SEARCH

SEARCH BY CITATION

Keywords:

  • Arabidopsis thaliana;
  • environmental effect;
  • fitness;
  • quantitative trait locus (QTL) analysis;
  • resistance to pathogens;
  • yield

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • • 
    Pathogens represent an important threat to plant communities and agriculture, and can shape many aspects of plant evolution. Natural variation in plant disease susceptibility is typically quantitative, yet studies on the molecular basis of disease resistance have focused mainly on qualitative variation.
  • • 
    Here we investigated the genetic architecture of quantitative susceptibility to the bacterium Pseudomonas syringae by performing a quantitative trait locus (QTL) analysis on the F2 progeny of two natural accessions of Arabidopsis thaliana under two nutrient treatments.
  • • 
    We found that a single QTL explains most of the variation (77%) in susceptibility between accessions Columbia (Col-0) and San Feliu-2 (Sf-2), and its effect is independent of nutrients. The Sf-2 allele at this QTL is dominant and can reduce the bacterial population size by 31-fold, much like a classical resistance (R) gene. However, minor QTLs, whose effects are altered by nutrient treatment, were also detected.
  • • 
    Surprisingly, we found that none of the QTLs for susceptibility had any effect on fruit production, suggesting that the use of resistance genes for crop improvement and evolutionary analysis of plant–pathogen interactions requires caution.

Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Research in the past decade suggests that the outcome of an interaction between a plant and a pathogen depends on a series of complex signaling pathways (Kunkel & Brooks, 2002). However, much of the research has been focused on the activation of defense signaling pathways by the interaction between resistance (R) genes in the host and the corresponding avirulence (avr) genes in the pathogen, also known as ‘gene-for-gene’ (GFG) resistance (reviewed by Dangl & Jones, 2001; Dangl & McDowell, 2006). While GFG resistance seems to be prevalent among plant–pathogen interactions (Hammond-Kosack & Jones, 1997), other mechanisms also contribute to plant defense (Glazebrook et al., 1996; Sticher et al., 1997; Dewdney et al., 2000; Jones & Takemoto, 2004; Diener & Ausubel, 2005). In particular, it has been shown that variation in resistance among crops and natural populations is often quantitative (Young, 1996; Kover & Caicedo, 2001; Denby et al., 2004). The fact that most natural interactions vary quantitatively, together with numerous examples of single-gene resistance being quickly overcome in crops (Strange & Scott, 2005), has led to the suggestion that quantitative resistance is more durable than qualitative resistance. Yet, the molecular basis of quantitative variation in disease resistance remains mostly unknown.

Pathogens represent one of the biggest challenges in increasing plant yield, because of their large effects on crop growth, productivity and quality (Strange & Scott, 2005). Characterization of the genetic basis of disease resistance offers a promising avenue for the development of new resistant varieties by transferring resistance genes into crops (Stuiver & Custers, 2001) or by performing marker-assisted selection (Strange & Scott, 2005). However, these avenues will only be successful if resistance genes directly cause an increase in fitness and are not costly, assumptions that are not usually tested in the model organisms used to identify resistance genes.

The interaction between Arabidopsis thaliana and Pseudomonas syringae has been one of the primary pathosystems in which the molecular basis of plant resistance has been studied (reviewed in Dangl & Jones, 2001; Katagiri et al., 2002; Thatcher et al., 2005). Pseudomonas syringae is a bacterial plant pathogen known to cause disease on a variety of important crop plants (Hirano & Upper, 2000), and it has also been observed in natural populations of A. thaliana (Jakob et al., 2002). The possibility of transforming the strain Pst DC3000 (isolated originally from tomato (Lycopersicon esculentum)) with single avr genes has allowed the identification of five R genes and a number of other genes involved in the resistance pathway (Katagiri et al., 2002; Meyers et al., 2003). Recently, a survey of 19 natural accessions of A. thaliana has shown the existence of quantitative variation in susceptibility to Pst DC3000 (Kover & Schaal, 2002). This finding provided the opportunity to investigate the molecular basis of quantitative variation in susceptibility and its possible relationship to qualitative variation.

