Neoantigens arising from somatic mutations are tumor specific and induce antitumor host T cell responses. However, their sequences are individual specific and need to be identified for each patient for therapeutic applications. Here, we present a proteogenomic approach for neoantigen identification, named Neoantigen Selection using a Surrogate Immunopeptidome (NESSIE). This approach uses an autologous wild-type immunopeptidome as a surrogate for the tumor immunopeptidome and allows human leukocyte antigen (HLA)–agnostic identification of both HLA class I (HLA-I) and HLA class II (HLA-II) neoantigens. We demonstrate the direct identification of highly immunogenic HLA-I and HLA-II neoantigens using NESSIE in patients with colorectal cancer and endometrial cancer. Fresh or frozen tumor samples are not required for analysis, making it applicable to many patients in clinical settings. We also demonstrate tumor prevention by vaccination with selected neoantigens in a preclinical mouse model. This approach may benefit personalized T cell–mediated immunotherapies.

The application of deep learning algorithms to predict the molecular profiles of various cancers from digital images of hematoxylin and eosin (H&E)-stained slides has been reported in recent years, mainly for gastric and colon cancers. In this study, we investigated the potential use of H&E-stained endometrial cancer slide images to predict the associated mismatch repair (MMR) status. H&E-stained slide images were collected from 127 cases of the primary lesion of endometrial cancer. After digitization using a Nanozoomer virtual slide scanner (Hamamatsu Photonics), we segmented the scanned images into 5397 tiles of 512 × 512 pixels. The MMR proteins (PMS2, MSH6) were immunohistochemically stained, classified into MMR proficient/deficient, and annotated for each case and tile. We trained several neural networks, including convolutional and attention-based networks, using tiles annotated with the MMR status. Among the tested networks, ResNet50 exhibited the highest area under the receiver operating characteristic curve (AUROC) of 0.91 for predicting the MMR status. The constructed prediction algorithm may be applicable to other molecular profiles and useful for pre-screening before implementing other, more costly genetic profiling tests.

Tumor-reactive CD4+ T cells often accumulate in the tumor microenvironment (TME) in human cancer, but their functions and roles in antitumor responses remain elusive. Here, we investigated the immunopeptidome of HLA class II–positive (HLA-II+) endometrial cancer with an inflamed TME using a proteogenomic approach. We identified HLA-II neoantigens, one of which induced polyclonal CD4+ tumor-infiltrating lymphocyte responses. We then experimentally demonstrated that neoantigen-specific CD4+ tumor-infiltrating lymphocytes lyse target cells in an HLA-II–dependent manner. Single-cell transcriptomic analysis of the TME coupled with T-cell receptor sequencing revealed the presence of CD4+ T-cell clusters characterized by CXCL13 expression. The CXCL13+ clusters contained two subclusters with distinct cytotoxic gene expression patterns. The identified neoantigen-specific CD4+ T cells were found exclusively in one of the CXCL13+ subclusters characterized by granzyme B and CCL5 expression. These results demonstrate the involvement of tumor-reactive CD4+ T cells with cytotoxic function in immune surveillance of endometrial cancer and reveal their transcriptomic signature.

The application of deep learning algorithms to predict the molecular profiles of various cancers from digital images of hematoxylin and eosin (H&E)-stained slides has been reported in recent years, mainly in gastric and colon cancers. In this study, we investigated the potential use of H&E-stained endometrial cancer slide images to predict the associated mismatch repair (MMR) status. H&E-stained slide images were collected from 127 cases of the primary lesion of endometrial cancer. After digitization using a Nanozoomer virtual slide scanner (Hamamatsu Photonics), we segmented the scanned images into 5,397 tiles of 512 × 512 pixels. The MMR proteins (PMS2, MSH6) were immunohistochemically stained, classified into MMR proficient/deficient, and annotated for each case and tile. We trained several neural networks, including convolutional and attention-based networks, using tiles annotated with the MMR status. Among the tested networks, ResNet50 exhibited the highest area under the receiver operating characteristic curve (AUROC) of 0.91 for predicting the MMR status. The constructed prediction algorithm may be applicable to other molecular profiles and useful for pre-screening before implementing other, more costly genetic profiling tests.

Table S6 from Identification and Phenotypic Characterization of Neoantigen-Specific Cytotoxic CD4+ T Cells in Endometrial Cancer

Fig. S8 | Cytokine production by neoantigen-specific CD4+ TIL clones.