anine transferase or lactate dehydrogenase in treated mice. Histopathological analysis of tissues from treated mice did not demonstrate any specific drug-related changes. GrB(C210A)-Fc-IT4 demonstrated excellent, specific cytotoxicity in vitro and impressive in vivo efficacy with no significant toxicity in normal murine models. These studies show GrB(C210A)-Fc-IT4 is an excellent candidate for further preclinical development.GrB(C210A)-Fc-IT4 demonstrated excellent, specific cytotoxicity in vitro and impressive in vivo efficacy with no significant toxicity in normal murine models. These studies show GrB(C210A)-Fc-IT4 is an excellent candidate for further preclinical development. To explore lymphocyte infiltration as a potential mechanism behind CXCL14-mediated tumor growth suppression in oral cavity squamous cell carcinoma (OSCC). We analyzed single cell RNA-sequencing (scRNA-seq) data from OSCC to identify expression changes among malignant cells in lymph nodes (LN) versus primary tumors. CXCL14 expression in murine OSCC cell lines was quantified using qRT-PCR. Short hairpin RNA knockdown of CXCL14 was performed in mouse oral cavity (MOC)1 cells, and CXCL14 overexpression was performed in MOC2 cells. Cells in each condition were injected into C57BL/6 mice with and without T cell depletion, and tumor volume was measured. At 30 days, tumors were dissociated and analyzed by flow cytometry for CD45 CD3 T cells. CXCL14 expression was correlated with gene expression signatures of tumor infiltrating lymphocytes (TIL) in scRNA-seq data, as well as TCGA tumors. scRNA-seq revealed CXCL14 as the most significantly downregulated gene among malignant cells in LNs relative to primary tum may discourage invasion and metastasis. STAT5-IN-1 In human scRNA-seq data, only malignant cell-specific CXCL14 was associated with TIL, suggesting a critical context-dependent effect of CXCL14 expression.Higher CXCL14 expression by tumor cells is associated with reduced tumor growth and increased TIL, supporting immune-mediated suppression of tumor growth in OSCC. Given that CXCL14 is downregulated in LN metastases compared with primary tumors, our data raise the possibility that CXCL14-mediated immune infiltration may discourage invasion and metastasis. In human scRNA-seq data, only malignant cell-specific CXCL14 was associated with TIL, suggesting a critical context-dependent effect of CXCL14 expression. Craniopharyngioma (CP) is a common refractory tumor of the central nervous system. However, little is known about the expression and clinical significance of B7 family ligands/receptors in CP patients. Thus, we conducted the present study to address this issue in a cohort of 132 CP cases. We mapped and quantified the expression of B7 family ligands/receptors molecules programmed cell death ligand 1 (PD-L1), B7-H3, B7-H4 and V-domain Ig-containing suppressor of T cell activation (VISTA) in 89 adamantinomatous-type CP and 43 papillary-type CP samples using immunohistochemistry and immunofluorescence. Associations between the marker levels, clinicopathological variables and survival were evaluated. The positive rates of PD-L1, B7-H3, B7-H4 and VISTA in the cohort of 132 CP cases were 76.5%, 100%, 40.2% and 80.3%, respectively. The cut-off values of PD-L1, B7-H3, B7-H4 and PD-L1 expression were determined by survival receiver operating characteristic (ROC) package, which was 70, 182, 0 and 20, respectively.targets when treating CPs.Cell crawling requires the generation of intracellular forces by the cytoskeleton and their transmission to an extracellular substrate through specific adhesion molecules. Crawling cells show many features of excitable systems, such as spontaneous symmetry breaking and crawling in the absence of external cues, and periodic and propagating waves of activity. Mechanical instabilities in the active cytoskeleton network and feedback loops in the biochemical network of activators and repressors of cytoskeleton dynamics have been invoked to explain these dynamical features. Here, I show that the interplay between the dynamics of cell-substrate adhesion and linear cellular mechanics is sufficient to reproduce many nonlinear dynamical patterns observed in spreading and crawling cells. Using an analytical formalism of the molecular clutch model of cell adhesion, regulated by local mechanical forces, I show that cellular traction forces exhibit stick-slip dynamics resulting in periodic waves of protrusion/retraction and propagating waves along the cell edge. This can explain spontaneous symmetry breaking and polarization of spreading cells, leading to steady crawling or bipedal motion, and bistability, where persistent cell motion requires a sufficiently strong transient external stimulus. The model also highlights the role of membrane tension in providing the long-range mechanical communication across the cell required for symmetry breaking.We discuss the current state of knowledge of stable homotopy groups of spheres. We describe a computational method using motivic homotopy theory, viewed as a deformation of classical homotopy theory. This yields a streamlined computation of the first 61 stable homotopy groups and gives information about the stable homotopy groups in dimensions 62 through 90. As an application, we determine the groups of homotopy spheres that classify smooth structures on spheres through dimension 90, except for dimension 4. The method relies more heavily on machine computations than previous methods and is therefore less prone to error. The main mathematical tool is the Adams spectral sequence.Modern practice for training classification deepnets involves a terminal phase of training (TPT), which begins at the epoch where training error first vanishes. During TPT, the training error stays effectively zero, while training loss is pushed toward zero. Direct measurements of TPT, for three prototypical deepnet architectures and across seven canonical classification datasets, expose a pervasive inductive bias we call neural collapse (NC), involving four deeply interconnected phenomena. (NC1) Cross-example within-class variability of last-layer training activations collapses to zero, as the individual activations themselves collapse to their class means. (NC2) The class means collapse to the vertices of a simplex equiangular tight frame (ETF). (NC3) Up to rescaling, the last-layer classifiers collapse to the class means or in other words, to the simplex ETF (i.e., to a self-dual configuration). (NC4) For a given activation, the classifier's decision collapses to simply choosing whichever class has the closest train class mean (i.