A simple leaf-scale model for assessing life-history traits of fungal parasites with growing lesions

Transcription

1 A simple leaf-scale model for assessing life-history traits of fungal parasites with growing lesions Melen Leclerc, Frédéric Hamelin & Didier Andrivon INRA - Institute for Genetics, Environment and Plant Protection ModStatSAP Network meeting - Paris - March 2018

2 Context Assessing life-history traits of pathogens is central for several purpose in phytopathology Understand the biology and the evolution of pathogens Evaluate the effects of plant resistance on the life-cyle of pathogens Optimise plant breeding for resistance Predict the spread, the emergence or the adaptation of plant pathogens...

3 Pathogen with growing lesions Empirical studies often rely on observations at the lesion scale The lesion scale is also central for studies aiming at scaling-up epidemic development ( Cunniffe at al., 2012 ; Segarra et al., 2001 ; Spijkerboer, 2004) When the pathogen spreads substantially in host tissues and induces a growing lesion, the direct measurement of key life-history traits such as latency period or sporulation dynamics is challenging One needs to disentangle the spatial spread of the pathogen and the epidemiological dynamics of each infinitesimal surface after infection ( Hethcote and Tudor, 1980 ; Powell et al., 2005 ; Segarra et al., 2001) It has received little attention in theoretical plant disease epidemiology (Powell et al.,2005)

4 Aims of the study Build a parsimonious mechanistic model for assessing life-history traits of pathogens with growing lesions Use this epidemiological model to analyse empirical data and provide estimates of the spread of the pathogen, the latency period and the sporulation dynamics Assess the effects of quantitative resistance on the pathogen

5 Introduction Experimental data Models Results Conclusion Perspectives Experimental data Phytophthora infestans which causes the late blight of potato 2 strains K = {BP3, BP6} 3 cultivars J = {Bintje, Möwe, Désiré} More than 100 inoculated leaves for each strain x cultivar pair For each destructive observation Oi, measurement of the size of the lesion at time since inoculation, the minor and major radii of the leaf, and the number of spores present on the leaf

6 A simple model for the growing lesion We consider that the leaf is an ellipse with radii R1 < R2

7 A simple model for the growing lesion Lesion growth model Leaf size L = R 1R 2π with R1 < R2 ellipse radii Let t 0 0 be the delay between inoculation and the initiation of the lesion. The surface of the lesion at time t is given by : l(t) = π min(ρ(t t 0), R 1) min(ρ(t t 0), R 2)

8 A simple model for the growing lesion

9 Sporulation model Spore production at the lesion scale Let be σ(a) a continuous time emission function giving the quantity of emitted spores at age since infection a per infinitesimal spatial infectious unit. The total number of spores produced at time t by the lesion is given by : s(t) = t l(t a)σ(a)da 0

10 Sporulation model Spore production at the lesion scale Let introduce a latency period t 1. The total number of spores produced at time t by the lesion is given by : s(t) = t t 1 0 l(t + t 0 t 1 a)σ(a)da

11 Sporulation model Spore production at the lesion scale Let introduce a latency period t 1. The total number of spores produced at time t by the lesion is given by : s(t) = t t 1 0 l(t + t 0 t 1 a)σ(a)da Emission function We define the continuous emission function σ(a) using a Rayleigh distribution : ( ) σ(a) = S a exp a2 µ 2 2µ 2

12 Sporulation model Then, we obtain : 0 if t t 0 δ = t 1, s s(t) = 1 (t, t) if t 0 δ t < T 1 δ, s 1 (t, T 1 ) + s 2 (t, t) if T 1 δ t < T 2 δ, s 1 (t, T 1 ) + s 2 (t, T 2 ) + s 3 (t) if T 2 δ t. s 3 (t) = LS 1 exp (t + δ T 2 )2 2µ 2. (2) (1) and (t T )2 s 2 (t, T ) = SπR 1 ρ (T + δ t 0 ) exp 2µ 2 (T 1 t 0 ) exp (t + δ T 1 )2 2µ 2 ( ( ( ) ( ))) π t T t + δ T1 + SπR 1 ρ µ erf erf, (3) 2 2µ 2µ s 1 (t, T ) = Sπρ 2 ( (t + δ t 0 ) 2 + 2µ 2) (T t)2 exp 2µ 2 2µ 2 exp (t + δ t 0 )2 2µ 2 + Sπρ 2 ( 2πµ(t + δ t 0 ) ( erf ( ) ( ))) t + δ t0 T t + erf. (4) 2µ 2µ where erf is the Gauss error function.

