henhao
Sign in
Systems BiologyCell Death and Differentiation
Peer-reviewed journal article

Pan-cancer proteogenomic interrogation of the Ubiquitin Proteasome System

Tania J González-RoblesiD, Maha Khan, Paul Sastourne-HaletouiD, Marisa Triola, Hua ZhouiD, Yuki KitoiD, Sharon KaisariiD, David FenyöiD, Gergely RónaiD, Yadira M. Soto‐FelicianoiD, Benjamin G. NeeliD, Kelly V. RugglesiD, Michele PaganoiD

Abstract

Somatic mutations rewire the ubiquitin-proteasome system (UPS) to support tumor growth, but the proteome-wide consequences of cancer-driver alterations on UPS composition remain incompletely understood. Using harmonized proteogenomic data from up to 11 CPTAC cohorts, we performed an integrated pan-cancer analysis of UPS protein dysregulation, prognostic associations, and mutation-driven remodeling. We show that mRNA poorly predicts UPS protein abundance, that a defined set of E3 ligases is recurrently dysregulated across cancers, and that somatic mutations (most strikingly TP53 loss) produce coherent UPS protein-quantitative trait locus (pQTL) signatures. Two case studies (UBR5 and TRIM28) illustrate orthogonal modes of UPS rewiring: a mutation-driven axis in which TP53-mutant tumors elevate UBR5 to support replication stress tolerance, and a lineage-driven axis in which TRIM28 engages tissue-restricted regulatory networks with opposing prognostic effects in glioblastoma versus head and neck cancer. Each axis exposes context-specific therapeutic vulnerabilities, including sensitivity to DNA damage response inhibitors (UBR5-high) and lineage-specific drug responses (TRIM28-high). Together, these analyses define a mechanistic framework for how cancer-driver mutations reshape proteostasis through the UPS and nominate mutation- and lineage-defined dependencies for precision degrader therapy. The harmonized pan-tissue atlas and the UbiDash interactive resource that underpin parts of this analysis are reported in our companion paper1. HighlightsO_LIProteome-centric profiling reveals UPS dysregulation invisible at the transcript level across human cancers C_LIO_LIPan-cancer differential expression identifies recurrently dysregulated E3 ligases with prognostic value C_LIO_LIGenome-wide pQTL analysis uncovers mutation-driven UPS remodeling, with TP53 loss as a dominant driver C_LIO_LIUBR5 and TRIM28 define orthogonal mutation- and lineage-driven axes of UPS rewiring with distinct therapeutic vulnerabilities C_LI

PDFDOICodeData
Full-text digest ↓

02 /Specification & Classification

Click any tag to browse matching papers
Citations
3
FWCI
6.24
Field percentile
top 3%
OpenAlex Topics

In-depth summary

✦ Multi-agent verified full-text synthesis

Claude Opus 5.5 · xhigh effort · 8 Nov 2026 · from the full text of version 2 (10 May 2026) · CC BY 4.0

Proteogenomic analysis of up to 11 CPTAC cohorts shows ubiquitin-proteasome proteins often diverge from transcripts; TP53-mutant tumors elevate UBR5 protein, while TRIM28 prognosis is lineage-dependent.

Ubiquitin-proteasome system (UPS) components make up nearly 20% of annotated cancer-driver genes, yet many pan-cancer UPS studies have used transcript readouts that incompletely capture E3 ligase abundance and substrate engagement. This work maps UPS protein dysregulation, survival links, and mutation-associated protein QTLs across CPTAC cohorts, separating mutation-driven (UBR5) from lineage-driven (TRIM28) E3 remodeling relevant to degrader design.

Key results

  1. ShownRecurrent somatic mutations broadly reshape UPS protein abundance, and TP53 point mutation yields the dominant signature: higher S-phase and mitotic drivers alongside loss of TP53-induced E3s tied to DNA repair.

