Stems of type A, without inflorescences, clusters and leaves were stored at 4°C. Then, samples were taken from the first three basal internodes for macroscopic and microscopic observations of vascular tissues.

Pieces of internodes 1, 2 and 3 (number 1 being closest to the base) were sectioned from stems of type B. The xylem area was detached from the phloem/bark (thereafter called “Xylem” or “X” and “Phloem”, “P” or “bark”, respectively) part, individually snap-frozen and then stored at -80°C. As small quantities of tissues were sampled by internode, we mixed tissues from the same internode of the five sampled vines. Each sample was ground in liquid nitrogen using a ball mill (Retsch MM400) and divided in two sub-samples: one for molecular analyzes (PCR and qRT-PCR), and one for metabolic analysis (GC-MS).

Stems of three first internodes from the five asymptomatic and symptomatic plants were sampled at 6 stages. Four pieces of approximately 3x1x1 mm in size from each internode were placed in Petri dishes containing malt agar medium (15 gL^-1 cristomalt, 20 gL^-1 agar, 200 mgL^-1 of chloramphenicol) and incubated in the dark at room temperature (Larignon and Dubos, 1997). Xylem and bark were weekly observed and pictures were acquired (Canon Power Shot G7) for 4 weeks for visual examination. The color, the hyphal and pycnidiospore morphology, and the aerial mycelium of the colony were examined under light microscope (Larignon et al., 2001).

The genomic DNA was extracted from molecular analyses sample powder of X and P tissues, by following the extraction protocol of Mundy et al. (2018). The genomic DNA of X and P samples was extracted using the DNeasy Plant Mini kit (Qiagen, Hilden, Germany). The quality of DNA was checked by agarose gel electrophoresis and the quantity was determined by measuring the absorbance at 260 nm.

Amplicons were generated by PCR using the Phusion High-Fidelity DNA polymerase, 20 ng of DNA, and the primers ITS1-F and ITS4, as described by White et al. (1990) and Udayanga et al. (2012) ( Supplementary Table 1 ). A nested PCR was performed taking 5 μL of the 1:100 diluted PCR1 template and using primers BOT100F and BOT472R, as described by Ridgway et al. (2011) ( Supplementary Table 1 ).

From the collected internodes 1, 2 and 3, cross sections were performed and observed under a macroscope (ZEISS, AXIOZoom.V16). Two images were acquired using the “SNAP” command (ZEN2.3 system software); one with samples exposed to white light (RGB experiment settings), and one with samples exposed to blue light only (DAPI experiment settings, excitation 365 nm, emission 445-450 nm).

A semi-automatic analysis was performed to establish the proportion of obstructed vessels among the total number of vessels. First, the total number of vessels (both obstructed or not obstructed) was counted automatically with an ImageJ macro. The image acquired with DAPI settings was processed with ImageJ 1.52 software with the Fiji package. Images were contrasted manually, then an automatic threshold was created (Auto Local Threshold parameter: method Phansalkar), thus converting the picture in either black or white pixels, without grey shades. Then using the “Analyze particles” setting, the total number of vessels was assessed (parameters: size: 35-1000; circularity: 0.35-1.00). The number of obstructed vessels was assessed manually on the RGB image using the multi-point count tool from the Fiji package.

The ratio of obstruction was calculated as follows:

Total RNA was isolated from 3 x 100 mg of X and P powder using the PureLink Plant RNA Purification Reagent (Invitrogen, Cergy Pontoise, France). The manufacter’s protocol was followed until the phase of separation with the chloroform: isoamylic alcohol (24:1). Then, one volume of ethanol 70% was added to the aqueous solution, before purification using the NucleoSpin RNA kit (Macherey-Nagel, Düren, Germany). The quality of RNA was checked by agarose gel electrophoresis and the quantity was determined by measuring the absorbance at 260 nm for each sample and adjusted to 100 ng µL-1. First-strand cDNA was synthetized from 150 ng of total RNA using the Verso cDNA Synthesis kit (Thermo Fisher Scientific, Inc., Waltham, MA, United States).

