FDp strain FD-PEY05 was used for inoculations. It was transmitted to broad bean (Vicia faba) by S. titanus leafhoppers, collected in 2005 in FD-infected vineyards in Peyrière, South-west France (Papura et al., 2009). The FD-PEY05 genotype belongs to the map-FD2 genetic cluster, and is widely distributed in the main outbreaks of Western Europe (Arnaud et al., 2007; Papura et al., 2009). The strain has been maintained since then by serial transmission on broad bean, using Euscelidius variegatus as a vector (Caudwell et al., 1972).

Details of V. vinifera cultivars, rootstock hybrids, and wild Vitis species used in this study are described in Table 1. Green growing arms of accessions were sterilized and introduced in vitro according to Perrin et al. (2004). For wild species, the concentration of calcium hypochlorite was reduced (2% final active chloride). Accessions were maintained by vegetative propagation on WPM medium (Lloyd and McCown, 1980), supplemented with 15 g.l^-1 sucrose and solidified with 6.5 g.l^-1 Difco bacto agar.

All graftings were done with in vitro grown plantlets. The rootstock consisted of a one-bud stem segment of CS or M (ca. 10 mm), onto which, a one bud-stem segment of Chardonnay scion was grafted (“English groove graft”). Plant vessels from the two parts joined within 4–6 days (data not shown), and plantlets were grown until the four to five internode developmental stage.

All in vitro plantlets (four to five internodes) were transferred for acclimatization in perlite-containing pots (Figures 1F,G). After 1 month of progressive weaning, plantlets were planted into 12 cm pots, filled with soil (Klassman n°5) for further development. Plants were grown at 25 ± 2°C, and 50–80% humidity, under illumination of 50 μEm^-2.s^-1 for a 16 h period (Osram, Lumilux), and under 400 W high pressure sodium lamps for a 14 h period, in vitro and in the greenhouse, respectively.

Three vineyards plots from Bordeaux area (Sites of Ambès, Baurech, and Beychac) were selected for growing side by side CS and M of the same age (except in Beychac), on the same rootstock (except in Beychac) and with important FD disease outbreaks (Figure 2). The mapping of the symptomatic plants was performed and symptom’s severity was recorded at the end of September 2010, and 2014 for the sites of Baurech and Ambès respectively and in mid-June 2011 for the site of Beychac. Stocks were classified in four categories: >0–25%, 26–50%, 51–75%, and >75% symptomatic branches on the plant. Twenty-eight to 35 symptomatic plants of each cultivar, distributed in the four categories, were randomly selected in the plot and 8–10 petioles from symptomatic leaves were sampled on the symptomatic branches at the end of September in Baurech and Ambès and at mid-June, mid-July, and mid-August in Beychac. Petioles were also sampled on the non-symptomatic parts of each stock and also on non-symptomatic stocks as controls.

After uprooting of vine plots, it is quite frequent to observe regrowth of rootstocks from undestroyed roots after years, and those invade neighboring canopies. Such wild rootstocks can also result from regrowth of cane cuttings which were thrown in the environment when, in the old times, grafting was directly performed in the plots. Such wild rootstocks regrowth, growing near FD-infected plots, were sampled at the end of the summer 2010 and 2012. They were mainly Vitis riparia, 161–49 V. riparia × Vitis berlandieri, and Clinton V. riparia × Vitis labrusca. Some symptomatic V. vinifera cultivars (CS, Semillon, Sauvignon) were also sampled in the neighboring FD-infected plots.

After washing, 0.5 g petioles were homogenized and frozen for further nucleic acid extraction and FDp quantification.

It was done as described in Maixner (2005), with a ratio of 0.5 g fresh weight (gFW) ground in 3 ml cetyltrimethylammonium bromide (CTAB) buffer. For the extraction procedure, 1 ml of the ground mix was used, and the nucleic acids were re-suspended in 80 μl TE 1× (Tris 10 mM, ethylenediaminetetraacetic acid (EDTA) 1 mM pH 8). For nucleic acid extraction from insects, a single individual was ground in 250 μl CTAB buffer, and the final pellet was re-suspended in 40 μl of TE 1×. Healthy grapevines were used as negative controls in each extraction series. Nucleic acid concentrations were determined with a NanoVue^TM Plus (GE Healthcare) and reduced to 40 ng.μl^-1 for qPCR analysis. Purity was assessed by calculating the ratio of the absorbance at 260 nm over the absorbance at 280 nm and at 230 nm.

