MHC one–two punch knocks out cancer

CPIs target molecular “brakes” that limit T cell activation, yet most tumors remain resistant—even those with heavy mutation burdens that generate abundant neoantigens. Most tumors resist CPI therapy, even those with high mutation burdens that generate numerous modified peptides that should escape self-T cell tolerance. Paradoxically, some tumors with fewer mutations can be effectively treated by CPI therapy (Chae et al., 2019).

CD4⁺ T cells are increasingly recognized as key orchestrators of effective anti-tumor immunity. They license dendritic cells (DCs), potentiate CD8⁺ T cell activation, and remodel the tumor microenvironment through cytokine secretion and costimulatory signals (Ahrends et al., 2017; Ferris et al., 2020). Without CD4⁺ help, CD8⁺ T cells enter an exhausted state—expressing PD-1 and other inhibitory receptors—and lose proliferative and cytotoxic capacity (Miller et al., 2019). A central challenge in cancer immunotherapy has, therefore, been to coordinate these two T cell arms. Zhang et al.’s elegant solution is to fuse their target peptides to ensure they are presented by a single DC.

Zhang et al. (2025) engineered tumor cells to express a complete T cell antigen (CTA): a single polypeptide with immunogenic class I – and II–restricted epitopes. When mice were inoculated with these CTA-expressing cells, even contralateral tumors lacking the antigen were rejected following CPI treatment—a striking, target antigen-independent adjuvant effect.

This potentiation vanished when the class I and II peptides were expressed as separate proteins or when given as exogenous protein immunogens or peptide-pulsed splenocytes, despite these also activating CD8⁺ T cells. The data imply that antigen linkage within the same expressed protein is required to drive optimal CD8⁺ T cell responses.

CTA immunization correlated with an influx of tumor-specific CD8⁺ T cells and a marked shift toward proliferative “stem-like” Slamf6⁺ CX3CR1⁺ subsets—cells long associated with durable CPI responses (Baden et al., 2020; Im et al., 2016). Expression of PD-1, TIM-3, and CD39 exhaustion markers decreased, while cytolytic mediators perforin and granzyme rose. Depletion of CD4⁺ T cells abolished these effects, underscoring the dependency of CD8⁺ responses on CD4⁺ help within this chimeric antigen framework.

CTA expression within tumor cell vaccines profoundly reprogrammed DCs in DLNs. Single-DC transcriptomics revealed upregulation of costimulatory molecules (CD40, CD80, and CD86) and activation of the ALCAM–CD6 signaling axis—a pathway that promotes stable DC–T cell synapses (Ibáñez et al., 2006)—accompanied by suppression of TIGIT-associated inhibitory genes. Flow cytometry validated these transcriptional shifts, confirming the emergence of a hyper-activating DC phenotype.

Functionally, CTA-primed DCs became highly efficient at activating both CD4⁺ and CD8⁺ T cells in DLNs, leading to robust expansion of tumor-infiltrating lymphocytes (TILs). These data define DCs as the central integrators of CTA-induced synergy, bridging dual-epitope presentation with checkpoint blockade responsiveness.

The striking feature of CTAs is the requirement that MHC class I and II peptides be expressed within a single endogenous protein. This configuration likely enables a single DC to simultaneously engage CD8⁺ and CD4⁺ T cells, providing optimal costimulation for CD8⁺ differentiation.

This dependency was revealed using dual- versus single-epitope constructs, both predicted to be rapidly degraded in the cytosol due to their small size and lack of stable structure. Such rapidly degraded immunogens are inefficient for classical exogenous cross-priming (Norbury et al., 2004; Wolkers et al., 2004). If such cross-priming dominates in this system, antigen availability should be limiting and require epitopes to be physically linked to increase the chances of a single DC co-expressing class I and class II peptides. Determining whether co-expression of two stable proteins, each containing a single immunogenic class I or class II peptide, will help clarify the underlying mechanism.

Another unresolved question is why cell expression of the CTA is required at all, since DCs efficiently generate and present class I and II peptides from exogenous proteins. One possibility is that tumor cells act as adjuvants, providing innate immune cues that license DCs. Testing whether exogenous protein mixed with tumor cell lysates (or subcellular fractions) can reproduce CTA effects could reveal the adjuvant signals necessary for optimal T cell priming. This would also enable CTA vaccination using defined antigens and molecular adjuvants, avoiding the complexity of cell-based immunogens.

Perhaps the study’s most intriguing insight is that tumors may already express natural CTAs. Frameshift mutations, arising from insertions or deletions, create aberrant open reading frames that generate relatively long nonself proteins containing both MHC-I and MHC-II epitopes. Whole-exome sequencing of MC38 tumor cells revealed 25 such frameshifts; one in Notch2 (Notch2^MUT) encodes a 216–amino acid neoprotein harboring 11 predicted class I and 14 predicted class II binders.

