Secondary metabolites influence Arabidopsis/Botrytis interactions: variation in host production and pathogen sensitivity


  • Daniel J. Kliebenstein,

    Corresponding author
    1. Department of Plant Sciences, University of California-Davis, One Shields Avenue, Davis, CA 95616, USA, and
      (fax +1 530 752 9659; e-mail
    Search for more papers by this author
  • Heather C. Rowe,

    1. Department of Plant Sciences, University of California-Davis, One Shields Avenue, Davis, CA 95616, USA, and
    Search for more papers by this author
  • Katherine J. Denby

    1. Department of Molecular and Cell Biology, University of Cape Town, Private Bag, Rondebosch 7701, South Africa
    Search for more papers by this author

(fax +1 530 752 9659; e-mail


Numerous studies have suggested that plant/pathogen interactions are partially mediated via plant secondary metabolite production and corresponding pathogen tolerance. However, there are inconsistent reports on the ability of particular compounds to provide resistance to a pathogen. Most of these studies have focused on individual isolates of a given pathogen, suggesting that pathogens vary in their sensitivity to plant-produced toxins. We tested variability in virulence among pathogen isolates, and the impact on this by plant production of, and pathogen tolerance to, secondary metabolites. Botrytis cinerea isolates showed differing sensitivity to purified camalexin, and camalexin-sensitive isolates produced larger lesions on camalexin-deficient Arabidopsis genotypes than on the wild type. In contrast, the camalexin-insensitive isolate produced lesions of similar size on wild-type and camalexin-deficient Arabidopsis. Additional analysis with Arabidopsis secondary metabolite biosynthetic mutants suggests that Botrytis also has variable sensitivity to phenylpropanoids and glucosinolates. Furthermore, Botrytis infection generates a gradient of secondary metabolite responses emanating from the developing lesion, with the Botrytis isolate used determining the accumulation pattern. Collectively, our results indicate that Arabidopsis/Botrytis interactions are influenced at the metabolic level by variations in toxin production in the host and sensitivity in the pathogen.


Secondary metabolites are defined as compounds present in specialized cells that are not necessary for the cell's survival, but are required for the plant's survival in the environment (Stahl, 1888). Secondary metabolites are believed to aid plant fitness by preventing insect herbivory and pathogen attack, as well as aiding reproduction through pollinator or seed-dispersal attraction via floral scent or coloration (Wink, 1988). Numerous secondary metabolites, including phenolics, hydroxycinnamic acids, sesquiterpenes and glucosinolates, have in vitro antimicrobial activity (Osbourn, 1999; Van Etten et al., 1989). However, in planta activity of these compounds has been difficult to verify because pathogens can differ in their sensitivity to specific compounds (Thomma et al., 1999; Tierens et al., 2001), possibly due to genotypic variation (Osbourn, 1999; Quidde et al., 1998). In addition, the plant can vary genetically or developmentally for the ability to produce a given compound (Brown et al., 2003; Kliebenstein et al., 2001a,b; Mert-Turk et al., 2003a; Rodman, 1980; Van Etten et al., 1982). Thus a well defined system examining genotypic variation in the pathogen and host could further our understanding of the in planta antimicrobial activities of secondary metabolites.

Arabidopsis thaliana is a well established system that contains diverse secondary metabolites, some with known in vitro antimicrobial activity (Chapple et al., 1994; Kliebenstein, 2004). Furthermore, Arabidopsis mutants that alter specific secondary metabolite pathways have been isolated (Alonso et al., 2003; Chapple et al., 1994; Kliebenstein, 2004). Using available genomics tools, it is possible to identify an Arabidopsis mutant collection specifically to test the in planta role of secondary metabolites. For example, glucosinolate mutants have been used to demonstrate that glucosinolate amount and structure can control plant/pathogen and plant/herbivore interactions (Jander et al., 2001; Lambrix et al., 2001; Tierens et al., 2001), and oat mutants lacking saponins are more susceptible to a variety of fungal pathogens (Papadopoulou et al., 1999). The Arabidopsis pad3 mutant has been used widely to test the role of camalexin, a major Arabidopsis phytoalexin, in mediating defense against a range of pathogens. Testing pathogen virulence on pad3 versus PAD3 plants has shown that camalexin affects only some pathogens (Ferrari et al., 2003; Glazebrook and Ausubel, 1994; Glazebrook et al., 1997; Mert-Turk et al., 2003b; Reuber et al., 1998; Thomma et al., 1999). While several studies have suggested that Botrytis cinerea is sensitive in planta to camalexin, others have not (Denby et al., 2004; Ferrari et al., 2003; Thomma et al., 1999). However, this discrepancy could be explained by intra-species variation in camalexin tolerance, as these studies utilized different B. cinerea isolates.

A factorial study of a plant secondary metabolite mutant collection with diverse pathogen genotypes, coupled with an efficient pathogen virulence assay, could elucidate which metabolites control the interaction, and whether pathogen isolates vary genetically for secondary metabolite tolerance. In the present study we obtained a collection of Arabidopsis mutants, altered in specific steps of secondary metabolism, and tested their interaction with different B. cinerea isolates selected for variation in in vitro camalexin sensitivity. Further, we analyzed the accumulation of the major methanol-extractable secondary metabolites in all Arabidopsis mutant × Botrytis isolate interactions to test for possible pleiotropy in the Arabidopsis mutant phenotypes. This allowed us to test the role of specific Arabidopsis secondary metabolites in controlling host resistance to Botrytis. We also tested the influence of genetic variation in B. cinerea on Arabidopsis secondary metabolite accumulation.


