A More Complete Story of the Genetic Bases of Resistance to the Rice Hoja Blanca Virus

Rice hoja blanca is one of the most serious diseases in rice growing areas in tropical Americas. Its causal agent is the Rice hoja blanca virus (RHBV), transmitted by the planthopper Tagosodes orizicolus Müir. Genetic resistance is the most effective and environment-friendly way of controlling the disease. So far, only one major quantitative trait locus (QTL) of Oryza sativa ssp. japonica origin, qHBV4.1, that alters incidence of the virus symptoms in two Colombian cultivars has been reported. This resistance has already started to be broken, stressing the urgent need for diversifying the resistance sources. In the present study we performed a search for new QTLs of O. sativa indica origin associated with RHBV resistance. We used four F 2:3 segregating populations derived from indica resistant varieties crossed with a highly susceptible japonica pivot parent. Beside the standard method for measuring disease incidence, we developed a new method based on computer-assisted image processing to determine the affected leaf area (ALA) as a measure of symptoms severity. Based on the disease severity and incidence scores in the F 3 families under greenhouse conditions, and SNP genotyping of the F 2 individuals, we identied four new indica QTLs for RHBV resistance on rice chromosomes 4, 6 and 11, namely qHBV4.2 WAS208 , qHBV6.1 PTB25 , qHBV11.1 and qHBV11.2. We also conrmed the wide-range action of qHBV4.1. Among the ve QTLs, qHBV4.1 and qHBV11.1 had the largest effects on incidence and severity, respectively. These results provide a more complete understanding of the genetic bases of RHBV resistance in the cultivated rice gene pool, and can be used to develop marker-aided breeding strategies to improve RHBV resistance. The power of joint- and meta- analyses allowed precise mapping and candidate genes identication, providing the basis for positional cloning of the two major QTLs qHBV4.1 and qHBV11.1.


Introduction
The rice hoja blanca (RHB) disease is one of the most important constraints to rice productivity in the tropical zone of Americas, causing yield losses in many countries, including Colombia, Costa Rica, Ecuador, Guyana, Panama, Peru, Dominican Republic, Nicaragua and Venezuela (Morales and Jennings 2010). Its causal agent is the Rice hoja blanca virus (RHBV), a Tenuivirus transmitted by the planthopper Tagosodes orizicolus M. (Hemiptera:Delphacidae). Genetic resistance to both the virus and its vector insect is key for a successful, environment-and consumers health-friendly, integrated crop management.
No real immunity has been found in the cultivated rice germplasm, and even in resistant materials, the plantlets (< 10 days old) can show susceptibility to the virus. Nonetheless, a handful of varietiesincluding the two Colombian cultivars, Fedearroz 50 and Fedearroz 2000 -with good resistance level have been bred in the past and have been important to stabilize rice production. Fedearroz 2000 is still the most resistant commercial variety, however, the incidence of the disease has increased under eld and controlled conditions. Fedearroz 50 has turned virus-susceptible in the past years in the eld, probably due to virus mutations that allowed it to overcome the resistance. Thus, there is an urgent need for diversifying the sources of genetic resistance in order to breed new varieties with more durable resistance . Genetic resistance to RHB disease can be decomposed into resistance to the virusitself, and resistance to its insect vector. In the present study, we focus on resistance to the virus.
In a previous study, we reported a major QTL on the short arm of chromosome four for resistance to RHBV, shared by Fedearroz 50 and Fedearroz 2000, abbreviated as FD 50 and FD 2000 in this paper (Romero et al. 2014). This QTL, called here qHBV4.1, controlled RHBV incidence, measured as the percentage of plants that show symptoms of the viral infection, no matter the level of the symptoms.
Incidence is of course an important parameter of the epidemics of a disease. Yet, its severity is certainly as much as important: if severity is low, a high incidence might have no signi cant impact on plant viability, panicle development, or grain yield. Additional to RHBV incidence, we thus designed -and report in this study -new experiments to decipher the genetic control of RHBV resistance seen as symptoms severity, measured by computer-aided image processing of the affected leaf area (ALA).
Moreover, a local ancestry analysis showed that, although FD 50 and FD 2000 are mostly of Indica genetic background, qHBV4.1 in those two cultivars was found to be of Japonica origin, and was surrounded by several hundred kilobase pairs of Japonica DNA. We found desirable to search for Indica resistance QTLs in order to ensure better compatibility with tropical irrigated materials in breeding programs, and also to increase the diversity of the available sources of genetic resistance. FD 2000). The main selection criteria were (1) selected sources should show a consistent, high resistance to RHBV and (2) they should cover the genetic diversity spectrum of the indica cluster. Then, the four resistance sources were crossed with the highly susceptible japonica line Bluebonnet 50 (IRGC 121874) (abb. BBT 50), using BBT 50 as the male parent. Candidate F 1 hybrids were checked for self-pollination of the mother plant with 48 SNP (single nucleotide polymorphism) markers distributed along the rice genome. F 2 (S 1 ) populations were derived from self-pollination of the veri ed F 1 hybrids. Each F 2 plant was self-pollinated to produce F 3 (S 2 ) families. All panicles involved in crosses or self-pollination were bagged to avoid out-crossing.

