Phenotypic and molecular genetic characterization indicate no major race-specific interactions between Xanthomonas translucens pv. graminis and Lolium multiflorum




Bacterial wilt of forage grasses, caused by the pathogen Xanthomonas translucens pv. graminis (Xtg), is a major disease of forage grasses such as Italian ryegrass (Lolium multiflorum). The plant genotype-bacterial isolate interaction was analysed to elucidate the existence of race-specific responses and to assist the identification of plant disease resistance genes. In a greenhouse experiment, 62 selected plant genotypes were artificially inoculated with six different bacterial isolates. Significant differences in resistance were observed among Lmultiflorum genotypes (< 0·001) and in virulence (intensity of disease symptoms) among Xtg isolates (< 0·001) using the area under the disease progress curve (AUDPC). No significant genotype-isolate interaction (> 0·05) could be observed using linear regression modelling. However, additive main effects and multiplicative interaction effects (ammi) analysis revealed five genotypes which did not cluster close to the origin of the biplot, indicating specific interactions between these genotypes and some bacterial isolates. Simple sequence repeat (SSR) markers were used to identify marker-resistance associations using the same plant genotypes and bacterial isolates. The SSR marker NFA027 located on linkage group (LG) 5 was significantly associated with bacterial wilt resistance across all six bacterial isolates and explained up to 37·4% of the total variance of AUDPC values. Neither the inoculation experiment nor the SSR analyses revealed major host genotype-pathogen isolate interactions, thus suggesting that Xtg resistance, observed so far, is effective across a broad range of different bacterial isolates and plant genotypes.


Bacterial wilt of forage grasses is caused by the pathogen Xanthomonas translucens pv. graminis (Egli et al., 1975; Vauterin et al., 1995). It is one of the most important diseases of forage grasses in temperate regions, and so far four different pathovars of X. translucens have been described that infect forage grasses. Xanthomonas translucens pv. graminis (Xtg) infects a broad range of forage grasses including ryegrasses, while the other pathovars are mostly restricted to one genus (Egli & Schmidt, 1982). Bacterial wilt is estimated to account for annual forage yield losses of 10–15% in cultivated grassland; however, depending on cultivar susceptibility, Xtg may also lead to complete yield losses in pure and mixed stands (Suter et al., 2005).

Xtg enters grass leaves and roots through wounded tissue, multiplies in the xylem and causes symptoms including wilting of leaves, yellow stripes along the vascular tissue and withering of entire tillers. In severe cases, the plants die within a few days after inoculation. Since treatments with antibiotics are undesired in forage production, and both the disinfection of mowing equipment and induced systemic resistance with epiphytic bacteria have not yielded satisfactory results (Schmidt, 1988), resistance breeding is currently the only applicable and accepted measure to prevent substantial yield losses. So far, resistance breeding programmes have been based on artificial seedling inoculation in the greenhouse and targeted recurrent phenotypic selection. Although considerable progress has been achieved, further advances in breeding programmes have often stagnated after several cycles of recurrent selection, and susceptible individuals have still been observed in advanced populations (Michel, 2001). Italian ryegrass (Lolium multiflorum) is a major forage crop for hay and silage production, and therefore particularly prone to infection through contaminated mowing equipment. Understanding the genetic basis of Xtg resistance is fundamental for the further improvement of disease resistance and for the development of molecular tools, which may help to enhance phenotypic selection through marker assisted selection (MAS).

