Aurora-A controls chromosome segregation during mitosis, and is amplified and/or over-expressed in a large proportion of human tumors. However, the frequency of Aurora-A overexpression in human cancers is much higher than that of its amplification. For example, in breast cancers, Aurora-A is amplified in 12% of cases whereas it is overexpressed in 98% of cases. The discrepancy between overexpression and amplification suggests that elevated Aurora-A levels are likely mediated through other mechanisms such as post-translational modifications. In this study, we demonstrate that two tumor suppressor genes, Fbxw7 and Pten, regulate Aurora-A levels in a post-translational manner.
Our results reveal that Aurora-A is targeted for ubiquitin-dependent degradation by FBXW7 in a process that is regulated by GSK3β. A recent study showed that overexpression of Aurora-A in AGS and SNU1 gastric cancer cells led to increased phosphorylation and inactivation of Gsk3β (24
), suggesting that a positive feedback loop exists between Gsk3β and Aurora-A. This process may be mediated through Akt since it has been shown that Aurora-A can activate Akt through phosphorylation (24
). As a result, it is possible that Aurora-A activates Akt followed by Akt inactivation of Gsk3β resulting in the increase of Aurora-A levels. Furthermore, based on our results that Pten negatively regulates Aurora-A through AKT/GSK3β, together with the recent report by Taga et. al. (26
) that long-term overexpression of Aurora-A in cells leads to degradation of Pten, it appears that yet another layer of positive feedback may be operating within this complex signaling network (Supplementary Figure S3
Unlike other FBXW7 targets, the consensus CDC4 phosphodegron motif was not found in Aurora-A. Using a series of Aurora-A truncated proteins, we have now identified residues 202 to 230 of the Aurora-A as the minimal region required for binding to FBXW7, which is located within Aurora-A catalytic domain. Moreover, the residues T217/E221 have been defined critical for binding to FBXW7 as our results showed that mutation of these two residues totally abolished the interaction between Aurora-A and FBXW7. Interestingly, recent studies showed that T217 is an important residue for developing small inhibitors that are highly selective for targeting Aurora-A (27
In this study, we also found two GSK3β binding sites in Aurora-A. Our results indicate that both sites are equivalently important for the interaction between Aurora-A and GSK3βMutation of either site alone is insufficient to block their interaction. Surprisingly, we found that the GSK3β binding sites in Aurora-A are distinct from the FBXW7 binding site, which differs from what is found in other FBXW7 targets, such as c-Myc and CCNE1. Mutation of the GSK3β binding sites within Aurora-A or pharmacologic inhibition of GSK3β reduces the binding affinity between Aurora-A and FBXW7, suggesting that GSK3β binding to Aurora-A promotes the interaction between Aurora-A and FBXW7. Nevertheless, both GSK3β and FBXW7 binding sites play an equivalent role in degradation of Aurora-A. Mutation of critical residues in either sites substantially stabilizes Aurora-A.
In conclusion, we have identified a novel interaction between the Fbxw7 and Pten tumor suppressor pathways that controls the levels of the oncoprotein Aurora-A kinase. We show that Fbxw7 degrades Aurora-A through physical interaction. Like most of the other known Fbxw7 substrates, ubiquitination of Aurora-A by Fbxw7 is a GSK3β-dependent event. Furthermore, we demonstrate that loss of Pten also increases the stability of Aurora-A through the Akt/GSK3β pathway. Using a genetic mouse model, we report that mice heterozygous for both Fbxw7 and Pten have significantly reduced tumor latency compared to mice heterozygous for either Fbxw7 or Pten, and normal tissues from Fbxw7+/− or Pten+/− mice have higher protein levels of Aurora-A than the equivalent tissue from wild-type mice. Moreover such higher starting levels of endogenous Aurora-A protein sustained in the tumors generated from Fbxw7+/− and Fbxw7+/−Pten+/− mice and in most of tumors from Pten+/− mice. In view of the toxicity of presently available inhibitors of Aurora-A kinase in clinical trials, these findings may lead to the design of alternative approaches to manipulate the levels of Aurora-A kinase protein for cancer therapy.