High throughput phenotyping and plant modelling : two legs for combined physiological and genetic approaches?

Transcription

1 High throughput phenotyping and plant modelling : two legs for combined physiological and genetic approaches? F. Tardieu, C. Welcker, C. Fournier, B. Muller, Th Simonneau, C. Granier J J xc Ψ evap R c Ψ cel Ψ xyl J xyl Ψ r R sp Ψ soil

2 1. Phenotyping The problem Phenotyping is the bottleneck in genomics. - Sequences of Arabidopsis, Rice, Sorghum, poplar, tomato, maize etc - Genotyping capacities for 1s of genotypes - Large collections of RILs, mutants, accessions Is this the proper way to address the problem? (defined by a need, not by a biological question) Are "genotyping" and "phenotyping" parallel activities?

3 1. Phenotyping The problem The 'problem' is that we can now measure traits in1s of plants in a robotised way Presentation C. Welcker Phenodyn (Sadok et al. 27 PCE) Presentation F. Baret Presentations T. Altman P Lejeune Lemnatec CropDesign Phenopsis Granier et al. 26 New Phyt Field Montes et al. Trends in pl sc 28 We need to refine biological questions avoiding reinventing the wheel Modelling can help...

4 1 Phenotyping The problem : controversial statements (food for discussion) Confusions of effects (examples with panels of lines) Phenotyping : the most expensive way to discover genes of flowering time while looking for other targets? early Leaf number Biomass early mid early semi late late late Drought tolerance semi late mid early Leaf area / biomass Water reserve in the soil at flowering time Modelling biomass accumulation and transpiration rate as a function of leaf area and time what is left once the trivial effects are accounted for? field C. Welcker

5 1 Phenotyping The problem : controversial statements (food for discussion) Confusions of effects (examples with panels of lines) Phenotyping : the most expensive way to discover genes of flowering time while looking for other targets? early Leaf number Biomass early mid early semi late late late Drought tolerance semi late mid early Leaf area / biomass Water reserve in the soil at flowering time Also valid for leaf area and cell size in At (Cookson et al 27 Ann Bot 99:73) effect of candidate genes on cell cycle (plenty of examples) etc. C. Granier

6 1 Phenotyping The problem : controversial statements (food for discussion) Confusions of effects (examples with panels of lines) Plant water content, an expensive way of measuring plant size? NIR sensor in plants which are rehydrating in a black chamber Lemnatec website IPK-Gatersleben

7 1 Phenotyping The problem : controversial statements (food for discussion) Parent et al Plant Phys , 2- Confusions of effects (examples with panels of lines) Plant water content, an expensive way of measuring plant size? NIR sensor in plants which are rehydrating in a black chamber J J xc Leaf water potential (MPa) S (small plants) WT (large plants) Ψ evap R xl Ψ xyl Lp r Ψ r R sp Ψsol J xyl R c Ψ cel Small plants recover water status more rapidly (this effect can be tested with a model)

8 1 Phenotyping The problem : controversial statements (food for discussion) Phenotyping platforms in controlled conditions not appropriate for plant performance (small pots, low light...) Field phenotyping platforms will never be numerous enough for a proper network of experiments (GxE, QTL x E)... Phenotyping platforms are for something else - detecting / analysing heritable traits involved in yield - in a modeller's language : parameters of models Phenodyn Lemnatec / CropDesign Phenopsis Field

9 1. Phenotyping : The problem Partial conclusion Roles for modelling? "Model assisted phenotyping" for : - Optimising designs to the biological question (variables, precision...) - Identifying heritable traits from massive (and messy) datasets - Scale up from individual traits to whole-plant performance in the real world.

10 1. Optimising designs to the biological question : which variables? Biomass = t Incident light * % intercepted * Radiation Use Efficiency (RUE) Duration biomass RUE Intercepted Radiation Plant architecture Leaf # Leaf size A / g s (WUE) g s Duration of growth max. rate sensitivity Monteith 1977

11 2. Optimising designs to the biological question : which variables? Biomass = t Incident light * % intercepted * Radiation Use Efficiency (RUE) Incident light Spatial variability of environmental conditions : acceptable or not? Modelling light at plant level Modelling phenotypic consequences Rosette leaf area (cm²) PPFD (mol m -2 d -1 ) Michael Chelle Inra Grignon Granier et al. 26 New Phyt.

