The wild tomato species Solanum chilense shows variation in pathogen resistance between geographically distinct populations

Wild tomatoes are a valuable source of disease resistance germplasm for tomato (Solanum lycopersicum) breeders. Many species are known to possess a certain degree of resistance against certain pathogens; however, evolution of resistance traits is yet poorly understood. For some species, like Solanum chilense, both differences in habitat and within species genetic diversity are very large. Here we aim to investigate the occurrence of spatially heterogeneous coevolutionary pressures between populations of S. chilense. We investigate the phenotypic differences in disease resistance within S. chilense against three common tomato pathogens (Alternaria solani, Phytophthora infestans and a Fusarium sp.) and confirm high degrees of variability in resistance properties between selected populations. Using generalised linear mixed models, we show that disease resistance does not follow the known demographic patterns of the species. Models with up to five available climatic and geographic variables are required to best describe resistance differences, confirming the complexity of factors involved in local resistance variation. We confirm that within S. chilense, resistance properties against various pathogens show a mosaic pattern and do not follow environmental patterns, indicating the strength of local pathogen pressures. Our study can form the basis for further investigations of the genetic traits involved.

231 Over the course of several months, we performed four biological replicates for each 232 population, , accumulating to about 450 -500 infected leafletsper population per 233 pathogen. The Alternaria infections were done on the adaxial side of the leaves, 234 Phytophthora and Fusarium infections were done on the abaxial side of the leaves. The 235 leaves were incubated at RT and scored after six to eight days, dependent seasonal 236 variability of temperature and light conditions in the lab, which slightly affected the rate 237 of symptom development..

239 Data analysis
240 All data analysis was done using R (version 3.2.3) (R foundation for statistical 241 computing). Generalised Linear Mixed Models (GLMM) were made using the glmer 242 option from the package lme4 (Bates et al., 2015, p. 4). To construct GLMM we used a 243 binomial variable (y) consisting of the number of successful and unsuccessful infection 244 events per leaf. The GLMM were constructed taking the leaf position in the box (leaf) 245 and a combination of the box number and experimental date (exp:box) into account as 246 random effects. For our first model populations names were used as fixed effects.
247 (model1 = y ~ accession +(1|leaf)+(1|exp:box)). For the next models, we hierarchically 248 tested different climatic and geographical parameters (e.g. model2 = y ~ geographic1 + 249 climatic1 + climatic2 + (1|leaf)+(1|exp:box)). Pairwise comparisons were examined 250 using an implementation of Tukey Honest Significant Difference test as provided by 251 function glht from the R package multcomp (Hothorn, Bretz & Westfall, 2008). glht 252 allows post-hoc hypothesis testing, similar to THSD, but is more suitable for general 253 linearised (mixed) models. The boxplots were drawn using the package ggplot2 254 (Wickham, 2009, p. 2 266 chilense population with Alternaria solani (st108) and counted the occurrence of 267 infected leaflets per leaf, as this represents the success rate of the pathogen to 268 establish itself and overcome genetic resistance. We scored infection events as either 269 negative (no infection or clear small necrotic lesions, indicating a hypersensitive 270 response) or positive (ranging from growth just outside the droplet area up to full 271 infection of the leaflet) ( Fig 1B)(All raw data can be found in S Data 3). We observed 272 variation within each population. In almost all instances at least one leaf was fully 273 infected whereas another was completely resistant. These outliers have large effect on 274 the calculated mean fraction. To allow good judgement we report the 1 st and 3 rd 275 quantile, the median value as well as the mean value for each population (Fig 2). The 283 With Fusarium we also saw differences in the infected fraction amongst populations.
286 Finally, for P infestans, the infected fractions again showed a different pattern. There 287 was a larger spread of the data as can be seen by the increased distance between the 288 1 st and 3 rd quartile and the lowest and highest mean and median fraction were closer 289 together, ranging from 0.30 and 0.21 for LA3111 to 0.60 and 0.70 for LA4330 ( Fig 2C). Manuscript to be reviewed 299 some populations for all three pathogens tested (S Data 2).

