compartments. It is innate immunity including C1q, C3c, C3d, C4c, C4d, also involved in protein clearance and acts as a iC3b (early-stage complement activation), C9 (late- receptor for neurotensin. In the superior frontal and stage complement activation) are found to be asso- temporal cortical areas, sortilin levels have been ciated with Ab plaques containing the fibrillar Ab peptide.206 Complement activation in plaques in the found to remain stable with disease progression 63 207 (NCI-MCI-AD). Furthermore, in MCI, levels do not inferior temporal gyrus and neocortical areas is differ by domain subtype (aMCI vs nMCI). However, found to increase with disease progression, with the frontal cortical levels do positively correlate with age MCI and NCI groups generally showing a similar and temporal levels positively correlate with Braak level of activation. In addition, the membrane attack staging and NIA-Reagan diagnosis.199 protein, Cb5-9, is rarely seen before the onset of clinical dementia.207 Therefore, upregulation of complement proteins does not appear to be involved CN/NFAT in the cognitive changes seen in early disease The multicomponent nuclear factor of activated T cell stages.63,207 (NFAT) transcription complex has roles in the The expression of cyclooxygenase-2, a marker of regulation of gene transcription and is dephosphory- 2 þ inflammation, increases with dementia progression lated by the Ca - and calmodulin-dependent phos- (CDR score) in hippocampal neurons in areas CA1, phatase calcineurin (CN). CN has roles in multiple CA2 and CA3.208 However, the expression in MCI systems including regulation of cyclic AMP-depen- cases (CDR = 0.5) does not differ from that in controls dent kinases and dephosphorylation of histones. across each hippocampal subregion. Rather, only at Isoform and brain regional-specific changes of NFAT X 200 later disease stages (CDR 1) is expression signifi- have been observed as a function of disease state. In cantly increased in areas CA2 and CA3. In area CA1, hippocampal nuclear fractions, NFAT1 is increased in significant elevation in expression is only observed at MCI relative to controls, with NFAT3 showing a slight severe, late stages of disease (CDR = 5). Cyclooxygen- decrease. In AD, NFAT1 is slightly decreased and ase-2 maybe a predictor of neurodegeneration, with NFAT3 markedly reduced relative to controls. These changes in expression correlating with moderate and patterns were not observed in cerebellar tissue. The severe NP and NFT pathology in the CA3 subdivision, pattern coincides with the nuclear accumulation of and to a lesser extent in area CA2. However, not all CN, suggesting translocation of NFATs from cytosol to inflammatory markers are found to be associated with the nucleus with disease progression. Furthermore, disease progression. Interleukin-6 and transforming accumulation of CN (subunit Aa) and NFAT3 corre- b 1À42 growth factor- 1 are found to be elevated in dementia, late with soluble Ab levels in the hippocampus. but only at the late severe stages of disease Therefore, such changes in CN/NFAT signaling may 209 1À42 (CDR = 5). Interleukin-6 was also increased in cases be driven by the Ab elevations and contribute to with high levels of NFT pathology. Microglia activa- the progression of dementia. Alternatively, both 1À42 tion measured using human leukocyte antigen-DR changes in Ab and CN/NFAT may be responding immunohistochemistry increases with AD progres- to other changes as yet unknown. sion (CDR score, NP and to a lesser extent NFTs) in both the gray and white matter of the EC, with Sirtuin 1 significant changes at moderate (CDR = 2) and severe Sirtuin 1 (SIRT1) is an NAD-dependent protein stages (CDR = 5).210 In the hippocampal area CA1, a deacetylase involved in many cellular processes, from significant change is observed earlier (CDR = 1), and regulating apoptosis201 to regulation of cellular meta- in the dentate gyrus, it is observed as early as MCI bolism and inflammatory and stress responses.202,203 (CDR = 0.5). Independent segments of the inflamma- SIRT1 directly affects processing of the amyloid tory cascade seem to be activated at different stages of precursor protein by upregulating a-secretase disease. Therefore, anti-inflammatory drugs need to ADAM10, leading to reduced production of be adapted depending on the stage of disease at which Ab.202,204 When comparing NCI, MCI and AD groups they are being targeted. levels of SIRT1 are found to be similar in the parietal cortex.205 Only at advanced AD symptomatic stages Heme degradation and in the presence of a high degree of Ab and tau Biliverdin reductase A (BVR-A) is part of the catabolic accumulation is a significant SIRT1 decrease pathway of hemoglobin which, together with heme observed.205 As such, a decrease in SIRT1 occurs at oxygenase, leads to the formation of bilirubin and late disease stages and therefore does not appear to be carbon monoxide. The action of heme oxygenase a target for MCI. However, as a protective factor, produce a ferrous iron, CO and biliverdin which SIRT1 is the focus of investigations that may lead to through the action of the BVR-A is converted into

Molecular Psychiatry Neuropathology of MCI BCM Stephan et al 1066 bilurubin. Total BVR-A protein levels are significantly scores and PE content varied across brain regions increased and activity (phosphorylation on Ser/Thr/ (except in the cerebellum). In addition, across the Tyr residues) is significantly decreased in the hippo- NCI-MCI-AD continuum, a significant decrease in campus (but not in the cerebellum) in aMCI and AD myelin basic protein immunoreactivity has been compared with controls.211 BVR-A in the hippocam- found.216 Myelin basic protein has roles in the pus (but not in the cerebellum) is also found to formation and maintenance of the axonal myelin undergo post-translational oxidative and nitrosative membrane and may also be involved in signaling path- modifications in AD and aMCI groups, including a ways in neural cells. However, although the level of significant reduction in protein carbonyl-derivatives change in the AD group is significantly different from of BVR-A in aMCI and AD compared with controls.212 that in the NCI group, change in the MCI group is not In both the AD and MCI groups in the hippocampus, distinguishable from either the AD or NCI groups. there is also a significant increase in 4-hydroxynone- nals (4-NE) levels relative to controls. In contrast, Sulfatides and ceramides BVR-bound 4-hydroxynonenals is not found to be Sulfatides are a class of sulfated galactosylceramides modified across NCI-MCI-AD groups in either the (STs) involved in several physiological functions (cell hippocampus or cerebellum.212 The increase in growth, signaling, neuronal plasticity). STs are BVR-A expression in AD and MCI could potentially synthesized primarily in the oligodendrocytes and represent an adaptive mechanism to compensate for incorporated into the myelin sheath of axons and the reduced enzymatic efficiency as demonstrated by a therefore present in both the white and the gray significant reduction in the phosphorylation of serine, matter. Altered levels of STs and ceramides (derived threonine and tyrosine residues and marked reduction from the degradation of STs and involved in the in BVR-A reductase activity. These adaptive mechan- modulation of inflammatory responses) in human isms could be induced by reducing insulin signaling, brain tissues may be invoked in the pathogenesis of which is associated with reduced phosphorylation of neurodegenerative disorders. A decrease in STs in BVR-A. A compensatory increase in heme oxygenase both the white and the gray matter was observed in activity can lead to an increased CO production with very mild dementia (CDR = 1), but an increase in CDR concommitant increases in blood flow and nutrient score was not associated with further changes in ST supply. Additionally, an increased concentration of levels.217 The NCI and MCI groups did not differ. The biliverdin may stimulate the expression of BVR-A. content of ceramides increased in the white matter in subjects with mild dementia (CDR = 1), but surpris- Autophagy ingly, a progressively decline was observed with an Autophagy, the pathway involved in intracellular increase in neuropathology (CDR = 2 and CDR = 3). degradation of long-lived proteins and organelles, has Ceramide content in the gray matter was not been implicated in the pathogenesis of AD. Dysfunc- associated with CDR scores. tion in the autophagy system is believed to result in 213 the accumulation of misfolded proteins. In MCI, Gene expression abnormalities Beclin 1, a protein that has a central role in autophagy and apoptosis, has been found to be decreased in the Using microarray analysis, differences in the pattern affected brain regions suggesting that reduced autop- of aberrantly expressed genes, namely downregula- hagy promotes AD progression.214 Therefore, enhan- tion, as a function of dementia severity (CDR score) cing autophagy by increasing Beclin 1 levels may be has been found.218 In MCI, the greatest transcriptional beneficial to preventing disease progression. abnormalities are observed in the temporal and parietal cortices, with the hippocampus and frontal Alterations in membrane lipid composition cortex less affected until later disease stages. As Modifications of the normal lipid and protein content disease progresses (CDRX2), more brain regions of cellular membranes and the myelin sheath may be become involved (that is, visual cortex and posterior linked to the pathogenesis of neurodegenerative cingulated gyrus) and an increase in the number of disorders. A decrease in ethanolamine plasmalogen upregulated genes is observed. The regional pattern of (PE) may be involved in AD-specific related neuro- alterations in gene expression is also found to change degeneration as PE contributes to the glyceropho- only slightly depending on the method used to spholipid fraction of myelin and plasma membrane. classify disease severity (that is, CDR score, Braak In NCI, PE content was higher in the white than in the stage or NP density). This study suggests that gene gray matter and they also differed in the type of therapies for AD targeted at early disease stages fatty-acid species incorporated in the PE molecule.215 would need to be adjusted for later disease stages in No difference was found in the distribution of PE terms of areas of vulnerability. However, the specific between brain areas. The PE content of the white targets were not outlined. matter in subjects with MCI (CDR = 0.5) decreased in all the measured regions compared with controls, but Neuropathology and cognitive profile an increase in CDR score was not associated with further decline in PE levels. On the contrary, the Mild cognitive impairment is a heterogeneous condi- decrease in the gray matter was proportional to CDR tion and each neuropsychological subtype (that is,