Resistance has been extensively studied in rice where X. oryzae pv. oryzae (Xoo) causes bacterial blight. At least 33 race-specific genes conferring resistance to Xoo have been identified and seven have been cloned (reviewed in White & Yang, 2009; Niño-Liu et al., 2006). Knowledge on this interaction may provide a suitable model for Xanthomonas resistance in L. multiflorum due to the syntenic relationships between species among the members of the grass family (Devos, 2005). Xanthomonas resistance genes in rice are mostly based on race-specific recognition of effector proteins secreted by pathogens. Some of these race-specific resistance genes are only effective against one or a few Xoo (e.g. Xa1), whereas others condition resistance to a wide spectrum of different isolates (e.g. Xa21 and Xa23). Race-specific (or qualitative) resistance is usually conferred by a single resistance gene, and therefore may exert high selection pressure on pathogen populations, which eventually evolve to overcome resistance. In contrast to Xoo resistance, no race-specific resistance genes have been found that are effective against X. oryzae pv. oryzicola (Xoc) isolates. Xoc resistance has been shown to be inherited quantitatively by 11 quantitative trait loci (QTL) that together explain 84·6% of the total phenotypic variance (Tang et al., 2000). Quantitative resistance controlled by various genes is considered to be more durable when compared to race-specific, qualitative resistance; however it carries the disadvantages of providing only partial resistance and being less effective when conditions are particularly favourable for the pathogen (Geiger & Heun, 1989).

In Lmultiflorum, QTL analysis of a pseudo-testcross family has revealed that bacterial wilt resistance was controlled by one major QTL on linkage group (LG) 4, explaining up to 84% of the total phenotypic variance (Studer et al., 2006). In addition to this major QTL, three minor QTL on LG 1, 5 and 6 were identified, explaining between 2·9% and 7·4% of the total phenotypic variance. In general, the existence of major QTL is often associated with the presence of one or only a few major resistance genes (Mutlu et al., 2005), and race-specificity is likely to develop for such genes. On the other hand, cultivar-isolate interactions of five commercially available L. multiflorum cultivars inoculated with four different Xtg isolates have revealed that the cultivar-isolate interaction was not significant (Michel, 2001). However, ryegrass cultivars are multi-genotype populations, and therefore these results have not allowed for race-specificity at the genotype level. Thus, the genetic control of Xtg resistance and the presence of race-specificity in the L. multiflorum-Xtg interaction remain unclear.

Recently, additive main effects and multiplicative interaction (ammi) analyses and the resulting biplot have been shown to be effective tools to understand and describe complex host-pathogen interactions even if large variability within host resistance or pathogen virulence is expected (Yan & Falk, 2002). ammi is a hybrid analysis that combines additive and multiplicative components of a two-way data structure. The biplot that results from the calculation of principal component analysis (pca) scores displays the host-by-pathogen interaction term and facilitates the interpretation of race-specific resistances.

The association of molecular markers and phenotypic traits such as disease resistance may be investigated in genetically diverse plant populations to allow for the development of diagnostic markers. In this approach, molecular marker alleles with frequencies correlating with the attributes of the phenotype are identified. For the selection of alleles to be included in the linear regression model, least absolute shrinkage and selection operator (lasso) subset selection is an efficient dimension-reduction technique, which produces a stable and interpretable model (Tibshirani, 1996). In addition to the identification of alleles associated with bacterial wilt resistance, significant associations would also reveal the existence of race-specific interactions.

The aims of the present study were to elucidate the existence of race-specificity in different L. multiflorum genotype-Xtg isolate combinations and to perform a marker-trait association analysis. A detailed investigation of the genotype-isolate interaction and the identification of molecular markers correlated with Xtg resistance may provide a starting point for the development of markers for marker-assisted breeding tools.

Materials and methods

Bacterial isolates and plant material

Six bacterial isolates collected in Switzerland and previously characterized for genetic diversity and virulence (Xtg3, Xtg5, Xtg8, Xtg9, Xtg19 and Xtg29; Kölliker et al., 2006) were selected in order to represent a broad genetic diversity and geographical distribution of original sampling sites. Bacteria were stored at −80°C in GYC (glucose 20 g L−1, yeast extract 10 g L−1, CaCO3 20 g L−1) broth containing 150 mL L−1 glycerol. Sixty-two plant genotypes selected from 12 cultivars, ecotypes or breeding populations were used for this study (populations LmA–LmL). The genotypes of population LmA are F2 progeny of a mapping population segregating for bacterial wilt resistance (Studer et al., 2006), genotypes of population LmB are Syn1 progeny of a polycross with nine elite genotypes, genotypes of the population LmC are from a cultivar candidate from Agroscope Reckenholz-Tänikon, LmJ, LmK and LmL are individuals from different commercially available cultivars (Turilo, Adret and Axis, respectively), and genotypes of the populations LmD, LmE, LmF, LmG, LmH, LmI are F2 progeny of ecotypes collected in different parts of Switzerland. Population LmA consisted of 12 genotypes, populations LmB, LmC, LmJ and LmL consisted of five genotypes, populations LmD, LmE, LmF, LmG, LmH and LmI consisted of four genotypes, and population LmK consisted of six genotypes. The genotypes were selected based on a pre-screening in order to represent different levels of resistance and susceptibility to Xtg by using a mixture of bacterial isolates (data not shown).

