B-cell-specific checkpoint molecules that regulate anti-tumour immunity.

The role of B cells in anti-tumour immunity is still debated and, accordingly, immunotherapies have focused on targeting T and natural killer cells to inhibit tumour growth1,2. Here, using high-throughput flow cytometry as well as bulk and single-cell RNA-sequencing and B-cell-receptor-sequencing analysis of B cells temporally during B16F10 melanoma growth, we identified a subset of B cells that expands specifically in the draining lymph node over time in tumour-bearing mice. The expanding B cell subset expresses the cell surface molecule T cell immunoglobulin and mucin domain 1 (TIM-1, encoded by Havcr1) and a unique transcriptional signature, including multiple co-inhibitory molecules such as PD-1, TIM-3, TIGIT and LAG-3. Although conditional deletion of these co-inhibitory molecules on B cells had little or no effect on tumour burden, selective deletion of Havcr1 in B cells both substantially inhibited tumour growth and enhanced effector T cell responses. Loss of TIM-1 enhanced the type 1 interferon response in B cells, which augmented B cell activation and increased antigen presentation and co-stimulation, resulting in increased expansion of tumour-specific effector T cells. Our results demonstrate that manipulation of TIM-1-expressing B cells enables engagement of the second arm of adaptive immunity to promote anti-tumour immunity and inhibit tumour growth.

Stress testing reveals gaps in clinic readiness of image-based diagnostic artificial intelligence models.

Artificial intelligence models match or exceed dermatologists in melanoma image classification. Less is known about their robustness against real-world variations, and clinicians may incorrectly assume that a model with an acceptable area under the receiver operating characteristic curve or related performance metric is ready for clinical use. Here, we systematically assessed the performance of dermatologist-level convolutional neural networks (CNNs) on real-world non-curated images by applying computational "stress tests". Our goal was to create a proxy environment in which to comprehensively test the generalizability of off-the-shelf CNNs developed without training or evaluation protocols specific to individual clinics. We found inconsistent predictions on images captured repeatedly in the same setting or subjected to simple transformations (e.g., rotation). Such transformations resulted in false positive or negative predictions for 6.5-22% of skin lesions across test datasets. Our findings indicate that models meeting conventionally reported metrics need further validation with computational stress tests to assess clinic readiness.