FDp cells were detected and quantified in plants and in insects, by N′,N′-dimethyl-N-[4-[(E)-(3-methyl-1,3-benzothiazol-2-ylidene)methyl]-1-phenylquinolin-1-ium-2-yl]-N-propylpropane-1,3-diamine (SYBR) Green absolute quantitative real-time PCR of the tuf gene, as it is present as a unique copy in the bacterial chromosome (Malembic-Maher et al., 2008). Determined primers 3Fbl (5′-TGAAGATCCAGTACGTGATTTAGAC-3′) and 3Rl (5′-TTTTAGTTTCTTTAATACCTATGATTTC-3′) are positioned at 456–480 and 613–586, respectively on the FD92 tuf sequence FN561880 and led to the amplification of a 158 bp fragment. The conditions and validation of the PCR on the total nucleic acid extracts from plants and insects are detailed in the Supplementary Material (Supplementary Figure S4). The FDp strains detected in field samples were genotyped by sequencing the map gene (Arnaud et al., 2007).

Box plots, histograms and curves representing FDp titers were obtained using Microsoft Excel, with calculations of the mean FDp titers (number of FDp cells.gFW^-1 ± standard error) for each Vitis genotype, and each sampling date or each plant section. Coefficient of variation (CV) corresponding to the standard error, divided by the mean value of the FDp titer was determined. Data comparison, and statistical significance evaluation, were performed using the χ² and the Wilcoxon rank sum tests with software R, and R commander package (Fox, 2005). A scatterplot was generated using R, to evaluate the relationships between the relative mean FDp titers (ratio between the mean FDp titer in the genotype and the mean FDp titer in the CS from the same experiment), and the relative proportion of infected plants (ratio between the proportion of infected plants in the genotype and the proportion of infected plants in CS from the same experiment) of the different accessions inoculated in the greenhouse. Hierarchical classification analysis was performed using FactoMineR package, combining the two factors (Lê et al., 2008) to reveal the similarities between the Vitis accessions, with respect to FDp titers and the proportion of infected plants.

In the three vineyard plots from the municipalities of Baurech, Beychac, and Ambès, we identified 38% (2411) and 3% (307) symptomatic CS and M, respectively (Figure 2; Supplementary Table S1). The proportion of symptomatic M was drastically lower than CS in each whole plot and also when only the 10 rows on each side of the border between CS and M were compared (Supplementary Table S1). To better characterize the differences between these cultivars, we evaluated the number of branches showing symptoms of FD disease (Figures 3A–C). Four classes of plants with >0–25%, 26–50%, 51–75%, and >75% symptomatic branches were defined. Of the CS stocks, 70% showed over 75% symptomatic branches whereas 55% of M stocks showed less than 25% symptomatic branches (Figure 3C). Similar results were obtained on each site independently, and even when only the bordering rows were compared (Supplementary Table S1). At the end of the summer in Ambès and Baurech, and throughout a full vegetative season in Beychac, we sampled petioles on CS and M stocks (28–35 stocks/cultivar/plot), equally selected in the four classes of symptoms and randomly distributed on each plot. FDp was detected in all symptomatic samples. The map gene sequence of FDp collected in Beychac was identical with the one of FDCAM-05 strain (Map-FD1, Papura et al., 2009) and FDp from Ambès and Baurech was genotyped as FDPEY-05 (Map-FD2).

In Beychac, FDp titers increased significantly between mid-June and mid-July for both cultivars and then stabilized until the end of August (Figure 4; Supplementary Table S1). At each time-point, FDp titers in M were inferior to those in CS with high statistical significance. The ratio between mean FDp titers in CS plants and M increased along the season from 5 times, in June, to 23 times, in autumn in the Beychac vineyard, while in Ambès and Baurech vineyards, ratios of 46 and 30 were reached, respectively (Figure 4). When analyses were performed on plants from the 10 rows of the border between CS and M, the ratio was 50.7 in Ambès and 30, in the site of Baurech. The plotting of FDp titers to the proportion of symptomatic branches for each M and CS plants shows that these variables are not linked (R² < 0.02; Supplementary Figure S1). We also sampled the green arms that remained non-symptomatic on each symptomatic stock studied. Ninety-one percent and 71% were positive for FDp in M and CS, respectively. However, phytoplasma was hardly detected, at the limit of quantification for both cultivars, suggesting that non-symptomatic branches from a symptomatic plant may contain very low FDp titers, whatever the cultivar. Finally, FDp was not detected in the completely non-symptomatic stocks from the same vineyard-plots.

In conclusion, our observations demonstrate that in the same vineyards, and even in the close bordering rows, most likely subjected to similar abiotic and biotic conditions, the proportion of symptomatic plants, the proportion of symptomatic branches, and the FDp titers in symptomatic branches were lower in M than in CS.