Experimental design and disease scoring

All L. multiflorum genotypes were clonally propagated by separating single tillers, which were transferred into soil-filled pots and grown in a greenhouse. Assessment of bacterial wilt symptoms was performed using four replications per genotype-isolate combination in a completely randomized block design. Replications were assigned to four sequentially grown blocks consisting of 372 plants each resulting from all possible combinations of 62 L. multiflorum genotypes with six Xtg isolates. After 8 weeks, the plants were inoculated with the six different Xtg isolates using the leaf clipping technique described by Kölliker et al. (2006). The inoculated plants were allowed to regrow in a greenhouse at average temperatures of 20/18°C (day/night), under long day conditions (16 h light, >100 μE m−2 s−1) at about 70% relative humidity. Scoring of bacterial wilt symptoms was performed on whole plants 15, 21 and 28 days after inoculation according to a scale ranging from 1 to 9. Disease scores were 1 = no symptoms, 2 = wilting or withering next to the cutting area, 3 = one entire leaf is wilting or is completely withered, 4 = half of the leaves and tillers are wilting, 5 = all leaves and tillers are affected, 6 = intermediate score, 7 = all leaves and tillers are affected and half of them are withering, 8 = plant is dried, base of tillers is still green, 9 = plant is dead, no green parts. Twenty-eight days after inoculation, the plants were cut with sterilized scissors and disease symptoms were again scored 21 days later (49 days after inoculation).

In addition, a control experiment was performed to test whether plants showing weak disease symptoms could be distinguished from control-treated plants (inoculation with NaCl solution). Therefore, the three isolates Xtg3, Xtg9 and Xtg19 were used to inoculate a selection of 13 different L. multiflorum genotypes in four replications grown simultaneously. None of the control-treated plants showed disease symptoms (score = 1), while disease scores of Xtg inoculated plants ranged from 2 to 9 and revealed significant differences among L. multiflorum genotypes in a Wilcoxon Sum rank test (< 0·001).

Data analysis and statistics

Area under the disease progress curve (AUDPC) values were calculated for each inoculated plant from the scores obtained at 15, 21 and 28 days after inoculation following the formula:


where Yi is the disease score at time point i and ti is the number of days after inoculation. Plants were assigned to three categories according to their AUDPC values: Resistant (AUDPC values ≤39; disease symptoms were never described with a score higher than 3), moderately susceptible (AUDPC values between 39 and 60; disease symptoms were never described with a score higher than 6), and susceptible (AUDPC values > 60). AUDPC values were used for all anova and ammi analyses.

The scores obtained at 49 days after inoculation (49d scores) showed an asymmetric distribution with many plants being either dead (score = 9) or healthy (scores = 2 or 3). The normal-score-transformation suggested by Tilquin (2003) was used to transform scores towards a normal distribution for subsequent anova. AUDPC values were fitted to a linear regression model including the factors replication, L. multiflorum genotype, Xtg isolate, and the interaction L. multiflorum genotype and Xtg isolate. The factor population was not included in the model, because L. multiflorum genotypes were not chosen at random and therefore do not represent the entire group.