12 2. Optimising designs to the biological question : which variables? Biomass = t Incident light * % intercepted * Radiation Use Efficiency (RUE) Architecture : Which variables for a genetic and GxE analyses Digitizing E Genetic / environmental analyses of parameters I II III IV V QTL analysis Karin Chenu and Christian Fournier, unpublished

13 2. Optimising designs to the biological question : which variables? Biomass = t Incident light * % intercepted * Radiation Use Efficiency (RUE) Stomatal conductance (impossible to measure at high throughput with gas exchange equipment) Environmental conditions, measured s Φ n + ρ a c p VPD air g a J w = λ [ s + γ (1+g a / g s )] * F(leaf area) Transpiration measured Estimated measured

14 2. Optimising designs to the biological question : which variables? Partial conclusion - Relevant variables for biological questions may need to be calculated from models (hidden variables, not raw phenotypes), - Modelling : testing environmental scenarios / variability

15 3 Organising the mess for genetic analyses - Experiments finished (different growing periods with different climates) - Environmental conditions differ between periods, genotypes,stages -1 4 to 1 6 datapoints per experiment, not that clean Handling data prior to genetic analyses Phenodyn Lemnatec, CropDesign Phenopsis Field

16 3 Organising the mess for genetic analyses Biomass = t Incident light * % intercepted * Radiation Use Efficiency (RUE) How to manage time in fluctuating conditions? - different growing periods with different temperatures - estimation of rates and durations biased by temperature fluctuations Temperature-compensated time and rates ('thermal time', 'degree days' 'GDD') Phenodyn Lemnatec, CropDesign Phenopsis Field

17 3 Organising the mess for genetic analyses : temperature compensation Fluctuating temperature (field, greenhouse) : Normalised rates Temperature Leaf elongation Seedlings elongation Cell division Leaf initiation Leaf appearance Germination Lehenbauer 1914 Ben Haj Salah 1995 Warrington 1983 Warrington We do not know the mechanisms of response - We do not know the mechanisms for coordination of processes - But extremely robust "metamechanism"

18 3 Organising the mess for genetic analyses : temperature compensation Normalised rates 2 1 Rice Temperature A.thaliana Temperature Leaf elongation Development Germination Yin 1996 Seedlings elongation Relative elongation rate Leaf initiation Leaf development Germination Orbovik 27 Granier 22 Rate = k.t.exp(-ea/rt) 1+ exp( S/R- H/RT)

19 3 Organising the mess for genetic analyses : temperature compensation LER (mm h -1 ) Apo Azucena CG14 IR64 Nipponbare Moroberekan Vandana Temperature ( C) Very low genetic variability of the response : - Rice, 7 lines - maize 35 lines, (Sadok et al. 27, PCE)

20 3 Organising the mess for genetic analyses : temperature compensation Model - Extremely robust - No idea of the mechanisms - A way to express time "as if" T was 2 C Rate = k.t.exp(-ea/rt) 1+ exp( S/R- H/RT) Rates 2 1 Maize Rice A.thaliana Parent et al J. Exp Bot Temperature ( C)

21 3 Organising the mess for genetic analyses : temperature compensation LER (mm Cd -1 ) Temperature compensation in a large range of T : rapidly fluctuating developmental variables (leaf elongation rate) LER (mm h -1 ) LER (mm h -1 ) Temperature ( C) LER (mm h 2 C -1 ) Time of day Temperature ( C) LER (mm h -1 2 C )

22 3 Organising the mess for genetic analyses : temperature compensation.1 mm initiation Initiation or end expansion 2 cm end expansion Time after germination (compensated) Serves to predict development of any leaf or to know the status of all developing leaves at a given time

23 3 Organising the mess for genetic analyses : temperature compensation Gc-14 C Gc-26 C Gh-2 C 1mm 1mm 1mm 6 d after initiation 1mm 1mm 1mm 14 Cj afterinitiation Temperature-compensated rates or durations allow joint analyses of several experiments and to identify patterns : essential for high throughput phenotyping (controlled and field Granier, Massonet, Turc, Muller and Tardieu, 22, Annals of Bot 89:595-64

24 3 Organising the mess for genetic analyses : rapidly fluctuating variables Some key variables with genetic variability fluctuate in minutes How to deal with them? Biomass = t Incident light * % intercepted * Radiation Use Efficiency (RUE) Leaf / root growth rate WUE g s A

25 3 Organising the mess for genetic analyses : rapidly fluctuating variables Example : leaf elongation rate x 2 RILs, 3-4 experiments Leaf elongation rate (mm. Cj -1 ) sensitivity evap. demand Time of day detail : sensitivity soil water deficit Time of day, progressive water deficit Sadok et al. 27 PCE 3,

