StarTrace Tracks Long-Term Drug Responses in Pooled Colorectal Cancer Organoids

Following Individual Models Within a Shared Culture

University of South Carolina researchers developed StarTrace to measure how different patient-derived colorectal cancer organoids respond when they are cultured together. Each established model carried a distinct DNA barcode, allowing its contribution to a mixed population to be followed during treatment. The study compared barcode measurements by nanopore sequencing with a targeted qPCR approach.

One pool contained 12 organoid models and was evaluated over 28 days with weekly sampling. A second pool contained 15 models and extended the approach to additional drug conditions. The assay tracked changes in each model's barcode signal relative to vehicle controls. Its long-term endpoint captured relative growth or survival within the experimental population over time.

Known Genotype–Drug Relationships Provide Test Cases

Across the models, 13 of 20 were resistant to the MDM2 inhibitor nutlin, and 12 of those 13 carried inactivating TP53 mutations. One resistant model had no detected TP53 mutation, leaving its resistance unexplained by that marker. The pool therefore reproduced a familiar relationship while retaining an exception.

A model carrying an RNF43 mutation was sensitive to PORCN inhibition when grown without an external WNT supplement. Providing external WNT reduced the inhibitor's activity, consistent with a drug that blocks endogenous WNT ligand production. The culture environment was therefore part of the genotype–drug comparison, rather than a neutral background condition.

The researchers also investigated an unexpected ibrutinib response in a model with an ERBB3 variant. In a mixture of tumor organoids and patient-matched normal material, the tumor-associated mutant allele declined after treatment. That experiment supported a tumor-associated response; it did not establish the ERBB3 variant itself as the direct molecular drug target.

Comparing Sequencing with qPCR

StarTrace qPCR used barcode-specific detection to quantify the individual models within a pool. Across the tested conditions, qPCR-derived and sequencing-derived changes were positively correlated. Reported Pearson correlations were 0.89 for nutlin, 0.79 for LGK-974 and 0.59 for etoposide. These comparisons examined the measurement of barcode response in the same experimental setting.

A separate two-model mixture compared responses to the PORCN inhibitors LGK-974 and Wnt-C59. The organoid with the relevant RNF43 background showed the expected sensitivity. These targeted comparisons complemented the larger pooled experiments by providing a simpler setting in which to inspect an individual response.

Miniaturization Preserves Some Responses but Loses Models

The investigators then moved the 15-model pool into a 96-well format and tested eight drugs at four concentrations. Twelve models produced sufficiently detectable signals for analysis. Three were undetectable at some untreated time points and were excluded. The smaller assay reproduced several expected sensitivities, including nutlin responses in TP53-wild-type models and the medium-dependent PORCN-inhibitor response.

The dropout is an operational limit of the pooled measurement: a model must remain detectable before its treatment-associated change can be estimated. Moreover, barcode abundance in a shared culture reflects the model's behavior in that competitive context. It provides a different perspective from a short single-model viability assay.

StarTrace offers a way to compare prolonged pharmacological responses while reducing the number of cultures maintained separately. The study did not link its pooled assay with a prospective patient-treatment outcome cohort. The authors and the University of South Carolina disclosed a patent application covering aspects of the platform.