Quantitative variation in disease susceptibility is typically polygenic and affected by the environment (Young, 1996). Thus, to better understand the genetic basis of the quantitative variation in susceptibility to Pst DC3000 in A. thaliana, we performed two quantitative trait locus (QTL) studies. The first study investigated the genetic basis of quantitative variation in susceptibility among the F2 progeny of accessions Nossen (No-0) and Columbia (Col-0) (Kover et al., 2005), and found a few QTLs of small effect on symptom severity score. Here, we report the results of the second QTL analysis, which was performed on the F2 progeny of a cross between accessions Col-0 and San Feliu-2 (Sf-2). These two accessions represent the extremes of the quantitative variation in disease susceptibility previously observed (Kover & Schaal, 2002), and are consequently more variable than No-0 and Col-0. In addition, Sf-2 has been previously reported to be resistant to a number of P. syringae strains, including Pst DC3000 (Whalen et al., 1991).

In this study we investigate the genetic basis of the difference in susceptibility between Sf-2 and Col-0, the possible interaction between QTLs for susceptibility and environmental conditions, and the effect of QTLs for susceptibility on fruit production (plant yield). In contrast to the first QTL study, we found that differences between accessions Col-0 and Sf-2 in susceptibility to Pst DC3000 can be largely explained by a single genetic factor mapping to the bottom of chromosome 5. This previously unknown genetic factor, which behaves in the manner typical of classical plant R genes, equally affects bacterial growth and disease symptom development, but has no effect on fruit production. Although this QTL explained most of the variation, two other QTLs of small and environmentally dependent effect were also detected on chromosomes 1 and 4. Although QTLs for fruit production were observed, QTLs for susceptibility did not have any effect on fruit production.

Materials and Methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Plant material

The seeds for the Arabidopsis thaliana (L.) Heynh. parental lines Sf-2 (CS6857) and Col-0 (CS6673) used in this experiment were obtained from the Ohio Stock Center. These accessions were chosen because they represent the extremes of the natural variation in resistance to Pseudomonas syringae observed by Kover & Schaal (2002) in a study of 19 accessions of A. thaliana.

To obtain a segregating F2 population, we crossed emasculated Col-0 flowers with pollen from Sf-2 flowers. F1 plants from this cross, confirmed to be heterozygous using three genetic markers (nga151, nga248, and nga168), were self-pollinated to produce F2 seeds. All seeds used in this experiment came from two maternal plants.

A total of 358 F2 seeds were planted in 3-inch pots containing a soil mixture of 1 part vermiculite: 1 part Redi-earth (Scotts, Maryville, OH, USA): 1 part Fafard light mix #2 (Fafard, Springville, MA, USA). Pots were then randomly distributed among 12 trays. Seedlings in 22 pots died before any data were collected. Seeds from the parental lines were also planted in two pots in each tray. All trays were cold-treated for 3 d at 4°C before being placed to germinate in a growth chamber set at 21°C, a light:dark cycle of 8 : 16 h, and 75% humidity. Each pot contained two seeds, which were thinned down to one plant when seedlings were 3 wk old. Trays were rotated every week to a new location in the growth chamber to reduce the possible effect of microenvironmental variation. Half of the trays were randomly assigned to a fertilizer treatment, where plants were bottom-watered with a 50 : 50 mixture of water and 150 ppm Peter's Light Special Fertilizer 15-16-17 (Scotts) for 6 wk. The other trays were assigned to a control treatment, where plants were provided with the same amount of water (without fertilizer), and at the same time as the trays in the fertilizer treatment.

Pathogen inoculation and disease assays

Plants were inoculated when they were 4 wk old by submerging them in a bacterial solution containing 10 mm MgCl2, 0.02% L-77 Silwet (GE Silicones, Friendly, WV, USA) and 108 cells ml−1 of the bacterium P. syringae pv. tomato strain DC3000 (Pst DC3000) as described in Whalen et al. (1991). Trays were covered with a plastic dome for the first 24 h after inoculation to maintain high humidity. To determine the number of bacterial cells present per area of leaf tissue, four discs of leaf tissue (0.25 cm2 area) were collected from each plant with a cork borer 3 d after inoculation. These discs were collected randomly from different leaves, before any symptoms were visible. Leaf discs were ground in 10 mm MgCl2 solution and plated, after appropriate dilutions were made, on NYG agar (17.5 g Bacto Agar, 5 g Bacto Peptone, 20 ml glycerol, 3 g yeast extract per liter of media) plates containing 1 mg ml−1 of Rifampicin. The number of colony-forming units (CFU) per plate was counted 48 h later. To determine the severity of disease symptoms, each plant was visually inspected for disease symptoms 5 d after inoculation. Symptoms were scored on a predefined scale that takes into consideration the presence and extent of chlorosis and necrotic disease lesions. The scale ranges from 1 (no signs of disease symptoms) to 5 (extensive chlorosis and patches of necrotic lesions). Each plant was score independently by two people, and the means of the two scores were used in the analysis (the two scores were tightly correlated: r2 = 0.94).

