Severe Stunting Symptoms upon Nepovirus Infection Are Reminiscent of a Chronic Hypersensitive-like Response in a Perennial Woody Fruit Crop

Virus infection of plants can result in various degrees of detrimental impacts and disparate symptom types and severities. Although great strides have been made in our understanding of the virus–host interactions in herbaceous model plants, the mechanisms underlying symptom development are poorly understood in perennial fruit crops. Grapevine fanleaf virus (GFLV) causes variable symptoms in most vineyards worldwide. To better understand GFLV-grapevine interactions in relation to symptom development, field and greenhouse trials were conducted with a grapevine genotype that exhibits distinct symptoms in response to a severe and a mild strain of GFLV. After validation of the infection status of the experimental vines by high-throughput sequencing, the transcriptomic and metabolomic profiles in plants infected with the two viral strains were tested and compared by RNA-Seq and LC-MS, respectively, in the differentiating grapevine genotype. In vines infected with the severe GFLV strain, 1023 genes, among which some are implicated in the regulation of the hypersensitive-type response, were specifically deregulated, and a higher accumulation of resveratrol and phytohormones was observed. Interestingly, some experimental vines restricted the virus to the rootstock and remained symptomless. Our results suggest that GFLV induces a strain- and cultivar-specific defense reaction similar to a hypersensitive reaction. This type of defense leads to a severe stunting phenotype in some grapevines, whereas others are resistant. This work is the first evidence of a hypersensitive-like reaction in grapevine during virus infection.


Introduction
The development of severe symptoms, rather than the presence of the virus itself, constitutes the main nuisance caused by viral infections in crop plants. It is now generally accepted that virus-host interactions are responsible for the development of symptoms, rather than the competition for resources [1][2][3][4]. These specific interactions between the viral pathogenicity determinants and the host components are thought to perturbate the host physiology [3]. The mechanisms underlying pathogenesis are diverse and include RNA silencing when the RNAs from the virus sequences share similarities with endogenous, coding or noncoding, RNAs [2,[5][6][7][8][9]. However, the symptoms are mainly determined by viral pathogenicity proteins through protein-protein interactions with a host factor [10,11]. While many viral pathogenicity factors, and their mutants, affecting symptom severity

Biological Material and Virus Inoculation
The GFLV strains, F13, GHu, B844, CO1, and CO2 (Supplemental Table S1), from different Vitis vinifera cvs [34,35,[37][38][39], were used in this study. All strains are composed of one RNA1 and one RNA2, except strain B844, which contains two RNA2 molecules designated RNA2a and RNA2b (Table S1). After multiple passages on herbaceous hosts, these five GFLV strains were transferred to the certified rootstock Kober 5BB clone 259 (Vitis berlandieri × Vitis riparia) by in vitro heterologous grafting. Briefly, rootstock cuttings were sterilized and assembled in tissue culture onto the stem fragments of Chenopodium quinoa infected by a specific GFLV strain, following mechanical inoculation, to allow the virus to translocate from the herbaceous host into the grapevine rootstock [37]. After removal of the C. quinoa stems, the grapevine rootstocks were further grown and micropropagated in tissue culture. The rootstocks were then tested for GFLV infection and the infected rootstocks were grafted with V. vinifera cv. Gewurztraminer (Gw; certified clone 643), or Chardonnay (Ch; certified clone 131), using green-grafting in an environmental growth chamber [40]. Each test vine consisted of a scion cultivar (Gw or Ch) grafted onto the Kober 5BB rootstock and will be named, herein, after the scion cultivar (Gw or Ch). The plants of Gw and Ch corresponding to scions grafted onto uninoculated rootstocks were used as controls. A total of 131 vines were established in four trials, as detailed in Table S2, either in an X. index-free field site or in a greenhouse. The vines in the field were trained according to the simple Guyot method. In the greenhouse, the plants were in individual pots and trained with two shoots at a 1.80 m height.

Detection of GFLV
The translocation of the virus from the rootstock to the cultivar was monitored over time by a double antibody sandwich-enzyme-linked immune-sorbent assay (DAS-ELISA) of leaf samples using a GFLV-specific serum (Bioreba, Reinach, Switzerland). For an estimation of the virus titer, the absorbance values were compared with the known amounts of purified GFLV serially diluted in the sap of healthy grapevines [34,35]. The presence of GFLV RNAs was confirmed by an immunocapture-reverse transcription-polymerase chain reaction-restriction fragment length polymorphism assay (IC-RT-PCR-RFLP), using StyI as described in [41]. The detection of viral siRNAs by northern blot was performed on RNA isolated from cortical scrapings and leaves as described in [42]. Long viral RNAs were detected by RT-qPCR [42].

RNA Extraction and RNA-Sequencing (RNA-Seq)
The total RNA from the leaf samples (200 mg) was isolated using the RNeasy Plant Mini Kit (Qiagen, Hilden, Germany), according to the manufacturer's instructions, with the addition of DNase. All library preparations, RNA treatments, and RNA sequencing steps were performed by the GeT-PlaGe platform of Genotoul (Genopole, Toulouse, France) using the TruSeq Stranded mRNA sample prep kit (Illumina, San Diego, CA, USA). Using poly(A) selection, the mRNAs were isolated, fragmented, and reverse-transcribed. Experiments were performed on an Illumina HiSeq 3000 (Illumina, San Diego, CA, USA), using a pairedend read length of 2 × 150 base pairs with Illumina Hiseq3000/4000 SBS sequencing kits.

