The evolutionarily conserved Ser-Thr protein kinase mTOR functions as the core catalytic component of two structurally and functionally distinct signaling complexes. mTOR complex 1 (mTORC1) regulates protein translation, autophagy and cell growth whereas mTOR complex 2 (mTORC2) regulates the actin cytoskeleton and cell survival (1
). mTORC1 and mTORC2 respond to upstream inputs such as growth factors, energetic status, and amino acid levels (3
) but relatively few downstream targets of mTOR have been identified.
Misregulated mTOR activity is a common feature of most cancers (1
) but clinical trials evaluating the mTORC1 selective inhibitor rapamycin as an anti-cancer agent have met with limited success (2
). Rapamycin resistance has emerged as a major challenge to its clinical use (4
), and is caused in part by feedback loops that activate the PI3K and ERK-MAPK signaling pathways in rapamycin treated cells through poorly understood mechanisms (5
). Identifying substrates of mTORC1 and mTORC2 will be important for understanding how mTOR signals downstream, and for defining components of feedback loops involved in rapamycin resistance.
We performed two sets of large-scale, quantitative phospho-proteomics experiments to characterize the signaling network downstream of mTOR (, S1, S2 and S3
). The first stable isotope labeling with amino acids in cell culture (SILAC) experiment (Rapa screen) was performed using Tsc2
−/− mouse embryonic fibroblasts (MEFs) (see supplemental text
for detailed description of the screen). We identified 4,484 and 6,832 unique phosphorylation sites on 1,615 and 1,866 proteins from two biological replicate experiments, respectively (Table S1
, Databases S1
Fig. 1 Sample preparation and data analysis for quantitative phosphoproteomic profiling of the mTOR downstream signaling networks. (A) Schematics of the two SILAC mass spectrometry experiments are shown with a plot highlighting the ratio distribution of phosphopeptides (more ...)
Several hundred peptides corresponding to 85 and 147 proteins in the two replicates (Database S3
and Fig. S1
) were determined to contain rapamycin-sensitive phosphorylation sites (defined as phosphorylated peptides in control cells whose abundance were more than twice that in samples from rapamyin-treated cells). Many known effectors of the canonical mTORC1 signaling pathway were identified in the downregulated population, including p70S6K, 4EBP1/2, Akt1s1 (PRAS40), rpS6, eIF4B, eIF4G1 and GSK3β (Table S2, Figs. 1C and 1D
). A representitive identification of the known rapamycin-sensitive phosphorylation sites on rpS6 is shown in . In addition, the identification of many kinases, e.g. unc-51-like kinase 1 (ULK1), in the downregulated proteins provides potential points for signal integration and crosstalk (Table S2, see supplementary text
and Table S3
for Gene Ontology (GO) analysis and detailed discussion of the hits).
Rapamycin is an allosteric inhibitor that only partially inhibits mTORC1 signaling and has no effect on the activity of mTORC2 under short-term treatment conditions (2
). In contrast, ATP-competitive mTOR inhibitors block the activity of both mTORC1 and mTORC2 (1
). To identify rapamycin-insensitive mTORC1, and mTORC2 substrates, we used the mTOR inhibitor Ku-0063794 and performed a second SILAC experiment (Ku screen) ( and S1
). In this experiment, one hundred proteins were determined to contain downregulated phosphopeptides after Ku-0063794 treatment (Database S3
, Table S2, see supplementary text
for detailed discussion).
One of the enriched GO classes of hits in the Rapa screen is the receptor protein tyrosine kinase (RTK) signaling pathway (P
= 0.01, Table S3
), suggesting that mTORC1 might modulate its upstream regulators by altering the activities of RTKs. In particular, phosphorylation of S501 and S503 on the growth factor receptor-bound protein 10 (Grb10) was strongly inhibited by a 2-hr rapamycin treatment ( and S4A, Table S2
). The intensity of a triply phosphorylated Grb10 peptide (T76, S96 and S104) also decreased after rapamycin treatment (Table S2
Fig. 2 Sensitivity of phosphorylation of Grb10 at S501 and S503 to rapamycin inhibition (A) Identification of a doubly-phosphorylated, rapamycin-sensitive Grb10 peptide (MNILSS*QS*PLHPSTLNAVIHR, asterisk indicates the site of phosphorylation at S501 and S503). (more ...)
