Design, synthesis and in vitro anticancer research of novel tetrandrine and fangchinoline derivatives.

Cancer treatment is one of the major public health issues in the world. Tetrandrine (Tet) and fangchinoline (d-Tet) are two bis-benzyl isoquinoline alkaloids extracted from Stephania tetrandra S. Moore, and their antitumor activities have been confirmed. However, the effective dose of Tet and d-Tet were much higher than that of the positive control and failed to meet clinical standards. Therefore, in this study, as a continuation of our previous work to study and develop high-efficiency and low-toxic anti-tumor lead compounds, twenty new Tet and d-Tet derivatives were designed, synthesized and evaluated as antitumor agents against six cancer cell lines (H460, H520, HeLa, HepG-2, MCF-7, SW480 cell lines) and BEAS-2B normal cells by CCK-8 analysis. Ten derivatives showed better cytotoxic effects than the parent fangchinoline, of which 4g showed the strongest cell growth inhibitory activity with an IC50 value of 0.59 µM against A549 cells. Subsequently, the antitumor mechanism of 4g was studied by flow cytometry, Hoechst 33258, JC-1 staining, cell scratch, transwell migration, and Western blotting assays. These results showed that compound 4g could inhibit A549 cell proliferation by arresting the G2/M cell cycle and inhibiting cell migration and invasion by reducing MMP-2 and MMP-9 expression. Meanwhile, 4g could induce apoptosis of A549 cells through the intrinsic pathway regulated by mitochondria. In addition, compound 4g inhibited the phosphorylation of PI3K, Akt and mTOR, suggesting a correlation between blocking the PI3K/Akt/mTOR pathway and the above antitumor activities. These results suggest that compound 4g may be a future drug for the development of new potential drug candidates against lung cancer.

Deep Regionlets: Blended Representation and Deep Learning for Generic Object Detection.

In this article, we propose a novel object detection algorithm named "Deep Regionlets" by integrating deep neural networks and a conventional detection schema for accurate generic object detection. Motivated by the effectiveness of regionlets for modeling object deformations and multiple aspect ratios, we incorporate regionlets into an end-to-end trainable deep learning framework. The deep regionlets framework consists of a region selection network and a deep regionlet learning module. Specifically, given a detection bounding box proposal, the region selection network provides guidance on where to select sub-regions from which features can be learned from. An object proposal typically contains three - 16 sub-regions. The regionlet learning module focuses on local feature selection and transformations to alleviate the effects of appearance variations. To this end, we first realize non-rectangular region selection within the detection framework to accommodate variations in object appearance. Moreover, we design a "gating network" within the regionlet leaning module to enable instance dependent soft feature selection and pooling. The Deep Regionlets framework is trained end-to-end without additional efforts. We present ablation studies and extensive experiments on the PASCAL VOC dataset and the Microsoft COCO dataset. The proposed method yields competitive performance over state-of-the-art algorithms, such as RetinaNet and Mask R-CNN, even without additional segmentation labels.