Phenotypic measurements

To determine whether the nutrient treatment had an effect on plant growth before inoculation, we counted the number of leaves on each plant when they were 25 d old. At that time, we also used a subset of the seedlings thinned to determine seedling biomass, and tested the effect of nutrients on biomass before inoculation.

Because A. thaliana is an annual plant, the effect of pathogen infection on lifetime fitness can be estimated by total seed production upon senescence. Because Col-0 and Sf-2 do not differ significantly in the number of seeds produced per fruit (P. X. Kover et al., pers. comm.), individual fitness was estimated by counting total number of fruits after plants had senesced.

QTL analysis

Leaf tissue from each F2 plant was collected shortly after the onset of flowering, and DNA was extracted using the DNeasy Kit™ from Qiagen (Valencia, CA, USA). Each F2 individual was genotyped with 35 microsatellite and four cleaved amplified polymorphic sequence (CAPS) markers (Fig. 1). All primers used to genotype these markers are described in The Arabidopsis Information Resource (TAIR; http://www.arabidopsis.org). Conditions for polymerase chain reaction (PCR) amplifications generally followed the protocol of Bell & Ecker (1994), but the annealing temperature was individually optimized for each primer pair. Specific annealing temperatures used are available upon request.

image

Figure 1. Recombination map for the F2 progeny between Arabidopsis thaliana Columbia (Col-0) and San Feliu-2 (Sf-2). The names and positions of all polymorphic markers used in this study are included, with their recombinational distances as estimated by mapmaker.

Download figure to PowerPoint

The genetic map for the F2 cross shown in Fig. 1 was generated using mapmaker 3.0b (Lander et al., 1987), and was based on the genotypes of all 336 individuals from both treatments. Because the A. thaliana genome has been completely sequenced, the chromosome assignment and physical order of markers were imposed in the analysis. Recombination distances between markers were estimated using the Kosambi function. The overall size of our map is 455 cm, and the average interval size between genetic markers is 11.7 cm. Because the physical positions of all genetic markers used are known and available through TAIR (Rhee et al., 2003), we can determine that our map encompasses 100 of the 125 Mb genome of the genome of the annotated A. thaliana.

QTL analysis was performed using the qtl cartographer software (Basten et al., 1994). We searched for QTLs using interval mapping, with tests being performed every 2 cm. Statistical significance for the logarithm of the odds (LOD) scores obtained in the interval mapping was determined individually for each trait by performing 10 000 permutations using the method developed by Churchill & Doerge (1994). Composite interval mapping (CIM) was used to further analyze traits for which more than one putative QTL was identified in the same chromosome when using interval mapping. CIM can improve QTL positioning by parceling out the effects of QTLs of large effect and linkage disequilibrium (Broman, 2001). We ran CIM models with five covariate markers and a window size of 10 cm for each trait separately. Marker covariates for CIM were identified using a stepwise forward and backward regression. Confidence intervals for QTL position were established as the region around the peak location where there is one order of magnitude (1 LOD) change in the probability of a false positive (Lander & Botstein, 1989).

To identify QTLs that affect resistance and life-history traits in each treatment, we performed two separate QTL analyses: one only with control plants, and the other only with plants in the nutrient treatment. Because there is limited power to detect QTLs (Beavis, 1994), this approach can only identify the most significant QTL in each treatment. Consequently, if the same QTL is not observed in both treatments it does not necessarily mean that a trait is being affected by different QTLs in the two treatments. To determine whether there is a QTL-by-treatment effect, i.e. whether the effect of QTLs is nutrient dependent, we performed a two-way analysis of variance (ANOVA) of the effect of treatment on the QTL effect, including all QTLs that had significant effects in at least one treatment. To determine which QTLs affect the measured traits independently of the nutrient treatment, we performed a QTL analysis with all plants together. The effect of nutrients was partialled out before this analysis because differences in the magnitude of the effect could increase the variance in phenotypic data, and give a false negative for a QTL that indeed affected the measured trait in both treatments.

