Arabidopsis myrosinases TGG1 and TGG2 have redundant function in glucosinolate breakdown and insect defense


*(fax +1 607 254 2958; email


In Arabidopsis and other Brassicaceae, the enzyme myrosinase (β-thioglucoside glucohydrolase, TGG) degrades glucosinolates to produce toxins that deter herbivory. A broadly applicable selection for meiotic recombination between tightly linked T-DNA insertions was developed to generate Arabidopsis tgg1tgg2 double mutants and study myrosinase function. Glucosinolate breakdown in crushed leaves of tgg1 or tgg2 single mutants was comparable to that of wild-type, indicating redundant enzyme function. In contrast, leaf extracts of tgg1tgg2 double mutants had undetectable myrosinase activity in vitro, and damage-induced breakdown of endogenous glucosinolates was apparently absent for aliphatic and greatly slowed for indole glucosinolates. Maturing leaves of myrosinase mutants had significantly increased glucosinolate levels. However, developmental decreases in glucosinolate content during senescence and germination were unaffected, showing that these processes occur independently of TGG1 and TGG2. Insect herbivores with different host plant preferences and feeding styles varied in their responses to myrosinase mutations. Weight gain of two Lepidoptera, the generalist Trichoplusia ni and the facultative Solanaceae-specialist Manduca sexta, was significantly increased on tgg1tgg2 double mutants. Two crucifer-specialist Lepidoptera had differing responses. Whereas Plutella xylostella was unaffected by myrosinase mutations, Pieris rapae performed better on wild-type, perhaps due to reduced feeding stimulants in tgg1tgg2 mutants. Reproduction of two Homoptera, Myzus persicae and Brevicoryne brassicae, was unaffected by myrosinase mutations.


Plants growing in nature are exposed to a wide variety of herbivores, pathogens and other biotic stresses. Dynamic physiological responses, including the production of defensive secondary metabolites, help plants to survive such environmental challenges. In plants of the order Capparales, which includes cruciferous crops and the model plant Arabidopsis thaliana, the glucosinolate–myrosinase system serves as a major chemical defense mechanism against insects, bacteria and fungi (Bones and Rossiter, 1996; Raybould and Moyes, 2001). Glucosinolates (β-thioglucoside-N-hydroxysulfates) are amino acid-derived secondary metabolites that can be cleaved by the enzyme myrosinase (β-thioglucoside glucohydrolase, TGG; EC; Bones and Rossiter, 1996; Wittstock and Halkier, 2002). Myrosinase and glucosinolates are localized in separate plant cells, myrosin cells and S-cells, respectively, or in separate intracellular compartments (Husebye et al., 2002; Koroleva et al., 2000; Thangstad et al., 2004). Herbivore attack, particularly by chewing insects, causes tissue disruption, thereby bringing glucosinolates into contact with myrosinase, resulting in the production of a variety of toxic degradation products, including isothiocyanates, nitriles, thiocyanates, oxazolidine-2-thiones and epithionitriles (Wittstock and Halkier, 2002), which tend to deter generalist herbivores, but can serve as attractants for crucifer specialists.

There is great diversity in the glucosinolate profiles of cruciferous plants. Secondary modifications of amino acid side chains produce >140 known glucosinolate structures (Fahey et al., 2001). In Arabidopsis, there are at least 37 different glucosinolates, with side chains derived from methionine, tryptophan, phenylalanine and isoleucine (Reichelt et al., 2002). In addition, regulated synthesis and breakdown of glucosinolates produces quantitative and qualitative changes during plant development (Brown et al., 2003; Petersen et al., 2002). Finally, although glucosinolates and myrosinase represent a preformed plant defense, various stresses, including insect and pathogen attack (Brader et al., 2001; Mewis et al., 2005), can induce alterations in the glucosinolate profile.

