Gastric Organoid Subtypes Can Diverge from Their Parent Tumors

A Smaller Gene Panel for a Model-Fidelity Question

University of Alberta researchers evaluated a 107-gene NanoString assay for translating gastric cancer molecular classifications into tissue and organoid studies. Their Genes, Chromosomes and Cancer paper used previously developed supervised classifiers to assign The Cancer Genome Atlas, or TCGA, subtypes and tumor-microenvironment, or TME, classes. The study then asked whether a culture received the same classification as the tumor from which it originated.

The prospective single-center cohort enrolled 38 patients with treatment-naive gastric adenocarcinoma. Twelve organoid cultures were generated from a subset, with a reported establishment rate of 75% among attempted cultures. One parental specimen did not contain cancer and its culture was excluded from drug-response analysis, although it remained usable for selected reference-test comparisons.

Reference-Test Agreement Did Not Guarantee Complete Subtype Fidelity

The assay measured expression in fixed tissue and fresh organoid material. RUV-III normalization used technical replicates to address unwanted variation, and feature-specific quantile normalization mapped the measurements to the classifier distributions. These processing steps were evaluated before interpreting subtype differences.

For ten paired samples, EBV-associated and microsatellite-instability classes were checked against EBER in situ hybridization and pentaplex PCR. The NanoString assignments were 100% accurate in this reference-tested set. Comparable gold-standard validation was not available for the chromosomal-instability and genomically stable classes.

The principal comparison involved 11 tumor–organoid pairs. Categorical concordance was 81.8% for TME classes and 63.6% for TCGA classes. Some organoids shifted toward a chromosomal-instability classification, and two pairs showed TME discordance. One EBV-positive, TME-high tumor yielded an organoid dominated by a TME-low, chromosomal-instability profile. The continuous MSI score also differed between paired tumors and organoids in an unadjusted analysis, with P = 0.04.

FLOT Response Was Not Significantly Linked to Subtype

The team tested a FLOT combination comprising 5-fluorouracil, leucovorin, oxaliplatin and docetaxel. Organoid cells grew for 24 hours before 48 hours of treatment. CCK-8 viability measurements across dose dilutions supplied fitted curves and a drug sensitivity score. An AGS gastric cancer cell-line experiment separately supported assay reproducibility across three independent runs.

The Results describe testing in 11 cultures with malignancy confirmed in the parent specimen; the figure presents ten organoid dose-response profiles. No significant association was found between drug sensitivity score and the molecular subtype scores. A positive correlation with the TME-high score was exploratory and did not establish a reliable subtype-based FLOT response rule.

The study had insufficient follow-up to compare the in vitro drug responses directly with progression-free or overall survival. It assessed molecular classification and short-term chemotherapy responses without clinical outcome validation.

Two-Dimensional Models Added Hypotheses with Their Own Limits

A separate analysis used 37 gastric cancer cell lines from the Cancer Cell Line Encyclopedia. Their TCGA and related subtype distributions broadly resembled those of a large clinical reference cohort, but the lines inadequately represented a TME-high signature. This difference is consistent with the limited immune and stromal context of conventional cell culture.

Drug-target-category analyses suggested subtype-dependent patterns for pathways such as EGFR and PI3K–mTOR signaling. At the individual-drug level, 48 unadjusted associations were found, but none remained significant after correction for multiple comparisons. These associations remain hypotheses for follow-up testing.

Sampling, processing and culture conditions can influence which tumor populations are represented in an organoid. In this series, molecular subtype sometimes changed despite agreement on selected reference classifications. Paired subtype analysis therefore adds information beyond lineage staining or mutation retention when these gastric models are used for drug research.