Zhu, Y., Weiss, T., Zhang, Q., Sun, R., Wang, B., Yi, X., Wu, Z., Gao, H., Cai, X., Ruan, G., Zhu, T., Pockley, A. G., Witzens-Harig, M., Wild, P. J., Aebersold, R., & Guo, T. (2019). High-throughput proteomic analysis of FFPE tissue samples facilitates tumor stratification. Molecular Oncology, 13(11), 2305–2328.

Zhu et al. (2019) — High-Throughput Proteomic Analysis of FFPE Tissue

Bibliographic Reference

Zhu, Y., Weiss, T., Zhang, Q., Sun, R., Wang, B., Yi, X., Wu, Z., Gao, H., Cai, X., Ruan, G., Zhu, T., Pockley, A. G., Witzens-Harig, M., Wild, P. J., Aebersold, R., & Guo, T. (2019). High-throughput proteomic analysis of FFPE tissue samples facilitates tumor stratification. Molecular Oncology, 13(11), 2305–2328. DOI: 10.1002/1878-0261.12570.

Core Argument

FFPE tissue proteomes, when accessed by an optimized PCT-SWATH (pressure cycling technology + sequential window acquisition of all theoretical mass spectra) workflow, produce proteome maps that are highly comparable to fresh-frozen (FF) counterparts and — critically for biomarker discovery — may be superior to FF samples because formalin cross-linking preserves protein integrity during long-term storage. This inverts the established narrative from DNA/RNA analysis where FFPE is universally detrimental: for proteomics, FFPE is not a degraded substrate but a stabilized one. The authors demonstrate this with the first systematic comparison of FFPE vs. FF proteomes from prostate cancer (PCa) and diffuse large B-cell lymphoma (DLBCL) samples, showing temporal stability across 1–15 years of storage and successful tissue stratification.

Methods

• PCT-SWATH workflow: Acidic hydrolysis (0.1% formic acid, 30 min) to reverse C–O methylol cross-links → alkaline hydrolysis (0.1 M Tris/HCl pH 10.0, 95°C, 30 min) to reverse C–N methylene cross-links → PCT-assisted protein extraction and digestion → SWATH-MS data-independent acquisition
• FFPE vs. FF comparison: 24 PCa patients with matched FFPE and FF tissue from radical prostatectomy (ProCOC cohort). FFPE: 3 replicate punches per sample (~900 µg dry mass total). FF: 1 punch per sample (~800 µg wet mass). Samples stored 4–8 years prior to analysis.
• Temporal stability: FFPE tissue microarrays (TMAs) from PCa patients stored 1–15 years
• Clinical application: PCa stratification (benign vs. tumor; Gleason grade groups) and DLBCL subtyping (PCNSL vs. eDLBCL vs. IVL)
• Validation: Independent PCa cohort (n=34) for biomarker confirmation by IHC
• LC-SWATH optimization: 24 conditions tested (3 gradient lengths × 8 window configurations) in duplicate

Key Findings

• FFPE and FF proteome maps are highly similar. The first systematic comparison of FFPE vs. FF tissue proteomes from the same patients demonstrated high concordance. Strongly overlapping protein sets were identified from both preservation methods, with highly correlated abundance patterns. Multiple proteins consistently regulated in PCa were confirmed in an independent cohort.

• FFPE outperforms FF for biomarker discovery. Counter to the universal narrative from DNA/RNA analysis, the authors report that FFPE samples provide superior proteomic data for biomarker discovery compared to FF samples. The mechanism: formalin cross-linking stabilizes proteins against degradation during long-term storage, while FF samples suffer continuous proteolytic degradation even at −80°C. The de-cross-linking chemistry developed in this study (acidic + alkaline hydrolysis) effectively reverses formalin-induced modifications, recovering intact, analyzable proteins.

• Proteome patterns are temporally stable over 1–15 years. FFPE tissue microarrays stored for 1–15 years showed no significant difference in proteome patterns — the protein signatures were stable across the entire storage range. This is in stark contrast to FFPE DNA, which degrades progressively with storage time (increased fragmentation, lower amplifiable template counts).

