A translational approach for in vitro monitoring of patient-derived tumor organoid responses to anti-cancer treatments.

Abstract

e15610 Background: Patient-derived organoids are often considered the 3D in vitro models that accurately recapitulate the architectural, biochemical, and biophysical features of human cancers compared to conventional cultures. Colorectal cancer (CRC) is a major global health burden, underscoring the urgent need for more predictive preclinical and patient-related models. Yet, the increased physiological relevance of 3D models introduces critical challenges for monitoring therapeutic responses, including limited optical accessibility, pronounced spatial heterogeneity, and the incompatibility of conventional endpoint assays originally designed for monolayer cultures. These limitations impede real-time, very low invasive, and functionally relevant assessments of drug efficacy, thereby reducing the predictive value of in vitro models in CRC research and clinical development. Methods: We developed an innovative in vitro platform that integrates CRC organoids with sensory neurons functioning under a “Neuron-as-a-Sensor” (NaaS) paradigm. In this approach, living neurons act as biological transducers, detecting and integrating tumour-derived biochemical and functional signals, and translating them into quantifiable electrophysiological signatures. The platform comprises a compartmentalized microfluidic device incorporating a customized microelectrode array (MEA), designed to spatially organize co-cultures of patient-derived CRC organoids embedded in a neuro-conductive hydrogel alongside sensory neurons. Results: CRC organoids were derived from a moderately differentiated metastatic adenocarcinoma (left colon, wild-type RAS/RAF, MSS), partially responsive to neoadjuvant treatment (ypT4 N1 M+). We observed robust, spatially organized interactions between neuronal projections and CRC organoids, consistent with an innervated tumour structure. Next, clinically relevant anti-CRC mono and combination therapies were applied, while the therapeutic efficacy was monitored by neuronal activity. MEA recordings revealed statistically significant treatment-dependent alterations in neuronal network activity compared to untreated controls. Notably, robust discrimination between treatment effects required the integration of multiple electrophysiological metrics, enabling the generation of multidimensional response maps that captured distinct therapeutic signatures. Conclusions: This work demonstrates the feasibility and value of a neuron-as-a-sensor strategy for label-free, real-time monitoring of therapeutic responses in complex 3D tumour models. By coupling patient-derived CRC organoids with neuronal biosensors in a microfluidic MEA platform, we provide a scalable, functionally rich approach with strong potential to enhance preclinical drug evaluation and improve the translational relevance of advanced in vitro CRC models.