Piscis: a novel loss estimator of the F1 score enables accurate spot detection in fluorescence microscopy images via deep learning.

Single-molecule RNA fluorescence in situ hybridization (RNA FISH)-based spatial transcriptomics methods have enabled the accurate quantification of gene expression at single-cell resolution by visualizing transcripts as diffraction-limited spots. While these methods generally scale to large samples, image analysis remains challenging, often requiring manual parameter tuning. We present Piscis, a fully automatic deep learning algorithm for spot detection trained using a novel loss function, the SmoothF1 loss, that approximates the F1 score to directly penalize false positives and false negatives but remains differentiable and hence usable for training by deep learning approaches. Piscis was trained and tested on a diverse dataset composed of 358 manually annotated experimental RNA FISH images representing multiple cell types and 240 additional synthetic images. Piscis outperforms other state-of-the-art spot detection methods, enabling accurate, high-throughput analysis of RNA FISH-derived imaging data without the need for manual parameter tuning.

Post-transcriptional splicing can occur in a slow-moving zone around the gene.

Splicing is the stepwise molecular process by which introns are removed from pre-mRNA and exons are joined together to form mature mRNA sequences. The ordering and spatial distribution of these steps remain controversial, with opposing models suggesting splicing occurs either during or after transcription. We used single-molecule RNA FISH, expansion microscopy, and live-cell imaging to reveal the spatiotemporal distribution of nascent transcripts in mammalian cells. At super-resolution levels, we found that pre-mRNA formed clouds around the transcription site. These clouds indicate the existence of a transcription-site-proximal zone through which RNA move more slowly than in the nucleoplasm. Full-length pre-mRNA undergo continuous splicing as they move through this zone following transcription, suggesting a model in which splicing can occur post-transcriptionally but still within the proximity of the transcription site, thus seeming co-transcriptional by most assays. These results may unify conflicting reports of co-transcriptional versus post-transcriptional splicing.

Lymphatic-preserving treatment sequencing with immune checkpoint inhibition unleashes cDC1-dependent antitumor immunity in HNSCC.

Despite the promise of immune checkpoint inhibition (ICI), therapeutic responses remain limited. This raises the possibility that standard of care treatments delivered in concert may compromise the tumor response. To address this, we employ tobacco-signature head and neck squamous cell carcinoma murine models in which we map tumor-draining lymphatics and develop models for regional lymphablation with surgery or radiation. We find that lymphablation eliminates the tumor ICI response, worsening overall survival and repolarizing the tumor- and peripheral-immune compartments. Mechanistically, within tumor-draining lymphatics, we observe an upregulation of conventional type I dendritic cells and type I interferon signaling and show that both are necessary for the ICI response and lost with lymphablation. Ultimately, we provide a mechanistic understanding of how standard oncologic therapies targeting regional lymphatics impact the tumor response to immune-oncology therapy in order to define rational, lymphatic-preserving treatment sequences that mobilize systemic antitumor immunity, achieve optimal tumor responses, control regional metastatic disease, and confer durable antitumor immunity.

Disruption of the HER3-PI3K-mTOR oncogenic signaling axis and PD-1 blockade as a multimodal precision immunotherapy in head and neck cancer.

Immune checkpoint blockade (ICB) therapy has revolutionized head and neck squamous cell carcinoma (HNSCC) treatment, but <20% of patients achieve durable responses. Persistent activation of the PI3K/AKT/mTOR signaling circuitry represents a key oncogenic driver in HNSCC; however, the potential immunosuppressive effects of PI3K/AKT/mTOR inhibitors may limit the benefit of their combination with ICB. Here we employ an unbiased kinome-wide siRNA screen to reveal that HER3, is essential for the proliferation of most HNSCC cells that do not harbor PIK3CA mutations. Indeed, we find that persistent tyrosine phosphorylation of HER3 and PI3K recruitment underlies aberrant PI3K/AKT/mTOR signaling in PIK3CA wild type HNSCCs. Remarkably, antibody-mediated HER3 blockade exerts a potent anti-tumor effect by suppressing HER3-PI3K-AKT-mTOR oncogenic signaling and concomitantly reversing the immune suppressive tumor microenvironment. Ultimately, we show that HER3 inhibition and PD-1 blockade may provide a multimodal precision immunotherapeutic approach for PIK3CA wild type HNSCC, aimed at achieving durable cancer remission.