Mechanisms regulating B cell development, activation, education in the germinal center (GC) and differentiation, underpin the humoral immune response. Protein arginine methyltransferase 5 (Prmt5), ...which catalyzes most symmetric dimethyl arginine protein modifications, is overexpressed in B cell lymphomas but its function in normal B cells is poorly defined. Here we show that Prmt5 is necessary for antibody responses and has essential but distinct functions in all proliferative B cell stages in mice. Prmt5 is necessary for B cell development by preventing p53-dependent and p53-independent blocks in Pro-B and Pre-B cells, respectively. By contrast, Prmt5 protects, via p53-independent pathways, mature B cells from apoptosis during activation, promotes GC expansion, and counters plasma cell differentiation. Phenotypic and RNA-seq data indicate that Prmt5 regulates GC light zone B cell fate by regulating transcriptional programs, achieved in part by ensuring RNA splicing fidelity. Our results establish Prmt5 as an essential regulator of B cell biology.
Most cancers become more dangerous by the outgrowth of malignant subclones with additional DNA mutations that favor proliferation or survival. Using chronic lymphocytic leukemia (CLL), a disease that ...exemplifies this process and is a model for neoplasms in general, we created transgenic mice overexpressing the enzyme activation-induced deaminase (AID), which has a normal function of inducing DNA mutations in B lymphocytes. AID not only allows normal B lymphocytes to develop more effective immunoglobulin-mediated immunity, but is also able to mutate nonimmunoglobulin genes, predisposing to cancer. In CLL, AID expression correlates with poor prognosis, suggesting a role for this enzyme in disease progression. Nevertheless, direct experimental evidence identifying the specific genes that are mutated by AID and indicating that those genes are associated with disease progression is not available. To address this point, we overexpressed Aicda in a murine model of CLL (Eμ-TCL1). Analyses of TCL1/AID mice demonstrate a role for AID in disease kinetics, CLL cell proliferation, and the development of cancer-related target mutations with canonical AID signatures in nonimmunoglobulin genes. Notably, our mouse models can accumulate mutations in the same genes that are mutated in human cancers. Moreover, some of these mutations occur at homologous positions, leading to identical or chemically similar amino acid substitutions as in human CLL and lymphoma. Together, these findings support a direct link between aberrant AID activity and CLL driver mutations that are then selected for their oncogenic effects, whereby AID promotes aggressiveness in CLL and other B-cell neoplasms.
Accumulation of the microtubule-associated protein Tau is linked to neuronal cell death in tauopathies, but how intraneuronal Tau levels are regulated in health and disease remains unclear. Here, we ...show that conditional inactivation of the trafficking adaptor protein Numb in retinal ganglion cells (RGCs) increases Tau levels and leads to axonal blebbing, which is followed by neuronal cell loss in aged mice. In the TauP301S mouse model of tauopathy, conditional inactivation of Numb in RGCs and spinal motoneurons accelerates neurodegeneration, and loss of Numb in motoneurons also leads to precocious hindlimb paralysis. Conversely, overexpression of the long isoform of Numb (Numb-72) decreases intracellular Tau levels and reduces axonal blebbing in TauP301S RGCs, leading to improved electrical activity in cultured neurons and improves performance in a visually guided behavior test in vivo. These results uncover Numb as a key regulator of intracellular Tau levels and identify Numb-72 as a potential therapeutic factor for tauopathies.
Recently, pre-trained foundation models have enabled significant advancements in multiple fields. In molecular machine learning, however, where datasets are often hand-curated, and hence typically ...small, the lack of datasets with labeled features, and codebases to manage those datasets, has hindered the development of foundation models. In this work, we present seven novel datasets categorized by size into three distinct categories: ToyMix, LargeMix and UltraLarge. These datasets push the boundaries in both the scale and the diversity of supervised labels for molecular learning. They cover nearly 100 million molecules and over 3000 sparsely defined tasks, totaling more than 13 billion individual labels of both quantum and biological nature. In comparison, our datasets contain 300 times more data points than the widely used OGB-LSC PCQM4Mv2 dataset, and 13 times more than the quantum-only QM1B dataset. In addition, to support the development of foundational models based on our proposed datasets, we present the Graphium graph machine learning library which simplifies the process of building and training molecular machine learning models for multi-task and multi-level molecular datasets. Finally, we present a range of baseline results as a starting point of multi-task and multi-level training on these datasets. Empirically, we observe that performance on low-resource biological datasets show improvement by also training on large amounts of quantum data. This indicates that there may be potential in multi-task and multi-level training of a foundation model and fine-tuning it to resource-constrained downstream tasks.
PDF