We developed a phospho-specific antibody (Figs. S5A and S5B
) and found rapamycin treatment induced rapid dephosphorylation of Grb10 at S501 and S503 (). Grb10 phosphorylation was also decreased in Tsc2
−/− MEFs deprived of amino acids (). To determine whether S501 and S503 of Grb10 can be phosphorylated by other kinases, we treated Tsc2
−/− cells with staurosporine, a broad-spectrum kinase inhibitor that, however, does not suppress mTOR activity (7
). No change in the phosphorylation of Grb10 was observed (). S6K activity was inhibited by staurosporine treatment, as shown by a complete loss of rpS6 phosphorylation, suggesting that Grb10 was directly phosphorylated by mTORC1 rather than by S6K.
In wt MEFs, insulin or serum stimulation increased Grb10 phosphorylation in a rapamycin-sensitive manner (). Akt inhibition also reduced Grb10 and S6 phosphorylation. Grb10 S503 can be phosphorylated by ERK in vitro
). Inhibition of the ERK-activating kinase, MEK with AZD6244 abolished phosphorylation of ERK but had no effect on phosphorylation of Grb10 (), indicating that phosphorylation of S501 and S503 on Grb10 is not mediated by ERK in vivo
. All other mTOR catalytic inhibitors tested, including LY294002, NVP-BEZ235, torin and pp242 (9
), also completely abolished Grb10 phosphorylation ().
To examine a potential interaction between Grb10 and the mTOR complexes in vivo, we expressed HA-tagged Grb10 with Myc-tagged raptor or rictor in human embryonic kidney (HEK) 293T cells. Grb10 interacted with raptor, but not rictor, indicating that Grb10 is a binding partner of mTORC1, but not mTORC2 (). Grb10 was also phosphorylated by recombinant mTOR at S501 and S503 in vitro ().
Fig. 3 Effect of mTOR-mediated phosphorylation to promote stability of Grb10. (A) Grb10 interacts with raptor, but not rictor. HA-tagged Grb10 was transfected with Myc-raptor or Myc-rictor into HEK293T cells. Cells were lysed in lysis buffer A and the lysates (more ...)
Grb10 was much more abundant in Tsc2
−/− and Tsc1
−/− MEFs than in their wild-type counterparts ( and S5C
) and the initial loss of Grb10 phosphorylation as a result of rapamycin treatment was followed by a decrease in Grb10 abundance and a smaller decrease in amount of Grb10 mRNA (). Exposure to a proteosome inhibitor MG-132 suppressed rapamycin-induced Grb10 protein degradation (Fig. S5D
). These results show that mTORC1 functions to promote accumulation of Grb10 both transcriptionally and post-translationally. Depletion of the mTORC1 component raptor also led to decreased abundance of Grb10 protein (). Furthermore, long-term treatment with mTOR catalytic inhibitors led to reduced levels of Grb10 in Tsc2
−/− MEFs (Fig. S5E
−/− MEFs (Fig. S5F
) and HeLa cells (Fig. S5G
To explore whether Grb10 S501 and S503 phosphorylation contributed to its stabilization and high expression, we transfected wt-Grb10, Grb10-S501A–S503A (AA) and Grb10-S501D–S503D (DD) into HEK293T cells. Exogenous wt and DD Grb10 proteins were expressed in similar amount, but the AA mutant of Grb10 was less abundant, perhaps due to protein instability (). To confirm this result, we generated Tsc2
−/− MEFs stably expressing the HA-tagged Grb10-DD. Long-term rapamycin treatment of these cells decreased amounts of the endogenous, wt Grb10 but had no effect on the abundance of the DD mutant protein (). This result appears not to result from protein overexpression, because in Tsc2
−/− cells expressing HA-tagged wt Grb10, rapamycin treatment decreased the abundance of both the endogenous and ectopically expressed Grb10 (Fig. S5H
). These data support a critical role for mTORC1 in stabilizing Grb10 through phosphorylation of the S501 and S503 residues. It is important to note that phospho-Akt levels still increased in these cells (), likely resulting from increased IRS1 levels after prolonged rapamycin treatment.
