1 Supporting Information 2 3 A transcriptomics based kinetic model for ethylene biosynthesis in tomato fruit: 4 development, validation and exploration of novel regulatory mechanisms 5 Bram Van de Poel, Inge Bulens, Maarten L.A.T.M. Hertog, Bart M. Nicolai, Annemie H. Geeraerd 6 7 8 Overview Supporting Figures 9 Supporting Figure S1. Model input data. ................................................................................................ 2 10 Supporting Figure S2. Illustration of the average calibration method..................................................... 3 11 12 Supporting Figure S3. Comparison qPCR results analyzed by the normal calibration method and the average calibration method. .................................................................................................................... 4 13 Supporting Figure S4. Input data adapted from literature. ...................................................................... 5 14 Supporting Figure S5. Distribution of ACC production by the three individual ACS isoforms. ............ 6 15 Supporting Figure S6. Monte Carlo evaluation of the model.................................................................. 7 16 17 Supporting Figure S7. The behaviour of the fTRAN and fDEG function during the entire fruit development period. ...................................................................................................................................................... 8 18 Supporting Tables 19 Supporting Table S1. Estimated parameter values and their units .......................................................... 9 20 Supporting Table S2. Estimated parameter values of fTRAN and fDEG. .................................................... 10 21 22 23 24 25 26 Supporting Methods Supporting Methods S1. Explanation of the average calibration method for qPCR analysis. .............. 11 Supporting Notes Supporting Notes S1. Overview of the model in Matlab code. ............................................................. 13 Supporting References Supporting References S1 ..................................................................................................................... 15 27 1 28 Supporting Figures 29 30 Supporting Figure S1. Model input data. Overview of the input data used by the model for (a) 31 SAM content (nmol mg protein-1), (b) ACO1 expression (relative copies), (c) ACS2 32 expression (relative copies), (d) ACS4 expression (relative copies), (e) ACS6 expression 33 (relative copies) and (f) MTN expression (relative copies) during tomato fruit development, 34 ripening and postharvest storage. A linear interpolation method is used between the data 35 points in order to get intermediate input values. 36 2 37 38 Supporting Figure S2. Theoretical illustration of the average calibration method. Two 39 different reference genes are analyzed (Ref gene A (green) and Ref gene B (red)) and for both 40 genes a dilution series was made ranging from 1 to 0.01 transcript copies (= gene specific 41 calibration curve). Both the qPCR profiles (a) and the calculated calibration curves (b) are 42 shown for both genes and their average expression (blue). For both genes an unknown sample 43 was also included (UA and UB). Based upon the gene specific calibration curve, a Cq value of 44 the unknown samples were calculated. Supporting Figure S3A shows how the average qPCR 45 profile is constructed from gene A and B for each dilution point. Supporting Figure S3b 46 shows the average calibration curve, calculated from the two independent calibration curves 47 of gene A and B. The average calibration curve allows to calculate an average expression 48 (Cq,Avg) level for the unknown samples UA and UB. This method shows that sample UA is ± 49 2.106 times more expressed than sample UB, relative to their average expression. Gene A is 50 thus identified as the most important gene in this theoretical example. 51 52 3 53 54 Supporting Figure S3. Comparison qPCR results analyzed by the normal calibration method 55 and the average calibration method. Relative gene expression profile of all ACO (a, b) and 56 ACS (c, d) isoforms with the gene specific calibration method (a, c) and the average 57 calibration method (b, d). Maturity stage annotations: S. small fruit, M. medium fruit, IMG. 58 immature green fruit, MG. mature green fruit, BR. breaker fruit, LO. light orange fruit, O. 59 orange fruit, P. pink fruit, R. red fruit, RR. red ripe fruit, RR + X. red ripe fruit + X days post 60 harvest (X = 3, 5, 7, 10 and 12 days). 61 4 62 63 Supporting Figure S4. Input data adapted from literature. Profiles of MACCT activity (nmol 64 h-1 mg protein-1) and SAMdc activity (nmol h-1 mg protein-1) used as model input and adopted 65 from (Martin et al., 1995; Morilla et al., 1996). 66 67 5 68 69 Supporting Figure S2. Distribution of ACC production by the three individual ACS isoforms 70 ACS2 (black), ACS4 (red) and ACS6 (blue). 