AbLecs: Targeted Glyco-Checkpoint Blockade for Cancer

Checkpoint inhibitors are now used in nearly half of US cancer patients, yet fewer than 20% benefit—and even among responders, at least 25% develop resistance. That gap isn’t a “more PD-1” problem. It’s a target problem.

A Nature Biotechnology paper published this week proposes a new target class and a new drug architecture to go after it: antibody-lectin chimeras (AbLecs) for glyco-immune checkpoint blockade. I think it’s one of the most practical ideas we’ve seen in a while for turning glycobiology from “interesting” into “buildable.”

What makes it especially relevant for pharma and drug discovery teams right now: AbLecs don’t just add another checkpoint. They make the tumor glycocalyx—a messy, heterogeneous layer that’s historically been hard to drug—behave like a tractable target. And that’s exactly the kind of complexity where AI-driven drug design can earn its keep.

Glyco-immune checkpoints: the suppression pathway most teams ignore

Cancer doesn’t only hide by flipping protein checkpoints (PD-L1, CTLA-4 ligands). It also hides by rewriting the sugars on its surface.

The “glyco-code” is immunology’s quiet control panel

Tumors often upregulate sialoglycans (sialic-acid–containing glycans). These bind inhibitory lectin receptors on immune cells—especially Siglecs (notably Siglec-7 and Siglec-9)—and dial down immune attack.

Here’s the practical implication: even if you recruit an immune effector cell with a therapeutic antibody, Siglec engagement at the immune synapse can blunt antibody effector functions like:

• ADCP (antibody-dependent cellular phagocytosis; macrophage “eat”)
• ADCC (antibody-dependent cellular cytotoxicity; NK cell “kill”)

That suppression is one reason “good antibodies” can look mediocre in certain tumor microenvironments.

Why glycan targets have stayed out of reach

Most companies get stuck here because glycans are hard targets:

1. Weak immunogenicity makes high-affinity, highly specific anti-glycan antibodies hard to generate.
2. Many lectin receptors have multiple ligands, and in cancer the identity of the relevant ligand can be unknown.
3. “Decoy receptor” lectin-Fc fusions keep native glycan recognition—but their affinities are typically µM to mM, far too weak for therapeutic use.
4. Systemic checkpoint blockade can create immune toxicity; existing checkpoint inhibitor toxicities occur in 50%+ of treated patients in some settings.

So the field has had a frustrating trade-off: broad biology, poor druggability.

What AbLecs are—and why the architecture matters

AbLecs are chimeric biologics that couple a tumor-targeting antibody (binding a surface antigen like HER2, CD20, EGFR) with a lectin receptor glycan-binding domain (like Siglec-7 or Siglec-9).

The idea is deceptively simple:

• The antibody arm provides high-affinity, antigen-specific localization to the tumor cell.
• That localization creates a high local concentration of the lectin binding domain right where the tumor’s inhibitory glycans sit.
• The lectin domain can then block glyco-immune checkpoint engagement at therapeutically relevant (nanomolar) doses—even if it’s intrinsically low affinity.

This is the key shift: AbLecs make a weak glycan interaction strong enough in context via avidity and proximity, rather than trying to “fix” glycans with an unrealistically perfect anti-glycan antibody.

Proof-of-concept: trastuzumab × Siglec AbLecs

The paper’s flagship examples fuse trastuzumab (HER2 antibody) with Siglec-7 or Siglec-9 domains:

• T7 AbLec (trastuzumab × Siglec-7)
• T9 AbLec (trastuzumab × Siglec-9)

In HER2+ cell models expressing Siglec ligands, these AbLecs:

• Bound tumor cells at low-nanomolar apparent K_D, similar to trastuzumab.
• Competed with fluorescent Siglec decoy receptors, effectively blocking Siglec engagement.
• Enhanced macrophage phagocytosis and NK/PMN killing compared with trastuzumab alone.

In a humanized in vivo lung colonization metastasis model (human Siglec-7/9 plus human FcγR biology), T9 reduced metastatic burden versus trastuzumab.

That’s not just incremental. It shows the platform can translate beyond a dish—while respecting the messy species differences in Siglec biology.

The “gain-of-function” insight: AbLecs work at the immune synapse

The most interesting mechanistic claim isn’t “bispecific binds two things.” It’s where the blockade happens.

Local blockade beats systemic blockade

When they benchmarked AbLecs against:

• trastuzumab + soluble Siglec-7-Fc
• trastuzumab + Siglec-blocking antibodies
• trastuzumab + enzymatic desialylation (sialidase)

…AbLecs held up extremely well and in some assays outperformed combinations, especially versus Siglec-blocking antibodies.

Their explanation: AbLecs exclude inhibitory Siglecs from the immunological synapse—the tight interface where FcγR signaling and “kill/eat” decisions occur.