The enzymatic mechanism of myrosinase has been well characterized in vitro (Bones and Rossiter, 1996), where the enzyme is activated by ascorbic acid and does not show substrate specificity for particular glucosinolates. Glucosinolate breakdown during herbivory is likely to be considerably more complicated. Myrosinases are part of a complex enzyme system that includes myrosinase-binding proteins (Falk et al., 1995) and myrosinase-associated proteins (Taipalensuu et al., 1997). However, with the exception of epithiospecifier proteins that promote hydrolysis of glucosinolates to nitriles rather than isothiocyanates (Lambrix et al., 2001), the function of the accessory proteins in these complexes remains unknown.

In the absence of mutations, research on the in planta defensive role of glucosinolate breakdown by myrosinase has been limited to correlative studies. For instance, one of the two major quantitative trait loci (QTL) controlling myrosinase activity in the Landsberg erecta (Ler) × Columbia (Col-0) Arabidopsis recombinant inbred line population (Lister and Dean, 1993) is negatively correlated with Trichoplusia ni (cabbage looper) herbivory (Kliebenstein et al., 2002). Similarly, although it has been proposed that myrosinase is involved in the developmental glucosinolate breakdown occurring during leaf senescence and seed germination, in the absence of myrosinase mutations there has been no conclusive evidence for or against this hypothesis (Brown et al., 2003).

The simple genetics and sequenced genome of Arabidopsis make this an excellent model organism for generating mutants and exploring the molecular mechanisms of the glucosinolate–myrosinase system. In contrast to various Brassicaceae species, which contain multiple forms of myrosinase genes divided into several subfamilies (Rask et al., 2000; Thangstad et al., 1993; Xue et al., 1992), only one myrosinase gene family (TGG1–TGG6) is present in Arabidopsis (Xu et al., 2004; D. Andersson and J. Meijer, Swedish University of Agricultural Sciences, Uppsala, Sweden, personal communication). There are two sets of paired myrosinases, each with one intervening gene of unknown function: TGG1 (At5g26000) and TGG2 (At5g25980), and TGG5 (At1g51470) and TGG6 (At1g51490). TGG3 (At5g48375) and TGG4 (At1g47600) are single genes.

In both Arabidopsis and Brassica napus (oilseed rape), TGG1 is expressed in stomatal guard cells and phloem cells in all above-ground tissues, but is not present in roots (Andreasson et al., 2001; Husebye et al., 2002; Thangstad et al., 2004). Mutations in the VAM3/SYP22 gene (At5g46860) lead to increased numbers of myrosinase-containing cells and over-expression of both TGG1 and TGG2 (Ueda et al., 2006). The remaining four Arabidopsis myrosinase genes have received relatively little attention. TGG3 is a frame-shifted pseudogene that is expressed in stamens and petals (Zhang et al., 2002). Expression of TGG4 and TGG5 has been detected only in root tissue, and TGG6 is only expressed in flowers, specifically in the stamens (Toufighi et al., 2005; Zimmermann et al., 2004).

Despite their chemical defenses, Arabidopsis and other Brassicaceae are consumed by both generalist and specialist insect herbivores. Larvae of T. ni, a generalist lepidopteran herbivore, tend to feed on the older, less well-defended parts of the plant (Broadway and Colvin, 1992). Myzus persicae (green peach aphid) is a broad generalist that feeds readily on crucifers (Blackman and Eastop, 2000) and avoids the effects of glucosinolate breakdown by an as yet unknown mechanism. Plutella xylostella (diamondback moth) larvae feed exclusively on crucifers and produce a gut sulfatase that cleaves glucosinolates to form inactive desulfoglucosinolates (Ratzka et al., 2002). A different detoxification mechanism is found in Pieris rapae (white cabbage butterfly) in the form of a gut enzyme that directs glucosinolate breakdown toward nitriles, which appear to be less toxic than isothiocyanates (Wittstock et al., 2004). Brevicoryne brassicae (cabbage aphid), a crucifer-feeding specialist, sequesters plant-derived glucosinolates and produces its own myrosinase as a defense against predators (Bridges et al., 2002; Pontoppidan et al., 2001). In addition to being able to detoxify glucosinolates, crucifer-specialized herbivores such as Pl. xylostella, Pi. rapae and B. brassicae often use glucosinolates or their breakdown products as an attractive signal for the identification of suitable host plants (Giamoustaris and Mithen, 1995).

