Antibody-Lectin Chimeras: A New Path for Immunotherapy

Immunotherapy is now a standard option for nearly half of cancer patients in the US, but the hard truth is that fewer than 20% of treated patients respond to today’s best-known checkpoint inhibitors—and at least 25% of responders later relapse. Those numbers should bother anyone building cancer drugs, designing clinical strategies, or funding the next platform bet.

A paper published this week (December 2025) in Nature Biotechnology introduces a practical answer to a problem the industry has been circling for years: tumors don’t only hide behind protein checkpoints (PD-1/PD-L1, CTLA-4). They also hide behind glycans—dense sugar structures on the cell surface that can switch immune cells off through “glyco-immune checkpoints.”

The headline innovation is a new biologic format called antibody-lectin chimeras (AbLecs). What makes this especially relevant to the AI in Pharmaceuticals & Drug Discovery series is that AbLecs turn glycans from a “nice biology story” into something that looks engineerable, screenable, and optimizable—exactly the kind of design space where AI-assisted drug discovery can win.

Why most companies underestimate glyco-immune checkpoints

Answer first: Glyco-immune checkpoints matter because tumors use altered glycosylation to engage inhibitory lectin receptors (like Siglecs) on immune cells, suppressing killing even when protein checkpoints are blocked.

Cancer cells often become hypersialylated—meaning they display more sialic-acid–containing glycans on their surface. Those glycans bind inhibitory receptors such as Siglec-7 and Siglec-9 on macrophages, NK cells, and other immune players. When Siglecs engage their ligands at the immune synapse, they can blunt antibody-dependent cellular phagocytosis (ADCP) and antibody-dependent cellular cytotoxicity (ADCC).

This isn’t a fringe pathway anymore. Multiple studies (and growing clinical interest) point to the Siglec–sialic acid axis as an immune suppression mechanism across solid tumors and hematologic cancers.

The problem: glycans are hard to drug like proteins

Drug discovery teams run into three recurring walls:

1. Anti-glycan antibodies are scarce and tricky. Glycans are often weakly immunogenic, and specificity is hard.
2. Lectin receptors have many ligands. Blocking one ligand may not matter if a tumor uses five others.
3. Decoy receptors bind too weakly. Many lectin–glycan interactions sit in the μM to mM affinity range—far from what most therapeutics can exploit.

If you’re leading discovery, this is why glycan biology often dies at the “cool target—no modality” stage.

What AbLecs are—and why the architecture is the real invention

Answer first: AbLecs work because they use a high-affinity antibody arm to concentrate a low-affinity lectin domain on the tumor surface, enabling glycan blockade at nanomolar dosing levels.

An AbLec combines:

• A tumor-targeting antibody (e.g., trastuzumab against HER2)
• A glycan-binding domain from a lectin immunoreceptor (e.g., Siglec-7 or Siglec-9 extracellular domain fused to Fc)

The key idea is local concentration. Even if the Siglec domain binds glycans weakly in free solution, the antibody arm anchors the molecule on the tumor cell. That creates a high effective local concentration of the lectin domain at the same membrane—enough to occupy inhibitory glycans at therapeutically realistic (nanomolar) concentrations.

In the paper’s HER2 example:

• Siglec-7/9 decoy receptors alone show low binding even at high concentrations.
• Trastuzumab × Siglec-7/9 AbLecs bind HER2+ tumor cells with low-nanomolar apparent affinity, comparable to the parent antibody.

This is an important design principle: AbLecs don’t need to “fix” lectin affinity. They route around it using multivalency and proximity.

Blocking at the synapse, not systemically

AbLecs also appear to deliver a gain-of-function: they can exclude Siglecs from the immunological synapse. That matters because inhibitory signaling often depends on synapse-local receptor clustering, not just receptor occupancy somewhere on the cell.

This is one of those mechanistic details that changes development strategy:

• Systemic blockade (e.g., circulating Siglec antibodies) can be effective but risks broader immune perturbation.
• Tumor-local blockade at the synapse can raise efficacy while plausibly improving safety.

What the experiments show (and what they imply for translation)

Answer first: In vitro and in vivo data suggest AbLecs can boost antibody effector functions beyond the parent antibody and may synergize with established checkpoint strategies.

The authors built AbLecs pairing trastuzumab with Siglec-7 (T7) or Siglec-9 (T9). They then asked a question drug teams care about: Does this make immune killing stronger than the parent antibody?

Stronger ADCP and ADCC across immune effectors

Across primary human immune cell assays:

• Macrophages: AbLecs increased phagocytosis (ADCP) versus trastuzumab.
• NK cells: T7 AbLec increased cytotoxicity (ADCC) versus trastuzumab.
• Neutrophils (PMNs): T7 and T9 improved tumor lysis compared to trastuzumab.

Crucially, the enhancement was shown to be:

• FcγR-dependent (so it’s still anchored in classic antibody biology)
• Siglec–glycan dependent (the benefit disappears if Siglecs are blocked already or tumor sialic acids are enzymatically removed)

So the story isn’t “a fancy bispecific is better because bispecific.” It’s “you removed a brake at the synapse while keeping the gas pedal.”