Molecular Psychiatry Neuropathology of MCI BCM Stephan et al 1067

Other Factors Cell Cycle, Signalling Cascades Synaptic Factors and Gene Expression Total Epo-R MCI AD Ab monomers, dimers and oligomers Glial Epo-R MCItc cyclins B, D and G1 MCI Glu pre-synaptic boutons MCI AD Pro-NGF NGF CDK2 MCI AD SYT, PSD95, NR2a, LRP1 or BDNF CDK5 MCI AD SYN MCI AD Trk-a, Trk-b, Trk-c CHAT MCI AD ZnT-1 PIN1 MCI DRB MCItc MCIfc ZnT-4, ZnT-6 PCNA MCI AD a7 nAChR expression Caspases 3, 6 miRNA (miR-107) MCI AD Other nAChRs MCI AD HSP27, HSP70 NFAT1 MCI AD A/SYN MCI AD BCL2, BAX NFAT3 MCI AD SOD CN in nucleus MCI AD SIRT1 MCI AD Beclin 1 Global Factors Metabolic activity Oxidative stress Impaired protein synthesis Proteasome activity Dendritic spine density LC pathology Tau pathology NB Vascular disease Axons and Cytoskeleton Hippocampal volume MBP Neuronal loss Phospho-Tau Synapse loss 3Rtau/4Rtau Endosomal lysosomal dysfunction PE (wm) Inflammation Sulfatides Autophagy

MCI Early to severe AD

Oxidative stress, Vascular disease, Neuronal loss, Tau Factors neuropathology in NB and LC, Aβ neuropathology in EC, Impaired Galanin, ZnT4, ZnT6, Cox2, IL6, increasing with protein synthesis, α7 nAChR subunit expression, Heat Shock TGFβ1, Ceramides in white disease Proteins HSP27 and HSP70, EpoR expression, PCNA, Cyclins D, matter, complement cascade, progression B, G, CDK2 and CDK5, Csp3 and Csp6, Pin1, Aβ monomers, rab3, rab4, rab24 dimers and oligomers, BCL2, BAX, NFAT1, Rab5, Rab7, BVR-A

Glutamatergic pre-synaptic Factors with Glutamatergic pre-synaptic boutons increase boutons decrease variable DRB in frontal cortex increase while DRB in temporal cortex DRB in frontal and temporal patterns of decrease cortex decrease change ChAT activity increase ChAT activity decrease

Factors Hippocampal volume, Dendritic spine density, Synapses, α-synuclein, SYN, SIRT1, decreasing with Proteosome activity, Autophagy, SOD, GST, GSH, Trka, Trkb, Trkc, ChAT reactive neuronal fibres, disease p75NRT in NB, MBP, PE in white matter, Sulfatides, PSD-95, NR2α, nACh receptor binding progression LRP1, BDNF, 3Rtau/4Rtau, PGC1α, ProNGF, NFAT3, Beclin1

Figure 1 Neuropathological changes associated with MCI and AD. Factors described in the text are summarised in panel a with reference to neuronal processes and location. The timings of changes to various factors described in the text with reference to disease stage are presented in panel b. aMCI vs nMCI) may possibly reflect a different performance. Furthermore, poor early-life linguistic neuropathological profile. Indeed, a single case ability (measured using idea density) is found to be study219 of a patient with executive MCI identified a associated with an increased likelihood of a neuro- pattern of pathology reflecting an early stage of the pathological diagnosis of AD, but not vascular-related frontal variant of AD that is distinct to the profile pathologies including infarcts (lacunar or large brain typically observed in aMCI including pathology in infarcts) and moderate-to-severe atherosclerosis.220,221 the temporal, EC, parietal and occipital cortices. Jicha When subtypes of MCI have been assessed, nMCI et al.73 distinguish a pre-LB dementia state from a pre- cases less often have a pathological diagnosis of AD or AD state based on cognitive and non-cognitive mixed AD-vascular pathologies compared with aMCI (psychiatric symptoms) symptoms again with varying cases. Only macroscopic infarcts without AD are neuropathological profiles. Riley et al.4 reported that more frequently observed in nMCI cases. In terms of the presence of memory impairment (vs non-memory pathological diagnosis, results have been found to be impairment) is associated with more severe AD similar for both aMCI and nMCI subtypes: pure AD pathology (NFTs indexed by Braak staging) and was the most frequent diagnosis and > 75% of increased severity of atherosclerosis than in indivi- individuals in both groups were found to have an duals with intact memory and impaired non-memory intermediate likelihood of AD (NIA-Reagan criteria).7

Molecular Psychiatry Neuropathology of MCI BCM Stephan et al 1068 Critique of studies and approach MCI. The results suggest that many different pro- cesses underlie cognitive decline and dementia and Large variability exists in the neuropathological that MCI cannot be understood within a single profile of the small number of cases that have come framework. Multiple neuronal processes, including to autopsy with a final diagnosis of MCI as shown in synaptic processes, neuronal homeostasis, intracellu- Supplementary Table 1. Cross-article comparison is lar signaling cascades and gene expression, have been problematic because of study design and methodolo- investigated and suggested to be affected at very early gical differences. Differences include: (1) sampling stages of cognitive decline as shown in Figure 1. To methods (population vs clinic vs volunteer samples) date there is no definitive profile that characterizes (2) MCI case selection criteria and nature of impair- the state of MCI. This will require representative brain ment (for example, aMCI, nMCI, CDR = 0.5) (3) tissue banks that use standardized MCI criteria, operationalization of criteria (with vs without clinical neuropathological protocols including staining and judgment) (4) comparison groups (for example, scoring techniques. This will enhance our under- dementia status can vary with respect to stage, standing of the pathological process involved in severity, likelihood and type) (5) exclusion criteria dementia, the risk factors for dementia progression (see Supplementary Table 1) (6) sample age (for and whether MCI can be distinguished from early AD. example, pathology–cognition associations change With population ageing, this is important, especially in the oldest-old population) (7) tissue handling to inform the development of effective early interven- procedures (differences in postmortem interval) (8) tion strategies. neuropathological methods (stereological vs semi- quantitative approaches vs ultra-structural studies; antibody staining differences) and regions/mechan- Conflict of interest isms of study interest, as well as (9) the interval between diagnosis and death and therefore the The authors declare no conflict of interest. possibility of MCI case transitions to dementia at the time of death. Agreement in MCI criteria for defining Acknowledgments groups and clearer reporting of sample demographics, operationalization of MCI criteria and handling Duncan Harris and David J Llewellyn are supported methods would assist the integration of findings. To by the NIHR Peninsula Collaboration for Leadership date, there is not enough consistency in any of these in Applied Health Research and Care (PenCLAHRC). parameters to be able to combine studies for analysis. BCMS is funded by the Joint European Post-Doctoral Programme: The European Research Area in Ageing (ERA-AGE) Network FLARE Programme. Summary The opportunity to examine the brains of individuals References who have been in the intermediate pre-clinical stage of cognitive decline when they die is relatively challen- 1 Matthews FE, Stephan BC, Bond J, McKeith I, Brayne C. Operationalization of mild cognitive impairment: a graphical ging. The pathological profile of MCI is heterogeneous approach. PLoS Med 2007; 4: 1615–1619. and many factors seem to be on a continuum rather 2 Petersen RC, Doody R, Kurz A, Mohs RC, Morris JC, Rabins PV than neuropathologically specific as anticipated.222 et al. Current concepts in mild cognitive impairment. Arch Most associations are confined to single studies and a Neurol 2001; 58: 1985–1992. few individuals. This may in part be due to the 3 Morris JC, Storandt M, Miller JP, McKeel DW, Price JL, Rubin EH et al. Mild cognitive impairment represents early-stage Alzheimer difficulty in operationalization of MCI at the end of disease. Arch Neurol 2001; 58: 397–405. life. Furthermore, there is the lack of a clear boundary 4 Riley KP, Snowdon DA, Markesbery WR. Alzheimer’s neurofi- between cognitive stages as dementia onset is gradual brillary pathology and the spectrum of cognitive function: by definition. A major obstacle in identifying the findings from the Nun Study. Ann Neurol 2002; 51: 567–577. 5 Hof PR, Bussiere T, Gold G, Kovari E, Giannakopoulos P, Bouras C pathological substrate of any disease is distinguishing et al. Stereologic evidence for persistence of viable neurons in changes that have a causative role from those that are layer II of the entorhinal cortex and the CA1 field in Alzheimer secondary to disease or are compensatory. Pathological disease. J Neuropathol Exp Neurol 2003; 62: 55–67. changes identified in MCI cases may suggest that this 6 Giannakopoulos P, Kovari E, Herrmann FR, Hof PR, Bouras C. state is already dementia. The concept of pre-MCI has Interhemispheric distribution of Alzheimer disease and vascular pathology in brain aging. Stroke 2009; 40: 983–986. instead been proposed as the earliest marker of 7 Schneider JA, Arvanitakis Z, Leurgans SE, Bennett DA. The cognitive impairment.26,223 However, the majority of neuropathology of probable Alzheimer disease and mild cogni- pre-MCI cases have been found to have neuropatholo- tive impairment. Ann Neurol 2009; 66: 200–208. gically confirmed AD, similar to findings reported in 8 Bennett DA, Schneider JA, Bienias JL, Evans DA, Wilson RS. 26 Mild cognitive impairment is related to Alzheimer disease aMCI and revised-MCI groups. pathology and cerebral infarctions. Neurology 2005; 64: 834–841. 9 Crystal H, Dickson D, Fuld P, Masur D, Scott R, Mehler M et al. Clinico-pathologic studies in dementia: nondemented subjects Outstanding questions and conclusions with pathologically confirmed Alzheimer’s disease. Neurology 1988; 38: 1682–1687. This review highlights the complexity inherent in the 10 Galvin JE, Powlishta KK, Wilkins K, McKeel Jr DW, Xiong C, study of neuropathological changes associated with Grant E et al. Predictors of preclinical Alzheimer disease

