Seven M. truncatula lines derived from natural accessions of distinct geographic origins around the Mediterranean basin ( Supplementary Table 1A ) and three varieties of M. sativa adapted for the Mediterranean region ( Supplementary Table 1B ) were used. These plant genotypes have been chosen also for their previously reported contrasted response towards V. alfalfae strain V31-2 (Molinéro-Demilly et al., 2007; Ben et al., 2013; Negahi et al., 2013). Seed germination, and plant culture in peat substrate (Jiffy-7, Jiffy International AS, Norway) were performed as described by Ben et al. (2013). The growth chamber conditions were 16 h of light (at 170 μmol m^–2 s^–1) at 25°C and 8 h of darkness at 23°C for one week before inoculation. Plants were inoculated with Sinorhizobium meliloti (strain RCR2011) two days after germination, by adding 200µl of a suspension at 10⁷ Colony Forming Unit.mL^-1 at the basis of the stem.

Twelve Verticillium strains from various host plants and geographic origins sampling different temperature zones, were used ( Supplementary Table 2 ), and were grown on Potato Dextrose Agar (PDA) medium at 24°C in the dark.

We selected hot-adapted strains using UV mutagenesis of the V. alfalfae V31-2 strain (Buxton and Hastie, 1962). Conidia were exposed to UV radiation (254 nm) during 12 seconds (LD50 previously determined, Supplementary Figure S1 ), resulting mycelium was grown at 24°C for two weeks and the newly formed conidia were used to inoculate the susceptible M. truncatula line F83005.5 at 28°C. The inoculated plants were maintained at 28°C and the symptoms were scored during 4 weeks. The disease severity was rated using a scale from 0 (no symptoms) to 4 (dead plants), as described in Figure 1 of Ben et al. (2013). The fungus was re-isolated from aerial parts of plants presenting symptom scores of ≥ 2. Re-isolated mycelium was incubated at 24°C to produce conidia, which were again submitted to a round of mutagenesis, conidia production, plant inoculation and re-isolation. This mutagenesis/selection/re-isolation cycle was repeated three times to increase selection intensity. After the third round, 180 randomly selected mono-conidium strains were isolated. Among these, strain AS38, which presented good growth, sporulation capacity and pathogenicity at 28°C was retained as a “warm temperature-adapted” strain.

Ten day-old plants were inoculated with a fungal spore suspension at a concentration of 10⁶ spores per ml, as described by Ben et al. (2013). Verticillium- and mock-inoculated plants were transferred into growth chambers with a photoperiod of 16 h and three different day/night regimes: 20°C/20°C, 25°C/23°C and 28°C/26°C. Individual plants’ symptoms were scored regularly during four weeks, and disease severity was rated as described in Ben et al. (2013). Eight plants per M. truncatula line were used for each condition (combinations of temperature regime and pathogen strains), and 100 plants for each M. sativa variety (only scoring dead plants). Three independent replicates of the whole design were done. We considered plants as fully resistant when pathogen growth in the xylem was very limited and few, if any, symptoms occurred. Susceptibility is defined as the state in which the fungus proliferates systemically throughout the plant leading to symptom expression and disease development. Partial resistance represents intermediate symptom expression. A plant is disease tolerant if extensive Verticillium colonization occurs systemically, but few, if any, symptoms are expressed (Beckman and Roberts, 1995; Gao et al., 1995; Roy and Kirchner, 2000).

Plugs of mycelium (diameter 5mm) from the periphery of growing cultures were placed on fresh PDA plates and the Petri dishes were incubated in the dark at 20°C, 25°C, 28°C and 30°C. Two Petri dishes were made for each fungal strain and condition, and three independent experiments were performed. Radial growth was measured on two perpendicular radii for each culture, every second day during two weeks, and the mean values for radius were calculated. Spore suspensions were obtained after two weeks of culture by flooding the Petri dishes with 20 ml of sterile water. Their concentration was determined with a hemocytometer.