Real-time polymerase reaction (PCR) was performed using Absolute Blue qPCR SYBR Green (Thermo Fisher Scientific, Waltham, MA, USA), in a CFX96 real-time PCR detection system (Bio-Rad, Hercules, CA, USA). The thermal profile was 10 s at 95°C (denaturation) and 45 s at 60°C (annealing/extension) for 40 cycles. The specificity of PCR amplification was checked using a heat dissociation curve from 65 to 95°C following the final cycle. For each experiment, PCR reactions were performed in duplicate. Expression of two reference genes (EF1-α and 60SRP) and of genes encoding enzymes involved in the sucrose metabolism and signalization (αAMY, βAMY, CWINV and SUC27), the plant defense responses (Cal-S7, GLU, GST5, PAL, POX4, PR6, STS, TL and WAT1), detoxification and stress tolerance (epoxH2 and SOD), vascular cambium (CDKB2, ERF5, MOL and WOX4) and cork cambium (APL, EBP1, HA3, SHR, PSKR1 and PSKR2) was tracked by quantitative reverse transcription-PCR (qRT-PCR) using the primers listed in Supplementary Table 2 .

The samples were fixed into a fresh mixture of 4% paraformaldehyde in potassium phosphate buffer (PBS 10mM, pH 7.4, with addition of tween 20 1‰ only in the first 1 h bath) and then embedded in paraffin (Colas et al., 2010). Wax sections at 10 µm thickness were obtained using a microtome and deposited on silanized slides. The samples were then dewaxed by 2 Q Path^® Safesolv baths (20 min each at room temperature), gently rinsed with a pipette in ethanol 90° (1 min at room temperature) and then gradually rehydrated; deposits of drops of ethanol 70°, 50° and 25° were gradually made and each was left for 1 min in contact with the sections. Finally, ultra-pure water was put in contact on the slide during 2 min in order to completely rehydrate the sample.

Aniline blue staining (1% in 3% acetic acid) was carried out by leaving the drop of dye 30 s on the samples, then rinsing once 10 s with 3% acetic acid and then twice 10 s with ultra-pure water in order to remove excess dye. Finally, the samples were observed under a bright-field light microscope (Leitz DM RB, Leica), and images were acquired using a camera (Nikon Digital Sight), with respect of the same shooting parameters for each observed sample.

The samples were fixed immediately in 2.5% glutaraldehyde (in 0.1 M sodium- phosphate buffer, pH 7.2, 1% sucrose, 1 ‰ Tween 20) for 20 to 25 min under vacuum, then overnight at 4°C with gentle rotation (without Tween 20). Samples were washed twice (10 min) in the same buffer and postfixed in 1% osmium tetroxide (OsO₄) in the same buffer for 1 h at 4°C. Then, samples were washed in phosphate buffer, dehydrated in a graded ethanol series, and treated with propylene oxide. Dehydrated samples were subsequently embedded in Epon (Merck, Darmstadt, Germany) and sectioned using a Reichert Ultracut E microtome (Leica, Reuil-Malmaison, France) with glass or diamond knives (Diatome, Bienne, Switzerland). Semithin (0.5 μm) sections from the tissue blocks were stained with 1% aqueous toluidine blue (in 1% sodium tetraborate) and examined under a bright-field light microscope (Leitz DM RB, Leica) in order to observe anatomical characteristics of both bark and xylem.

Data analysis were carried out using Past 4.08 software. A Chi-square (X²) test of homogeneity was used to compare the distribution of obstructed to non-obstructed vessels in shoot of asymptomatic and symptomatic stems. Moreover, a Kruskal Wallis test (p<0.05) was used to analyze and compare the proportions of obstructed vessels between internodes and sampling period.

