Cited by 122Alzheimer's disease research and treatmentsDementia and Cognitive Impairment ResearchAdvanced Proteomics Techniques and Applications

The DOI matches its Crossref record; 209 of the 225 references it cites were found in a registry or an index; 2 of the 282 sources its claims cite were checked for support.

Systems-based proteomics to resolve the biology of Alzheimer’s disease beyond amyloid and tau

Open access

Abstract

The repeated failures of amyloid-targeting therapies have challenged our narrow understanding of Alzheimer’s disease (AD) pathogenesis and inspired wide-ranging investigations into the underlying mechanisms of disease. Increasing evidence indicates that AD develops from an intricate web of biochemical and cellular processes that extend far beyond amyloid and tau accumulation. This growing recognition surrounding the diversity of AD pathophysiology underscores the need for holistic systems-based approaches to explore AD pathogenesis. Here we describe how network-based proteomics has emerged as a powerful tool and how its application to the AD brain has provided an informative framework for the complex protein pathophysiology underlying the disease. Furthermore, we outline how the AD brain network proteome can be leveraged to advance additional scientific and translational efforts, including the discovery of novel protein biomarkers of disease.

Overview

AI snapshot

Written from the paper’s full text.

TL;DR

This review argues that Alzheimer’s disease pathophysiology extends beyond amyloid and tau and that network-based proteomics of human brain, CSF, and plasma can resolve conserved protein modules reflecting inflammatory, myelination, RNA-binding/splicing, synaptic, mitochondrial, and cytoskeletal changes. It synthesizes evidence that these systems-level proteomic signatures can guide biomarker discovery and cell-type-specific investigation.

Key findings

  • Six core, highly conserved modules are reported across AsymAD and AD cohorts: three consistently increased in AD brain network proteome (inflammatory, myelination, RNA binding/splicing) and three consistently decreased (synaptic, mitochondrial, cytoskeleton).
  • Protein and transcriptomic networks are only partly overlapping: only 30–40% of AD brain network proteome modules overlap with transcriptomic modules, differential protein expression in overlapping modules is reasonably concordant (R~0.5), and only half of disease-related variance in the proteome is reflected at the transcriptome level.
  • TMT-MS quantified 6,533 proteins across 47 brain tissues compared with 2,736 proteins by LFQ-MS on the same samples; a deep DLPFC proteome quantified >8,000 proteins and yielded 44 modules, and a consensus AD brain network was derived from hundreds of brain samples.
  • In biomarker translation, 15 of 44 brain modules strongly overlapped the CSF proteome, collectively harboring nearly 300 proteins altered in AD CSF; these were segregated into five systems-based panels and validated in >500 CSF samples, narrowing to 60 proteins, while ~20 proteins from the MSN module showed significant elevations in AD CSF across two cohorts totaling nearly 400 spinal fluid samples.

Methods

  • Narrative review, not a primary research study; it synthesizes network-based proteomic analyses of postmortem AD brain, CSF, and plasma and discusses their translational applications.
  • Describes bottom-up mass spectrometry workflows including DDA/LFQ, isobaric labeling (TMT/iTRAQ), DIA/SWATH, and targeted SRM/MRM/PRM, plus network construction with WGCNA and module enrichment for cell types, pathways, and GWAS risk factors.
  • Data discussed include AMP-AD efforts leveraging over 2,000 brain tissues and multiple AD cohorts spanning control, AsymAD, and AD cases across regions such as DLPFC, precuneus, and temporal cortex, with integration to CSF and plasma proteomes.

Limitations

  • Paper-stated: bulk human brain proteomics is limited by inter- and intra-regional variability, tissue heterogeneity, averaging of cellular microenvironments, reference cell-type data derived from healthy brain that may not reflect disease, and an inability to directly probe disease mechanisms.
  • Paper-stated: local/cell-type proteomic approaches have limitations summarized in Table 2, including lower proteome coverage and possible contamination for LCM, inability to isolate intact neurons for MACS/FACS, and potential cell activation or shear stress during sorting.
  • Evident from the text: as a narrative review, it presents no new primary data or systematic meta-analysis, so its conclusions depend on the cited studies and the authors’ consensus interpretation.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1038/s41386-020-00840-3

