NexAb maps antibody binding across human tissues and monkey models
A new multiplex protein microarray platform is designed to answer a stubborn preclinical question: where do therapeutic antibodies bind outside their intended target, and how reliably does that binding translate across species? The system, called NexAb, combines a 7,015-spot tissue lysate array with antigen-oriented protein arrays to profile tissue binding and cross-reactivity before a drug reaches human studies. In practical terms, it gives developers a faster way to see whether an antibody recognizes the intended antigen cleanly or lights up unrelated tissues in a way that could complicate safety assessment.
The platform was applied to four anti-HER2 antibodies: trastuzumab, pertuzumab, trastuzumab emtansine (T-DM1), and trastuzumab deruxtecan (T-DXd). ERBB2 mRNA data were integrated into the analysis, adding a transcript-level reference point for interpreting protein binding patterns. The result is not simply a catalog of spots on a chip; it is a preclinical map of how antibody structure, epitope recognition, and payload chemistry can reshape tissue interaction.
On the antigen-oriented arrays, cell-synthesized HER2 bound trastuzumab and pertuzumab strongly and in an epitope-dependent manner. Wheat-germ HER2 did not bind. That contrast matters because it points to the importance of antigen presentation: when HER2 is produced in a cellular system, its conformation appears compatible with the antibody binding sites, while the wheat-germ product fails to reproduce the same molecular display. For antibody developers, this is more than a technical curiosity. It speaks to whether an assay is capturing the biology of the native target or a distorted version of it.
T-DM1 tracked trastuzumab closely, while T-DXd broadened the tissue signal
The tissue lysate array revealed a striking divergence between the anti-HER2 conjugates. Trastuzumab and T-DM1 bound HER2-positive cancer cell lines and tumor tissues with nearly identical profiles, with a correlation coefficient of r = 0.991. That degree of concordance suggests that adding the emtansine payload did not materially alter the antibody’s tissue recognition pattern in this assay. In a preclinical safety framework, such near-superimposable profiles make sense: the antibody backbone remains the main determinant of where binding occurs, and the conjugate behaves as a closely related version of the parent molecule.
The pattern changed sharply with T-DXd. This antibody bound ERBB2-low cancer spots and 38.8% of normal human tissue spots, a profile not conserved in cynomolgus monkeys. The broader signal in human tissue implies either expanded recognition of low-abundance or structurally distinct HER2-related targets, altered accessibility of epitopes, or tissue interactions influenced by the deruxtecan conjugate itself. In antibody-drug conjugates, payload chemistry can reshape distribution and apparent binding, especially when the linker-drug complex affects charge, hydrophobicity, or local tissue retention. Here, the signal was not confined to overt HER2-rich malignant tissue, which raises an immediate translational concern for preclinical toxicology programs.
One unexpected result stood out plainly: the human tissue signal for T-DXd was far wider than the monkey signal.
Species differences may complicate toxicology extrapolation
Trastuzumab, T-DM1, and pertuzumab showed no cross-reactivity with cynomolgus monkey tissues. T-DXd bound 3.1% of monkey tissue spots. The gap between human and monkey tissue reactivity is the most consequential finding in the dataset because nonhuman primates are often used to estimate on-target and off-target risk before first-in-human exposure. When a tissue-binding profile is not conserved across species, safety interpretation becomes less straightforward. A monkey model may underestimate the range of human tissues exposed to or recognized by a therapeutic antibody, particularly when the human signal arises from subtle differences in antigen expression, post-translational modification, or tissue architecture.
This is exactly where a multiplex platform becomes clinically and developmentally useful. Preclinical safety evaluation for antibodies has long depended on a combination of tissue cross-reactivity studies, target biology, and animal toxicology. Yet those methods are often limited by sparse sampling, low throughput, or a narrow view of the proteome. NexAb compresses those constraints into a quantitative assay that can compare human tumor, normal human tissue, and monkey tissue side by side. For programs advancing therapeutic antibodies and antibody-drug conjugates, the ability to distinguish conserved from nonconserved binding may influence species selection, dose planning, and the interpretation of unexpected organ toxicity later in development.
Epitope presentation and payload chemistry sit at the center of the assay readout
The findings also reinforce a broader biological principle: antibodies do not bind merely to a gene name, they bind to a molecular surface shaped by folding, glycosylation, and cellular context. The strong binding of trastuzumab and pertuzumab to cell-synthesized HER2, paired with the absence of binding to wheat-germ HER2, suggests that antigen source can determine whether an assay faithfully reproduces therapeutic recognition. In a preclinical setting, that can spell the difference between a useful safety screen and a misleading one.
The nearly identical profiles of trastuzumab and T-DM1 point to a different lesson. Payload attachment does not always disrupt specificity, especially when the antibody scaffold remains intact and the conjugate preserves the parent drug’s epitope preference. T-DXd behaves differently. Its broader binding footprint across human tissues suggests a more complex interaction surface, one that may be more difficult to predict from HER2 expression alone. In the context of drug development, that makes cross-reactivity profiling more than a checkbox exercise. It becomes a way to read the chemistry of the molecule against the anatomy of the host.
A preclinical tool with immediate safety relevance
NexAb is positioned as a quantitative platform for preclinical cross-reactivity profiling, and the dataset supports that role. The assay distinguishes strong target-aligned binding from broader tissue engagement, separates human from monkey reactivity, and shows that not all HER2-directed antibodies behave alike once they are moved from target validation into translational safety work. For oncology drug developers, that distinction can shape how a candidate is advanced and how much confidence can be placed in a given animal model.
The study remains preclinical and has not undergone peer review. Even so, its central message is direct: multiplex tissue binding assays can expose species-specific and molecule-specific behavior that standard target-centric readouts may miss. For antibodies such as trastuzumab deruxtecan, where human tissue binding extends well beyond the expected HER2-positive compartment, the assay creates a sharper boundary between intended pharmacology and potential off-target risk. Future work will need to determine how these binding patterns relate to in vivo toxicity and whether the same platform can be generalized across other therapeutic antibody classes.