High throughput phenotyping and plant modelling · Lemnatec CropDesign Phenopsis Granier et al....

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

Ψsoil

Ψr

Ψxyl

Ψevap Ψcel

Jxyl

Jxc

J

Rsp

Rc

1. Phenotyping The problem

Phenotyping is the bottleneck in genomics.- Sequences of Arabidopsis, Rice, Sorghum, poplar, tomato, maize etc- Genotyping capacities for 1000s 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 ?

1. Phenotyping The problem

The 'problem' is that we can now measure traits in1000sof plants in a robotised way

We need to refine biological questionsavoiding reinventing the wheel

Modelling can help...

Phenodyn (Sadok et al. 2007 PCE)

LemnatecCropDesign

PhenopsisGranier et al. 2006 New Phyt

Field Montes et al. Trends in pl sc 2008

PresentationsT. AltmanP Lejeune

PresentationC. Welcker

PresentationF. Baret

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 ?

Modelling biomass accumulation and transpiration rate as a function of leaf area and timewhat is left once the trivial effects are accounted for ?

Leaf

num

ber

earlymid earlysemi latelate

Leaf area / biomass

Biomass

early

semi late mid early

late

Water reserve in the soilat flowering time

Drought tolerance

C. Welcker

field



Phenotyping : the most expensive way to discover genes of flowering time while looking for other targets ?

Also valid for leaf area and cell size in At (Cookson et al 2007 Ann Bot 99:703)effect of candidate genes on cell cycle (plenty of examples) etc.

Leaf

num

ber

earlymid earlysemi latelate

Leaf area / biomass

Biomass

early

semi late mid early

late

Water reserve in the soilat flowering time

Drought tolerance

C. Granier

Plant water content, an expensive way of measuring plant size ? NIR sensor in plants which are rehydrating in a black chamber



Lemnatecwebsite

IPK-Gatersleben

00

-0.4

Leaf

wat

er p

oten

ti al

(MP

a)

-0.8 WT (large plants)

S (small plants)

-1 0 1 32

Ψevap

Jxyl

Lpr

Rsp

Ψsol

Ψr

Ψxyl

Ψcel

Jxc

J

Rxl Rc

Small plants recover water status more rapidly(this effect can be tested with a model)

Parent et al Plant Phys 2009 149, 2000-



Plant water content, an expensive way of measuring plant size ? NIR sensor in plants which are rehydrating in a black chamber

Phenotyping platforms in controlled conditions not appropriate for plant performance (small pots, low light...)

Field phenotyping platforms will never be numerous enoughfor a proper network of experiments (GxE, QTL x E)

... Phenotyping platforms are for something else


Phenodyn

Phenopsis

Phenodyn

Phenopsis

Phenodyn

Lemnatec / CropDesign PhenopsisField

- detecting / analysing heritable traits involved in yield- in a modeller's language : parameters of models

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.

1. Optimising designs to the biological question : which variables ?

Durationof growth

max. rate

sensitivity

Leaf # Leaf size gsA / gs(WUE)

Plantarchitecture

Biomass = Incident light * % intercepted * Radiation Use Efficiency (RUE) ∫t

0

Intercepted Radiation

biom

ass

RUEDuration

Monteith 1977

Incident light Spatial variability of environmental conditions : acceptable or not ?

14

15

16

17

18

19

20

21

0 50 100 150 200

0

20

40

6080

100

Michael Chelle Inra Grignon

Modelling light at plant level


Ros

ette

leaf

area

(cm

²)

0 3 6 9 12 150

5

10

15

20

PPFD (mol m-2 d-1)

Granier et al. 2006New Phyt.

Modelling phenotypic consequences


0Biomass = Incident light * % intercepted * Radiation Use Efficiency (RUE) ∫

t

0∫t

0

Architecture : Which variables for a genetic and GxE analyses

EDigitizing

Genetic / environmental analyses of parameters

I II III IV V


Biomass = Incident light * % intercepted * Radiation Use Efficiency (RUE) Biomass = Incident light * % intercepted * Radiation Use Efficiency (RUE) ∫t

0∫t

0

Karin Chenu and Christian Fournier, unpublishedQTL analysis

Stomatal conductance(impossible to measure at high throughput with gas exchange equipment)


s Φn + ρa cp VPDair gaJw = -------------------------λ [ s + γ (1+ga / gs)]

s Φn + ρa cp VPDair gaJw = -------------------------λ [ s + γ (1+ga / gs)]