Molecular Psychiatry Neuropathology of MCI BCM Stephan et al 1069 and dementia: a clinicopathologic study. Arch Neurol 2005; 62: 31 von Gunten A, Kovari E, Rivara CB, Bouras C, Hof PR, 758–765. Giannakopoulos P. Stereologic analysis of hippocampal 11 Giannakopoulos P, Hof PR, Kovari E, Vallet PG, Herrmann FR, Alzheimer’s disease pathology in the oldest-old: evidence for Bouras C. Distinct patterns of neuronal loss and Alzheimer’s sparing of the entorhinal cortex and CA1 field. Exp Neurol 2005; disease lesion distribution in elderly individuals older than 90 193: 198–206. years. J Neuropathol Exp Neurol 1996; 55: 1210–1220. 32 von Gunten A, Kovari E, Bussiere T, Rivara CB, Gold G, Bouras C 12 Green MS, Kaye JA, Ball MJ. The Oregon brain aging study: et al. Cognitive impact of neuronal pathology in the entorhinal neuropathology accompanying healthy aging in the oldest old. cortex and CA1 field in Alzheimer’s disease. Neurobiol Aging Neurology 2000; 54: 105–113. 2006; 27: 270–277. 13 Haroutunian V, Schnaider-Beeri M, Schmeidler J, Wysocki M, 33 Williams TI, Lynn BC, Markesbery WR, Lovell MA. Increased Purohit DP, Perl DP et al. Role of the neuropathology of levels of 4-hydroxynonenal and acrolein, neurotoxic markers Alzheimer disease in dementia in the oldest-old. Arch Neurol of lipid peroxidation, in the brain in mild cognitive impairment 2008; 65: 1211–1217. and early Alzheimer’s disease. Neurobiol Aging 2006; 27: 14 Hof PR, Bierer LM, Perl DP, Delacourte A, Buee L, Bouras C et al. 1094–1099. Evidence for early vulnerability of the medial and inferior aspects 34 Yang Y, Mufson EJ, Herrup K. Neuronal cell death is preceded by of the temporal lobe in an 82-year-old patient with preclinical cell cycle events at all stages of Alzheimer’s disease. signs of dementia. Regional and laminar distribution of neurofi- J Neurosci 2003; 23: 2557–2563. brillary tangles and senile plaques. Arch Neurol 1992; 49: 35 Wang S, Simon BP, Bennett DA, Schneider JA, Malter JS, Wang 946–953. DS. The significance of Pin1 in the development of Alzheimer’s 15 Ikonomovic MD, Mufson EJ, Wuu J, Cochran EJ, Bennett DA, disease. J Alzheimers Dis 2007; 11: 13–23. DeKosky ST. Cholinergic plasticity in hippocampus of 36 Gold G, Kovari E, Herrmann FR, Canuto A, Hof PR, Michel JP individuals with mild cognitive impairment: correlation with et al. Cognitive consequences of thalamic, basal ganglia, and deep Alzheimer’s neuropathology. J Alzheimers Dis 2003; 5: 39–48. white matter lacunes in brain aging and dementia. Stroke 2005; 16 McKee AC, Au R, Cabral HJ, Kowall NW, Seshadri S, Kubilus CA 36: 1184–1188. et al. Visual association pathology in preclinical Alzheimer 37 Assaraf MI, Diaz Z, Liberman A, Miller Jr WH, Arvanitakis Z, Li Y disease. J Neuropathol Exp Neurol 2006; 65: 621–630. et al. Brain erythropoietin receptor expression in Alzheimer 17 Mitchell TW, Mufson EJ, Schneider JA, Cochran EJ, Nissanov J, disease and mild cognitive impairment. J Neuropathol Exp Han LY et al. Parahippocampal tau pathology in healthy aging, Neurol 2007; 66: 389–398. mild cognitive impairment, and early Alzheimer’s disease. Ann 38 Schipper HM, Bennett DA, Liberman A, Bienias JL, Schneider JA, Neurol 2002; 51: 182–189. Kelly J et al. Glial heme oxygenase-1 expression in Alzheimer 18 Mufson EJ, Chen EY, Cochran EJ, Beckett LA, Bennett DA, disease and mild cognitive impairment. Neurobiol Aging 2006; Kordower JH. Entorhinal cortex beta-amyloid load in indivi- 27: 252–261. duals with mild cognitive impairment. Exp Neurol 1999; 158: 39 Mesulam M, Shaw P, Mash D, Weintraub S. Cholinergic nucleus 469–490. basalis tauopathy emerges early in the aging-MCI-AD continuum. 19 Petersen RC, Parisi JE, Dickson DW, Johnson KA, Knopman DS, Ann Neurol 2004; 55: 815–828. Boeve BF et al. Neuropathologic features of amnestic mild 40 Troncoso JC, Martin LJ, Dal Forno G, Kawas CH. Neuropathology cognitive impairment. Arch Neurol 2006; 63: 665–672. in controls and demented subjects from the Baltimore Long- 20 Price JL, Morris JC. Tangles and plaques in nondemented itudinal Study of Aging. Neurobiol Aging 1996; 17: 365–371. aging and ‘preclinical’ Alzheimer’s disease. Ann Neurol 1999; 41 Caselli RJ, Walker D, Sue L, Sabbagh M, Beach T. Amyloid load in 45: 358–368. nondemented brains correlates with APOE e4. Neurosci Lett 21 Prohovnik I, Perl DP, Davis KL, Libow L, Lesser G, Haroutunian V. 2010; 473: 168–171. Dissociation of neuropathology from severity of dementia in late- 42 Kok E, Haikonen S, Luoto T, Huhtala H, Goebeler S, Haapasalo H onset Alzheimer disease. Neurology 2006; 66: 49–55. et al. Apolipoprotein E-dependent accumulation of Alzheimer 22 Sabbagh MN, Shah F, Reid RT, Sue L, Connor DJ, Peterson LK disease-related lesions begins in middle age. Ann Neurol 2009; et al. Pathologic and nicotinic receptor binding differences 65: 650–657. between mild cognitive impairment, Alzheimer disease, and 43 Jack Jr CR, Dickson DW, Parisi JE, Xu YC, Cha RH, O’Brien PC normal aging. Arch Neurol 2006; 63: 1771–1776. et al. Antemortem MRI findings correlate with hippocampal 23 Saito Y, Murayama S. Neuropathology of mild cognitive impair- neuropathology in typical aging and dementia. Neurology 2002; ment. Neuropathology 2007; 27: 578–584. 58: 750–757. 24 Smith JL, Xiong S, Markesbery WR, Lovell MA. Altered expres- 44 Bussiere T, Friend PD, Sadeghi N, Wicinski B, Lin GI, Bouras C sion of zinc transporters-4 and -6 in mild cognitive impairment, et al. Stereologic assessment of the total cortical volume early and late Alzheimer’s disease brain. Neuroscience 2006; 140: occupied by amyloid deposits and its relationship with cognitive 879–888. status in aging and Alzheimer’s disease. Neuroscience 2002; 112: 25 Sonnen JA, Larson EB, Crane PK, Haneuse S, Li G, 75–91. Schellenberg GD et al. Pathological correlates of dementia in a 45 Morris JC, Storandt M, McKeel Jr DW, Rubin EH, Price JL, longitudinal, population-based sample of aging. Ann Neurol Grant EA et al. Cerebral amyloid deposition and diffuse plaques 2007; 62: 406–413. in ‘normal’ aging: Evidence for presymptomatic and very mild 26 Storandt M, Grant EA, Miller JP, Morris JC. Longitudinal course Alzheimer’s disease. Neurology 1996; 46: 707–719. and neuropathologic outcomes in original vs revised MCI and in 46 Morris JC, McKeel Jr DW, Storandt M, Rubin EH, Price JL, pre-MCI. Neurology 2006; 67: 467–473. Grant EA et al. Very mild Alzheimer’s disease: informant-based 27 Sultana R, Butterfield DA. Regional expression of key cell cycle clinical, psychometric, and pathologic distinction from normal proteins in brain from subjects with amnestic mild cognitive aging. Neurology 1991; 41: 469–478. impairment. Neurochem Res 2007; 32: 655–662. 47 Rubin EH, Morris JC, Grant EA, Vendegna T. Very mild senile 28 Wang DS, Bennett DA, Mufson E, Cochran E, Dickson DW. dementia of the Alzheimer type. I. Clinical assessment. Arch Decreases in soluble alpha-synuclein in frontal cortex correlate Neurol 1989; 46: 379–382. with cognitive decline in the elderly. Neurosci Lett 2004; 359: 48 Giannakopoulos P, von Gunten A, Kovari E, Gold G, 104–108. Herrmann FR, Hof PR et al. Stereological analysis of neuropil 29 Wang DS, Uchikado H, Bennett DA, Schneider JA, Mufson EJ, threads in the hippocampal formation: relationships Wu J et al. Cognitive performance correlates with cortical iso- with Alzheimer’s disease neuronal pathology and cognition. peptide immunoreactivity as well as Alzheimer type pathology. Neuropathol Appl Neurobiol 2007; 33: 334–343. J Alzheimers Dis 2008; 13: 53–66. 49 Fernandez-Vizarra P, Fernandez AP, Castro-Blanco S, Serrano J, 30 Wang J, Markesbery WR, Lovell MA. Increased oxidative damage Bentura ML, Martinez-Murillo R et al. Intra- and extracellular in nuclear and mitochondrial DNA in mild cognitive impairment. Abeta and PHF in clinically evaluated cases of Alzheimer’s J Neurochem 2006; 96: 825–832. disease. Histol Histopathol 2004; 19: 823–844.