Evaluation of RHBV resistance
To evaluate the level of resistance -or susceptibility -of the F 3 plants, we looked at two complementary aspects of the disease: the incidence and the severity of the symptoms. Incidence is simply de ned by the proportion of diseased plants in a population, while severity is the area or volume of plant tissue that is visibly diseased (Campbell and Neher 1994 Virulent insect colonies T. orizicolus virulent insects were obtained from colonies maintained at CIAT. The RHBV harboring colony contained insects that were fed on RHBV-infected plants and allowed to reproduce on BBT 50. To determine the percentage of virulent insects in this colony, 200 individual nymphs were tested for virulence on separate, caged, RHB-susceptible 8-day-old seedlings.

Incidence
Here we consider incidence as the percentage of plants showing any level of RHBV symptoms. We consider the absence of symptoms as a sign of absence of infection, not as extreme tolerance. F 3 materials and controls were planted in plastic trays containing 17 furrows of 20 plants each. Each tray contained a furrow of each parent -BBT 50 also served as a susceptible control -and the controls FD 2000 (resistant) and Colombia 1 (intermediate). The trays were placed in a mesh cage 18 days after sowing, and infested by mass release of T. orizicolus with virulence between 50% and 65%, with an average of four nymphs per plant. The nymphs were allowed to feed for three days on the plants, after which they were eliminated with water rinsing. A randomized complete block design was used with three replicates (20 plants from each F 3 family per replicate), where each block represented a cage. Incidence assessment was made 35 days after infestation (DAI) by counting the plants showing disease symptoms, per row.

Severity
In order to test the severity of RHBV, a novel methodology based on the affected leaf area (ALA) by the virus was developed using an image processing approach. Brie y, images of infected plants were taken using a reference scale (a ruler) and a contrasting background under homogeneous light exposure. Raw images of plants with symptoms ( Figure 1A) are processed using the ImageJ software in two steps: in the rst step, brightness is decreased to eliminate the areas affected by the disease ( Figure 1B). Then, the image is binarized, i.e., converted to black and white ( Figure 1C). The number of black pixels is thus representing the healthy area of the leaf, according to the previously calibrated reference scale. In the second step, a binarization process allows to calculate the total area ( Figure 1D). The affected area is then calculated as the difference between total and healthy areas.

