T Cell Engager: Antengene Adds CD2 Co-Stimulation for Solid Tumors

T Cell Engager: Antengene Adds CD2 Co-Stimulation for Solid Tumors

Antengene’s T Cell Engager Adds a Second Signal to Reach Solid Tumors

Antengene has presented preclinical results for AnTenGager-D2, a costimulatory T cell engager platform, in a rapid oral abstract presentation at SITC 2026, and the conference selected the study among its TOP150 abstracts. The proposition is specific: CD3-based T cell engagers work in blood cancers and have struggled in solid tumours, and Antengene’s answer is to add a second co-stimulatory signal through CD2 while masking CD3 so the molecule stays inert until it reaches its target. The candidate built on the platform, ATG-109, targets KLK2 for prostate cancer. All of it is preclinical.

Why solid tumors resist T cell engagers

A CD3-based T cell engager works by bringing a T cell into contact with a tumour cell. One arm binds CD3 on the T cell, the other binds a tumour-associated antigen, and the forced proximity triggers the T cell to kill. That mechanism has produced approved medicines in B-cell malignancies, where the target is unambiguous and the tumour cells circulate accessibly.

Solid tumours present different problems. The microenvironment exhausts T cells that arrive, and a T cell engaged through CD3 alone receives signal one without signal two. In normal T cell biology, full activation requires both: an antigen signal through the T cell receptor and a co-stimulatory signal through a receptor such as CD28 or CD2. A T cell receiving only the first signal does not activate properly. It becomes anergic or dysfunctional.

Antengene’s argument is that this is a signalling problem rather than a delivery one, and that a T cell engager supplying only CD3 engagement asks a T cell to do something biology does not permit.

What AnTenGager-D2 adds

The platform integrates three elements: an antibody targeting the CD3γε and CD3δε heterodimers, a steric-hindrance mechanism masking CD3 until the molecule engages its tumour antigen, and an engineered binder for CD2 that supplies the second signal.

The sequence matters. The molecule binds its tumour-associated antigen first. That crosslinking event relieves the CD3 mask, allowing CD3 engagement, and the CD2 arm provides co-stimulation at the same moment. A T cell therefore receives both signals only where tumour antigen is present.

The specificity of the CD3 arm deserves a note. Screening 6,108 full-length human proteins and 403 human heterodimers using Retrogenix’s cell microarray technology, Antengene reports its parental anti-CD3 antibody bound the CD3δ+CD3ε and CD3γ+CD3ε heterodimers with high specificity. It also reports fast-on-fast-off binding kinetics in surface plasmon resonance and lower cytokine release than benchmark CD3 clones, which is the property that matters for the cytokine release syndrome limiting this drug class.

The platform lineage

AnTenGager-D2 extends an existing platform rather than opening a new company direction. AnTenGager is Antengene’s T cell engager 2.0 platform, described as featuring bivalent 2+1 binding for low-expressing targets, steric hindrance masking and proprietary CD3 sequences with fast on/off kinetics to reduce cytokine release. Seven programmes built on it appear in the pipeline, several with partners: CD19 × CD3 partnered with UCB, CDH6 × CD3 partnered with K2 Therapeutics established by MPM BioImpact, and programmes against ALPPL2, LY6G6D, GPRC5D, LILRB4 and FLT3.

The D2 variant adds CD2 co-stimulation to that architecture. That matters commercially as well as scientifically, because a platform a company has already partnered becomes more valuable when a second generation extends it, and the existing partnerships signal that the first generation passed someone else’s diligence.

The masking is the safety argument

Masking CD3 addresses the failure mode that has hurt this class. An unmasked T cell engager can engage T cells anywhere in the body, and the consequences run from circulating cytokine to systemic T cell activation. A molecule that stays inert until it meets its antigen concentrates the effect at the tumour.

Antengene reports limited antigen-independent T cell activation and no T cell-to-T cell crosslinking across the platform, and identifies one format, F2, as having the largest safety window among its DLL3-targeting constructs. Those are preclinical readouts in cell systems rather than in patients, and the gap between the two is where most T cell engager programmes have discovered their safety window was narrower than the assay suggested.

Co-stimulation adds a second safety question the release does not address. A molecule delivering both signals is more potent than one delivering only the first, and potency is what a therapeutic index measures.

