Aggregated Evidence

 Quick Analysis Plan Copied

 Long Analysis Plan

What the supplied evidence does — and does not — establish

No supplied record contains colorectal cancer biomarker data. None of the eight provided papers report HSPA5/GRP78 protein vs mRNA concordance, CRC patient cohorts, or survival associations. A quantitative CRC meta-analysis therefore cannot be computed from these records; it requires external data (TCGA COAD/READ RNA-seq, CPTAC colorectal proteomics, matched IHC cohorts). What the records can support is the mechanistic framework and analysis plan for the discordance study.

Three sourced mechanisms that predict protein–mRNA discordance

• Transcriptional feedback is substrate-bound, not load-blind. Only BiP molecules bound to ER substrates trigger GRP78/BiP gene induction: HNT2-expressing CV-1 cells showed ~8-fold mRNA induction while non-binding mutants (HNT3, HNDT3) showed none, so mRNA reflects folding demand rather than chaperone abundance alone. BiP mRNA induction requires substrate binding
• Post-transcriptional stabilization changes protein without changing mRNA. Chaperone coexpression strongly increased E6/E7 protein levels without altering mRNA, and cycloheximide chase showed half-lives extending from ~45→119 min (E6) and ~56→154 min (E7). Chaperone-mediated protein stabilization without mRNA change This is the canonical discordance axis: mRNA quantifies synthesis, protein quantifies degradation-resistance.
• Protein functional state is invisible to mRNA. HOCl oxidation of GRP78 Lys353 abolished ZBM-H–induced ATPase increases (K353A mutant) while K352A did not, and GRP78–AMPK signaling depends on this modification. Lys353 oxidation controls GRP78 functional state Additionally, a separate plasma-membrane GRP78 interactome (10–18 proteins) governs cancer stemness independently of ER abundance. Surface GRP78 interactome drives stemness

Reported qualitative/semi-quantitative values from Fig. 7 of the 1992 study; blue bars mark constructs that co-immunoprecipitated BiP, red bars those that did not. The paired observation—binding constructs induce mRNA, non-binding do not—is the sourced signal; bar heights use midpoints of the reported ranges. BiP mRNA induction fold-changes by construct

Meta-analysis plan: protein vs mRNA discordance

• Data acquisition: TCGA COAD/READ HSPA5 TPM (matched normal), CPTAC colorectal tumor proteomics (HSPA5 iBAQ/log2 ratios), THPA/interrogated IHC cohorts with H-scores; require matched mRNA+protein from the same patient-sample where possible.
• Concordance modeling: per-patient Spearman correlation of HSPA5 mRNA vs protein; classify each patient concordant-high, discordant (RNA-high/protein-low and reverse), concordant-low; test discordance vs MSI/CIN status, CMS, tumor stage as moderators.
• Statistical aggregation: random-effects meta-analysis of Fisher-z correlations across cohorts (statsmodels/metafor-style), I² heterogeneity, leave-one-out sensitivity; preregister that IHC (semiquantitative) vs MS (isobaric, batch effects) are non-equivalent assays — this measurement heterogeneity is the primary confounder of apparent discordance.
• Mechanistic annotation: overlay UPR activity signatures (XBP1s/ATF4 targets) to test whether discordance tracks substrate-bound feedback (mRNA driver) vs protein half-life/stabilization axis, consistent with the chaperone-stabilization model above.
• Known unknowns: surface/secreted GRP78 pools are invisible to bulk transcriptomics yet functionally decisive (interactome evidence above); PTM state (Lys353-like oxidation) is unmeasured in standard proteomics; no CRC-specific record exists in the supplied evidence, so effect sizes remain unknown until external data are retrieved.