Expression of Notch2^MUT in diverse tumor lines triggered spontaneous rejection and durable memory. When combined with anti–PD-L1 therapy, it eradicated otherwise resistant tumors. Both CD4⁺ and CD8⁺ T cells recognized Notch2^MUT-derived peptides, confirming its dual MHC restriction.

This discovery may explain why cancers with high long frameshift burdens—such as microsatellite instability–high tumors—respond disproportionately well to checkpoint blockade (Maby et al., 2016). Zhang et al. posit that such tumors naturally generate CTA neoantigens, enabling intrinsic dual T cell activation.

Yet caveats remain. MC38 cells, derived by methylcholanthrene mutagenesis, contain mutation loads far exceeding most human tumors. Notch2^MUT is unusually rich in predicted MHC-binding peptides (25 total), which, given an overall accuracy of ~70%, predicts 18 MHC-binding peptides, with further filters (levels of CTA expression, proteolytic liberation, TAP transport, and TCR recognition) predicted to greatly limit immunogenicity (Yewdell, 2006). Moreover, MC38 cells—despite harboring at least one endogenous CTA—still form tumors unless engineered to express a novel CTA. These observations highlight that endogenous CTAs may not universally induce tumor rejection.

Mining the Cancer Genome Atlas, Zhang et al. (2025) found a remarkable correlation: colorectal cancer patients whose tumors contained long frameshift-derived peptides (>120 amino acids) had nearly double the 10-year survival (66 vs. 38%) compared with patients lacking such mutations. Importantly, total mutational burden alone did not predict survival.

This suggests that the structure of mutations—specifically, the generation of extended, multi-epitope frameshifts—may determine the efficacy of anti-tumor immunity. If validated in other malignancies, frameshift length could serve as a biomarker to predict CPI responsiveness and guide personalized vaccine design.

Building on this concept, Zhang et al. (2025) demonstrated that DNA vaccines encoding chimeric CTA constructs could induce potent anti-tumor immunity. Mice immunized intramuscularly with DNA encoding a dual MHC-I/II immunogenic antigen rejected established MC38 tumors, whereas immunization with a class I–only construct had minimal impact. Similarly, DNA encoding the endogenous Notch2^MUT protein synergized with anti–PD-L1 therapy, quadrupling TIL infiltration, expanding self-renewing CD8⁺ T cells, and shrinking tumor mass by ∼30%.

The CTA effect can be viewed as a “DC checkpoint” inhibitor—a mechanism that unleashes DCs to prime balanced CD4⁺ and CD8⁺ responses alongside CPI therapy. Translating this approach to humans will require tailoring constructs to individual HLA haplotypes, though there are a number of class II epitopes known to be broadly immunogenic (Alexander et al., 1994). More simply, viral proteins such as SARS-CoV-2 Spike are natural “Super-CTAs” with dozens of epitopes that bind multiple class I and II molecules.

Indeed, in the early 1900s, reprising Coley, there were anecdotal reports of cancer regression following respiratory infection, tuberculosis, and rabies vaccination (Dock, 1904). By the 1970s, studies had shown that RNA virus–infected tumor lysates could induce tumor-specific immunity (Lindemann, 1974). More recently, case reports described lymphoma regression following SARS-CoV-2 infection (Challenor and Tucker, 2021). Strikingly, mRNA-based SARS-CoV-2 vaccination within 100 days of CPI initiation nearly doubles 3-year survival in lung cancer and melanoma (Grippin et al., 2025).

Zhang et al. provide a mechanistic foundation for harnessing dual MHC-I and MHC-II peptide expression to strengthen anti-tumor T cell cooperation. Their findings open multiple experimental and translational avenues:

• •

  Breadth: How broadly will CTAs function across HLA haplotypes and tumor types?

• •

  Specificity: Should vaccines incorporate epitopes that prime naïve responses, recall memory responses, or exploit persistent immunity to CMV or other herpesviruses?

• •

  Safety: To what extent do CTA responses provoke autoimmunity?

• •

  Delivery: Will optimal platforms be mRNA, DNA, or cell-based formulations?

Ultimately, CTAs demonstrate that a deceptively simple structural innovation—linking class I and II epitopes within one continuous sequence—can reshape the landscape of tumor immunotherapy. In immunology as in life, the whole can indeed be far greater than the sum of its parts.

This research was supported by the Intramural Research Program of the National Institutes of Health (NIH). The contributions of the NIH authors are considered works of the United States Government. The findings and conclusions presented in this paper are those of the authors and do not necessarily reflect the views of the NIH or the US Department of Health and Human Services.

Author contributions: Amir Ghorbani: conceptualization, visualization, and writing—original draft, review, and editing. Jonathan W. Yewdell: conceptualization, resources, supervision, visualization, and writing—original draft, review, and editing.