Ten days after inoculation (i.e. onset of wilting symptoms in the most susceptible M. truncatula lines), the first unifoliate leaf was harvested and DNA was extracted as described by Ghérardi et al. (1998). Verticillium DNA was amplified using vert853F and vert927R primers described by Larsen et al. (2007), and M. truncatula DNA was amplified with actin primers (Ariel et al., 2010). Reactions were carried out in two replicates per sample in a 384-well plate containing 5µl of KAPA SYBR Fast ABI Prism Readymix Kit (Clinisciences, France), 2 µl primers (0.35µM) and using respectively 50ng/µl and 150ng/µl of DNA matrix for Medicago and Verticillium amplifications in 10 μl total reaction volume. Relative fungal biomass was estimated as the ratio of fungal DNA per unit of plant DNA, each determined by their cycle threshold, Ct (ΔCt = Ct Verticillium DNA – Ct M. truncatula DNA). Pools of eight plants were analysed in that way per M. truncatula x temperature condition, in two independent experiments.

Thirty days after inoculation, 2-cm-long fragments of the main stem was cut above the first node, and was successively surface-sterilized in 70% ethanol for 15 seconds and in 0.96% commercial bleach for 6 minutes then rinsed thrice in sterile water for 1 minute. The stem sections were then incubated on PDA medium containing 50mg/ml streptomycin during three days at 25°C. The number of stem ends showing outgrowth of white mycelium was determined and percentage of re-isolation was calculated. This experiment was repeated three times independently. Presence of the fungus in reproductive parts (podes and seeds) was checked by disinfecting whole pods and seeds in the same way. Seeds were scarified with a sterile scalpel. Emptied pods and seeds were incubated on medium as described above.

After four weeks of symptom scoring in phytotron at 20°C, 25°C and 28°C, six plants (three mock-inoculated and three Verticillium-inoculated) were chosen randomly for each M. truncatula line and each temperature, and transferred into the greenhouse for assessment of fitness in response to strain V31-2. Each plant was potted in soil-sand (2:1) mixture in 2L pots. This experiment was repeated twice in two following years, with three blocks per replicate. After two months of growth, each plant was wrapped in a net in order to recover all pods of an individual plant. Plants were harvested six months after transfer into the greenhouse. Number of pods, pod weight and aerial dry biomass per plant, as well as germination capacity of seeds were measured to assess plant fitness.

Area Under the Disease Progress Curve and Area Under the Growth Progress Curve (AUDPC and AUGPC, respectively, Shaner and Finney, 1977) were computed based on the disease scores and fungal radial growth respectively, using the ‘agricolae’ R package (De Mendiburu and Muhammad Yaseen, 2020; R Core team, 2022). For sporulation, data were analyzed using generalized linear model for counts data using Poisson regression (glm function of R). For Verticillium mutagenesis, LD50 was determined after fitting a generalized linear model for counts data as a function of time of exposition to UV light, with a binomial link function (glm function of R). For disease symptom scores (disease severity), data were analyzed using proportional-odds models (McCullagh, 1980; Agresti, 2012), using the orm function of the ‘rms’ R package (Harrell, 2015). The aim is to relate the dependence of the ordinal response (disease symptom scores) on discrete (pathogen strains and plant lines) or continuous (temperature) covariates. Effects of covariates were tested using analysis of deviance. Multiple comparisons of odd-ratio for symptom scores were done using least square means and Bonferonni correction using the ‘lsmeans’ R package (Lenth, 2016) after fitting a proportional-odds model for each temperature level. For disease incidence in M. sativa (percentage of dead plants at 28 dpi), data were analyzed using a generalized linear model for proportions using the glm function with the logit link. Fitness experiment in common garden was analyzed with ANOVA in a Randomized Complete Block Design model. When required, data transformations were applied to achieve normality and homoscedasticity of ANOVA residuals. Pairwise treatment differences were determined by a Tukey’s test using package ‘lsmeans’ or Newman-Keuls test using the ‘agricolae’ package. Graphs were drawn using the ‘ggplot2’ package (Wickham, 2009).

