Dysregulation of androgen signaling and pericellular proteolysis is necessary for prostate cancer progression, but the links between them are still obscure. In this study, we show how the membrane-anchored serine protease TMPRSS2 stimulates a proteolytic cascade that mediates androgen-induced prostate cancer cell invasion, tumor growth, and metastasis. We found that matriptase serves as a substrate for TMPRSS2 in mediating this proinvasive action of androgens in prostate cancer. Further, we determined that higher levels of TMPRSS2 expression correlate with higher levels of matriptase activation in prostate cancer tissues. Lastly, we found that the ability of TMPRSS2 to promote prostate cancer tumor growth and metastasis was associated with increased matriptase activation and enhanced degradation of extracellular matrix nidogen-1 and laminin b1 in tumor xenografts. In summary, our results establish that TMPRSS2 promotes the growth, invasion, and metastasis of prostate cancer cells via matriptase activation and extracellular matrix disruption, with implications to target these two proteases as a strategy to treat prostate cancer.

No abstract

TMPRSS2 is an important membrane-anchored serine protease involved in human prostate cancer progression and metastasis. A serine protease physiologically often comes together with a cognate inhibitor for execution of proteolytically biologic function; however, TMPRSS2's cognate inhibitor is still elusive. To identify the cognate inhibitor of TMPRSS2, in this study, we applied co-immunoprecipitation and LC/MS/MS analysis and isolated hepatocyte growth factor activator inhibitors (HAIs) to be potential inhibitor candidates for TMPRSS2. Moreover, the recombinant HAI-2 proteins exhibited a better inhibitory effect on TMPRSS2 proteolytic activity than HAI-1, and recombinant HAI-2 proteins had a high affinity to form a complex with TMPRSS2. The immunofluorescence images further showed that TMPRSS2 was co-localized to HAI-2. Both KD1 and KD2 domain of HAI-2 showed comparable inhibitory effects on TMPRSS2 proteolytic activity. In addition, HAI-2 overexpression could suppress the induction effect of TMPRSS2 on pro-HGF activation, extracellular matrix degradation and prostate cancer cell invasion. We further determined that the expression levels of TMPRSS2 were inversely correlated with HAI-2 levels during prostate cancer progression. In orthotopic xenograft animal model, TMPRSS2 overexpression promoted prostate cancer metastasis, and HAI-2 overexpression efficiently blocked TMPRSS2-induced metastasis. In summary, the results together indicate that HAI-2 can function as a cognate inhibitor for TMPRSS2 in human prostate cancer cells and may serve as a potential factor to suppress TMPRSS2-mediated malignancy.

No abstract

Reinforcement learning (RL) can enable task-oriented dialogue systems to steer the conversation towards successful task completion. In an end-to-end setting, a response can be constructed in a word-level sequential decision making process with the entire system vocabulary as action space.Policies trained in such a fashion do not require expert-defined action spaces, but they have to deal with large action spaces and long trajectories, making RL impractical. Using the latent space of a variational model as action space alleviates this problem. However, current approaches use an uninformed prior for training and optimize the latent distribution solely on the context. It is therefore unclear whether the latent representation truly encodes the characteristics of different actions. In this paper, we explore three ways of leveraging an auxiliary task to shape the latent variable distribution: via pre-training, to obtain an informed prior, and via multitask learning. We choose response auto-encoding as the auxiliary task, as this captures the generative factors of dialogue responses while requiring low computational cost and neither additional data nor labels. Our approach yields a more action-characterized latent representations which support end-to-end dialogue policy optimization and achieves state-of-the-art success rates. These results warrant a more wide-spread use of RL in end-to-end dialogue models.