1 Table S1. Factors that affect theextinction ofthe powderymildewpathogen Podosphaera plantaginis duringthe off-season (2011/2012) in its host populations in theÅland Islands, southwestern Finland. Shown arethe results from aspatial Bayesian model usingLAPLACE approximation. Estimatesin bold aresignificant as based on adecreasein DIC. Thespatial pattern of winter extinction is shown in Fig. 2b in themain manuscript. Factor mean 0.025quant 0.5quant 0.975quant DIC (Intercept) 1.1185 0.0595 1.1390 2.0728 Plant coverage(m2) -0.0180 -0.0380 -0.0173 -0.0018 246.31 Plant dryness (%) -0.0722 -0.1290 -0.0705 -0.0248 247.37 July rainfall (mm) -0.0222 -0.0406 -0.0225 -0.0021 253.52 Populationage (yr) -0.3331 -0.7259 -0.3248 0.0136 245.39 Full model 243.95 2 Table S2. Factors that affect theJulyabundance(year 2012) of thepowderymildewpathogen Podosphaera plantaginisin infected host populations in the Åland Islands, southwestern Finland. Shown arethe results from a spatial Bayesian model usingLAPLACE approximation. Estimates in bold aresignificant as based on adecreasein DIC. Factor mean 0.025quant 0.5quant 0.975quant DIC (Intercept) 1.7342 1.2615 1.7564 2.0721 Plant coverage(m2) 0.0030 0.0013 0.0030 0.0048 222.82 Distancetoshore(m) -0.0001 -0.0003 -0.0001 0.0000 216.41 Spatialconnectivity -0.0045 -0.0118 -0.0046 0.0035 214.57 Fullmodel 215.63 3 Table S3. Factors that affect thefraction ofinfected leaves with restingstructures of thepathogen Podosphaera plantaginisin its host populations inthe Åland Islands, southwestern Finland. Shown arethe results from year-specificspatial Bayesian models usingLAPLACE approximation. Estimates in bold aresignificant as based on adecreasein DIC. Thespatial pattern of restingspore production is shown in Fig. 4 in themain manuscript. 2010 Factor mean 0.025quant 0.5quant 0.975quant DIC (Intercept) 0.8689 0.5444 0.8722 1.1744 Plant dryness (%) 0.0042 0.0010 0.0042 0.0075 -114.46 Augustrainfall (mm) -0.0050 -0.0093 -0.0051 -0.0005 -77.99 Fullmodel -388.90 2011 Factor mean 0.025quant 0.5quant 0.975quant DIC (Intercept) 0.3838 0.1695 0.3837 0.5988 Habitat openness (0/1) 0.3681 0.2048 0.3680 0.5321 -99.23 Spatial connectivity 0.0021 0.0004 0.0021 0.0039 -121.72 July rainfall (mm) 0.0020 0.0003 0.0020 0.0036 -114.15 Augustrainfall (mm) -0.0029 -0.0051 -0.0029 -0.0006 -125.35 Populationage (yr) -0.0238 -0.0378 -0.0238 -0.0097 -105.98 Full model -139.30 4 2012 Factor mean 0.025quant 0.5quant 0.975quant DIC (Intercept) 0.3351 0.2515 0.3351 0.4183 Plant coverage(m2) 0.0000 0.0000 0.0000 0.0000 -141.38 Habitat openness (0/1) 0.0004 -0.0012 0.0004 0.0021 -134.29 Populationage (yr) 0.0001 -0.0013 0.0001 0.0015 -137.84 Fullmodel -139.85 5 Table S4. Theimpact of population oforigin and off-season storagelocation on the abilityof restingstructures to infect individuallycaged plants in spring. SeeFig. 5 ofthe main manuscript for avisual depiction of thedata. Factor Infection Infectedleaves / Total number of leaves Pathogen populationof origin X2=0.00, p=1.00 X2 = 0.38, p = 0.54 Storagelocation X2=0.00, p=1.00 X2=0.00, p=1.00 Pathogen populationof originx Storage location X2=15.43,p<0.001 Micro-site (Storage location) X2=3.66, p=0.06 X2=3.87, p=0.05 Plant (Pathogen populationof origin) X2=2.83, p=0.09 X2=62.05,p<0.001 Receivingplant genotype X2=19.90,p<0.001 Number of restingstructures F1,161= 4.71; p=0.03 F1,137=26.13, p<0.001 6 Fig. S1. Restingstructures of thepowderymildewPodosphaera plantaginis. Panel A shows the conspicuous brown/black spores as detected in field surveys. Panel Bdepicts the mature(brown) and immature(white/green) restingstructures as seen under astereomicroscope. Panel C shows a restingstructurerevealingits single ascus with up to eight ascospores. Photo credits: Anna-Liisa Laineand RiikkaAlanen. 7 Fig. S2. Locations where restingstructures werecollected in autumn and/or stored duringthe off- season. Red triangles indicate pathogen populations whererestingstructures werecollected in autumn 2010 and stored under various indoor conditions. Green squares indicate populations where restingstructures werecollected in autumn 2010, which weresubsequentlystored duringthe off- season in both thegreen and purplesquares. Thepink circles show thelocation whereresting structures werecollected in autumn 2011, and subsequentlystored reciprocallyin thesesame five locations. In autumn 2011, wecaged c. ten plants in each of six populations (pink and yellow circles). 