    9,495 mutations present in at least 11 tumors across 10 CPTAC cohorts yielded 17,925 UPS protein-mutation associations (FDR < 10%) in purity- and cohort-adjusted linear models; TP53 point mutants altered 111 proteins (FDR < 0.05), raising CDT2 (DTL), CDC20 and PLK1 and lowering DDB2 and FBXO22. · Cancer-associated mutations reshape the UPS…; UPS protein–mutation quantitative trait locus…

  2. ShownTumor-versus-normal proteomics reveals selective, cohort-variable UPS dysregulation, with some E3s recurrently shifted in four or more cancer types and others, such as KEAP1, changing direction by tissue.

    Of 1,287 UPS proteins (670 E3s), 1,025 (562 unique) UPS and 503 (295 unique) E3 calls at |LFC| > 0.5, p-adj < 0.05; range 33 in COAD (2.6%) to 357 in GBM (27.7%); UPS changed less than proteome (Fisher odds ratio < 1, p-adj < 0.0001); PRKN, RNF123 down, SKP2, CDT2 up; KEAP1 LFC = −0.2 in LUAD vs 0.4 in LSCC. · Differential protein expression reveals UPS…; Tumor vs. normal-adjacent tissue comparisons

  3. ShownUPS protein abundance stratifies overall survival mainly in a tumor-type-specific manner; few E3s predict outcome pan-cancer, and TRIM28 shows opposite prognostic direction in different tissues.

    Quartile split of 922 UPS proteins (411 E3s) in 11 cohorts: 153 E3s tied to survival in at least one cohort (log-rank p < 0.05); pan-cancer, high FBXL18 worse and FBXL3 better; high FBXO33 and UBR5 adverse in LUAD and PDAC; high TRIM28 favorable in LUAD (p = 0.031), trending in GBM (p = 0.059), adverse in HNSCC. · UPS protein expression levels stratify cancer…; UPS stratification and survival analysis

  4. SuggestedUBR5 and TRIM28 exemplify orthogonal UPS remodeling: UBR5 rises in TP53-mutant tumors, often post-transcriptionally, alongside replication and DNA-repair programs, whereas TRIM28 follows lineage, with distinct co-dependencies and drug sensitivities in CNS versus head and neck lines.

    UBR5 up in TP53-mutant tumors in seven cohorts (Wilcoxon P < 0.05), protein-only in 4 of 7, with higher protein-to-mRNA ratios; UBR5-high lines more sensitive to JNJ-64619178 (PRMT5) and MIW-815 (STING). TRIM28 lacks mutation links; TRIM28-high CNS lines sensitive to SRT2104 and VX-984, HNSCC lines to ibutamoren and belizatinib. · Cancer-associated mutations reshape the UPS…; UBR5 defines a mutation-associated UPS…; TRIM28 defines a lineage-specific UPS remodeling…

  5. SuggestedMatched tumor mRNA and protein levels are often discordant, and UPS components (like transcription factors and signaling proteins) frequently show weak or inverse concordance, supporting protein-level analysis.

    Gene-wise mRNA-protein Spearman correlations were mostly positive but included many negative values; least concordant genes (ρ < −0.2; P < 0.01) were enriched for post-transcriptional processes such as translation efficiency and protein half-life, whereas metabolic and housekeeping genes were most concordant. · Divergence of RNA–protein establishes the need…

Methods, data and reproducibility

Secondary analysis of CPTAC proteogenomic data: DEqMS/limma tumor-versus-normal testing, quartile-based Kaplan-Meier and Cox survival models, and purity- and cohort-adjusted linear mutation-to-protein QTL (pQTL) models, followed by GSEA, DepMap co-dependency, and PRISM drug-response correlations for UBR5 and TRIM28.