Verticillium strains from different geographic origins and host plants were screened for their optimal temperature in order to identify strains adapted to cool, temperate and warm climate. In vitro growth parameters were analysed at 20, 25, and 28°C. Hyphal growth as calculated by AUGPC showed that the 12 strains differed in temperature optima and in their range of tolerance to high temperatures ( Figure 1A ). Statistical analysis demonstrates highly significant strain × temperature interactions with P<0.001 ( Supplementary Table 3 ). The optimal temperature for hyphal growth was 25°C except for strain Vdp5 (potato isolate from Israel) which grew best at 28°C and strain VA1 (potato isolate from Canada) which grew best at 20°C. Sporulation of these strains, determined after two weeks of growth, was also affected by temperature ( Figure 1B and Supplementary Table 4 ). The temperature optimum for sporulation was shown to be 25°C for eight strains and 20°C for the four remaining strains (namely Vdp1, Vdp5, Vd89 and VA1); none sporulated better at 28°C.

To test for optimal temperature of disease development, these 12 Verticillium strains were inoculated on four M. truncatula lines (A17, DZA45.5, F83005.5 and DZA315.16, Figure 2 ), at different day-night temperature regimes. A17 and DZA45.5 were previously described as resistant, and F83005.5 and DZA315.16 as susceptible to V31-2 at 20°C (Ben et al., 2013). The results show that Verticillium isolates from other host plants, e.g. V. dahliae V200 from cotton and V. non-alfalfae VA1 from potato, are able to cause disease on M. truncatula. Statistical analysis of the disease score at the end of the experiment (herein called Maximum Symptom Score, MSS) using proportional-odds models reveals a highly significant line × strain × temperature interaction, indicating that plant’s response to the pathogen strain depends on the temperature. (P<0.0001, Supplementary Table 5 ). In susceptible genotypes, the optimum temperature for disease induction was 20°C or 25°C depending on the fungal strain. The aggressiveness of strain V31-2 on the susceptible M. truncatula lines, and that of most other strains was highest at 25°C and 20°C, and lowest at 28°C. Strain Vdp5 which was well adapted to high temperature for in vitro growth, was weakly pathogenic on M. truncatula susceptible accessions ( Supplementary Table 6 ). Results obtained with the AUDPC parameter were similar and did not allow identifying a Verticillium strain adapted to higher temperature (data not shown). Based on temperature optima for the three parameters in vitro growth, sporulation capacity and aggressiveness, none of these 12 strains seemed to be truly adapted to higher temperatures, to be used for studies of plant responses in a scenario of climate change.

V. alfalfae (previously named V. albo-atrum ‘alfalfa strains’) has been reported to survive in the Mojave Desert (Erwin et al., 1988; Erwin and Howell, 1998) at temperature higher than 35°C. However, these strains are no longer available (A. Howell, personal communication). Hence we created a strain adapted to 28°C by successive cycles of UV mutagenesis, plant inoculation at 28°C and fungus re-isolation. As can be seen in Figure 3 , the aggressiveness of mutated V. alfalfae increased with each cycle of inoculation/reisolation on the susceptible line F83005.5. At the end of this accelerated evolution, 86% of plants inoculated with the V. alfalfae mutants presented symptom scores of 4 (dead plants), whereas with the wild type strain the maximum score observed was 2 at 28°C and 60% of the plants were symptomless. Mono-conidium isolates were obtained from mutants after 3 rounds of mutagenesis. One-hundred eighty of these mono-conidium isolates from the collection were inoculated on M. truncatula line A17 (resistant to V31-2) and F83005.5 (susceptible to V31-2) at 28°C. This preliminary experiment revealed that all isolates showed adaptation to high temperature ( Supplementary Figure S2 ), and that some of them seemed to have acquired the ability to cause mild symptoms on line A17, such as AS11, AS38 or AS90. The mono-conidium isolate named AS38 which showed similar discrimination between the two genotypes as the wild type was retained for further studies. Despite its infectious adaptation to higher temperatures, optimal temperature for in vitro growth and sporulation of this mutant was still 25°C ( Supplementary Figure S3 ).