Variation for tolerance to camalexin in B. cinerea

Pathogenic fungi have two main avenues to tolerate plant-produced toxins: either active tolerance (enzymatic detoxification); or passive tolerance (avoidance). Several studies using different B. cinerea isolates have produced contrasting data on the role of camalexin in protecting Arabidopsis against B. cinerea infection (Denby et al., 2004; Ferrari et al., 2003; Thomma et al., 1999). We hypothesized that the differences between studies were due to variations in camalexin sensitivity among B. cinerea isolates. We assayed three different B. cinerea isolates for in vitro camalexin sensitivity during both spore germination and hyphal growth. This assay identified one isolate, DKUS-1, as being more tolerant to camalexin during hyphal growth than the other isolates (Figure 1). This tolerance was also observed during spore germination (Figure 1). Botrytis cinerea therefore has genetic variation in its ability to tolerate camalexin.

Figure 1.

Variable camalexin tolerance among Botrytis isolates.
Hyphal growth and spore germination of three Botrytis cinerea isolates (DKUS-1, DGUS-1 and GLUK-1) in the presence of increasing camalexin concentration. 100% indicates values in the absence of camalexin. Standard error bars are shown from three independent experiments.

To confirm this variation in camalexin tolerance in planta, we inoculated three Arabidopsis mutants defective in camalexin production (pad3, cyp79B2 and cyp79B2/B3 double mutant: Glawischnig et al., 2004; Zhao et al., 2002; Zhou et al., 1999; Figure 2) with the three Botrytis isolates (DKUS-1, DGUS-1 and GLUK-1). The cyp79B3 single mutant was not included in this comparison as it is not defective in camalexin accumulation (Glawischnig et al., 2004). Three days post-inoculation the camalexin-sensitive isolates DGUS-1 and GLUK-1 both showed increased growth on the camalexin-deficient mutants (Table 1). In contrast, the camalexin-tolerant isolate DKUS-1 showed no significant difference in lesion development between the wild-type Col-0 and the camalexin biosynthetic mutants (Table 1). Similar results were obtained after 2 days (data not shown). Camalexin accumulation still occurs after infection with the DKUS-1 isolate (5.8 pmol mm−1 lesion circumference in infected Col-0 plants compared with no detectable camalexin in uninfected plants), hence these results indicate that DKUS-1 is camalexin-tolerant both in vitro and in planta.

Figure 2.

Biosynthetic location of secondary metabolite mutants used in this study.
Arabidopsis secondary metabolites and schematic biosynthetic pathways with the specific biosynthetic location of the mutants used in this study are shown in bold italics. Single arrows between compounds show single biosynthetic steps, but pathways have been simplified for ease of presentation. Figure based on previous representations of these pathways (Glawischnig et al., 2004; Hemm et al., 2003; Shirley et al., 1995, 2001). TRP, tryptophan; MET, methionine; PHE, phenylalanine.

Table 1.  Lesion development on Arabidopsis secondary metabolite mutants by three B. cinerea isolates
Arabidopsis GenotypeDKUS-1DGUS-1GLUK-1
  1. NS indicates a P-value greater than 0.10. Bold font indicates a lesion size significantly different to that of the corresponding wildtype using Tukey's HSD test.

  2. aMean is the mean lesion diameter in mm, 3 days after infection.

  3. bP is the P-value that the mean lesion size for that given Arabidopsis genotype × Botrytis isolate pair is significantly different than the relevant Arabidopsis wildtype × Botrytis isolate pairing as determined by anova.


Impact of secondary metabolites on B. cinerea lesion development

Glucosinolates and phenylpropanoids, including both flavonols and hydroxycinnamic acids, can provide Botrytis resistance in other species but have not been tested in Arabidopsis (Goetz et al., 1999; Terry et al., 2004; Wurms et al., 2003). To test if the presence of glucosinolates or phenylpropanoids alters the Arabidopsis/Botrytis interaction, we inoculated the same three B. cinerea isolates on mutants with decreased glucosinolate production: cyp79B3 and ref2-3 (Glawischnig et al., 2004; Hemm et al., 2003); flavonol production: tt5 (Shirley et al., 1995); and sinapate production: sng1, sng2 and fah1 (Lehfeldt et al., 2000; Lorenzen et al., 1996; Meyer et al., 1996; Shirley et al., 2001) (Figure 2). This mutant collection should allow testing of specific compounds or compound classes for the ability to alter Arabidopsis/Botrytis interactions.

In the Col-0 background, the GLUK-1 isolate showed increased lesion size on the ref2-3 mutant, but no significant differences were detected for any of the B. cinerea isolates on the sinapate biosynthetic mutants (Table 1). In the Ler background, the tt5 mutant showed increased susceptibility to the GLUK-1 isolate (Table 1). This suggests that aliphatic glucosinolates and Arabidopsis flavonols, or other compounds produced via chalcone isomerase, provide resistance to the B. cinerea isolate GLUK-1. However, before drawing such conclusions it was necessary to check the metabolite profile of these biosynthetic mutants.