Sample Infection Status and GFLV Diversity
Analyses of the sequence datasets were performed using CLC Genomics Workbench 11.0 software (Qiagen Digital Insights, Aarhus, Denmark), as described in [43]. After the trimming and quality check, only reads longer than 70 nucleotides were kept. Using loose parameters, with a length fraction of 0.5 and similarity of 0.7, a direct mapping was performed with a curated set of references based on grapevine viruses and viroids [27]. To study the genetic diversity, the de novo assembly tool from CLC Workbench (Qiagen Digital Insights, Aarhus, Denmark) was used and the contigs were mapped against GFLVcoding regions. The GFLV RNA1 and RNA2 5 and 3 untranslated regions (UTRs) were excluded in this study because they contain sequences shared by the two RNAs [29,44]. Multiple rounds of mapping and assembly were implemented until complete coding regions were obtained.
Alignment analysis of the nucleic acid sequences and neighbor-joining-based phylogenetic trees were performed using CLC Workbench, with bootstrapping analyses of 1000 replicates.
Gene categorization was performed by an enrichment of the Gene Ontology (GO) terms on the DEGs using the TopGo R package (v2.32.0) [51]. Revigo online-specific tools were used to hierarchize the enriched GOs, providing a more suitable representation [52].

Reverse Transcription Quantitative PCR (RT-qPCR)
Specific primers (Table S3) were designed from sequences retrieved from the RNA-Seq data. The reverse transcription step was performed on 1 µg of total RNA using Superscript III enzyme (Invitrogen, Carlsbad, CA, USA) and oligo-d(T) primer. Real-time quantitative PCRs were carried out using a LightCycler480 thermocycler and a LightCycler 480 SYBR Green I Master (both from Roche, Basel, Switzerland). The thermal cycling conditions were: 5 min at 95 • C; 55 cycles of 10 s at 95 • C; 15 s at 60 • C; 15 s at 72 • C; and a final step of 30 s at 40 • C. Primer specificity was checked by a melting curve at the end of each cycle. The results were normalized to the expression of the more stable gene defined with geNorm [53] and Norm Finder [54] software, namely, NEMFI. The primer efficiency was determined by LinRegPCR software (v.2017.0) [55]. We used the Pfaffl method [56] to calculate the relative expression, using the control samples as calibrators.

Metabolomic Analyses
Ultrahigh-pressure liquid chromatography and mass spectrometry (UHPLC-MS) analyses were performed as described previously [57], with some modifications. Metabolites were extracted with methanol (25 µL/mg frozen fresh leaves) supplemented with an external standard (apigenin, 1 mg/L). An amount of 50 mg of each sample were sonicated in an ultrasonic bath (Fisher Scientific, Hampton, NH, USA) for 15 min. After centrifugation at 12,000× g for 10 min, 200 µL of the supernatants were recovered for LC-MS analyses. The analyses were performed using a Dionex Ultimate 3000 UHPLC system (Thermo Fisher Scientific, Waltham, MA, USA). The chromatographic separations were performed on a Nucleodur C18 HTec column (150 × 2 mm, 1.8 µm particle size; Macherey-Nagel, Düren, Germany) maintained at 30 • C. The mobile phase consisted of acetonitrile/formic acid (0.1%, v/v) (eluant A), and water/formic acid (0.1%, v/v) (eluant B), at a flow rate of 0.25 mL/min. The gradient elution program was as follows: 0 to 4 min, 80% to 70% B; 4 to 5 min, 70% to 50% B; 5 to 6.5 min, 50% B isocratic; 6.5 to 8.5 min, 0% B; and 8.5 to 10 min, 0% B isocratic. The sample volume injected was 1 µL. The UHPLC system was coupled to an Exactive Orbitrap mass spectrometer (Thermo Fischer Scientific, Waltham, Viruses 2021, 13, 2138 5 of 23 MA, USA), equipped with an electrospray ionization (ESI) source operating in positive or negative mode. The parameters were set at 360 • C for the ion transfer capillary temperature, and 3500 V (positive mode) and 2500 V (negative mode) for the needle voltages. Nebulization with nitrogen sheath gas and auxiliary gas were maintained at 60 and 15 arbitrary units, respectively. The spectra were acquired within the m/z mass range of 110 to 1200 atomic mass units (a.m.u.), using a resolution of 50,000 at m/z 200 a.m.u. The system was calibrated internally using dibutyl phthalate as lock mass at m/z 279.1591, giving a mass accuracy lower than 1 ppm in the positive mode. The system was calibrated externally using calibration solution (Pierce™ LTQ ESI Negative Ion solution, Thermo Fisher Scientific), giving a mass accuracy lower than 5 ppm in the negative mode. The instruments were controlled using the Xcalibur software. The exact m/z and retention time of each metabolite were used for targeted metabolomic analyses using the Xcalibur software. A targeted metabolomic strategy was used to quantify the relative amounts of a total of 101 metabolites in grapevine leave extracts. The selected metabolites were grouped into eight major chemical or functional families, including amino acids and derivatives, flavonoids, hormones, hydroxycinnamic acids and derivatives, organic acids, proanthocyanidins, stilbenoids, and terpenoids. For some metabolites, identity was confirmed with the corresponding authentic standard provided by Sigma-Aldrich (Saint-Quentin-Fallavier, France) or Extrasynthese (Genay, France).

