^{1}

^{*}

^{1}

^{1}

^{2}

^{3}

^{4}

^{5}

^{6}

^{1}

Conceived and designed the experiments: FMN FJP SNT WO CAG. Performed the experiments: AB WSF FMN. Analyzed the data: FMN. Wrote the paper: FMN FJP SNT WO CAG DJB.

The authors have declared that no competing interests exist.

Heterogeneity in host populations is an important factor affecting the ability of a pathogen to invade, yet the quantitative investigation of its effects on epidemic spread is still an open problem. In this paper, we test recent theoretical results, which extend the established “percolation paradigm” to the spread of a pathogen in discrete heterogeneous host populations. In particular, we test the hypothesis that the probability of epidemic invasion decreases when host heterogeneity is increased. We use replicated experimental microcosms, in which the ubiquitous pathogenic fungus

Pathogen spread and epidemic invasion in plant, animal and human populations depend on host properties (infectivity, susceptibility) that can vary amongst hosts within the same population. However, such host variability (or heterogeneity) is typically difficult to control experimentally, and little explicit research has been done on its effects on pathogen invasion. We present the first systematic investigation on the spread of a pathogen (the fungal plant pathogen

Host heterogeneity is receiving increasing attention as one of the factors affecting the dynamics of epidemic spread. The properties of individual hosts, such as contact rate, susceptibility, or infectiousness, can vary across a population as a result of environmental

The experimental results presented here test for the first time the existence of a link between host heterogeneity and epidemic thresholds in a broad, relevant class of spatially-extended systems, thereby confirming recent theoretical predictions

Previous experiments

The experiment described in the present paper is inspired by a model by Neri

We consider systems of two different sizes, on triangular lattices with the same topology as for the populations used in the experiment. The graphs were obtained with numerical simulations (see

We use replicable microcosms

We set up a series of notional experimental treatments (replicated populations), designed in such a way to ensure an appropriate range for the average and variance of the transmissibility. The notional values of the parameters are chosen according to the theoretical predictions of Neri

Six experimental treatments, labelled from A to F, were designed (see below and

Spatio-temporal maps of fungal colonisation dynamics were used, in order to count the cumulative number of colonised sites over time, and to identify those replicates in which the fungus spreads invasively (

(A–F) Individual colonisation curves for each of the replicates of a treatment (thin lines) and the average over all the replicates (bold solid line). The upper limit of the vertical axis in all the panels concides with the total number of sites in the population (i.e.,

The variability in the final number of colonised sites amongst replicates of the same treatment (

In order to estimate

The analysis of posterior distributions for

Red crosses and blue circles correspond to invasive and non-invasive replicates, respectively (error bars not shown here). The green thick line is the discriminant function separating the invasive and non-invasive regimes. The purple dash-dotted line is the phase boundary for an infinite system with the same topology (see

The data presented in

A multiple logistic regression test (function

We have shown experimentally that between-host variability affects the nearest-neighbour spread of a pathogen in a population: when the variability is increased, the probability of epidemic invasion decreases. From a broad point of view, our results answer a very general question: what is the effect of individual variability on disease spread?

Here, for the first time, the approach and quantitative predictions of Neri

Our analysis showed that within-treatment variability can be large enough to mask the effects of experimental treatments in replicated populations (see

Our results have a potentially high impact in finding control strategies for the spread of disease. Let us consider a homogeneous system, with the same topology as for our experimental microcosms (

The treatments presented here are slightly idealized versions of the real experimental treatments A, C, F (the condition

In particular, if we assume that the amount of control agent is equal to

In the experimental design and the subsequent analysis, the parameters were evaluated in two steps, before and after the population experiment. Before the population experiment, the dependence of

In our experimental systems, each agar dot consisted of a small aliquot (10

The fact that the transition between the invasive and the non-invasive regime is “smeared out” (

The value of

We analysed the posterior distribution for the replicate transmissibility

(A) Posterior distributions for

(PDF)

(PDF)

(PDF)

(PDF)

We thank R. Stutt for providing tools for data acquisition.

_{0}as a predictor of disease invasion in structured populations.