The four receptors of the Notch family are widely expressed transmembrane proteins that function as key conduits through which mammalian cells communicate to regulate cell fate and growth. Ligand binding triggers a conformational change in the receptor negative regulatory region (NRR) that enables ADAM protease cleavage at a juxtamembrane site that otherwise lies buried within the quiescent NRR. Subsequent intramembrane proteolysis catalysed by the gamma-secretase complex liberates the intracellular domain (ICD) to initiate the downstream Notch transcriptional program. Aberrant signalling through each receptor has been linked to numerous diseases, particularly cancer, making the Notch pathway a compelling target for new drugs. Although gamma-secretase inhibitors (GSIs) have progressed into the clinic, GSIs fail to distinguish individual Notch receptors, inhibit other signalling pathways and cause intestinal toxicity, attributed to dual inhibition of Notch1 and 2 (ref. 11). To elucidate the discrete functions of Notch1 and Notch2 and develop clinically relevant inhibitors that reduce intestinal toxicity, we used phage display technology to generate highly specialized antibodies that specifically antagonize each receptor paralogue and yet cross-react with the human and mouse sequences, enabling the discrimination of Notch1 versus Notch2 function in human patients and rodent models. Our co-crystal structure shows that the inhibitory mechanism relies on stabilizing NRR quiescence. Selective blocking of Notch1 inhibits tumour growth in pre-clinical models through two mechanisms: inhibition of cancer cell growth and deregulation of angiogenesis. Whereas inhibition of Notch1 plus Notch2 causes severe intestinal toxicity, inhibition of either receptor alone reduces or avoids this effect, demonstrating a clear advantage over pan-Notch inhibitors. Our studies emphasize the value of paralogue-specific antagonists in dissecting the contributions of distinct Notch receptors to differentiation and disease and reveal the therapeutic promise in targeting Notch1 and Notch2 independently.
Polycomb repressive complex 2 (PRC2) consists of three core subunits, EZH2, EED and SUZ12, and plays pivotal roles in transcriptional regulation. The catalytic subunit EZH2 methylates histone H3 lysine 27 (H3K27), and its activity is further enhanced by the binding of EED to trimethylated H3K27 (H3K27me3). Small-molecule inhibitors that compete with the cofactor S-adenosylmethionine (SAM) have been reported. Here we report the discovery of EED226, a potent and selective PRC2 inhibitor that directly binds to the H3K27me3 binding pocket of EED. EED226 induces a conformational change upon binding EED, leading to loss of PRC2 activity. EED226 shows similar activity to SAM-competitive inhibitors in blocking H3K27 methylation of PRC2 target genes and inducing regression of human lymphoma xenograft tumors. Interestingly, EED226 also effectively inhibits PRC2 containing a mutant EZH2 protein resistant to SAM-competitive inhibitors. Together, we show that EED226 inhibits PRC2 activity via an allosteric mechanism and offers an opportunity for treatment of PRC2-dependent cancers.
Overexpression and somatic heterozygous mutations of EZH2, the catalytic subunit of polycomb repressive complex 2 (PRC2), are associated with several tumor types. EZH2 inhibitor, EPZ-6438 (tazemetostat), demonstrated clinical efficacy in patients with acceptable safety profile as monotherapy. EED, another subunit of PRC2 complex, is essential for its histone methyltransferase activity through direct binding to trimethylated lysine 27 on histone 3 (H3K27Me3). Herein we disclose the discovery of a first-in-class potent, selective, and orally bioavailable EED inhibitor compound 43 (EED226). Guided by X-ray crystallography, compound 43 was discovered by fragmentation and regrowth of compound 7, a PRC2 HTS hit that directly binds EED. The ensuing scaffold hopping followed by multiparameter optimization led to the discovery of 43. Compound 43 induces robust and sustained tumor regression in EZH2 preclinical DLBCL model. For the first time we demonstrate that specific and direct inhibition of EED can be effective as an anticancer strategy.
Hardware accelerators are being increasingly deployed to boost the performance and energy efficiency of deep neural network (DNN) inference. In this paper we propose Thundervolt, a new framework that enables aggressive voltage underscaling of high-performance DNN accelerators without compromising classification accuracy even in the presence of high timing error rates. Using post-synthesis timing simulations of a DNN accelerator modeled on the Google TPU, we show that Thundervolt enables between 34%-57% energy savings on stateof-the-art speech and image recognition benchmarks with less than 1% loss in classification accuracy and no performance loss. Further, we show that Thundervolt is synergistic with and can further increase the energy efficiency of commonly used run-time DNN pruning techniques like Zero-Skip.
Due to their growing popularity and computational cost, deep neural networks (DNNs) are being targeted for hardware acceleration. A popular architecture for DNN acceleration, adopted by the Google Tensor Processing Unit (TPU), utilizes a systolic array based matrix multiplication unit at its core. This paper deals with the design of faulttolerant, systolic array based DNN accelerators for high defect rate technologies. To this end, we empirically show that the classification accuracy of a baseline TPU drops significantly even at extremely low fault rates (as low as 0.006%). We then propose two novel strategies, fault-aware pruning (FAP) and fault-aware pruning+retraining (FAP+T), that enable the TPU to operate at fault rates of up to 50%, with negligible drop in classification accuracy (as low as 0.1%) and no run-time performance overhead. The FAP+T does introduce a one-time retraining penalty per TPU chip before it is deployed, but we propose optimizations that reduce this one-time penalty to under 12 minutes. The penalty is then amortized over the entire lifetime of the TPU's operation.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
customersupport@researchsolutions.com
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.