Interactions between Lmultiflorum genotypes and pathogen isolates were studied by applying an ammi model with additive effects for plant genotypes (G) and isolates (I) and a multiplicative term for G-I interaction (De Mendiburu, 2009). The ammi analysis combines the anova and the singular value decomposition (SVD) and has been explained in detail by Gollob (1968). In this analysis, multiplicative effects for G-I are fitted by principal component analysis (pca). The additive interaction in the anova model is obtained by multiplication of genotype PC scores by isolate PC scores. A reduction of PCs to one or two axes (PC1 and PC2) is used to plot the interaction effect via the PC scores for genotypes and isolates. The ammi model is described as:


where Yij is the AUDPC value of the ith genotype inoculated with the jth isolate, gi is the mean AUDPC value of the ith genotype minus the grand mean, ej is the effect of the jth isolate, λk is the square root of the eigenvalue of the pca axis k, inline image and Yjkare the principal component scores for the pca axis k of the ith genotype and the jth isolate and Rijis the residual. In order to facilitate visualization of the relationships among isolates and genotypes, a polygon was drawn on the genotypes farthest from the biplot origin such that all other genotypes were contained within the polygon.

Molecular marker and association analysis

Genomic DNA was extracted from lyophilized plant leaves using the Corbett x-tractor Gene roboter (Qiagen) and the Nucleospin 96 plant kit (Macherey-Nagel) following the manufacturers’ recommendations. DNA concentrations and purities were assessed with a Nanodrop spectrophotometer (Thermo Fisher Scientific) and the ND-1000 software. The simple sequence repeat (SSR) primers used for PCR amplification were selected to be localized on the same LGs where QTL for bacterial wilt resistance were identified previously in glasshouse and field experiments (Studer et al., 2006; Table 1). PCR reactions were conducted in a total volume of 20 μL containing 10 ng of genomic DNA, 4 μL of 5× PCR buffer, 0·4 μmol L−1 labelled (FAM, HEX or NED) forward and unlabelled reverse primer, 2 mmol L−1 MgCl2, 0·2 mmol L−1 of each dNTP and 0·75 U GoTaq® Flexi DNA Polymerase (Promega). PCR conditions were as follows: initial denaturation for 5 min at 94°C, followed by 12 cycles of touchdown PCR consisting of 30 s at 94°C, 1 min at 72°C (decreasing by 1°C per touchdown cycle) and 1 min at 72°C. This was followed by 30 cycles of 30 s at 95°C, 1 min at 60°C, 1 min at 72°C and a final extension of 5 min at 72°C. PCR fragments were separated on an ABI Prism 3100 Genetic Analyzer (Applied Biosystems) and analysed using the ROX HD400 standard (Applied Biosystems) and the GeneMarker® software version 1.51 (SoftGenetics). SSR alleles were scored for presence (1) or absence (0) and entered into a binary matrix for each genotype. In order to assess genetic variation within and between populations, an analysis of molecular variance (amova) was performed with the software Arlequin (Excoffier et al., 2005).

Table 1.   Names, sequences and references of simple sequence repeat (SSR) markers used to genotype 62 Lolium multiflorum plants differing in resistance to bacterial wilt caused by Xanthomonas translucens pv. graminis. Linkage groups and positions were determined in the L. multiflorum mapping population as described by Studer et al. (2006)
LocusLinkage groupPosition (cM)Forward primer sequence (5′→3′)Reverse primer sequence (5′→3′)Reference

Based on this matrix and the AUDPC values assessed in the greenhouse, all alleles were submitted to a least absolute shrinkage and selection operator (lasso) analysis (Tibshirani, 1996). This shrinkage method represents an efficient tool for subset selection of parameters to be included in the regression model. lasso subset selection was carried out for each bacterial isolate separately. The alleles selected by lasso were included in a linear regression model with the log-transformed AUDPC values for each plant genotype. Alleles with a significant effect on bacterial wilt resistance were identified using anova.

All statistical analyses were performed with the r statistical software (The R Development Core Team, 2008) using the packages: stats, graphics, coin (Hothorn et al., 2006), agricolae (De Mendiburu, 2009) and lars (Hastie & Efron, 2007).