Statistical treatment
Since ALA was measured as the ratio of affected/total leaf area for each family in each block, a weighted ALA was calculatedwith a generalized linear mixed model (MLGM) using the gamma distribution. The incidence variable was analyzed with a MLGM using the binomial distribution. The MLGM procedure is used for variables that are not necessarily normally distributed. For each variable, an analysis of variance was performed to test for differences between families and to evaluate the effect of the blocks.
Subsequently, adjusted means were calculated for each family and a mean comparison test was performed to identify those families that were more resistant or equal to the resistant parent. The

DNA extraction
Leaf tissue was collected from 15-days-old F2 plants. Samples were frozen in liquid nitrogen and stored at -80°C until processed. Plant DNA was isolated in 96-racked tubes using a modi ed version of a method previously described (Risterucci et al. 2000) as follows: 480 µl extraction buffer was added to 150 mg ground frozen leaf tissue. The buffer was 100 mM Tris (pH = 8.0), 2M NaCl, 20 mM EDTA (pH = 8.0), MATAB 2%, sodium bisul te 0.5%, and PEG 8000 1%. This mixture was homogenized and incubated in a water bath at 74°C for 30 min. Subsequently, 480 µl chloroform:isoamyl-alcohol (24:1) was added and the mixture was centrifuged at 3000rpm. Supernatants were precipitated with 270 µl isopropanol at -20°C for one hour and centrifuged at 3000rpm. The pellets were washed with 200 µl 80% ethanol and allowed to dry at 40°C by inverting the tubes for one hour. DNA was resuspended in Tris-EDTA [10 mM Tris-HCl, pH8 and 1 mM EDTA, pH8) containing 40µg/ml of RNAse A and spectrophotometric quanti cation was done using a multimode plate reader Synergy H1 (Biotek). DNAs were normalized at 60ng/µl for subsequent processing.
Genotype scoring F 2 individuals and parents were scored for polymorphic SNPs in each population using the Fluidigm nano uidic genotyping platform, according to the manufacturer's protocols. Brie y, pre-ampli cation of targets regions was done in a 5µl multiplex-PCR reaction, used as template for allele speci c-PCR that was conducted in 48 ´ 48 integrated uidic circuit (IFCs) of the Fluidigm ® genotyping platform. Fluorescent images of the IFCs were acquired in an EP1 reader and analyzed with the Fluidigm Genotyping Analysis Software.

Linkage maps construction
Genetic maps were calculated with MapDisto 2.0 (Lorieux 2012; Heffel nger et al. 2017) (http://mapdisto.free.fr). Goodness-of-t to Mendelian segregation (1:2:1) was tested for each marker by computing the chi-squared (c2) statistic with the 'Segregation c2s' function. As the SNP markers were de ned from WGS, we kept their order on the physical map or rice. For each cross separately, we checked the data for singletons (e.g., the "B" in "AAAAAAABAAAAAA" is a singleton) with the 'Replace errors by anking genotypes' function, with a maximum probability of a singleton to occur of 0.001. A few missing data were inferred using the 'Replace missing data by anking genotypes' function, with the same threshold. Recombination fractions were converted to centimorgans (cM) with the Kosambi mapping function (Kosambi 1944).

