Helixon Develops First-in-Class CDH17/CEACAM5 Trispecific T-Cell Engager for Advanced Colorectal Cancer

For patients with advanced colorectal cancer who have exhausted standard therapies, the options are grim. The objective response rate for single-agent treatments in third-line metastatic colorectal cancer stands at a mere 1% to 2%. Even the latest Phase III data combining trifluridine-tipiracil with bevacizumab achieved only a 6.1% response rate—a reminder of how stubbornly this disease resists treatment.

Now Helixon, a Beijing-based biotech formerly known as Huashen Zhiyao, is betting that the answer lies not in targeting one cancer protein, but two. The company has developed what it calls the first trispecific T-cell engager designed to simultaneously hit CEACAM5 and CDH17—two molecules that frequently appear together on colorectal tumor cells but have largely been pursued separately by drugmakers.

The approach addresses a fundamental problem in oncology: tumors are heterogeneous. Not every cancer cell expresses the same targets at the same levels. A drug designed to hit one antigen may work initially, only for resistant cells lacking that antigen to multiply and drive relapse. By engineering a molecule that can bind two different targets, Helixon aims to cast a wider net.

A Delicate Balancing Act

The molecule, designated tsTCE3, carries three binding arms: one Fab fragment targeting CEACAM5, and two single-chain variable fragments targeting CDH17 and CD3, the latter being the receptor on T cells that triggers killing. The architecture evolved through several iterations.

Early versions used a 2Fab-plus-1scFv format. The company tested two configurations that differed only in the order of the tumor-targeting arms. The version with CEACAM5 first, then CDH17, showed stronger killing activity than the reverse order—and outperformed a cibisatamab-like molecule, an existing CEACAM5-targeting bispecific. The spatial arrangement of antibody arms, it turns out, is not interchangeable.

The final design swapped one Fab for a scFv, creating a 1Fab-plus-2scFv format. This change, combined with affinity fine-tuning of each arm, produced a roughly 18-fold increase in killing activity compared with the cibisatamab-like comparator in HT55 colorectal cancer cells.

Division of Labor

The three arms don't contribute equally. In cells expressing both targets at high levels, the CDH17 arm drives most of the killing activity, while the CEACAM5 arm primarily enhances tumor binding and immune synapse formation. Binding experiments showed that molecules carrying the CEACAM5 arm bound to tumor cells 10 to 12 times more strongly than those relying on CDH17 alone.

This division of labor becomes critical when tumors vary in antigen expression. Testing across a panel of cell lines with different expression patterns—some high in CEACAM5 but negative for CDH17, others the reverse—showed that tsTCE3 maintained killing activity comparable to or better than bispecific antibodies in each scenario.

The Real Test: Low CDH17 Tumors

The true value of dual targeting emerged in mouse models. Using human PBMC-reconstituted NCG mice implanted with human tumor xenografts, Helixon tested three models dosed at equimolar levels with twice-weekly injections for six doses.

In HT55 tumors expressing both targets at high levels, tsTCE3 achieved a tumor growth inhibition of 77.97%, slightly below the 83.09% seen with a CDH17 bispecific that targets two different epitopes on the same molecule. The company was careful not to claim superiority over that bispecific, noting the comparison was limited to cibisatamab-like and CEACAM5 bispecific molecules.

In SW403 tumors with low CEACAM5 expression, both tsTCE3 and the CDH17 bispecific achieved roughly 59% TGI—essentially equivalent.

But in LS174T tumors with low CDH17 expression, the trispecific's advantage became clear. TsTCE3 achieved 94.5% TGI, significantly outpacing the CDH17 bispecific's 74.6%, and far exceeding the cibisatamab-like molecule at 29.0% and the CEACAM5 bispecific at 28.1%.

That gap is precisely what the CEACAM5 arm was designed to address: when CDH17 expression is insufficient to drive killing on its own, the second target picks up the slack.

Safety by Design

The affinity of each arm was deliberately kept moderate, ranging from 40 to 52 nanomolar, with the CD3 arm at 51.9 nM and monovalent to avoid overstimulation. Compared with cabotamig, Arbele's CDH17-CD3 bispecific that uses bivalent CD3 binding, tsTCE3 showed more restrained cytokine release. Cabotamig's analog triggered the strongest T-cell activation and killing, but also elevated TNF-α and IFN-γ levels substantially. TsTCE3 maintained moderate cytokine release while still achieving potent tumor killing—a trade-off the company views as a safety feature.

The anti-CEACAM5 antibody was selected for its specificity: it binds the B3 domain of CEACAM5 and shows no cross-reactivity with CEACAM1, CEACAM6, or CEACAM8, family members widely expressed on normal epithelial and granulocyte cells. The CDH17 antibody was chosen through functional screening across multiple structural domains, ultimately selecting one that binds the EC5 domain.

For Helixon, the trispecific represents a proof of concept that structural optimization—shifting formats, fine-tuning affinities, and ordering arms deliberately—can meaningfully enhance efficacy without resorting to maximum-strength binding. Whether that translates into clinical benefit for colorectal cancer patients, who have so few options left, remains to be seen in human trials.