Molecular Psychiatry Neuropathology of MCI BCM Stephan et al 1070 50 Sinka L, Kovari E, Gold G, Hof PR, Herrmann FR, Bouras C et al. 69 Woltjer RL, Cimino PJ, Boutte AM, Schantz AM, Montine KS, Small vascular and Alzheimer disease-related pathologic Larson EB et al. Proteomic determination of widespread determinants of dementia in the oldest-old. J Neuropathol Exp detergent-insolubility including Abeta but not tau early in Neurol 2010; 69: 1247–1255. the pathogenesis of Alzheimer’s disease. FASEB J 2005; 19: 51 McKeel Jr DW, Ball MJ, Price JL, Smith DS, Miller JP, Berg L et al. 1923–1925. Interlaboratory histopathologic assessment of Alzheimer 70 Jicha GA, Bowser R, Kazam IG, Davies P. Alz-50 and MC-1, a new neuropathology: different methodologies yield comparable monoclonal antibody raised to paired helical filaments, recognize diagnostic results. Alzheimer Dis Assoc Disord 1993; 7: 136–151. conformational epitopes on recombinant tau. J Neurosci Res 52 Berg L, McKeel Jr DW, Miller JP, Baty J, Morris JC. Neuropatholo- 1997; 48: 128–132. gical indexes of Alzheimer’s disease in demented and nondemented 71 Jeganathan S, Hascher A, Chinnathambi S, Biernat J, persons aged 80 years and older. Arch Neurol 1993; 50: 349–358. Mandelkow EM, Mandelkow E. Proline-directed pseudo-phos- 53 Bouras C, Hof PR, Giannakopoulos P, Michel JP, Morrison JH. phorylation at AT8 and PHF1 epitopes induces a compaction of Regional distribution of neurofibrillary tangles and senile the paperclip folding of tau and generates a pathological (MC-1) plaques in the cerebral cortex of elderly patients: a quantitative conformation. J Biol Chem 2008; 283: 32066–32076. evaluation of a one-year autopsy population from a geriatric 72 Luna-Munoz J, Garcia-Sierra F, Falcon V, Menendez I, hospital. Cereb Cortex 1994; 4: 138–150. Chavez-Macias L, Mena R. Regional conformational change 54 Bussiere T, Gold G, Kovari E, Giannakopoulos P, Bouras C, Perl involving phosphorylation of tau protein at the Thr231, precedes DP et al. Stereologic analysis of neurofibrillary tangle formation the structural change detected by Alz-50 antibody in Alzheimer’s in prefrontal cortex area 9 in aging and Alzheimer’s disease. disease. J Alzheimers Dis 2005; 8: 29–41. Neuroscience 2003; 117: 577–592. 73 Jicha GA, Schmitt FA, Abner E, Nelson PT, Cooper GE, 55 Haroutunian V, Purohit DP, Perl DP, Marin D, Khan K, Lantz M et Smith CD et al. Prodromal clinical manifestations of neuropatho- al. Neurofibrillary tangles in nondemented elderly subjects and logically confirmed Lewy body disease. Neurobiol Aging 2010; mild Alzheimer disease. Arch Neurol 1999; 56: 713–718. 31: 1805–1813. 56 Leuba G, Saini K, Zimmermann V, Giannakopoulos P, Bouras C. 74 Guillozet AL, Weintraub S, Mash DC, Mesulam MM. Neuro- Mild amyloid pathology in the primary visual system of fibrillary tangles, amyloid, and memory in aging and mild nonagenarians and centenarians. Dement Geriatr Cogn Disord cognitive impairment. Arch Neurol 2003; 60: 729–736. 2001; 12: 146–152. 75 Mufson EJ, Ma SY, Dills J, Cochran EJ, Leurgans S, Wuu J et al. 57 Tremblay C, Pilote M, Phivilay A, Emond V, Bennett DA, Calon F. Loss of basal forebrain P75(NTR) immunoreactivity in subjects Biochemical characterization of Abeta and tau pathologies in with mild cognitive impairment and Alzheimer’s disease. mild cognitive impairment and Alzheimer’s disease. J Alzheimers J Comp Neurol 2002; 443: 136–153. Dis 2007; 12: 377–390. 76 Haroutunian V, Perl DP, Purohit DP, Marin D, Khan K, Lantz M 58 Ghoshal N, Garcia-Sierra F, Wuu J, Leurgans S, Bennett DA, Berry et al. Regional distribution of neuritic plaques in the nondemen- RW et al. Tau conformational changes correspond to impairments ted elderly and subjects with very mild Alzheimer disease. Arch of episodic memory in mild cognitive impairment and Neurol 1998; 55: 1185–1191. Alzheimer’s disease. Exp Neurol 2002; 177: 475–493. 77 Naslund J, Haroutunian V, Mohs R, Davis KL, Davies P, Greengard 59 Bussiere T, Giannakopoulos P, Bouras C, Perl DP, Morrison JH, P et al. Correlation between elevated levels of amyloid beta- Hof PR. Progressive degeneration of nonphosphorylated neuro- peptide in the brain and cognitive decline. JAMA 2000; 283: filament protein-enriched pyramidal neurons predicts cognitive 1571–1577. impairment in Alzheimer’s disease: stereologic analysis of 78 Bennett DA, Schneider JA, Wilson RS, Bienias JL, Arnold SE. prefrontal cortex area 9. J Comp Neurol 2003; 463: 281–302. Neurofibrillary tangles mediate the association of amyloid load 60 Gold G, Kovari E, Corte G, Herrmann FR, Canuto A, Bussiere T with clinical Alzheimer disease and level of cognitive function. et al. Clinical validity of A beta-protein deposition staging in Arch Neurol 2004; 61: 378–384. brain aging and Alzheimer disease. J Neuropathol Exp Neurol 79 Bennett DA, Schneider JA, Wilson RS, Bienias JL, Arnold SE. 2001; 60: 946–952. Education modifies the association of amyloid but not tangles 61 Ikonomovic MD, Wecker L, Abrahamson EE, Wuu J, Counts SE, with cognitive function. Neurology 200527; 65: 953–955. Ginsberg SD et al. Cortical alpha7 nicotinic acetylcholine 80 Cummings BJ, Pike CJ, Shankle R, Cotman CW. Beta-amyloid receptor and beta-amyloid levels in early Alzheimer disease. deposition and other measures of neuropathology predict Arch Neurol 2009; 66: 646–651. cognitive status in Alzheimer’s disease. Neurobiol Aging 1996; 62 Ho L, Guo Y, Spielman L, Petrescu O, Haroutunian V, Purohit D 17: 921–933. et al. Altered expression of a-type but not b-type synapsin 81 Parvathy S, Davies P, Haroutunian V, Purohit DP, Davis KL, Mohs isoform in the brain of patients at high risk for Alzheimer’s RC et al. Correlation between Abetax-40-, Abetax-42-, and disease assessed by DNA microarray technique. Neurosci Lett Abetax-43-containing amyloid plaques and cognitive decline. 2001; 298: 191–194. Arch Neurol 2001; 58: 2025–2032. 63 Loeffler DA, Camp DM, Bennett DA. Plaque complement 82 Forman MS, Mufson EJ, Leurgans S, Pratico D, Joyce S, Leight S activation and cognitive loss in Alzheimer’s disease. J Neuroin- et al. Cortical biochemistry in MCI and Alzheimer disease: flamm 2008; 5:9. lack of correlation with clinical diagnosis. Neurology 2007; 68: 64 Archer HA, Schott JM, Barnes J, Fox NC, Holton JL, Revesz T 757–763. et al. Knight’s move thinking? Mild cognitive impairment in a 83 Neuropathology Group of the Medical Research Council chess player. Neurocase 2005; 11: 26–31. Cognitive Function Ageing Study (MRC CFAS).. Pathological 65 Thal DR, Holzer M, Rub U, Waldmann G, Gunzel S, Zedlick D correlates of late-onset dementia in a multicentre, community- et al. Alzheimer-related tau-pathology in the perforant path target based population in England and Wales. Lancet 2001; 357: zone and in the hippocampal stratum oriens and radiatum 169–175. correlates with onset and degree of dementia. Exp Neurol 2000; 84 Price JL, McKeel Jr DW, Buckles VD, Roe CM, Xiong C, Grundman 163: 98–110. M et al. Neuropathology of nondemented aging: presumptive 66 Uboga NV, Price JL. Formation of diffuse and fibrillar tangles in evidence for preclinical Alzheimer disease. Neurobiol Aging aging and early Alzheimer’s disease. Neurobiol Aging 2000; 21: 2009; 30: 1026–1036. 1–10. 85 Haroutunian V, Hoffman LB, Beeri MS. Is there a neuropathology 67 Delacourte A, David JP, Sergeant N, Buee L, Wattez A, difference between mild cognitive impairment and dementia? Vermersch P et al. The biochemical pathway of neurofibrillary Dialogues Clin Neurosci 2009; 11: 171–179. degeneration in aging and Alzheimer’s disease. Neurology 1999; 86 Pham E, Crews L, Ubhi K, Hansen L, Adame A, Cartier A et al. 52: 1158–1165. Progressive accumulation of amyloid-beta oligomers in Alzhei- 68 Haroutunian V, Davies P, Vianna C, Buxbaum JD, Purohit DP. Tau mer’s disease and in amyloid precursor protein transgenic mice is protein abnormalities associated with the progression of Alzhei- accompanied by selective alterations in synaptic scaffold pro- mer disease type dementia. Neurobiol Aging 2007; 28: 1–7. teins. FEBS J 2010; 277: 3051–3067.