Differential Symptoms in Vitis vinifera cv. Gewurztraminer Infected by Five Different GFLV Strains
To gain insights into the GFLV symptom determinants, the vines infected with a single GFLV strain were produced using a two-step grafting procedure to separately transfer five GFLV strains (F13, GHu, B844, CO 1 and CO 2 , Table S1) from mechanically inoculated C. quinoa plants into scions of cultivar Chardonnay (Ch) or Gewurztraminer (Gw) grafted onto rootstock Kober 5BB. Non-inoculated graft assemblies served as controls. In the first trial (Trial 1, Table S2), six to eight vines of each cultivar-virus strain combination (for a total of 90 vines) were established in an experimental vineyard devoid of the nematode vector, X. index, that was maintained according to local commercial practices. From 2012 to 2014, the symptoms were visually scored and the fruits were collected at harvest time. All of the GFLV strains induced symptoms in both the Ch and Gw vines. On the Ch vines, all strains essentially caused flower abortion, known as coulure, leaf discolorations, such as mosaic and mottling, and a reduction in the fruit yield ( Figure 1, Table S4). The yield reduction was more variable over time than among strains ( Figure 1c, Table S5). On Gw vines, all strains induced coulure and very similar discolorations; however, strain B844 induced leaf deformation and a severe stunting (Figure 1b). These symptoms were already visible during the third growing season (in 2009) and persisted over time (Figure 2a,b). Moreover, strain B844 had a more severe effect on the fruit yield compared with the other strains, with the reduction consistently reaching approximately 80% (Figure 1d, Table S5). Since strain F13 behaved similarly to strains GHu, CO1, and CO 2 on both grapevine cultivars, subsequent work focused on GFLV-F13 and GFLV-B844.
Starting in 2012, we quantified the phenotypes of vines infected by GFLV-F13 or GFLV-B844, in addition to visually scoring the symptoms (Figure 1). Six to eight plants per plant-virus combination were monitored for at least three years. The results are presented as cumulative data in Figure 2 (and in detail, according to years, in Figure S1). On the Ch vines, both strains affected fruit production by reducing the mean cluster weight, the mean number of berries per cluster, and the mean weight of the berries, but not the mean number of clusters (Figure 2d-g). No significant difference was consistently observed between the two strains. On the contrary, strains F13 and B844 caused distinct phenotypes on the Gw vines. Strain F13 had fairly the same effect on Ch vines, whereas B844 caused a more pronounced reduction of the mean cluster weight and number of berries per cluster. It also occasionally caused a reduction of the mean number of clusters per plant (Figure 2d-g). The measurements of pruning wood weight in winter, as an indication of vine growth during the preceding season [58,59], revealed a differential effect of the two strains on Gw vines: while GFLV-B844 induced a 50% reduction compared with uninfected vines, GFLV-F13 increased the wood weight by a factor of 1.2 to 1.3 ( Figure 2h, Table S6). In summary, strain B844 had a more drastic effect than strain F13 on fruit production in Gw vines and drastically reduced growth.
To test whether the symptom severity was related to the virus titer, the level of virus accumulation was measured by semiquantitative DAS-ELISA in the young leaves of Ch and Gw scions (Figure 2i). Over the years, the virus titer was quite variable, but generally not significantly different, between the two strains (although GFLV-B844 always accumulated less than GFLV-F13), suggesting that virus accumulation is not directly associated with symptom severity. The control vines always tested negative for GFLV. Altogether, this work allowed us to identify B844 and F13 as severe and mild strains of GFLV on Gw grapevines.  Starting in 2012, we quantified the phenotypes of vines infected by GFLV-F13 or GFLV-B844, in addition to visually scoring the symptoms (Figure 1). Six to eight plants per plant-virus combination were monitored for at least three years. The results are presented as cumulative data in Figure 2 (and in detail, according to years, in Figure S1). On the Ch vines, both strains affected fruit production by reducing the mean cluster weight, the mean number of berries per cluster, and the mean weight of the berries, but not the mean number of clusters (Figure 2d-g). No significant difference was consistently observed between the two strains. On the contrary, strains F13 and B844 caused distinct phenotypes on the Gw vines. Strain F13 had fairly the same effect on Ch vines, whereas B844 caused a more pronounced reduction of the mean cluster weight and number of berries per cluster. It also occasionally caused a reduction of the mean number of clusters per plant (Figure 2d-g). The measurements of pruning wood weight in winter, as an indication of vine growth during the preceding season [58,59], revealed a differential effect of the two strains on Gw vines: while GFLV-B844 induced a 50% reduction compared with uninfected vines, GFLV-F13 increased the wood weight by a factor of 1.

RNA-Sequencing and Read Mapping on Virus and Grapevine Genomes
To shed light on the processes underlying the onset of mild or severe symptoms, the transcriptomes of F13-and B844-infected Gw grapevines were compared to those of the uninoculated controls. We sampled leaves after budburst, in the spring of 2016, on similar shoots and at a similar phenological stage (visible clusters, 4 to 6 expanded leaves). Three plants were analyzed per treatment: non-inoculated Gw scions (control, vines n • B31, B34, and C51), GFLV-F13 infected Gw scions (F13, vines n • B52, B53, and C38), and GFLV-B844infected Gw scions (B844, vines n • B47, C39, and C40). Representative pictures of the three  Figure S2A. To characterize the stunting symptoms, we counted the open leaves and measured the size of the leaves and shoots. The results ( Figure S2b) show that plants B47 and C40 infected by GFLV-B844 developed fewer leaves, and that these leaves, and their corresponding shoots, were shorter compared to non-inoculated plants. It is worth noting that the plants inoculated with strain B844 exhibited more variable phenotypes than the plants of the other two treatments.
Viruses 2021, 13, x FOR PEER REVIEW 7 of 25 to 1.3 ( Figure 2h, Table S6). In summary, strain B844 had a more drastic effect than strain F13 on fruit production in Gw vines and drastically reduced growth.
To test whether the symptom severity was related to the virus titer, the level of virus accumulation was measured by semiquantitative DAS-ELISA in the young leaves of Ch and Gw scions (Figure 2i). Over the years, the virus titer was quite variable, but generally not significantly different, between the two strains (although GFLV-B844 always accumulated less than GFLV-F13), suggesting that virus accumulation is not directly associated with symptom severity. The control vines always tested negative for GFLV. Altogether, this work allowed us to identify B844 and F13 as severe and mild strains of GFLV on Gw grapevines.  The total RNA extracted from the leaves was subjected to sequencing. For each plant, 52 to 85 million paired-end reads of 150 nucleotides were obtained, with the exception of plant B47 (B844-inoculated), for which we obtained 165 million reads, indicating an overamplification (Tables 1, S7 and S8).