Data structure and distribution

No complete resistance was observed on any plant genotype and all scores between 2 and 9 were used to describe bacterial wilt symptoms. Average disease scores ranged from 2·67 ± 0·97 (15 days after inoculation) to 3·39 ± 1·16 (21 days after inoculation) and reached a maximum value of 3·66 ± 1·37 (28 days after inoculation). At 28 days after inoculation, the plants were cut using sterile scissors and allowed to regrow for 21 days. The resulting average 49d scores were 2·86 ± 1·78. The AUDPC values varied greatly among some L. multiflorum genotypes (Fig. 1). Comparison of the mean AUDPC values across all Xtg isolates revealed that genotype LmK-01 was the most susceptible genotype with a mean AUDPC value of 83·72, whereas genotype LmG-04 was the most resistant genotype with a mean AUDPC value of 30·06 (Fig. 1). The bacterial isolates also varied significantly in their virulence, here defined and measured as the intensity of symptoms of disease. Xtg9 and Xtg8 were the most virulent isolates, whereas Xtg19 and Xtg3 were the least virulent. Xtg5 and Xtg29 were intermediately virulent.

Figure 1.

 Mean area under the disease progress curve (AUDPC) values from four replications for each of 62 Lolium multiflorum (Lm) genotypes inoculated with six different Xanthomonas translucens pv. graminis (Xtg) isolates (standard error = 0·37). The AUDPC values are ranked according to mean AUDPC values calculated across all Xtg isolates. Left: most resistant genotype, right: most susceptible genotype.

Regression modelling and race-specificity

Significant differences in the AUDPC values among Lmultiflorum genotypes (< 0·001) were observed using anova for the days 15, 21 and 28 days after inoculation (Table 2) and the normal-transformed 49d scores (Table 3). In addition, the factors replication and Xtg isolate (< 0·001) had considerable influence on the observed variance of AUDPC values. The factors included in the anova explained 73% of the total variance of AUDPC values. F-values of replication and Lmultiflorum genotype were high (62·94 and 39·18), underlining their strong contribution to variance effects (Table 2). The interaction of bacterial isolate and plant genotype was not significant in the linear regression analysis of AUDPC values, and the F-value was very low (> 0·05). Furthermore, no significant influence of the Xtg–Lmultiflorum interaction was observed with anova of normal-transformed 49d scores (> 0·05; Table 3).

Table 2.   Analysis of variance for area under disease progress curve (AUDPC) values of 62 Lolium multiflorum genotypes inoculated with six Xanthomonas translucens pv. graminis (Xtg) isolates
Source of variationDFMean squaresF-statistics
Replication  3472562·94 (< 0·001)
L. multiflorum genotype 61294139·18 (< 0·001)
Xtg isolate  5 99913·31 (< 0·001)
Lmultiflorum genotype- Xtg isolate 305 630·95 (= 0·97)
Residuals1113 75 
Table 3.   Analysis of variance of normal-transformed disease scores obtained at 49 days after inoculation of 62 Lolium multiflorum genotypes inoculated with six Xanthomonas translucens pv. graminis (Xtg) isolates
Source of variationDFMean squaresF-statistics
Replication 32·437·53 (< 0·001)
L. multiflorum genotype 616·7520·87 (< 0·001)
Xtg isolate 53·059·43 (< 0·001)
Lmultiflorum genotype- Xtg isolate3050·331·03 (= 0·36)

Using the AUDPC value based classification of resistant (R), moderately susceptible (MS) and susceptible (S) genotypes, 11 (17·7%) L. multiflorum genotypes were resistant, four (6·4%) were moderately susceptible, and three (4·8%) were susceptible across all bacterial isolates (Table 4). The other 44 (71·1%) Lmultiflorum genotypes showed variable resistance to one or more bacterial isolates, but there were only two genotypes (3·2%) that were classified as resistant to one or more bacterial isolates and also susceptible to one or more other bacterial isolates. There were also three genotypes that showed significantly different AUDPC values across different isolates according to a pairwise t-test (< 0·05). This analysis implies that some genotype-isolate specific interactions could exist. Therefore, the interaction between pathogen isolates and plant genotypes was evaluated by ammi and visualized using a biplot (Fig. 2). The first two principal component axes of the biplot accounted for 35·7% (PC1) and 26·2% (PC2) of the total variation of the genotype-isolate interaction. In this biplot, genotypes LmA-08, LmA-09, LmI-04, LmJ-04, LmF-01, LmC-01 and LmA-03 were located on the vertices of the polygon, indicating differential response to some specific isolates. On the other hand, genotypes LmA-03, LmF-01 and LmC-01 were located very close to the origin of the biplot indicating no differential response. Therefore, only five out of 62 genotypes (LmJ-04, LmI-04, LmA-09, LmA-08 and LmJ-01) that did not cluster close to the origin (PC1 and PC2: −1·5 < or >1·5) were considered to show potential race-specific resistance. For example, genotype LmA-09 was especially susceptible to the isolates Xtg5 and Xtg9, but only moderately susceptible to the other isolates, while genotype LmI-04 was particularly susceptible to isolates Xtg8 and Xtg9 but only moderately susceptible or resistant to the other isolates (Table 4).