Molecular Psychiatry Neuropathology of MCI BCM Stephan et al 1071 87 Molano J, Boeve B, Ferman T, Smith G, Parisi J, Dickson D et al. 108 Grudzien A, Shaw P, Weintraub S, Bigio E, Mash DC, Mild cognitive impairment associated with limbic and neocor- Mesulam MM. Locus coeruleus neurofibrillary degeneration in tical Lewy body disease: a clinicopathological study. Brain 2010; aging, mild cognitive impairment and early Alzheimer’s disease. 133(Part 2): 540–556. Neurobiol Aging 2007; 28: 327–335. 88 Mizutani T, Shimada H. Neuropathological background 109 Lyness SA, Zarow C, Chui HC. Neuron loss in key cholinergic of twenty-seven centenarian brains. J Neurol Sci 1992; 108: and aminergic nuclei in Alzheimer disease: a meta-analysis. 168–177. Neurobiol Aging 2003; 24: 1–23. 89 Kovari E, Gold G, Herrmann FR, Canuto A, Hof PR, Michel JP 110 Nagy Z, Esiri MM, Smith AD. The cell division cycle and the et al. Cortical microinfarcts and demyelination significantly pathophysiology of Alzheimer’s disease. Neuroscience 1998; 87: affect cognition in brain aging. Stroke 2004; 35: 410–414. 731–739. 90 Kovari E, Gold G, Herrmann FR, Canuto A, Hof PR, Bouras C 111 Currais A, Hortobagyi T, Soriano S. The neuronal cell cycle as a et al. Cortical microinfarcts and demyelination affect cognition in mechanism of pathogenesis in Alzheimer’s disease. Aging cases at high risk for dementia. Neurology 2007; 68: 927–931. (Albany NY) 2009; 1: 363–371. 91 Bouras C, Kovari E, Herrmann FR, Rivara CB, Bailey TL, von 112 Arendt T. Synaptic plasticity and cell cycle activation in neurons Gunten A et al. Stereologic analysis of microvascular morphology are alternative effector pathways: the ‘Dr Jekyll and Mr. Hyde in the elderly: Alzheimer disease pathology and cognitive status. concept’ of Alzheimer’s disease or the yin and yang of J Neuropathol Exp Neurol 2006; 65: 235–244. neuroplasticity. Prog Neurobiol 2003; 71: 83–248. 92 Gomez-Isla T, Price JL, McKeel Jr DW, Morris JC, Growdon JH, 113 Ou CY, Poon VY, Maeder CI, Watanabe S, Lehrman EK, Fu AK Hyman BT. Profound loss of layer II entorhinal cortex neurons et al. Two cyclin-dependent kinase pathways are essential for occurs in very mild Alzheimer’s disease. J Neurosci 1996; 16: polarized trafficking of presynaptic components. Cell 2010; 141: 4491–4500. 846–858. 93 Kordower JH, Chu Y, Stebbins GT, DeKosky ST, Cochran EJ, 114 Lalioti V, Pulido D, Sandoval. Cdk5, the multifunctional Bennett D et al. Loss and atrophy of layer II entorhinal cortex surveyor. Cell Cycle 2010; 9: 284–311. neurons in elderly people with mild cognitive impairment. Ann 115 Lopes JP, Oliveira CR, Agostinho P. Neurodegeneration in an Neurol 2001; 49: 202–213. Abeta-induced model of Alzheimer’s disease: the role of Cdk5. 94 Dickerson BC, Goncharova I, Sullivan MP, Forchetti C, Aging Cell 2010; 9: 64–77. Wilson RS, Bennett DA et al. MRI-derived entorhinal and 116 Cheung ZH, Ip NY. The roles of cyclin-dependent kinase 5 in hippocampal atrophy in incipient and very mild Alzheimer’s dendrite and synapse development. Biotechnol J 2007; 2: 949–957. disease. Neurobiol Aging 2001; 22: 747–754. 117 Evans GJ, Cousin MA. Activity-dependent control of slow 95 Stoub TR, Rogalski EJ, Leurgans S, Bennett DA, deToledo-Morrell synaptic vesicle endocytosis by cyclin-dependent kinase 5. L. Rate of entorhinal and hippocampal atrophy in incipient and J Neurosci 2007; 27: 401–411. mild AD: relation to memory function. Neurobiol Aging 2010; 31: 118 Tan TC, Valova VA, Malladi CS, Graham ME, Berven LA, Jupp OJ 1089–1098. et al. Cdk5 is essential for synaptic vesicle endocytosis. Nat Cell 96 Gosche KM, Mortimer JA, Smith CD, Markesbery WR, Snowdon Biol 2003; 5: 701–710. DA. Hippocampal volume as an index of Alzheimer neuropathol- 119 Lai KO, Ip NY. Recent advances in understanding the roles of ogy: findings from the Nun Study. Neurology 2002; 58: 1476–1482. Cdk5 in synaptic plasticity. Biochim Biophys Acta 2009; 1792: 97 Frisoni GB, Fox NC, Jack CR, Scheltens P, Thompson PM. The 741–745. clinical use of structural MRI in Alzheimer disease. Nat Rev 120 Fu AK, Fu WY, Cheung J, Tsim KW, Ip FC, Wang JH et al. Cdk5 is Neurol 2010; 6: 67–77. involved in neuregulin-induced AChR expression at the 98 Akram A, Christoffel D, Rocher AB, Bouras C, Kovari E, Perl DP et neuromuscular junction. Nat Neurosci 2001; 4: 374–381. al. Stereologic estimates of total spinophilin-immunoreactive 121 Zhang S, Edelmann L, Liu J, Crandall JE, Morabito MA. spine number in area 9 and the CA1 field: relationship with the Cdk5 regulates the phosphorylation of tyrosine 1472 NR2B and progression of Alzheimer’s disease. Neurobiol Aging 2008; 29: the surface expression of NMDA receptors. J Neurosci 2008; 28: 1296–1307. 415–424. 99 Scheff SW, Price DA, Schmitt FA, DeKosky ST, Mufson EJ. 122 Albrecht S, Bourdeau M, Bennett D, Mufson EJ, Bhattacharjee M, Synaptic alterations in CA1 in mild Alzheimer disease and mild LeBlanc AC. Activation of caspase-6 in aging and mild cognitive cognitive impairment. Neurology 2007; 68: 1501–1508. impairment. Am J Pathol 2007; 170: 1200–1209. 100 Scheff SW, Price DA, Schmitt FA, Scheff MA, Mufson EJ. 123 Bader Lange ML, Cenini G, Piroddi M, Abdul HM, Sultana R, Synaptic loss in the inferior temporal gyrus in mild cognitive Galli F et al. Loss of phospholipid asymmetry and elevated impairment and Alzheimer’s disease. J Alzheimers Dis 2011; 24: brain apoptotic protein levels in subjects with amnestic mild 547–557. cognitive impairment and Alzheimer disease. Neurobiol Dis 101 Scheff SW, Price DA, Schmitt FA, Mufson EJ. Hippocampal 2008; 29: 456–464. synaptic loss in early Alzheimer’s disease and mild cognitive 124 Pompl PN, Yemul S, Xiang Z, Ho L, Haroutunian V, Purohit D impairment. Neurobiol Aging 2006; 27: 1372–1384. et al. Caspase gene expression in the brain as a function of the 102 Terry RD, Masliah E, Salmon DP, Butters N, DeTeresa R, Hill R clinical progression of Alzheimer disease. Arch Neurol 2003; 60: et al. Physical basis of cognitive alterations in Alzheimer’s 369–376. disease: synapse loss is the major correlate of cognitive impair- 125 Raina AK, Hochman A, Zhu X, Rottkamp CA, Nunomura A, ment. Ann Neurol 1991; 30: 572–580. Siedlak SL et al. Abortive apoptosis in Alzheimer’s disease. Acta 103 Arendt T. Synaptic degeneration in Alzheimer’s disease. Acta Neuropathol 2001; 101: 305–310. Neuropathol 2009; 118: 167–179. 126 Rissman RA, Poon WW, Blurton-Jones M, Oddo S, Torp R, 104 Carmel G, Mager EM, Binder LI, Kuret J. The structural basis of Vitek MP et al. Caspase-cleavage of tau is an early event in monoclonal antibody Alz50’s selectivity for Alzheimer’s disease Alzheimer disease tangle pathology. J Clin Invest 2004; 114: pathology. J Biol Chem 1996; 271: 32789–32795. 121–130. 105 Goedert M, Jakes R, Vanmechelen E. Monoclonal antibody AT8 127 Guillozet-Bongaarts AL, Garcia-Sierra F, Reynolds MR, Horowitz recognises tau protein phosphorylated at both serine 202 and PM, Fu Y, Wang T et al. Tau truncation during neurofibrillary threonine 205. Neurosci Lett 1995; 189: 167–169. tangle evolution in Alzheimer’s disease. Neurobiol Aging 2005; 106 Porzig R, Singer D, Hoffmann R. Epitope mapping of mAbs AT8 26: 1015–1022. and Tau5 directed against hyperphosphorylated regions of the 128 Sultana R, Banks WA, Butterfield DA. Decreased levels of PSD95 human tau protein. Biochem Biophys Res Commun 2007; 358: and two associated proteins and increased levels of BCl2 and 644–649. caspase 3 in hippocampus from subjects with amnestic mild 107 Ginsberg SD, Che S, Counts SE, Mufson EJ. Shift in the ratio of cognitive impairment: insights into their potential roles for loss three-repeat tau and four-repeat tau mRNAs in individual of synapses and memory, accumulation of Abeta, and neurode- cholinergic basal forebrain neurons in mild cognitive impairment generation in a prodromal stage of Alzheimer’s disease. and Alzheimer’s disease. J Neurochem 2006; 96: 1401–1408. J Neurosci Res 2010; 88: 469–477.