To initiate an analysis of myrosinase activity, we identified T-DNA insertions that inactivate TGG1 and TGG2, the two major above-ground myrosinases of Arabidopsis. A double knockout line was generated using a selection for meiotic recombination between the T-DNA insertions in these tandemly duplicated genes. Further research with the single and double myrosinase mutants was directed toward answering fundamental questions about the role of myrosinases in plant–insect interactions and glucosinolate metabolism. What is the role of myrosinases in plant defense? Are myrosinases involved in the developmental glucosinolate decreases observed during senescence and germination? Do myrosinases show glucosinolate specificity in degradation after tissue damage? Are the TGG1 and TGG2 myrosinases functionally redundant?


TGG1 and TGG2 expression patterns

Different expression patterns could explain the need for multiple myrosinase genes in Arabidopsis. As has been reported previously (Andreasson et al., 2001; Husebye et al., 2002; Thangstad et al., 2004), we observed TGG1 promoter:GUS expression in all stomatal guard cells and in phloem-associated cells in above-ground tissues (Figure 1a–d). In contrast, TGG2 promoter:GUS fusion lines show GUS expression only in phloem-associated cells (Figure 1e–h). This expression pattern was present only in above-ground tissues, and the same phloem-associated cells were stained as in the TGG1–GUS experiment. Rare cases of TGG1 or TGG2 fusion lines showing GUS expression in root tissue were probably artifacts of the insertion site, as such expression has not been reported previously. No transcripts of either gene were detected by RT-PCR in root tissue of wild-type seedlings grown on agar plates (data not shown).

Figure 1.

 Spatial distribution of TGG1 and TGG2 expression in leaves, the flower stalk, the stamen and carpel as determined by TGG1 and TGG2 promoter-driven GUS expression.
(a–d) Histochemical GUS staining in transgenic lines carrying the TGG1 promoter:GUS construct.
(e–h) GUS expression in TGG2 promoter:GUS lines. Tissue of 4-week-old plants was used in all cases, except leaves, which were taken from 2-week-old plants. Arrows indicate expression in phloem cells. GC, guard cells. Bar 80 μm.

Isolation of homozygous tgg1 and tgg2 mutants

To explore the physiological function of TGG1 and TGG2 myrosinases in glucosinolate breakdown and insect defense, mutant plants with defects in both genes were identified. Homozygous T-DNA insertions were identified in lines from both the SALK (Alonso et al., 2003) and SAIL (Sessions et al., 2002) collections. DNA sequencing showed that the base pair positions of the insertions relative to the ATG start codon are 2018 bp for tgg1-2 (SALK_069615), 1697 bp for tgg2-1 (SALK_038730; Ueda et al., 2006), 2149 bp for tgg1-3 (SAIL_786_B08), and −246 bp for tgg2-2 (SAIL_237_G11). Semi-quantitative RT-PCR was used to show that the tgg1 and tgg2 mutants lack the respective mRNAs (Figure 2). To confirm allelism, the tgg1-2 and tgg1-3, and tgg2-1 and tgg2-2 mutants were crossed to one another, and the absence of the respective transcripts in the F1 progeny was verified by RT-PCR. Unless specifically stated otherwise, all subsequent experiments were done with both alleles of each mutant gene, and produced the same results in every case.

Figure 2.

 Levels of TGG1 and TGG2 transcripts in wild-type and homozygous tgg1, tgg2 and tgg1tgg2 double mutants.
Amounts of transcripts were determined by amplifying specific TGG1 and TGG2 cDNA fragments from 1 μg total RNA using semi-quantitative RT-PCR. Shown are cDNA fragments of three wild-type replicates, three independent homozygous isolates of the tgg1 and tgg2 mutants from the SAIL and SALK T-DNA collections, and three independently isolated tgg1tgg2 double mutants. UBQ10 served as an internal control. Primers used in this analysis as well as their respective locations are described in Table S1.