Infection Status of the Plant Samples
To establish the virome of the vines, the trimmed reads were aligned to known virus and virus-like sequences found in grapevine. The vast majority of sequences matched to GFLV (Figure 3 and Table S7). Intra-lane contamination, with a low quantity of nonexpected GFLV sequences, as previously reported in [42,43], was found in all samples. In addition, grapevine rupestris stem-pitting-associated virus (GRSPaV), grapevine yellow speckle viroid (GYSVd), and hop stunt viroid (HSVd) were detected. This virus and two viroids are ubiquitous in the grapevine worldwide [60].
To further confirm the identity of GFLV sequences, the reads matching the GFLV genome were de novo assembled. The resulting contigs, corresponding to viral genomic sequences, were named after the identification of the plant they originated from. Phylogenetic trees were then built by aligning RNA1 and RNA2 coding sequences. The sequences retrieved from the plants inoculated with GFLV-F13 (B52, B53, and C38), or GFLV-B844 (B47, C39, and C40), were grouped together and shared more than 99.64% of their nucleotide sequence identity with the GFLV-F13 and GFLV-B844 reference sequences, respectively ( Figure 3b). Thus, the reads mapping to the GFLV sequences indicated a perfect correlation with the inoculated GFLV strain, and a genome diversity within the range observed earlier that is attributable to the quasi-species nature of the viruses [41,43].
Altogether, these results show that the analyzed samples were confidently infected with the expected GFLV strains, and no superinfection took place during the course of the experiment.

•
Mapping to the grapevine reference genome.
The sequence reads were aligned against the grapevine reference genome [61,62]. A total of 82 to 88% reads mapped to the grapevine genome (Tables 1 and S8) and 81 to 86% corresponded to uniquely mapped reads. A total of 77 to 82% reads were assigned to a PN40024 gene. Globally, reads were obtained for 25,255 genes, representing about 80% of the total (31,842) annotated genes of the reference genome.
With the exception of plant C40 (GFLV-B844-infected), the samples were grouped according to the experimental conditions, as seen by the principal component analysis (PCA) based on the read counts after variance-stabilized transformation (VST) (Figure 4a). Variability was observed between the biological replicates of a given treatment. However, the treatments were more discriminant than the replicates. It is worth noting that the differences were more prominent between the control and infected vines than between Viruses 2021, 13, 2138 9 of 23 vines inoculated by the two GFLV strains. Because sample C40 was an outlier, we decided to exclude it from further analyses. The estimation of dispersion showed that no information was lost by eliminating this sample, and a narrower estimated dispersion resulted (Figure 4b,c).
Phylogenetic trees were then built by aligning RNA1 and RNA2 coding sequences. The sequences retrieved from the plants inoculated with GFLV-F13 (B52, B53, and C38), or GFLV-B844 (B47, C39, and C40), were grouped together and shared more than 99.64% of their nucleotide sequence identity with the GFLV-F13 and GFLV-B844 reference sequences, respectively (Figure 3b). Thus, the reads mapping to the GFLV sequences indicated a perfect correlation with the inoculated GFLV strain, and a genome diversity within the range observed earlier that is attributable to the quasi-species nature of the viruses [41,43].
Altogether, these results show that the analyzed samples were confidently infected with the expected GFLV strains, and no superinfection took place during the course of the experiment.   Table S1 for accession numbers).
To better understand the processes underlying the stunting phenotype of vines infected with GFLV strain B844, we analyzed the differentially expressed genes (DEGs) in the GFLV-F13-and GFLV-B844-infected vines compared to the controls. From the 25,255 genes identified, low counts, representing 4407 genes, were filtered out. A total Viruses 2021, 13, 2138 10 of 23 of 3981 genes were differentially expressed (FDR ≤ 0.05), with 1458 genes (4.6% of the annotated PN40024 12X.v2 genes) deregulated by both virus strains (Figure 5a). This set of genes corresponded to the core genes reprogramed during GFLV infection. Noteworthy, from the 1458 shared genes, a single gene (VIT_208s0007g01690, predicted as a probable sugar phosphate translocator) was deregulated in the opposite direction, upregulated in the F13-and downregulated in B844-infected Gw. Thus, the phenotypic differences between the two strains are mainly associated with differentially deregulated genes rather than to the direction of the regulation of common genes.