QTL mapping
Data les were prepared using the 'Export map and data' function of MapDisto 2.0. Analyses of distribution of the phenotypic traits as well as QTL detection were performed using the Qgene 4.0 program (Joehanes and Nelson 2008) (http://www.qgene.org). For QTL detection, the LOD score statistic was calculated with the following different methods were compared: single-marker regression (SMR), simple interval mapping (SIM) and composite interval mapping (CIM). The forward cofactor selection option was used in CIM. Empirical thresholds to declare the presence of a QTL were obtained using the resampling by permutation method, performing 10,000 (SMR) or 1,000 (CIM) iterations for each traitchromosome combination. In order to correct for possible erroneous phenotypic data corresponding to escape, mis-scoring, or incomplete penetrance, all positive QTLs were additionally con rmed by analysis of outliers in the trait distribution using the 'Plot trait vs. genotype' module of MapDisto. This module allows to calculate corrected single-marker regression F-test values after detecting and removing outlier data in each marker genotypic sub-class.
In the case of a secondary LOD score peak linked to a major peak of a QTL, in order to determine if the secondary peak corresponded to a true QTL or to an artifact -or "fake QTL" -a detailed analysis of the distribution of recombination fractions along the chromosome was performed, owing the method of (Lorieux 2018). The analysis looked for restriction of recombination fractions that could induce arti cial linkage disequilibrium between the major and the secondary LOD score peaks. If arti cial linkage disequilibrium was detected, then the secondary peak was declared an artifact. Interaction or epistasis was tested using the R/qtl 'Scantwo' function (Prins et al. 2010).
When a QTL is found in more than one cross, joint or meta-analysis can increase the precision of the QTL location. As different markers segregated in the four populations, we created a genotype matrix from the union of the four individual data sets. Markers with no data in some of the populations were imputed using the R/qtl 'Argmax' function. Joint-and meta-QTL analyses were then ran using MapDisto 2.0.
To identify candidate genes in the ne-mapped regions, the MSU Rice genome annotation database and the Overview of Functionally Characterized Genes in Rice Online database (OGRO) (Yamamoto et al. 2012) were used. A 1-LOD support interval was used to de ne the search region for each QTL.

Results
Genetic maps Table 1 summarizes the main statistics of the individual genetic maps obtained for each cross.Overall, we obtained genetic map sizes coherent with their expected size according to ten high-quality maps In the cross WAS 208 ´ BBT 50, the parental lines showed 24% and 98% of incidence and 25.8% and 69.5% of RHBV severity in WAS 208 and BBT 50, respectively. The F 3 families (each representing one F 2 parental plant) exhibited a continuous variation of incidence ranged between 21.4% and 100%, as well as a variation of severity ranged between 9% and 68% ( Figure 2). The high incidence score for the susceptible parent BBT 50 -almost 100% -and the maximum values in the F 3 families indicate a perfect infection e ciency. The similarity between the severity score for BBT 50 and the maximum scores in F 3 families indicates that the experimental design, and in particular the population size, were adequate to capture the range of variation between the parental values. Interestingly, some F 3 families exhibited a signi cant lower severity than the resistant parent WAS 208.
In the cross Badka ´ BBT 50, the resistant and susceptible parents exhibited 32.8% and 98.9% of RHBV incidence, respectively. In the same order, these parents displayed 35.3% and 82% of severity. The incidence in the F 3 families ranged between 13% to 98%, and the severity varied between 11% and 80% ( Figure 2). A positive correlation between incidence and severity of the disease was detected (r=0.56, p<0.0001), suggesting a common partial genetic control for both assessments of RHBV disease. Some F 3 families exhibited lower incidence and severity than the resistant parent Badka.
In the cross PTB 25 x BBT 50, the incidence of the resistant and susceptible parents was 3.4% and 98.2%, respectively, and ranged between 5% to 96% in the F 3 families. Furthermore, PTB 25 and BBT 50 showed 11.8% and 68% of severity, respectively. The same trait varied between 7% and 69% in the F 3 families.
In the cross FD 2000 ´ BBT 50, the resistant and susceptible parents showed 20.8% and 75.4% of RHBV incidence, and 3.4% and 50.1% of severity, respectively. In the F 3 families; incidence ranged between 10% and 90%, and severity between 0% and 59%. The incidence and severity for the susceptible parent BBT 50 were notably lower than in the crosses before mentioned, indicating a low infection e ciency that can affect the power of QTL detection and the QTL effects estimation. In contrast, the incidence of FD 2000 was higher than in previous works (Romero et  Incidence and severity showed correlation between 0.42 and 0.66 (Table 2), indicating either a common genetic control of the two traits, or that they are interdependent.
The major QTL qHBV4.1 for RHBV incidence is present in most donors A major QTL, qHBV4.1, on chromosome 4 for RHBV incidence was identi ed by SMR, SIM and CIM in the crosses Badka ´ BBT 50 (LOD=20.97, R 2 =0.63), PTB 25 ´ BBT 50 (LOD=21.26, R 2 =0.60) and FD 2000 ´ BBT 50 (LOD=9.11, R 2 =0.34) ( Table 3). It was not detected in the cross WAS 208 ´ BBT 50, although a fake QTL analysis (Lorieux 2018) showed that it could be present in WAS 208, but with a much smaller effect (data not shown). The QTL support intervals are overlapping in the three populations, suggesting that the same QTL is shared by the three resistance donors. Joint analysis gave a LOD=47.00 and R 2 =0.42 at the position ~3.56 Mbp. The qHBV4.1 position also corresponds to a previously identi ed locus characterized as the major contributor to RHBV resistance in FD 2000 and FD 50 (Romero et al., 2014), con rming the wide range of action of this QTL. It explained 34-63% of the trait variance, indicating that qHBV4.1 is a major regulating factor of incidence of RHBV infection. The estimation of QTL effects for qHBV4.1 showed that this QTL is mostly of the additive type. The same genomic region was also associated with RHBV severity in the same crosses, however with lower LOD scores and R 2 values (LOD=6.69-8.50, R 2 =25-31%).
A new major QTL for RHBV incidence, qHBV4.2, identi ed in WAS 208 The control of RHBV incidence in WAS 208 was mainly explained by a different QTL on chromosome 4, designated as qHBV4.2 WAS208 (LOD=15.49, R 2 =0.52) by SMR, SIM and CIM. This QTL was not found in the other crosses (Table 3). This newly discovered QTL was located between 21.29 and 21.81 Mbp and explained 52% of the incidence variance. qHBV4.2 is therefore another major QTL for RHBV incidence that seems less frequent than qHBV4.1 in the rice germplasm.