Wild rootstocks growing near FD-infected vine plots were sampled at the end of the summer in the Bordeaux area. Petioles were collected mainly on branches of V. riparia showing leaf yellowing, and on some branches of 161-49C (V. riparia × V. berlandieri) and Clinton (V. riparia × V. labrusca) that showed leaf yellowing and/or discrete rolling. Among 56 rootstocks analyzed, 37.5% were positive for FDp, with a mean titer of 2 × 10⁷ ± 6 FDp cells.gFW^-1, which is similar to the titer observed in CS at the end of the season.

To further characterize grapevine susceptibility, we attempted to translate what occurs in the vineyards into controlled conditions, under confinement constraints (Figure 1). All accessions introduced in vitro developed after a 6-month period. Plants became juvenile and cuttings shared very homogenous plant development. After transfer in soil, plants developed homogenously too, thus favoring insect-mediated transmission (Figure 1).

Through the inoculation experiments performed across 4 years, the proportion of infected S. titanus collected at the end of each transmission experiment varied from 83.6 to 100% and the mean FDp titers ranged from 4.5 × 10⁷ ± 9.1–3.4 × 10⁸ ± 4.8 FDp cells.insect^-1. In such conditions, transmission efficiency in the control CS was high, with a mean proportion of 92.9% infected plants (CV 5.7%) and the first symptoms were observed 6 wpt in all experiments. For FDp titers in CS plants, the CVs were higher, ranging from 51.2 to 113.7% for each experiment and 78.3% between experiments. Inter-experiment variations were normalized by calculating the percentage of infected plants and the mean FDp titers relative to CS for each experiment [cf. Materials and Methods (M&M)]. The whole procedure lasted 24–32 weeks, and proved to be efficient on a large collection of accessions.

For the time course experiment, infectious insects were encaged on each whole CS and M plants. The percentage of infected plants (n = 5–10), symptom development and mean FDp titers in the entire aerial parts were evaluated 3, 5, 10, and 16 wpt (Figure 5). The proportion of infected M (34.4%) was lower than CS (96.9%), as observed in the vineyards. In CS, the mean FDp titers strongly increased after transmission, with a multiplication factor of 38 between 3 and 10 wpt, and a maximum of 7 × 10⁷ ± 5.4 FDp cells.gFW^-1 10 wpt. The FDp titers were statistically different between all time points. The symptoms appeared between 6 and 7 wpt, close to the end of the FDp growing phase and reached their maximum intensity 12 wpt (Figure 5). Ten wpt, FDp titers tended to decrease, with a mean of 2.2 × 10⁷ ± 1.4 FDp cells.gFW^-1, 16 wpt and, during this phase, plants began to decay with leaves drying and falling, with 62.5% of CS presenting dead parts of the plant. In M, FDp titers followed a similar growth curve as in CS, but with no statistical significance between sampling times. In the experimental conditions, M plants did not show any symptoms and were still growing 16 wpt (Supplementary Figure S3). Notably, FDp titers measured in M remained statistically lower than in CS. Indeed ratios between means FDp titers in CS vs M were of 224 and 241, 5 and 10 wpt, respectively, thus resembling data obtained in the vineyards.

To study the distribution of FDp in CS and M plants, we encaged infectious S. titanus on the 3rd or 4th leaf from the apex. Ten wpt, CS and M plants were cut in sections, as schematized in Figure 6A and FDp was quantified (Figure 6B). The CS sections 2, including the leaf that received the infected S. titanus, were 100% infected with a high mean FDp titer of 1.3 × 10⁷ ± 1.8 FDp cells.gFW^-1 (Figure 6B). The other sections from the main stem were infected in the same proportion, however, with mean FDp titers 99 times higher in the apical (section 3) than in the basal parts (sections 1) (p-value < 0.001). The proportion of infected basal regrown arms (R1) was lower than for the other sections (38%), with a titer similar to that measured in the apex. In contrast, 63% of the M sections 2 were infected with low mean FDp titers of 8.9 × 10⁴ ± 12 FDp cells.gFW^-1 (Figure 6B), while none of the other sections were infected, suggesting that FDp is almost restricted to the leaf used for transmission. Taken together, these data suggest that FDp circulates within the CS plant and multiplies preferably close to the main apical meristem or to the meristem of axillary branches initiated post-transmission, whereas the phytoplasma is arrested in M, in the site of infection.