Secondary metabolite mutant pleiotropy

To test whether our collection of mutants exhibited pleiotropy in their secondary metabolite phenotypes, we measured the content of camalexin, all detectable glucosinolates, all detectable sinapic acid esters and all detectable flavonols in the mutants, before and after infection with each Botrytis isolate. This analysis showed that both Arabidopsis genotype and Botrytis isolate had a significant impact on the accumulation of nearly every compound tested (Table 2; Tables S1 and S2).

Table 2.  Statistical analysis of the impact of variation in Arabidopsis genotypes (mutants versus wild type) and treatment (control or Botrytis isolate infection) on secondary metabolite accumulation
AccnaTraitArabidopsis genotypeb TreatmentcGenotype × treatmentd
  1. Bold font indicates anova model components that have a significant impact on the given trait using anova.

  2. aWhether trait is tested in wild-type and mutant collection from Col or Ler accession background; there are different mutant subsets for each accession background.

  3. bWhether anova identified a significant impact of genotype variation (different Arabidopsis mutants) on the trait.

  4. cWhether anova identified a significant impact of treatment variation (uninfected control or infection with different Botrytis isolates) on the trait. Control and three Botrytis isolates were treated as four different treatments for all variables, except for camalexin, for which only the three Botrytis isolate treatments were utilized as there was no measurable camalexin accumulation in control leaves.

  5. dWhether anova identified a significant interaction of genotype (different Arabidopsis mutants) and treatment in controlling the trait (that different treatments cause different effects on particular genotypes).

  6. eF-value obtained for specified anova model component on the given trait.

  7. fP-value obtained for specified anova model component on the given trait.

 3-methylsulfinylpropyl GS1.38NS8.16<0.0010.90NS
 4-methylsulfinylbutyl GS1.67NS20.80<0.0010.99NS
 3-methylthiopropyl GS3.030.00211.32<0.0010.76NS
 7-methylsulfinylheptyl GS1.48NS18.38<0.0010.95NS
 4-methylthiobutyl GS2.410.01321.51<0.0011.13NS
 8-methylsulfinyloctyl GS1.46NS18.74<0.0011.22NS
 indol-3-ylmethyl GS18.10<0.00158.38<0.0012.66<0.001
 N-methoxy-indol-3-ylmethyl GS15.82<0.00112.38<0.0013.62<0.001
 Aliphatic GS2.890.00823.28<0.0010.92NS
 Indolic GS20.22<0.00143.05<0.0013.31<0.001
 Total GS3.090.00229.16<0.0011.11NS
 Sinapoyl malate63.57<0.00131.08<0.0011.68NS
 Total flavonol13.63<0.0017.67<0.0010.94NS
 Flavonol to phenylpropanoid137.56<0.00125.15<0.0011.28NS
 3-hydroxypropyl GS2.730.05012.40<0.0010.59NS
 3-methylsulfinylpropyl GS0.66NS1.83NS1.31NS
 4-methylsulfinylbutyl GS2.11NS0.99NS1.14NS
 7-methylsulfinylheptyl GS5.800.0019.84<0.0011.91NS
 8-methylsulfinyloctyl GS1.04NS10.63<0.0010.89NS
 Indol-3-ylmethyl GS3.750.01549.69<0.0010.46NS
N-methoxy-indol-3-ylmethyl GS2.11NS19.58<0.0011.51NS
 Aliphatic GS2.730.05012.66<0.0010.69NS
 Indolic GS3.080.03341.84<0.0010.46NS
 Total GS2.43NS16.80<0.0010.68NS
 Sinapoyl malate9.15<0.00127.96<0.0010.81NS
 Total flavonol33.96<0.0012.55NS0.93NS
 Flavonol to phenylpropanoid21.99<0.00111.58<0.0010.40NS

Ref2-3 affected the accumulation of many compounds (Figure 3; Table S1) but, surprisingly, did not show the reduction in aliphatic glucosinolates previously described (Hemm et al., 2003). This is possibly due to different growth conditions and/or age of the leaf material. Furthermore, ref2-3 showed an unexpected reduction in camalexin accumulation, suggesting that the increased B. cinerea GLUK-1 lesion size in ref2-3 plants (Table 1) could be due to this camalexin deficiency. The ref2-3 genotype was confirmed by direct sequencing of the mutation.

Figure 3.

Secondary metabolite accumulation in Arabidopsis biosynthetic mutants.
Secondary metabolites were assayed in 5-week-old leaf tissue in Arabidopsis biosynthetic mutants before and after infection with Botrytis cinerea. Except for camalexin and indole glucosinolates, mean and standard error from all leaves per genotype are presented (mean value combining both control and infected samples). This is a valid comparison when the relative level of metabolites between different genotypes is the same for all treatments. For camalexin, the mean was calculated only from infected samples, as there is no detectable camalexin in uninfected leaves. For indolic glucosinolates, the mean from only uninfected leaves was calculated because relative levels between treatments vary depending on the plant genotype (there is a significant genotype × treatment interaction; Table 2). Asterisks indicate significant differences from the respective wild type at P = 0.05 as determined using Tukey's HSD test. Each mutant was assayed with four leaves per plant, six plants per replicate, and two independent replicates for a total of 48 leaves per mutant. 4-methylthiobutyl is a Col-0-specific compound and is not detectable in Ler or Ler mutants. Indolic GS, sum of all tryptophan-derived glucosinolates; aliphatic GS,: sum of all methionine-derived glucosinolates.

