Self-clustered GAN for precipitation nowcasting.

This paper proposes a novel GAN framework with self-clustering approach for precipitation nowcasting (ClusterCast). Previous studies have primarily captured the motion vector using only a single latent space, making the models difficult to adapt to disparate space-time distribution of precipitation. Environmental factors (e.g., regional characteristics and precipitation scale) have an impact on precipitation systems and can cause non-stationary distribution. To tackle this problem, our key idea is to train a generator network to predict future radar frames by learning a sub-network that automatically labels precipitation types from a generative model. The training process consists of (i) clustering the hierarchical features derived from the generator stem using a sub-network and (ii) predicting future radar frames according to the self-supervised labels, enabling heterogeneous latent representation. Additionally, we attempt an ensemble forecast that prescribes random perturbations to improve performance. With the flexibility of representation learning, ClusterCast enables the model to learn precipitation distribution more accurately. Results indicate that our method generates non-blurry future frames by preventing mode collapse, and the proposed method demonstrates robustness across various precipitation scenarios. Extensive experiments demonstrate that our method outperforms four benchmarks on a 2-h prediction basis with a mean squared error (MSE) of 8.9% on unseen datasets.

Asxl1 ablation in mouse embryonic stem cells impairs neural differentiation without affecting self-renewal.

Additional sex comb-like1 (Asxl1) is known as a chromatin modulator that plays dual functions in transcriptional regulation depending on the cell type. Recent studies using Asxl1 knockout mice revealed that Asxl1 is important for the proliferation and differentiation of hematopoietic progenitor cells, and the development of organs. Although we previously reported Asxl1 as a Sox2 target gene, its function in embryonic stem cells (ESCs) remains largely unknown. For this purpose, we isolated ESCs from the blastocyst inner cell mass of Asxl1-/- mice. Asxl1 deficiency in ESCs exhibited no effect on cell proliferation, expression of core pluripotent transcription factors, or alkaline phosphatase activity, suggesting dispensability of Asxl1 for self-renewal of ESCs. By contrast, the differentiation of Asxl1-/- ESCs was significantly affected as shown by size reductions of embryoid bodies accompanied with apoptosis, aberrant expression of differentiation genes, downregulation of bivalent neurogenesis genes, and abnormal axon formation in neurons. Overall, our findings indicated that Asxl1 played a critical role in regulating genes associated with neural differentiation without affecting self-renewal of mouse ESCs.

Asxl1 exerts an antiproliferative effect on mouse lung maturation via epigenetic repression of the E2f1-Nmyc axis.

Although additional sex combs-like 1 (ASXL1) has been extensively described in hematologic malignancies, little is known about the molecular role of ASXL1 in organ development. Here, we show that Asxl1 ablation in mice results in postnatal lethality due to cyanosis, a respiratory failure. This lung defect is likely caused by higher proliferative potential and reduced expression of surfactant proteins, leading to reduced air space and defective lung maturation. By microarray analysis, we identified E2F1-responsive genes, including Nmyc, as targets repressed by Asxl1. Nmyc and Asxl1 are reciprocally expressed during the fetal development of normal mouse lungs, whereas Nmyc downregulation is impaired in Asxl1-deficient lungs. Together with E2F1 and ASXL1, host cell factor 1 (HCF-1), purified as an Asxl1-bound protein, is recruited to the E2F1-binding site of the Nmyc promoter. The interaction occurs between the C-terminal region of Asxl1 and the N-terminal Kelch domain of HCF-1. Trimethylation (me3) of histone H3 lysine 27 (H3K27) is enriched in the Nmyc promoter upon Asxl1 overexpression, whereas it is downregulated in Asxl1-deleted lung and -depleted A549 cells, similar to H3K9me3, another repressive histone marker. Overall, these findings suggest that Asxl1 modulates proliferation of lung epithelial cells via the epigenetic repression of Nmyc expression, deficiency of which may cause hyperplasia, leading to dyspnea.

Fibroblast growth factor receptor 3-mediated reactivation of ERK signaling promotes head and neck squamous cancer cell insensitivity to MEK inhibition.

Recurrent or metastatic head and neck squamous cell carcinoma (HNSCC) has been a longstanding challenge for head and neck oncologists, and current treatments still have limited efficacy. ERK is aberrantly overexpressed and activated in HNSCC. Herein, we aimed to investigate the cause of the limited therapeutic effect of selumetinib, a selective inhibitor of MEK in HNSCC, as MEK/ERK reactivation inevitably occurs. We assessed the effects of combining selumetinib with fibroblast growth factor receptor 3 (FGFR3) inhibitor (PD173074) on tumor growth. Selumetinib transiently inhibited MAPK signaling and reactivated ERK signaling in HNSCC cells. Rebound in the ERK and Akt pathways in HNSCC cells was accompanied by increased FGFR3 signaling after selumetinib treatment. Feedback activation of FGFR3 was a result of autocrine secretion of the FGF2 ligand. The FGFR3 inhibitor PD173074 prevented MAPK rebound and sensitized the response of HNSCC cells to selumetinib. These results provided rational therapeutic strategies for clinical studies of this subtype of patients that show a poor prognosis with selumetinib. Our data provide a rationale for combining a MEK inhibitor with inhibitors of feedback activation of FGFR3 signaling in HNSCC cells. ERK rebound as a result of the upregulation of FGFR3 and the ligand FGF2 diminished the antitumor effects of selumetinib, which was overcome by combination treatment with the FGFR3 inhibitor.

Asxl1 deficiency in embryonic fibroblasts leads to cellular senescence via impairment of the AKT-E2F pathway and Ezh2 inactivation.

Although ASXL1 mutations are frequently found in human diseases, including myeloid leukemia, the cell proliferation-associated function of ASXL1 is largely unknown. Here, we explored the molecular mechanism underlying the growth defect found in Asxl1-deficient mouse embryonic fibroblasts (MEFs). We found that Asxl1, through amino acids 371 to 655, interacts with the kinase domain of AKT1. In Asxl1-null MEFs, IGF-1 was unable to induce AKT1 phosphorylation and activation; p27Kip1, which forms a ternary complex with ASXL1 and AKT1, therefore remained unphosphorylated. Hypophosphorylated p27Kip1 is able to enter the nucleus, where it prevents the phosphorylation of Rb; this ultimately leads to the down-regulation of E2F target genes as confirmed by microarray analysis. We also found that senescence-associated (SA) genes were upregulated and that SA ß-gal staining was increased in Asxl1 -/- MEFs. Further, the treatment of an AKT inhibitor not only stimulated nuclear accumulation of p27Kip1 leading to E2F inactivation, but also promoted senescence. Finally, Asxl1 disruption augmented the expression of p16Ink4a as result of the defect in Asxl1-Ezh2 cooperation. Overall, our study provides the first evidence that Asxl1 both activates the AKT-E2F pathway and cooperates with Ezh2 through direct interactions at early embryonic stages, reflecting that Asxl1 disruption causes cellular senescence.