Cites

209 of 209 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Fig. 1 Unbiased network-based proteomics to characterize the complex biochemical and cellular endophenotypes of Alzheimer’s disease (AD). Traditional “bottom-up” mass spectrometry (MS)-based techniques, such as label-free quantitation (LFQ) or isobaric labeling, are used for protein identification and quantification in a large cohort of postmortem brain tissues from control, asymptomatic AD (AsymAD), and AD cases. Following fractionation, the deep discovery proteomic data generated from these large cohorts are organized into biologically meaningful groups, or modules, of proteins with sophisticated analytical methods, such as weighted gene correlation network analysis (WGCNA). The co-expression modules are assessed for enrichment with specific cell types, organelles, biological pathways, and genetic risk factors generated from genome-wide association studies (GWAS). In addition, abundance changes at the module level can be correlated with disease status, clinical features, and neuropathological measures. Using this framework, six core and highly conserved modules across AsymAD and AD cohorts with reproducible links to specific cell types, organelles, and biological functions have been identified. Three of the modules are consistently increased in the AD brain network proteome: inflammatory, myelination, and RNA binding/splicing, while the remaining three are consistently decreased: synaptic, mitochondrial, and cytoskeleton.
Fig. 1 Unbiased network-based proteomics to characterize the complex biochemical and cellular endophenotypes of Alzheimer’s disease (AD). Traditional “bottom-up” mass spectrometry (MS)-based techniques, such as label-free quantitation (LFQ) or isobaric labeling, are used for protein identification and quantification in a large cohort of postmortem brain tissues from control, asymptomatic AD (AsymAD), and AD cases. Following fractionation, the deep discovery proteomic data generated from these large cohorts are organized into biologically meaningful groups, or modules, of proteins with sophisticated analytical methods, such as weighted gene correlation network analysis (WGCNA). The co-expression modules are assessed for enrichment with specific cell types, organelles, biological pathways, and genetic risk factors generated from genome-wide association studies (GWAS). In addition, abundance changes at the module level can be correlated with disease status, clinical features, and neuropathological measures. Using this framework, six core and highly conserved modules across AsymAD and AD cohorts with reproducible links to specific cell types, organelles, and biological functions have been identified. Three of the modules are consistently increased in the AD brain network proteome: inflammatory, myelination, and RNA binding/splicing, while the remaining three are consistently decreased: synaptic, mitochondrial, and cytoskeleton.CC BY · PubMed Central
Fig. 2 Local proteomics approaches to define the cellular basis of Alzheimer’s disease. Summary of approaches: Laser capture microdissection (LCM) is a method that can procure subpopulations of cells or very small regions of interest under direct microscopic visualization. LCM (pink panel) has been used to micro-dissect neurons from both fresh frozen human brain and formalin-fixed paraffin embedded tissue sections for proteomic analyses. Prior to dissection, immunohistochemical or histological staining of fixed tissue sections is performed to identify a specific cell type or population of cells in a region of interest without compromising protein quality. For acute isolation of brain cell types (green panel), a fresh brain sample needs to be processed to yield a single-cell suspension which is then subjected to magnetically-activated cell sorting (MACS) or fluorescent-activated cell sorting (FACS). For MACS, the desired cell type is labeled with a 50 nm magnetic microbead conjugated to an antibody specific to cell-surface receptor. After incubation, the sample is placed on a magnet to drain unbound cells and retain desired cell type within the column. Once the column is removed from the magnet, the bound cells are released and collected for downstream analyses. Analogous to MACS, in FACS, the single-cell suspension is incubated with a fluorophore-conjugated antibody specific to a cell-surface receptor. Subsequently, the desired cell type is sorted based on their size and fluorescent signal directly into a buffer amenable for downstream proteomic analyses. In vivo biorthogonal amino acid tagging (BONCAT) of proteins (purple panel) is achieved by expressing MetRS* under a cell type-specific promoter in a mouse (Camk2a-Cre-MetRS*). MetRS* harbors a mutation (L247G) in the amino acid binding site which preferentially tags nascent proteins with an azide-tagged methionine analog, azidonorleucine (Anl). After treating the mice with tamoxifen (Tmx) to facilitate Cre-mediated recombination, Camk2a cells express MetRS* and nascent proteins are tagged with Anl. The azide residue of Anl is amenable to copper-catalyzed azide-alkyne cycloaddition or “click” chemistry. Following protein extraction, Anl-tagged proteins residues are “clicked” with a PEG-biotin-alkyne and then purified using avidin beads or avidin resin for subsequent MS analyses. Proximity labeling (orange panel) is achieved by various enzymes that biotinylate proximal endogenous proteins. The BioID approach uses the Escherichia coli biotin ligase, BirA*, with a catalytic site mutation (R118G). The mutation destabilizes the enzyme, facilitating active biotin molecules (biotinoyl-5′-AMP) to dissociate and bind to primary amines of exposed lysine residues on adjacent proteins. APEX catalyzes the oxidation of biotin-phenol to the short-lived (<1 ms) biotin-phenoxyl radical in the presence of hydrogen peroxide, which then reacts with electron-rich amino acids, such as tyrosine, in neighboring proteins. TurboID was developed by taking advantage of yeast display-based directed evolution. TurboID retains the promiscuous biotinylation property of BioID, but rapidly labels proteins in 10 min compared to the 18–24 hrs by BioID. Subsequent to biotin labeling and protein extraction from a sample, proteins can be affinity captured by streptavidin beads or matrices for downstream proteomic analyses by MS.
