CD155 downregulation in solid tumors weakens TIGIT blockade

What the work claims

This is a primary mechanistic study that maps how CD155 and CD112 ligands on tumor cells compete for binding to four NK-cell receptors: the inhibitory receptors TIGIT, CD96, CD112R and KIR2DL5, and the activating receptor CD226. Using the A498 renal carcinoma cell line in 2D co-culture and 3D spheroids, the authors argue that the solid tumor microenvironment lowers CD155 expression and TIGIT levels on infiltrating NK cells, which reduces the efficacy of TIGIT blockade. At the same time, CD226 retains plasticity by switching from CD155 to CD112, and CD226-CD112 ligation specifically promotes NK migration from the spheroid periphery into the core.¹

The central claim is that effective immunotherapy design must account for ligand-receptor hierarchy in three-dimensional solid tissue, not just receptor expression on immune cells in suspension. The paper points toward CD112R blockade in primary tumors as a way to both enhance NK killing and promote infiltration.

How it works

The authors first dissected receptor dominance in 2D co-cultures of interleukin-15-activated human NK cells with A498 renal carcinoma variants engineered to lack CD155, CD112, or both. In TIGIT-positive NK cells, CD155 delivered the dominant inhibitory signal; removing CD155 increased degranulation, while removing CD112 alone did not. CD112R blockade worked only when CD112 was present. CD226 blockade impaired NK activation against wild-type, CD155-knockout and CD112-knockout targets, but not the double knockout, confirming that CD226 binds both ligands. A key observation followed: when CD155 was absent, CD226 switched to CD112, as shown by CD112 accumulation at the NK-tumor interface.¹

The 3D spheroid experiments are where the organoid-relevant insight sits. A498 spheroids formed by ultrasonic standing wave showed a 20 percent drop in CD155 expression at 24 hours and a 50 percent drop at 48 hours, stabilizing thereafter, while CD112 expression remained unchanged. The authors attribute the drop to maturation of cell-to-cell contacts in solid tissue rather than a viability artifact, because spheroid viability matched monolayer viability and CD155 levels were independent of the 3D culture method.¹

Against these spheroids, TIGIT blockade enhanced NK-mediated killing, but the effect was delayed compared with monolayers and incomplete. Residual viable tumor cells persisted in the spheroid core. The authors found that NK cells retrieved from spheroids had lower TIGIT expression than NK cells kept in suspension, suggesting that the tumor microenvironment itself downregulates the target of the therapeutic antibody. CD112R blockade, by contrast, improved killing at six hours, when TIGIT blockade had not yet shown efficacy. At twelve hours both helped, but combination did not add benefit, consistent with the idea that the dominant inhibitory axis shifts from TIGIT-CD155 in monolayers to CD112R-CD112 in spheroids.¹

Prior drug exposure compounded the effect. Lenalidomide and axitinib both reduced spheroid viability when combined with NK cells, but adding anti-TIGIT did not improve outcomes because the drugs lowered TIGIT expression on NK cells. A 3D infiltration assay quantified NK distribution in 10-micrometer radial shells: 44 percent of NK volume stayed within the outer 10 micrometers, 33 percent occupied the 10-to-20-micrometer shell, 15 percent the 20-to-30-micrometer shell and only 8 percent penetrated beyond 30 micrometers. CD226 blockade reduced total infiltration, and in the presence of CD112 it redistributed NK cells toward the periphery and away from the core, indicating that CD226-CD112 ligation specifically drives inward migration.¹

Where a skeptic should push

The model system is a single renal carcinoma cell line. A498 is convenient because it expresses both ligands, but kidney cancer is not the indication driving most TIGIT clinical trials. Whether lung, breast or gastrointestinal tumors show the same ligand downregulation in 3D, or the same receptor hierarchy, needs direct testing. The authors use spheroids rather than patient-derived tumor organoids, so stromal, metabolic and hypoxic features of real tumors are absent.

The mechanistic interpretation also relies partly on siRNA-mediated CD155 knockdown and pharmacologic receptor blockade. While informative, these are strong perturbations. A more physiological question is whether native human tumors present low enough CD155 and TIGIT to make TIGIT blockade intrinsically weak, or whether the effect is specific to the A498 spheroid context. The infiltration assay is elegant but uses a single time point and one tumor line, and the authors note that they could not resolve a competing hypothesis involving CD112R inhibition in CD112-knockout spheroids.

Finally, the paper's therapeutic conclusion - favor CD112R blockade in primary tumors - is extrapolated from in vitro killing and migration assays. No animal or patient data are shown. CD112R antibodies are less advanced clinically than TIGIT antibodies, and the safety and pharmacokinetic profiles needed to achieve the inferred mechanism in patients are unknown.

What this changes for tumor organoid immunotherapy models

The non-obvious implication is that tumor organoid-immune co-culture assays need to be built in three dimensions and, where possible, with prior drug exposure, or they will misrank immunotherapy candidates. A 2D dissociated-organoid killing assay would have seen high CD155, high TIGIT and a dominant TIGIT-CD155 axis, and would have concluded that TIGIT blockade is highly effective. The spheroid system reveals that the same tumor, simply allowed to form compact tissue, lowers both ligand and receptor enough to make TIGIT blockade delayed and incomplete. For any foundry building immune-organoid platforms for checkpoint inhibitor screening, this is a direct design requirement: test in 3D, and measure ligand expression in the assay itself.

The opportunity is a more predictive preclinical filter. If TIGIT blockade looks weaker in 3D tumor spheroids than in 2D or suspension assays, the discrepancy itself becomes a biomarker. Lines or patients whose tumors retain high CD155 in solid tissue may be the real responders, while those with low CD155 may need CD112R co-blockade or a different strategy. Organoid assays that report ligand density, receptor density and infiltration depth could stratify patients before clinical trials.

The threat is overbuilding complexity on a narrow foundation. The study is built on one cell line, one NK-cell source and one spheroid method. If the field starts engineering tumor organoid-immune assays around A498-derived rules without validating across cancer types and donor-derived models, it could optimize for a renal carcinoma artifact. The deeper risk is interpretive: a beautiful mechanistic story in spheroids can be mistaken for a patient-stratification rule before it is tested against clinical response data.

The bottom line

Established: in A498 renal carcinoma spheroids, CD155 expression falls in 3D while CD112 remains stable, TIGIT blockade is delayed and incomplete, CD112R blockade becomes the earlier effective intervention, and CD226-CD112 ligation specifically promotes NK infiltration from the spheroid periphery to the core. Prior exposure to lenalidomide or axitinib lowers TIGIT and blunts the added benefit of anti-TIGIT. Unestablished: whether the same hierarchy applies across cancer types, patient-derived tumor organoids and in vivo tumors, and whether CD112R blockade translates into clinical benefit. What would strengthen the argument is replication in patient-derived tumor organoids with autologous NK cells and a direct comparison with clinical checkpoint blockade response. What would weaken it is evidence that A498 spheroids are a special case with uniquely low CD155. For the foundry, the paper is a clear reminder that immunotherapy readouts in 3D can diverge sharply from 2D, and that ignoring the divergence risks advancing the wrong drugs.