Mapping to the grapevine reference genome.
The sequence reads were aligned against the grapevine reference genome [61,62]. A total of 82 to 88% reads mapped to the grapevine genome (Tables 1 and S8) and 81 to 86% corresponded to uniquely mapped reads. A total of 77 to 82% reads were assigned to a PN40024 gene. Globally, reads were obtained for 25,255 genes, representing about 80% of the total (31,842) annotated genes of the reference genome.
With the exception of plant C40 (GFLV-B844-infected), the samples were grouped according to the experimental conditions, as seen by the principal component analysis (PCA) based on the read counts after variance-stabilized transformation (VST) (Figure 4a). Variability was observed between the biological replicates of a given treatment. However, the treatments were more discriminant than the replicates. It is worth noting that the differences were more prominent between the control and infected vines than between vines inoculated by the two GFLV strains. Because sample C40 was an outlier, we decided to exclude it from further analyses. The estimation of dispersion showed that no information was lost by eliminating this sample, and a narrower estimated dispersion resulted (Figure 4b,c).
To better understand the processes underlying the stunting phenotype of vines infected with GFLV strain B844, we analyzed the differentially expressed genes (DEGs) in the GFLV-F13-and GFLV-B844-infected vines compared to the controls. From the 25,255 genes identified, low counts, representing 4407 genes, were filtered out. A total of 3981 genes were differentially expressed (FDR ≤ 0.05), with 1458 genes (4.6% of the annotated PN40024 12X.v2 genes) deregulated by both virus strains (Figure 5a). This set of genes corresponded to the core genes reprogramed during GFLV infection. Noteworthy, from the 1458 shared genes, a single gene (VIT_208s0007g01690, predicted as a probable sugar phosphate translocator) was deregulated in the opposite direction, upregulated in the F13and downregulated in B844-infected Gw. Thus, the phenotypic differences between the A total of 2481 and 2958 genes were deregulated by GFLV-B844 and GFLV-F13, respectively. Whereas up-and downregulated genes were almost in even proportion following GFLV-F13 infection (48.2% upregulated vs 51.8% downregulated), a greater bias was observed towards downregulated genes following GFLV-B844 infection (40.4% upregulated vs 59.6% downregulated). Most of the significant changes (around 77%) are contained in a −1 and +1 Log2FC range (Figure 5b).

•
Pathways differently affected by the two GFLV strains.
To identify pathways differently affected by the two GFLV strains, we further analyzed the DEGs with a special focus on the 1500 and 1023 genes specifically deregulated in the F13 and B844 treatments compared to the controls, respectively (Figure 5a). In order to decipher the affected biological processes, we performed a Gene Ontology (GO) terms enrichment analysis ( Figure S3). The core DEGs shared by the F13 and B844 conditions ( Figure S3a) belong to general-biological-process GOs, such as translation (ribosome biogenesis), growth (cell proliferation, regulation of meristem growth), and sugar metabolism (maltose metabolic process, starch biosynthetic process).
Globally, B844 DEGs ( Figure S3d) focused on expression (positive regulation of transcription, DNA and RNA processing, translation), metabolism (carotenoid biosynthetic process), and defense pathways (response to chitin, regulation of plant-type hypersensitive response, RNA processing). The DEGs of the F13 treatment ( Figure S3b) belong to expression pathways (DNA replication, RNA methylation) and the cell cycle and growth processes.
Remarkably, B844-specific DEGs (Figure 5c) participate in defense pathways, such as the regulation of HR, which is known to occur in plants in response to virus infection alongside specific symptom development.
F13-specific DEGs ( Figure S3c) direct reprogramming to focus on replication, cellular development, and growth, and could be directly linked to the virus hijacking the cellular machinery for its multiplication.
These first results point to an HR-like response specifically occurring in Gw upon GFLV-B844 infection.
Viruses 2021, 13, x FOR PEER REVIEW two strains are mainly associated with differentially deregulated genes rather th direction of the regulation of common genes.
A total of 2481 and 2958 genes were deregulated by GFLV-B844 and G respectively. Whereas up-and downregulated genes were almost in even p following GFLV-F13 infection (48.2% upregulated vs 51.8% downregulated), bias was observed towards downregulated genes following GFLV-B844 infectio upregulated vs 59.6% downregulated). Most of the significant changes (around contained in a −1 and +1 Log2FC range (Figure 5b).  Pathways differently affected by the two GFLV strains.
To identify pathways differently affected by the two GFLV strains, w analyzed the DEGs with a special focus on the 1500 and 1023 genes sp deregulated in the F13 and B844 treatments compared to the controls, respective

•
Validation of the RNA-Seq differential genes expression by RT-qPCR.
To validate the RNA-Seq results, we randomly selected 20 genes (AFC1, APL1, BCP, CDT1A, GAPDH, GPAT6, GRP, KCS5, NA_420, NA_1090, NA_1410, NEMFI, NQO1, PDF1, PDF2, PP2A, PP7, RD22, TIP41, and SYP121) with different expression levels and regulation profiles and analyzed their expression by RT-qPCR using control samples as calibrators ( Figure S4). Seven of these genes (NEMFI, GAPDH, PP2A, NQO1, TIP41, AFC1, PP7) were tested as potential reference genes using geNorm and Normfinder algorithms [53,54] to calculate their stability across all samples. All seven genes were indeed stable, with TIP41 and NEMFI being the most stable across our eight samples. As already described in other virus-host pathosystems [63][64][65], GAPDH, which is commonly used as reference, did not perform well in our experimental setup. The RT-qPCR results of the remaining 19 genes were then analyzed using NEMFI as the reference gene. The overall expression profiles obtained by RT-qPCR were consistent with the RNA-Seq data ( Figure S4), although the degree of modulation differed between the two techniques ( Figure S4a-p). The results were contradictory only for three genes (APL1, GAPDH, and NA_1090, Figure S4q-s), possibly because of the higher variability of the RT-qPCR results for the F13 compared with the B844 treatment for these genes.