• PCT-SWATH enables tissue-sparing analysis from biopsies. The optimized workflow processed tissue punches mimicking needle biopsies (width <1 mm, length ~2–3 mm, ~300–400 µg dry mass), generating ~60 µg peptide per mg FFPE tissue — comparable to yields from FF wet tissue (~50 µg/mg). This enables proteomic analysis from the small tissue quantities typical of clinical biopsy archives.

• Successful stratification of two cancer types. In PCa, the FFPE proteome distinguished benign from tumor tissue and stratified Gleason grade groups. In DLBCL, the proteome classified clinically relevant subtypes (primary CNS lymphoma vs. extracerebral DLBCL vs. intravascular lymphoma) that are morphologically indistinguishable but therapeutically distinct.

Concepts Introduced or Used

FFPE proteomics, PCT-SWATH, pressure cycling technology, data-independent acquisition (DIA), formalin de-cross-linking, acidic hydrolysis, alkaline hydrolysis, protein stability in FFPE, temporal proteome stability, tissue stratification, biomarker discovery

Entities Referenced

• ProCOC cohort — prospective prostate cancer cohort (Zurich)
• SWATH-MS — sequential window acquisition of all theoretical mass spectra
• PCT — pressure cycling technology for protein extraction
• Prostate cancer (PCa), diffuse large B-cell lymphoma (DLBCL)
• CRYAB, DCN, DES, GOLM1, MAP1A, MDH2, MPO, SORD, UCHL1 — differentially regulated proteins

Limitations (as stated by authors)

• The FFPE vs. FF comparison used samples stored 4–8 years; results may differ for freshly collected specimens where FF proteins haven’t yet degraded
• Sample sizes were modest (24 PCa patients for FFPE/FF comparison; TMA validation on independent cohort)
• The PCT-SWATH workflow requires specialized equipment (Barocycler for PCT) not available in all proteomics facilities
• De-cross-linking chemistry may not completely reverse all formalin modifications — some protein modifications may persist and affect specific peptides
• Limited to two cancer types; generalizability to other tumor types not demonstrated

Relevance to Clonal Evolution

This paper forces a re-evaluation of the FFPE-as-degraded-substrate narrative established by the DNA-focused literature. Its relevance to clonal evolution is threefold:

1. Analyte-specific FFPE bias. Greytak et al. (2015) and Do & Dobrovic (2015) documented that FFPE degrades DNA and RNA, introducing sequence artifacts and expression biases. Zhu et al. demonstrate that the opposite is true for proteins: formalin cross-linking preserves proteins from the degradation that affects FF samples even at −80°C. This means that FFPE introduces analyte-dependent biases — detrimental for nucleic acid analysis, neutral-to-beneficial for proteomic analysis. For multi-omic clonal evolution studies that integrate DNA, RNA, and protein data from the same FFPE block, the biases operate in opposite directions for different analytes.

2. Protein-level ITH is accessible from FFPE archives. The wiki’s framework recognizes that intratumor-heterogeneity has both genetic and non-genetic dimensions (transcriptional, phenotypic). Zhu et al. demonstrate that the phenotypic dimension — the proteomic state of tumor cells — can be recovered from FFPE archives with fidelity comparable to or better than FF samples. This opens the possibility of retrospective proteomic ITH analysis from archival cohorts where only FFPE tissue exists.

3. The pre-analytical variable hierarchy is analyte-specific. The wiki now documents a rank order of FFPE impact: DNA > RNA >> protein. DNA suffers the most severe artifacts (C→T deamination, fragmentation, crosslinking). RNA is intermediate — miRNA is more stable than mRNA, but both show significant preservation-dependent biases. Protein is the most stable analyte in FFPE — formalin fixation, far from being a problem, is actually a preservation strategy for proteins. This hierarchy has practical implications: when only FFPE tissue is available (the majority of clinical cohorts), proteomic measurements are more trustworthy than genomic measurements, and within genomic measurements, higher-coverage NGS with UDG pretreatment and molecular tagging is essential to mitigate artifacts.