For each gene and for each modality, a mean Cq value was obtained. Relative gene expression (RE) was determined with the 2-ΔΔCq method using CFX Manager 3.0 software. For every sample, ΔΔCq was the ΔCq difference between 2 samples (diseased vs control). These values were used to generate a gene expression heat map through Past 4.08 software. Hierarchical clustering (ward’s method, Euclidian distance) was applied to group samples with similar expression level.

GC-MS data were processed as described in Moret et al. (2020). Briefly, Data files in NetCDF format were analyzed with AMDIS software. A home retention indices/mass spectra library and standard compounds were used for metabolite identification. Peak areas were determined with the Targetlynx software (Waters). AMDIS, Target Lynx in splitless and split 30 mode data were compiled into a single Excel file for comparison and peak areas were normalized to ribitol and fresh weight. GC-MS data were then pre-processed and filtered RSD (relative standard deviation) on QC samples was calculated for each compound. Only features with RSD > 30% and non-aberrant QC were conserved for further analysis. Missing values were replaced by half the lowest value, corresponding to limit detection of the method. PCA (Principal Component Analysis) and PLS (Partial Least Square) analyses were made on the processed data. Mean, fold change and two-sided Student’s T-test p-value (p<0.05) were calculated and compiled into a data frame that was used for the Volcano plots (Goodacre et al., 2007; Vinaixa et al., 2012; Worley and Powers, 2013; Schiffman et al., 2019). Significant compounds highlighted by Volcano plots (T-test p<0.05) were verified and compounds with aberrant concentrations across samples were removed.

Table 2 summarizes the analysis of 1200 fragments of either bark or xylem. In 2018, Botryosphaeriaceae were isolated whatever the sampling time, with a maximum detection at maturity in both phloem (bark) and xylem. Botryosphaeriaceae isolates were more abundant from bark samples than from xylem samples (i.e., 14 times more at the “inflorescences visible” stage). The number of isolates was similar in both tissues at flowering stage, but two times more abundant in the bark than in the xylem at maturity stage.

Botryosphaeriaceae were present in both control (C) and diseased (D) vines, in both phloem (P) and xylem (X), whatever the internode (1, 2 and 3) and the sampling period ( Table 3 ). The detection of Botryosphaeriaceae is less frequent at T3 (late veraison stage; 6 out of 12 samples) than at T1 (flowering; 10 out of 12 samples) and T2 (cluster closure; 11 samples out of 12).

In xylem (X) and phloem (P), targeted genes on sucrose metabolism and signalization, plant defense response and detoxification process, and secondary meristems were studied on the 3 internodes before and during the expression of foliar symptoms. A hierarchical clustering (heatmap) was made in order to group samples, in P or X tissues, with similar expression levels ( Figure 3 ). Regardless of the tissue observed (P or X), the samples from times T1, T2 and T3 were clearly separated. Overall and regardless of the plant tissue, defense genes such as TL, PR6, GLU and STS were among the most up-regulated, followed by genes related to detoxification (GST5) and to the degradation of storage or transit sugars (αAMY and CWINV).

Even if the most regulations (up and down) were observed for the time T2 (start of appearance of symptoms), it is interestingly to note that regulations (up and down) were also detected before the expression of leaf symptoms (T1; Figure 3 , light grey). At this stage, genes related to carbohydrate metabolism (aAMY and CWINV) and defense responses (GLU, PR6, GTS5, POX4, STS and TL) were up-regulated in both X and P tissues. Moreover, the Cal-S7 gene was up-regulated in P tissues ( Figure 3A ) and down-regulated in X tissues ( Figure 3B ). At T1, no gene modifications related to the formation of the secondary meristem were detected except for APL, repressed in X tissues. At the early symptomatic stage (emergence of symptoms at T2), the expression of the majority of targeted genes was modified in X and P tissues of D vines ( Figure 3 ). A down-regulation of genes associated to the cambium (WOX4, CDKB2 and MOL), the phellogen (APL and SHR) and sucrose signalization (SUC27), and an up-regulation of genes associated to carbohydrates (aAMY, βAMY and CWINV), plant defense (GLU, PR6, GTS5, STS, PAL and TL) and phellogen (HA3, and PSKR2) were observed ( Figure 3 , medium grey). Changes at the late symptomatic stage (T3) were similar to changes at T2, except for genes involved in the formation of secondary meristems and showing less changes ( Figure 3 , dark grey).