13 Models fitting Fraction of symptomatic tissues (%) Bintje*BP6 Bintje*BP3 Möwe*BP6 Möwe*BP3 Désiré*BP6 Désiré*BP Time since inoculation (h)

14 Models fitting (a) (b) (c) (d) (e) (f) Time since inoculation (h) Residuals BP6 BP3 Bintje Möwe Désiré

15 Models fitting Sporulation function Bintje*BP6 Bintje*BP3 Möwe*BP6 Möwe*BP3 Désiré*BP6 Désiré*BP Time since sporulation (h)

16 Models fitting e+05 0e+00 6e+05 (a) e+05 0e+00 6e+05 (b) e+05 0e+00 6e+05 (c) e+05 0e+00 6e+05 (d) e+05 0e+00 6e+05 (e) e+05 0e+00 6e+05 (f) Time since inoculation (h) Residuals BP6 BP3 Bintje Möwe Désiré

17 Models fitting Parameter Unit BP3 BP6 Bintje Möwe Désiré Bintje Möwe Désiré t 0 hours ρ cm.hours t 1 hours s spores.cm µ hours

18 Statistical analysis Pairwise comparison of treatments using F-tests (Gilligan, 1990) : F = [RSS CM RSS X ]/[df CM df X ] [RSS SM1 + RSS SM2 ]/df SM Non-significant parameters : BP3 vs BP6 t 1 on Bintje, µ on Möwe, latency on Désiré BP3 t 1 for Bintje vs Désiré and Möwe vs Désiré BP6 t 0 for Bintje vs Désiré, {t 1, s, µ} for Möwe vs Désiré

19 Conclusion A simple modelling framework for assessing life-history traits of pathogens with growing lesions ( we have an analytical solution!) Some discrepancies but it captures the main processes Using mechanistic models for analysing the observation of plant lesions provides a better understanding of the effects of plant resistance on pathogens For the late blight of potato : strain x cultivar interaction quantitative resistance : reduces the speed of the spread of the lesion and the number of emitted spores, induces a change in the emission function (delayed mode) less effects on the latency period

20 1 Investigate other pathosystems : Perspectives Mycosphaerella pinodes & Phoma medicaginis on pea (Ascochyta blight of pea disease complex) Ontogenetic and disease-induced changes in host susceptibility (plant senescence - yellowing) (Richard et al., 2012) 2 Improve experimental measurements : Multi-modal imaging : RGB, fluorescence, hyperspectral (data assimilation with filtering methods??) Molecular techniques for measuring the density of pathogens and plant responses

21 Perspectives 3 Spatially-explicit models : Reaction-diffusion models (homogeneous and/or inhomogeneous) u =.(D(t, x).u) + f (t, x, u) t where u is the cryptic density of the pathogen

22 Perspectives 3 Spatially-explicit models : Nonlinear diffusions (ontogenetic and disease-induced changes in the susceptibility) u =.(Du(t, x, u, v).u) + fu(t, x, u) t v =.(Dv (t, x, u, v).v) + fv (t, x, v) t D u(t, x, u, v) = g u(v) D v (t, x, u, v) = g v (u) where u is the pathogen and v related to host physiology/susceptibility

23 Perspectives 3 Spatially-explicit models : Nonlinear cross-diffusion (co-infections) u t = Du 2 u + D uv 2 u + f u(u) + g u(u, v) v t = Dv 2 v + D vu 2 u + f v (v) + g v (u, v) where D u and D v are the diffusion coefficients of respectively u and v, D uv and D v u are cross diffusion coefficients of u and v respectively.

24 Perspectives 4 Relax the hypotheses on the distribution of latency and infectious periods Generic Integral equation models (Hethcote et al. 1980, Segarra et al. 2001) Easier if (Generalised) Erlang distributions... ds ( n ) = β dt k=1 I S, k de 1 ( n ) = β dt k=1 I S mγe k 1, de i = mγe i mγe i 1, (1 i m 1) dt di 1 = mγem nµi 1, dt di i = nµi i nµi i 1, (1 i n 1) dt dr = nµin. dt 5 Improve statistical inference (e.g. heteroscedasticity), optimal design of experiments for parameter estimation and model selection

25 Thank you for your attention!