Molecular Psychiatry Neuropathology of MCI BCM Stephan et al 1072 129 Chong ZZ, Li F, Maiese K. Stress in the brain: novel cellular 148 Shao C, Xiong S, Li GM, Gu L, Mao G, Markesbery WR et al. mechanisms of injury linked to Alzheimer’s disease. Brain Res Altered 8-oxoguanine glycosylase in mild cognitive impairment Rev 2005; 49: 1–21. and late-stage Alzheimer’s disease brain. Free Radic Biol Med 130 Calabrese V, Butterfield DA, Scapagnini G, Stella AM, Maines 2008; 45: 813–819. MD. Redox regulation of heat shock protein expression by 149 Ansari MA, Scheff SW. Oxidative stress in the progression of signaling involving nitric oxide and carbon monoxide: relevance Alzheimer disease in the frontal cortex. J Neuropathol Exp to brain aging, neurodegenerative disorders, and longevity. Neurol 2010; 69: 155–167. Antioxid Redox Signal 2006; 8: 444–477. 150 Cenini G, Sultana R, Memo M, Butterfield DA. Effects of 131 Di Domenico F, Sultana R, Tiu GF, Scheff NN, Perluigi M, oxidative and nitrosative stress in brain on p53 proapoptotic Cini C et al. Protein levels of heat shock proteins 27, 32, 60, 70, protein in amnestic mild cognitive impairment and Alzheimer 90 and thioredoxin-1 in amnestic mild cognitive impairment: disease. Free Radic Biol Med 2008; 45: 81–85. an investigation on the role of cellular stress response in 151 Sultana R, Reed T, Perluigi M, Coccia R, Pierce WM, Butterfield the progression of Alzheimer disease. Brain Res 2010; 1333: DA. Proteomic identification of nitrated brain proteins in 72–81. amnestic mild cognitive impairment: a regional study. J Cell 132 Butterfield DA, Poon HF, St Clair D, Keller JN, Pierce WM, Klein Mol Med 2007; 11: 839–851. JB et al. Redox proteomics identification of oxidatively modified 152 Sultana R, Piroddi M, Galli F, Butterfield DA. Protein levels and hippocampal proteins in mild cognitive impairment: insights activity of some antioxidant enzymes in hippocampus of subjects into the development of Alzheimer’s disease. Neurobiol Dis 2006; with amnestic mild cognitive impairment. Neurochem Res 2008; 22: 223–232. 33: 2540–2546. 133 Counts SE, Nadeem M, Lad SP, Wuu J, Mufson EJ. Differential 153 Weissman L, Jo DG, Sorensen MM, de Souza-Pinto NC, expression of synaptic proteins in the frontal and temporal cortex Markesbery WR, Mattson MP et al. Defective DNA base excision of elderly subjects with mild cognitive impairment. J Neuro- repair in brain from individuals with Alzheimer’s disease and pathol Exp Neurol 2006; 65: 592–601. amnestic mild cognitive impairment. Nucleic Acids Res 2007; 35: 134 Glavan G, Schliebs R, Zivin M. Synaptotagmins in neuro- 5545–5555. degeneration. Anat Rec (Hoboken) 2009; 292: 1849–1862. 154 Fernandez-Vizarra P, Fernandez AP, Castro-Blanco S, Encinas JM, 135 Head E, Corrada MM, Kahle-Wrobleski K, Kim RC, Sarsoza F, Serrano J, Bentura ML et al. Expression of nitric oxide system in Goodus M et al. Synaptic proteins, neuropathology and cognitive clinically evaluated cases of Alzheimer’s disease. Neurobiol Dis status in the oldest-old. Neurobiol Aging 2009; 30: 1125–1134. 2004; 15: 287–305. 136 Ho L, Fivecoat H, Wang J, Pasinetti GM. Alzheimer’s disease 155 Sultana R, Perluigi M, Butterfield DA. Oxidatively modified biomarker discovery in symptomatic and asymptomatic patients: proteins in Alzheimer’s disease (AD), mild cognitive impairment experimental approaches and future clinical applications. Exp and animal models of AD: role of Abeta in pathogenesis. Acta Gerontol 2010; 45: 15–22. Neuropathol 2009; 118: 131–150. 137 Ho L, Sharma N, Blackman L, Festa E, Reddy G, Pasinetti GM. 156 Mangialasche F, Polidori MC, Monastero R, Ercolani S, From proteomics to biomarker discovery in Alzheimer’s disease. Camarda C, Cecchetti R et al. Biomarkers of oxidative and Brain Res Rev 2005; 48: 360–369. nitrosative damage in Alzheimer’s disease and mild cognitive 138 Canto´ C, Auwerx J. PGC-1[alpha], SIRT1 and AMPK, an energy impairment. Ageing Res Rev 2009; 8: 285–305. sensing network that controls energy expenditure. Curr Opin 157 Butterfield DA, Reed T, Newman SF, Sultana R. Roles of amyloid Lipidol 2009; 20: 98–105. beta-peptide-associated oxidative stress and brain protein mod- 139 Qin W, Haroutunian V, Katsel P, Cardozo CP, Ho L, Buxbaum JD ifications in the pathogenesis of Alzheimer’s disease and mild et al. PGC-1alpha expression decreases in the Alzheimer cognitive impairment. Free Radic Biol Med 2007; 43: 658–677. disease brain as a function of dementia. Arch Neurol 2009; 66: 158 Lovell MA, Markesbery WR. Oxidative DNA damage in mild 352–361. cognitive impairment and late-stage Alzheimer’s disease. Nucleic 140 Butterfield DA, Reed T, Perluigi M, De Marco C, Coccia R, Cini C Acids Res 2007; 35: 7497–7504. et al. Elevated protein-bound levels of the lipid peroxidation 159 Lovell MA, Markesbery WR. Oxidative damage in mild cognitive product, 4-hydroxy-2-nonenal, in brain from persons with mild impairment and early Alzheimer’s disease. J Neurosci Res 2007; cognitive impairment. Neurosci Lett 2006; 397: 170–173. 85: 3036–3040. 141 Butterfield DA, Reed TT, Perluigi M, De Marco C, Coccia R, Keller 160 Bonda DJ, Wang X, Perry G, Nunomura A, Tabaton M, Zhu X et al. JN et al. Elevated levels of 3-nitrotyrosine in brain from subjects Oxidative stress in Alzheimer disease: a possibility for preven- with amnestic mild cognitive impairment: implications for the tion. Neuropharmacology 2010; 59: 290–294. role of nitration in the progression of Alzheimer’s disease. Brain 161 Sultana R, Butterfield DA. Role of oxidative stress in the Res 2007; 1148: 243–248. progression of Alzheimer’s disease. J Alzheimers Dis 2010; 19: 142 Lovell MA, Markesbery WR. Oxidatively modified RNA in mild 341–353. cognitive impairment. Neurobiol Dis 2008; 29: 169–175. 162 Markesbery WR, Lovell MA. Damage to lipids, proteins, DNA, and 143 Cecarini V, Ding Q, Keller JN. Oxidative inactivation of RNAinmildcognitiveimpairment.Arch Neurol 2007; 64: 954–956. the proteasome in Alzheimer’s disease. Free Radic Res 2007; 163 Guglielmotto M, Giliberto L, Tamagno E, Tabaton M. Oxidative 41: 673–680. stress mediates the pathogenic effect of different Alzheimer’s 144 Cenini G, Sultana R, Memo M, Butterfield DA. Elevated levels of disease risk factors. Front Aging Neurosci 2010; 2:3. pro-apoptotic p53 and its oxidative modification by the lipid 164 Butterfield DA, Lange ML. Multifunctional roles of enolase in peroxidation product, HNE, in brain from subjects with amnestic Alzheimer’s disease brain: beyond altered glucose metabolism. mild cognitive impairment and Alzheimer’s disease. J Cell Mol J Neurochem 2009; 111: 915–933. Med 2008; 12: 987–994. 165 Ding Q, Dimayuga E, Keller JN. Oxidative damage, protein 145 Keller JN, Schmitt FA, Scheff SW, Ding Q, Chen Q, Butterfield synthesis, and protein degradation in Alzheimer’s disease. Curr DA et al. Evidence of increased oxidative damage in subjects Alzheimer Res 2007; 4: 73–79. with mild cognitive impairment. Neurology 2005; 64: 1152–1156. 166 Ding Q, Markesbery WR, Cecarini V, Keller JN. Decreased RNA, 146 Reed T, Perluigi M, Sultana R, Pierce WM, Klein JB, Turner DM and increased RNA oxidation, in ribosomes from early et al. Redox proteomic identification of 4-hydroxy-2-nonenal- Alzheimer’s disease. Neurochem Res 2006; 31: 705–710. modified brain proteins in amnestic mild cognitive impairment: 167 Ding Q, Markesbery WR, Chen Q, Li F, Keller JN. Ribosome insight into the role of lipid peroxidation in the progression and dysfunction is an early event in Alzheimer’s disease. J Neurosci pathogenesis of Alzheimer’s disease. Neurobiol Dis 2008; 30: 2005; 25: 9171–9175. 107–120. 168 Keller JN. Interplay between oxidative damage, protein synthesis, 147 Sultana R, Perluigi M, Newman SF, Pierce WM, Cini C, Coccia R and protein degradation in Alzheimer’s disease. J Biomed et al. Redox proteomic analysis of carbonylated brain proteins in Biotechnol 2006; 2006: 12129. mild cognitive impairment and early Alzheimer’s disease. 169 Ginsberg SD, Alldred MJ, Counts SE, Cataldo AM, Neve RL, Antioxid Redox Signal 2010; 12: 327–336. Jiang Y et al. Microarray analysis of hippocampal CA1 neurons