TGG1 transcript levels in tgg2 mutants and TGG2 transcripts in tgg1 mutants appear to be slightly higher than in the wild-type (Figure 2; Table S1). Quantitative analysis of mRNA levels in the three genotypes revealed that tgg2 mutants have approximately 25% more TGG1 mRNA than the wild-type (nWT = 9, ntgg2 = 17, P < 0.05, Student's t-test). In tgg1 mutants, TGG2 levels were similarly elevated, but the effect was not significant (nWT = 9, ntgg1 = 17, P > 0.05, Student's t-test).

Generation of a tgg1tgg2 double mutant

Whereas SALK T-DNA insertion lines contain the NPTII gene, conferring kanamycin (Kan) resistance, SAIL insertion lines harbor the BAR gene, conferring Basta resistance. This permitted selection for double mutants using a crossing scheme similar to one that was used previously to find intragenic recombination between point mutations in the Arabidopsis CSR1 gene (Figure 3; Mourad et al., 1994). The homozygous tgg1-3 mutant was chosen as pollen donor for this selection, because the tgg1-1 (SALK_130474, Ueda et al., 2006) and tgg1-2 mutants showed silencing of the NPTII gene.

Figure 3.

 Illustration of the crossing scheme used to generate homozygous tgg1tgg2 double mutants by selecting for recombination between the tgg1 and tgg2 mutations, which are 5800 bp apart.
Boxes with an inverted triangle (bsl00072) indicate a T-DNA insertion in the respective gene. BastaR, Basta resistance; KanR, kanamycin resistance

Crossing homozygous tgg1-3 and tgg2-1 mutants to one another resulted in F1 progeny that contained both mutations in repulsion phase, i.e. the mutations are on each of the two homologous chromosomes (Figure 3, step 1). Assuming that 1 cM genetic distance in the Arabidopsis genome corresponds to a physical distance of 250 000 bp (Lukowitz et al., 2000), approximately one seed in 4300 was predicted to be generated by a pollen grain with meiotic recombination between T-DNA insertions separated by 5800 bp, but half of these recombinants would carry neither resistance gene. In order to identify the rare recombination events that bring tgg1-3 and tgg2-1 into coupling phase, 50 000 outcross progeny, generated by crossing the F1 plants from the first cross to the male-sterile ap3-6 mutant (Yi and Jack, 1998), were screened for resistance to both Kan and Basta (step 2). This identified 25 lines that are potentially heterozygous for both mutations. Progeny from eight of the 25 lines were analyzed by PCR for homozygosity of both tgg1-3 and tgg2-1 (step 3). F2 progeny that were homozygous tgg1-3 tgg2-1 and either AP3/ap3 or AP3/AP3 were found for three of the eight lines. These plants showed neither TGG1 nor TGG2 gene expression (Figure 2) and completely lacked myrosinase activity in leaves (Figure 4), verifying the successful isolation of double mutants. Despite the complete gene knockouts, visual inspection of single and double myrosinase mutants did not identify discernable growth or morphological differences compared with wild-type (data not shown).

Figure 4.

 Myrosinase activity during development and in various tissues.
(a) Myrosinase activity in rosette leaves of four developmental stages.
(b) Myrosinase activity in tissues from 28-day-old plants.
(c) Myrosinase activity in roots harvested from 14-day-old seedlings of tgg1-3, tgg2-1 and tgg1-3tgg2-1 double mutants grown on MS plates. Values represent means ± SE of eight (wild-type; a, b), 16 (mutants; a, b), or four (c) independently grown samples. Significant differences in comparison to the wild-type: *P < 0.05, xP < 0.001, Student's t-test.

We identified homozygous progeny from only three of eight lines tested in step 3. This may be due to PCR genotyping inefficiencies, and no attempt was made to reconfirm F2 lines that were not unequivocally the desired homozygous segregants. However, false positives in step 2 of the selection could also be generated by pollen carrying two copies of chromosome 5. Such aneuploid progeny would not result in the isolation of homozygous tgg1-3 tgg2-1 double mutants in step 3. If this is the case, aberrant pollen arose with a frequency of one in 3000 during our selection. Similar frequencies of meiotic chromosomal non-disjunction have been reported previously (Bretagnolle and Thompson, 1995; Henry et al., 2005; Ramsey and Schemske, 1998).