The phenylpropanoid mutants also had a pleiotropic impact with altered glucosinolate and phenylpropanoid accumulation (Figure 3; Tables S1 and S2), making definitive conclusions about the role of secondary metabolites difficult. Indole glucosinolates, aliphatic glucosinolates and sinapoyl malate levels were all reduced in tt5 mutants compared with wild type, in addition to the expected reduction in flavonol content. However, glucosinolate levels were not significantly different between tt5 and sng2, and drastically reduced sinapoyl malate levels in fah1 and sng1 did not affect lesion size. This suggests that the increased sensitivity of tt5 to Botrytis GLUK-1 could be due to the large reduction in flavonol content in addition to any slight impact of decreased glucosinolate and sinapoyl malate levels (Table 1; Figure 3).

The cyp79 mutants affected both camalexin and glucosinolate accumulation, however, only the camalexin content correlated with Botrytis lesion size. The camalexin-defective genotypes (cyp79B2 and cyp79B2/B3) showed increased lesion size, while the camalexin overaccumulating genotype (cyp79B3) showed a trend towards decreased lesion size, although not significantly (Figure 3; Table 1). All three genotypes had reduced indolic glucosinolate levels (Figure 3). This suggests that the major factor leading to altered Botrytis lesion development in these mutants is altered camalexin accumulation.

Pad3 was the least pleiotropic mutant in this collection, affecting only camalexin accumulation (Figure 3). The fact that the pad3 Botrytis lesion size mirrors the cyp79B2 and cyp79B2/B3 results further suggests that the effects of cyp79 mutations on Botrytis lesion development (Table 1) are due to altered camalexin accumulation. Pad3 and cyp79B2/B3 mutants both have extremely low levels of camalexin, yet the cyp79B2/B3 line showed greater Botrytis susceptibility than pad3 after infection with isolates DGUS-1 and GLUK-1. This may be the result of both camalexin and glucosinolate loss, as the cyp79B2/B3 mutant alters indole and aliphatic glucosinolate accumulation in addition to camalexin.

In summary, mutations in biosynthetic pathways can have pleiotropic effects on levels of plant metabolites. HPLC analysis of secondary metabolite production allowed us to verify the impact of mutations on secondary metabolism, and to strengthen the link between specific metabolites and altered Botrytis lesion development. In addition to the major role of camalexin, glucosinolates and flavonols appear to play a minor role in determining susceptibility to B. cinerea in an isolate-dependent fashion.

Spatial secondary metabolite accumulation

Botrytis cinerea can avoid plant toxins by spatially repressing their accumulation in a zone near developing lesions. Spatial regulation of plant compounds near the infection site is a common theme in plant/pathogen interactions (Mansfield et al., 1974; Rahe, 1973). It is possible that Botrytis lesion development leads to a repression of Arabidopsis secondary metabolism, allowing the developing pathogen to avoid antimicrobial effects of these secondary metabolites. To test whether repression of secondary metabolism occurs in infected Arabidopsis leaves, and to determine if this varies with lesion development by different B. cinerea isolates, we analyzed accumulation of the major methanol-soluble Arabidopsis secondary metabolites (camalexin, glucosinolates, flavonols and sinapates) in a spatial context surrounding the developing lesion.

Leaves from Col-0 plants inoculated with B. cinerea GLUK-1 isolate were incubated until approximately 25% of the leaf had been consumed by the developing lesion. The leaves were then radially dissected from the lesion (Figure 4). Each leaf disk was individually extracted, and detectable methanol-extractable secondary metabolites were measured. The data were analyzed by anova. Distance from the lesion had the greatest effect on compound accumulation, while individual leaves did not differ significantly for most compounds (Table S3). In no case was there a significant interaction of leaf and distance (Table S3), therefore the data were combined from all leaves to estimate the metabolite spatial regulation (Figure 4; Table S3).

Figure 4.

Spatial regulation of Arabidopsis secondary metabolites around a developing Botrytis cinerea lesion.
Schematic diagram of tissue samples at defined distances from a developing GLUK-1 lesion and level of secondary metabolites in these leaf disks. The lesion + 1 mm samples contained approximately 1 mm healthy tissue. Distance given for other leaf disks is that from center of leaf disk to edge of lesion. Mean of 10 leaves from 10 wild-type Col-0 plants used in this analysis and standard error are shown. Columns with different letters are significantly different from one another at P = 0.05 within a compound, as determined using Tukey's HSD test. Indolic GS, sum of all tryptophan-derived glucosinolates; aliphatic GS, sum of all methionine-derived glucosinolates.