Two new QTLs for RHBV incidence identi ed in WAS 208 and PTB 25
Two additional QTLs, although of lesser effect, were detected for RHBV incidence (Table 4) (Tables 3 and 4). The QTL effects in the populations involving WAS 208 and Badka indicate an additive behavior of qHBV11.1, while in FD 2000 ´ BBT 50 it seems to be more dominant (Table 3). Due to the correlation between incidence and severity, qHBV11.1 was also signi cant for RHBV incidence in the cross FD 2000 ´ BBT 50 although with lower statistics (LOD=7.43, R 2 =0.29), con rming that this QTL is controlling primarily the RHBV severity.

The qHBV4.1 and qHBV11.1 QTLs show strong interaction
Testing interaction between QTL regions with the R/qtl "scantwo" function produced a strong signal between the two QTLs qHBV4.1 and qHBV11.1 (LOD>15) (Figure S1), revealing either an epistasis relationship between the two regions, or a simple interdependency between the two traits. This is coherent with the positive correlation observed between severity and incidence.

Joint-and meta-analyses provide good candidates for QTL cloning
The joint and meta-analyses approaches allow pooling populations, providing more resolution for QTL mapping. This allowed us identifying candidate genes that could underlie two of the QTLs we discovered.

The qHBV11.1 region contains a gene for durable resistance to Rice stripe virus
The genomic region of qHBV11.1 contains many nucleotide binding site-leucine-rich repeat (NBS-LRR) genes, which are broadly known to confer resistance to multiple diseases (McHale et al. 2006). Other types of genes are also found in the interval, and remarkably the STV11 gene (MSU:LOC_Os11g30910, Chr11:17,984,964-17,986,719b; RAP-DB: Os11g0505300), which confers durable resistance to Rice stripe virus (RSV), one of the most devastative viral diseases of rice in Asia ). Interestingly, this virus belongs to the same genus as RHBV and it is also transmitted by planthoppers (Laodelphax striatellus Fallen). It should be also noted that, very close to the qHBV11.1 support interval, there are several paralog histidine kinase/Hsp90-like ATPase genes (MSU:LOC_Os11g31480, MSU:LOC_Os11g31500) that also confer resistance to RSV (Hayano-Saito and Hayashi, 2020).

