Scientists design first selective TNKS2 inhibitor to shut down WNT signaling in 8p-deleted tumors

WNT signaling is one of the most powerful engines of cancer growth, yet every attempt to drug it systemically has stumbled over the same problem: healthy tissue needs WNT too. A new study published in Nature Chemical Biology reports a way around that barrier by exploiting a genetic weakness found in roughly half of common epithelial cancers.

The strategy rests on a concept known as collateral vulnerability. Large genomic deletions in chromosome 8 are frequent in many epithelial cancer types. Analysis of The Cancer Genome Atlas PanCancer Atlas and International Cancer Genome Consortium whole-genome sequencing data showed that approximately half of colon, lung, breast, liver and prostate cancers carry heterozygous or homozygous 8p deletions. This segment contains the TNKS1 gene, and loss of the region correlated with reduced TNKS1 transcript and protein expression. Because TNKS1 and TNKS2 are partially redundant enzymes that both restrain WNT signaling by destabilizing the AXIN1 protein, a tumor that has discarded one copy or both copies of TNKS1 becomes unusually dependent on the remaining TNKS2 to sustain that activity.

The team first confirmed this dependency genetically. In 8p diploid DLD1 colorectal cancer cells, both TNKS enzymes had to be silenced to stabilize AXIN1, but in the 8p-deleted lung cancer line HCC1171, silencing TNKS2 alone was sufficient. Using CRISPR engineering, the researchers generated isogenic DLD1 clones lacking one or both TNKS1 alleles. In homozygous knockout cells, TNKS2 silencing stabilized AXIN1 to the same degree as dual suppression, and in heterozygous clones the effect appeared at intermediate levels, indicating that TNKS1 is haploinsufficient: a single copy cannot fully support cellular TNKS activity.

Mouse experiments reinforced the tumor-selective logic. Homozygous deletion of both tankyrase genes causes embryonic lethality, whereas TNKS2 knockout mice are viable and fertile, hinting that systemic TNKS2 targeting should spare normal tissue. To test this directly, the researchers created mice carrying an inducible shRNA against Tnks2 and crossed them with animals lacking one or both copies of Tnks1. In the intestine, a tissue with high WNT pathway activity, silencing both tankyrases significantly reduced stem-cell-driven organoid formation from isolated crypts, while Tnks2 suppression alone produced only a mild, statistically insignificant decrease. When Tnks1 was deleted, however, doxycycline-induced Tnks2 silencing caused a marked reduction in Lgr5 expression, a key WNT target and intestinal stem cell marker, and a significant defect in organoid formation. Heterozygous animals showed intermediate effects, mirroring the dose-dependent behavior seen in the engineered cell lines.

Designing a drug that hits TNKS2 but not TNKS1 was the central chemical challenge. The catalytic PARP domains of the two enzymes are highly similar, particularly around the nicotinamide and adenosine subsites where more than 50 previously reported tankyrase inhibitors bind. When the team resynthesized and tested four published TNKS2-selective compounds, each showed less than tenfold selectivity in a commercial PARylation assay, contradicting earlier claims. Scanning available crystal structures, the researchers spotted two residues near the enzyme’s acceptor site, alanine 1112 and leucine 1136 in TNKS2, that are not conserved in TNKS1. They reasoned that a ligand extending from the nicotinamide subsite into the acceptor site could exploit this difference.

Using structure-based design with docking and validated free energy calculations, the team synthesized more than 300 small molecules based on a quinazolin-4(3H)-one scaffold derived from the pan-tankyrase inhibitor XAV939. From an initial screen, 88 compounds advanced to full dose-response testing against both enzymes. Seven achieved high potency against TNKS2, with half-maximal inhibitory concentrations below 50 nanomolar, and more than 25-fold selectivity over TNKS1. Cellular validation narrowed the field further: two candidates were inactive in cells up to 10 micromolar, and one compound, TDI-012804, emerged as the most potent and clearly selective. In an AXIN1-GFP reporter assay, TDI-012804 stabilized AXIN1 in tankyrase-1-depleted cells but not in control cells, while XAV939 and G007-LK behaved identically in both. In a TOPFlash transcriptional reporter in isogenic DLD1 cells, only TDI-012804 remained potent, with an EC50 below 50 nanomolar, and TNKS2 selective. The compound also showed no inhibition of the related PARP-family enzymes PARP1, PARP2 and PARP3.