ATG-109 and the prostate target

The candidate molecule is ATG-109, a KLK2 × CD2 × CD3 construct. KLK2 is a kallikrein expressed in prostate tissue, which makes it a rational target for prostate cancer and a crowded one, since several programmes pursue it.

The reported results follow a conventional preclinical package. ATG-109 showed lower binding to CD3-positive Jurkat cells in the absence of KLK2 crosslinking, which is the masking working. It showed enhanced T cell-dependent cellular cytotoxicity against the KLK2-positive VCaP line compared with a benchmark, and no non-specific cytotoxicity against the KLK2-negative Du145 line. In vivo, it produced potent efficacy in a human PBMC re-established prostate cancer cell-derived xenograft model.

The developability data are as consequential as the efficacy data for a molecule of this format. Antengene reports a melting temperature above 68 degrees Celsius and stability at 10 milligrams per millilitre by size-exclusion chromatography after four weeks at 40 degrees, with a half-life of about 183.5 hours in human FcRn mice. A construct with three binding arms is harder to manufacture and stabilise than a simple bispecific, and those figures are the first ones a development team reads.

The target choice

KLK2 is a reasonable choice of target and a difficult one. Prostate tissue expresses the kallikrein and its expression rises in prostate cancer, which gives a tumour-selective handle. It also sits in a field where PSMA has drawn most of the attention and most of the clinical investment, so a KLK2 programme competes not only against other KLK2 programmes but against a better-established target with imaging agents, radioligand therapies and cell therapies attached to it.

Antengene does not address that positioning. Antengene makes the preclinical case for ATG-109 against an unnamed benchmark rather than against the clinical landscape it would enter.

The preclinical ceiling

Everything in this release is preclinical, and the word appears in the headline, the opening paragraph and the conclusion. That is the right disclosure and it also defines the limit of what anyone can say.

T cell engagers carry a well-documented history of looking promising in mouse models and meeting different biology in humans. A xenograft model using re-established human PBMC is informative and it is not a patient: the T cells are human, the mouse is not, and the tumour microenvironment is engineered rather than grown. Cytokine release, which the platform aims to reduce, is a clinical phenomenon that first appeared in patients, and no assay fully predicts it.

The release also reports results across several constructs, including DLL3-targeting molecules that are not the lead candidate. That is normal for a platform presentation and it makes the headline claim harder to attribute to a single asset. The release also does not say whether the company has tested the co-stimulation approach in a model with an intact immune system, which is where the solid tumour problem lives. A PBMC re-established model introduces human T cells into a mouse that lacks them, while the exhaustion defeating T cell engagers in patients arises from chronic antigen exposure in a suppressive microenvironment, and that is the condition the platform must survive.

The company context

The company context shapes how to read the announcement. Antengene reports 35 investigational new drug approvals across the United States and Asia, and new drug application approvals in 10 Asia Pacific markets, with XPOVIO approved in mainland China, Taiwan, Hong Kong, Macau, South Korea, Singapore, Malaysia, Thailand, Indonesia and Australia and included in national insurance schemes in five of those. That is a commercial-stage company with a revenue-generating asset, not a discovery-stage startup presenting a poster.

It is also a company that has taken seven T cell engager programmes into a pipeline, several with partners, which means AnTenGager-D2 is a next step on a road already under construction rather than a standalone bet on a single molecule.

T Cell Engager: Antengene Adds CD2 Co-Stimulation for Solid Tumors
Photo by National Cancer Institute on Unsplash

What is missing

No link to the abstract or poster appears, so nobody can read the underlying data. The release describes the benchmark molecules without naming them, which prevents comparison. The release states no IND timeline, no first-in-human date and no clinical plan for ATG-109, and it does not say which indication a trial would pursue first or how the company would design it. The file as distributed also contains the release twice, with datelines of 8 and 9 November 2026, which is a distribution error rather than a content one, and worth flagging to the sender for that reason.

What to watch

Whether the abstract and poster become accessible, because a conference presentation without the underlying data is a claim rather than evidence. Whether ATG-109 reaches an IND filing and on what timeline, since a preclinical platform with a named lead candidate and one with a clinical plan are different propositions. And, hether co-stimulation produces a wider or narrower therapeutic index in humans, because potency cuts both ways. And whether the KLK2 prostate target proves differentiated against the other programmes pursuing it, which the release does not address at all and a reader will ask first.