Temporal perturbation of ERK dynamics reveals network architecture of FGF2/MAPK signaling

• DOI: 10.15252/msb.20198947
• Publication Date: 2019-11-20T07:37:50Z
• Authors: Yannick Blum; Jan Mikelson; Maciej Dobrzyński; Hyunryul Ryu; Marc‐Antoine Jacques; Noo Li Jeon; Mustafa Khammash; Olivier Pertz
• Keywords: 0301 basic medicine; 570; Medicine (General); QH301-705.5; MAP Kinase Signaling System; microfluidics; 610; PC12 Cells; 03 medical and health sciences; R5-920; Methods; Animals; Biology (General); Extracellular Signal-Regulated MAP Kinases; 0303 health sciences; cell fate determination; ERK signaling dynamics; mechanistic modeling; microfluidics; parameter estimation; Dose-Response Relationship, Drug; Bayes Theorem; Microfluidic Analytical Techniques; Receptors, Fibroblast Growth Factor; mechanistic modeling; Rats; ERK signaling dynamics; 570 Life sciences; biology; Fibroblast Growth Factor 2; parameter estimation; cell fate determination; Heparan Sulfate Proteoglycans

Article19 November 2019Open Access Source DataTransparent process Temporal perturbation of ERK dynamics reveals network architecture FGF2/MAPK signaling Yannick Blum Institute Cell Biology, University Bern, Switzerland Search for more papers by this author Jan Mikelson Department Biosystems Science and Engineering, ETH Zurich, Basel, Maciej Dobrzyński orcid.org/0000-0002-0208-7758 Hyunryul Ryu Advanced Machinery Design, Seoul National University, Seoul, Korea Marc-Antoine Jacques Noo Li Jeon Mustafa Khammash Olivier Pertz Corresponding Author [email protected] orcid.org/0000-0001-8579-4919 Information Blum1,‡, Mikelson2,‡, Dobrzyński1,‡, Ryu3,4, Jacques1, Jeon3, Khammash2 *,1 1Institute 2Department 3Institute 4Present address: Research Laboratory Electronics, Massachusetts Technology, Cambridge, MA, USA ‡These authors contributed equally to work *Corresponding author. Tel: [phone]; E-mail: Molecular Systems Biology (2019)15:e8947https://doi.org/10.15252/msb.20198947 PDFDownload PDF article text main figures. Peer ReviewDownload a summary the editorial decision including letters, reviewer comments responses feedback. ToolsAdd favoritesDownload CitationsTrack CitationsPermissions ShareFacebookTwitterLinked InMendeleyWechatReddit Figures & Info Abstract Stimulation PC-12 cells with epidermal (EGF) versus nerve (NGF) growth factors (GFs) biases distribution between transient sustained single-cell activity states, proliferation differentiation fates within cell population. We report that fibroblast GF (FGF2) evokes distinct behavior consists gradually changing population transient/sustained states in response increasing inputs dose response. Temporally controlled perturbations MAPK applied using microfluidics reveal wider mix emerges through combination an intracellular feedback, competition FGF2 binding FGF receptors (FGFRs) heparan sulfate proteoglycan (HSPG) co-receptors. show latter experimental modality is instructive model selection Bayesian parameter inference. Our results provide novel insights into how different receptor tyrosine kinase (RTK) systems differentially wire fine-tune fate decisions at level. Synopsis Analyses MAPK/ERK temporally EGF, NGF, stimulations compared EGF/NGF. A mathematical accounts these presented. leads distributions dynamic Increasing shifts states. new structure. FGFR/HSPG interactions enable Introduction Signaling dynamics, rather than steady have been shown control (Levine et al, 2013). For multiple (RTK), heterogeneity can explain variability observed (Cohen-Saidon 2009; Chen 2012). Both biological noise extrinsic individual intrinsic networks shape fate. It has proposed nature signal transduction enables accurate information transmission presence (Wollman, 2018). Measuring therefore key understanding cellular correlate specific decisions. The extracellular signal-regulated (ERK) regulator such as differentiation. functions mitogen-activated protein (MAPK) pathway which factor (GF) activate membrane-resident Ras GTPase subsequently triggers cascade leading activation (Avraham Yarden, 2011). Rat adrenal pheochromocytoma widely used system study regulation (Marshall, 1995). EGF or NGF population-averaged specifically trigger Thus, duration determinant 1995; Santos 2007). These result from GF-dependent (Santos 2007), negative positive feedback producing all-or-none adaptive bistable outputs, respectively (Xiong Ferrell, 2003; 2007; Avraham More recently, assays indicated EGF/NGF induces heterogeneous across (Ryu 2015). While responses, isogenic due expression components receptor-dependent modulation loops. This might induce differentiating proliferating (Chen Further support stems model-based prediction stimulation schemes synthetic patterns determine independently identity An additional GF, FGF2, also activates regulates processes angiogenesis, wound healing, development (Ornitz Itoh, Upon stimulation, FGFR dimerizes, autophosphorylates, recruits adaptors, Ras/RAF/MEK/ERK In cells, activity, correlates (Qui Green, 1992). FGF–FGFR are further regulated co-receptor (Ornitz, 2000; Matsuo Kimura-Yoshida, initially binds HSPGs high-affinity interaction, followed 2nd lower affinity interaction HSPG/FGF2/FGFR trimeric complex. dimerizes dimer trimer complex autophosphorylate downstream marked contrast exhibit sigmoidal elicits biphasic where intermediate concentration higher outputs low high concentrations. example, bell-shaped neuronal (Williams 1994), (Zhu 2010) dose–response challenges systems. 