Prediction of homologous recombination deficiency from routine histology with attention-based multiple instance learning in nine different tumor types.

BACKGROUND: Homologous recombination deficiency (HRD) is recognized as a pan-cancer predictive biomarker that potentially indicates who could benefit from treatment with PARP inhibitors (PARPi). Despite its clinical significance, HRD testing is highly complex. Here, we investigated in a proof-of-concept study whether Deep Learning (DL) can predict HRD status solely based on routine hematoxylin & eosin (H&E) histology images across nine different cancer types. METHODS: We developed a deep learning pipeline with attention-weighted multiple instance learning (attMIL) to predict HRD status from histology images. As part of our approach, we calculated a genomic scar HRD score by combining loss of heterozygosity (LOH), telomeric allelic imbalance (TAI), and large-scale state transitions (LST) from whole genome sequencing (WGS) data of n = 5209 patients across two independent cohorts. The model's effectiveness was evaluated using the area under the receiver operating characteristic curve (AUROC), focusing on its accuracy in predicting genomic HRD against a clinically recognized cutoff value. RESULTS: Our study demonstrated the predictability of genomic HRD status in endometrial, pancreatic, and lung cancers reaching cross-validated AUROCs of 0.79, 0.58, and 0.66, respectively. These predictions generalized well to an external cohort, with AUROCs of 0.93, 0.81, and 0.73. Moreover, a breast cancer-trained image-based HRD classifier yielded an AUROC of 0.78 in the internal validation cohort and was able to predict HRD in endometrial, prostate, and pancreatic cancer with AUROCs of 0.87, 0.84, and 0.67, indicating that a shared HRD-like phenotype occurs across these tumor entities. CONCLUSIONS: This study establishes that HRD can be directly predicted from H&E slides using attMIL, demonstrating its applicability across nine different tumor types.

Direct prediction of Homologous Recombination Deficiency from routine histology in ten different tumor types with attention-based Multiple Instance Learning: a development and validation study.

Background: Homologous Recombination Deficiency (HRD) is a pan-cancer predictive biomarker that identifies patients who benefit from therapy with PARP inhibitors (PARPi). However, testing for HRD is highly complex. Here, we investigated whether Deep Learning can predict HRD status solely based on routine Hematoxylin & Eosin (H&E) histology images in ten cancer types. Methods: We developed a fully automated deep learning pipeline with attention-weighted multiple instance learning (attMIL) to predict HRD status from histology images. A combined genomic scar HRD score, which integrated loss of heterozygosity (LOH), telomeric allelic imbalance (TAI) and large-scale state transitions (LST) was calculated from whole genome sequencing data for n=4,565 patients from two independent cohorts. The primary statistical endpoint was the Area Under the Receiver Operating Characteristic curve (AUROC) for the prediction of genomic scar HRD with a clinically used cutoff value. Results: We found that HRD status is predictable in tumors of the endometrium, pancreas and lung, reaching cross-validated AUROCs of 0.79, 0.58 and 0.66. Predictions generalized well to an external cohort with AUROCs of 0.93, 0.81 and 0.73 respectively. Additionally, an HRD classifier trained on breast cancer yielded an AUROC of 0.78 in internal validation and was able to predict HRD in endometrial, prostate and pancreatic cancer with AUROCs of 0.87, 0.84 and 0.67 indicating a shared HRD-like phenotype is across tumor entities. Conclusion: In this study, we show that HRD is directly predictable from H&E slides using attMIL within and across ten different tumor types.

Swarm learning for decentralized artificial intelligence in cancer histopathology.

Artificial intelligence (AI) can predict the presence of molecular alterations directly from routine histopathology slides. However, training robust AI systems requires large datasets for which data collection faces practical, ethical and legal obstacles. These obstacles could be overcome with swarm learning (SL), in which partners jointly train AI models while avoiding data transfer and monopolistic data governance. Here, we demonstrate the successful use of SL in large, multicentric datasets of gigapixel histopathology images from over 5,000 patients. We show that AI models trained using SL can predict BRAF mutational status and microsatellite instability directly from hematoxylin and eosin (H&E)-stained pathology slides of colorectal cancer. We trained AI models on three patient cohorts from Northern Ireland, Germany and the United States, and validated the prediction performance in two independent datasets from the United Kingdom. Our data show that SL-trained AI models outperform most locally trained models, and perform on par with models that are trained on the merged datasets. In addition, we show that SL-based AI models are data efficient. In the future, SL can be used to train distributed AI models for any histopathology image analysis task, eliminating the need for data transfer.

Enhanced L1CAM expression on pancreatic tumor endothelium mediates selective tumor cell transmigration.

L1 cell adhesion molecule (L1CAM) is a transmembrane cell adhesion molecule initially defined as a promigratory molecule in the developing nervous system that appears to be also expressed in some endothelial cells. However, little is known about the functional role of L1CAM on endothelial cells. We observed that L1CAM expression was selectively enhanced on endothelium associated with pancreatic adenocarcinoma in situ and on cultured pancreatic tumor-derived endothelial cells in vitro. L1CAM expression of endothelial cells could be augmented by incubation with immunomodulatory cytokines such as tumor necrosis factor alpha, interferon gamma, or transforming growth factor beta 1. Antibodies to L1CAM and the respective ligand neuropilin-1 blocked tube formation and stromal cell-derived factor 1beta induced transmigration of tumor endothelial cells in vitro. L1CAM expression on tumor-derived-endothelial cells enhanced Panc1 carcinoma cell adhesion to endothelial cell monolayers and transendothelial migration. Our data demonstrate a functional role of L1CAM expression on tumor endothelium that could favor metastasis and angiogenesis during tumor progression.