Biological Validation of an HR-Like Pathway Activation by GFLV-B844 in Gw Vines Expression Analysis of Candidate Genes in Relation to PLANT Defense Pathways
In order to test the differential deregulation of the plant-type hypersensitive response by the two strains of GFLV, we extended our study to samples from the same field trial (Trial 1, Table S2), collected at two different time points, and samples from a second trial, established in a greenhouse (Trial 2). The symptoms developing on the plants from Trial 2 were consistent with our previous observations, with B844-and F13-infected vines showing stunting and mild symptoms, respectively. In total, 54 samples (36 from the field trial, including those analyzed by RNA-Seq, and 18 from the greenhouse) were subjected to RT-qPCR assays. Candidate genes (NDF6, AGAL2, NF-YC4, AIL6, ERF4, SYP121, and CYP94) were selected for their specific deregulation by GFLV-B844, with the highest induction or repression rates and/or their belonging to the GO "Regulation of hypersensitive response" (ERF4, SYP121, and CYP94; Supplemental Table S9). We also chose PIN5 and ROXY2, which are deregulated by the two virus strains, but more substantially by B844 than F13. Finally, we chose PR10 and RDR1, two defense genes that showed a low induction by both viruses in our transcriptomic analysis.
Significant variations in the expression of all eleven genes selected for this study were apparent in relation to the environmental conditions and the phenological stages, suggesting highly dynamic processes in gene regulation (Figures 6 and S5). However, some differences were virus-strain-specific, as shown by the RT-qPCR results analyzed either by gene (Figure 6a), or globally by the PCA of the eleven genes that enables the discrimination of the virus strains, regardless of the trial and time point considered (Figure 6b). Notably, the expression of PR10 (Figure 6a) appeared differently regulated by strains B844 and F13. Altogether, these results illustrate that some genes involved in the HR, or its regulation, are differently expressed in plants infected with GFLV-B844 compared with plants infected with GFLV-F13.
To gain further insights into the grapevine response to GFLV infection, a metabolomic analysis was performed by UHPLC-MS on leaf samples collected at different time points from all three treatments grown in three different trials (Table S2). In total, 112 samples from 36 plants were subjected to a targeted metabolomic analysis in order to quantify the relative amounts of 101 compounds belonging to major chemical or functional families, including amino acids and derivatives, flavonoids, hormones, hydroxycinnamic acids and derivatives, organic acids, proanthocyanidins, stilbenoids and terpenoids. A pairwise comparison of the significant metabolite changes in the control leaves and leaves from the GFLVinfected vines revealed major differences in the plant responses to the B844 or F13 strains (Figures 7 and S6). The leaves of the control and the GFLV-F13-infected vines exhibited very similar profiles, indicating that infection by this mild GFLV strain had very little impact on the selected leaf metabolites. Similarly, infection with GFLV-B844 had no significant effect on the majority of the selected metabolites. However, GFLV-B844 impacted a small number of metabolites belonging to specific families, when compared with both the control and GFLV-F13-infected plants. GFLV-B844 had a moderate impact on some metabolites belonging to flavonoids and hydroxycinnamic acids. Interestingly, both salicylic acid (SA) and zeatin were significantly increased by factors of 5 and 2.5, respectively, in GFLV-B844-inoculated vines, whereas jasmonic acid (JA) remained unaffected. Nevertheless, the most remarkable impact of GFLV-B844 infection was a very significant accumulation of phytoalexins, such as piceid and resveratrol, with a sixty-fold increase in the transresveratrol amount compared with the control or GFLV-F13-infected plants.
Viruses 2021, 13, x FOR PEER REVIEW 13 o Significant variations in the expression of all eleven genes selected for this study w apparent in relation to the environmental conditions and the phenological stag suggesting highly dynamic processes in gene regulation (Figures 6 and S5). Howev some differences were virus-strain-specific, as shown by the RT-qPCR results analyz either by gene (Figure 6a), or globally by the PCA of the eleven genes that enables discrimination of the virus strains, regardless of the trial and time point conside (Figure 6b). Notably, the expression of PR10 (Figure 6a) appeared differently regulated strains B844 and F13. Altogether, these results illustrate that some genes involved in HR, or its regulation, are differently expressed in plants infected with GFLV-B compared with plants infected with GFLV-F13.  Table S2 for a description of the samples). Error bars = standard deviation. Statistically significant differences (Mann-Whitney rank test) are indicated with asterisks: *** p < 0.005. (b) Principal component analysis (PCA) of the expression level of the eleven tested  Table S2 for a description of the samples). Error bars = standard deviation. Statistically significant differences (Mann-Whitney rank test) are indicated with asterisks: *** p < 0.005.   Table S2 for a precise description of the samples). Statistical analysis was performed using Tukey's honest significant difference method, followed by a false discovery rate (FDR) correction, with FDR < 0.05. For FDR ≥ 0.05, Log2 fold changes were set to 0. Glycosides are indicated as glucoside (Glc), apiosylglucoside (GlcAp), glucuronide (GlcA), rhamnoside (Rha), or  Table S2 for a precise description of the samples). Statistical analysis was performed using Tukey's honest significant difference method, followed by a false discovery rate (FDR) correction, with FDR < 0.05. For FDR ≥ 0.05, Log2 fold changes were set to 0. Glycosides are indicated as glucoside (Glc), apiosylglucoside (GlcAp), glucuronide (GlcA), rhamnoside (Rha), or rutinoside (Rut). * indicates that metabolite identity was confirmed with the corresponding authentic standard. The application of the cytokine zeatin has been shown to decrease symptoms caused by the bacteria, Pseudomonas syringae, in tobacco plants, in relation to increased levels of SA accumulation [66]. Resveratrol and its glycosylated derivatives, trans-piceids, have been shown to have antimicrobial properties, among others [67], and resveratrol has been more precisely associated to the hypersensitive response in Vitis cells [68][69][70].
Altogether, these results suggest that GFLV-B844, the severe strain, stimulates the chronic accumulation of specific metabolites related to the plant defense pathways in a way typical of a biotrophic pathogen.