To get a detailed evaluation of the impact of temperature increase on the plant-pathogen interaction, the two V. alfalfae strains (V31-2 and its derived mutant AS38) were inoculated at three temperature regimes (20/20°C, 25/23°C, 28/26°C) on a set of seven M. truncatula lines that have been observed to show contrasted behavior with strain V31-2 at 20°C (Ben et al., 2013; Negahi et al., 2013 and this study). Statistical analysis of MSS using proportional-odds model showed a clear effect of temperature on disease severity with significant interactions with strains and plant accessions ( Table 1 ). This confirms that the interaction of the plants with the wild strain V31-2 and with its warm-temperature-adapted mutant AS38 depends on temperature.

Strain V31-2, considered to be adapted to a temperate climate, was most aggressive at 25°C, with MSS ≥ 3 on the three susceptible lines F83005.5, DZA315.16 and SA3780 ( Figure 4A ); but at 28°C these values decreased. The remaining lines A17, DZA45.5, and PRT180-A were resistant and the line SA9048 was partially resistant. The mutant strain AS38 which has been adapted to high temperatures was more aggressive than V31-2, notably at 28°C, and induced symptoms in the three lines susceptible to strain V31-2 (DZA315.16, F83005.5 and SA3780) and in addition in line SA9048, with MSS >3.5; the remaining three lines A17, DZA45.5 and PRT180-A were resistant ( Figure 4B ). At 20°C, the disease severity of V31-2 and AS38 is not significantly different ( Figure 4C ). At 25°C, the line SA9048 was significantly more affected with AS38 compared to V31-2. At 28°C, three lines exhibited a pattern of increased disease severity: SA9048, F83005.5, DZA315.16 ( Figure 4C and Supplementary Table 7 ). In all cases, the line SA3780 is severely affected, with non-significant but noticeable increase of MSS when infected by AS38. MSS and AUDPC were positively correlated and AUDPC analysis revealed a similar pattern of variation (data not shown).

In order to evaluate if plant colonization capacity was modified in mutant AS38 compared to its wild type, and the putative influence of temperature, colonization of aerial parts by the fungus was assessed by microbiological re-isolation, and by quantification of fungal DNA by qPCR, in four lines inoculated by strain V31-2, in two independent experiments. Supplementary Figure S4 represents fungal colonization quantified by qPCR as ΔCt versus quantification by re-isolation rates. ΔCt values were significantly and negatively correlated with colonization estimated by re-isolation (r = 0.79, P<0.01), at the three temperatures, i.e. less fungal DNA was detected (high ΔCt) when tissue colonisation was low. qPCR analysis is a non-destructive test which at an early stage (10 dpi) allows predicting plant colonization. Yet, as re-isolation has both the advantage of revealing alive pathogen and being more practicable, we systematically performed fungal re-isolation to assess plant response and pathogen fitness for all combinations of line × strain × temperature ( Table 2 ). Analysis of deviance indicates very significant effect of line × strain × temperature interaction on re-isolation percentage ( Supplementary Table 8 ), and shows that in all cases (except A17), re-isolation percentage at 28°C is significantly greater with AS38 compared to V31-2 strain. This suggests an improved fitness of the hot-adapted AS38 mutant at 28°C.

Combining MSS data and plant colonization assessment allows a more detailed characterization of QDR at the different temperatures in response to the two pathogen strains. Joint representation of MSS data and re-isolation percentages of strain V31-2 from inoculated plants confirmed the classification of susceptible (DZA315.16, F83005.5 and SA3780) and resistant (A17, DZA45,5, SA9048 and PRT180-A) lines ( Figure 5A ), with some variability in the colonization percentage among lines exhibiting low symptoms. Notably, the colonization of susceptible plants was always highest at 25°C and lowest at 28°C. Among the resistant lines, PRT180-A showed a noticeable colonization (percentage of 20%) at 28°C. For the hot-adapted mutant strain AS38 ( Figure 5B ), it appeared that lines DZA315.16, F83005.5, SA9048 and SA3780 were susceptible, with high colonization at 28°C. Significant differences in the plant responses to the wild type strain and its temperature-adapted mutant were observed: Line SA9048, which was partially resistant to V31-2 was highly susceptible to the mutant AS38 at 25°C and 28°C, and moderately susceptible at 20°C. Line DZA45.5 which was resistant to V31-2, could be considered as disease tolerant to AS38 at 25 and 28°C, since it was significantly colonized with highest colonization rate at 28°C despite low symptom scores (MSS<1.5). With the noticeable exception of A17, the resistant lines all exhibited higher colonization rates when infected with AS38, in particular at 28°C. Incidentally, these results also show that disease symptoms are always associated with plant colonization by the fungus, and thus not triggered by a toxin release by the fungus without plant colonization (Palmer et al., 2005).