301 Pairwise comparisons reveal individual differences between different pathogens
302 To further determine which populations were different from each other, we performed 303 pairwise comparisons using a variant of Tukey's Honest Significant Difference test (see 304 methods). The observed pairwise differences were clearly distinct between the three 305 pathogens. Figure 3 shows a summary of the pairwise differences, with corresponding 306 difference estimates for each comparison. Cells with significant differences (p< 0.001) 307 are highlighted in green. All pairwise differences with their 95% confidence intervals are 308 plotted in S Data 5. Of the 63 pairwise comparisons, 32 showed a significant difference 309 in infection ratio. Overall, there were more significant differences between populations 310 when it came to Fusarium infection (15) than to Alternaria infection (10) or Phytophthora 311 (7). Interestingly, some populations showed the same result for all pathogens: there are 312 no differences between LA1963 and LA2931 (both central) nor for LA2931 and LA4107 313 (central and south coast) or LA4107 and LA4117 (south coast and south mountain).
314 Also, LA1963 was always more susceptible than LA2932, and LA4117 was always 315 more susceptible than LA4330. In some cases a population in a pair was more resistant 316 to one pathogen and more susceptible to another. LA4330 was more resistant than 317 LA3111 to Fusarium, but less resistant to Alternaria and Phytophthora. 318 319 320 A mix of climatic and geographic variables affect pathogen resistance 321 To see whether a change in certain geographic and climatic conditions could be linked 322 to an increase or decrease of resistance rates between populations, we built new 323 GLMM using such data. First we made a simple model for resistance to Alternaria, 324 testing the infection counts (y) against either latitude or longitude, a combination of both 325 or an interaction of both. This showed that both latitude and longitude had a significant 326 effect (p < 0.001, Table 1). The quality of the models is reported by the statistical 327 software using the Akaike Information Criterion (AIC), where a lower AIC, indicates a 328 better relative quality of the model. A model with both parameters showed a better AIC.
329 However, a model with an interaction only shows significance of the latitude parameter.
330 We extended the model to include both parameters (longitude + latitude) as well as 331 several environmental parameters. We obtained the best AIC (2641.8) for a model 332 containing altitude, annual precipitation, the temperature in the wettest and the 333 temperature in the coldest quarter. Additions of other climatic data did not yield an 334 improvement of the model. A selection of tested models with their AIC and significance 335 is shown in Table 1. When closely investigating the best fitting model for resistance to 336 Alternaria (Table 1, model 6), we noted that whereas all variables contribute significantly 337 to the model, the estimated effect size differs greatly. Table 2 shows that of all effects, 338 longitude was the strongest effect, followed by the mean minimum temperature in winter 339 (TempB), the annual precipitation and altitude. It should be noted that models that only 340 take temperature effects into account did not account for significance (Table 1) 347 Similar to resistance to Alternaria, we tested all variables for resistance to Phytophthora 348 and resistance to Fusarium. The pattern seen for resistance to Phytophthora is almost 349 identical to that of Alternaria resistance. The AIC values are generally lower, but the 350 trends are the same. Interestingly, Fusarium resistance showed a different picture.  Manuscript to be reviewed 368 exchange between some of the southern most populations that are separated by the 369 extremely dry Atacama desert. This leads to the conclusion that S. chilense can be 370 divided in a northern, a central and two southern genotype groups . 447 These findings are in line with several inter-species studies in wild potato, where no 448 correlation could be found between geographical location of the species and resistance 449 against P. infestans (Khiutti et al., 2015) or A. solani .    Manuscript to be reviewed   Manuscript to be reviewed  Table 1 Summary of GLMM results. AIC is a measure of relative quality of the model, with a lower AIC (within one species) indicating a better model. AICs in bold represent models that are significant (p < 0.001). TempA denotes the temperature in the wettest quarter, TempB in the coldest quarter and TempC the annual mean temperature. Manuscript to be reviewed