We selected four V. vinifera L. cvs based on their different susceptibility levels to esca disease (Feliciano et al., 2004; Bruez et al., 2013; Travadon et al., 2013; Murolo and Romanazzi, 2014). One year-old dormant grape cuttings cvs Merlot (M), Chardonnay (Ch), Cabernet Sauvignon (CS), and Thompson Seedless (TS) were provided by the Foundation Plant Services (Supplementary Table 1). Cuttings were prepared as described by Pouzoulet et al. (2013b) and potted in 4 l pots containing UC-mix soil. Four weeks after potting, homogeneous plants were selected on the basis of height and stem diameter. Inoculations were performed in the dorsal or ventral part of the stem using the drill method. Inoculum was prepared as described by Travadon et al. (2013) using P. chlamydospora voucher isolate UCR-Pc4 (Morales-Cruz et al., 2015). Inoculum consisted in one 20 μl drop of sterile 10 mM phosphate buffer (20 μm filtered, pH = 7) containing a total of 1000 P. chlamydospora conidia. Control plants received 20 μl drop of 10 mM phosphate buffer. Plants (cultivars and treatments) were organized randomly in a glasshouse (temperature ranged from 21 to 32°C) and watered three times a week. A total of 128 plants were used per year with 32 plants per cultivar (16 inoculated and 16 control plants). Half of the plants (8 inoculated and 8 control plants) were used for the measurement of necrotic lesions and the other half was used for the quantification of fungal DNA in wood fragment above the inoculation site. The experiment was repeated in 2013 and 2014.

Fungal colonization in planta was evaluated 10 weeks post-inoculation by measuring necrotic lesion length in the xylem and quantifying the amounts of fungal DNA using a qPCR assay. For the measurement of lesions, stems were longitudinally cut using a scalpel in a plane striking across the wound. Pictures of lesions developed above the inoculation site were taken using a stereo-microscope (M165C, Leica microsystems CMS GmbH, Wetzlar, Germany) and measured using LAS v4.2 software (Leica microsystems CMS GmbH, Wetzlar, Germany). For the quantification of P. chlamydospora DNA in xylem fragments, stem samples from the wounded non-inoculated control and P. chlamydospora inoculated plants were frozen and subjected to 48 h of lyophilization using Labconco freezone 2.5 l (Kansas City, MO, United States). Samples were cut longitudinally in the plan crossing the bud in order to recover only the half part of stems carrying the infection. For P. chlamydospora inoculated plants, two fragment lengths were cut above the inoculated point, the first one from point of inoculation up to 15 mm (level 1: L1) and the second one 15–30 mm (level 2: L2) distant from the inoculation point. Bark and pith were removed from the wood sections using a sterile scalpel. Samples were ground with a mixer mill (MM 400, Retsch GmbH, Haan, Germany) and DNA was extracted as described previously (Pouzoulet et al., 2013b). Concentration of total DNA extracted was determined using a Qubit^TM fluorometer (Invitrogen, Carlsbad, CA, United States) and the Quant-it dsDNA high-sensitivity reagent (Invitrogen) according to the manufacturer protocol. The qPCR reactions proceeded in a final volume of 25 μl, and reaction mixtures contained 12.5 μl of 2X SYBR^® Green Quantitect^® Master Mix (Qiagen, Venlo, Netherlands). Primers PchQF (5′-CTCTGGTGTGTAAGTTCAATCGACTC-3′)/PchQR (5′-CCATTGTAGCTGTTCCAGATCAG-3′) were used at a final concentration of 0.5 μM. Two μl of DNA template were used per reaction. Experiments were conducted with a CFX96 Real-Time PCR cycler using CFX manager software v3.1 (Bio-Rad, Irvine, CA, United States). The cycling program consisted of (1) an initial denaturation step at 95°C for 15 min, (2) 40 cycles of 15 s at 95°C (for denaturation) followed by 45 s at 62°C (for both annealing and extension), and (3) an additional melting analysis of 40 min from 60 to 95°C. Preparation and use of standard solutions for the absolute quantification of P. chlamydospora isolate UCR-Pc4 DNA was done as previously described (Pouzoulet et al., 2013b). Average of absolute amounts of P. chlamydospora DNA determined by qPCR in three independent technical replicates were standardized on the amounts of input DNA and used for further statistical analysis. The qPCR values observed in levels L1 and L2, cultivars and years were subjected to an ANOVA using SAS Proc Mixed (version 9.4; SAS Institute, Cary, NC, United States) followed by multiple comparisons of means using Fisher’s protected least significant difference test.