As a first step to analyze the significance of the plants’ response to Verticillium at different temperatures, the fitness of plants inoculated with V31-2, and corresponding controls, at the three temperature regimes was studied. After four weeks, plants were transferred into the greenhouse where they continued to grow all under the same conditions. Pod number per plant, pod weight per plant and dry biomass of aerial parts per plant were assessed as proxy for fitness after six months. The germination rate of progeny seeds was 100% in all M. truncatula lines and for all conditions. Furthermore, the number of seeds per pod was not significantly different between lines and between infected and non-infected plants with an overall average number of five seeds per pods (data not shown). Hence the number of pods per plant seemed to be an accurate measure of total lifetime fitness.

Statistical analysis of the three fitness parameters showed that the effect of temperature observed during the first phase had disappeared ( Tables 3A–C ). However, a significant interaction line × treatment was observed, as expected for differences between susceptible and resistant lines in a compatible pathogenic interaction. Pod number was strongly affected in the susceptible lines F83005.5, DZA315.16 and SA3780, and moderately in the partially resistant line SA9048. This effect was not significant for the three remaining resistant lines A17, PRT180-A and DZA45.5 ( Figure 6A and Supplementary Table 9 ). Pod weight and aerial biomass were significantly decreased by inoculation in the three susceptible lines. ( Figures 6B, C and Supplementary Tables 10 , 11 ). As a general tendency, the fitness consequences of infection by V31-2 were always independent from the temperature overcame during the first month i.e. at early stages of infection.

To test if the disease can spread via plants’ reproductive structures, Verticillium re-isolation experiments from pods and seeds from inoculated plants showed that the fungus was present inside the pods, but not inside the seeds, of susceptible lines, and absent from pods and seeds of resistant or partially resistant lines (data not shown), whatever the temperature.

The effect of temperature increase on Verticillium wilt and of potentially evolving new strains was studied in alfalfa (M. sativa), a cultivated crop phylogenetically close relative from the wild species M. truncatula. Three commercialized varieties were inoculated with the two Verticillium strains V31-2 and AS38, and their susceptibility was assessed by mortality rate at the end of the experiment (four weeks after inoculation). The combined analysis of deviance showed a very significant strain × temperature and strain × line interactions, but no line × temperature interaction, indicating that differences in disease response are not due to a differential response of the lines to temperature increase ( Table 4 ). When the alfalfa cultivars were inoculated with strain V31-2 cultivar Magali was the most susceptible with 80% mortality, whereas Lifeuil and Prunelle were partially resistant with not more than 30% mortality ( Figure 7 ). The highest mortality was observed at 25°C, and the lowest at 28°C ( Table 4 ). When the alfalfa cultivars were inoculated with strain AS38, cultivar Magali was the most susceptible with 60% and 65% mortality at 25°C and 28°C respectively, whereas Lifeuil and Prunelle were partially resistant with not more than 40% mortality ( Figure 7 ). The highest mortality was observed at 25°C and 28°C, and the lowest at 20°C. These results confirm that temperature is a significant determinant of Verticillium aggressiveness towards M. sativa. They also demonstrate that AS38, a pathogenic strain which was initially adapted on M. truncatula for higher temperatures, is able to infect another species M. sativa with the same optimum temperature for disease onset (28°C).