We searched for pairwise epistatic interactions using the canonical correlations analysis in sas (SAS Institute, Cary, NC, USA) as described in Cheverud & Routman (1995) and Routman & Cheverud (1997). The significance threshold used to evaluate the significance of epistatic effects was determined by performing a Bonferroni correction for the number of independent comparisons possible given the marker locations (as described in Cheverud, 2001). Pairs of loci that showed significant interaction (P < 0.05) were further analyzed with a full regression model where phenotypic values were regressed on additive, dominant, additive-by-additive (A × A), additive-by-dominant (A × D), dominant-by-additive (D × A) and dominant-by-dominant (D × D) genotypic effects.

Total phenotypic variance explained by the significant single-locus and epistatic QTL effects was determined using a regression model. Calculations were performed using proc reg in sas.

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Phenotypic differences between the parental lines and among the F2 progeny

As expected, Sf-2 was significantly less susceptible to Pst DC3000 than Col-0. Sf-2 had almost no disease symptoms 5 d after inoculation, and supported a significantly smaller bacterial population 3 d after inoculation than Col-0 (Table 1). Thus, from a qualitative perspective, there was no effect of treatment on the response of the two accessions to Pst DC3000. However, a two-way ANOVA indicated a significant effect of accession, as well as a significant interaction between accession and treatment (F = 363.9, P < 10−4, and F = 6.4, P = 0.02, respectively). The actual differences in scores between treatments were quite small, and their biological significance is hard to interpret. In contrast, when bacterial population size was analyzed, we only observed a significant effect of accession (F = 137.1, P < 10−4).

Table 1.  Mean values (± standard error) for the traits measured in the two parental lines of Arabidopsis thaliana (Columbia (Col-0) and San Feliu-2 (Sf-2)), and in the F2 progeny
TraitF2aCol-0aSf-2a P b
  • a

    Numbers in parentheses indicate the sample size.

  • b

    P-value indicating the significance of a t-test comparing the two parental lines.

  • CFU, colony-forming units.

Disease symptom score
 Control2.3 ± 0.09 (168)3.8 ± 0.15 (12)1.8 ± 0.13 (12)10−9
 Fertilizer treatment2.2 ± 0.09 (168)4.1 ± 0.14 (12)1.4 ± 0.06 (12)10−11
Log CFU cm2
 Control5.8 ± 0.08 (168)6.7 ± 0.18 (12)4.9 ± 0.12 (12)10−7
 Fertilizer treatment5.8 ± 0.08 (168)6.8 ± 0.18 (12)4.8 ± 0.13 (12)10−8
Number of fruits per plant
 Control180.6 ± 5.42 (161)206.2 ± 12.17 (12)200.4 ± 9.48 (12)0.710
 Fertilizer treatment533.8 ± 15.63 (168)549.6 ± 35.83 (12)730.6 ± 41.45 (12)0.003

By contrast, fruit production under pathogen pressure heavily depended on the nutrient treatment (Table 1). In the control treatment, where no nutrients were added, fruit production in Col-0 plants and that in Sf-2 plants were not significantly different. However, in the fertilizer treatment, Sf-2 produced significantly more fruits than Col-0. These results could suggest that the less susceptible parent can increase fruit production at a higher rate than the susceptible parent in the presence of nutrients. Alternatively, Sf-2 plants may have other loci that always improve fruit production in the presence of fertilizer, even in the absence of pathogens. Because in this experiment all plants were inoculated, we cannot determine the effect of resistance (i.e. the difference in fruit production between an inoculated and a healthy plant), or distinguish these two hypotheses.

The distribution of traits among the F2 progeny under both nutrient treatments is shown in Fig. 2. While all traits showed quantitative variation typical of complex inheritance, the distribution in symptom scores is also compatible with the segregation of a single, dominant Mendelian gene. If we consider plants with a disease symptom score below 3 (i.e. plants that show little chlorosis, and no water-soaked lesions) as ‘resistant’, and all the plants that received symptoms scores of 3 and above as ‘susceptible’, we obtain a ratio of 246 resistant: 90 susceptible plants. This ratio is not significantly different (χ2 = 0.59, P = 0.74) from the 3 resistant: 1 susceptible plant ratio expected when a dominant R gene is segregating in a F2 progeny. The same results were obtained if we scored plants in each nutrient treatment separately (data not shown). Thus, it is possible that a single R gene that confers resistance to Pst DC3000 can explain the variation in susceptibility between Col-0 and Sf-2. This does not exclude the possibility that other genes of small effect also contribute to the observed variation.

image

Figure 2. Frequency distributions of measured traits. (a, c, e) Arabidopsis thaliana plants grown in the control treatment; (b, d, f) plants grown in the fertilizer treatment. Gray bars, trait distribution within F2 progeny; speckled bars, Columbia (Col-0); open bars, San Feliu-2 (Sf-2). CFU, colony-forming units.

