Lung cancer organoid-immune co-cultures dissect ADC mechanism before the clinic

What the work claims

The study, led by the Guangzhou Institute of Respiratory Disease, is designed as a ten-patient, prospective, case-control observational protocol. Its claim is that patient-derived organoids of HER2- or TROP2-positive non-small cell lung cancer, co-cultured with the patient's own tumor-infiltrating lymphocytes or peripheral blood mononuclear cells, can reproduce the cellular and immune mechanisms that determine antibody-drug conjugate activity. Rather than relying on organoid killing alone, the protocol layers pharmacodynamic, imaging, transcriptomic, proteomic, and immunophenotyping readouts to distinguish target expression, payload internalization, apoptosis, antibody-dependent cellular cytotoxicity, and bystander effects. If the readouts correlate as expected, the platform could provide a mechanistic rationale for selecting patients and combinations for HER2- and TROP2-directed ADCs.¹

How it works

The starting material is clinically resected tumor tissue or malignant pleural effusion from patients with HER2- or TROP2-positive non-small cell lung cancer. The protocol states that organoids will be validated for histological fidelity and phenotypic stability by hematoxylin and eosin staining and TTF-1 immunohistochemistry, then stratified by target expression intensity. Autologous immune cells, either tumor-infiltrating lymphocytes or peripheral blood mononuclear cells, will be added to the co-culture. Their purity and activation status will be checked by flow cytometry, and effective immune-tumor crosstalk will be confirmed by ELISA for cytokines.

The pharmacodynamic readouts are deliberately stacked. ATP-based viability assays, live imaging of organoid viability, and Caspase-3/7 apoptosis detection will measure tumor killing. The protocol specifies that ADCs will be tested at concentrations designed to mimic peak plasma levels after clinical administration, so the ex vivo exposure is meant to map onto in vivo dosing. The ADCs named in the protocol are trastuzumab deruxtecan, trastuzumab emtansine, and sacituzumab govitecan, tested alone and in combination with tyrosine kinase inhibitors or immune checkpoint blockers.

The mechanistic layer is what sets the protocol apart. pHrodo-conjugated ADCs will be tracked by high-content imaging at 15-minute intervals for 24 hours to quantify cellular uptake and intracellular trafficking. Single-cell RNA sequencing will be performed 96 hours after co-culture and treatment to profile transcriptional changes and identify subpopulations that are sensitive or resistant. Multiplex immunofluorescence and Olink proteomics will add spatial and secreted-protein information on inflammatory cytokines. Secondary endpoints include ELISA for IL-6, TNF-alpha, and IFN-gamma in the supernatant, plus flow cytometry for T cell activation markers CD69 and CD25 and exhaustion markers PD-1 and TIM-3.¹

Where a skeptic should push

The single most load-bearing assumption is that an ex vivo co-culture of ten patients can capture the biology that determines ADC response in the clinic. The protocol language describes observed ADC responses and mechanistic insights, yet the trial is listed as recruiting with an estimated enrollment of ten and a start date of late 2025. That mismatch is worth flagging: what is being presented as trial-record text may include preliminary data, planned analyses, or aspirational language, and the reader cannot easily tell which is which without a linked peer-reviewed publication.

Ten patients is a small observational cohort. Even if every readout works technically, the study will be underpowered for robust correlations between organoid response and clinical outcome. The case-control design compares strong versus weak HER2/TROP2 expression, but it does not randomize treatment, so any association between organoid sensitivity and patient benefit will be hypothesis-generating at best.

Another concern is generalization. HER2- and TROP2-positive non-small cell lung cancer are molecularly heterogeneous subsets. The protocol does not state how many patients will fall into each target-expression stratum, or whether the organoid-immune co-culture will be validated against the original tumor microenvironment. Autologous PBMCs are not tumor-resident immune cells, and their reactivity in a dish may not mirror the exhausted, suppressive tumor infiltrate that limits ADC efficacy in patients. Finally, the use of peak plasma concentrations is a pragmatic choice, but it ignores drug exposure over time, tumor penetration gradients, and systemic metabolism.

Implication for organoid models of human organs and drug-discovery work

For organoid-based drug discovery, the value of this protocol is its explicit attempt to link a compound's pharmacodynamic effect to its mechanism of entry and the immune response it triggers. Most PDO drug-sensitivity studies report an IC50 and stop there. This study asks a harder question: why does the ADC kill, and why does it fail? By combining pHrodo internalization, Caspase-3/7 apoptosis, single-cell transcriptomics, and immune-cell phenotyping, it moves toward a causal model of ADC action rather than a purely correlative sensitivity map.

The opportunity is a more rational approach to combination therapy. If the organoid-immune co-culture shows that an ADC's bystander effect depends on T cell presence, or that checkpoint blockade restores exhausted T cell killing, sponsors would have a human-relevant justification for pairing ADCs with immunotherapy. The same platform could be repurposed for other solid-tumor ADCs once the target and payload readouts are swapped in.

The threat is that the model may be too complicated to reproduce. Co-cultures introduce donor-to-donor immune variation, batch effects in organoid-immune cell ratios, and readout interactions that make standardization difficult. A ten-patient feasibility study cannot validate all of that. There is also a real risk of over-interpreting mechanism: an ADC that internalizes well and triggers apoptosis in an organoid may still fail in patients because of tumor penetration, antigen heterogeneity, or systemic toxicity that the ex vivo system does not model.

The less obvious implication is governance. As organoid-immune co-cultures move from research tools toward predictive assays for clinical trials, regulators and ethics committees will need to decide how much weight to give a multi-modal ex vivo result. This protocol is a early example of the kind of data package that could eventually support a companion-diagnostic-like claim for organoid-guided ADC selection, but only if the mechanistic readouts are reproducibly tied to clinical benefit in much larger studies.

The bottom line

This is an ambitious observational protocol, not a completed result. Its strength is the stacked mechanistic design: pharmacodynamic, imaging, transcriptomic, proteomic, and immunophenotyping readouts applied to the same patient-derived organoid-immune co-culture. Its weakness is the small planned enrollment and the absence, so far, of peer-reviewed data linking the ex vivo signatures to clinical outcomes. What would confirm the model is a prospective validation showing that organoid ADC sensitivity, internalization depth, and immune activation jointly predict patient response. What would weaken it is evidence that the co-culture readouts are dominated by technical variables, such as immune-cell preparation or organoid batch, rather than by target biology. For now, the protocol is best read as a detailed blueprint for how organoid models can be used to decompose ADC mechanism, not as proof that they already do.