Evidence of in vivo benefit (in a humanized model)

Mouse modeling is notoriously awkward for Siglec biology because mouse Siglecs don’t map cleanly to human Siglec-7/9. The team used humanized Siglec-7/9 and human FcγR mice and showed reduced lung metastatic burden—particularly with the Siglec-9 AbLec.

That’s meaningful because it suggests the mechanism survives contact with in vivo complexity.

Synergy with CD47 blockade

AbLecs also showed synergy with CD47 blockade in phagocytosis assays. Practically, this positions AbLecs as part of a trend: myeloid checkpoint combinations rather than yet another PD-1 add-on.

And it points to a commercial logic: if AbLecs can enable lower-dose CD47 blockade, they might help mitigate some of the toxicity and program failures that have plagued high-dose CD47 strategies.

Where AI-assisted drug discovery fits (and why it’s not just a buzzword here)

Answer first: AbLecs create a modular design space—antigen arm, lectin domain, Fc engineering, geometry—where AI can speed candidate selection, predict developability, and optimize efficacy-risk tradeoffs.

Glyco-immune checkpoints are messy because the ligand landscape is messy. That’s exactly where AI approaches tend to outperform human intuition—provided the modality is buildable.

Here are practical, high-leverage AI opportunities the AbLec concept opens up.

1) Target selection: matching tumors, antigens, and glyco-checkpoints

A core AbLec decision is: which tumor antigen should localize the chimera?

AI can help prioritize targets by combining:

• Tumor antigen density and heterogeneity (single-cell, spatial omics)
• Immune infiltration patterns and FcγR-bearing effector presence
• Glycosylation signatures (proteomics, glycomics, lectin staining proxies)

The goal isn’t merely “pick HER2.” It’s “pick an antigen that co-localizes AbLecs to the right cells in the right microenvironment for the right effector mechanism.”

2) Architecture search: geometry is a parameter, not an afterthought

AbLecs are an architectural idea: one arm binds a protein antigen strongly, the other binds glycans weakly but effectively through proximity.

That means geometry and avidity matter. AI-guided design can explore:

• Linker lengths and flexibility
• Fc formats and self-assembly constraints
• Valency and spatial reach to maximize synapse-local blockade

This is one of the few times “structure-guided modeling” isn’t just a slide—because the synapse-local mechanism makes spatial organization central to potency.

3) Developability and manufacturing prediction

Bispecific-like constructs can die on developability: aggregation, expression yield, stability, immunogenicity.

Machine learning models for developability can triage candidates early by predicting:

• Aggregation risk from sequence/structure features
• Viscosity and formulation risk
• Expression and purification likelihood

This matters for AbLecs because platform scalability will decide whether they remain a “cool paper” or become a pipeline.

4) Biomarker strategy: who should get an AbLec first?

If you’re thinking about clinical translation, biomarkers are not optional.

AI can support companion biomarker design by identifying predictors of benefit, such as:

• High tumor sialylation signatures
• High Siglec-7/9 ligand expression proxies
• Myeloid-rich microenvironments where ADCP is a dominant killing route

A blunt but useful stance: AbLecs should not go into all-comers trials. The mechanism is specific enough that enrichment is realistic—and likely necessary.

What to do with this as a drug discovery leader

Answer first: Treat AbLecs as a platform for tumor-local glycan blockade and build a feasibility plan around target selection, synapse biology, and biomarker-driven development.

If you’re in pharma/biotech discovery or translational strategy, here’s a practical checklist I’d use to evaluate AbLecs as an internal program or partnership target.

A pragmatic AbLec evaluation checklist

1. Is the tumor antigen actually targetable in your indication? Look for surface density, internalization behavior, and prior antibody success.
2. Is there evidence the Siglec axis is active in that tumor microenvironment? Prioritize indications with myeloid suppression signatures.
3. Which effector function do you need most—ADCP, ADCC, or both? This informs Fc choices and model systems.
4. Do you have a biomarker plan that can be deployed before Phase 2? If not, you’re building risk into the timeline.
5. Can you manufacture it reliably? Platform formats fail fast when CMC doesn’t keep up.

Where AbLecs could go next

AbLecs look extensible in two valuable directions:

• More tumor antigens: the paper demonstrates HER2, CD20, EGFR.
• Dual checkpoint designs: examples include PD-1 × galectin-9 and PD-L1 × Siglec-7 formats.

The highest upside bet is a clinically validated combination: tumor targeting + glyco-checkpoint blockade + a known checkpoint backbone (PD-1/PD-L1, CD47, etc.), guided by biomarkers and optimized with AI-assisted drug discovery workflows.

The bigger point for this topic series is simple: AI isn’t just accelerating “more of the same” antibodies. It’s increasingly about making new therapeutic mechanisms practical—by navigating complexity that humans can’t brute force.

If glyco-immune checkpoint blockade becomes a real pillar alongside PD-1 and CD47, AbLecs may be remembered as one of the formats that made it manufacturable, targetable, and testable.

If you’re building in this space, the question worth asking now isn’t whether glycans matter. It’s whether your discovery stack—data, models, and platform strategy—is ready to treat glycans like first-class drug targets.