Download figure to PowerPoint

The effect of fertilizer on susceptibility and fruit production

We observed no statistically significant difference between treatments in the average number of leaves per plant at 25 d (7.1 vs 7.0 leaves in the control and nutrient treatment, respectively). While we found that the biomass of plants in the control treatment was on average larger than the biomass of plants in the nutrient treatment (29 and 25 mg per plant, respectively), the difference was not statistically significant (P = 0.07). These results indicate that, at the time of inoculation, plants receiving the nutrient treatment did not have any size advantage over control plants. Although the experimental design precluded obtaining further destructive samples, visual observation indicated that plants in the nutrient treatment were clearly larger by the time plants were 6 wk old.

To determine the effect of nutrients on susceptibility and fruit production, we performed a two-way ANOVA on the data from the F2 progeny, with treatment and tray as factors. Tray was considered a good indicator of microenvironmental variation because all pots in a given tray remained together during the whole experiment, and therefore experienced more similar microenvironmental conditions. Because all plants in a tray received the same nutrient treatment, tray effect was nested within treatment effect. We found that tray grouping only significantly affected the early growth of a plant, as measured by the number of leaves present when plants were 25 d old (F10,323 = 23.4, P = 10−32). As early growth was not correlated with susceptibility or yield (data not shown), we assume microenvironmental effects did not significantly affect our results. The fertilizer treatment did not have a statistically significant effect on symptom score or bacterial population. However, the fertilizer treatment clearly led to a significant increase in fruit production (F1,10 = 476.4, P = 10−64; also see Table 1).

QTL analysis for symptom severity and bacterial population size

The initial analysis of symptom severity scores revealed a QTL of very large effect at the bottom of chromosome 5 when plants from either treatment were used (Table 2). This QTL by itself explained between 71 and 75% of the observed variance in disease symptoms. A subsequent ANOVA did not detect any QTL-by-treatment interaction effect (Table 2), confirming that this QTL had the same effect in both treatments.

Table 2.  List of quantitative trait loci (QTLs) identified for susceptibility and yield, with their positions, significance and effects
TraitTreatmentChromosomePosition (cm)Confidence interval (cm)LODa R 2 b Additive effectcDominance effectQTL by treatment?d
  • a

    Significance of the detected QTL.

  • b

    Proportion of the variance in the trait measured that is explained by the QTL.

  • c

    The additive effect indicates the effect of San Feliu-2 (Sf-2) alleles in comparison to a Columbia (Col-0) allele. Thus, when this value is negative, the sf-2 allele reduces susceptibility.

  • d

    The type of effect by treatment interaction is indicated in parentheses: A × T, additive effect; D × T, dominance effect.

  • LOD, logarithm of the odds; ns, not significant.

Disease scoreControlI29.119.1–33.74.50.12  0.35−0.07 (ns)No
FertilizerI13.184.0–21.15.60.10  0.30 0.03 (ns) 
CombinedI21.113.1–27.18.90.11  0.33−0.01 (ns) 
ControlIV61.954.4–68.73.60.09  0.07 (ns) 0.42Yes (D × T)
ControlV49.647.6–51.653.90.71 −1.31−0.83No
FertilizerV55.451.6–57.448.80.75 −1.28−0.72 
CombinedV51.649.6–53.4100.50.72 −1.26−0.78 
Bacterial population sizeControlIV61.948.4–68.7 3.00.08  0.27 0.22Yes (D × T)
ControlV51.647.6–53.425.00.48 −0.95−0.30Yes (D × T)
FertilizerV55.451.6–57.433.70.61 −1.10−0.74 
CombinedV51.649.6–53.456.10.52 −1.00−0.53 
Number of fruitsFertilizerI80.680.3–90.63.10.10−72.9113.34 (ns)Yes (A × T)
FertilizerIII18.716.3–26.72.90.03 45.79 (ns)80.32No

At the same location, we also found a QTL of large effect on the size of the bacterial population growing in leaves (Table 2). Bacterial population size over the whole F2 progeny was highly correlated with symptom score (R = 0.77, P < 0.001). While this QTL for bacterial population size was also significant in both treatments, we detected a significant QTL-by-treatment interaction. This interaction was mostly attributable to differences in the magnitude of the dominance effects (the Sf-2 allele had a larger dominance effect in the nutrient treatment), as the directions of the effects of the alleles were the same in both treatments (Tables 2, 3).