Fig. 2 Local proteomics approaches to define the cellular basis of Alzheimer’s disease. Summary of approaches: Laser capture microdissection (LCM) is a method that can procure subpopulations of cells or very small regions of interest under direct microscopic visualization. LCM (pink panel) has been used to micro-dissect neurons from both fresh frozen human brain and formalin-fixed paraffin embedded tissue sections for proteomic analyses. Prior to dissection, immunohistochemical or histological staining of fixed tissue sections is performed to identify a specific cell type or population of cells in a region of interest without compromising protein quality. For acute isolation of brain cell types (green panel), a fresh brain sample needs to be processed to yield a single-cell suspension which is then subjected to magnetically-activated cell sorting (MACS) or fluorescent-activated cell sorting (FACS). For MACS, the desired cell type is labeled with a 50 nm magnetic microbead conjugated to an antibody specific to cell-surface receptor. After incubation, the sample is placed on a magnet to drain unbound cells and retain desired cell type within the column. Once the column is removed from the magnet, the bound cells are released and collected for downstream analyses. Analogous to MACS, in FACS, the single-cell suspension is incubated with a fluorophore-conjugated antibody specific to a cell-surface receptor. Subsequently, the desired cell type is sorted based on their size and fluorescent signal directly into a buffer amenable for downstream proteomic analyses. In vivo biorthogonal amino acid tagging (BONCAT) of proteins (purple panel) is achieved by expressing MetRS* under a cell type-specific promoter in a mouse (Camk2a-Cre-MetRS*). MetRS* harbors a mutation (L247G) in the amino acid binding site which preferentially tags nascent proteins with an azide-tagged methionine analog, azidonorleucine (Anl). After treating the mice with tamoxifen (Tmx) to facilitate Cre-mediated recombination, Camk2a cells express MetRS* and nascent proteins are tagged with Anl. The azide residue of Anl is amenable to copper-catalyzed azide-alkyne cycloaddition or “click” chemistry. Following protein extraction, Anl-tagged proteins residues are “clicked” with a PEG-biotin-alkyne and then purified using avidin beads or avidin resin for subsequent MS analyses. Proximity labeling (orange panel) is achieved by various enzymes that biotinylate proximal endogenous proteins. The BioID approach uses the Escherichia coli biotin ligase, BirA*, with a catalytic site mutation (R118G). The mutation destabilizes the enzyme, facilitating active biotin molecules (biotinoyl-5′-AMP) to dissociate and bind to primary amines of exposed lysine residues on adjacent proteins. APEX catalyzes the oxidation of biotin-phenol to the short-lived (<1 ms) biotin-phenoxyl radical in the presence of hydrogen peroxide, which then reacts with electron-rich amino acids, such as tyrosine, in neighboring proteins. TurboID was developed by taking advantage of yeast display-based directed evolution. TurboID retains the promiscuous biotinylation property of BioID, but rapidly labels proteins in 10 min compared to the 18–24 hrs by BioID. Subsequent to biotin labeling and protein extraction from a sample, proteins can be affinity captured by streptavidin beads or matrices for downstream proteomic analyses by MS.CC BY · PubMed Central
Fig. 3 Conceptual framework for translating brain protein networks into clinical biomarkers. In this framework, multidimensional discovery-driven proteomics data collected from local cell type-specific approaches and antemortem biofluids will be integrated with the AD brain network proteome to identify systems-based panels of promising CSF and/or plasma biomarkers. Target prioritization will rest on the significance, magnitude, reproducibility, and ease of detection of the candidate biomarker in disease. Prioritized targets will also require links to disease mechanisms, informed in part by localized and cell type-specific proteomics. These network-based biomarker panels will then be validated using targeted quantitation approaches, including mass spectrometry (MS) and immunoassays. Both validation methods offer highly sensitive and accurate quantitation, though MS may offer certain advantages, such as highly selective target detection using unique peptides and the ability to cost-effectively analyze large panels of proteins in an initial verification phase prior to more expensive validation efforts. Validated biomarker panels representing a wide range of pathophysiologies could serve a variety of clinical uses, including preclinical profiling, disease monitoring, measuring therapeutic response, and confirming target engagement.
Fig. 3 Conceptual framework for translating brain protein networks into clinical biomarkers. In this framework, multidimensional discovery-driven proteomics data collected from local cell type-specific approaches and antemortem biofluids will be integrated with the AD brain network proteome to identify systems-based panels of promising CSF and/or plasma biomarkers. Target prioritization will rest on the significance, magnitude, reproducibility, and ease of detection of the candidate biomarker in disease. Prioritized targets will also require links to disease mechanisms, informed in part by localized and cell type-specific proteomics. These network-based biomarker panels will then be validated using targeted quantitation approaches, including mass spectrometry (MS) and immunoassays. Both validation methods offer highly sensitive and accurate quantitation, though MS may offer certain advantages, such as highly selective target detection using unique peptides and the ability to cost-effectively analyze large panels of proteins in an initial verification phase prior to more expensive validation efforts. Validated biomarker panels representing a wide range of pathophysiologies could serve a variety of clinical uses, including preclinical profiling, disease monitoring, measuring therapeutic response, and confirming target engagement.CC BY · PubMed Central

