IGFBP5-Conditioned Vasculature Changes Drug Responses in Colorectal Tumoroid Chips

Building a Vascular Component for Antiangiogenic Drug Testing

Antiangiogenic treatments act on a tumor’s blood-vessel environment, which tumor-only cultures cannot reproduce. Researchers at Jinling Hospital and affiliated university programs developed a colorectal tumoroid-on-a-chip that combines patient-derived tumor cultures with a perfusable endothelial network. Their Cell Reports Medicine study asks whether making that network more tumor-like changes drug-response measurements and their agreement with patient outcomes.

The chip places colorectal tumoroids in a central compartment surrounded by human umbilical vein endothelial cells, or HUVECs. The endothelial cells self-assemble into connected vessels before tumoroids are introduced. Fluorescent dextran and microbead experiments confirmed perfusion, and a 1:1 mixture of endothelial and tumoroid media supported both compartments. Selected tumor and endothelial markers did not differ significantly from the respective single-medium controls.

HUVECs provide a practical vascular source but are not native colorectal tumor endothelial cells. To address that difference, the investigators analyzed two public single-cell datasets and identified IGFBP5 as a highly enriched factor in tumor-associated endothelium. Tissue staining supported its association with colorectal tumor vessels. They then conditioned HUVECs with 50 ng/mL IGFBP5 and compared them with unconditioned cells.

The Endothelial State Alters Both Structure and Drug Response

Conditioned cells showed altered tube formation and gene expression, with closer transcriptomic similarity to isolated colorectal tumor endothelial cells than untreated HUVECs. On-chip networks became denser and more branched, with shorter vessel segments and a wider distribution of diameters. Dextran leakage and microbead escape increased, while microbead flow slowed. These are measured features of the experimental vascular interface rather than a reconstruction of every aspect of tumor angiogenesis.

The study also tested potential confounding effects. Tumoroid co-culture alone did not significantly increase endothelial IGFBP5 expression. Adding IGFBP5 directly to tumoroids did not significantly change the measured colorectal markers, viability or proliferation. These comparisons supported the use of exogenous conditioning to modify the vascular compartment.

Drug testing compared original and conditioned-vascular models from the same patients. The main regimen was fluorouracil, leucovorin and oxaliplatin (FOLFOX) plus bevacizumab. Vascular density and relative tumoroid diameter were assessed after four days of exposure. Both endpoints varied between patients. The conditioned model generally appeared less sensitive than the original model, and selected resistant clinical cases were represented more closely by the conditioned network.

Comparing Chip Results with Patient Outcomes

The principal outcome comparison used 22 patients with clinical FOLFOX–bevacizumab response information. Complete and partial responses were grouped as clinical response; stable and progressive disease were grouped as resistance. Thresholds for vessel density and tumoroid diameter were selected from these data using the Youden index and bootstrap sampling.

The optimized model achieved 81.82% sensitivity, 90.91% specificity and 86.36% accuracy, compared with 54.55%, 90.91% and 72.73% for the original model. Here, sensitivity means correctly identifying clinically resistant patients. Specificity means correctly identifying clinically responsive patients.

The conditioned model also had lower coefficients of variation for the measured endpoints. This supports better experimental consistency within the reported comparison, while the same small cohort was used to choose thresholds and assess clinical agreement. It is not an independent multicenter validation.

Additional experiments examined several small-molecule antiangiogenic agents and showed patient-dependent vascular and tumoroid responses. Those exploratory tests broaden the possible use of the system but do not establish clinical recommendations for untested treatment choices.

The study’s main contribution is an experimental demonstration that endothelial state changes a drug-screen result. A vascularized model can therefore require more than simply adding endothelial cells to a tumor culture. The authors identify the 22-patient sample and the need for larger, clinically annotated multicenter cohorts as central limitations. Their findings support further evaluation of this interface while leaving its general predictive utility and resistance mechanisms unresolved.