Table 4.   Classification of 62 Lolium multiflorum genotypes according to their resistance to six bacterial Xanthomonas translucens pv. graminis (Xtg) isolates. Genotype-isolate combinations with different letters showed significantly different area under disease progress curve (AUDPC) values according to a pairwise t-test (< 0·05) determined on four replications per genotype-isolate combination. P-value adjustment was performed according to Holm (1979)
Plant genotypesBacterial isolatesa
  1. aR: resistant (AUDPC values ≤39); MS: moderately susceptible (39< AUDPC values <60); S: susceptible (AUDPC values ≥60).

LmA-07R aMS bMS abMS abR aR ab
LmC-03R aMS abMS bMS bMS abR a
LmG-02R abMS bR abR abR aR ab
Figure 2.

 Biplot of the first two interaction principal component (PC) axes scores derived from additive main effects and multiplicative interaction effects (AMMI) analysis of area under the disease progress curve (AUDPC) values of 62 Lolium multiflorum genotypes (Lm) inoculated with six Xanthomonas translucens pv. graminis (Xtg) isolates. The dashed line connects the L. multiflorum genotypes located on the vertices of the polygon. Lolium multiflorum genotypes not clustering close to the origin (PC1 and PC2 scores: −1·5< or >1·5) are indicated with names. Arrows indicate direction and distance from the origin of PC scores of the Xtg isolates.

SSR genotyping and association analysis

EST-derived SSR markers were used to perform marker-trait associations and to complement the linkage map of the L. multiflorum population initially developed by Studer et al. (2006; Fig. 3). With the 20 SSR primer pairs (Table 1), 122 different alleles were detected across the 62 genotypes of L. multiflorum. The number of alleles per SSR locus varied from two to 15. Some of these alleles were detected in many different plant genotypes, whereas several alleles were rare and only scored in one or two plant genotypes. amova revealed that 80·55% of the variance in allele frequencies was due to variation between genotypes within populations, and 19·45% to variation between populations. After lasso analysis, selected alleles (factors) and the AUDPC values (response) were fitted to a linear regression model followed by anova. Of the 122 scored alleles, 72 different alleles were significantly correlated with resistance to at least one Xtg isolate and the three alleles NFA027 (187 bp), NFA027 (193 bp) and G03_013 (193 bp) were significantly correlated with resistance to all six bacterial isolates. Most alleles only explained between 0·001 and 5% of the observed variance for AUDPC values. However, 11 alleles each explained more than 5% of the variance of AUDPC values for resistance to at least one bacterial isolate (Table 5). Five of the alleles explaining more than 5% of the observed variance of AUDPC values were scored on the same NFA027 locus on LG 5. Most notably, NFA027 (187 bp) explained between 32·3% and 37·4% of the total observed variance of AUDPC values across all isolates. Further, one additional allele on LG 5, i.e. NFA059 (125 bp), four alleles on LG 4, i.e. G03_013 (203 bp), G04_034 (162 bp), G05_121 (226 bp) and G05_122 (381 bp), and one allele on LG 1, i.e. G02_037 (197 bp), were observed that explained more than 5% of the observed variance of AUDPC values. Alleles associated with the QTL observed in the initial L. multiflorum mapping population were not or only occasionally observed in individuals of LmA, the F2 progeny of the mapping population, most likely due to the small sample size of only 12 individuals. Across all L. multiflorum genotypes used in the present study, none of the alleles from marker loci that mapped close to the major QTL on LG 4 were significantly correlated with bacterial wilt resistance.