Molecular Psychiatry Neuropathology of MCI BCM Stephan et al 1073 implicates early endosomal dysfunction during Alzheimer’s 188 Counts SE, Nadeem M, Wuu J, Ginsberg SD, Saragovi HU, disease progression. Biol Psychiatry 2010; 68: 885–893. Mufson EJ. Reduction of cortical TrkA but not p75(NTR) protein 170 Ginsberg SD, Mufson EJ, Counts SE, Wuu J, Alldred MJ, Nixon in early-stage Alzheimer’s disease. Ann Neurol 2004; 56: RA et al. Regional selectivity of rab5 and rab7 protein upregula- 520–531. tion in mild cognitive impairment and Alzheimer’s disease. 189 Mufson EJ, Ikonomovic MD, Styren SD, Counts SE, Wuu J, J Alzheimers Dis 2010; 22: 631–639. Leurgans S et al. Preservation of brain nerve growth factor in 171 Davis KL, Mohs RC, Marin D, Purohit DP, Perl DP, Lantz M et al. mild cognitive impairment and Alzheimer disease. Arch Neurol Cholinergic markers in elderly patients with early signs of 2003; 60: 1143–1148. Alzheimer disease. JAMA 1999; 281: 1401–1406. 190 Peng S, Wuu J, Mufson EJ, Fahnestock M. Precursor form of 172 DeKosky ST, Ikonomovic MD, Styren SD, Beckett L, brain-derived neurotrophic factor and mature brain-derived Wisniewski S, Bennett DA et al. Upregulation of choline neurotrophic factor are decreased in the pre-clinical stages of acetyltransferase activity in hippocampus and frontal cortex of Alzheimer’s disease. J Neurochem 2005; 93: 1412–1421. elderly subjects with mild cognitive impairment. Ann Neurol 191 Peng S, Wuu J, Mufson EJ, Fahnestock M. Increased proNGF 2002; 51: 145–155. levels in subjects with mild cognitive impairment and mild 173 Bierer LM, Haroutunian V, Gabriel S, Knott PJ, Carlin LS, Purohit Alzheimer disease. J Neuropathol Exp Neurol 2004; 63: 641–649. DP et al. Neurochemical correlates of dementia severity in 192 Bruno MA, Mufson EJ, Wuu J, Cuello AC. Increased matrix Alzheimer’s disease: relative importance of the cholinergic metalloproteinase 9 activity in mild cognitive impairment. deficits. J Neurochem 1995; 64: 749–760. J Neuropathol Exp Neurol 2009; 68: 1309–1318. 174 Schaeffer EL, Gattaz WF. Cholinergic and glutamatergic altera- 193 Mufson EJ, Counts SE, Fahnestock M, Ginsberg SD. Cholino- tions beginning at the early stages of Alzheimer disease: trophic molecular substrates of mild cognitive impairment in the participation of the phospholipase A2 enzyme. Psychopharma- elderly. Curr Alzheimer Res 2007; 4: 340–350. cology (Berl) 2008; 198: 1–27. 194 Davis KL, Mohs RC, Marin DB, Purohit DP, Perl DP, Lantz M et al. 175 Gilmor ML, Erickson JD, Varoqui H, Hersh LB, Bennett DA, Neuropeptide abnormalities in patients with early Alzheimer Cochran EJ et al. Preservation of nucleus basalis neurons disease. Arch Gen Psychiatry 1999; 56: 981–987. containing choline acetyltransferase and the vesicular acetylcho- 195 Bell KF, Bennett DA, Cuello AC. Paradoxical upregulation of line transporter in the elderly with mild cognitive impairment and glutamatergic presynaptic boutons during mild cognitive impair- early Alzheimer’s disease. J Comp Neurol 1999; 411: 693–704. ment. J Neurosci 2007; 27: 10810–10817. 176 Ginsberg SD, Che S, Wuu J, Counts SE, Mufson EJ. Down 196 Owen JB, Di Domenico F, Sultana R, Perluigi M, Cini C, Pierce regulation of trk but not p75NTR gene expression in single WM et al. Proteomics-determined differences in the concanava- cholinergic basal forebrain neurons mark the progression of lin-A-fractionated proteome of hippocampus and inferior parietal Alzheimer’s disease. J Neurochem 2006; 97: 475–487. lobule in subjects with Alzheimer’s disease and mild cognitive 177 Ikonomovic MD, Mufson EJ, Wuu J, Bennett DA, DeKosky ST. impairment: implications for progression of AD. J Proteome Res Reduction of choline acetyltransferase activity in primary visual 2009; 8: 471–482. cortex in mild to moderate Alzheimer’s disease. Arch Neurol 197 Citron BA, Suo Z, SantaCruz K, Davies PJ, Qin F, Festoff BW. 2005; 62: 425–430. Protein crosslinking, tissue transglutaminase, alternative splicing 178 Dubelaar EJ, Mufson EJ, ter Meulen WG, Van Heerikhuize JJ, and neurodegeneration. Neurochem Int 2002; 40: 69–78. Verwer RW, Swaab DF. Increased metabolic activity in nucleus 198 Lovell MA, Smith JL, Xiong S, Markesbery WR. Alterations in basalis of Meynert neurons in elderly individuals with mild zinc transporter protein-1 (ZnT-1) in the brain of subjects with cognitive impairment as indicated by the size of the Golgi mild cognitive impairment, early, and late-stage Alzheimer’s apparatus. J Neuropathol Exp Neurol 2006; 65: 257–266. disease. Neurotox Res 2005; 7: 265–271. 179 Perry E, Martin-Ruiz C, Lee M, Griffiths M, Johnson M, Piggott M 199 Mufson EJ, Wuu J, Counts SE, Nykjaer A. Preservation of cortical et al. Nicotinic receptor subtypes in human brain ageing, sortilin protein levels in MCI and Alzheimer’s disease. Neurosci Alzheimer and Lewy body diseases. Eur J Pharmacol 2000; 393: Lett 2010; 471: 129–133. 215–222. 200 Abdul HM, Sama MA, Furman JL, Mathis DM, Beckett TL, 180 Counts SE, He B, Che S, Ikonomovic MD, DeKosky ST, Ginsberg Weidner AM et al. Cognitive decline in Alzheimer’s disease is SD et al. Alpha7 nicotinic receptor up-regulation in cholinergic associated with selective changes in calcineurin/NFAT signaling. basal forebrain neurons in Alzheimer disease. Arch Neurol 2007; J Neurosci 2009; 29: 12957–12969. 64: 1771–1776. 201 Vaziri H, Dessain SK, Ng Eaton E, Imai SI, Frye RA, Pandita TK 181 Schmidt WE, Kratzin H, Eckart K, Drevs D, Mundkowski G, et al. hSIR2(SIRT1) functions as an NAD-dependent p53 Clemens A et al. Isolation and primary structure of pituitary deacetylase. Cell 2001; 107: 149–159. human galanin, a 30-residue nonamidated neuropeptide. Proc 202 Bonda DJ, Lee HG, Camins A, Pallas M, Casadesus G, Smith MA et al. Natl Acad Sci USA 1991; 88: 11435–11439. The sirtuin pathway in ageing and Alzheimer disease: mechanistic 182 Chan-Palay V. Neurons with galanin innervate cholinergic cells and therapeutic considerations. Lancet Neurol 2011; 10: 275–279. in the human basal forebrain and galanin and acetylcholine 203 Monteiro JP, Cano MI. SIRT1 deacetylase activity and the coexist. Brain Res Bull 1988; 21: 465–472. maintenance of protein homeostasis in response to stress: an 183 Counts SE, Chen EY, Che S, Ikonomovic MD, Wuu J, Ginsberg SD overview. Protein Pept Lett 2011; 18: 167–173. et al. Galanin fiber hypertrophy within the cholinergic nucleus 204 Donmez G, Wang D, Cohen DE, Guarente L. SIRT1 suppresses basalis during the progression of Alzheimer’s disease. Dement beta-amyloid production by activating the alpha-secretase gene Geriatr Cogn Disord 2006; 21: 205–214. ADAM10. Cell 2010; 142: 320–332. 184 Counts SE, Perez SE, Mufson EJ. Galanin in Alzheimer’s disease: 205 Julien C, Tremblay C, Emond V, Lebbadi M, Salem Jr N, Bennett neuroinhibitory or neuroprotective? Cell Mol Life Sci 2008; 65: DA et al. Sirtuin 1 reduction parallels the accumulation of tau in 1842–1853. Alzheimer disease. J Neuropathol Exp Neurol 2009; 68: 48–58. 185 Miller MA, Kolb PE, Leverenz JB, Peskind ER, Raskind MA. 206 Akiyama H, Barger S, Barnum S, Bradt B, Bauer J, Cole GM et al. Preservation of noradrenergic neurons in the locus ceruleus that Inflammation and Alzheimer’s disease. Neurobiol Aging 2000; coexpress galanin mRNA in Alzheimer’s disease. J Neurochem 21: 383–421. 1999; 73: 2028–2036. 207 Zanjani H, Finch CE, Kemper C, Atkinson J, McKeel D, Morris JC 186 Chu Y, Cochran EJ, Bennett DA, Mufson EJ, Kordower JH. Down- et al. Complement activation in very early Alzheimer disease. regulation of trkA mRNA within nucleus basalis neurons in Alzheimer Dis Assoc Disord 2005; 19: 55–66. individuals with mild cognitive impairment and Alzheimer’s 208 Ho L, Purohit D, Haroutunian V, Luterman JD, Willis F, Naslund J disease. J Comp Neurol 2001; 437: 296–307. et al. Neuronal cyclooxygenase 2 expression in the hippocampal 187 Mufson EJ, Ma SY, Cochran EJ, Bennett DA, Beckett LA, Jaffar S formation as a function of the clinical progression of Alzheimer et al. Loss of nucleus basalis neurons containing trkA immunor- disease. Arch Neurol 2001; 58: 487–492. eactivity in individuals with mild cognitive impairment and 209 Luterman JD, Haroutunian V, Yemul S, Ho L, Purohit D, Aisen PS early Alzheimer’s disease. J Comp Neurol 2000; 427: 19–30. et al. Cytokine gene expression as a function of the clinical

Molecular Psychiatry Neuropathology of MCI BCM Stephan et al 1074 progression of Alzheimer disease dementia. Arch Neurol 2000; protein to age-related cognitive decline. Neurosci Res 2004; 48: 57: 1153–1160. 93–100. 210 Xiang Z, Haroutunian V, Ho L, Purohit D, Pasinetti GM. Microglia 217 Han X, D MH, McKeel Jr DW, Kelley J, Morris JC. Substantial activation in the brain as inflammatory biomarker of Alzheimer’s sulfatide deficiency and ceramide elevation in very early disease neuropathology and clinical dementia. Disease Markers Alzheimer’s disease: potential role in disease pathogenesis. 2006; 22: 95–102. J Neurochem 2002; 82: 809–818. 211 Barone E, Di Domenico F, Cenini G, Sultana R, Cini C, Preziosi P 218 Haroutunian V, Katsel P, Schmeidler J. Transcriptional vulner- et al. Biliverdin reductase—a protein levels and activity in the ability of brain regions in Alzheimer’s disease and dementia. brains of subjects with Alzheimer disease and mild cognitive Neurobiol Aging 2009; 30: 561–573. impairment. Biochim Biophys Acta 2011; 1812: 480–487. 219 Johnson JK, Vogt BA, Kim R, Cotman CW, Head E. Isolated 212 Barone E, Di Domenico F, Cenini G, Sultana R, Coccia R, Preziosi P executive impairment and associated frontal neuropathology. et al. Oxidative and nitrosative modifications of biliverdin reduc- Dement Geriatr Cogn Disord 2004; 17: 360–367. tase-A in the brain of subjects with Alzheimer’s disease and amne- 220 Snowdon DA, Kemper SJ, Mortimer JA, Greiner LH, Wekstein stic mild cognitive impairment. J Alzheimers Dis 2011; 25: 623–633. DR, Markesbery WR. Linguistic ability in early life and cognitive 213 Funderburk SF, Marcellino BK, Yue Z. Cell ‘self-eating’ (autophagy) function and Alzheimer’s disease in late life. Findings from the mechanism in Alzheimer’s disease. Mt Sinai J Med 2010; 77: 59–68. Nun Study. JAMA 1996; 275: 528–532. 214 Pickford F, Masliah E, Britschgi M, Lucin K, Narasimhan R, Jaeger 221 Snowdon DA, Greiner LH, Markesbery WR. Linguistic ability in PA et al. The autophagy-related protein beclin 1 shows reduced early life and the neuropathology of Alzheimer’s disease and expression in early Alzheimer disease and regulates amyloid beta cerebrovascular disease. Findings from the Nun Study. Ann N Y accumulation in mice. J Clin Invest 2008; 118: 2190–2199. Acad Sci 2000; 903: 34–38. 215 Han X, Holtzman DM, McKeel Jr DW. Plasmalogen deficiency in 222 Brayne C. Clinicopathological studies of the dementias from an early Alzheimer’s disease subjects and in animal models: epidemiological viewpoint. Br J Psychiatry 1993; 162: 439–446. molecular characterization using electrospray ionization mass 223 Chao LL, Mueller SG, Buckley ST, Peek K, Raptentsetseng S, spectrometry. J Neurochem 2001; 77: 1168–1180. Elman J et al. Evidence of neurodegeneration in brains of older 216 Wang DS, Bennett DA, Mufson EJ, Mattila P, Cochran E, Dickson adults who do not yet fulfill MCI criteria. Neurobiol Aging 2010; DW. Contribution of changes in ubiquitin and myelin basic 31: 368–377.