  • CPTAC pan-cancer proteogenomic cohorts: tumor/normal proteomes, transcriptomes, somatic mutations, and clinical annotations · human · n = up to 11 cohorts · dbGaP phs001287.v16.p6
  • Tumor vs normal-adjacent differential protein expression subset (eight matched cohorts plus one unmatched GBM cohort) · human · n = nine cohorts; 1,287 UPS proteins (670 E3s)
  • Survival subset with overall survival time and vital status from CPTAC annotations · human · n = 11 tumor types; 922 UPS proteins (411 E3s)
  • Mutation-to-protein pQTL subset restricted to mutations seen in at least 11 tumors · human · n = 10 cohorts; 9,495 of 19,932 mutations
  • Proteostasis Network UPS gene annotation (downloaded June 22, 2024) · n = 1,286 UPS genes
  • DepMap Public 25Q2 CRISPR-Cas9 gene-effect scores for CNS/Brain and Head and Neck lineages · human (DepMap cancer cell lines) · n = 89 CNS/Brain; 77 Head and Neck cell lines
  • PRISM Repurposing drug-response (viability AUC) screens, analyzed in Head and Neck and CNS/Brain lines · human (cancer cell lines) · n = 578 cell lines; 4,518 compounds
  • CCLE cell line protein (or mRNA when protein missing) abundance matched to PRISM lines · human (cancer cell lines)
Methods & tools
DEqMS v1.27.1 (spectraCounteBayes), limma, survival and survminer R packages, Cox proportional-hazards models (coxph), fgsea v1.35.6, MSigDB via msigdbr v25.1.1, Hmisc::rcorr v5.2.3, ESTIMATE tumor purity, DepMap CRISPR Chronos gene effects (Public 25Q2), PRISM Repurposing drug screens, BioGRID interaction annotations, UbiDash R Shiny app
Code
https://github.com/ruggleslab/ubiDashhttps://ruggleslab.shinyapps.io/UbiDash/
Data & accessions
https://pdc.cancer.gov/pdc/cptac-pancancerhttps://portal.gdc.cancer.gov/https://dataservice.datacommons.cancer.gov/https://www.proteostasisconsortium.com/pn-annotation/https://www.depmap.org/
Software & environment
R v4.2.2 (survival analyses), R v4.5.1 (co-dependency visualization); DEqMS v1.27.1, fgsea v1.35.6, msigdbr v25.1.1, Hmisc v5.2.3; no container or lockfile described.
Reproducibility gaps
Controlled CPTAC genomic data require approved dbGaP access (phs001287.v16.p6); some additional data are available only from the lead contacts on request.
Limitations (6)
  • Bulk tumor proteogenomics lacks single-cell and spatial resolution, so it cannot capture cell-type-specific or microenvironment-dependent UPS remodeling, or temporal and therapy-induced dynamics. · stated by the authors
  • CPTAC skews toward female-specific tumor types (breast, endometrial, ovarian) and has few male-specific cancers like prostate adenocarcinoma, potentially restricting generalizability. · stated by the authors
  • Proteomic missing values and bias toward abundant proteins may hide low-abundance or tissue-restricted ligases. · stated by the authors
  • GBM had no matched normal-adjacent tissue, possibly inflating its differential burden; survival models ignore treatment history and disease stage; E3 auto-ubiquitylation may make protein changes partly reflect self-turnover. · stated by the authors
  • pQTL models adjust only for tumor purity and cohort, not tumor mutational burden, co-mutation, or subtype, so some of the 17,925 associations may reflect hypermutation or subtype structure rather than gene-specific effects. · noted in this summary
  • Drug associations use nominal thresholds (P < 0.05, |ρ| > 0.3) within a PRISM screen of 4,518 compounds, and no drug or dependency hit was tested experimentally, so nominated UBR5 and TRIM28 vulnerabilities remain hypotheses. · noted in this summary

Who should read:Read this if you study E3 ligases, deubiquitylases, or targeted protein degradation in cancer and want protein-level, mutation- and lineage-resolved CPTAC evidence to prioritize context-specific UPS targets.

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
THISMartínez-Ji '18Kim '26Yan '26Belal '25Klink '26Elsasser '26
63% semantic matchBioinformatics
Francisco Martínez-Jiménez, Ferran Muiños, Erika López‐Arribillaga+2 more·bioRxiv·Dec 28, 2018· 2 citations