henhaobio
Sign in

Nature CommunicationsBiochemistryPeer-reviewed journal article

Human-mouse proteomics reveals the shared pathways in Alzheimer's disease and delayed protein turnover in the amyloidome

Jay M. YarbroiD, Xian HaniD, Abhijit DasguptaiD, Ka YangiD, Danting LiuiD, Him K. ShresthaiD, Masihuz ZamaniD, Zhen WangiD, Kaiwen YuiD, Dong Geun LeeiD, David VanderwalliD, Mingming NiuiD, Huan Sun, Boer Xie, Ping‐Chung CheniD, Yun JiaoiD, Xue ZhangiD, Zhiping WuiD, Surendhar Reddy ChepyalaiD, Yingxue FuiD, Yuxin LiiD, Zuo‐Fei YuaniD, Xusheng WangiD, Suresh PoudeliD, Barbora Vagnerova, , Junmin PengiD

PDFDOIPreprintCodeData
In-depth full-text digest available · Jump to digest ↓

Abstract

Murine models of Alzheimers disease (AD) are crucial for elucidating disease mechanisms but have limitations in fully representing AD molecular complexities. We comprehensively profiled age-dependent brain proteome and phosphoproteome (n > 10,000 for both) across multiple mouse models of amyloidosis. We identified shared pathways by integrating with human metadata, and prioritized novel components by multi-omics analysis. Collectively, two commonly used models (5xFAD and APP-KI) replicate 30% of the human protein alterations; additional genetic incorporation of tau and splicing pathologies increases this similarity to 42%. We dissected the proteome-transcriptome inconsistency in AD and 5xFAD mouse brains, revealing that inconsistent proteins are enriched within amyloid plaque microenvironment (amyloidome). Determining the 5xFAD proteome turnover demonstrates that amyloid formation delays the degradation of amyloidome components, including A{beta}-binding proteins and autophagy/lysosomal proteins. Our proteomic strategy defines shared AD pathways, identify potential new targets, and underscores that protein turnover contributes to proteome-transcriptome discrepancies during AD progression.

In-depth summary

Claude Opus 5.5 · xhigh effort · 7 Nov 2026 · from the full text of version 1 (25 Oct 2024) · CC BY 4.0

Alzheimer's mouse model proteomes capture 30% (amyloid-only models) to 42% (adding tau and splicing models) of consensus human AD protein changes; in 5xFAD, plaques slow degradation of plaque-associated proteins.

Amyloidosis mouse models are central to Alzheimer's disease (AD) research, yet none reproduces the full human disease, and earlier proteomic surveys of these mice were often shallow, covered single models or compared little with human data. This work profiles deep cortical proteomes (most above 10,000 proteins) across five models (three at multiple ages), scores them against consensus human AD protein changes, and adds plaque proteomes and in vivo turnover to explain protein-RNA discordance.

Key results

  1. ShownPooling the amyloidosis models 5xFAD and NLGF recovers 30% of consensus human AD protein changes; adding 3xTG (tau) and BiG (U1 snRNP splicing) raises this to 42%, and both amyloid models resemble late AD more than mild cognitive impairment (MCI).

    Of 866 human differentially expressed proteins (DEPs) consistent across deep AD proteomic studies (654 detected in mice), 196/654 are DEPs in 5xFAD or NLGF and 275/654 across four models; each model alone shares <25%. Correlation with late AD: R = 0.32 (5xFAD), 0.23 (NLGF) vs 0.00 and -0.09 with MCI. 108 DEPs are shared by human AD, 5xFAD and NLGF. · Human-mouse comparison identifies shared AD…; Additional pathologies beyond amyloidosis in…; A comprehensive multi-layered proteomics…

  2. ShownAbout a third of altered proteins in human AD and 5xFAD brains change without a matching RNA change, and the RNA-independent increases shared by both species are found mainly in amyloid plaques.

    Protein-RNA correlation R = 0.40 (human, n = 10,781) and 0.46 (5xFAD, n = 8,840); 35% (human) and 36% (mouse) of altered proteins are RNA-independent. Of 31 shared RNA-independent upregulated proteins, 23 were detected in laser-captured plaque proteomes (5,364 proteins, 4 mice) and 17 (74%) were among 438 plaque-enriched proteins, none among 191 non-plaque-enriched. · Transcriptome-proteome inconsistency occurs in…

  3. ShownIn 5xFAD, Aβ turns over far more slowly than full-length human APP, most proteins with altered half-life become longer-lived, and 32 RNA-independent, upregulated plaque proteins (e.g., Apoe) show prolonged half-lives.

    pSILAC in 30 mice gave corrected half-lives (T50) for 8,492 proteins. hAβ vs hAPP: apparent T50 132.0 vs 40.1 d; corrected 18.2 vs <0.5 d. Of 84 proteins with changed T50, 75% were longer in 5xFAD (APOE ΔT50 4.2 d, VTN 3.5 d); Apoe <0.5 d in WT vs 8.85 d in 5xFAD. The 32 plaque proteins enrich for amyloid matrisome, autophagy/lysosome and neurogenesis. · Delayed protein degradation in amyloidome…; Molecular insights from multi-omics integration…

  4. Shown5xFAD (transgenic) and NLGF (knock-in) mice share much of an age-dependent proteome change that scales with Aβ load, while NLF changes little; their shared phosphorylation changes mostly occur without a change in protein abundance.