That’s a strong design principle for next-gen immunotherapies:

If inhibition is organized spatially at the synapse, your drug should be designed to win spatially—not just biochemically.

This is also where safety could improve. A drug that blocks glyco-checkpoints only when bound to a tumor antigen has a very different risk profile than systemic glycan editing or systemic lectin blockade.

Why AbLecs are a big deal for AI in drug discovery

Glycosylation is high-dimensional biology. AbLecs turn it into something AI can optimize with clear objectives.

1) AI can model the glycan landscape AbLecs are exploiting

The AbLec concept is “agnostic to exact glycan identity,” but development programs aren’t. You still need to decide:

• Which tumors are hypersialylated (or galectin-driven)?
• Which immune compartments express the relevant inhibitory receptors (Siglec-7 vs Siglec-9 vs Siglec-10)?
• Which patient subgroups show glyco-checkpoint signatures tied to resistance?

AI helps by integrating multi-omic and imaging signals into actionable biomarkers:

• glycoproteomics / glycomics
• single-cell RNA + spatial transcriptomics
• digital pathology features correlated with myeloid suppression
• clinical response labels from checkpoint therapy

The output pharma teams actually need is not “a glycan atlas.” It’s a deployable rule like:

• “High Siglec-9 ligand activity + neutrophil-rich lung metastasis → prioritize Siglec-9 AbLec.”

2) AI can optimize AbLec design variables faster than trial-and-error

AbLecs introduce tunable parameters that are perfect for computational design loops:

• antibody target choice (HER2 vs EGFR vs CD20)
• lectin domain choice (Siglec-7/9/10, galectins)
• Fc engineering (effector function balance)
• geometry and linker/topology that governs synapse exclusion
• avidity vs manufacturability trade-offs (aggregation, stability)

The study even uses a structure-guided multivalent interaction model to explain why AbLecs are potent: a massive increase in effective concentration for the second binding interaction (a proximity/avidity effect). That modeling mindset is exactly where modern AI/ML approaches can scale: simulate thousands of architectures before you express the top 20.

3) AbLecs give a credible path to overcoming immunotherapy resistance

Resistance to PD-1/PD-L1 is often framed as T-cell exhaustion or antigen presentation failure. Those are real—but myeloid suppression is a major driver, and glyco-checkpoints sit right in that lane.

The paper demonstrates synergy between glyco-checkpoint blockade and established checkpoints like CD47/SIRPα (and designs dual blockade AbLecs that include PD-1 or PD-L1 components).

Clinically, that matters because combination therapy is the norm—but toxicity is the limiter. If AbLecs allow lower-dose checkpoint combinations by localizing suppression relief to the synapse, that’s not just efficacy. It’s a development strategy.

Practical takeaways for pharma teams evaluating AbLecs

If you’re screening this platform for a pipeline fit, these are the questions I’d start with.

1) Pick indications where myeloid effector function matters

AbLecs are built to improve Fc-mediated killing. That suggests strong fit for:

• antibody-driven solid tumor programs where ADCC/ADCP are key
• myeloid-heavy microenvironments (certain metastasis sites)
• settings where CD47 combos are already being explored

2) Treat antigen heterogeneity as a design constraint, not an exclusion

A pleasant surprise in the data: AbLecs retained improved function even with low antigen expression in engineered models. That hints they may be useful where antigen density is a known failure mode.

But the business reality remains: target selection has to balance tumor coverage and off-tumor expression. AbLecs don’t remove that constraint—they raise the value of choosing well.

3) Build biomarker strategy early (glyco-checkpoints aren’t “one test”)

Glyco-immune checkpoint blockade will likely require composite biomarkers. You’ll want a plan that measures:

• tumor antigen level
• Siglec ligand activity (direct glycan assays or proxies)
• immune context (Siglec receptor expression on infiltrates)

AI-powered biomarker discovery is not optional here; it’s how you avoid running expensive trials in biologically mismatched patients.

Where AbLecs could go next

This paper positions AbLecs as modular, and the modularity is real. But the next wave of value will come from program decisions, not the concept.

• Which lectin axes matter most by tumor type? Siglec-9 may dominate in some contexts; Siglec-7 in others.
• Which combinations are worth the toxicity budget? CD47 has a history here; AbLecs might enable lower, safer doses.
• Can we predict synapse-level behavior from structure? That’s where design and AI meet: geometric constraints, multivalency, and spatial exclusion effects.

The big bet is that glycosylation becomes a controllable variable in immuno-oncology drug design, not an afterthought.

If you’re building an AI in pharma roadmap, AbLecs are exactly the kind of platform that benefits from AI: high-dimensional biology, modular design space, clear functional assays, and immediate clinical relevance.

The question isn’t whether glyco-immune checkpoints matter. The question is whether we’ll finally build drugs that can target them precisely.