An automated detection analysis of vessels and measurement of their arithmetic diameter were performed on stem cross-sections as described by Scholz et al. (2013). Six internode segments (three stems from two different mother-vines) of 8 to 10 mm of diameter and 100 to 120 mm in length were sampled within cutting bundles of each cultivar (Supplementary Table 1). A 10 mm long fragment was sampled in the middle part of the each internode. Samples were fixed in ethyl-alcohol 80% at 4°C during 48 h, after which ethyl-alcohol solution was replaced and samples were kept at 4°C. High definition micrographs (430 pixels mm^-1) covering a quarter of each Toluidine O stained stem cross sections (70 μm thick) were obtained as described by Pouzoulet et al. (2014). For each stems, number of vessels and vessel density (vessel count mm^-2) were determined from six consecutive fascicular portions within the dorsal or ventral area using ImageJ v1.48¹. Arithmetic vessel diameter was computed according to Scholz et al. (2013) and vessel distribution per class of diameter was determined using Excel 2010 (Microsoft Corporation, Redmond, WA, United States). The density of vessels for different classes of diameter (20 μm steps) observed in the four cultivars studied was subjected to an ANOVA using SAS Proc Mixed (version 9.4; SAS Institute, Cary, NC, United States) followed by multiple comparisons of means using Fisher’s protected least significant difference test.

Shoots of 1 year-old grapevine cv CS were potted in 8 l container and grown in glasshouse for 2 months. Soil was watered to saturation every 2 days during the period of the experiment. Plants developed 4 m long shoots with a basal diameter of 8 to 10 mm. Shoots were wounded in dorsal area using a drill method (2 mm of diameter) at the third node. Five plants were inoculated with a 20 μm filtered solution of phosphate buffer (pH = 7), and the five plants were inoculated with a suspension of P. chlamydospora spore as described previously. Eight weeks after treatment, internode fragments above the inoculation wound were sampled and fixed in FAA solution (Formaldehyde, Acetic acid, Ethyl Alcohol; 5/5/9) for 48 h at 4°C, then rinsed and conserved in 80% Ethyl alcohol at 4°C. Sections (70 μm thick) were obtained as previously described. Staining of specimen with Toluidine O (0.05%, pH = 4.3, Sigma–Aldrich, St Louis, MO, United States), IKI (potassium iodide-iodine, Ricca Chemical Company, Arlington, TX, United States), ruthenium red (0.005%, Sigma–Aldrich), Phloroglucinol/HCl (Sigma–Aldrich), and Sudan IV (0.01% in 70% ethyl-alcohol, Sigma–Aldrich) were adapted from Ruzin (1999). Light microscopy works were done using a Leica DM 4000 upright microscope (Leica Microsystems CMS GmbH, Wetzlar, Germany). Epifluorescent micrographs were obtained using an inverted epifluorescent microscope (Arcturus XT, Life Technologies, Carlsbad, CA, United States) by overlaying micrographs from the ‘UV’ (excitation: 325–375 nm, emission: >420 nm) and ‘green’ filters (excitation: 510–560 nm, emission: >590 nm) with ImageJ v1.48. For the co-visualization of fungal hyphae and plant material by epifluorescent and confocal microscopy, plant sections were placed in a 2 ml plastic tube containing 1 M potassium hydroxide and incubated 1 h in boiling water. Sections were then rinsed three times during 10 min in deionized water. Sections were stained using an aqueous solution of Auramine O (0.1%, Sigma–Aldrich) for 5 min for the staining of plant material followed by 20 min of staining with Calcofluor M2R White (Sigma–Aldrich) for the labeling of the fungal cell wall. Sections were briefly rinsed and mounted in deionized water. Epifluorescent and confocal microscopy works were realized at the Microscopy Core/Center for Plant Cell Biology at the Institute for Integrative Genome Biology at the University of California, Riverside. Epifluorescence micrographs were obtained with an epifluorescent microscope Arcturus XT^TM using UV filter. Confocal micrographs were obtained using a Leica SP2 (Leica microsystems CMS GmbH, Wetzlar, Germany) equipped for UV microscopy. Two channels were used, one for the recovery of calcofluor M2R white related signal (excitation = 364 nm, detection = 410–505 nm, displayed in green), and one for the recovery of the Auramine O related signal (excitation = 488 nm, detection = 505–590 nm, displayed in magenta). Acquisition of channels was done in sequence. Acquisition of Z-stacks varying from 50 to 100 μm of depth were done using a water immersion objective (20×), micrographs being taken approximately every 1 μm. Z-projections of micrographs were done with ImageJ v1.48 using the standard deviation method. Both inoculations and microscopic observations were replicated twice on different sets of plants, and an interval of 6 weeks.