In vitro myrosinase activity of homozygous tgg1, tgg2 and tgg1tgg2 mutants

The availability of both single and double myrosinase mutants allowed us to determine the physiological contributions of TGG1 and TGG2 to total myrosinase activity. Soluble activity was measured in leaves at four developmental stages, in various above-ground tissues of 4-week-old plants, and in roots of 14-day-old seedlings (Figure 4). Myrosinase activity varies during development, with activity increasing from 14-day-old seedlings to mature plants (21- and 28-day-old plants) and decreasing in leaves undergoing senescence (35-day old plants; Figure 4a). The highest activity was detected in flowers, followed by rosette and cauline leaves. The flower stalk, immature siliques (Figure 4b) and roots (Figure 4c) exhibit comparatively low activity in wild-type plants. In tgg1 mutants, there is only about 5% of wild-type myrosinase activity in above-ground tissue, showing that TGG1 contributes the majority of the detectable activity. Interestingly, myrosinase activity in tgg2 mutants is generally elevated by 30% in leaves (nWT = 31, ntgg2 = 16, P < 0.01, Student's t-test) and 40% in flowers (nWT = 8, ntgg2 = 18, P < 0.001, Student's t-test), which correlates with the higher TGG1 transcript abundance in tgg2 mutants (Figure 2). The only myrosinase activity in the above-ground tissue of tgg1-3tgg2-1 double mutants was in flowers, with about 1% of wild-type activity remaining (Figure 4b). In contrast to the above-ground tissues, the already very low myrosinase activity in roots was only slightly decreased in myrosinase mutants (Figure 4c).

Our assays, which measure only soluble myrosinase, would not detect membrane-bound or otherwise insoluble enzyme activity (Matsushima et al., 2003). Therefore, myrosinase assays were performed using fractionated leaf extracts: (i) unpurified leaf extracts, i.e. ground leaf tissue in extraction buffer, (ii) the leaf extract pellet, and (iii) the supernatant, with and without Sephadex purification. Activity in the pellet, the fraction that would contain insoluble proteins, was very low in wild-type, tgg1 and tgg2 mutants, and most likely derived from residual soluble protein in the pellet fraction (data not shown). Myrosinase activity in the unpurified total leaf extract and the supernatant fractions was not significantly different from that in Figure 4(a), indicating that activity was not lost in the Sephadex columns. Hence, we conclude that there is little or no insoluble myrosinase in Arabidopsis leaves.

Damage-induced glucosinolate breakdown in myrosinase mutants

To complement the in vitro measurements (Figure 4), myrosinase activity in crushed leaves of wild-type and tgg mutant plants was determined. Figure 5 depicts the time-course of total aliphatic glucosinolate and total indole glucosinolate breakdown in crushed leaves over 5 min. In wild-type and the tgg2-1 mutant, most aliphatic (Figure 5a) and indole glucosinolates (Figure 5b) are degraded after 15 sec. At t = 15 sec, significantly more glucosinolates remained in the tgg1-3 mutant than in wild-type (n = 4, P < 0.01, Student's t-test), but breakdown is complete after 1 min. These results suggest that the 5% remaining myrosinase activity in tgg1-3 mutant (Figure 4) may not be limiting for glucosinolate degradation in crushed tissue. In contrast, in the tgg1-3tgg2-1 mutant, the amount of intact aliphatic glucosinolates at t = 5 min was not significantly different from the amount in this mutant at t = 0 (n = 4, P > 0.05, Student's t-test, Figure 5a). Surprisingly, breakdown of indole glucosinolates still proceeds, although at a greatly reduced rate with a half-life of about 5 min (Figure 5b). Degradation kinetics of individual glucosinolates are similar to the grouped aliphatic and indole glucosinolates (Figure S1), indicating that the two Arabidopsis myrosinases do not exhibit a high degree of glucosinolate specificity.

Figure 5.

 Time course of glucosionlate degradation in wild-type, tgg1-3, tgg2-1 and tgg1-3tgg2-1 double mutants.
(a) Total content of aliphatic glucosinolates (GS) per gram fresh weight (FW).
(b) Total content of indole glucosinolates. Leaves of 21-day-old plants were crushed in water to allow breakdown of glucosinolates by myrosinase. Myrosinase activity was stopped by heat inactivation at the indicated time points and the remaining glucosinolates were extracted. Values represent means ± SE of four independently grown samples. SE is not shown where it is smaller than the symbols.

