MDM2 promoter P1/P2 switching and colorectal cancer lineage plasticity: deep-learning morphology classification at 98.5% accuracy, with higher P2 index in TP53 wild-type tumors and greater Nutlin-3a sensitivity

MDM2 dual-promoter P1/P2 usage co-varies with the lineage identity of colorectal cancer cells and with the secretory, mucin-rich character of the tumor tissue, consistent with a molecular-switch role alongside p53 suppression. The correspondence between MDM2 promoter usage, colorectal cancer lineage plasticity, and tissue morphology had not been systematically characterized; this work links promoter choice to morphological phenotype and consensus molecular subtypes. Based on 63 organoid samples from 22 patients, external validation in TCGA-COAD/READ (n=624) and GSE39582 (n=536) totaling 1,160 cases, and an MDM2 Splicing Index derived from expression arrays; evidence comes from the multi-cohort validation design described in the abstract.

Deep-learning (VGG16) morphological classification reached 98.5% test accuracy (64/65), and morphology corresponded to P1/P2 isoform usage: Type1 compact glandular morphology predominated in P1-dominant samples (median Type1 fraction 0.826 versus 0.444), while non-Type1 cystic mucinous morphology corresponded to P2 dominance (AUC 0.79). The study directly maps automated morphological classification onto promoter usage as a molecular feature, providing quantitative evidence for a morphology-molecule association. Test accuracy of 64/65; the morphology-promoter association is reported as a median fraction comparison and an AUC of 0.79, in samples from 63 organoids and 22 patients.

The P2 signature was higher in mismatch repair deficient (microsatellite-unstable) tumors, both signatures differed across the four consensus molecular subtypes, and the directly quantified P2_index was higher in TP53 wild-type tumors, consistent with P2 being p53-inducible. It aligns promoter usage with three existing molecular classification axes: CMS subtype, MSI status, and TP53 mutation status, extending the classification meaning of the P1/P2 ratio. Based on external validation in TCGA-COAD/READ and GSE39582 totaling 1,160 cases and on comparison of P2_index with TP53 status; the abstract does not report specific effect sizes or statistics.

TP53 wild-type cell lines were more sensitive to the MDM2 inhibitor Nutlin-3a and more dependent on MDM2 in the DepMap CRISPR screen; among 198 GDSC2 drugs, Nutlin-3a correlated most strongly with the P2 score; and in the independent biobank, TP53 wild-type lines (17) were more sensitive to nutlin-3 than mutant lines (48) (median log(IC50) 1.386 versus 4.283). It moves promoter usage from a classification marker toward a therapeutic-stratification hypothesis and provides a quantitative sensitivity contrast in an independent patient-derived organoid biobank. From the GDSC2 and DepMap public cell line panels and 65 lines of an independent patient-derived organoid biobank; in the independent biobank, 17 TP53 wild-type versus 48 mutant lines, median log(IC50) 1.386 versus 4.283.

The results are aimed at molecular classification and therapeutic stratification research in colorectal cancer, applicable in laboratory and translational settings with organoid or expression-profiling resources; the P1/P2 ratio is described as measurable by RT-qPCR or targeted NGS, so its potential application setting is biopsy-level classification and stratification. The reported morphology-molecule correspondence, CMS/MSI/TP53 associations, and Nutlin-3a sensitivity differences rest on the organoid cohort, public cohorts, and cell line panels described in the abstract, and can serve as a starting point for testing in samples and workflows closer to clinical decision-making.

This reading is incomplete, based only on the abstract and the competing interest statement; the main text, figures, and supplementary materials were not included, so statistics, confidence intervals, multiple-testing correction, and sample composition details for each comparison could not be verified. The abstract gives no specific effect sizes for the P2_index versus TP53 status comparison or for the differences of both signatures across the four consensus molecular subtypes, and the strength of the Nutlin-3a correlation with the P2 score among 198 GDSC2 drugs is not quantified. In addition, the clinical utility of the MDM2 P1/P2 ratio as a biomarker still needs testing in independent, prospective patient cohorts, and the causal direction of its relationship with p53 pathway suppression remains an open question.