Colon tumoroids with their macrophages attached

What the work claims

This is an observational, prospective, case-only study with a drug-design ambition attached, and the distinction matters. The stated goal is to create novel small-molecule inhibitors of the CD47-SIRPa interaction, using structure-based drug design, to reverse tumor-associated macrophage mediated immune suppression in colorectal cancer. The record claims that such small molecules could target the tumor microenvironment with greater efficacy and lower toxicity than CD47-targeting antibodies. To test activity on the tumor immune microenvironment, the investigators propose organoids derived from patient colon cancer biopsies, on the grounds that tumoroids preserve the patient's tumor stroma, including myeloid cells, and therefore model tumor-immune interaction more faithfully than tumor-only cultures.¹

What is actually registered as measurable is more modest. Enrollment is estimated at 115 adults with colon cancer undergoing surgery. The biospecimen plan is specific: 24 ml of peripheral blood, plus tumor and adjacent healthy tissue taken in parallel from the surgical specimen. The single primary endpoint is the establishment of patient-derived organoids, a minimum of 30 lines, resembling the primary tissue sample, scored by a histology comparison at 18 months from study launch. Secondary endpoints are a validated cryobank of at least 30 lines assessed through 4 passages with a pre- and post-cryopreservation growth score, a T-cell biobank built by FACS isolation of CD3-positive cells over 48 months, and an assessment of immunotherapy effects on the antitumor activity of autologous peripheral-blood T cells, read out by tumor viability using CellTiter-Glo and flow cytometry.¹

How it works

The CD47-SIRPa axis is a macrophage checkpoint. CD47 is a surface protein expressed broadly, including on red blood cells, that engages SIRPa on macrophages and delivers a do-not-eat-me signal that suppresses phagocytosis. Many tumors upregulate CD47 to escape immune clearance, and tumor-associated macrophages are abundant in the colorectal cancer microenvironment, typically polarized toward a tumor-supporting state. Blocking the interaction should restore macrophage phagocytosis of tumor cells, which is the rationale for the inhibitor program. The record's claim that small molecules may beat antibodies rests on pharmacology: an antibody against CD47 circulates systemically and hits CD47 wherever it is expressed, including on red blood cells, whereas a small molecule designed against the SIRPa-CD47 interface could in principle be tuned for tumor-microenvironment exposure, though the record itself does not demonstrate that property for any specific compound.¹

The model innovation is the point. A conventional tumor organoid enriches epithelial tumor cells and loses stromal and immune compartments within a few passages. The MACROSWITCH design assumes its tumoroids retain myeloid cells, and it hedges that assumption intelligently: the cryobank endpoint scores growth through 4 passages, which is exactly where stromal and myeloid dropout would show up as drift between the pre- and post-cryopreservation phenotype. The functional assay then pairs the tumoroid with autologous T cells isolated from the same patient's blood, with tumor viability as the shared readout.¹

Where a skeptic should push

The most load-bearing assumption is that a myeloid-preserving tumoroid can be built and kept stable long enough to screen drugs, and here the registry is candid in an uncomfortable way: the primary endpoint is exactly that assumption, with a floor of 30 established lines. Everything downstream, the inhibitor screening, the immunotherapy readouts, the drug-design loop, is contingent on a model whose feasibility is still the open question. The 48-month timelines on the functional endpoints and an overall completion estimate of 2027-01 against an original primary-completion estimate of 2025-01, with the record last verified in 2024-10 and still recruiting, tell you the program is running slower than its own registry arithmetic. Slippage is not failure, but at 115 consented surgical patients the yield will matter: if establishment with intact myeloid stroma lands near typical tumoroid success rates, the surviving lines may be a small, biology-selected subset.

Second, the observational frame means there is no clinical outcome at the end of this chain. No patient receives a MACROSWITCH-selected drug; the endpoints are histology scores, growth scores, and ex vivo viability. The leap from reversing macrophage suppression in a dish to efficacy in metastatic colorectal cancer crosses pharmacokinetics, tumor penetration, on-target hematologic exposure, and the adaptive remodeling of the myeloid compartment, none of which a tumoroid contains. Structure-based small-molecule programs also fail routinely at developability long after the binding assay looks clean.

Third, myeloid retention cuts both ways as an assay property. A tumoroid that keeps patient macrophages is more physiological, but it is also more variable: macrophage state depends on the patient's treatment history, sampling site, and time in culture, so well-to-well and patient-to-patient variance will be high, and the study registers no power calculation because, as a feasibility study, it has no formal hypothesis. Readers of any eventual positive screen should ask how the variance was partitioned before believing a hit.

Stroma-preserving tumoroids as immunotherapy assay

The non-obvious implication for organoid models of human organs is that this study inverts the usual critique. The field's recurring failure mode is using tumor-only organoids to rank drugs whose mechanism requires immune effectors, an assay-drug mismatch that quietly invalidates the ranking. MACROSWITCH starts from the drug's mechanism and engineers the model backward from it: because the target is a macrophage checkpoint, the model must keep macrophages. That inversion, mechanism-first assay design, is the transferable idea, and it applies wherever organoids are being sold into immunology-driven drug discovery, from macrophage-targeting agents to T-cell engagers to bispecifics.

The opportunity is a validated, cryobanked, stroma-preserving colon tumoroid panel with paired adjacent-normal lines and autologous T cells from the same patients. Paired tumor and matched normal from one surgical specimen is a genuinely strong design: the adjacent-normal organoid is the control that most tumoroid studies lack, and a cryobank that survives 4 passages with a documented growth score would be an infrastructure contribution other groups could build on regardless of what happens to the inhibitor chemistry. If the myeloid compartment really persists and can be standardized, the panel becomes a screening service with a defensible mechanistic basis.

The threat is silent drift in the one compartment the whole design depends on. Myeloid cells are exquisitely sensitive to culture conditions, and a tumoroid that loses its macrophages between establishment and screening quietly becomes the tumor-only model the study was designed to avoid, while still being called a tumoroid. The registered growth score through 4 passages is the right control only if it is reported per compartment, epithelial, stromal, and myeloid, not as an aggregate. And the generalization caveat applies in full: one French surgical cohort, one culture pipeline, one tumor type. A feasibility success in Marseille would be a license to replicate, not a property of stroma-preserving tumoroids as a class.

The bottom line

Established by the registry record: a recruiting observational study at Institut Paoli-Calmettes that will collect tumor, adjacent normal tissue, and blood from an estimated 115 colon cancer patients, with a primary endpoint of establishing at least 30 histology-faithful patient-derived organoid lines in 18 months, a cryobank scored through 4 passages, an autologous CD3 T-cell biobank, and ex vivo immunotherapy readouts on tumor viability over 48 months, in support of a structure-based small-molecule CD47-SIRPa inhibitor program. Asserted, not established: that the tumoroids retain a functional myeloid compartment long enough to screen drugs, that small molecules will achieve tumor-restricted checkpoint blockade where antibodies could not, and that any of this predicts clinical efficacy. What would confirm the approach: per-compartment validation of myeloid retention across passages, hit reproducibility across patients with variance reported, and a lead compound with demonstrated tumor-microenvironment selectivity. What would break it: establishment success coupled to myeloid dropout, or ex vivo activity that evaporates once exposure and hematologic toxicity are considered. The model work is worth watching; the drug claims should wait for the model.