In the course of the experiments, we observed lower survival rates of S. titanus at the end of the transmission period on M, when compared to CS (Supplementary Table S2). This discrepancy could lead to a lower transmission efficiency within M. To elucidate whether this differential response is due to a plant–insect or FDp–plant interaction, we grafted a scion of Chardonnay, a highly susceptible cultivar, onto CS and M rootstocks. Plant tissues readily welded within 1–4 days, so that vascular tissues looked continuous. After 2 months’ further development of the grafted plants, S. titanus were encaged for transmission onto the Chardonnay scion solely. The Chardonnay shoots expressed symptoms 5–6 wpt for both cultivars. The proportions of infected Chardonnay scions were 71 and 83% for CS and M, respectively. The FDp titers, 15 wpt, were not statistically different in Chardonnay shoots grafted onto CS or M, suggesting that the rootstock does not influence FDp multiplication into the scion (Figure 7). The FDp likely diffused into the plants, through the graft, and then multiplied, reaching a high titer 15 wpt in the CS rootstocks, with a mean of 1.7 × 10⁷ ± 2.2 FDp cells.gFW^-1 and 100% infection rate. In contrast, a very low titer of 2.4 × 10³ was found in a single M rootstock, whereas the other five remained FDp-free. So, although we do not exclude an influence of plant cultivar on the transmission step by S. titanus, here the FDp–plant interaction per se can explain the lower susceptibility of M, when compared to CS.

A collection of 28 Vitis accessions was introduced and cultivated in vitro (Table 1). Accessions were free of virus Grapevine virus A (GVA), grapevine leafroll-associated virus (GLRaV) 1–3, and grapevine fanleaf virus (GFLV), except CS 337, which was infected with GLRaV 2 (data not shown). The percentage of infected plants, symptom development, and mean FDp titers in the entire aerial parts were evaluated 5 and 10 wpt. A total of seven transmission experiments were performed (Supplementary Table S2), with the responses of the control CS being used as a reference, in each experiment. The scatterplot (Figure 8) integrated for each accession the mean relative FDp titer in infected plants and the proportion of infected plants, relative to CS. On the basis of these values, a hierarchical clustering suggested three categories (Figure 8B): (1) accessions with high FDp titers and high proportion of infected plants, (2) accessions with intermediate FDp titers and high proportion of infected plants, and (3) accessions with intermediate to low FDp titers and low proportion of infected plants (red, black, and green circles, respectively in Figure 8A). V. vinifera cultivars were distributed throughout these three categories. Sauvignon was classified as highly susceptible with FDp titers and proportion of infected plants not statistically different from those measured for CS (Supplementary Table S2). On the opposite, M, Syrah, and Magdeleine were classified as poorly susceptible, showing FDp titers and proportion of infected plants much lower than in CS, with high statistical significance. Pinot Noir, although included in the same hierarchical cluster than M, showed a more susceptible phenotype, with low proportion of infected plants but intermediate FDp titers. Finally, Grenache, Chardonnay, and Cabernet Franc were classified as intermediate, with intermediate FDp titers, but high proportion of infected plants. The cultivars with high or intermediate FDp titers showed symptoms (Figure 8; Supplementary Figure S3) while those with low titers did not show any symptoms 10 wpt (Figure 8).

For the rootstocks, with the exception of 5BB and Nemadex, which were classified as poorly susceptible, all accessions were grouped in the category of “intermediate susceptibility.” They showed high proportions of infected plants and intermediate relative FDp titers, ranging from 0.04 for 41B to 0.25 for 3309C. The 5BB was less susceptible, with significantly lower proportion of infected plants and FDp titers. Finally, in Nemadex, a complex hybrid with a Muscadinia rotundifolia genetic background, FDp was barely detectable and plants did not show any symptoms (Supplementary Figure S2). Interestingly, none of the rootstocks exhibited FD symptoms in our experimental conditions (Figure 8), even those having the highest FDp titers.

The 12 wild Vitis species analyzed were distributed among the three categories, with diverse susceptibilities. As observed for the rootstocks, they could be infected and have high FDp titers, without developing any symptoms 10 wpt, except for Vitis doaniana and Vitis rubra which showed yellowing and rolling of the leaves 7–9 wpt. V. berlandieri, Vitis rupestris, and V. riparia (RGM), the species most used decades ago in rootstock breeding schemes to produce the main rootstocks used today, showed high proportions of infected plants and intermediate to high FDp titers. However, none showed symptoms 10 wpt. A high proportion of Vitis pentagona and Vitis coignetiae were infected, but the FDp titers were low. Conversely, low proportions of infected plants, but with high FDp titers were obtained with Vitis simpsonii. These observations suggest that the efficiency of transmission, on one hand, and the FDp titers, on the other hand, are likely associated with distinct plant traits. The same stands true for the FDp titers and symptom development, which are unlinked in all rootstocks studied.