Discussion
Successful genotype-phenotype association studies require precise methodologies to assess phenotypic variables. The latest technological advances in image acquisition have allowed the development of computer tools for picture processing and analysis that provide accurate data. Image processing for quantifying damage caused by plant diseases is particularly useful as it advantageously replaces traditional scoring scales, which are too dependent on the observer. The RHB disease has been commonly evaluated using the percentage of diseased plants at a given time -that is, the incidence. However, the main drawback of this approach is that it classi es all the plants with symptoms in the same class, without considering the level of damage -the severity. Understanding the genetic mechanisms of resistance to the hoja blanca disease therefore requires proper assessment of not only its incidence but also its severity. In this sense, we proposed for the rst time the use of digital images to evaluate the severity of damage caused by RHBV by estimating the affected leaf area. We showed that severity varies continuously among F 3 families, a characteristic behavior of quantitative features.
Severity and incidence evaluation allowed us to identify different QTLs, allowing us to better explain the complexity of the genetics of resistance to RHBV. Keeping in mind that the parents of the four populations were among the most resistant of a larger panel (Cruz et al. 2018), and although signi cant differences between incidence and severity were found among the parents of the four populations, complete resistance was not observed, con rming a previous observation by ( Morales and Jennings 2010) about the lack of immunity to RHBV in the O. sativa genepool.