Figure 3.

 Linkage Groups (LG) 1, 4, 5 and 6 of the genetic linkage map of a Lolium multiflorum reference population (Studer et al., 2006) constructed using 368 amplified fragment length polymorphism (AFLP) and simple sequence repeat (SSR) markers complemented with expressed sequence tag (EST)-derived SSR markers (Gxx_xxx). SSR markers in bold were used for marker-trait association analysis (Table 1). Scale units are given in centiMorgan (cM). Positions of Quantitative Trait Loci (QTL) for bacterial wilt resistance were re-calculated using multiple QTL model (MQM) mapping based on least square means of resistance from greenhouse data of Studer et al. (2006). The maximum logarithm of the odds (LOD) score position is indicated with a horizontal line, bars represent the interval between two positions obtained at LOD scores two units lower than the maximal score.

Table 5.   Alleles significantly correlated with bacterial wilt resistance, the percentage of the variance for area under the disease progress curve (AUDPC) values they explained, and the linkage group (LG) they mapped to on the Lolium multiflorum reference map (Studer et al., 2006). The alleles listed explained at least 5% of the phenotypic variance by anova and were selected with least absolute shrinkage and selection operator (LASSO) at (< 0·001)
AlleleLGBacterial isolates
Xtg 3Xtg 5Xtg 8Xtg 9Xtg 19Xtg 29
NFA027 (179 bp)5  8·0 8·3 8·2 8·3 8·1
NFA027 (183 bp)512·3 9·5  9·5 9·9 9·6
NFA027 (186 bp)5 12·517·212·4  
NFA027 (187 bp)534·934·937·434·632·332·4
NFA027 (193 bp)510·1 5·9 5·4 5·810·310·5
NFA059 (125 bp)5      6·0
G02_037 (197 bp)1  3·03·3 3·9 9·6 
G03_013 (203 bp)4    5·8 9·0 
G04_034 (162 bp)4  5·5   9·3
G05_121 (226 bp)4  1·03·4   8·4
G05_122 (381 bp)410·7 0·20·6 0·6  0·3


A resistance-based strategy for sustainable control of bacterial wilt in forage grasses would strongly benefit from a more thorough understanding of the existence of race-specific resistance mechanisms. This understanding may be especially important for durable disease control, since race-specific resistance usually provides protection against subpopulations of the pathogen only and may be prone to changes in pathogen structure due to genetic changes or migration from other geographic areas. However, the fact that anova revealed no statistically significant interaction of Lmultiflorum genotypes and Xtg isolates and disease severity varied more or less continuously and independently of Xtg isolates, suggests that Xtg resistance is not conferred by race-specific resistance genes in this species.

These results are congruent with observations from Michel (2001), where the L. multiflorum cultivar-bacterial isolate interaction was not significant. However, in the mentioned study, the cultivars used consisted of genetically diverse individuals. Thus, the results do not preclude the existence of plant genotype-bacterial isolate interactions. However, even though highly diverse plant genotypes were used, representing a broad range of cultivars and ecotypes, and bacterial isolates representing the broadest range of genetic diversity currently available from Switzerland, no evidence was found for race-specific interactions. Although such results are always limited to the plant genotypes and bacterial isolates used for investigation, the materials used in this study are representative for grassland agriculture in many temperate regions.

The significant differences observed among replications can be explained by varying outside temperatures across the sequentially grown replications and the inability to cool the greenhouse. Growth rates and xanthan production of xanthomonads are affected by temperature and influence disease development (Imaizumi et al., 1999). However, correlation among the four replications was high and significant, confirming the suitability of the experimental approach used.

The ranking in virulence of the six bacterial isolates in the present study was for the most part congruent with the differences in virulence observed by Kölliker et al. (2006) which was based on three different L. multiflorum cultivars. This provides strong support that the bacterial isolates used are reproducibly virulent on different plant material and during different experiments.