Phase Ib evaluation of a self-adjuvanted protamine formulated mRNA-based active cancer immunotherapy, BI1361849 (CV9202), combined with local radiation treatment in patients with stage IV non-small cell lung cancer.

BACKGROUND: Preclinical studies demonstrate synergism between cancer immunotherapy and local radiation, enhancing anti-tumor effects and promoting immune responses. BI1361849 (CV9202) is an active cancer immunotherapeutic comprising protamine-formulated, sequence-optimized mRNA encoding six non-small cell lung cancer (NSCLC)-associated antigens (NY-ESO-1, MAGE-C1, MAGE-C2, survivin, 5T4, and MUC-1), intended to induce targeted immune responses. METHODS: We describe a phase Ib clinical trial evaluating treatment with BI1361849 combined with local radiation in 26 stage IV NSCLC patients with partial response (PR)/stable disease (SD) after standard first-line therapy. Patients were stratified into three strata (1: non-squamous NSCLC, no epidermal growth factor receptor (EGFR) mutation, PR/SD after ≥4 cycles of platinum- and pemetrexed-based treatment [n = 16]; 2: squamous NSCLC, PR/SD after ≥4 cycles of platinum-based and non-platinum compound treatment [n = 8]; 3: non-squamous NSCLC, EGFR mutation, PR/SD after ≥3 and ≤ 6 months EGFR-tyrosine kinase inhibitor (TKI) treatment [n = 2]). Patients received intradermal BI1361849, local radiation (4 × 5 Gy), then BI1361849 until disease progression. Strata 1 and 3 also had maintenance pemetrexed or continued EGFR-TKI therapy, respectively. The primary endpoint was evaluation of safety; secondary objectives included assessment of clinical efficacy (every 6 weeks during treatment) and of immune response (on Days 1 [baseline], 19 and 61). RESULTS: Study treatment was well tolerated; injection site reactions and flu-like symptoms were the most common BI1361849-related adverse events. Three patients had grade 3 BI1361849-related adverse events (fatigue, pyrexia); there was one grade 3 radiation-related event (dysphagia). In comparison to baseline, immunomonitoring revealed increased BI1361849 antigen-specific immune responses in the majority of patients (84%), whereby antigen-specific antibody levels were increased in 80% and functional T cells in 40% of patients, and involvement of multiple antigen specificities was evident in 52% of patients. One patient had a partial response in combination with pemetrexed maintenance, and 46.2% achieved stable disease as best overall response. Best overall response was SD in 57.7% for target lesions. CONCLUSION: The results support further investigation of mRNA-based immunotherapy in NSCLC including combinations with immune checkpoint inhibitors. TRIAL REGISTRATION: ClinicalTrials.gov identifier: NCT01915524 .

Low-dose irradiation programs macrophage differentiation to an iNOS⁺/M1 phenotype that orchestrates effective T cell immunotherapy.

Inefficient T cell migration is a major limitation of cancer immunotherapy. Targeted activation of the tumor microenvironment may overcome this barrier. We demonstrate that neoadjuvant local low-dose gamma irradiation (LDI) causes normalization of aberrant vasculature and efficient recruitment of tumor-specific T cells in human pancreatic carcinomas and T-cell-mediated tumor rejection and prolonged survival in otherwise immune refractory spontaneous and xenotransplant mouse tumor models. LDI (local or pre-adoptive-transfer) programs the differentiation of iNOS⁺ M1 macrophages that orchestrate CTL recruitment into and killing within solid tumors through iNOS by inducing endothelial activation and the expression of TH1 chemokines and by suppressing the production of angiogenic, immunosuppressive, and tumor growth factors.

Intratumoral cytokines and tumor cell biology determine spontaneous breast cancer-specific immune responses and their correlation to prognosis.

Spontaneous immune responses in cancer patients have been described. Yet their clinical relevance and the conditions for their generation remain unclear. We characterized conditions that determine immune responses in primary breast cancer patients. We used tetramer analysis, ex vivo IFN-gamma ELISPOT, cytotoxicity assays, and ELISA in 207 untreated patients and 12 Her-2/neu-specific CD8 T-cell lines to evaluate tumor-specific T cells (TC) in the bone marrow or MUC1-specific antibodies in the blood. Multiplex analysis was performed to quantify 27 intratumoral cytokines, chemokines, and growth factors. Results were compared with multiple pathologic and clinical parameters of the patients and tumors. Forty percent of the patients showed tumor-specific TC responses. These correlated with tumors of high differentiation, estrogen receptor expression, and low proliferative activity, and with a reduced cancer mortality risk. High tumor cell differentiation correlated with increased intratumoral, but not plasma, concentrations of IFN-alpha and reduced transforming growth factor (TGF)beta1. In an in vitro priming experiment these two cytokines increased or inhibited, respectively, the capacity of dendritic cells to induce tumor-reactive TC. Tumor-specific B-cell responses, mainly of IgM isotype, were detectable in 50% of the patients and correlated with advanced tumor stage, increased TGFbeta1, reduced IFN-alpha, and absence of TC responses. We show here that different types of immune responses are linked to distinct cytokine microenvironments and correlate with prognosis-relevant differences in tumor pathobiology. These findings shed light on the relation between immune response and cancer prognosis.