Cryo-electron microscopy revealed why the drug discriminates. The researchers determined a roughly 2.2-angstrom-resolution structure of TDI-012804 bound to the TNKS2 catalytic domain assembled as a helical filament. The ligand occupies both the nicotinamide donor site and the acceptor site, anchored by hydrogen bonds to serine 1068 and the glycine 1032 backbone, a π-π stacking interaction with tyrosine 1071, and, in the acceptor subsite, a hydrogen bond to glutamate 1138. The decisive feature is the B-loop flanking the acceptor site. The terminal phenyl ring of TDI-012804 forces this loop into an open conformation, whereas XAV939 or no ligand leaves it closed. In the closed state, leucine 1134 would clash with the phenyl ring, blocking binding. Structural comparisons showed the TNKS2 B-loop is dynamic, sampling open and closed forms, while the TNKS1 B-loop is modeled almost exclusively in a rigid closed conformation. A single amino acid, leucine 1136 in TNKS2 versus tyrosine 1289 in TNKS1, distinguishes the loops. Swapping leucine 1136 for tyrosine significantly reduced TDI-012804’s potency against TNKS2 without affecting XAV939, and the reciprocal Y1289L substitution increased TNKS1’s sensitivity to the drug. The substitutions did not fully invert selectivity, implying additional features also contribute, but cellular experiments confirmed the mechanism: TDI-012804 stabilized AXIN1 in tankyrase-1 knockout cells expressing wild-type TNKS2 but not in cells expressing the L1136Y mutant.

Functional tests in organoids carrying a common truncating APC mutation, which models events seen in familial and sporadic colorectal cancer, demonstrated tumor-restricted WNT suppression. In Apc-mutant intestinal organoids lacking both Tnks1 copies, TDI-012804 produced a transcriptional response nearly identical to the pan-tankyrase inhibitor G007-LK, with reduced stem and progenitor cell signatures and upregulated differentiation markers. Organoids with intact Tnks1 showed no significant gene expression changes compared to vehicle controls. RNA-seq comparisons found essentially no differentially expressed genes between TDI-012804 and G007-LK in knockout organoids, and only a single differentially expressed gene in wild-type organoids, strongly suggesting the compound lacks measurable off-target activity at effective doses. Over six days of treatment, TDI-012804 strongly suppressed growth of Tnks1 knockout organoids with an EC50 of 59.1 nanomolar, while Tnks1-intact organoids were 19-fold less sensitive, at 1.1 micromolar, and heterozygous organoids showed a 5.5-fold lower EC50 than wild type. Prolonged 20-day treatment caused no cumulative detriment in wild-type cultures. The compound also selectively curbed proliferation of Tnks1 knockout organoids carrying Apc, Kras and p53 mutations, with 13-to-25-fold lower EC50 values than in wild-type lines.

Human cancer cell lines and patient-derived organoids extended the picture. In the 8p heterozygous colorectal line SW403, TDI-012804 stabilized AXIN1 and reduced colony formation similarly to G007-LK. In 8p diploid H23 lung cancer cells, the compound alone did nothing, but depletion of TNKS1 sensitized them, an effect enhanced by the CDK4/6 inhibitor palbociclib. OVCAR4 ovarian cancer cells responded to TDI-012804 only after TNKS1 depletion. Among two APC-mutant colorectal cancer patient-derived organoids, the one carrying a heterozygous 8p deletion was significantly more sensitive to TDI-012804 than the 8p diploid line, with a P value of 0.02.

The authors caution that the work is a proof of concept rather than a therapy. Initial attempts to find a suitable vehicle for in vivo delivery of TDI-012804 were unsuccessful, so the safety of this class of molecules in animals and patients remains undefined, although the viability of TNKS2 knockout mice suggests it would be well tolerated. The study also notes limits on who could benefit: colorectal cancers with short APC truncations or downstream β-catenin mutations often do not respond to tankyrase inhibition, and drug resistance driven by lineage change can relieve WNT dependence, though such resistance frequently involves YAP signaling, which tankyrases also regulate. Transcriptomic data showed YAP signaling among the top downregulated pathways after TNKS2 inhibition in knockout organoids, raising the possibility that the drug could counter both pathways simultaneously.

Subject of Research: Development of a TNKS2-selective inhibitor that exploits chromosome 8p loss to suppress WNT signaling in tumors

Article Title: A selective TNKS2 inhibitor for tumor-restricted WNT suppression

Article References: Zimmerman, J., Malone, B. F., Chu, K., Finkin-Groner, E., Sun, S., Liang, R., Bahadur, S., Foronda, M., Schatoff, E. M., Granowsky, E., Madorsky, F., Goswami, S., Katti, A., Leach, B., Alcorn, H., Tammela, T., Hopkins, B., Fukase, Y., Khan, T., … Dow, L. E. (2026). A selective TNKS2 inhibitor for tumor-restricted WNT suppression. Nature Chemical Biology. https://doi.org/10.1038/s41589-026-02358-z

Image Credits: AI Generated

DOI: 10.1038/s41589-026-02358-z

Keywords: TNKS2, TNKS1, WNT signaling, chromosome 8p deletion, tankyrase inhibitor, collateral vulnerability, AXIN1 stabilization, cryo-EM structure, APC-mutant organoids, colorectal cancer, PARP family, drug selectivity