Partial resistance to bacterial leaf streak disease was concluded to be plant genotype-dependent in an experiment with wheat genotypes inoculated with X. translucens pv. cerealis (previously named X. campestris pv. cerealis) isolates, as no significant genotype-isolate interaction was detected (El Attari et al., 1996). In contrast, highly significant genotype-Xanthomonas isolate interactions have been observed in rice and cassava. Knowledge of these interactions has been crucial for the identification of plant genotypes with specific resistances and pathogen isolates with specific virulence (Wydra et al., 2004; Nayak et al., 2008). However, these two Xanthomonas species (Xoo infecting rice and X. axonopodis pv. manihotis (Xam) infecting cassava) possess a very high level of genetic diversity (60–90%) indicating that they underwent substantial genetic differentiation (Restrepo et al., 2000; Hua et al., 2007). On the other hand, Xtg isolates have been shown to possess a high level of genetic similarity (>86%) based on amplified fragment length polymorphism (AFLP) markers and low genetic diversity (Kölliker et al., 2006). Further, it has been shown that pathogen variation is suppressed by mixtures of genetically diverse host plants (Zhu et al., 2000). This has been shown for rice cultivars grown in monocultures compared to mixtures of different rice varieties and the rice blast causing pathogen Magnaporthe grisea. Low genetic diversity of Xtg and genetically diverse host plant populations may therefore represent an impediment to adaptation. That is, the evolutionary potential of Xtg during the interaction with its host may be limited and the frequency of occurrence of race-specific interactions may also be reduced. Cultivars of forage grasses are usually composed of numerous individual genotypes containing several different resistance genes and they are often grown not in pure stands but in mixtures with other grasses and legume species. Therefore, selection pressure may be too low for the development of race-specific resistance in the L. multiflorum-Xtg interaction. Qualitative race-specific resistance has been shown to exist in L. multiflorum against other pathogens such as the crown rust fungus, Puccinia coronata (Schubiger et al., 2006). Therefore, not only the genetic composition of the host, but also characteristics of the pathogen such as the evolutionary potential may be responsible for the absence of race-specific interactions.

Based on the insignificant L. multiflorum-Xtg interaction, one might expect of the marker-trait associations that all alleles would correlate with bacterial wilt resistance across all isolates. This was true for one allele of the NFA027 locus (187 bp) which mapped to LG 5 and explained up to 37·4% of the total observed variance of AUDPC values across all isolates. This is a further indication of major broad spectrum resistance not limited to specific isolates. The major QTL on LG 4 described by Studer et al. (2006) could not be confirmed in the F2 progeny derived from the initial mapping population due to the small number of parental alleles observed in the investigated individuals. However, the intention was rather to investigate whether such a QTL was present in more distantly related germplasm. The markers that mapped close to the major QTL on LG 4 (Fig. 3) observed by Studer et al. (2006) did not have major effects on the total observed variance of AUDPC values across all genotypes, indicating the existence of additional resistance genes or QTL. Four alleles that mapped to LG 4, but not in the vicinity of the major QTL, explained up to 10·7% of the total observed variance of AUDPC values, suggesting that there may be other genes with minor contributions to Xtg resistance. In addition, the alleles that explained between 5·5% and 9·3% of variance of AUDPC values across only one or two bacterial isolates may be an indication of the existence of several minor race-specific resistance genes with only small effects.

In summary, it is concluded from the greenhouse experiment and the marker-trait associations that in a wide range of L. multiflorum genotypes, bacterial wilt resistance is effective against a broad range of Xtg isolates with different virulence. Consequently, no major race-specific resistance seems to exist in the Xtg-L. multiflorum interaction. Therefore, resistance breeding may be based on one or a few Xtg isolates in order to select for L. multiflorum genotypes with high levels of Xtg resistance without compromising durability of resistance.


This project was funded by the Swiss National Science Foundation (SNF; Project number: 3100A0-112582). We appreciate the laboratory work contributions of Denise Portmann. We would also like to acknowledge Lukas Rosinus for excellent statistical support, Philipp Streckeisen for technical support and Michael Winzeler for critical review of the manuscript.