We evaluated the impact of xylem morphology on quantity of tyloses during vessel occlusion. Five cuttings from all four grapevine cvs were wounded and grown for 10 weeks as described previously. Xylem segments above the wound were sampled and fixed in FAA and 70 μm cross-sections of xylem region located 5 mm above the wound were obtained as described previously. Sections were dehydrated in 100% ethanol, dried at room temperature for 15 min and observed using a tabletop scanning electron microscope (Hitachi TM-1000, Tokyo, Japan). The arithmetic vessel diameter, the number of tyloses (N_tyl) and the outer tylosis surface length (OTSL) were measured with ImageJ v1.48 software. Preliminary analyses indicated that the relationships between the vessel diameter (D, μm) and the number of tylosis, and the vessel diameter and the OTSL (μm) was of non-linear nature, and could be described with an exponential curve (Eqn. 1):

where α and β were the model parameters. By applying a natural logarithm transformation to Eqn. 1 we obtained the following equation (Eqn. 2):

It was possible to parameterize the model with an ordinary least square (OLS) approach. Eqn. 2 was parameterized by Generalized Reduced Gradient optimization (Lasdon et al., 1978), using the Excel 2010 Solver tool (Frontline Systems, Incline Village, NV, United States). Eqn. 2 described the global (i.e., for all four cvs) relationships. The robustness of the global model was tested using a leave-one-group (i.e., cvs)-out cross-validation (Martens and Dardenne, 1998). Note that curves presented correspond to the back-transformed data.

Necrotic lesion length and fungal DNA amount in wood tissue were used as criteria to assess the level of resistance of grape cultivars to P. chlamydospora. There were no significant interactions between cultivar and year for the necrotic lesion length (Table 1) and data from both years were pooled together (Figure 1A). After 10 weeks of incubation, P. chlamydospora inoculated plants developed wood necrotic lesions significantly longer than their respective controls for the four cultivars tested. Post hoc analysis classified all mock-inoculated plants in a same statistical group, while inoculated plants were classified in different groups (Figure 1A). Among all four cultivars, wood necrotic lesions that developed in M were significantly smaller than the other cultivars and were classified on this basis as the most resistant to P. chlamydospora. In contrast, TS developed the longest necrotic lesions and was classified as the most susceptible cultivar. CS showed intermediate lesion length and differ significantly from both M and TS. Ch also showed an intermediate susceptibility and differed significantly from M, but not from either CS or TS (Figure 1A).

Phaeomoniella chlamydospora DNA amount was measured in wood tissue close (L1) and distant (L2) from the inoculation point (Figure 1B). Fungal DNA was consistently detected in xylem tissue the most distant from the inoculation point (L2) for all four cultivars despite the fact that these fragments were mostly asymptomatic (i.e., no wood necrotic lesions beyond 15 mm; Figure 1A). L1 and L2 were analyzed separately for statistical purposes. In addition, only values from inoculated plants were considered for its statistical analysis because no P. chlamydospora DNA was detected in control plants. For L1, statistical analysis of P. chlamydospora DNA amount detected a significant interaction between cultivar and year (Table 1 and Supplementary Table 2). Differences between the 2 years were observed with overall lower DNA amounts in 2014 vs. 2013. Statistical classification for grape cultivars was not consistent across the 2 years. For L2, statistical analyses of the P. chlamydospora DNA amounts also revealed an interaction between cultivar and year (Table 1 and Supplementary Table 3). As with L1, our data showed overall lower fungal DNA amount measured in wood for 2014 compared to 2013 but with a similar trend in both year, a better statistical segregation of cultivars being observed in 2013 than 2014 (Supplementary Table 3). Finally, L2 pooled data from 2013 and 2014 separated M, CS and TS in three groups of resistance, M being the most resistant, followed by CS and TS, the most susceptible, whereas Ch belonged to an intermediate group (Figure 1B). Observations from wood necrotic lesions and amounts of P. chlamydospora DNA in L2 were found to correlate significantly (R² = 0.60, p = 0.025, n = 8), thus supporting overall similar conclusion on the level of resistance of grape cultivars in our bioassays.

Wood morphology was analyzed on the dorsal or ventral area of grapevine stems, because this is where P. chlamydospora inoculations were performed. Vessel distributions of the plant materials were overall consistent between 2013 and 2014 (Figure 2). For the number of vessels per diameter class, ANOVA shows a significant effect of cultivars for many diameter classes (e.g., 80–99, 100–119, 140–159, 160–1879, and 180–199 μm), but also some interactions between cultivar and year (i.e., 60–79, d > 200; Supplementary Tables 4, 5). Our results indicated that M, the most resistant cv, displayed the highest number of narrow diameter vessels (i.e., from 80 to 119 μm), and the lowest number of wide diameter vessels (i.e., above 120 μm). In contrast TS, the most susceptible cv, displayed the highest number of wide diameter vessels (Figure 2). Ch and CS, of intermediate resistance, displayed an intermediate number of wide diameter vessels (Figure 2).