Table 3.  The average phenotypic value for each of the three traits measured, given the genotype at marker ciw9 (linked to quantitative trait loci (QTL) for susceptibility on chromosome 5)
Genotype at marker Ciw9Symptom severity scoreaBacterial population size (CFU cm2)No. of fruits per plant
  1. a Symptom scores vary from 1 (less susceptible) to 5 (most susceptible).

  2. Col-0, Columbia; Sf-2, San Feliu-2.

Col-0/Col-0 (n = 94)3.762791341
Col-0/Sf-2 (n = 158)1.83 90376
Sf-2/Sf-2 (n = 81)1.34 38342

As the same QTL was being detected under both nutrient treatments, we performed a QTL analysis combining the data from both treatments. With a larger number of plants we can obtain more precise confidence intervals for the location of this QTL and a better estimate of the effect of this QTL independent of the nutrient treatment. This analysis indicated that the most likely position of this QTL is at 51.6 cm on our recombinational map (Fig. 1), between markers Ciw9 and AthPHYC. Assuming a linear relationship between the recombinational distances observed on our map and the known physical distances between these markers (Rhee et al., 2003), we could determine that this QTL is located approx. 400 kb upstream from the genetic marker ciw9. In both treatments this QTL behaved in a semidominant fashion, with the allele from Sf-2 (the less susceptible accession) reducing the symptom score, and being dominant over the Col-0 allele. The effect of this QTL on plant susceptibility is better illustrated in Table 3, where we show the average symptom score of plants with each of the three possible genotypes at the ciw9 marker. The presence of a single Sf-2 allele at this marker reduced the bacterial population size by 31-fold, while a plant homozygous for the Sf-2 allele at the ciw9 marker had a bacterial population size 73 times smaller than a plant homozygous for the Col-0 allele. Such an effect is comparable to the effect of other known R genes, which typically reduce bacterial growth by 102 (Katagiri et al., 2002).

Because the QTL on chromosome 5 had such a large effect, it could potentially have masked the effect of any other QTL segregating in the population. Thus, we performed a second set of QTL analyses, partialling out the effect of marker ciw9. This analysis revealed two additional QTLs for disease susceptibility on chromosomes 1 and 4.

The QTL on chromosome 1 affected symptom score under both nutrient treatments, and explained between 10 and 12% of the variance in symptom scores (after removing the variation attributable to the QTL on chromosome 5). At this QTL, the Sf-2 allele actually increased the susceptibility of the plant and had no significant dominance (see Table 2). When data from both nutrient treatments were combined, it was found that the same location significantly contributed to the variation in symptom score, and no QTL-by-environment interaction effect was detected (Table 2). The best estimate for the location of this QTL (from the data on combined treatments) is at 21.1 cm on our map, which corresponds to approx. 2000 kb downstream from marker Nga63.

The QTL on chromosome 4 affected both disease symptoms and bacterial population growth (Table 2), but was only significant in the control treatment. This QTL is located at 61.9 cm (approx. 611 kb downstream from marker Nga1139), and only had a significant dominant effect. The effect at this QTL was such that the heterozygote had a significantly higher symptom score (more susceptible) than either homozygote. As expected, a significant QTL-by-environment interaction was detected when plants from both treatments were combined. However, this QTL only explained a small proportion of the variation in symptom scores and bacteria population size (8 or 9%, respectively).

We found only one statistically significant interaction between QTLs on chromosomes 1 and 5, which affected symptom severity (Table 4). The statistical significance of this interaction resulted from the fact that the effect of the QTL on chromosome 1 was much more pronounced when the QTL on chromosome 5 was homozygous for the Col-0 (or susceptible) allele (Table 5). Still, this interaction explained only a relatively small amount of variance (6%). No significant epistatic interactions were detected for bacterial growth, and removing the effect of the large QTL on chromosome 5 did not reveal any other additional interactions.

Table 4.  List of significant epistatic interactions detected
TraitTreatmentChromosomes and positions of interacting
QTLs LODa P a Typeb R 2 c
  • a

    Significance of the detected quantitative trait loci (QTLs).

  • b

    The type of effect by treatment interaction is indicated in parentheses: A × A, additive × additive interaction; A × D, additive × dominance interaction; D × D, dominance × dominance interaction.