Glucosinolate content of tgg mutants during development

Glucosinolate content varies greatly during Arabidopsis development. Seed-specific glucosinolates are broken down during the early stages of seedling growth (Brown et al., 2003). In older plants, glucosinolate levels show a peak at the onset of flowering and decrease due to transport or breakdown in leaves undergoing senescence (Brown et al., 2003; Kim et al., 2004; Petersen et al., 2002). In order to assess the previously unknown contribution of the two myrosinases to developmental changes in glucosinolate content, we measured leaf glucosinolate levels at key stages during plant growth. Glucosinolate content was determined in dry seeds, 5-, 7- and 10-day-old seedlings (leaves only) and in 14-, 21-, 28- and 35-day-old rosette leaves. For the latter, tissue was harvested from the first and the second rosette leaves (first foliage pair) and the third and fourth rosette leaves (second foliage pair).

The total content of aliphatic glucosinolates in dry seeds is similar in wild-type and tgg mutants (Figure 6a). However, double mutants contain approximately 30% less benzoyloxy glucosinolates (n = 4, P < 0.001, Student's t-test) than wild-type (Figure 6b). Furthermore, the content of almost all individual seed glucosinolates is different in tgg1-3 tgg2-1 double mutants in comparison to the wild-type (Figure S2). Despite these differences, above-ground glucosinolate content decreased similarly in the four genotypes during the early stages of seedling development, showing that the TGG1 and TGG2 myrosinases are not involved in this process (Figure 6a; Figure S2).

Figure 6.

 Alterations of the glucosinolate content during seedling development and in rosette leaves of four developmental stages of wild-type, tgg1-3, tgg2-1 and tgg1-3tgg2-1 double mutants.
(a) Total content of aliphatic glucosinolates (GS) per gram dry weight (DW) in seeds (time 0) and in seedlings germinated on MS plates.
(b) Total content of benzoyloxy glucosinolates in seeds and seedlings.
(c) Total content of aliphatic glucosinolates of the second foliage pair.
(d) Total content of indole glucosinolates of the second foliage pair. Means ± SE of four or eight (tgg1-3tgg2-1) independently grown samples are shown. SE is not shown where it is smaller than the symbols.

Glucosinolate levels of the first foliage pair are marginally, but not significantly elevated in tgg mutants (Figure S3). However, in 14-day-old leaves of the second foliage pair, the content of aliphatic and indole glucosinolates in the tgg1-3tgg2-1 mutant is threefold higher than in wild-type (nWT = 4, ntgg1,tgg2 = 8, P < 0.001, Student's t-test; Figure 6c,d). Developing leaves of tgg1-3 and tgg2-1 single mutants also appear to have slightly elevated glucosinolate levels in comparison to the wild-type but this effect is significant only for tgg2-1 in Figure 6(d) (n = 4, P < 0.01, Student's t-test). Significantly higher glucosinolate levels in the second leaf pair of myrosinase mutants at 2 weeks of age were observed in a repeat of this experiment with independently grown plants (data not shown). Moreover, a higher glucosinolate level in undamaged leaves of myrosinase mutants is also seen in the third leaf pair of 3-week-old plants in Figure 5, at the t = 0 time point. All three mutant lines have slightly elevated glucosinolate levels, but the effect is significant only for aliphatic glucosinolates in the tgg1-3tgg2-1 mutant (Figure 5a; n = 4, P < 0.05, Student's t-test) and indole glucosinolates in the tgg1-3 mutant (Figure 5b; n = 4, P < 0.01, Student's t-test).

The developmental glucosinolate profile in wild-type leaves (Figure 6), i.e. highest levels in maturing leaves followed by a senescence-associated decrease, is similar to what has been reported previously (Brown et al., 2003; Petersen et al., 2002). As these leaves of the double mutant senesce, glucosinolate content decreases rapidly, reaching levels similar to the wild-type and single tgg mutants (Figure 6c,d; Figure S4). Hence, the senescence-associated leaf glucosinolate decrease is independent of TGG1 and TGG2.