    Summed over ages: 1,382 DEPs in 5xFAD, 1,142 in NLGF, 7 in NLF (FDR < 0.05, |log2FC| > 2 SD); 610 of 1,914 (32%) overlap, 597 in the same direction; DEP number vs Aβ accumulation R = 0.86. Of 80 phosphoproteins (122 phosphopeptides) changed in both models, only 30 (37.5%) changed at protein level; kinase-substrate enrichment analysis infers raised CSNK, MAPK, PAK and AKT activity. · Proteome profiling of multiple AD amyloidosis…; Phosphoproteome profiling of AD models…

  5. SuggestedMost proteins altered in both the mouse models and human AD have little AD literature; rank aggregation over 12 omics datasets puts established AD genes near the top and flags understudied candidates.

    Of 275 mouse-human shared DEPs, 86% have <20 AD-related publications. Order-statistics ranking across GWAS (1), transcriptome (2), proteome (6), phosphoproteome (2) and interactome (1) datasets ranks APP, APOE, GFAP, TREM2, MAPT and CLU high, with MDK, NTN1, SFRP1, OLFML3, PTPRC/CD45, SMOC1, CD180 and PTN also in the top 20. · Additional pathologies beyond amyloidosis in…; Prioritization of proteins/genes based on…

Methods, data and reproducibility

Cortical proteomes of five AD mouse models versus age-matched wild-type mice (phosphoproteomes for two) were quantified by tandem mass tag (TMT) two-dimensional LC-MS/MS and compared with consensus human AD proteomic changes. In 5xFAD, laser-captured plaque proteomes and in vivo half-lives (pulsed heavy-lysine labeling, pSILAC, corrected for lysine recycling in JUMPt) were added.

  • Cortex proteomes of 5xFAD (3, 6, 12 months), NLF (3, 12 months), NLGF (3, 6, 12, 18 months) and age-matched wild-type mice · mouse · n = 66 brains (16 conditions)
  • Cortex proteomes of 3xTG and BiG (5xFAD x N40K) mice with wild-type controls, about 6 months old · mouse · n = 37 brains (text); legend lists 3xTG 19, WT 8, BiG 4, WT 4
  • Phosphoproteomes (TiO2 enrichment) of 5xFAD (12 months), NLGF (3-18 months) and wild-type controls · mouse · n = 36 mice
  • Laser-captured amyloid plaque and non-plaque regions from 5xFAD brain sections · mouse · n = 4 mice (8 samples)
  • Pulsed heavy-lysine (13C6) labeling time course (0, 4, 8, 16, 32 days) in 5xFAD and wild-type littermates, about 9 months old · mouse · n = 30 mice (3 replicates)
  • Consensus human AD differentially expressed proteins compiled from published deep proteomics studies · human · n = 866 proteins (654 detected in mice); more than 30 studies
  • Matched transcriptome and proteome fold changes, human AD and 5xFAD versus controls · human; mouse · n = n = 10,781 (human), n = 8,840 (5xFAD)
  • Raw mass spectrometry data deposited for this study (dataset mapping in Supplementary Table 2) · PRIDE PXD018590, PXD007974, PXD023395, PXD031545, PXD031732, PXD031734, PXD031735, PXD031769, PXD053314
Methods & tools
TMT / TMTpro labeling, two-dimensional LC-MS/MS (Q-Exactive HF Orbitrap), JUMP software suite (incl. JUMPl with phosphoRS), JUMPsilactmt v1.0.0, JUMPt v1.0.0, pulsed SILAC (13C6-lysine diet), laser-capture microdissection (Arcturus XT), MKmisc (moderated t-test), KSEAapp v0.99.0 (PhosphoSitePlus, NetworKIN), PANTHER, custom PPI database (InWeb_IM, STRING, BioPlex), Cytoscape
Software & environment
JUMPsilactmt v1.0.0 (Python), JUMPt v1.0.0, KSEAapp v0.99.0 under RStudio (v4.1.2); no container or requirements file described.
Experimental conditions
Instrument time about 4 days per TMT batch with at least 40 LC fractions per batch; computing hardware not stated.
Reproducibility gaps
No code linked for differential expression, pathway activity scoring or the order-statistics ranking; PRIDE accessions map to datasets only via Supplementary Table 2; the JUMPsilactmt repository is named Version-0.0.1 while the text cites version 1.0.0.
Limitations (4)
  • Mouse models capture only part of human AD: most lack substantial neuronal loss, each shares <25% of consensus human DEPs, and the 58% of human DEPs not covered (mitochondrial function, cell morphogenesis, lipid regulation) may reflect less neuronal death and species differences. · stated by the authors
  • Corrected half-lives are less reliable for proteins turning over very fast (<0.5 d) or very slowly (>100 d). · stated by the authors
  • Several headline half-life contrasts (hAPP corrected T50 <0.5 d; Apoe <0.5 d in WT) fall in the range the authors flag as less reliable, so their magnitudes are uncertain. · noted in this summary
  • The plaque-turnover link rests on overlapping 5xFAD datasets (plaque proteome from 4 mice) without perturbation or human turnover data; 3xTG and BiG were profiled at one age (about 6 months), BiG with 4 mice per group. · noted in this summary

Read this if you pick or interpret Alzheimer's mouse models, or want deep mouse brain proteome, phosphoproteome, plaque and in vivo turnover data to compare against human AD changes.

Citations
41
FWCI
22.89
Field percentile
top 1%

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISRayaprolu '26Boeddrich '23Levites '24Savas '17Chu '26Ball '17
74% semantic matchNeuroscience
Sruti Rayaprolu, Sneha Malepati, Eric B. Dammer+15 more·bioRxiv·Nov 6, 2026