We report that chlamydiae, which are obligate intracellular bacterial pathogens, possess a novel antiapoptotic mechanism. Chlamydia-infected host cells are profoundly resistant to apoptosis induced by a wide spectrum of proapoptotic stimuli including the kinase inhibitor staurosporine, the DNA-damaging agent etoposide, and several immunological apoptosis-inducing molecules such as tumor necrosis factor-α, Fas antibody, and granzyme B/perforin. The antiapoptotic activity was dependent on chlamydial but not host protein synthesis. These observations suggest that chlamydia may encode factors that interrupt many different host cell apoptotic pathways. We found that activation of the downstream caspase 3 and cleavage of poly (ADP-ribose) polymerase were inhibited in chlamydia-infected cells. Mitochondrial cytochrome c release into the cytosol induced by proapoptotic factors was also prevented by chlamydial infection. These observations suggest that chlamydial proteins may interrupt diverse apoptotic pathways by blocking mitochondrial cytochrome c release, a central step proposed to convert the upstream private pathways into an effector apoptotic pathway for amplification of downstream caspases. Thus, we have identified a chlamydial antiapoptosis mechanism(s) that will help define chlamydial pathogenesis and may also provide information about the central mechanisms regulating host cell apoptosis.

No abstract

No abstract

Training robust deep video representations has proven to be much more challenging than learning deep image representations. This is in part due to the enormous size of raw video streams and the high temporal redundancy; the true and interesting signal is often drowned in too much irrelevant data. Motivated by that the superfluous information can be reduced by up to two orders of magnitude by video compression (using H.264, HEVC, etc.), we propose to train a deep network directly on the compressed video.This representation has a higher information density, and we found the training to be easier. In addition, the signals in a compressed video provide free, albeit noisy, motion information. We propose novel techniques to use them effectively. Our approach is about 4.6 times faster than Res3D and 2.7 times faster than ResNet-152. On the task of action recognition, our approach outperforms all the other methods on the UCF-101, HMDB-51, and Charades dataset.