Grb10 functions as a negative regulator of insulin signaling. In Grb10 null mice, PI3K-Akt pathway hyperactivation was observed in insulin-sensitive tissues (10
). We therefore examined the possibility that mTORC1-mediated Grb10 phosphorylation and accumulation activated a negative feedback loop from mTORC1 to the PI3K-Akt pathway. The PI3K-Akt and ERK-MAPK pathways were both refractory to insulin or IGF stimulation in Tsc2
−/− cells, as a result of constitutively elevated mTORC1 signaling (5
). In contrast, phosphorylation of both Akt and ERK was increased in Grb10-depleted cells deprived of serum or stimulated with insulin or IGF ( and S6A
). Conversely, overexpression of Grb10 in HEK293 cells suppressed activation of PI3K (Fig. S6B
) by inhibiting insulin receptor-dependent phosphorylation of insulin receptor substrate (IRS) and its subsequent recruitment of PI3K (Figs. S6C and S6D
). Knockdown of Grb10 in Tsc2
−/− MEFs to a level close to that in wt cells did not completely restore the sensitivity of PI3K to insulin stimulation (Fig. S6E
), suggesting additional mechanisms (e.g. lower IRS levels in Tsc
2−/− MEFs) contribute to the feedback inhibition (Fig. S6F
). Our data complement the previous findings and suggest that activation of mTORC1-S6K promotes negative feedback inhibition of PI3K through a two-prong mechanism: first, mTORC1-S6K-mediated phosphorylation and degradation of a positive regulator of PI3K signaling, IRS (6
); second, mTORC1-mediated phosphorylation and accumulation of a negative regulator of PI3K signaling, Grb10.
Fig. 4 Grb10 is involved in the feedback inhibition loop from mTORC1 to PI3K and ERK-MAPK, and GRB10 mRNA expression is decreased in abundance in many cancers and is negatively correlated with PTEN expression. (A) Knockdown of Grb10 in Tsc2−/− (more ...)
We next asked if PI3K activation in Grb10-depleted cells would promote survival against stress-induced apoptosis. In response to either staurosporine or etoposide, reduced caspase 3 cleavage was observed in Grb10 knockdown cells compared to that of control cells, indicating that Grb10 depletion is sufficient to protect cells from apoptosis ( and S6G
). Because rapamycin can protect cells from energy stress-induced death (15
), these results provide additional possible explanations for the cytostatic rather than cytotoxic effects of rapamycin in some cancers, and suggest a complete understanding of the feedback inhibition control will be critical in designing combination therapies involving rapamycin analogues.
Comprehensive meta-analysis of published microarray data revealed that the abundance of GRB10
was decreased in many tumor types compared to that in normal tissue counterparts (). Given that loss of Grb10 results in activation of the PI3K-Akt pathway (), we performed correlation analysis and found that there was a significant (p < 0.05) negative correlation between GRB10
expression (). This correlation was only observed in tumor samples but not in normal tissue controls (). PIK3CA
mutations and PTEN
loss are mutually exclusive in breast cancer (16
), suggesting that increased abundance of PIP3
resulting from of genetic alteration of either PIK3CA
relieves selective pressure targeting the other gene. Similarly, loss of Grb10, which results in PI3K activation, might provide the cells with growth and survival advantages that are redundant with respect to PTEN loss-of-function, suggesting that Grb10 might be a tumor suppressor that is regulated by mTORC1. These data point to the prospect of targeting Grb10 stability in cancer therapy.