Hierarchical Cluster Analysis (HCA) analysis performed on the whole sample set allows to clearly separate the phloem (P) and xylem (X) samples in two clusters ( Supplementary Figure 3 ). PCA analyses were therefore performed on GC-MS data of each tissue ( Figure 4 ). Samples were separated depending on time (PC1, 32.2%), and to a lesser extent depending on health state (PC2, 25.7%). For both tissues, at time T1 (before symptom expression), there is no separation of C and D samples (red and green ellipses, respectively), suggesting a similar metabolic profile. At T2 and T3, all groups are different and different from T1, indicating that their metabolic profile evolves over time with disease expression. For both tissues, variability is low whatever the modality and the sampling time.

Volcano plants represent the distribution of compounds for which the concentration is significantly different in X and P tissues between C and D samples ( Figure 5 , Supplementary Tables 3 , 4 ). The number of compounds differentially accumulated in P and X tissues is of 1 and 0 for P and X, respectively at T1, of 42 and 39 for P and X, respectively, at T2, and of 22 and 27 for P and X, respectively, at T3. At T1, glycosylsalicylate was less than two-times more accumulated in P tissues of D plants than in P tissues of C plants. At T2, there was two-times more compounds accumulated in samples from D plants than in samples from C plants. In P tissues, 1-kestose, digalactosylglycerol and threonate were more accumulated in tissues from C plants than from D plants, and mannitol, trans- and cis-resveratrol and leucine were more accumulated in tissues from D plants than from C plants ( Supplementary Table 3 ). In X tissues, 1-kestose and galactinol were more accumulated in tissues from C plants than from D plants, and trans- and cis-resveratrol, mannitol and piceid were more accumulated in tissues from D plants than from C plants ( Supplementary Table 4 ). At T3, there was more compounds accumulated in tissues from D plants than from C plants. In P tissues, 1-kestose, threonate and phenylalanine were more accumulated in tissues from C plants than from D plants, and glutamine and isoleucine were more accumulated in tissues from D plants than from C plants ( Supplementary Table 3 ). In X tissues, raffinose, digalactosylglycerol and 1-kestose were more accumulated in tissues from C plants than from D plants, and trans- and cis-resveratrol, and piceid were more accumulated in tissues from D plants than from C plants ( Supplementary Table 4 ). The concentration of 1-kestose was determined in all samples as significantly accumulated in tissues from C samples ( Figure 6 ). At T1, the concentration of 1-kestose was similar in X and P tissues from C and D samples (C/D ratio from 0.8 to 1, and from 1 and 1.9 in P and X tissues, respectively). At T2, the concentration sharply decreased in X and P tissues from D plants (not detected in internode 1 of P). At T3, the concentration increased (except in X of internode 3: not detected) in X and P tissues, without reaching T1 values (C/D ratio from 4.6 to 7.5, and from 4.1 to 5.1 in P and X tissues, respectively).

At T1, the herbaceous stems were collected before the onset of leaf symptom expression ( Figure 7 ).

Regarding its anatomy, the overall view of the C sample ( Figure 7A ) has a characteristic structure of the Euvitis section of Vitis genus (Metcalfe and Chalk, 1950; Esau, 1965). The secondary phloem appears in blocks separated from another one by wide rays ( Figure 7A , r). Within the blocks, tangential strata of sieve elements ( Figure 7A , s) with associated companion cells alternate with tangential bands of fibers ( Figure 7A , fi). Separating the secondary P from the secondary X, the presence of the vascular cambium is clearly observed ( Figure 7A , ca). At the level of the secondary P ( Figure 7B ), the sieve tubes and their companion cells are distinguished, indicating that this tissue is functional in the transport of the elaborate sap. Moreover, functional sieve plates can also be observed ( Figure 7C , arrow). The secondary X presents functional, unobstructed vessels ( Figures 7A, D ), and many amyloplasts could be observed in parenchyma rays ( Figure 7D , arrow).