Transpirationmeasured

Environmental conditions, measured

* F(leaf area)

measuredEstimated


0∫t

0


Partial conclusion

- Relevant variables for biological questions may need to be calculatedfrom models (hidden variables, not raw phenotypes),

- Modelling : testing environmental scenarios / variability

3 Organising the mess for genetic analyses

Phenodyn

Phenopsis

Phenodyn

Phenopsis

Phenodyn

Lemnatec, CropDesign PhenopsisField

- Experiments finished (different growing periods with different climates)- Environmental conditions differ between periods, genotypes,stages- 104 to 106 datapoints per experiment, not that clean

Handling data prior to genetic analyses

3 Organising the mess for genetic analyses

Phenodyn

Phenopsis

Phenodyn

Phenopsis

Phenodyn

Lemnatec, CropDesign PhenopsisField

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')


0∫t

0

0

1

2

3010 20 40Temperature

Nor

mal

ised

rate

s Leaf elongation

Seedlings elongation

Cell division

Leaf initiation

Leaf appearance

Germination

Lehenbauer 1914

Ben Haj Salah 1995

Warrington 1983

Warrington 1983

Fluctuating temperature (field, greenhouse) :

- We do not know the mechanisms of response- We do not know the mechanisms for coordination of processes- But extremely robust "metamechanism"

3 Organising the mess for genetic analyses : temperature compensation

Leaf elongationRelative elongation rateSeedlings elongation

Leaf developmentGermination

Leaf initiationGermination

Development

1

2

03010 20

0

0.5

1

1.5

3010 20

Nor

mal

ised

rate

s

Temperature Temperature

Rice A.thaliana

Orbovik 2007

Granier 2002Yin 1996

k.T.exp(-Ea/RT)Rate =1+ exp(∆S/R- ∆H/RT)


Very low genetic variability of the response : - Rice, 7 lines- maize 350 lines , (Sadok et al. 2007, PCE)

0

2

4

0

2

4

5 15 25 35 5 15 25 35 5 15 25 35Temperature (°C)

LER

(mm

h-1

)

Apo Azucena CG14

IR64 Nipponbare Moroberekan Vandana

5 15 25 35


k.T.exp(-Ea/RT)Rate =1+ exp(∆S/R- ∆H/RT)

Model- Extremely robust - No idea of the mechanisms

15 20 25 30 350

1

2

Temperature (°C) 15 20 25 30 35

0

1

2

MaizeRiceA.thaliana

Rat

es

- A way to express time "as if" T was 20°C


Parent et al J. Exp Bot 2010.

LER (mm h20°C-1)

3LE

R

(mm

h 20°C-1)

Temperature (°C

)

LER

(m

m C

d-1)

LER (mm h-1)Temperature (°C)

LER

(m

m h

-1)

2

1

0

12

8

4

0

6

4

2

0

20

10

0

4

Time of day4 4418 1818


Temperature compensation in a large range of T : rapidly fluctuating developmental variables (leaf elongation rate)

Serves to predict development of any leaf or to know the status of all developing leaves at a given time

Initi

atio

n or

end

expa

nsio

n

Time after germination (compensated)

0.1 mm

2 cm

end expansioninitiation


6 d afterinitiation

140°Cj afterinitiation

Gc-14°C Gc-26°C Gh-20°C1mm1mm

1mm 1mm

1mm

1mm

Granier, Massonet, Turc, Muller and Tardieu, 2002, Annals of Bot 89:595-604


Temperature-compensated rates or durations allow joint analyses of several experiments and to identify patterns :

essential for high throughput phenotyping (controlled and field

3 Organising the mess for genetic analyses : rapidly fluctuating variables

Leaf / rootgrowth rate

Some key variables with genetic variability fluctuate in minutes

How to deal with them ?

gsWUE


0A

Sadok et al. 2007 PCE 30, 135–146


Example : leaf elongation rate

x 200 RILs, 3-4 experiments

0.5

1.5

2.5

3.5

4.5

6 12 18 0Time of day

-

0

Leaf

elon

gatio

nra

te(m

m.°

Cj-1

)

sensitivityevap. demand

Time of day, progressive water deficit

sensitivitysoil water deficit

detail : http://bioweb.supagro.inra.fr/phenodyn/

dl/dt = a - b VPDla - c Ψ

LER

(m

m.°

Cd-1

)

