Engineering Precision to Overcome On-Target Off-Tumor Toxicity

This direction is split into other directions: on-target off-tumor toxicity. Its last version stays readable and citable here.

Version 1 of 150 papers58 conceptsfrozen 2026-10-04

The central challenge of modern cancer immunotherapy is not just killing cancer cells, but distinguishing them from the healthy "self." Most potent tumor antigens are not unique to the malignancy; they are often proteins that are simply overexpressed in tumors or shared with vital organs. This lack of absolute specificity leads to on-target, off-tumor toxicity, where engineered immune cells or antibodies attack healthy tissues like red blood cells, kidneys, or the bone marrow.

Researchers are now moving beyond simple antigen recognition, developing sophisticated bioengineering strategies—from affinity tuning and logic-gated receptors to targeting tumor-specific post-translational modifications—to widen the therapeutic window.

The Problem of the Antigen Sink

Many promising targets, such as the "don't eat me" signal CD47, suffer from what is known as an "antigen sink." Because CD47 is ubiquitously expressed on healthy red blood cells and platelets, traditional high-affinity antibodies are "soaked up" by these healthy cells before they can reach the tumor. As Chenyang Jiang and colleagues note in Targeting the CD47/SIRPα pathway in malignancies, this leads to severe systemic side effects like hemolytic anemia and hemagglutination, which have caused several high-profile clinical failures.

To bypass this, researchers like Hang Ke and Dianze Chen are employing affinity-balancing. In the development of the bispecific antibody HX009 and other CD38 x CD47 agents, the antibodies are engineered with low affinity for the ubiquitous target (CD47) but high affinity for a second, more tumor-restricted marker (like PD1 or CD38). This ensures the drug only binds strongly when both targets are present, effectively "ignoring" the single-target healthy cells.

Exploiting the Tumor's "Sugar Coating"

If a protein's amino acid sequence is identical in both healthy and cancerous cells, how can a drug tell them apart? The answer often lies in glycosylation—the way cells attach sugar chains to proteins.

Tumors often have "stripped-down" or aberrant glycosylation patterns. Researchers have leveraged this by developing antibodies like PODO447, which targets a tumor-restricted glycoepitope on the protein podocalyxin. While podocalyxin is essential for kidney function, Diana Canals Hernaez demonstrated that PODO447-ADC could eliminate ovarian tumors without damaging the kidneys because it only recognizes the specific "sugar signature" found in the tumor microenvironment. Similarly, Jun-Li Liu developed NXA01, an antibody that targets a hypoglycosylated version of CD24, sparing healthy B cells and granulocytes.

Programmable Immunity and Logic Gates

When a single marker isn't enough, synthetic biology allows us to treat immune cells as programmable computers. As Bo Zhu and Ming-Ru Wu explain in Synthetic biology approaches for improving the specificity and efficacy of cancer immunotherapy, we can now engineer CAR-T cells with logic gates.

An "AND" gate, for example, requires the T cell to recognize two different antigens before it triggers a kill signal. This prevents the "on-target, off-tumor" destruction of organs that might express only one of those antigens. Other approaches include "remote-controlled" systems, such as the doxycycline-inducible CAR-T cells for liver cancer, which allow clinicians to toggle the therapy's activity on and off to manage toxicity in real-time.

Mapping the Immune Blind Spots

To design better receptors, we must first understand what the immune system already knows. Gurinder S. Atwal and Lili M. Blumenberg have used single-cell RNA sequencing of the human thymus to identify genes that are "missing" from central tolerance. These genes represent a "blind spot" in immune training; because the body hasn't learned to tolerate them, they make ideal targets for immunotherapy with a lower risk of triggering autoimmune cross-reactivity.

This structural understanding is being paired with deep learning frameworks like LSMTCR, developed by Ruihao Zhang, which designs T-cell receptors (TCRs) from scratch. By computationally modeling the binding energy between a receptor and the MHC-peptide complex, researchers can ensure the receptor binds to the specific cancer mutation rather than the conserved MHC framework shared by all healthy cells, as highlighted by Matthew I. J. Raybould.

Papers in this version

History & citationVersion 1 · 2026-10-04

Split into other directions · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_engineering_precision_to_overcome_on_tar_6dea0ea2,
  title = {Engineering Precision to Overcome On-Target Off-Tumor Toxicity},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/engineering-precision-to-overcome-on-target-off-tumor-toxicity/rdr_6dea0ea2faf049629bcbb8f0424ad687},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-21},
}

This exact version (version 1, frozen text):

@misc{lacuna_engineering_precision_to_overcome_on_tar_6dea0ea2_v1,
  title = {Engineering Precision to Overcome On-Target Off-Tumor Toxicity},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/engineering-precision-to-overcome-on-target-off-tumor-toxicity/rdr_6dea0ea2faf049629bcbb8f0424ad687/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-21},
}