In the stems sampled on declining vines (pre-symptomatic), clear anatomical alterations were observed in both secondary P and X ( Figure 7E ). In the secondary P, even if the layers of fibers appear conform to what is observed in C samples, vascular cells are disorganized. Indeed, in some areas the sieve plates appeared blocked by a material more or less intensely stained with toluidine blue ( Figures 7F, G , arrows). In the meantime, the P vascular cells appeared collapsed in some layers, crushed and therefore non-functional ( Figure 7G ). Finally, in the secondary X some vessels were obstructed by the presence of tylosis ( Figures 7E, H ). It clearly also appeared that, unlike C samples, the parenchyma rays were empty of any amyloplast ( Figure 7H , arrow).

At T2 and in normal development (C plants, Figure 8.1 ), the stem will transform from an immature state with a green cortex (herbaceous stage), to a stem with collapsed and brownish epidermis created by the developing periderm from an active phellogen (the cork cambium, a lateral meristem that creates the periderm). Indeed, on the 5 C stems taken, all had a periderm (data not shown). Structural observations of C stem ( Figure 8A ) revealed that the initial periderm is formed, normally, within the secondary phloem ( Figure 8A ). Periderm is the secondary protective tissue that replaced the epidermis. It consisted of phellem (cork), phellogen (cork cambium) and only few layers of phelloderm (Esau, 1948). Under the periderm, we observed the same tissues as those described previously with the secondary phloem, separated from the secondary X by the vascular cambium ( Figure 8A , ca). As for the previous sampling time, the secondary phloem appears functional ( Figure 8B ), with the presence of sieve tubes and unobstructed sieve plates separating neighboring vascular elements ( Figure 8C , arrow). In addition, the xylem also appears to be functional ( Figure 8D ), with the presence of empty, unobstructed vessels and a large number of amyloplasts in the parenchymal rays ( Figure 8E , arrow), suggesting an efficient starch storage.

In D vines, periderm development was complete for two canes, incomplete for one and not differentiated for two (data not shown). As example, green islands found on the stem of a D cane ( Figure 8.2 ) consist on an incomplete periderm formation and subsequent uneven cane maturation. At the structural level, we therefore observed a defect in the placement of the phellogen, reflected by the absence of this meristem at the level of the outermost secondary phloem layers ( Figure 8F ). Moreover, as at the previous sampling time, and whatever the inter-node, the secondary phloem appears non-functional from a vascular point of view ( Figures 8F–H ). Indeed, the vascular cells are totally collapsed and it even becomes almost impossible to distinguish them individually ( Figures 8G, H ). Occasionally, the presence of sieve plates is still observed which appears obstructed by a component weakly stained with toluidine blue ( Figure 8H , arrows). In the secondary xylem we also observed the presence of vessels obstructed by tylosis ( Figures 8I, J ) and the lack of amyloplast in the parenchymal rays ( Figure 8J , arrow), indicating a defective storage in D vines. Finally, in the internodes at the base of the canes (internode 1), we were able to demonstrate the presence of fungal hyphae in the secondary phloem ( Supplementary Figures 4C-E , arrows). In all cases, the presence of these hyphae was found in the tissue layers closest to the vascular cambium and not in the most peripheral or internal layers.

The potential alterations in the secondary P were observed and identified after aniline blue staining from samples of C and D plants included in paraffin. In samples from C plants, the sieve tubes appear functional ( Supplementary Figures 5A, B ), the sieve plates are visible ( Supplementary Figure 5B , arrows) and a simple pale blue staining (ie. background staining) is detected. At T1, in samples of D plants, an intense blue staining is observed at the level of the cell walls of the sieve tubes whatever the internode, suggesting the presence of callose ( Supplementary Figures 5C, D ) and a potential alteration of the transversal elaborate sap conduction.