To confirm that cv resistance was related to xylem morphology, regressions were calculated between the density of vessel superior to a fixed diameter (e.g., d > 100, d > 120, ..., d > 200) and P. chlamydospora DNA amount (Table 2). We found significant correlations between the amount of P. chlamydospora DNA (Log_e transformed) from the L2 and the number of vessels superior to 120, 140, 160, and 180 μm diameter per mm² (Table 2). These results supported our hypothesis, whereby the diameter of xylem vessels affects P. chlamydospora compartmentalization outcome in grapevine and in turn cultivar resistance.

To better understand how vessel diameter could affect the level of resistance to P. chlamydospora in grapevine, we first characterized the plant response to infection and the fungal colonization strategy in a single cultivar. CS was used due to its intermediate level of resistance. Observation of tangential sections next to the point of inoculation in control plants showed a mass of scar tissue (a.k.a. bark ridge) that filled the wound and xylem vessel occlusion with tyloses that extended about 1 mm above the wound (Figure 3A). In contrast, no scar tissue could be found in P. chlamydospora inoculated plants and occlusion of vessels extended far beyond the inoculation point (Figure 3B). Shoot cross-sections 20 mm above the inoculation point showed typical necrotic wood lesions caused by P. chlamydospora (Figures 3D,F) while absent in mock-inoculated plants (Figures 3C,E). Symptomatic area displayed occluded vessels, either partially or completely, with tyloses and gels (Figures 3F–H). The use of Ruthenium Red revealed staining mostly of partially occluded vessels, although occasionally some reactions occurred with apparently fully occluded vessels (Figures 3G,I), suggesting the presence of immature tyloses and/or pockets of pectin-rich gels. The use of IKI stain revealed that the response to infection was associated with a depletion of starch storage in ray parenchyma cells next to occluded vessels (Figure 3H). However, starch depletion was only partial next to fully occluded vessels, whereas it was much more severe next to vessels partially occluded. Staining with Toluidine O indicated the presence of non-structural phenolics in the lumen of ray parenchyma and living fiber for both mock and P. chlamydospora inoculated plants (Figures 3A,B,J). However, the accumulation of phenolics often observed in cells next to partially occluded vessels suggests a local response to infection (Figure 3J). Characterization of cell wall modifications occurring in symptomatic area using the Mäule test indicated no noticeable change in structural lignin (Figure 3K). The use of Sudan IV and UV epifluorescence microscopy, however, revealed the deposition of suberin in ray parenchyma cells next to partially occluded vessels, indicating that a second line of local defense took place in tissue surrounding infected vessels (Figure 3L).

To better understand the colonization strategy of P. chlamydospora in vessels, the fungus was located in situ within symptomatic area of the xylem of CS shoots. For the co-visualization of P. chlamydospora hyphae and plant material by epifluorescent and confocal microscopy, best results were obtained using Auramine O (for the labeling of the plant material) counter stained with Calcofluor White M2R (for labeling of the fungal cell wall) (Figure 4). Observation of mock-inoculated plants showed no fungal structures in the xylem vessels (Figure 4A) indicating that those observed in the inoculated plants likely belonged to P. chlamydospora (Figures 4B–E). In vessels next to the point of inoculation, P. chlamydospora was detected in association with both damaged and intact tyloses (Figure 4B). In some vessels of wide diameter where occlusion failed, a dense network of hyphae was often observed in the xylem lumen along the vessel wall (Figure 4C). In other large vessels where occlusion was apparently successfully, the presence of P. chlamydospora hyphae was associated with deteriorated clusters of tyloses, where it progressed in between the network of mature tyloses walls and also between tyloses and the vessel wall (Figures 4D,E).

We first investigated how the diameter of vessels could impact the quantity of tyloses that were formed during occlusion in grapevine cvs. We observed that the occlusion in narrow xylem vessels featured a compact cluster of tyloses whereas in wide vessels a loose cluster of tyloses with pockets of pectic gels was often observed (Figures 5A–C). Scanning electron microscopy images revealed that the number of tyloses and the apparent tyloses cell wall surface increased with vessel diameter and this was consistent for all grape cultivars (Figures 5D,E). These observations were quantified and our data showed a strong positive correlation between the numbers of tyloses found in vessel cross-sectional area with the increase of xylem vessel diameter (Figure 6A). A strict correlation (R² = 0.73) was also found between the diameter of vessels and the outer tylosis surface length (OTSL) (Figure 6B). Really high dependence (R² = 0.97) was found between the number of tylosis and the OTSL. To determine whether these relations could differ amongst grape cultivars, a leave-one-group-out cross-validation scheme was performed. For every combination of grape cultivars, the regressions calculated with the data collected from three grape cultivars were found to explain with confidence (P < 0.05) the data observed on the fourth one. These observations confirmed that the influence of the diameter on the resulting tyloses network occluding a vessel was a commonly shared physiological process among the four cultivars studied.