Supplementary Information accompanies the paper on the Molecular Psychiatry website (http://www.nature.com/mp)

APPENDIX 1: ABBREVIATIONS BVR-A, biliverdin reductase A CAII, carbonic anhydrase II a7nAChR, alpha7 nicotinic acetylcholine receptor CAA, cerebral amyloid angiopathy DTau, cleavage of tau at Asp421 CAT, Catalase 3-NT, 3-nitrotyrosine CBF, cholinergic basal forebrain 3Rtau, three-repeat tau CDK, cyclin-dependent kinases 4-NE, 4-hydroxynonenals cDNA, complementary DNA 4Rtau, four-repeat tau CDR, Clinical Dementia Rating Score 5-OH-cytosine, 5-hydroxycytosine Cer, ceramides 8-OH-adenine, 8-hydroxyadenine CER, cerebellum 8-OHdG, 8-hydroxy-2-deoxy guanosine CERAD, Consortium to Establish a Registry for 8-OH-guanine, 8-hydroxyguanine Alzheimer’s Disease 8-OHG, 8-hydroxyguanosine ChAT, choline acetyltransferase A–Z ChAT-ir, ChAT-like immunoreactive Ab, amyloid beta-peptide CI-CVD, cognitive impairment with lacunes ABCA1, ATP Binding Cassette Transporter A1 CIND, Cognitive Impairment no Dementia AChE, acetylcholinesterase CN, calcineurin AD, Alzheimer’s disease COX-2, cyclooxygenase-2 Adv-AD, advanced AD CRF, corticotptrophin-releasing factor AGD, argyrophilic grain disease Csp, caspase aMCI, amnestic mild cognitive impairment Cu,Zn-SOD, superoxide dismutases enzyme with AMY, amygdala copper and zinc APOE E, apolipoprotein E Cyt-c, cytochrome c APP, amyloid precursor protein Def-AD, Definite AD ASYMAD, asymptomatic Alzheimer’s disease (no DEM, dementia cognitive decline but AD pathology) Diff., difference ATP, adenosine triphosphate Dihydro DEH, dihydrolipoyl DEH BACE1, beta-site amyloid precursor protein-cleaving DP, diffuse plaques enzyme 1 DRB, drebrin Bax, Bcl-2-associated X protein DRP-2, dihydropyrminidase like-2 b/w, between DWMD, deep white matter demyelination Bcl-2, B-cell lymphoma 2 EAAT2, excitatory amino acid transporter 2 Bcl-xL, antiapoptotic member of the Bcl-2 family E-AD, early AD BER, base-excision repair EC, entorhinal cortex BDNF, brain-derived neurotrophic factor EfTu, elongation factor Tu

Molecular Psychiatry Neuropathology of MCI BCM Stephan et al 1075 ENO1, enolase NADPH, carbonyl reductase EpoR, erythropoietin receptor NB, nucleus basalis ERK1/2, extracellular signal-regulated kinases-1/2 NCI, no cognitive impairment ERK, extracellular-regulated kinase nDNA, nuclear deoxyribonucleic acid (DNA) F(2)-IsoP, F(2)-isoprostane NFAT, nuclear factor of activated T-cells F(4)-NP, F(4)-neuroprostane NFT, intraneuronal fibrillary tangles FBA-C, fructose bisphospate aldolase C NGF, nerve growth factor FC, frontal cortex NIA-Reagan, National Institute on Aging-Reagan FRAP, ferric reducing ability of plasma Institute criteria G-6-PD, glucose-6-phosphate nMCI, non-amnestic mild cognitive impairment G6PDH, glucose-6-phosphate dehydrogenase nNOS, neuronal nitric oxide synthase GA, Golgi apparatus NOX, nicotinamide adenosine dinucleotide phos- GalC, galactocerebroside(s) phate oxidase GAPDH, glyceraldehyde-3-phosphate dehydrogenase NP, neuritic plaque GFAP, glial fibrillary acidic protein NPrG, 1-N2-propanodeoxyguanosine GLUL, glutamine synthetase NR2A, N-methyl-D-aspartate receptor subunit 2A GPx, glutathione peroxidase OGG-1, oxoguanine glycosylase 1 GR, glutathione reductase p75NRT, p75 neurotrophin GRP-78, glucose-regulated protein 78 PARP, poly (ADP-ribose) polymerase Grp., group Pc, parietal cortex GSH, glutathione PC, choline glycerophospholipid GSSG, oxidized form glutathione disulfide PCAD, Pre-clinical Alzheimer’s disease GST, glutathione-S-transferase PCNA, proliferating cell nuclear antigen GSTM3, glutathione-S-transferases Mu PD, Parkinson’s disease HHE, protein-bound 4-hydroxhexenal PE, ethanolamine plasmalogen HIPP, hippocampus PEBP1, phosphatidylethanolamine binding protein 1 HLA-DR, human leukocyte antigen DR-1 PGC-1a, peroxisome proliferator-activated receptor HNE, 4-hydroxy-2-nonenal gamma coactivator-1alpha HO-1, heme oxygenase-1 PHF, paired helical filaments HPG, parahippocampal gyrus PHG, parahippocampal gyrus HS, hippocampal sclerosis Pin1, peptidyl-prolyl cis/trans isomerase HSP, heat-shock protein PK-M1, pyruvate kinase M1-type IL-6, interleukin-6 PKM2, pyruvate kinase M2 iNOS, inducible nitric oxide synthase PM1, carbonylation of phosphoglycerate mutase 1 IPL, inferior parietal lobe PMS, postmitochondrial supernatant L-AD, late stage AD Po-AD, possible AD LB, Lewy body PPARG, peroxisome proliferator-activated receptor gamma LBD, Lewy body dementia PR VI, peroxiredoxin 6 LC, locus coeruleus Pr-AD, probable AD LOAD, late onset AD proBDNF, precursor of brain-derived neurotrophic LRP1, lipoprotein receptor-1 factor MAPKI, mitogen-activated protein kinase I proNGF, pro-form of nerve growth factor MAP2KI, mitogen-activated protein kinase kinase I PSD-95, postsynaptic density-95 MAPs, microtubule-associated proteins PtdSer, phosphatidylserine MBP, myelin basic protein pTyr-BVR-A, phosphorylation for Tyr residues of MCI, mild cognitive impairment biliverdin reductass-A MCI-LB, pre-clinical Lewy body dementia state PVD, periventricular demyelination MDH, mdh2 encoded malate dehydrogenase rMCI, revised MCI Mi-AD, mild AD RNS, reactive nitrogen species miRNA (miR-107), micro RNA miR-107 ROS, reactive oxygen species MLA, [3H]methyllycaconitine rRNA, ribosomal RNA MMSE, Mini Mental State Examination S-AD, severe AD mAChRs, muscarinic AChRs SBP1, synaptic binding protein I MMPs, matrix metalloproteinases Sds22, protein phosphatase-related protein Sds-22 MMP-9, extracellular collagenase metalloproteinase-9 SFC, superior frontal cortex Mo-AD, moderate AD Sig., Significant mRNA, messenger ribonucleic acid (RNA) SIRT1, sirtuin 1 MRP3, multidrug resistant protein SIVD, subcortical ischemic vascular dementia mtDNA, mitochondrial deoxyribonucleic acid SLI, somatostatin-like immunoreactivity (DNA) SMTG, superior middle temporal gyri nAChR, nicotinic acetylcholine receptor SNAP25, synaptosomal-associated protein 25

Molecular Psychiatry Neuropathology of MCI BCM Stephan et al 1076 SOD, superoxide dismutase TGF-b, transforming growth factor beta pathway Sp-ir, spinophilin immunoreactivity Trk, tyrosine kinase ST, sulfatides tRNA, transfer ribonucleic acid (RNA) STC, superior temporal cortex tTG, tissue transglutaminase STs, sulfated galactosylceramides UCHL1, ubiquitin carboxy-terminal hydrolase STS, superior temporal sulcus VAChT, vesicular acetylcholine transporter SYN, synaptophysin VAMP2, vesicle-associated membrane protein 2 SYT, synaptotagmin Zn, zinc TauDCsp-6, tau cleaved by caspase 6 ZnT-1, Zn transporter protein-1 TBARS, thiobarbituric acid reactive substances ZnT-4, Zn transporter protein-4 TGF, transforming growth factor ZnT-6, Zn transporter protein-6

Molecular Psychiatry