RHBV resistance is regulated by multiple QTLs
The QTLs found in this study show that resistance to RHBV, assessed both by affected leaf area and incidence, is controlled by multiple genes. The co-location of LOD score peaks for incidence in the same region of chromosome 4 in three different crosses suggests that it is likely the same QTL, namely qHBV4.1. Although qHBV4.1 was associated to both phenotypic variables, its association was greater with incidence where it explained up to 63.6% of the phenotypic variance. These results suggest that the main action of qHBV4.1 could take place in the rst phase of the interaction of the virus with the plant, preventing the virus to enter and propagate in the plant, which is re ected in a smaller number of plants with symptoms. In addition, the little variation in nucleotides -less than 1% -in the region of qHBV4.1, between Badka and PTB 25 indicates a probable common local ancestry ( Figure S2). The differences in the LOD score (5.83 vs. 8.26) and in R 2 values (21.4% vs. 36.6%) for ALA, in these two populations, could be explained by the difference in the percentage of virulence of the vector colonies, which were ~69% and 46% for Badka ´ BBT 50 and PTB 25 ´ BBT 50, respectively. Also, the phenotypic evaluation trials were performed at different times of the year for the different populations. This involves the use of different vector colonies, and implies variations in microenvironmental factors such as day and night temperature, radiation or humidity that might affect the behavior of the insect and its ability to transmit the virus.
A major QTL had been previously identi ed in the region of qHBV4.1 in the varieties FD 2000 and Fedearroz 50 by incidence assessment (Romero et al. 2014), which suggests a predominant, common mechanism of resistance to RHBV in the different clades of Oryza sativa. The genetic variation between the parents observed in this region shows that they carry different alleles, of Japonica and Indica origins. This has implications for crop improvement, allowing breeders to broaden the genetic base of resistance in elite germplasm. A QTL for resistance to Rice stripe virus (RSV) has been identi ed between 4.4 and 6.9 Mpb in the N22 Aus variety (Wang et al. 2013). The qHBV4.1 QTL region might thus contain at least two tenuivirus resistance genes. As mentioned above, in the region of qHBV4.1 there is a putative gene that encodes for the AGO-4 Argonaute protein (LOC_Os04g06770). AGO proteins, in addition to be regulatory factors of endogenous gene expression, also play a critical role in the defense against viruses through interference with small RNA of viral origin which bind to AGOs and serve as a guide for it to cut new viral RNA particles (Mallory and Vaucheret 2011)(Silva-Martins et al., 2020). This system is a common defense mechanism against pathogens, and AGO-4 might well be associated with resistance to RHBV. We are currently carrying CRISPR-Cas9 knock-out experiments in AGO-4 in order to test this hypothesis.
A distinct QTL for RHBV incidence on chromosome 4, qHBV4.2 WAS , was identi ed In the WAS 208 ´ BBT 50 cross. This QTL was also a major one, explaining ~50% of the phenotypic variation. The region contains genes associated with cold tolerance (SAPK7) (Basu and Roychoudhury, 2014) and photooxidative stress (OsAPX7) (Caverzan et al. 2014), and a more interesting gene encoding for a zinc ngertype C3H protein (LOC_Os04g35800). This class of proteins has been found to be associated with resistance viruses in animals through degradation of viral RNA (Gao et al. 2002;Mao et al. 2013). Although there are no reports so far of its involvement in resistance to viruses in plants, it has been found to be associated with response to other types of pathogens (and abiotic stress) (Deng et al. 2012;Jan et al. 2013), which makes it an interesting candidate for RHBV resistance. Another candidate is a gene encoding for a MAPKKK kinase, which belongs to a signaling cascade that plays a major role in disease resistance in eukaryotes. This gene is also an interesting candidate because the activation of MAPK proteins occurs in the earliest stages after the interaction of the pathogen with the plant (Meng and Zhang 2013), which coincides well with the identi cation of this QTL through the assessment of incidence, but not severity, indicating that it -like qHBV4.1exerts its action in the early stages of the interaction of RHBV with the plant.
In contrast with the two QTLs identi ed on chromosome 4 that are predominantly associated with incidence, qHBV11.1 has a greater effect on disease severity. This could suggest a later action, in a second stage of the interaction of the plant with the virus, after the virus manages to enter the cells and to propagate. In this scenario qHBV11.1 would have a key role in diminishing the damage caused by the virus. This is thus more a "tolerance QTL".
It has been proposed that the resistance to RHBV comes mostly from Japonica germplasm (Morales and Jennings 2010). If this was the case, and to explain that the resistance sources studied here are of Indica type, one would need to assume that the region of qHBV11.1 was inherited from Japonica through introgression into Indica. This is not supported by the clustering analysis, which showed that the WAS 208 and Badka parents, for this region, are grouped together with Indica type accessions such as IR8. Therefore, this QTL does not originate from the Japonica cluster.
A search for candidate genes for qHBV11.1 evidenced a co-localization with a QTL for resistance to Rice  (Maeda et al. 2006). Whether those QTLs correspond to the same locus or to different, linked loci still needs to be clari ed. The hypothesis of a common defense mechanism acting against RHBV and RSV seems plausible, as both are RNA viruses of the same genus, although it has been found that RHBV is more related to other tenuiviruses like Echinochloa hoja blanca virus (EHBV) and Urochloa hoja blanca virus (UHBV) than to RSV (Fauquet et al., 2005), besides not being serologically related (Morales and Jennings 2010). Further studies of ne mapping and cloning of qSTV11 KAS showed that the gene responsible for resistance to RSV encodes a sulfotransferase that catalyzes the conversion of salicylic acid to its sulfated form. Although it is not clear how this process confers resistance against the virus, it was found that the susceptible allele of this gene is not capable of inducing this conversion . Salicylic acid has been found to be essential in the initiating signal to activate systemic resistance against Tobacco mosaic virus (Zhu et al. 2014), and also plays a central role in the hypersensitive response against Potato virus Y (PVY) (Baebler et al. 2014). It has even been reported as a direct inhibitor of the replication of Tomato bush dwarf virus (TBSV) (Tian et al. 2015). It is thus plausible that the STV11 gene is involved in resistance to RHBV. We are currently analyzing knockout lines obtained by CRISPR-Cas9 in STV11 to verify this hypothesis.
A second QTL on chromosome 11, qHBV11.2, was identi ed for disease incidence in the two crosses involving WAS 208 and PTB 25. Based on the variation found in the clustering analysis -and assuming the QTL is the same in the two crosses -it is possible that WAS 208 and PTB 25 have different allelic variants of qHBV11.2 with a different effect on disease resistance, which could explain that in PTB 25 it was also associated with severity. The region could also correspond to two distinct, linked QTLs in the two populations, since their support intervals are quite large (> 3 Mpb). To answer this question, a ne mapping study of the region using larger F 2 populations would be needed.
An additional QTL for RHBV incidence, qHBV6, was identi ed In the PTB 25 ´ BBT 50 cross on chromosome 6. In a previous GWAS experiment, the same region was identi ed in the genotypes PTB 25 and Pokkali (Cruz-Gallego et al. 2018), also supported by a biparental QTL analysis in the F 2:3 cross FD 50 ´ WC 366 (our unpublished data). A search for genes in this region found the OsBBI1 gene that participates in various biological processes, among which is the innate immune response. OsBBI1's expression is induced by the fungus Magnaporthe oryzae and chemical inducers such as salicylic acid (Li et al. 2011), which, as discussed above, is also associated with systemic resistance to viruses. OsBBI1 is thus a good candidate gene for qHBV6.
Altogether these results show that resistance to RHBV disease is controlled by multiple quantitative genetic factors of different origins, with varying effects and action mode. The identi cation of strong candidate genes underlying the detected QTLs supports the idea that resistance is mediated by different defense mechanisms such as viral gene silencing and the salicylic acid pathway. This hypothesis has proven true for other study models such as in the Citrus tristeza virus (CTV) where it has been found that silencing of key genes in these defense pathways increases the spread of the virus and its accumulation in the plant(Gómez-Muñoz et al. 2017). A considerable amount of work is still needed to understand the ne mechanisms behind defense against RHBV.

