Culture Medium Rewrites Drug Sensitivity in Patient-Derived Tumor Organoids

Patient-derived tumor organoids have become one of the most promising tools in translational oncology: tiny three-dimensional cultures grown from a patient’s own tumor that preserve its histology, mutations and cellular complexity outside the body. But a new study from researchers at Antwerp University Hospital and the University of Antwerp, published in Experimental Hematology & Oncology, delivers a sobering warning to laboratories that rely on these models. The choice of culture medium, the team reports, is not a passive background variable. It actively reshapes the transcriptional state, morphology and drug sensitivity of the organoids, sometimes enough to flip a tumor model from sensitive to resistant against an entire class of targeted therapies.

The study compared two culture formulations side by side. The conventional approach, used in most organoid labs worldwide, supplements the medium with Wnt, R-spondin and noggin, growth factors that sustain the stem-like, self-renewing behavior of epithelial tumor cells. The alternative is OncoPro Tumoroid Culture medium, a proprietary, serum-free formulation introduced by Thermo Fisher Scientific that deliberately omits Wnt-3A, R-spondin, noggin and small-molecule inhibitors, offering standardized formulations and broader commercial freedom to operate. The researchers tested both media across 36 patient-derived tumor organoid lines spanning pancreatic, colorectal, gastric, gastroesophageal junction, lung, biliary tract, head and neck, breast cancers and a carcinoma of unknown primary.

The first practical question was whether organoids could even grow in the new medium. When existing Wnt-grown organoid lines were transferred to OncoPro, 83.3 percent, or 15 of 18 lines, adapted successfully, with 95 percent confidence intervals running from about 61 to 94 percent. De novo establishment from fresh tumor tissue told a different story: only 11.1 percent of samples, 2 of 18, established successfully in OncoPro, compared with 33.3 percent, 6 of 18, in Wnt medium. Notably, none of the de novo lines survived exclusively in OncoPro, although organoids that did take hold there often showed faster early expansion, with larger diameters and cystic morphologies at first passage. That early advantage, however, did not translate into better long-term establishment.

One tumor type proved a persistent problem. Non-small cell lung cancer organoids emerged rapidly in both media, but immunohistochemical profiling showed they had lost the marker signature of the parental tumors. A squamous-cell carcinoma sample that should have been TTF-1 negative and p63 positive shifted to TTF-1 positivity in culture, while an adenocarcinoma that should have expressed Napsin A did not. In both cases the pattern pointed to overgrowth by non-malignant epithelial cells, a known challenge in lung organoid culture that OncoPro did not overcome. These cultures were excluded from the downstream comparisons and reported as a model-specific limitation.

For the organoids that did grow well in both conditions, the deeper question was what the media were actually doing to the tumor cells. Whole-exome sequencing on nine matched lines showed that key oncogenic driver alterations were retained across both media, and germline fingerprinting confirmed that parental, OncoPro-grown and Wnt-grown samples were genetically related with no evidence of cross-contamination. Bulk RNA sequencing of 11 matched lines found that samples clustered predominantly by patient of origin rather than by medium, and in most tumor types fewer than 10 percent of genes were differentially expressed between conditions. Tumor identity, in other words, survived the medium switch.

What changed instead was the epithelial state. Organoids grown in Wnt medium upregulated proliferation-associated genes such as STMN1 and CDCA7 along with canonical Wnt-responsive stem cell regulators including LGR5 and LRIG1. Organoids grown in OncoPro showed higher expression of adhesion-associated genes such as CEACAM6, COL17A1, AMIGO2 and AHNAK2, and gene set enrichment analysis revealed activation of interferon-alpha, interferon-gamma and TGF-beta signaling programs, although with stricter statistical thresholds only the interferon-alpha response remained significant, indicating that the pathway-level differences across the heterogeneous cohort were modest. Morphology diverged in parallel: Wnt-grown organoids tended to be rounded, cohesive and smooth, while OncoPro-grown organoids were irregular, crenulated and structurally heterogeneous.