MeristemTemperature

(°C)

10 20 300

4

2

6

Evaporative demand VPD (kPa)

(WW conditions, day)

0 1 2 3Predawn Leaf Water

Potential (MPa)(W Deficit, night

-0.4-0.20.0

intrinsic growth rateof the genotype

sensitivity to evapor. demand

sensitivity to soil water deficit

Response curve of each RIL of mapping populationseach genotype, one set of parameters (field, chamber, greenhouse)



-1 0 10 20 30

F-2 x IoF-2 x F252P1 x P2

LER

(mm

.h-1 ) 0 10 20 30

b

b

b

b

b

b

b

b

c

c

c

c

cc

c

c

cc

cc

c

c

Reymond et al. 2003 Plant phy 114, 893-Welcker et al. 2007 J. Exp Bot 58, 339Sadok et al 2007 PCE 30, 135–

QTLs of sensitivity

Reymond et al. 2003 Plant Phy

0

1

2

3

4

0

1

2

3

4

A Experimental fits for each RIL r2 = 0.83

B QTL model r2 = 0.80

Pre

dict

ed L

ER

mm

h-1

Observed LER mm h-1 0 1 2 3 4

0

1

2

3

4 C QTL model with new RILs or PLs r2 = 0.74

Virtual genotypes under any scenario

It is feasible to carry out an in silicoprevision of genotypes

dl/dt = a - b VPDla - c Ψ

intrinsic growth rateof the genotype

QTLs

sensitivity to evapor. demand

QTLs

sensitivity to soil water deficit

QTLs


Does the model hold ?

3 Organising the mess for genetic analyses : Partial conclusion

Strategy for data mining : we need specific methods aimed at answering questions

Data mining for response curves in litteratureH Poorter's talk

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.

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

0 200 400 600 800 1000 1200

02

46

810

12

Thermal Time

Leaf

#


Evaporative demand VPD (kPa)

0 1 2 3Soil water Potential

(MPa)

-1.0-0.50.0 -1.5Evaporative demand

VPD (kPa)

0 1 2 3Soil water Potential

(MPa)

-1.0-0.50.0 -1.5

Climatic data

+Morphogenetic programme

Sensitivity to environ. conditions+

virtual plant / genotype(with effect of QTLs)

Chenu et al. 2008 Plant Cell Environment 31, 378–Col. Hammer, Chapman U. Queensland

20 40 60 80 100

Well wateredwater deficit

Final lamina length (cm)0 20 40 60 80 100

Leaf

rank

0

5

10

15

Wet airDry air

Observed (symbols) and simulated (lines) leaf lengths under contrasting climatic scenarios in the fieldChenu et al. 2008 Plant Cell Environment 31, 378–Col. Hammer, Chapman U. Queensland



virtual plant / genotype(with effect of QTLs)

effect of allelic composition on plant performance

Biomass = Incident light * % intercepted * Radiation Use Efficiency (RUE) Biomass = Incident light * % intercepted * Radiation Use Efficiency (RUE)

Climatic data

calculated feedbacks of plants onenvironment (e.g. soil depletion)

Leaf

are

a

0

2

4

Effects of leaf QTLs on simulated yield

0

4

8

Yie

ld (t

Ha-1

)

Maintaining leaf area most often favourableexcept in terminal stress Chenu et al. 2009

Genetics

Alleles which maintain growth under stress

Tester Alleles which decrease growth under stress

Leaf

elo

ngat

ion

rate

or tr

ansp

iratio

n ra

te

Soil water potential (MPa)

0

4

2

6

-1.-0.50.0wet dry

Leaf

elo

ngat

ion

rate

or tr

ansp

iratio

n ra

te


0

4

2

6

-1.-0.50.0wet dry

Leaf

elo

ngat

ion

rate

or tr

ansp

iratio

n ra

te


0

4

2

6

-1.-0.50.0wet dry

Wet air dry air

WW

WW

early

dro

ught

flow

erin

g dr

.

term

inal

dr

early

dro

ught

flow

erin

g dr

.

term

inal

dr


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


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 foridentifying 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 100s of virtual experiments in different scenarios

High throughput phenotyping and plant modelling · Lemnatec CropDesign Phenopsis Granier et al....

Documents

Transcript of High throughput phenotyping and plant modelling · Lemnatec CropDesign Phenopsis Granier et al....