2010; Kanodia 2014). ability emerge (Kanodia However, FGF2-dependent significantly less defined NGF. important question field RTKs specify network. Here, we explore controls level cells. find those increase input modulates Using perturb network, logic behind data together modeling underlying FGF2/FGFR/HSPG layer coupled simple conclude differently decode their cognate engaging structures. suggest evolved translate gradual changes be regulate during interpretation morphogen gradient. Results induced EGF/NGF-triggered studied studies revealed much complexity previously anticipated asked if potentially To level, line stably expressing EKAR2G, fluorescence resonance energy transfer (FRET)-based biosensor endogenous, cytosolic activity. EKAR2G extensively validated elsewhere (Harvey 2008; Fritz extract temporal patterns, CellProfiler-based (Kamentsky 2011) image analysis pipeline segmentation tracking single computation per-cell average FRET ratio. computer-programmable microfluidic device pulses (Fig 1A). Figure 1. A. Flow-based, delivery. Computer-controlled, pressure pump mixing medium GFs deliver cultured device. right panel illustrates typical patterns. B. Representative ratio images treated 25 ng/ml FGF2. Ratio color-coded so warm/cold colors represent high/low levels. Scale bar = 50 μm. C, D. Population averages (C), challenge 0.25, 2.5, 25, 250 (D). Single-cell time series were normalized own means before t [0, 40]. Red curve bottom C indicates profile measured simultaneously Alexa 546-labeled dextran. N [48, 120] per concentration. 2′ intervals. E. Hierarchical clustering pooled (N 983) focus on relevant trimmed x-axis [36, 100] min. Each row heatmap corresponds cell. warping Ward's linkage method building dendrogram, was then cut distinguish 6 clusters left. F. Average identified (E), (E). G. Distribution trajectories panels (E F) dosages. H. Separability populations calculated area under Jeffries–Matusita distance along (Materials Methods, Appendix Fig S2B). dendrogram created complete-linkage method. Data information: (C, D, F), gray band 95% CI mean, representative 3 replicates. (D black horizontal stimulation. available online figure. 1 [msb198947-sup-0006-SDataFig1.zip] Download figure PowerPoint First, stimulated EGF/NGF/FGF2 1B C). fluorescent dextran quality delivery 1C, red trace). As expected, when evaluated led while peak amplitude peak. contrast, sharper one evoked EGF. After fast adaptation, increases over time. Since determination variety 2014), tested our system. 0.25–250 range, previous works 2014; 1D). On average, all concentrations triggered initial identical but faster adaptation 0.25 only moderate without 2.5 after small almost indistinguishable profiles both Indeed, robust peak, whereas clearly transient. At transient, again slow recovery. described evident consider point Additionally, observe 1st highly similar identity/concentration EV1A). Click here expand EV1. Raw 10 min (T 30′ 50′). Lower upper hinge correspond 25th 75th percentiles. whiskers 1.5*IQR hinge. Replicates: 40] mean; 1F. hierarchical clusters. (B) same approach longer interval (t 200] min). left column cluster 4 clusters, associated Heterogeneous overlaid EV1B). examine heterogeneity, (EGF, FGF2—4 concentrations) ranging shortly 60′ 1E). (DTW) (Giorgino, 2009) classes DTW calculates similarity two matching features may shifted series. Visual inspection obtained procedure us identify major patterns/clusters, highlighted vertical color bars 1E. 1F summarizes each cluster, EV1C displays cluster. Even though differ amplitude, recognize 1, 3, 4, 2, 5, 6. computed conditions 1G). Low-amplitude activities (clusters 2) largely absent stimulations, indicating signaling. With replaced 5 lowest low-amplitude responses. High high-amplitude decreasing contribution low- decreased favor (high adaptation). Then, increased dosage strongly amplitudes. Intrigued fact recovery measurements 1D), case purpose, repeated individually experiments addition 160′ EV1D). displayed levels 1–3), Importantly, three present adaptation. phenomenon not assess response, PCA decomposition accumulated pairwise distances points (Figs 1H EV2). approaches showed pattern dosages separated (Appendix Text). EV2. Principal component (PCA) dataset first account 85% overall variability. use [36,100] Note untreated shown, although they (red points). Schematic calculation Two noisy consist 100 each, trajectory spanning 5] T 0.1 interval. Step 1: every point, calculate (dJM) quantity. 2: Calculate (AUC) dJM express it fraction maximum AUC dJM, 2dxN, length number points. construct 1H. pair (0.25, ng/ml) [36,200] EV1B. Decoding properties sought structure explains FGFR/MAPK dynamically perturbed delivering lengths captures salient accessible and, many cases, population-homogeneous simpler interpret 3′, 10′, four previously. plotted pulse EV3). 2. single-pulse A–C. (A), (B), (C) Solid lines—population [39, 166], replicates: EGF: NGF: FGF: replicates; bands—95% bars—duration 2 [msb198947-sup-0007-SDataFig2.zip] EV3. Clustering regimes (C). highlight branches clustered independently; thus, share unrelated. pulsed consistent observations Population-averaged exhibited full-amplitude 3′ 10′ pulse, except 2A) pronounced. kinetics dose. full occurred concomitantly washout. doses pulsing EV3A). did yield any scheme 3B). Above concentration, (achieved and/or duration) profile. contributing EV3B). 3. multi-pulse 3′–20′ Cluster performed Manhattan method; (B, D) (F) visualization. [52, 91], [msb198947-sup-0008-SDataFig3.zip] Intriguingly, varying 2C). threshold input, pattern, whereby rebound decayed slowly. visible pulse. shorter activities. well characteristic EV3C). enriched inputs. even emerged 2.5–250 concentrations, ensued washed away. stronger decay confirms rebound, probe architectural timescales test responds subjected 20′ pauses 3A). Multiple timescale phase puls