•
Virus restriction HR is mostly associated with a restriction of the pathogen's systemic spread within the plant, although the induction of HR has been described to be insufficiently effective at limiting the propagation of some viruses [35,71]. In another field trial, established in 2013 (Trial 4, Table S2), some B844-inoculated vines showed an unusual phenotype (Figure 8a

Discussion
GFLV causes variable symptoms in the vineyard, depending on the virus strain, grapevine genotype, cultural practices, and environmental conditions [28]. A study of the virus symptomatology on this perennial woody fruit crop in the vineyard is complicated by the frequency of multiple virus infections. In this work, we describe a unique experimental setting where two grapevine genotypes were graft-inoculated with five distinct GFLV strains and monitored during three to five growing seasons. This setting allowed us to distinguish the GFLV strain, B844, which caused a drastic yield reduction To check for the presence of the virus, three randomly chosen symptomless plants were further analyzed by RT-qPCR and northern blot using RNA extracted from cortical scrapings, after removing the bark of: (i) the rootstock below the graft union; (ii) the scion immediately above the graft union; and (iii) the leaves of the scion. The presence of viral RNAs and their catabolic products (viral small RNAs, vsiRNAs) was investigated by RT-qPCR and northern blot, respectively (Figure 8b,c). VsiRNAs originating from both RNA1 and RNA2 were detected in the rootstock and in the scion immediately upstream of the graft union point but were absent from the apical leaves (Figure 8b). These results were confirmed by RT-qPCR analyses (Figure 8c), showing that the severe strain, GFLV-B844, was present in all rootstocks but was unable to translocate to the upper parts of the scion of some plants. In these plants, the virus was efficiently restricted to the rootstock and the plants remained symptomless, showing a high degree of resistance toward the severe strain.

Discussion
GFLV causes variable symptoms in the vineyard, depending on the virus strain, grapevine genotype, cultural practices, and environmental conditions [28]. A study of the virus symptomatology on this perennial woody fruit crop in the vineyard is complicated by the frequency of multiple virus infections. In this work, we describe a unique experimental setting where two grapevine genotypes were graft-inoculated with five distinct GFLV strains and monitored during three to five growing seasons. This setting allowed us to distinguish the GFLV strain, B844, which caused a drastic yield reduction and stunted growth in Gw grapevines from other strains. From these observations, GFLV-B844 is considered hyper-aggressive on Gw, whereas F13 is a milder strain, thus partially confirming previous results (Legin et al., 1993). The virus titer did not consistently and significantly vary with the viral strain and phenotype, showing that the virus concentration cannot explain the severity of the GFLV-B844-induced symptoms.
We compared the transcriptomic changes induced by the severe strain, GFLV-B844, and the mild strain, F13, using HTS because (i) it is an unbiased approach to quantifying gene expression, and (ii) it can ascertain the infection status of grapevines. Apart from some GRSPaV, GYSVd, and HSVd reads, we did not retrieve any unexpected sequences from virus or viroid origin, validating our conditions to produce vines with single-GFLV-strain infections, and controlled settings that prevented a superinfection of the experimental vines maintained in the open and in the greenhouse. Thus, although we cannot completely rule out a synergistic effect of GRSPaV, GYSVd, or HSVd on GFLV symptoms, it is very likely that the GFLV-B844 is responsible for the more severe symptoms on Gw vines. This represents the first example of a study on virus-induced symptoms in grapevine with a nontargeted search for viral sequences. By comparing the transcriptomic changes caused by the hyper-aggressive strain, B844, with those caused by the milder strain, F13, we identified 2481 deregulated unigenes, 1458 of which were shared by the two strains, and 1023 that were specifically modulated by strain B844 (419 induced and 604 repressed genes). The RNA-Seq results were confirmed by RT-qPCR on 16 genes out of the 19 randomly chosen gene candidate targets. Although an outlier plant was detected and removed from further analyses, no DEG was lost.
Among the GOs of deregulated genes found in this work, some are often found in the transcriptomic profiling of viral infections. This is the case for genes involved in photosynthetic processes, ribosome biogenesis, sugar metabolisms, and translation [72][73][74]. The perturbation of photosynthesis and the chloroplasts has already been described for viral infections in general [75,76], and GFLV in particular [77,78]. The stunting symptoms of strain B844 could be consistent with the repressed genes implicated in the regulation of meristem growth. The two top GOs of the 43 B844-DEGs (namely, the response to chitin and the regulation of the plant-type hypersensitive response) are related to plant defense and, more precisely, to the hypersensitive-type response. This was of particular interest because it led to the attractive hypothesis that HR might have taken place in GFLV-B844-infected 'Gewurztraminer' vines. This hypothesis is reinforced by the specific accumulation of the major vine phytoalexin, resveratrol, and the signaling hormone, SA, in GFLV-B844inoculated Gw vines, and by the limited virus spread in some plants. Phytoalexins are low molecular mass secondary metabolites with large spectrum antimicrobial activities. Resveratrol, the main phytoalexin in the Vitis spp., has been associated with hypersensitive or hypersensitive-like responses in Vitis cells upon fungus elicitation [68][69][70]. An antiviral effect of resveratrol or its derivatives has been mentioned for human cells infected with the now-called SARS-CoV-1 [79] and, very recently, against tobacco mosaic virus infection in herbaceous plants [80]. Whether resveratrol or trans-resveratrol has a direct antiviral effect and contributes to the success of HR against GFLV remains to be tested.
HR occurs when the product of a cellular resistance gene (R) interacts with a pathogenic avirulence factor (Avr). This interaction generally leads to cell death and a limitation of the virus movement within the plant [81][82][83][84]. However, cell death and virus restriction represent separate processes that have been temporally, spatially, or genetically uncoupled for a few plant-virus pathosystems [71,[85][86][87][88][89]. The efficiency of resistance was suggested to be correlated with the speed and the intensity of the host response. A rapid reaction of the plant would thus promote an efficient restriction of the virus without severe necrosis, whereas a slower reaction would be unable to stop virus movement, and this may increase the intensity of the reaction and result in systemic HR. Cell death could, therefore, be regarded as a side effect of a delayed defense response [71]. No necrosis was observed in our pathosystem. However, the virus entered the Gw genotype through the vascular system via the graft-inoculation procedure used in this study. Thus, we do not know whether necrosis took place in some plants or not.
A study of the specific role of deregulated genes could now give a better picture of the reactions leading to the plant's response and the observed phenotype. This specific involvement is difficult to predict from the literature because (i) our pathosystem results in a long-lasting contact between the virus and its host and it is, therefore, difficult to evaluate an induction step from a feed-back regulation of the induced reaction, and (ii) some genes have been described with the opposite function in different pathosystems. This is well-exemplified by the transcription factor, WRKY 11, found to be decreased in the GFLV B844-infected grapevines of the present work. It has recently been established as a positive regulator of resistance to Botrytis cinerea in strawberry fruit [90], and is known like WRKY7, as a negative regulator of resistance to P. synringae in Arabidopsis thaliana [91,92]. Thus, although some of the DEGs have described functions in A. thaliana regarding stress responses, it is very speculative to predict their role in our pathosystem.
Timing and R-Avr interaction (affinity or quantity) are predicted to determine the outcome of the plant-virus interaction in a continuum of possible responses, ranging from extreme resistance to systemic HR [71]. Hence, HR and its downstream regulation are not contradicted by the systemic infection observed in many of the plants of our experiment. The grafting procedure represents a high-pressure virus inoculation method [18] that could influence the onset and efficiency of the defense response. Moreover, we might have introduced a strong bias against the proportion of plants actually restricting the virus because we generally select plants for the virus presence after grafting and before greenhouse or field establishment. An HR response to GFLV, only capable of efficiently restricting the virus in a fraction of the inoculated plants, has been shown to occur in Nicotiana occidentalis when inoculated with strain F13. It is not known what differs between N. occidentalis plants, where resistance occurs, and plants showing a so-called trailing HR, where the pathogen is not completely stopped and initiates a reaction in all the tissues where it moves. In this herbaceous model plant, the molecular markers of the HR were detected starting three days post inoculation [35]. Because the inoculation procedure of grapevines requires the in vitro heterologous grafting of a rootstock onto an infected C. quinoa stem, followed by the herbaceous grafting of a grapevine cultivar onto the infected rootstock, the experiments lasted several years after the inoculation. This may explain why our transcriptomic analysis did not show an induction of the characteristic HR genes, for instance, the homologs of HSR203J, NDR1, or EDS1, as these genes are expressed at early time points after pathogen challenge [93]. To confirm that GFLV-B844 induces an HR in V. vinifera cv. Gewurztraminer, a model of this pathosystem, allowing a direct inoculation of the virus, needs to be developed.
Stunting, as a consequence of plant defense in perennial or woody plants, where mechanistic data is scarce, has only begun to be mentioned in the literature. A recent paper describes the upregulation of genes implicated in the biosynthesis of phytoalexins in response to apple replant disease [94]. Another example is a stunting phenotype that has been described in hop plants infected with a viroid [95]. These plants show a deregulation of HR-related genes. Here, by comparing plants infected with a mild and a severe GFLV strain, we provide an additional indication that HR could be associated with severe stunting symptoms. A comparison of the plants exhibiting effective virus resistance with the plants presenting an ineffective defense response can now be envisioned in both the herbaceous model plant, N. occidentalis, and the woody crop grapevine. This should help understand what component of the HR response is actually responsible for the virus resistance.