71 6 72 73 Supporting Figure S3. Monte Carlo simulation of the model. Heat plots representing the 74 distribution results from a Monte Carlo simulation for (a) ethylene production (nmol h-1 mg 75 protein-1), (b) ACC levels (nmol mg protein-1), (c) MACC levels (nmol mg protein-1), (d) 76 MTA levels (nmol mg protein-1), (e) ACO activity (nmol h-1 mg protein-1), (f) ACS activity 77 (nmol h-1 mg protein-1) and (g) MTN activity (nmol h-1 mg protein-1) during tomato fruit 78 development, ripening and postharvest storage. 79 80 7 81 82 Supporting Figure S4. The behaviour of the fTRAN (a) and fDEG (b) function during the entire 83 fruit development period. 84 85 8 86 Supporting Tables 87 88 Supporting Table S1. Estimated parameter values and their units; including initial values of model independent variables. Abbreviation Estimated value Unit ± Std. deviation 0.22 ± 0.07 nmol h-1 mg protein-1 d-1 relative copies-1 kt,ACS 0.028 ± 0.003 nmol h-1 mg protein-1 d-1 relative copies-1 kt,MTN 34.9 ± 8.2 nmol h-1 mg protein-1 d-1 relative copies-1 1.39 ± 0.4 d-1 kpd,ACS2 0.21 ± 0.02 d-1 kpd,ACS4 4.25 ± 0.66 d-1 kpd,ACS6 0.22 ± 0.04 d-1 kpd,MTN 0.16 ± 0.04 d-1 0.0126 ± 0.0009 mg protein h nmol-1 d-1 kACS2 2.2 ± 0.3 mg protein h nmol-1 d-1 kACS4 0.63 ± 0.58 mg protein h nmol-1 d-1 kACS6 0.016 ± 0.005 mg protein h nmol-1 d-1 kMACC 0.073 ± 0.004 mg protein h nmol-1 d-1 kSAMDC 0.22 ± 0.05 mg protein h nmol-1 d-1 kMTN 0.24 ± 0.04 mg protein h nmol-1 d-1 kDACC 0.008 ± 0.011§ d-1 0.04 ± 0.01 nmol h-1 mg protein-1 [MTNini] 75.7 ± 6.7 nmol h-1 mg protein-1 [MTAini] 1.2 ± 0.2 nmol mg protein-1 0.35 ± 0.03 nmol mg protein-1 4.4 ± 1.3 h-1 Protein synthesis kt,ACO Protein degradation kpd,ACO1 Metabolite synthesis kACO Initial values [ACS6ini] Basal MTA level [MTAb] Ethylene diffusivity kdiff 89 § 90 omitted from the model. Additional biochemical information about DACC formation could improve the estimation of this parameter. 91 Standard deviation was larger than the estimated parameter value. This indicates that the estimated value of kDACC is uncertain and could be All other initial values were fixed (see Supporting Notes S1 for values). 92 9 93 Supporting Table S2. Estimated parameter values of fTRAN and fDEG. Abbreviation Estimated value ± Std. deviation Unit 0.29 ± 0.03 / 0.037 ± 0.006 / 0.09 ± 0.02 / fDEGini 3.7 ± 0.8 / fDEGmin 0.93 ± 0.02 / fDEGmax 4±1 / fTRAN kTRAN fTRANini fDEG kDEG 94 95 10 96 Supporting Methods 97 Supporting Methods S1. Explanation of the average calibration method for qPCR analysis. 98 The implementation of the average calibration method allows cross-comparison of gene 99 expression levels of different genes. The theoretical explanation of the average calibration 100 method is supported by Supporting Figure S3 that shows real-time qPCR profiles for two 101 theoretical genes (genes A and B) and their calibration curves. If one analyzes the expression 102 level of sample A with a calibration curve designed in the range of gene A only, it would have 103 a Cq,A value of 0.025. This value is much lower than the Cq,B value (0.38) of unknown sample 104 B calculated with a gene B-specific calibration curve. For the average calibration method the 105 average Cq values for all dilutions points of the gene specific calibration curve are calculated. 106 One can now reanalyze the expression level of unknown samples A and B with the new 107 average calibration curve. The new expression levels of unknown sample A and B have a 108 Cq,Avg value of 14.2 and 0.000065 respectively. This method shows that gene A is much more 109 expressed than gene B, relative to the average expression level of both genes, identifying gene 110 A as more important than gene B. The method can now be extended and improved by 111 calculating the average calibration curve for many stable reference genes. This allows a 112 rescaling of unknown genes according to the average expression level of a group of reference 113 genes. 114 In our study, all performed qPCR runs included a gene specific calibration curve for relative 115 quantification. This calibration curve consisted out of a dilution series of a mixed sample, 116 allowing quantification of unknown samples in the same expression range of the mixed 117 sample. Unfortunately, it is impossible to compare transcription levels of different genes, 118 since the calibration curve is gene specific (assessed with the same primers). To overcome 119 this, the new average calibration curve was calculated based upon the average expression of 11 120 five different reference genes (GAPDH, ACT, ELF1α, RPL2 and PP2Ac). This new 121 calibration curve represents the average expression level of the five reference genes. The 122 expression level of all unknown samples was reanalyzed relatively to the average expression 123 of the five reference genes and, as such, allows a cross-comparison of relative expression 124 levels between different genes. 