8 Fig. S3. Posterior distribution ofthe spatial rangeestimateforaspatial Bayesian model forA) winterextinction ofthe pathogen Podosphaera plantaginis from 2011 to 2012 and B) pathogen abundancein July2012. Thespatial pattern of winterextinction is shown in Fig. 2b in themain manuscript. 9 Fig. S4. Posterior distribution ofthe spatial rangeestimateforthe fraction of infected leaves with restingstructures for thepathogen Podosphaera plantaginis in its host populations in Åland, southwestern Finland. Shown arethe results from year-specificspatial Bayesian models using LAPLACE approximation. Thespatial pattern of restingsporeproduction is visualized in Fig. 4 in the main manuscript. 10 Fig. S5. Fraction ofinfected cageplants in the overwinteringtrial experiment. Leaves werestored indoors under multiplecontrolled temperatures (-10°C, -5.5°C, 0.1°C, 5°C, 10°C and field°C; the latterreflected theaveragemonthlytemperatureduringwinter in Åland from Octoberto April: 8.8°C, 4.0°C, 1.9°C, 2.0°C, 2.0°C, 0.8°C, and 4.7°C). Theeight outdoor locations (populations 1043, 1413, 490, 1290, 4450, 876, 4698 and 8538)weredistributed across Åland (squares in Fig. S2). 11 Notes S1. A trial experiment on overwinteringsurvival usingindoor and outdoor overwintering sites Aims Wecarried out atrial overwinteringexperiment to simultaneouslyi) provide afirst demonstration that restingstructures are able to infect plants in springin this pathosystem, and ii) assess the impact of winter conditions on the survival ofrestingstructures collected from different populations. Materials and methods Wehaphazardlycollected leaves bearingrestingstructures in autumn 2010 from several populations (triangles and squares in Fig. S2). As asingle leaf and population can contain multiple pathogen genotypes (Tollenaere et al., 2012), wenote that such an experiment cannot measure variation amongindividual pathogen genotypes for overwintering, but it mayindicatewhetherthere is differentiation amongpathogen populations. Westored leaves from threepopulations (red triangles in Fig. S2) under multiplecontrolled indoor conditions (-10°C, -5.5°C, 0.1°C, 5°C, 10°C and ‘field temperature’; thelatterreflected the averagemonthlytemperatureduringwinterin Åland from Octoberto April: 8.8°C, 4.0°C, 1.9°C, 2.0°C, 2.0°C, 0.8°C, and 4.7°C). Leaves from an additional four populations (green squares in Fig. S2) werestored reciprocallyin each of thefour populations from which theleaves werecollected, as well as in an additional fourpopulations (purplesquares in Fig. S2). Leaves werestored individuallyin polyester pollination bags (PBS International) within theaforementioned indoor and field locations. In April 2011, leaves were taken from theindoor storages and field locations. To assess whether leaves bearingresting structures wereableto infect plants in spring, wethen hungtwo leaves bearingrestingstructures from thesame ‘population oforigin / overwinteringsite’combination aboveasingleplant individual usingtwo vertical sticks and horizontal iron wire. Plants wereindividuallycaged usinga polyester pollination bag(PBS International 10-1; 1-window; 255 x 510 mm), as previous work has shown that infection develops well in thesebags, and spores cannot leaveor enter(Laine, 2011). Plants werescored on 21 Juneforthe presenceof powderymildewinfection. Results Environmental conditions duringthe off-season had amajorimpact on viabilityand spring germination ofthe restingstructures. Theexperiment revealed that none of the restingstructures stored indoors wereable to infect caged plants in spring, even though theywerestored at a wide rangeof temperatures (-10 to +10 °C; Fig. S5). This is in strikingcontrast with an averageinfection 12 percentageof 34%(11 out of 32 caged plants) when restingstructures werestored overwinterin natural populations (Fig. S5). Thereis also an indication that off-season survival in thefield varied amongthe overwintering sites (Fig. S5). 