26 3 Organising the mess for genetic analyses : rapidly fluctuating variables LER (mm. Cd -1 ) Response curve of each RIL of mapping populations each genotype, one set of parameters (field, chamber, greenhouse) dl/dt = a - b VPD la - c Ψ intrinsic growth rate of the genotype sensitivity to evapor. demand sensitivity to soil water deficit Meristem Temperature ( C) Evaporative demand VPD (kpa) (WW conditions, day) Predawn Leaf Water Potential (MPa) (W Deficit, night

27 3 Organising the mess for genetic analyses : rapidly fluctuating variables -1 QTLs of sensitivity c b b c c c c c c b b F-2 x Io F-2 x F252 P1 x P2 c c c b b c c LER (mm.h c b b c -1 ) Reymond et al. 23 Plant phy 114, 893- Welcker et al. 27 J. Exp Bot 58, 339 1Sadok 2 3 et al 27 PCE 3, 135

28 3 Organising the mess for genetic analyses : rapidly fluctuating variables Does the model hold? dl/dt = a - b VPD la - c Ψ 4 3 A Experimental fits for each RIL r 2 =.83 intrinsic growth rate of the genotype QTLs sensitivity to evapor. demand QTLs Virtual genotypes under any scenario It is feasible to carry out an in silico prevision of genotypes sensitivity to soil water deficit QTLs Predicted LER mm h B QTL model r 2 =.8 4 C QTL model with new RILs or PLs 3 r 2 = Reymond et al. 23 Plant Phy Observed LER mm h -1

29 3 Organising the mess for genetic analyses : Partial conclusion Strategy for data mining : we need specific methods aimed at answering questions Temperature compensation, a key question (to which variables? ) Rapidly fluctuating variables : "hidden variables" can - synthesise genotypic behaviours - be analysed genetically - allow reconstruction of the original phenotype. Data mining for response curves in litterature H Poorter's talk

30 4. From platform to fields I have QTLs or associations for traits in phenotyping platforms (controlled or field) Can they predict performance and G x E in a large range of scenarios? Needs field experiments BUT - Never enough situations to test the interest of an allele - Usually partly fails : where and why? Predicting the allele effect with model AND comparing with fields

31 4. From platform to fields Climatic data + Morphogenetic programme Leaf # Thermal Time + Sensitivity to environ. conditions virtual plant / genotype (with effect of QTLs) Evaporative demand VPD (kpa) Soil water Potential (MPa) Chenu et al. 28 Plant Cell Environment 31, 378 Col. Hammer, Chapman U. Queensland

32 4. From platform to fields 15 Leaf rank 1 5 Wet air Dry air Final lamina length (cm) Well watered water deficit Observed (symbols) and simulated (lines) leaf lengths under contrasting climatic scenarios in the field Chenu et al. 28 Plant Cell Environment 31, 378 Col. Hammer, Chapman U. Queensland

33 4. From platform to fields Climatic data virtual plant / genotype (with effect of QTLs) effect of allelic composition on plant performance calculated feedbacks of plants on environment (e.g. soil depletion) iomass = Incident light * % intercepted * Radiation Use Efficiency (RUE)

34 Leaf elongation rate or transpiration rate 4. From platform to fields Effects of leaf QTLs on simulated yield Wet air dry air wet dry Soil water potential (MPa) 4 2 Leaf area Alleles which maintain growth under stress Tester Alleles which decrease growth under stress 8 4 Yield (t Ha -1 ) WW early drought flowering dr. terminal dr WW early drought flowering dr. terminal dr Maintaining leaf area most often favourable except in terminal stress Chenu et al. 29 Genetics

35 4. From platform to fields Effects of leaf QTLs on simulated yield Test requires - Adequate environmental measurements in the field - Methods for field phenotyping + model - Multi-environmental tests F. Baret's presentation

36 CONCLUSION - A big risk if phenotyping is not considered via biological questions (phenotyping a new name for whole plant physiology? ) - Phenotyping plaforms are not 'easier fields' : best use for identifying heritable parameters or 'hidden variables' used in models. - Analysis of large messy datasets needs a framework of analysis - temperature-compensated rates (joint analysis of experiments - Hidden variables which synthesise time courses - Platform to field : needs a theoretical framework - helps interpretation - allows 1s of virtual experiments in different scenarios