The phytoalexin camalexin showed a dramatic increase only in the region immediately surrounding the lesion (Figure 4). In contrast to camalexin, glucosinolate, flavonol and sinapate concentrations decreased closer to the lesion, with almost complete disappearance in tissue directly proximal to the lesion (Figure 4; Table S3). For all three isolates, lesion material that did not contain any intact, healthy tissue had no detectable level of any of the glucosinolates, phenylpropanoids or camalexin (data not shown). These results suggest that on infection by Botrytis, the plant undergoes a major metabolic re-organization away from the production of glucosinolates and phenylpropanoids towards production of camalexin and possibly other compounds.

There were significant differences in how Arabidopsis secondary metabolism changed in response to infection with different isolates, demonstrating that each isolate does not simply induce a generic Botrytis pattern of secondary metabolite accumulation (Figure 5; Tables S1 and S2). Camalexin was highly induced by the camalexin-sensitive isolates, but not as dramatically by the tolerant isolate (Figure 5). DKUS-1 did not repress flavonol accumulation to the same extent as DGUS-1 or GLUK-1 (Figure 5). However, Arabidopsis/DKUS-1 interactions are not marked by a general lack of secondary metabolite response, as sinapoyl malate, aliphatic and indolic glucosinolates all decreased as in the Arabidopsis/GLUK-1 interaction (Figure 5). The DGUS-1 and GLUK-1 isolates showed similar patterns for camalexin and phenylpropanoid levels, but DGUS-1 had an enhanced glucosinolate repression (Figure 5).

Figure 5.

Secondary metabolite accumulation in Arabidopsis leaves after infection with different Botrytis cinerea isolates.
Secondary metabolite levels following infection with three different Botrytis isolates, showing standard errors for each mean. Different letters indicate values significantly different from one another at P = 0.05 within a compound utilizing Tukey's HSD test. Data from all Col-0 genotypes were pooled to estimate mean and standard error for all compounds apart from indolic glucosinolates. This is valid as the relative levels of these metabolites between treatments did not vary in different genotypes. For indolic glucosinolates, only the wild-type Col-0 data were used because, on particular genotypes, relative levels between treatments changed. Col-0 genotypes were used to minimize any potential impact of variation in genomic background, however the Ler genotypes showed similar patterns (Table S2). Indolic GS, sum of all tryptophan-derived glucosinolates; aliphatic GS, sum of all methionine-derived glucosinolates. ND, not detectable.


The major findings of this work are that B. cinerea contains genetic variation for sensitivity to camalexin, and that changes in Arabidopsis secondary metabolism vary after infection by different isolates. In addition, the use of Arabidopsis biosynthetic mutants has identified secondary metabolites with a role in resistance against B. cinerea.

Variable camalexin tolerance in B. cinerea

It has long been debated whether phytoalexins are the cause or consequence of induced plant pathogen defenses (Muller, 1961; Muller et al., 1939; Peuppke and VanEtten, 1976). Some Arabidopsis/Botrytis studies showed that camalexin provides Botrytis resistance, and others did not (Denby et al., 2004; Ferrari et al., 2003; Thomma et al., 1999). However, these studies differed in the Botrytis isolate used. In this study, we show that B. cinerea isolates differ in their camalexin tolerance, and this difference determines the isolate's ability to produce lesions on camalexin-deficient mutants (Figure 1; Table 1). Hence, when the pathogen is phytoalexin-sensitive, the phytoalexin contributes to host resistance, whereas against phytoalexin-tolerant pathogens, phytoalexins are a non-functional consequence of induced defenses.

This theory is further supported by the finding that Arabidopsis still induces camalexin production in response to the camalexin-tolerant DKUS-1 Botrytis isolate (Figure 5). Interestingly, this isolate induces the least camalexin accumulation of the isolates tested (Figure 5). This may be because DKUS-1 can convert camalexin to other non-toxic compounds, and camalexin is rapidly broken down by the pathogen, leading to lower detectable levels. This form of camalexin tolerance has been found in several other plant pathogens (Pedras and Ahiahonu, 2002; Pedras and Khan, 1997), but we did not find any of the expected catabolism products predicted by this model (data not shown). Alternatively, the rapid growth of DKUS-1 lesions (nearly twice that of the other two isolates) may kill plant tissue before large amounts of camalexin can be synthesized. It is also possible that this B. cinerea isolate is not effectively recognized by the plant, resulting in poor induction of defense responses. Different stimuli can induce varying levels of camalexin in Arabidopsis (e.g. Zhao et al., 1998); however, repression of the other measured secondary metabolites by DKUS-1 is similar to that of GLUK-1. The identification of additional camalexin-tolerant isolates with differing lesion growth rates will help resolve this issue.

There was no camalexin detectable in the lesions produced by any of the three Botrytis isolates. This could indicate that all three isolates are able to degrade camalexin (despite differences in camalexin sensitivity); however, lesion material also has undetectable levels of glucosinolates and phenylpropanoids, suggesting a more generic explanation. Cells within the wet lesion are unlikely to be intact, and it is known that during Botrytis infection high levels of reactive oxygen species accumulate (Govrin and Levine, 2000), creating an environment in which these plant compounds could be non-enzymatically oxidized or otherwise degraded.

