Restoring functional β-cell mass is an important therapeutic goal for both type 1 and type 2 diabetes (1). While proliferation of existing β-cells is the primary means of β-cell replacement in rodents (2), it is unclear whether a similar principle applies to humans, as human β-cells are remarkably resistant to stimulation of division (3,4). Here, we show that 5-iodotubercidin (5-IT), an annotated adenosine kinase inhibitor previously reported to increase proliferation in rodent and porcine islets (5), strongly and selectively increases human β-cell proliferation in vitro and in vivo. Remarkably, 5-IT also increased glucose-dependent insulin secretion after prolonged treatment. Kinome profiling revealed 5-IT to be a potent and selective inhibitor of the dual-specificity tyrosine phosphorylation–regulated kinase (DYRK) and cell division cycle–like kinase families. Induction of β-cell proliferation by either 5-IT or harmine, another natural product DYRK1A inhibitor, was suppressed by coincubation with the calcineurin inhibitor FK506, suggesting involvement of DYRK1A and nuclear factor of activated T cells signaling. Gene expression profiling in whole islets treated with 5-IT revealed induction of proliferation- and cell cycle–related genes, suggesting that true proliferation is induced by 5-IT. Furthermore, 5-IT promotes β-cell proliferation in human islets grafted under the kidney capsule of NOD-scid IL2Rgnull mice. These results point to inhibition of DYRK1A as a therapeutic strategy to increase human β-cell proliferation.

Anti-CD19 chimeric antigen receptor (CAR) T cells showed significant anti-leukemic activity in B-precursor acute lymphoblastic leukemia (ALL). Allogeneic, HLA-mismatched off-the-shelf 3rd-party donors may offer ideal fitness of the effector cells but carry the risk of graft-versus-host disease. Knockout of the endogenous T-cell receptor (TCR) in CD19-CAR-T cells may be a promising solution. Here, we induced a CRISPR/Cas9-mediated knockout of the TCRb-chain in combination with a 2nd-generation retroviral CAR transduction including a 4-1BB costimulatory domain in primary T cells. This tandem engineering led to a highly functional population of TCR-KO-CAR-T cells with strong activation (CD25, IFN-γ), proliferation and specific killing upon CD19 target recognition. TCR-KO-CAR-T cells had a balanced phenotype of central memory and effector memory T cells. KO of the endogenous TCR in T cells strongly ablated alloreactivity in comparison to TCR-expressing T cells. In a patient-derived xenograft model of childhood ALL, TCR-KO-CAR-T cells clearly controlled CD19+ leukemia burden and improved survival in vivo. However, co-expression of endogenous TCR plus CAR led to superior persistence of T cells and significantly prolonged leukemia control in vivo, confirmed by a second in vivo model using NALM6 leukemia cells. These results point towards an essential role of the endogenous TCR for longevity of the response at the price of alloreactivity. In conclusion, anti-CD19 CAR-T cells with a CRISPR/Cas9-mediated TCR-KO are promising candidates for non-matched third-party adoptive T-cell transfer with high anti-leukemic functionality in the absence of alloreactivity, but long-term persistence in vivo is better in the presence of the endogenous TCR.

The histone methyltransferase G9a is overexpressed in a variety of cancer types, including pancreatic adenocarcinoma, and promotes tumor invasiveness and metastasis. We recently reported the discovery of BRD4770, a small-molecule inhibitor of G9a that induces senescence in PANC-1 cells. We observed that the cytotoxic effects of BRD4770 were dependent on genetic background, with cell lines lacking functional p53 being relatively resistant to compound treatment. To understand the mechanism of genetic selectivity, we used two complementary screening approaches to identify enhancers of BRD4770. The natural product and putative BH3 mimetic gossypol enhanced the cytotoxicity of BRD4770 in a synergistic manner in p53-mutant PANC-1 cells but not in immortalized non-tumorigenic pancreatic cells. The combination of gossypol and BRD4770 increased LC3-II levels and the autophagosome number in PANC-1 cells, and the compound combination appears to act in a BNIP3 (B-cell lymphoma 2 19-kDa interacting protein)-dependent manner, suggesting that these compounds act together to induce autophagy-related cell death in pancreatic cancer cells.

Mesenchymal stromal cells (MSCs) have initially been characterized as a fibroblastlike cell population that can be expanded readily in vitro, and is able to support hematopoiesis in vitro and in vivo. By serendipity it was discovered that MSCs can also be administered into the bloodstream. This mode of application formed a major breakthrough in the clinical use of MSCs, because MSC transplantation was found to cure severe immune hyperactivation states such as graft-versus-host disease after allogeneic bone marrow transplantation, or bacterial sepsis. However, MSCs were found difficult to trace and consensus to date is lacking in the scientific community as to where transplanted MSCs end up in the body and which major principles are responsible for the therapeutic effects of MSCs. This chapter gives an overview of the current knowledge on interactions of freshly transplanted MSCs with the cells in the blood stream and the vessel wall, with major organs such as lung, liver, gut, and spleen, and discusses the limitations of the methodologies used to trace transplanted MSCs. The findings will be put into perspective on how therapeutically applied, culture-expanded MSCs may exert beneficial effects.

Recent advances in deep learning enable using chemical structures and phenotypic profiles to accurately predict assay results for compounds virtually, reducing the time and cost of screens in the drug discovery process. The relative strength of high-throughput data sources - chemical structures, images (Cell Painting), and gene expression profiles (L1000) - has been unknown. Here we compare their ability to predict the activity of compounds structurally different from those used in training, using a sparse dataset of 16,979 chemicals tested in 376 assays for a total of 542,648 readouts. Deep learning-based feature extraction from chemical structures provided a remarkable ability to predict assay activity for structures dissimilar to those used for training. Image-based profiling performed even better, but requires wet lab experimentation. It outperformed gene expression profiling, and at lower cost. Furthermore, the three profiling modalities are complementary, and together can predict a wide range of diverse bioactivity, including cell-based and biochemical assays. Our study shows that, for many assays, predicting compound activity from phenotypic profiles and chemical structures is an accurate and efficient way to identify potential treatments in the early stages of the drug discovery process.