Resistance QTLs in the susceptible parent
In the PTB 25 ´ BBT 50 population, two QTLs for RHBV incidence were identi ed on chromosomes 9 and 10, of which the allele that reduces the affected leaf area is brought by BBT 50. One explanation could be that PTB 25 carries susceptibility alleles at these QTLs. However, this is unlikely since inactivation of a susceptibility gene by the resistant allele results in increased resistance. Genetically, the resistant allele of a susceptibility gene is therefore generally recessive. This is not what we observed, since the mean ALA of the heterozygous class AB is more similar to the susceptible class AA (PTB 25) than to the BB class (BBT 50). Therefore, the most likely explanation is that the three other resistant parents (FD 2000, WC 366 and Badka) and the susceptible parent (BBT 50) carry resistant alleles at these QTL, while the PTB 25 parent does not.

Integration of QTLs into a RHBV disease resistance model
Based on the QTLs associated with RHBV resistance, we propose a simplistic model that draws the possible processes involving these loci. The QTLs qHBV4.1 and qHBV4.2 were found to be most associated with disease incidence, so it is likely that they are involved in the rst phase of virus-plant interaction. These QTL may be inhibiting the translation of viral RNA, either by direct action on them or through other genes for degradation. The effect of these QTLs on the incidence, indicates that this mechanism might be the most important for the resistance of the plant to the virus. In the scenario where the favorable alleles of qHBV4.1 or qHBV4.2 are not present or that the virus manages to overcome this barrier and to propagate in the plant, an additional mechanism involving qHBV11.1 would be hampering the multiplication of the virus. This could occur by direct action, inhibiting the synthesis of new viral particles, or more likely, serving as a trigger signal for the activation of defense genes of the systemic resistance system. Although this mechanism would not provide complete resistance to the virus, it would considerably reduce the damage -meaning less leaf area affected. The qHBV6 QTL is also related to the severity of the disease, so it is possible that it also participates in this same mechanism. Although the role of the susceptibility QTLs qHBV9 and qHBV10 is more di cult to interpret, it is likely that they are negatively regulating the activation of defense genes against the virus. Availability of supporting data