Secondary metabolite/Botrytis interactions

We developed a single extract methodology to assay known methanol-extractable secondary metabolites in Arabidopsis to test for pleiotropic effects of mutations and propose roles for specific compounds in Arabidopsis/Botrytis interactions. This approach confirmed the effect of ref2-3 on sinapoyl malate and highlighted a previously unreported increase in flavonol production in this mutant (Figure 3). Our data also indicated an association between the glucosinolate and phenylpropanoid pathways, as the phenylpropanoid mutants sng1 and sng2 displayed slight, but significant, decreases in glucosinolate accumulation (Figure 3). Thus there appears to be cross-talk between the glucosinolate and phenylpropanoid pathways, although they are postulated to be spatially separated in the plant.

We also identified previously unknown associations within phenylpropanoid metabolism whereby mutants altering sinapate metabolism also altered flavonol production. This effect was mutant-specific in that sng1 (which shifts sinapate production from sinapoyl malate to sinapoyl glucose; Lehfeldt et al., 2000) significantly increased flavonol production, while fah1 (which blocks synthesis of all sinapates; Meyer et al., 1996) decreased flavonol accumulation (Figure 3). We were able to circumvent this interpretation difficulty somewhat by using multiple Arabidopsis mutants and analyzing secondary metabolite levels.

Inoculating multiple camalexin-deficient mutants with several B. cinerea isolates confirms previous conclusions that camalexin plays a major role in resistance against this pathogen (Denby et al., 2004; Ferrari et al., 2003). Assaying a range of Arabidopsis biosynthetic mutants for Botrytis susceptibility also indicated that glucosinolates and flavonols may influence lesion development. For example, the flavonol-deficient tt5 line had increased susceptibility to isolate GLUK-1. This susceptibility could be phenylpropanoid-related, as the glucosinolate and sinapate impacts of this mutant are similar to those found in sng2 (has similar reduction in indole glucosinolates) and sng1 (has similar reduction in aliphatic glucosinolates), which do not alter Botrytis lesion size or flavonol accumulation (Figure 5; Table 1). However, it is also possible that the slight reduction in both aliphatic and indole glucosinolates in tt5 is sufficient to alter Botrytis susceptibility. The enhanced susceptibility of cyp79B2/B3 compared with pad3, along with metabolite measurements in these mutants, suggests that indole glucosinolates do contribute to resistance against B. cinerea.

Glucosinolates can deter insects and some Arabidopsis pathogens (Jander et al., 2001; Kliebenstein et al., 2002; Tierens et al., 2001). Purified glucosinolate breakdown products were not toxic to one B. cinerea isolate (Tierens et al., 2001), but there may be genetic variation in the pathogen for tolerance to these compounds. The single knockout cyp79B3, which has reduced indolic glucosinolate levels, did not significantly alter B. cinerea lesion size. However, camalexin levels in these mutants were increased (Figure 3), possibly masking the effects of lowered indolic glucosinolate content. The analysis of glucosinolate-specific mutations in a camalexin-deficient background would enable a rigorous test of the impact of glucosinolates on Arabidopsis/Botrytis interactions.

Localized host responses to Botrytis

Previous work has shown that PR-1 expression is induced locally at the edge of the developing Botrytis lesion (Ferrari et al., 2003). Additionally, phytoalexin responses have typically been found in a spatially limited zone around Botrytis infections (Mansfield et al., 1974, 1980; Rahe, 1973). Our analysis of secondary metabolites in Arabidopsis showed that camalexin also exhibits a narrow zone of induction adjacent to the lesion (Figure 4). In contrast to this local camalexin induction, we found a localized reduction in the majority of other compounds assayed (Figure 4). The reduction was dependent on the distance from the lesion, with the decrease in metabolite level greatest near the lesion.

The spatial response gradient in the leaf suggests that a signal regulating metabolite levels is produced at the edge of the developing lesion and diffuses outwards towards uninfected tissue. This signal may be produced by the host or pathogen, or both. Changes in plant secondary metabolite accumulation varied after infection by the different Botrytis isolates, possibly because isolates produce a different signal or different intensity of signal, or because isolates induce differential synthesis of a plant signal. This phenomenon of pathogen genotype affecting secondary metabolite accumulation has been shown recently for Pseudomonas syringae, a bacterial pathogen. Antimicrobial secondary metabolites are secreted from Arabidopsis roots infected with a non-pathogenic strain of P. syringae. However, the pathogenic strain P. syringae pv. tomato DC 3000 blocks secretion of these metabolites in a type III secretion system-dependent manner (Harsh et al., 2005).

It is not clear whether this metabolic reprogramming is an active process driven by Botrytis to evade plant toxins (potentially glucosinolates and flavonols), or a plant response to synthesize camalexin preferentially at the expense of other metabolites. As the degree of metabolic reprogramming appears to be inversely correlated with susceptibility to B. cinerea, it suggests the latter. Infection with the DGUS-1 isolate led to the smallest lesions and greatest change in secondary metabolite levels. However, it is equally plausible that there is interplay between host and pathogen such that DGUS-1 drives metabolic reprogramming to evade plant toxins, but the large accumulation of camalexin in the host limits pathogen growth.