A clue to the mechanism came from an exploratory cytokine assay measuring 49 proteins. The OncoPro supplement itself contained high levels of total TGF-beta, meaning the cytokine is introduced directly by the proprietary formulation rather than produced solely by the organoids. Active TGF-beta was absent at baseline but appeared during culture exclusively in OncoPro supernatants, consistent with activation of the latent TGF-beta pool, while no active TGF-beta emerged in Wnt cultures. The supplement also contained detectable IL-15, and chemokines such as IP-10 and IL-8 accumulated over time, far more in OncoPro than in Wnt medium. Because these measurements were not normalized for organoid size or cellularity, the authors caution they should be read as hypothesis-generating rather than definitive.

To ask whether these medium-imposed states resemble anything found in actual patients, the team projected their OncoPro-up and Wnt-up gene signatures onto public single-cell RNA sequencing atlases of pancreatic ductal adenocarcinoma and colorectal cancer. The result was striking. In pancreatic cancer, the OncoPro signature mapped predominantly to malignant epithelial cells, with the highest scores in the classical subtype and intermediate scores in basal-like cells, and was largely absent from non-malignant acinar and ductal epithelium. The Wnt signature did the opposite, aligning with non-malignant epithelial populations. In colorectal cancer the pattern repeated, with the OncoPro signature enriched in malignant cells. With only three matched pancreatic and four matched colorectal organoid pairs, the authors stress this analysis is exploratory, but it suggests OncoPro may capture malignant programs that conventional medium dilutes.

The pharmacological consequences were the study’s most consequential finding. Screening a 33-drug panel in three matched pancreatic and colorectal organoid lines, the researchers found that drug-response profiles clustered by culture medium rather than by individual organoid line. Across all paired measurements, OncoPro and Wnt responses correlated only moderately, with a Pearson correlation of roughly 0.59, and the distribution skewed toward higher drug effect in Wnt medium. Wnt-grown organoids were globally more sensitive in the screened subset, particularly to MAPK-axis inhibitors such as trametinib and ASTX029 and to the apoptosis-sensitizer birinapant, while OncoPro-grown organoids showed relative resistance. Taxanes and classical chemotherapies including 5-fluorouracil and irinotecan centered near zero medium effect, and platinum agents showed flat dose-response curves in both media, indicating limited ex vivo activity regardless of formulation.

Crucially, the sensitivity shifts appear reversible rather than reflecting permanent clonal selection. When the team transferred a Wnt-established pancreatic line to OncoPro and back again, and when they grew two independent lines from the same colorectal tumor in each medium and then switched them, drug responses moved toward greater sensitivity after return to Wnt medium across the panel, arguing against irreversible loss of sensitive subpopulations and favoring a dynamic, medium-dependent epithelial state. The authors argue this may help explain why organoid-guided trials of targeted therapies have struggled clinically: historical screens may have tested targeted agents in culture conditions that induce artificial hypersensitivity absent from the patient’s tumor. They do not advocate a single best medium. Instead, they suggest Wnt media for expansion and stemness-vulnerability discovery, OncoPro for malignant-state and targeted-therapy profiling, and side-by-side testing when in doubt, alongside prospective studies linking parallel medium screens to actual patient outcomes.

Subject of Research: How culture medium composition shapes phenotype, transcriptome and drug response in patient-derived tumor organoids

Article Title: Comparative characterization of OncoPro™ and Wnt-based media reveals distinct phenotypic and pharmacologic states in patient-derived tumor organoids

Article References: Seghers, S., Le Compte, M., Rodrigues Fortes, F., Baroen, J., Ott, J., Roeyen, G., Hartman, V., Kumar-Singh, S., Coppens, L., de Maat, M., Komen, N., Van Den Broeck, S., Valk, J., Tjalma, W., Hendriks, J. M. H., Van Schil, P., Van Haesendonck, G., Peeters, S. L. J., Prenen, H., & Deben, C. (2026). Comparative characterization of OncoPro™ and Wnt-based media reveals distinct phenotypic and pharmacologic states in patient-derived tumor organoids. Experimental Hematology & Oncology, 15(1), Article 110. https://doi.org/10.1186/s40164-026-00831-y

Image Credits: AI Generated

DOI: 10.1186/s40164-026-00831-y

Keywords: patient-derived tumor organoids, OncoPro medium, Wnt signaling, drug sensitivity, TGF-beta, pancreatic cancer, colorectal cancer, single-cell RNA sequencing, transcriptional states, organoid drug screening, precision oncology, epithelial plasticity