Hybrid glasses connect the emerging field of metal-organic frameworks (MOFs) with the glass formation, amorphization and melting processes of these chemically versatile systems. Though inorganic zeolites collapse around the glass transition and melt at higher temperatures, the relationship between amorphization and melting has so far not been investigated. Here we show how heating MOFs of zeolitic topology first results in a low density ‘perfect' glass, similar to those formed in ice, silicon and disaccharides. This order–order transition leads to a super-strong liquid of low fragility that dynamically controls collapse, before a subsequent order–disorder transition, which creates a more fragile high-density liquid. After crystallization to a dense phase, which can be remelted, subsequent quenching results in a bulk glass, virtually identical to the high-density phase. We provide evidence that the wide-ranging melting temperatures of zeolitic MOFs are related to their network topologies and opens up the possibility of ‘melt-casting' MOF glasses.

The transcription factor KLF5 is highly expressed in basal-like breast cancer and promotes breast cancer cell proliferation, survival, migration and tumour growth. Here we show that, in breast cancer cells, KLF5 is stabilized by the deubiquitinase (DUB) BAP1. With a genome-wide siRNA library screen of DUBs, we identify BAP1 as a bona fide KLF5 DUB. BAP1 interacts directly with KLF5 and stabilizes KLF5 via deubiquitination. KLF5 is in the BAP1/HCF-1 complex, and this newly identified complex promotes cell cycle progression partially by inhibiting p27 gene expression. Furthermore, BAP1 knockdown inhibits tumorigenicity and lung metastasis, which can be rescued partially by ectopic expression of KLF5. Collectively, our findings not only identify BAP1 as the DUB for KLF5, but also reveal a critical mechanism that regulates KLF5 expression in breast cancer. Our findings indicate that BAP1 could be a potential therapeutic target for breast and other cancers.

Soil moisture is fundamental to ecosystem sustainability in semi-arid regions, and characterizing the response of temporal soil moisture variation to different vegetation types is important for assessing the sustainability of vegetation restoration. In this study, the soil moisture among eight typical types of vegetation is investigated and compared during three rainy seasons. The temporal variations of soil moisture in the near-surface (0-0.4 m), sub-surface (0.4-1.0 m), and deep layers (1.0-2.0 m) are explored to evaluate the ecohydrological effect of vegetation restoration in the semi-arid Loess Plateau of China. The results show that soil moisture content decreases drastically after vegetation restoration, with no significant difference in near-surface soil moisture among the vegetation types but significant differences in the sub-surface and deep soil layers. Introduced vegetation is the main factor affecting the soil moisture deficit below near-surface layers. Secondly, soil moisture is temporally stable in the sub-surface and deep layers, especially in introduced vegetation. This indicates that introduced vegetation consumes excessive amount of soil moisture and induces temporally stable soil desiccation. Soil desiccation with temporal stability cannot provide enough available soil moisture for plants and will inevitably threaten the sustainability of vegetation restoration and the associated ecosystem services. Lastly, high planting density is the main cause of severe soil moisture deficit on a long-term temporal scale. Our study results suggest that the current planting density of introduced vegetation is too high in specific cases and should be optimized with local soil moisture conditions in semi-arid regions.

Triple-negative breast cancer (TNBC) is currently the most malignant subtype of breast cancers without effective targeted therapies. Mifepristone (MIF), a drug regularly used for abortion, has been reported to have anti-tumor activity in multiple hormone-dependent cancers, including luminal type breast cancers. In this study, we showed that MIF suppressed tumor growth of the TNBC cell lines and patient-derived xenografts in NOD-SCID mice. Furthermore, MIF reduced the TNBC cancer stem cell (CSC) population through down-regulating KLF5 expression, a stem cell transcription factor over-expressed in basal type TNBC and promoting cell proliferation, survival and stemness. Interestingly, MIF suppresses the expression of KLF5 through inducing the expression of miR-153. Consistently, miR-153 decreases CSC and miR-153 inhibitor rescued MIF-induced down-regulation of the KLF5 protein level and CSC ratio. Taken together, our findings suggest that MIF inhibits basal TNBC via the miR-153/KLF5 axis and MIF may be used for the treatment of TNBC.

No abstract

Out of the breast cancer subtypes, triple-negative breast cancer (TNBC) has the poorest prognosis without effective targeted therapies. Metformin, a first-line drug for type 2 diabetes mellitus, was demonstrated to target breast cancer stem cells selectively. However, the efficiency and the mechanism of action of metformin in TNBC are unclear. In this study, we demonstrated that metformin decreased the percentage of TNBC stem cells partially through the downregulation of the expression of the stem cell transcription factor Krüppel-like factor 5 (KLF5) and its downstream target genes, such as Nanog and FGF-BP1, in TNBC cell lines. Metformin induced glycogen synthase kinase-3β (GSK3β)-mediated KLF5 protein phosphorylation and degradation through the inhibition of protein kinase A (PKA) activity in TNBC cells. Consistently, PKA activators increased the expression levels of KLF5. We observed a positive correlation between p-CREB, p-GSK3β, KLF5 and FGF-BP1 protein levels in human TNBC samples. These findings suggest that metformin suppresses TNBC stem cells partially through the PKA-GSK3β-KLF5 signaling pathway.

Convolutional neural networks (CNN) are increasingly used in many areas of computer vision. They are particularly attractive because of their ability to "absorb" great quantities of labeled data through millions of parameters. However, as model sizes increase, so do the storage and memory requirements of the classifiers, hindering many applications such as image and speech recognition on mobile phones and other devices. In this paper, we present a novel network architecture, Frequency-Sensitive Hashed Nets (FreshNets), which exploits inherent redundancy in both convolutional layers and fully-connected layers of a deep learning model, leading to dramatic savings in memory and storage consumption. Based on the key observation that the weights of learned convolutional filters are typically smooth and low-frequency, we first convert filter weights to the frequency domain with a discrete cosine transform (DCT) and use a low-cost hash function to randomly group frequency parameters into hash buckets. All parameters assigned the same hash bucket share a single value learned with standard backpropagation. To further reduce model size, we allocate fewer hash buckets to high-frequency components, which are generally less important. We evaluate FreshNets on eight data sets, and show that it leads to better compressed performance than several relevant baselines.

Training neural network language models over large vocabularies is computationally costly compared to count-based models such as Kneser-Ney. We present a systematic comparison of neural strategies to represent and train large vocabularies, including softmax, hierarchical softmax, target sampling, noise contrastive estimation and self normalization. We extend self normalization to be a proper estimator of likelihood and introduce an efficient variant of softmax. We evaluate each method on three popular benchmarks, examining performance on rare words, the speed/accuracy trade-off and complementarity to Kneser-Ney.