In conclusion, the interaction between Botrytis and Arabidopsis is complex and is influenced by the genotype of both pathogen and host. Botrytis cinerea possesses genetic variation for tolerance to plant secondary metabolites, and the host varies in secondary metabolite accumulation in infected leaves. Furthermore, the interaction between specific host genotypes and pathogen genotypes affects the accumulation of secondary metabolites, and ultimately the severity of disease symptoms.

Experimental procedures

Analysis of Botrytis in vitro camalexin tolerance

Three B. cinerea isolates were used in this study. Isolate GLUK-1 was obtained from Dr Gary Loake, University of Edinburgh, UK, and was labeled ‘pepper’ in a previous study (Denby et al., 2004). DGUS-1 was obtained from Dr Doug Gubler, University of California, Davis, CA, USA; DKUS-1 was from single-spore propagation of a field-collected strawberry from San Diego County, CA, USA, infected with B. cinerea. All three isolates were verified as B. cinerea utilizing genus-specific ELISA and species-specific DNA sequence analysis (Dewey et al., 2000; Staats et al., 2005). The three isolates had differential growth rates on various defined media. However, the relative rankings between the isolates varied depending on the media utilized, and did not correlate with in planta virulence (data not shown).

The in vitro camalexin tolerance of Botrytis was measured during both spore germination and hyphal growth. At the hyphal growth stage, tolerance was measured as previously described (Ferrari et al., 2003). Briefly, this involves infusing potato dextrose agar plates with defined amounts of camalexin, inoculating with a disk of Botrytis hyphae and measuring growth over several days. Tolerance was measured as percentage inhibition with regard to growth on control plates without camalexin. Germination of Botrytis spores was analyzed by harvesting Botrytis spores as described (Denby et al., 2004). A defined amount of camalexin was added to a suspension of 1 × 106 spores ml−1 and spore germination was followed microscopically over 24 h. Camalexin tolerance was measured as percentage inhibition of germination of 200 spores compared with control suspensions without added camalexin.

Plant growth

All plants were grown for 5 weeks in Premier Pro-Mix B mix soil under 150 μmol m−2 sec−1 photosynthetic photon flux density light with a 12-h photoperiod. The plants were grown in 104-cell planting trays at a single plant per cell, providing a density of 1600 plants m−2.

Botrytis cinerea infection of secondary metabolite mutants and analysis of secondary metabolite accumulation

We set up a factorial experiment to test simultaneously how the Botrytis isolates interacted with the different Arabidopsis genotypes. The experiment described below allowed us to measure the interaction at both lesion development and secondary metabolite-accumulation levels. Seven plants of each Arabidopsis genotype were planted in a randomized fashion and grown for 5 weeks. Three independent planting replicates of the entire experiment were carried out.

At 5 weeks of age, four leaves of a similar developmental age were harvested from each plant and one leaf from each plant was placed into each of four plastic humidity trays containing 1.5% phytagar to a depth of 1 cm. Leaves in one tray were inoculated with a 4-μl drop of half-strength grape juice as a control. Leaves in the remaining three trays were inoculated with a 4-μl drop of one of the B. cinerea isolates, GLUK-1, DKUS-1 or DGUS-1, as described (Denby et al., 2004). Total leaf area and diameter of the developing lesion were measured after 48 and 72 h using imagej and high-resolution digital pictures of the leaves. A 1-cm scale was included in each image, which enabled calibration of the measurements. One leaf of each Arabidopsis genotype from each of the three inoculated trays was used for an ELISA assay to test for the presence of B. cinerea (Dewey et al., 2000). Botrytis cinerea was detected at similar levels per unit lesion area in all lesions tested (data not shown).

After the 72 h time point, remaining leaves were weighed and placed into microtiter tubes for secondary metabolite extraction, as described below. Glucosinolates, flavonols and sinapates are repressed near the lesion, therefore they were expressed as amount per mg healthy tissue. The area of healthy tissue was measured from digital images of leaves and the healthy mass was calculated from total mass after determining that the average mass per mm2 healthy tissue was 10-fold higher than that of infected tissue. Camalexin production is confined to a narrow zone around the lesion, hence the camalexin values were expressed as amount per mm lesion circumference as most of the leaf tissue would not be accumulating camalexin.

The lesion and secondary metabolite data were divided into two data sets, one for the Arabidopsis genotypes from the Col-0 background, the other for the Arabidopsis genotypes from the Ler genotypic background. This was done to minimize the impact of variation in genomic background. Leaves were classified by plant genotype, treatment and replicate number. The control and different B. cinerea isolate infections were considered independent treatments. Replicate was not a significant component of any model, and therefore was dropped from further analysis. The anova model was secondary metabolite level or lesion size = plant genotype + treatment + plant genotype × treatment + error. This model tests the significance of the influence of plant genotype, treatment (control and different Botrytis isolates), or a specific interaction between plant genotype and one or more treatments on lesion size and secondary metabolite levels. This model was used to obtain the F-value, significance, means and standard error for each variable (Table 2). Using all the data to obtain means and standard errors for each factor (genotype or treatment) provides better statistical significance. It is a valid analysis when there is no interaction between experimental factors, for example, sinapate levels in sng1 are low compared with wild type in all treatments. Sinapate levels which were high in sng1 in one treatment (and low in the others) would indicate an interaction between genotype and treatment. Data from all leaves were therefore used for the analysis reported in Table 2; Figures 3 and 5, with the following exceptions. For lesion size (Table 1) and camalexin (Figure 3), means and standard errors were re-calculated using only the Botrytis-treated samples, as control samples had no lesions or camalexin accumulation. Means and standard errors for indolic glucosinolate levels in Figure 3 were re-calculated using just control (uninfected) samples, and in Figure 5 using just wild-type samples. This was because there is a significant interaction between genotype and treatment for indolic glucosinolate levels (Table 2), meaning that the relative levels between treatments vary depending on the genotype of the plant. All means were tested for a significant difference from the appropriate wild-type control utilizing Tukey's honestly significant difference (HSD) test. All possible comparisons with wild type were conducted to identify all significant differences.

