Diedrichs DR et al. (2018). A data-entrained computational model for testing the regulatory logic of the vertebrate unfolded protein response

June 2019, model of the month by Thawfeek M Varusai
Original model: BIOMD0000000703


     The vertebrate unfolded protein response (UPR) is essential for cells to adapt to endoplasmic reticulum (ER) stress but also commits the cell to apoptosis if this stress is too severe. A complex signaling network regulates underlying this dual conflicting role of the UPR process. Three pathways are involved in UPR signaling: serine/threonine-protein kinase/endoribonuclease (IRE1), Eukaryotic translation initiation factor 2-alpha kinase 3 (PERK) and Cyclic AMP-dependent transcription factor ATF-6 alpha (ATF6). These pathways are interconnected with multiple overlapping feedback and feedforward regulations among them (Figure 1A). IRE1 regulates degradation of ER-associated mRNAs and PERK mediates translational arrest [1,2]. All three pathways also participate in transcriptional control regulating protein folding mechanisms [3]. Together this intricate interplay of regulations decide the cell response to ER stress. Diedrichs DR et al. (2018) computationally investigate the dynamic nature of the UPR system (BIOMD0000000703) and explain the different cellular responses [4].


     UPR signaling needs to balance between the ability of cells to adapt to ER stress and to trigger apoptosis if the stress is severe. Surprisingly, the same pathways in UPR signaling control these two extreme scenarios. Furthermore, specific sub-pathways are not decisive for adaptation or apoptosis. Regulatory proteins in UPR including DNA damage-inducible transcript 3 protein (CHOP) and Endoplasmic reticulum chaperone BiP (BiP) have unclear roles in cell response. Transient levels rather than steady state levels of protein expression seem to be more predictive of cell fate.


Figure 1

Figure 1. Modeled relationships among UPR components. (A) Schematic (B) Wiring diagram. Figures adapted from [4].

Mathematical modelling

     Diedrichs DR et al. (2018) created an ODE model of the UPR network including activation signals of CHOP and BiP and downstream outcomes of the IRE1, PERK and ATF6 (Figure 1B). This system consists of 14 mass-action modelled equations and 61 parameters. Parameter estimation was based on time course data from quantitative RT-PCR and immunoblot experiments in wild type MEF cell lines under different stress doses. COPASI’s LSODAR solver was used to simulate the model and the numerical optimization algorithms were used to fit parameters to experimental data.


     Diedrichs DR et al. (2018) provide an empirically entrained ODE model of the vertebrate UPR system. This model was used to test the regulatory logic of the network by exploring the functional consequences of perturbing various pathway components. In silico experiments conclude that eukaryotic translation initiation factor 2 subunit 1 (eIF2α) initiates an auto-regulatory negative feedback loop. Another model prediction is that crosstalk between PERK and ATF6 pathways facilitate a sensitive CHOP response to ER stress.

Predictive power of the model

     The UPR model is comprehensive in that it takes into account activation and the output of the signaling network. Model parameters were estimated using time course experiments in wild type MEF cell lines under a low and a high stress dose. Furthermore, in silico perturbations in the model were empirically validated using knockout studies. Taken together, these factors impart a high predictive power to this UPR model.

Significance of the model

     To understand the significance of mathematical modelling in this study, we can try to imagine what would be the case without it. To arrive at the same conclusions the authors would require homologous-directed genome editing at precise positions in genes. Furthermore, target clones would need identification, selection and propagation. Alternate to this would be mutating stem cells and subsequently differentiate them into fibroblasts. All of these techniques are laborious and expensive. In contrast, building reliable models to computationally predict the outcome is effective and powerful. Thus, the use of dynamic modelling in this work is quite resourceful.

Scientific Value Added

     The ODE model in this study has reliably predicted the role of the different components and interactions in the UPR signaling system. Two prominent outcome of the work are the identification of an eIF2α auto-regulatory negative feedback loop and the finding that the crosstalk between PERK and ATF6 pathways regulate CHOP sensitivity to stress. Furthermore, there is now a template model available for the UPR network for other researchers to work with. This model can be parameterized with empirical data from other cell types or conditions. Additional components can be added to the model to expand its scope.


  1. Hollien J, Weissman JS (2006), Decay of endoplasmic reticulum-localized mRNAs during the unfolded protein response, Science 313, 104–107.
  2. Harding HP, Zhang Y, Ron D (1999), Protein translation and folding are coupled by an endoplasmic-reticulum-resident kinase, Nature 397, 271–274.
  3. Walter P, Ron D (2011), The unfolded protein response: from stress pathway to homeostatic regulation, Mol Biosyst. 2012;8:1806–1814.
  4. Diedrichs DR, Gomez JA, Huang CS, Rutkowski DT, Curtu R (2018), A data-entrained computational model for testing the regulatory logic of the vertebrate unfolded protein response, Mol Biol Cell. 29, 1502-1517.