The central player on the tumor side of this story is HSP90, a heat shock protein that functions as a molecular chaperone. Chaperones are the quality-control workers of the cell: they help newly made proteins fold into their correct three-dimensional shapes, stabilize damaged or mutated proteins that would otherwise be degraded, and buffer the cell against stresses ranging from heat to chemical toxicity. In cancer, HSP90 takes on an outsized role. Because tumor cells accumulate mutations and copy-number changes that produce unstable, misfolded proteins, they become unusually dependent on chaperone buffering to keep their aberrant signaling machinery running. This phenomenon, sometimes described as adaptive proteostasis buffering, allows evolving tumors to survive therapeutic pressure, including exposure to multikinase inhibitors used against hepatocellular carcinoma, the most common form of primary liver cancer.

The researchers asked a deceptively simple question: if HSP90 is what allows tumors to adapt under drug pressure, what happens to the tumor’s stress-response system when that buffering is deliberately disrupted? Using evolving tumor models of multikinase inhibitor-resistant hepatocellular carcinoma, together with patient-derived samples, they found that low-dose HSP90 inhibition produced a striking and precise effect. Rather than simply killing the cells or broadly denaturing their proteins, the treatment triggered the relocalization of a second chaperone, glucose-regulated protein 78, better known as GRP78, from its usual home in the endoplasmic reticulum to the outer surface of the tumor cell.

This relocalization matters because of where GRP78 normally sits. In healthy cells, GRP78 is an endoplasmic reticulum resident protein, part of the machinery that monitors protein folding inside the organelle and initiates the unfolded protein response when folding capacity is overwhelmed. It is not supposed to be visible on the cell surface. Yet under severe stress, some GRP78 escapes the endoplasmic reticulum and appears on the plasma membrane, where it can act as a signaling receptor and promote survival. Crucially, because surface GRP78 is absent from most normal adult tissues but enriched on stressed, evolving tumor cells, it has long been an attractive candidate target for antibody-based therapies. The problem has been getting enough of it onto the cell surface, reliably and selectively, to make such targeting effective.

The new study identifies the trafficking mechanism behind this relocalization. The researchers found that the movement of GRP78 to the tumor cell surface depends on DNAJC1, a member of the DnaJ heat shock protein family, also known as Hsp40. These co-chaperones typically assist HSP70-family proteins, and in this setting DNAJC1 appears to act as the trafficking factor that escorts GRP78 out of the endoplasmic reticulum and toward the plasma membrane when HSP90-dependent buffering is knocked down. When the team blocked HSP90 at low doses, the resulting proteostatic stress activated this DNAJC1-dependent route, converting GRP78 from an intracellular housekeeping protein into a surface-exposed marker of the stressed tumor state.

The authors give this relocalized protein a new conceptual name: a stress-induced tumor antigen, abbreviated SITA. The idea is that the very stress-response programs tumors deploy to survive therapy can be rewired so that they display targetable molecules on their surface. In other words, the adaptive machinery that normally confers drug resistance becomes a beacon for immune attack. This reframing is what distinguishes the study from earlier work on surface GRP78. Instead of relying on whatever baseline amount of the protein happens to be present on a tumor, the approach actively induces the antigen with a pharmacological intervention, then attacks it with engineered immune cells, creating a sequential strategy in which the first treatment sets up the second.

The immunological half of the strategy relies on CAR-based effector cells. Chimeric antigen receptors are synthetic receptors engineered into a patient’s own or donor-derived T cells or natural killer cells, allowing them to recognize a specific surface antigen and kill the cell carrying it. CAR-T cell therapies have produced dramatic results in blood cancers, but solid tumors have proven far harder, in part because researchers struggle to find antigens that are abundant on tumor cells and absent from essential healthy tissue. The Rutgers team generated GRP78-directed CAR-NK cells and tested them against tumor cells whose surface GRP78 had been induced by HSP90 inhibition. The stress-induced surface GRP78 made the resistant hepatocellular carcinoma cells markedly more susceptible to CAR-NK-mediated killing, turning a previously inaccessible intracellular chaperone into a legitimate immunotherapy target.

The findings were not confined to cell culture. In xenograft models, in which human tumor cells are implanted into immunodeficient mice, and in syngeneic models, in which intact immune systems are present, the combination of proteostasis perturbation and CAR-based therapy produced stronger anti-tumor effects than either approach alone. Disrupting the HSP90-GRP78 axis enhanced the activity of both CAR-NK and CAR-T cells. Perhaps most intriguingly, the treatment reshaped the tumor microenvironment itself: the researchers observed increased infiltration of natural killer cells and macrophages within tumor niches that expressed high levels of surface GRP78. This suggests that inducing the stress antigen does more than provide a docking site for engineered cells; it appears to create immunologically favorable niches that recruit and support innate immune players as well.

The clinical logic of the approach is worth unpacking. Multikinase inhibitor-resistant hepatocellular carcinoma represents a setting with few good options, and the study’s use of evolving, drug-resistant models means the strategy was tested against exactly the kind of tumors that defeat conventional targeted therapy. Because the HSP90 inhibitor is used at low doses, the goal is not maximal cytotoxicity, which has historically limited chaperone inhibitors in the clinic due to toxicity in normal tissues, but rather a calibrated perturbation that induces surface GRP78 without wholesale proteotoxic collapse. The selectivity then comes from the CAR step, since the engineered cells attack only cells displaying the induced antigen. In principle, this two-step design separates the toxicity of the sensitizing drug from the specificity of the immune kill.

As with any preclinical study, the path from these models to patients will require careful translation. Dosing schedules, the kinetics of GRP78 relocalization, the behavior of the strategy in heterogeneous human tumors, and the safety of combining chaperone inhibition with adoptive cell therapy all remain to be established in trials. The authors themselves frame the work as uncovering a previously unrecognized concept rather than a ready-made regimen. Still, the study, published open access on 3 October 2026 with the DOI 10.1186/s12943-026-02807-2, adds a genuinely new idea to the solid-tumor immunotherapy toolbox: that tumor evolution under therapeutic pressure, the very process that produces drug resistance, can be hijacked to display a stress-induced tumor antigen, and that CAR-engineered immune cells can be aimed at that antigen to destroy tumors that thought they had already adapted to everything medicine could throw at them.

Subject of Research: HSP90 inhibition induces DNAJC1-dependent GRP78 relocalization to the tumor cell surface, creating a stress-induced tumor antigen that enhances CAR-based immunotherapy in drug-resistant hepatocellular carcinoma.

Article Title: HSP90 inhibition drives GRP78 relocalization to potentiate CAR-based immunotherapy during tumor evolution

Article References: Sabha, Y., Cho, J. H., Zhang, X., Haggerty, J., Ai, A., Mayorga, A., Rajpoot, S., Chen, C.-H., Bulatowicz, J. J., Byrne, M., Cruz, M. A., Xynogala, A., Liu, C., Chen, F., Gause, W. C., Wang, X., Fu, K., Fritzky, L., Qin, Y., … Liu, D. (2026). HSP90 inhibition drives GRP78 relocalization to potentiate CAR-based immunotherapy during tumor evolution. Molecular Cancer. https://doi.org/10.1186/s12943-026-02807-2

Image Credits: AI Generated

DOI: 10.1186/s12943-026-02807-2

Keywords: HSP90, GRP78, CAR-NK cells, CAR-T cells, hepatocellular carcinoma, proteostasis, molecular chaperones, tumor evolution, drug resistance, immunotherapy, DNAJC1, stress-induced tumor antigen