LetterCell Biology

Comment on “A Dynamic Network Model of mTOR Signaling Reveals TSC-Independent mTORC2 Regulation”: Building a Model of the mTOR Signaling Network with a Potentially Faulty Tool

See allHide authors and affiliations

Science Signaling  10 Jul 2012:
Vol. 5, Issue 232, pp. lc3
DOI: 10.1126/scisignal.2003250


In their study published in Science Signaling (Research Article, 27 March 2012, DOI: 10.1126/scisignal.2002469), Dalle Pezze et al. tackle the dynamic and complex wiring of the signaling network involving the protein kinase mTOR, which exists within two distinct protein complexes (mTORC1 and mTORC2) that differ in their regulation and function. The authors use a combination of immunoblotting for specific phosphorylation events and computational modeling. The primary experimental tool employed is to monitor the autophosphorylation of mTOR on Ser2481 in cell lysates as a surrogate for mTOR activity, which the authors conclude is a specific readout for mTORC2. However, Ser2481 phosphorylation occurs on both mTORC1 and mTORC2 and will dynamically change as the network through which these two complexes are connected is manipulated. Therefore, models of mTOR network regulation built using this tool are inherently imperfect and open to alternative explanations. Specific issues with the main conclusion made in this study, involving the TSC1-TSC2 (tuberous sclerosis complex 1 and 2) complex and its potential regulation of mTORC2, are discussed here. A broader goal of this Letter is to clarify to other investigators the caveats of using mTOR Ser2481 phosphorylation in cell lysates as a specific readout for either of the two mTOR complexes.

Much of the complexity, and often confusion, with the mammalian or mechanistic target of rapamycin (mTOR) signaling network stems from the fact that the kinase mTOR exists within two physically and functionally distinct complexes that are indirectly linked through various signaling events. The Dalle Pezze et al. study (1) focuses on mTOR complex 2 (mTORC2), the more recently discovered of the two mTOR-containing complexes, which is stimulated by insulin and other growth factors through unknown mechanisms (2). From their combined analyses, the authors conclude that the TSC1-TSC2 complex, an established negative regulator of mTOR complex 1 (mTORC1) has no input into the regulation of mTORC2. This conclusion opposes the findings of previous biochemical and genetic studies indicating that loss of function of the TSC1-TSC2 complex in mouse and human cell lines and tumors results in an attenuation of mTORC2 kinase activity and downstream signaling (3, 4). The positive influence of the TSC1-TSC2 complex on mTORC2 activation appears to be mediated through both inhibitory effects on mTORC1-driven feedback mechanisms (5-10) and an mTORC1-independent mechanism (3, 4), the molecular nature of which remains to be elucidated. Dalle Pezze et al. conclude from their study that none of these mechanisms affect mTORC2.

The likely explanation for the discrepancy in conclusions between the Dalle Pezze et al. study and previous studies is discussed here. In explaining their experimental design, the authors emphasize the important point that monitoring the phosphorylation status of the mTORC2 substrate Akt on Ser473 in cell lysates is not a specific or reliable readout of the activation state of mTORC2, an argument that was put forward previously (11). An assay developed by Sarbassov and colleagues (2), the direct measurement of mTORC2 kinase activity after specific immunoprecipitation of the complex, has emerged as the gold standard in the field. It was this assay that was used as the primary basis for the previous conclusion that the TSC1-TSC2 complex is required for the growth factor–stimulated activation of mTORC2 (3). However, the authors chose, instead, to use a more high-throughput approach based on mTOR autophosphorylation in cell lysates.

