Theoretical and experimental analysis links isoform‐ specific ERK signalling to cell fate decisions

Abstract

Cell fate decisions are regulated by the coordinated activation of signalling pathways such as the extracellular signal‐regulated kinase (ERK) cascade, but contributions of individual kinase isoforms are mostly unknown. By combining quantitative data from erythropoietin‐induced pathway activation in primary erythroid progenitor (colony‐forming unit erythroid stage, CFU‐E) cells with mathematical modelling, we predicted and experimentally confirmed a distributive ERK phosphorylation mechanism in CFU‐E cells. Model analysis showed bow‐tie‐shaped signal processing and inherently transient signalling for cytokine‐induced ERK signalling. Sensitivity analysis predicted that, through a feedback‐mediated process, increasing one ERK isoform reduces activation of the other isoform, which was verified by protein over‐expression. We calculated ERK activation for biochemically not addressable but physiologically relevant ligand concentrations showing that double‐phosphorylated ERK1 attenuates proliferation beyond a certain activation level, whereas activated ERK2 enhances proliferation with saturation kinetics. Thus, we provide a quantitative link between earlier unobservable signalling dynamics and cell fate decisions. ### Synopsis Cell fate decisions such as proliferation, differentiation, and survival are controlled by the activation of signalling networks. A paradigm of a complex signalling pathway is the Raf/MEK/ERK cascade that integrates information from receptors on the cell membrane to elicit cellular responses. Although the components and the wiring of this network are known, there is still a lack of information on its kinetic behaviour and how activation of signalling proteins relates to cellular responses. In particular, two highly related isoforms, ERK1 and ERK2, are expressed in mammalian cells. Although some redundancy seems to exist between these isoforms, ERK1 knockout mice are viable but ERK2 knockout mice are embryonic lethal. Thus, the contribution of these isoforms on signalling and cell fate decision needs to be analysed at a quantitative level. Earlier mathematical models addressing the ERK cascade focused on receptor tyrosine kinases (RTKs)‐induced ERK signalling. RTKs elicit strong and nearly complete phosphorylation of ERK. In contrast, cytokine receptors that have a major role in cell proliferation and the prevention of apoptosis in the haematopoietic system induce only weak phosphorylation of ERK suggesting major difference in systems properties. Activation of ERK involves phosphorylation on two residues, which could be achieved by (i) a processive mechanism, with the upstream kinase binding to ERK and phosphorylating both residues before dissociation or (ii) a distributive mechanism, with the upstream kinase binding to ERK, phosphorylating one residue, dissociating, binding a second time, phosphorylating the second residue and then dissociating. There is in vitro evidence that ERK phosphorylation may be distributive, however, scaffolding proteins might favour a processive mechanism in vivo . Although processive ERK activation would lead to fast signalling because of only one binding step, a distributive ERK activation would lead to signal amplification because of a larger sensitivity of the reaction to the upstream kinase. To address these challenging questions, we established a mathematical model of ERK signalling through a cytokine receptor, the erythropoietin receptor (EpoR). To calibrate our model, we generated time‐resolved data by quantitative immunoblotting in primary erythroid progenitor (colony‐forming unit erythroid stage, CFU‐E) cells after stimulation with erythropoietin (Epo). Distinct from cell lines that are commonly used for biochemical studies, intracellular signalling networks are mostly unperturbed in primary cells. To represent events at the plasma membrane, we tested several receptor models and selected a model that can describe both the temporal and the dose–response kinetics of phosphorylated JAK2. To resolve the underlying mechanism for ERK activation, we established two models: (i) model with a processive activation mechanism and (ii) model with a distributive mechanism. Only the distributive model could explain our time‐resolved data sufficiently ([Figure 1][1]). Both mechanisms can be distinguished by the extent of mono‐ and double‐phosphorylated ERK being generated. Therefore, we predicted the concentrations of non‐, mono‐, and double‐phosphorylated ERK after Epo stimulation in silico and quantified the respective concentrations in Epo‐stimulated CFU‐E cells by label‐free mass spectrometric analyses. The experimental results were in line with the predictions of our distributive model, thus showing that in vivo ERK phosphorylation occurs by a distributive mechanism. Furthermore, the mathematical model allowed us to analyse information processing through this signalling network by calculating the concentration of activated molecules at each step of the pathway. The results indicate a substantial signal attenuation between the receptor and the membrane‐associated factors SOS, Ras, and Raf, as expected for cytokine receptors. However, strong signalling amplification is observed between Raf, MEK, and ERK. Thus, signal processing of the EpoR‐induced ERK cascade represents a phosphorylation bow tie, resembling regulatory bow‐tie structures that are common motifs in biology. The main source of fragility is the knot in the bow tie, in this case SOS, Ras, and Raf. These results were confirmed by sensitivity analyses of the mathematical model: changes in the initial protein concentrations of these proteins have the largest impact on ERK phosphorylation. The sensitivity analysis provided another unexpected result regarding the interlinked wiring of the signalling network: increasing the...