Conclusions
To conclude, our work used a unique experimental setting of mono-infected vines for the identification of GFLV strains with distinct symptoms on Gewurztraminer. This has provided the tools to compare the transcriptome of plants infected by a severe (B844), and a mild strain (F13), of GFLV in order to shed light on the mechanisms underlying pathogenesis. From the specifically deregulated genes identified in our study, we hypothesize that an HR-type reaction might be induced by the severe strain causing stunting. The accumulation of SA and phytoalexins, and the partial resistance observed in some GFLV-B844 infected grapevines, strengthen our hypothesis. This defense reaction might lead to either a severe stunting or to resistance, most probably by the action of SA, which is involved in the defense towards viral diseases, as well as in dwarfing symptoms during virus infection in herbaceous plants [75]. Thus, SA likely represents the major player in the growth/immunity tradeoff in grapevine, in a manner comparable to what has been described in Arabidopsis [96,97]. To the best of our knowledge, this is the first indication of HR in virus-infected grapevines and the potential mechanisms underlying stunting in a perennial woody fruit crop.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/ 10.3390/v13112138/s1: Figure S1: Annual comparison of grapevines infected with GFLV strains F13 and B844; Figure S2: Phenotype of the plants subjected to RNA-Seq analysis; Figure S3: GO categorization of differently regulated genes; Figure S4: Comparison of RNA-Seq and qRT-PCR determination of transcript accumulation; Figure S5: Transcript accumulation of five candidate genes, not related to defense, in infected vines grown in different conditions and collected at different time points; Figure S6: Principal component analysis of the global impact of GFLV strains F13 and B844 on leaf metabolites of GFLV-infected vines; Table S1: GenBank accessions of grapevine fanleaf virus strains used in this study; Table S2: Trials and plant analyzed in this study; Table S3: Primers used in RT-qPCRs performed on cellular transcripts; Table S4: Annual symptom scores of the five GFLV strains on the two grapevine cultivars; Table S5: Annual and cumulative fruit yield reduction caused by the five GFLV strains on the two grapevine cultivars; Table S6: Effect of strains F13 and B844 on the growth of 'Gewurztraminer' and 'Chardonnay' grapevines; Table S7: Read counts and RPKM of viral and viroidal sequences obtained in the RNA-Seq experiment; Table S8: Details on excluded reads; Table S9: Grapevine genes specifically deregulated in response to GFLV-B844 infection of 'Gewurztraminer' vines.