Please visit the new BioModels platform to access the latest content. This website is no longer updated and will be retired on 31 May 2019.
BioModels Database logo

BioModels Database


Dalle Pezze et al., (2016). A systems study reveals concurrent activation of AMPK and mTOR by amino acids.

June 2017, model of the month by Thawfeek Mohamed Varusai
Original model: BIOMD0000000640


The mammalian target of Rapamycin (mTOR) is a serine/threonine kinase that plays a prime role in the regulation of metabolism and cellular growth. mTOR forms a part of two multi-protein complexes, namely mTOR complex 1 (mTORC1) and mTOR complex 2 (mTORC2). Multitudes of upstream regulators control the activity of mTORC1, which in turn regulates several effectors involved in a range of cellular processes including autophagy and protein synthesis. Autophagy is also regulated by AMP-activated protein kinase (AMPK). In this paper [1, BIOMD0000000640], the authors study the regulation of autophagy upon nutrient (amino acids) stimulation using computational approaches and experimentally validating their predictions.


Known to be activated by nutrient addition and insulin stimulation, mTORC1 can then inhibit autophagy by phosphorylating Unc-51-like kinase 1 (ULK1) at serine 757. On the contrary, AMPK is activated by nutrient and energy shortage can trigger autophagy by phosphorylating ULK1 at serine 317. Thus, mTORC1 and AMPK are perceived to be antagonists that function towards inhibiting and stimulating autophagy under the presence and absence of nutrients (2). However, autophagy is also required during nutrient availability to provide amino acids and metabolite intermediates for biosynthetic processes (3). It is assumed that AMPK is in an activated state under nutrient stimulation, which can facilitate autophagy. Nevertheless, this is still an unexplained hypothesis lacking experimental validation.

Mathematical Modelling

The authors deterministically model amino acid signaling in the mTOR/AMPK network using ordinary differential equations (Figure 1). The model contains 31 species and 48 reactions in total. Mass action kinetics was used to describe the phosphorylation and dephosphorylation events in the system. Extensive time course and dose response data was obtained using gene knock down and inhibitor experiments. The Matlab toolbox PottersWheel was used for designing and estimating the rate constants of the model. Additional amino acid input points were added in a step-wise manner to determine the model variant with the best empirical fit. Furthermore, module-based parameter estimation was also performed to overcome non-identifiability and computation-time issues. Copasi 4.8.35 was used to simulate inhibition experiments in the network. Model predictions were experimentally validated thus confirming the findings.


The findings of the mathematical model and supporting experiments are as follows:

  1. Amino acid, which is believed to trigger mTORC1 activity, may have more than one input points in the mTOR signaling network via IRS/PI3K, AMPK and mTORC2.
  2. Amino acids activate AMPK independently of PI3K, Akt and mTORC2 and via the Ca2+/calmodulin-dependent protein kinase kinase β (CaMKKβ).
  3. Amino acid-activated AMPK sustain ULK1 activity and autophagy do not inhibit mTORC1. The most likely mechanism is the inhibition of the c-Jun by CaMKKβ-AMPK. c-Jun is known to be an autophagy suppressor.

Predictive Power of the Model

The authors have developed a dynamic model of the mTOR signaling network and perform experiments to evaluate model parameters and validate the predictions. To study the different amino acid input points in the system, the authors have investigated 70 different model variants for best fit with the empirical data. Appropriate initial values were set to unphosphorylated proteins to avoid protein phosphorylation. Initial amount of phosphorylated proteins was set to zero. Together this avoids basal signaling effects of the system under starvation conditions. The model with 96 parameters and 31 initial values was fit to 12 observables. Few of the observables are combinations of several model species. To overcome complications in parameter estimation and identification, the authors used constrains on the combined species. This reduced the number of estimated parameters and allowed better fitting. All parameters were shown to be identifiable. Taken together, this suggests that the mTOR model in the study may closely resemble the biological system and has a high predictive power.

Figure 1

Figure 1 mTOR network model with four amino acid inputs and an extended p70-S6K module fits the experimental data. This figure shows the Graphical model of the mTOR network activated by insulin and four amino acid inputs. Figure taken from (1).

Significance of the Model

The significance of a biological model may be determined by the contribution of the model to the obtained knowledge. Mathematical modelling plays an indispensable role in this study. The impact of an mTOR dynamic model here is as follows.

  1. The model was unable to explain the empirical data upon amino acids stimulation and this initiated the investigation of amino acids signaling.
  2. 70 different model variants were tested to predict the amino acid input points. These predictions were then experimentally confirmed.
  3. The route of AMPK activation by amino acids was computationally determined using an elimination strategy.

Had it not been for computational approaches, it may not be possible or may be experimentally laborious to make the above findings. Thus, this work demonstrates a strong case for the application of computational modelling to understand cell signaling dynamics.

Scientific Value Added

This paper sheds light on the amino acid signal transduction in the mTOR signaling network. One of the major contribution of this paper is the idea that amino acids can influence the mTOR network in an mTORC1-independent fashion. The popular notion is that amino acids can only trigger mTORC1 activity directly via the RAG GTPases and other mechanisms. Here the authors show that amino acids can also activate AMPK in an mTORC1-independent mechanism. They also provide evidence of possible mechanisms of amino acid-AMPK activation. Physiologically, this study tries to explain the molecular nature of autophagy under nutrient starvation, which may be required to maintain protein homeostasis and deliver metabolite intermediates for biosynthetic processes.


  1. Dalle Pezze P. et al. A systems study reveals concurrent activation of AMPK and mTOR by amino acids.. Nat Commun. 2016 Nov 21;7:13254.
  2. Hindupur SK. et al. The opposing actions of target of rapamycin and AMP-activated protein kinase in cell growth control.. Cold Spring Harb Perspect Biol. 2015 Aug 3;7(8):a019141.
  3. Kaur J. et al. Autophagy at the crossroads of catabolism and anabolism.. Nat Rev Mol Cell Biol. 2015 Aug;16(8):461-72.