Verification

Claims the paper makes

157 Not checked yet · 43 Own interpretation · 18 No basis given · 4 Own result · 1 Supported so far · 1 Own method

Verify all 280 references — free account

Show all 141 claims

  • It may state two separate claims as one. 57
  • Probably not a substantial claim. 39
83 of 83
  • The paper relies on this without support.

    Protein co-expression modules reflect various molecular, cellular, and circuit-level phenotypes.

    “This approach organizes complex proteomic data into unbiased groups or “modules” of protein co-expression that reflect various molecular, cellular, and circuit-level phenotypes.”
  • The paper relies on this without support.

    Modules enriched with AD risk factors reflect causal, rather than reactive, disease mechanisms.

    “Modules enriched with AD risk factors are considered reflective of causal, rather than reactive, disease mechanisms.”
2 of 2

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

1 of 1

Citations

Cites

Items 201–209 of 209

Research works printed but not linked: 1

  1. Rayaprolu S, Gao T, Xiao H, Ramesha S, Duong DM, Dammer EB, et al. Quantitative multiplexed proteomics of mouse microglia by flow-cytometric sorting reveals a core set of highly-abundant microglial proteins. bioRxiv. https://doi.org/10.1101/802694. · DOI · Paper page

Cited by

Sorted by citations

See all 122

Metadata

Authors
Rayaprolu, S., Higginbotham, L., Bagchi, P., Watson, C. M., Zhang, T., Levey, A. I., Rangaraju, S., & Seyfried, N. T.
Affiliations
Emory University (US)
Corresponding author
Seyfried, Nicholas T.
Venue
Neuropsychopharmacology
Published
2020-09-08
Volume
46
Issue
1
Pages
98-115
ISSN
  • 0893-133X (ISSN-L)
  • 1740-634X
OpenAlex
W3084203681
Cited by
118
References
225
Citations by year
2026: 9 · 2025: 17 · 2024: 30 · 2023: 25 · 2022: 22 · 2021: 13 · 2020: 2
FWCI
2.01
Type
journal article
Language
en
Open access
Locations
  • Neuropsychopharmacology (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
Funders
  • Alzheimer's Association
  • Weston Brain Institute
  • Versus Arthritis (11060)
  • National Institute on Aging (RF1AG057470, R01AG057911, U01AG061357, RF1AG062181, U01AG046161, F32AG064862, RF1AG057471, P50AG025688, R01AG053960, R01AG061800, P30AG066511)
  • National Institute of Neurological Disorders and Stroke (K08NS099474, R01NS114130, P30NS055077)
Topics
  • Life Sciences › Neuroscience › Cellular and Molecular Neuroscience › Alzheimer's disease research and treatments
  • Health Sciences › Medicine › Psychiatry and Mental health › Dementia and Cognitive Impairment Research
  • Physical Sciences › Chemistry › Spectroscopy › Advanced Proteomics Techniques and Applications
Keywords
Alzheimer's disease, amyloid, tau protein, protein biomarkers, brain proteome, systems biology
MeSH terms
Alzheimer Disease, Brain, Humans, Biomarkers, Amyloid beta-Peptides, tau Proteins, Proteomics

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review