13 Methods S1. A detailed description of thestatistical methods. To analysethe impact of environmental and spatial factors on the spatial pattern ofwinter extinction, Julyabundanceand the proportion ofinfected leaves with restingstructures, wefitted a Bayesian spatial model usingthe integrated nested Laplaceapproximation (Cameletti et al., 2012) as implemented in the packageINLA (Rueet al., 2009; Lindgren et al., 2011)in Rversion 2.15.1 (R CoreTeam, 2012). Theadvantageof this method is that it efficientlyand accuratelyestimates both covariates and the spatial rangeof autocorrelation (as based on Euclidean distancebetween populations). For both overwinteringsurvival and restingstructureformation, weincluded the environmental variables distanceto shore, plant dryness, patch shadow, habitat openness, July rainfall, August rainfall and population age(i.e. how manyyears ago the pathogen population had been established bycolonization, with a maximum valueof 5)and thespatial factors host plant coverage, road presenceand host plant spatial connectivityas explanatorycovariates. Theaverage rainfall in Julyand August was estimated separatelyforeach population usingdetailed radar- measured rainfall data. To reducethenumberof covariates in themodel wepre-selected the covariates to beincluded usingalinear / logistic model and thefunction stepAIC with theoption ‘backwards’ (packageMASS). Significanceof the explanatoryvariables was then assessed based on the devianceinformation criterion (DIC) in thespatial Bayesian model. To analysethe experimental data, weused theframework of generalized linear mixed-effects models (Littell et al., 2006). All models werefitted with procedureGLIMMIX in SAS 9.3. For binomial data, weassumed abinomial distribution with alogit link. For models with multiple interactions, weused the principle ofbackwards stepwise model simplification to arriveat aminimum adequate model, wherevariables were retained when p<0.1 (Crawley, 2007). Significanceforfixed and random effects was assessed usingF-tests and log-likelihood ratio tests, respectively(Littell et al., 2006). In the overwinteringexperiment weaimed to investigate theimpact of pathogen population of origin and overwinteringsiteon overwinteringsuccess. To analysethis experiment, wemodelled the response variables Infection (0/1)and Proportion of infected leaves (numberof infected / total leaves) as afunction ofthe fixed variable ‘Number of resting structures’ (representingthe impact of restingstructurequantity) and the random variables ‘Pathogen population of origin’, ‘Overwintering site’, and theirinteraction. Wefurther added thevariable Micro-site(nested within ‘Overwintering site’)to account formicro-environmental variation duringthe off-season and ‘Plant individual’ (nested within ‘Pathogen population of origin’) to account forvariation among pathogens collected from different plants within thesame population. Finally, weadded thevariable ‘Receiving plant genotype’ to represent theimpact of receivinghost plant genotypeduringinfection 14 in spring. When the interaction between ‘Pathogen population of origin’ and ‘Overwintering site’ was significant (as it was fordiseaseintensity, see Results), weprobed foraglobal pattern of local adaptation usingtwo alternativemodels. In model 1, weconstructed asimilar model as above, wherewespecified all factors as fixed variables. Wethen used aleast-squares means contrast to test forglobal local adaptation (bycontrastingleast-squares means forsympatric[local] and allopatric [non-local] pathogen population of origin / overwinteringsitecombinations). In model 2, we modified the original model byincludingthe fixed variable ‘Sympatry’ (and removingthe interaction term between ‘Overwintering site’ and ‘Population of origin’). Theterm ‘Sympatry’(0/1) would capturevariation in diseaseintensityresultingfrom infection byrestingstructures that were stored in sympatryor allopatry(i.e. restingstructures that wereoverwintered in thelocation from which theywerecollected or in anon-local location, respectively). Finally, weinvestigated therelationship between thepresenceof restingstructures and off- season survival at two spatial scales. At theplant level, infection (0/1) and diseaseintensity (number ofinfected leaves / total numberof leaves) in springweremodelled as afunction ofthe numberof leaves with restingstructures. Thepathogen population was used as random factor to account for variation amongpopulations in overwinteringsuccess. At thepatch level, off-season survival and Julyabundanceweremodelled as afunction ofthe fraction of infected leaves with restingstructures in theprevious autumn.