  • c

    Proportion of the variance in the trait measured that is explained by the QTL.

  • LOD, logarithm of the odds.

Symptom scoreCombinedI (39.8 cm)V (53.4 cm)6.030.0001A × A0.06
No. of fruitsControlI (39.8 cm)II (8.0 cm)6.150.0001A × D, D × D0.12
I (60.3 cm)IV (57.9 cm)7.961 × 10−5A × A, A × D, D × A0.14
Table 5.  Illustration of the effect of the interaction between quantitative trait loci (QTLs) detected on chromosomes 1 and 5 for symptom severity scores on the average score for Arabidopsis thaliana plants with each of the nine genotypic combinations at the two markers closest to the identified QTL
Ciw9Nga392
Col-0/Col-0aCol-0/Sf-2aSf-2/Sf-2a
  1. a Numbers in parentheses indicate the number of plants in each genotypic class.

  2. Col-0, Columbia; Sf-2, San Feliu-2.

Col-0/Col-03.4 (24)3.6 (47)4.3 (33)
Col-0/Sf-21.7 (37)1.8 (84)1.8 (37)
Sf-2/Sf-21.3 (18)1.4 (47)1.4 (18)

QTL analysis for fruit production under pathogen pressure

We were only able to detect QTLs for fruit production for plants grown under the fertilizer treatment. Two QTLs were detected: one at the bottom of chromosome 1, and one at the top of chromosome 3 (Table 2). They explained 10 and 3% of the variance in fruit production, respectively. When plants from both treatments were combined, a significant QTL effect by treatment effect was observed. These two QTLs for fruit production did not coincide with any QTL identified for symptom score or bacterial growth.

Although no QTLs of main effect were found in the control treatment, two significant epistatic interactions were detected (Table 4). The locations of the QTLs involved in these interactions were not significant per se in either treatment. In addition, these interactions were not significant when the two treatments were combined, suggesting that the effects of these QTLs are not consistent across treatments. Still, the two epistatic interactions together explained a large proportion of the phenotypic variance (46%).

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

A new R-gene-like factor in A. thaliana?

Our study provides strong evidence for the existence of a single gene of large effect in the accession Sf-2 capable of conferring resistance to Pst DC3000. Because of the strength of this QTL, we were able to obtain a very tight confidence interval when plants from both treatments were combined. We estimate this QTL to be located between 16 083 and 17 060 kb on chromosome 5, with its most likely position being ∼463 kb upstream from ciw9. This interval contains 321 predicted genes, including five genes with a TIBS-NBS-LRR domain typical of previously identified R genes: At5g40910, At5g41540, At5g41550, At5g41740 and At5g41750. In this interval we also find a gene analogous to cf-4 (resistance gene against Cladosporium fulvum), an R gene identified in tomato against the fungus Cladosporium fulvum. While these genes are the most obvious candidates, it is also possible that a locus with a different function and structure underlies the observed QTL on chromosome 5.

While our study is the first to provide evidence that resistance in Sf-2 to Pst DC3000 is mainly attributable to an R-gene-like factor; a previous study reported that the Sf-2 accession was resistant to 30 different strains of P. syringae (Whalen et al., 1991). While it is possible that this accession has accumulated a large number of R genes over its evolutionary history, this hypothesis seems very unlikely. Further mapping of this region is ongoing to narrow down the most appropriate candidate gene for this QTL. We are also determining whether this QTL is specific against Pst DC3000, or whether it represents some form of a broad-spectrum resistance factor.

The generation of transgenic Pst DC3000 strains containing single avr genes has been one of the main avenues for the identification of new R genes in A. thaliana mutants or natural accessions (Katagiri et al., 2002; Meyers et al., 2003). More and more researchers are turning to natural variation in A. thaliana to unravel the genetic basis of complex traits. However, the existence of a natural accession with resistance to Pst DC3000 suggests that more than one virulent strain should be routinely used when assaying for natural variation.

The effect of nutrients on the susceptibility of A. thaliana to Pst DC3000

Studies on quantitative variation of disease resistance in plants have often found that environmental variation plays an important role in the expression of disease resistance (e.g. Huber & Watson, 1974; Mingeot et al., 2002; Nam et al., 2006). Identification of resistance QTLs that are environment independent is of particular importance to the agricultural industry because it allows the creation of crop varieties appropriate to a broad range of environments.