All of the major conclusions made in the Dalle Pezze et al. study hinge on the phosphorylation of mTOR on Ser2481 in cell lysates being a very specific readout of mTORC2 activity that is not influenced by mTOR within mTORC1. Several issues arise with this assumption. First, mTOR is reportedly phosphorylated on Ser2481 within both mTORC1 and mTORC2, the amount of which varies based on cellular growth conditions (12, 13). In the Dalle Pezze et al. study, the minimal validation of Ser2481 phosphorylation as a readout of mTORC2 activity in their chosen system (HeLa cells) is only performed at a steady-state condition or single time point of stimulation, distinct from every other time-course experiment in the study. Even under this condition, Ser2481 phosphorylation on mTORC1 is evident in their data. The stimulation paradigm employed in the rest of the study involves starving the cells of all growth factors and amino acids, followed by a 2-hour time course of stimulation with both. It is well known that mTORC1 and mTORC2 are distinct in their response to growth factors and amino acids [reviewed in (14)], and how Ser2481 phosphorylation within the two complexes changes dynamically over time after acute stimulation with both is not determined in this study.

Second, disrupting or differentially activating one of the complexes will affect the abundance, activity, and phosphorylation state of the other complex. In all experiments using inhibitors, stimulators, and RNA interference knockdowns, the authors are measuring a combination of phosphorylation on mTORC1 and mTORC2, thereby greatly complicating a study on the relationship between these two dynamically regulated complexes. Throughout the study, the authors fail to measure the phosphorylation of mTOR specific to each complex by immunopurifying the complexes, and they disregard the fact that the relative phosphorylation of mTOR within each complex will change under the conditions assayed. For instance, knockdown of the mTORC1 component Raptor will enhance Ser2481 phosphorylation on mTOR within mTORC2, as more mTOR will be available to form this complex. Because mTORC1 is no longer contributing to the overall signal, this can easily be misread as no change in phosphorylation. On the other hand, knockdown of the mTORC2 component Rictor could decrease the phosphorylation of both complexes, because mTORC1 can be activated downstream of mTORC2 through Akt. With respect to the main findings regarding involvement of the TSC1-TSC2 complex, Fingar and colleagues (13) have shown that Ser2481 phosphorylation of mTOR within mTORC1 is robustly increased in TSC1-knockout cells or those transiently overexpressing the downstream target of the TSC1-TSC2 complex Rheb. Therefore, it is impossible to assess mTORC2 activity in response to TSC2 knockdown based on mTOR Ser2481 phosphorylation alone, because it is expected to be increased within mTORC1, thereby masking any effects on mTORC2.

A third consideration is that no study to date has determined how autophosphorylation of Ser2481 relates to the kinase activity of mTOR toward known direct substrates of either mTORC1 or mTORC2, such as ribosomal S6 kinase 1 (S6K1) and Akt, respectively. In fact, the experiments in the Dalle Pezze et al. paper demonstrate that mTOR Ser2481 phosphorylation tracks with the mTORC1-dependent phosphorylation of S6K1 rather than the mTORC2-dependent phosphorylation of Akt. The authors conclude that there must be a second Akt Ser473 kinase stimulated by insulin, rather than suggesting that this autophosphorylation event may have little to do with the regulated activity of mTORC2 toward its endogenous substrates. A knockdown of the mTORC2 component Rictor would have resolved this.

Finally, and most importantly, the authors do not directly measure the kinase activity of mTORC2, using the gold-standard assay, as validation of their final conclusions. In a previous study, the kinase activity of mTORC2 was decreased in both Tsc1−/− and Tsc2−/− mouse embryo fibroblasts, and its activity was fully recovered by re-expressing these genes (3). In fact, the defect in insulin-stimulated mTORC2 activity upon loss of the TSC1-TSC2 complex was further confirmed with short hairpin RNA–mediated knockdown of TSC2 in HeLa cells, the same cell line used in the Dalle Pezze et al. study.

Studies, such as that by Dalle Pezze et al., combining dynamic measurements of signaling events with in silico modeling can be powerful and should enhance our understanding of complex signaling networks. However, the strength of computational models produced from experimental data is intimately linked to the specificity and accuracy of that data. Ultimately, this is a thorough study on the kinetics of mTOR phosphorylation on Ser2481, an autophosphor­ylation site of unknown function. Questions remain regarding the relevance of these kinetic changes to the specific regulatory mechanisms affecting the activation status of mTORC2.


View Abstract

Navigate This Article