BackgroundCircPVT1 is demonstrated to promote cancer progression in esophageal squamous cell carcinoma (ESCC). However, the role and potential functional mechanisms of circPVT1 in regulating 5-fluorouracil (5-FU) chemosensitivity remain largely unknown.MethodsESCC cells resistant to 5-FU were induced with continuous increasing concentrations of 5-FU step-wisely. A cell counting kit-8 assay was used to analyze the viability of ESCC cells. LDH release assay kit was used to evaluate the cytotoxicity. RT-qPCR was used to assess the expression level of non-coding RNAs and cDNAs. Luciferase was used to confirm the interaction between non-coding RNAs and targets. Western blotting was used to detect the expression of downstream signaling proteins. Flow cytometry and ferroptosis detection assay kit were utilized to measure the ferroptosis of ESCC cells.ResultsCircPVT1 was significantly upregulated in ESCC cells resistant to 5-FU. Knockdown of circPVT1 enhanced the 5-FU chemosensitivity of ESCC cells resistant to 5-FU by increasing cytotoxicity and downregulating multidrug-resistant associated proteins, including P-gp and MRP1. Luciferase assay showed that circPVT1 acted as a sponge of miR-30a-5p, and Frizzled3 (FZD3) was a downstream target of miR-30a-5p. The enhanced 5-FU chemosensitivity by circPVT1 knockdown was reversed with miR-30a-5p inhibitor. Besides, the increased 5-FU chemosensitivity by miR-30a-5p mimics was reversed with FZD3 overexpression. Furthermore, knockdown of circPVT1 increased ferroptosis through downregulating p-β-catenin, GPX4, and SLC7A11 while miR-30a-5p inhibition and FZD3 overexpression reversed the phenotype by upregulating p-β-catenin, GPX4, and SLC7A11.ConclusionsThese results suggested a key role for circPVT1 in ESCC 5-FU-chemosensitivity in regulating the Wnt/β-catenin pathway and ferroptosis via miR-30a-5p/FZD3 axis, which might be a potential target in ESCC therapy.

No abstract

No abstract

Hardware accelerators are essential to the accommodation of ever-increasing Deep Neural Network (DNN) workloads on the resource-constrained embedded devices. While accelerators facilitate fast and energy-efficient DNN operations, their accuracy is threatened by faults in their on-chip and off-chip memories, where millions of DNN weights are held. The use of emerging Non-Volatile Memories (NVM) further exposes DNN accelerators to a non-negligible rate of permanent defects due to immature fabrication, limited endurance, and aging. To tolerate defects in NVM-based DNN accelerators, previous work either requires extra redundancy in hardware or performs defect-aware retraining, imposing significant overhead. In comparison, this paper proposes a set of algorithms that exploit the flexibility in setting the fault-free bits in weight memory to effectively approximate weight values, so as to mitigate defect-induced accuracy drop. These algorithms can be applied as a one-step solution when loading the weights to embedded devices. They only require trivial hardware support and impose negligible run-time overhead. Experiments on popular DNN models show that the proposed techniques successfully boost inference accuracy even in the face of elevated defect rates in the weight memory.