Single-extract secondary metabolite extraction and HPLC analysis

To measure the major methanol-extractable plant secondary metabolites, either a single leaf or single leaf disk was harvested into a 1.5-ml tube in a 96-well microtiter plate (Micronics, ISC Bioexpress USA, Kaysville, UT, USA) containing 500 μl 90% methanol and one 3.8-mm stainless steel ball bearing (National Precision Bearing Group of Mechatronics, Inc., Preston, WA, USA). The lid was replaced, the microtiter plate placed into a paint shaker, and the tissue homogenized by shaking for 5 min (Fluid Management, Chicago, IL, USA). The capped plate was incubated at 65°C for 15 min. The microtiter plates were then centrifuged at 3220 g for 15 min to precipitate any particulate matter. Each extract was split into three 100-μl fractions using a multichannel pipette. Two fractions were placed in separate microtiter plates (one for camalexin analysis and one for phenylpropanoid analysis), while the third was pipetted into a microtiter multi-channel column plate containing diethylaminoethyldextran (DEAE) sephadex to allow for glucosinolate purification.

Camalexin analysis

Fifty microliters of the extract was run on an Agilent Lichrocart 250-4 RP18e 5-μm column using an Agilent 1100 series HPLC (Agilent, Wilmington, DE, USA). Camalexin was detected using a DiodeArray (DAD) at 330 nm and with a fluorescence detector at emission 318 nm/excitation 385 nm. Separation was achieved using the following program with aqueous acetonitrile: 5-min gradient from 63 to 69% acetonitrile, 30-sec gradient from 69 to 99% acetonitrile, 2 min at 99% acetonitrile and a post-run equilibration of 3.5 min at 63% acetonitrile (Denby et al., 2004). Purified camalexin was used to produce a standard curve to identify and quantify camalexin.

Phenylpropanoid analysis

Fifty microliters of the extract was run on an Agilent Hypersil 250-4 ODS 5-μm column using an Agilent 1100 series HPLC (Borevitz et al., 2000). Phenylpropanoids were detected using a DAD at 254 and 330 nm. Separation was achieved using the following program with 1% phosphoric acid and acetonitrile. Starting conditions were 15% acetonitrile and 85% 1% phosphoric acid. After injection, there was an 11-min gradient to 32.5% acetonitrile. The column was then flushed with a 2-min gradient to 99% acetonitrile and a 3-min hold at 99% acetonitrile. The column was returned to starting conditions and allowed to equilibrate for 6 min. This was the shortest equilibration time to maintain peak shape with this column and packing material. The major flavonols and sinapates were purified and identified by liquid chromatography/mass spectrometry (LC-MS-MS). These purified standards were then used to facilitate compound quantification.

Glucosinolate analysis

The third fraction was used to purify and analyze glucosinolates using the basic high-throughput sephadex/sulfatase protocol described previously (Hogge et al., 1988; Kliebenstein et al., 2001a,b). Glucosinolates were identified and quantified in comparison with purified standards (Kliebenstein et al., 2001a,b; Reichelt et al., 2002).

Spatial analysis of secondary metabolites

Two leaves of similar developmental age from six different Col-0 wild-type plants were detached and placed in a humidity tray containing 1.5 l 1.5% phytagar. Leaves were drop-inoculated with 4 μl B. cinerea isolate GLUK-1 at 5000 spores ml−1 in half-strength grape juice (Denby et al., 2004). Lesions were allowed to develop for 3 days until about 25% of the leaf surface had been infected. One leaf from each of the six different plants was dissected into leaf disks using a 6-mm cork borer, and the unused leaves were discarded (Figure 4). This provided on average 24 samples per leaf. Disks were collected radially, starting at the edge of the lesion. Lesions were also harvested as either including approximately 1 mm healthy tissue, or just lesion tissue with no healthy leaf material. All lesions were verified as containing B. cinerea using an ELISA-based detection assay (data not shown) (Dewey et al., 2000). Four leaves were sacrificed to verify that there was no detectable B. cinerea outside the lesion borders (data not shown). Leaf disks were individually extracted and analyzed for secondary metabolites as described above. Disks were classified by leaf number and lesion proximity. The results were analyzed using anova as described above, using the model: secondary metabolite level = leaf + distance + leaf × distance + error. This tested the impact of the leaf and distance from lesion on the accumulation of various secondary metabolites, and generated the mean and standard error for each compound at each distance.