125 12 126 Supporting Notes 127 Supporting Notes S1. Overview of the model in Matlab code. 128 The full model is written in Matlab code and is shown below. It can be run by the OptiPa 129 toolbox in a Matlab environment (Hertog et al., 2007). OptiPa is a graphical user interface 130 freeware tool designed for solving ordinary differential equations and parameter optimization 131 for kinetic modeling. 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 %-------------------------------------------------------------------------% Ethylene Biosynthesis model %-------------------------------------------------------------------------switch status case 'step1' %------------------------------------------------------------------------------------% Initialize variables and parameters %------------------------------------------------------------------------------------output_var = { 'ACO1' 'ACS2' 'ACS4' 'ACS6' 'MTN' 'ACC' 'MACC' 'DACC' 'MTA' 'ETHint' 'ETHpro' 'ACS'} ; %------------------------------------------------------------------------------------% Definition of all model parameters. One parameter per row. % For consistency, these parameters values are set equal to the optimized values shown in SI Table SI. In practice, initial estimates are to be provided here. %------------------------------------------------------------------------------------param_DEF = { 'ktaco' 0.22;... 'ktacs' 0.028;... 'ktmtn' 34.9;... 'kpdaco1' 'kpdacs2' 'kpdacs4' 'kpdacs6' 'kpdmtn' 1.39;... 0.21;... 4.25;... 0.22;... 0.16;... 'kaco1' 'kacs2' 'kacs4' 'kacs6' 'kmacc' 'kdacc' 'ksamdc' 'kmtn' 0.0126;... 2.2;... 0.63;... 0.016;... 0.073;... 0.008;... 0.22;... 0.24;... 'ACS60' 'MTN0' 0.04;... 75.7;... 13 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 'MTA0' 1.2;... 'MTAb' 0.35;... 'kdiff' 4.4}; %------------------------------------------------------------------------------------case 'step2' %------------------------------------------------------------------------------------% Initialize ODEs %-------------------------------------------------------------y0 = [ 1.5 0 0 ACS60 MTN0 0 1 1 MTA0 0.008] ; %------------------------------------------------------------------------------------case 'step3' %------------------------------------------------------------------------------------% Transform experimental data %------------------------------------------------------------------------------------case 'step4' %------------------------------------------------------------------------------------% ODE model definition %------------------------------------------------------------% defining local model constants %------------------------------------------------------------------------------------% SAM content SAM=interp1(t_cond,SAM,t); SAM0=SAM(1); % Gene expression data ACO1e=interp1(t_cond,ACO1e,t); ACS2e=interp1(t_cond,ACS2e,t); ACS4e=interp1(t_cond,ACS4e,t); ACS6e=interp1(t_cond,ACS6e,t); MTNe=interp1(t_cond,MTNe,t); % SAM decarboxylase activity SAMdc=interp1(t_cond,SAMdc,t); SAMdc0=SAMdc(1); % MACC transferase activity MACCT=interp1(t_cond,MACCT,t); %------------------------------------------------------------------------------------% Definition of ODEs %------------------------------------------------------------------------------------% Protein synthesis dACO1dt=ktaco*ACO1e-kpdaco1*ACO1; dACS2dt=ktacs*ACS2e-kpdacs2*ACS2; 14 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 dACS4dt=ktacs*ACS4e-kpdacs4*ACS4; dACS6dt=ktacs*ACS6e-kpdacs6*ACS6; dMTNdt=ktmtn*MTNe-kpdmtn*MTN; % Metabolite production dACCdt=kacs2*SAM*ACS2+kacs4*SAM*ACS4+kacs6*SAM*ACS6kaco1*ACC*ACO1-kmacc*ACC*MACCT-kdacc*ACC; dMACCdt=kmacc*ACC*MACCT; dDACCdt=kdacc*ACC; dMTAdt=kacs2*SAM*ACS2+kacs4*SAM*ACS4+kacs6*SAM*ACS6kmtn*(MTA-MTAb)*MTN+ksamdc*SAM*SAMdc; % Ethylene production dETHintdt=kaco1*ACC*ACO1-kdiff*ETHint; %------------------------------------------------------------------------------------% Returning all ODE's together as dydt %------------------------------------------------------------------------------------dydt = [dACO1dt dACS2dt dACS4dt dACS6dt dMTNdt dACCdt dMACCdt dDACCdt dMTAdt dETHintdt]'; %------------------------------------------------------------------------------------case 'step5' %------------------------------------------------------------------------------------% Transform ODE model output %------------------------------------------------------------------------------------ETHpro=ETHint*kdiff; ACS=ACS2+ACS4+ACS6; MTA=MTA+MTAb; %------------------------------------------------------------------------------------end Supporting References 254 255 1. Martin MN, Cohen JD, Saftner RA (1995). A New 1-Aminocyclopropane-1-Carboxylic AcidConjugating Activity in Tomato Fruit. Plant Physiol 109: 917-926. 256 257 2. Morilla A, Garcia JM, Albi MA (1996). Free polyamine contents and decarboxylase activities during tomato development and ripening. J Agr Food Chem 44: 2608-2611. 258 259 260 261 3. Hertog MLAT, Verlinden BE, Lammertyn J, Nicolai BM (2007) OptiPa, an essential primer to develop models in the postharvest area. Comput Electron Agr 57: 99-106. 15