We investigated in particular the effect of nutrient concentration because disease resistance in A. thaliana has been shown to be costly (Mauricio, 1998; Tian et al., 2003). We hypothesized that a larger defense response would be mounted when more nutrients were available. However, we found that the majority of the variance in susceptibility to P. syringae was explained by a single QTL with nutrient-independent effects. A QTL-by-environment interaction was only observed for the QTL of smaller effect on chromosome 4.

The relationship between susceptibility and plant yield

The identification of resistance genes offers a promising avenue for transformation of crops with resistance genes from nonrelatives (Stuiver & Custers, 2001; Strange & Scott, 2005). Implicit in this research program is the assumption that plants carrying disease resistance will experience less reduction in yield when infected, because of the pleiotropic effects of resistance genes.

A previous study also failed to find a correlation between susceptibility and productivity (Kover & Schaal, 2002). However, the effect of resistance on plant yield is better tested in a segregating population where genes that affect fruit production should segregate independently from genes that confer resistance, unless they are pleiotropic or physically linked. We found that the specific chromosomal segment that contains the QTL on chromosome 5 did not have any effect on yield despite significantly reducing symptoms and bacterial population size. Plants homozygous for Sf-2 alleles at the ciw9 marker produced on average the same number of fruits as plants homozygous for the Col-0 allele.

From an evolutionary perspective, plants can respond to pathogens through resistance or tolerance traits. While resistance traits reduce pathogen growth and spread, tolerance mediates fitness improvement in the presence of pathogens (e.g. upregulation of photosynthesis; Inglese & Paul, 2006). Our results suggest that research on the genetic basis of tolerance traits might provide a more useful avenue to elucidate fruit production under pathogen pressure in A. thaliana. The same approach may also be more appropriate to improve yield in crops.

The genetic basis of quantitative variation in disease resistance

Despite the importance of quantitative variation in susceptibility in natural populations and crop plants (reviewed in Young, 1996; Kover & Caicedo, 2001), its molecular basis remains unclear. In particular, it is not clear whether the genetic pathways that mediate quantitative and qualitative variation in susceptibility are the same or involve completely different genes. Dissection of the genetic basis of quantitative variation in susceptibility in A. thaliana offers a good opportunity to investigate the molecular basis of quantitative variation and its relationship to qualitative resistance, because so much is known already about the molecular basis of qualitative variation.

The two QTL studies we performed in the interaction between A. thaliana and P. syringae gave seemingly very different results. In the No-0 × Col-0 cross we detected only QTLs of small effect (Kover et al., 2005), while in this study we found a potential R gene of large effect. However, in both studies we observed a QTL in the same location on chromosome 1, where the allele from Col-0 affects only symptoms by reducing it. Thus it is possible that this QTL in common is attributable to the same underlying loci. While the confidence interval on chromosome 1 is quite large, the peak of this QTL is located very closely to RPS5, one of the five characterized R genes for P. syringae (Warren et al., 1998). The effect of the QTL on chromosome 1 is compatible with the Col-0 allele being a dominant allele that confers resistance, as would be expected for RPS5. It is also interesting that the effect of the QTL on chromosome 1 was more pronounced when the QTL of major effect was homozygous for the susceptible allele (Table 5). Thus, the QTL interaction can mask some of the underlying quantitative variation.

While it would be premature to draw firm conclusions about the relationship between quantitative and qualitative variation in susceptibility, our results support the hypothesis that quantitative resistance results from the interaction of multiple R genes of small effect, and that classical R genes are just alleles with particularly extreme effects at the same loci (Gebhardt & Valkonen, 2001; Katagiri et al., 2002). Co-location of QTL and R genes has been previously observed in other systems (Gebhardt & Valkonen, 2001; Kover & Caicedo, 2001; Wilson et al., 2001; Wisser et al., 2005) but it has not been possible to discard the hypothesis that genes with common function (i.e. disease resistance) are clustering in the same location (Wisser et al., 2005). Further exploitation of the region of chromosome 5 identified in this study will provide an exciting new avenue for investigating the molecular basis of quantitative variation and its relationship to classical R genes in the more genetically tractable plant A. thaliana.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

We thank B. Kunkel for invaluable technical training, and for many useful comments on early versions of this manuscript. We also thank J. B. Wolf for helpful discussions, and R. Collier and V. Brown for technical assistance. This project was funded by the Monsanto Corporation under the Washington University-Monsanto Company Plant Science Agreement. Support was also provided by a NSF minority postdoctoral grant to PXK.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References