Herein, we evaluated the interplay between pVHL and mTORC1 pathways in ccRCC. We show that REDD1 is upregulated in VHL-deficient ccRCC tumors, that VHL disruption is sufficient to upregulate REDD1, and that REDD1 upregulation in ccRCC depends on pVHL and can be mediated by either HIF-1 or HIF-2. Furthermore, our data show that whereas REDD1 is involved in restraining mTORC1 activity in some ccRCCs, in others, mechanisms have evolved to uncouple mTORC1 from REDD1 inhibition. One such mechanism involves the disruption of the TSC1/TSC2 complex and our results implicate TSC1 as a novel tumor suppressor gene in sporadic ccRCC.
REDD1 is broadly upregulated in VHL
-deficient ccRCCs and in vivo
as well as in vitro
experiments indicate that REDD1 regulation is VHL
-dependent. As VHL
inactivation and HIF stabilization are thought to be among the earliest molecular events in renal tumorigenesis, it is not surprising that REDD1 can already be observed upregulated in renal cysts in von Hippel-Lindau patients (data not shown). REDD1 upregulation in ccRCC requires HIF, and despite that differences exist in ccRCC depending upon whether HIF-1α or HIF-2α is expressed (24
), both HIF-1 and HIF-2 are involved in the regulation of REDD1, and as supported by our ChIP studies, REDD1
is directly acted upon by both. Nonetheless, should preferential regulation by HIF-1 occur in tumors, these data could contribute to explain why HIF-1 may function as a tumor suppressor in ccRCC.
Importantly, our results show that mTORC1 is regulated by REDD1 in ccRCC in a context-dependent manner. The two extremes are illustrated by Caki-2 and 786-O cell lines. In Caki-2 cells, REDD1 is engaged in mTORC1 inhibition suggesting that REDD1 functions in a negative feedback loop to downregulate mTORC1 following VHL
disruption. These data are consistent with the findings that acute disruption of Vhl
in immortalized and primary MEFs has previously been shown to lead to senescence (44
). We conjecture that mTORC1 remains responsive to REDD1 in Caki-2 cells because, as a consequence of a partially active pVHL, the level of REDD1 upregulation is modest, such that the selective pressure to uncouple mTORC1 from REDD1 is lower than in 786-O cells. By contrast, REDD1 levels are high in 786-O cells, but mTORC1 is insensitive to it.
Loss of function mutations in TSC1 activate mTORC1 and result in its uncoupling from REDD1 as shown here with the TSC1 splice site mutant. Three additional TSC1 mutations were identified in the 77 tumors, including two truncating mutations that were confirmed to be somatically acquired. To our knowledge, this is the first report implicating TSC1 as a tumor suppressor in sporadic ccRCC. As for PTEN, one somatically acquired inactivating mutation was found in REDD1, suggesting that, although rarely, mutations in REDD1 may similarly contribute to ccRCC development.
By contrast to TSC1
, somatic mutations in TSC2
were not identified. A potential explanation for these findings would be furnished by the existence of a second tumor suppressor gene in the proximity of TSC1
such that regional deletions may result in the simultaneous loss of the remaining wild-type copy for the two (or more) tumor suppressor genes. In fact, precedent for this exists in ccRCC and a tumor suppressor gene, PBRM1
, was recently identified in relative proximity to VHL
mTORC1 has been previously reported to be activated in 60–85% of ccRCCs (1
). While these studies were based on phospho-S6S235/236
signal, which is thought to be less specific than phospho-S6S240/244
), we found that approximately 80% of ccRCC were positive for phospho-S6S240/244
, adding support to the notion that mTORC1 is broadly activated in this tumor type.
Given the low frequency of mutations in TSC1
, other mechanisms must exist to prevent mTORC1 inhibition by REDD1. While targeted sequencing studies have failed to identify activating mutations in Rheb
), recently, mutations in the mTOR
gene itself were identified in ccRCC (46
). Furthermore, tumor associated mTOR
mutants have been found to lead to mTORC1 activation and diminished mTORC1 inhibition by hypoxia (48
). Nevertheless, these mutations are also rare. It is also possible that inactivation of a single TSC1
allele, and TSC1
is found in 9q, a region that is deleted in approximately 20% of ccRCC (31
), may be sufficient to activate mTORC1 and render it unresponsive to REDD1. In keeping with this idea, modest depletion of TSC2 appears to be sufficient to block REDD1-induced mTORC1 inhibition (8
Understanding how mTORC1 is deregulated in ccRCC may pave the way for the identification of patients most likely to benefit from mTORC1 inhibitors. It would be expected that only tumors with active mTORC1 would respond to its inhibition and this is supported by a small retrospective correlative study (49
). While the phosphorylation state of mTORC1 effector proteins may be used as the readout for mTORC1 activation, TORC1 is a critical regulator of ribosomal biogenesis and nucleolar size (50
) and conceivably, nucleolar dimensions could serve as a surrogate for mTORC1 activity in tumors. In RCC, nucleolar prominence is a major determinant of the prognostic Fuhrman grading scale (52
). The regulation of nucleolar size by mTORC1 provides an explanation for the positive correlation previously reported between phospho-S6 and tumor grade (1
). This raises the possibility that the activation state of mTORC1 may contribute to the prognostic significance of the Fuhrman grading scale. Furthermore, it is possible that nucleolar size could serve as a pharmacodynamic indicator of mTORC1 inhibition in RCC. Should nucleolar prominence be dynamically regulated by mTORC1, the exposure of patients to mTORC1 inhibitors could affect Fuhrman grading. Supporting this concern, rapamycin was previously shown to reduce nucleolar size in both mammalian and yeast cells (51
Driver mutations in tumors may reflect a state of addiction that could be exploited therapeutically. In this context, and despite that the TSC1/TSC2 complex likely regulates other processes besides mTORC1 (53
), mutations in TSC1
in ccRCC could portend a state of addiction to mTORC1. While this represents a single instance, it is noteworthy that the index patient who had the original TSC1
splice site mutation we evaluated had an extraordinary response to everolimus in the second line. Whereas the median progression-free interval on everolimus in the second line in the pivotal phase III clinical trial was 4 months (4
), the patient remained on everolimus without progression for 13 months, and this was despite progression to sunitinib in 3 months.
In summary, this work begins to unravel the complexity of signaling pathways linking pVHL and mTORC1 in ccRCC and suggests that mechanisms have evolved in tumors to escape growth suppressive signals resulting from VHL loss and REDD1 upregulation.