Responses of specialist and generalist herbivores to tgg mutations

In the absence of mutants, it has been difficult to study the role of myrosinase-mediated glucosinolate breakdown in herbivore defense. To take advantage of the unique opportunity presented by a cruciferous plant without foliar myrosinases, we performed feeding experiments with two crucifer-specialist Lepidoptera (Pi. rapae and Pl. xylostella), a Brassicaceae-feeding generalist lepidopteran (T. ni), a facultative Solanaceae specialist lepidopteran (Manduca sexta, tobacco hornworm), a crucifer-specialist homopteran (B. brassicae) and a generalist homopteran (My. persicae). Although only single experiments are shown in Figure 7, all insect experiments were run at least twice with similar results.

Figure 7.

 Growth of insect herbivores on tgg1, tgg2 and tgg1tgg2 double mutants in no-choice feeding experiments.
Wild-type and mutant plants were planted in one pot to allow pair-wise comparisons.
(a–d) Weight gain (dry weight) of newly hatched larvae. The inset in (c) shows leaf consumption by T. ni during the feeding period.
(e, f) Progeny produced by a single first-instar My. persicae or B. brassicae aphid after 10 days. Values represent means ± SE. *P < 0.05, ***P < 0.001, paired Student's t-test.

The two crucifer-specialist Lepidoptera showed differing responses to Arabidopsis myrosinase mutations. Growth of Pl. xylostella larvae was not significantly different on wild-type and tgg mutant plants (Figure 7a). Unknown environmental factors caused overall Pl. xylostella growth in the wild-type versus tgg1-3tgg2-1 mutant comparison to be lower than in the single myrosinase mutant experiments, which were run at a different time. In contrast to Pl. xylostella, Pi. rapae showed a 40% reduction in weight gain on double, but not single mutant plants (Figure 7b). Similarly, in leaf plug choice tests, 110 out of 186 Pi. rapae larvae showed a preference for wild-type over tgg1-3 tgg2-1 mutant leaves (P < 0.05, χ2-test).

The two Lepidoptera that are not crucifer specialists showed improved growth in the absence of myrosinase. T. ni larvae were 70% larger on tgg1-3tgg2-1 double mutants than on wild-type, which correlated with the more pronounced leaf damage (Figure 7c inset). Somewhat surprisingly, larvae feeding on tgg1 and tgg2 single mutants gained slightly less weight than on wild-type, but no difference in leaf damage was observed (data not shown). Single myrosinase mutations had little effect on Ma. sexta weight gain, with only a slight, but significant, increase on tgg1 mutants (Figure 7d). However, Ma. sexta larvae gained almost nine times as much weight on tgg1-3tgg2-1 mutants as on wild-type (Figure 7d). Observation of Ma. sexta feeding showed that they consumed only senescing leaves of wild-type, but ate the entire tgg1-3tgg2-1 mutant plants. Unlike on wild-type and single mutants, larvae were able to reach pupation on tgg1-3tgg2-1 mutants. Choice tests showed similar results to the weight gain experiments: 33 of 37 Ma. sexta larvae preferred tgg1-3tgg2-1 mutant leaf plugs over wild-type (P < 0.001, χ2-test), consuming on average threefold greater mutant leaf area than wild-type leaf area (P < 0.05, Student's t-test).

Both aphid species tested, the generalist My. persicae and the specialist B. brassicae, reproduced similarly on wild-type and tgg mutant plants (Figure 7e,f). One possible explanation is that these aphids are able to avoid the generation of toxic glucosinolate breakdown products. The paths of aphid stylets penetrating plant epidermis and parenchyma tissue to reach phloem sieves are mainly intercellular (Miles, 1999; Powell et al., 2006), and may avoid damage of the myrosinase and glucosinolate-containing cells around the phloem. Furthermore, My. persicae secrete substantial amounts of glucosinolates in their honeydew (Kim and Jander, unpublished data) and B. brassicae sequester intact glucosinolates in their bodies (Bridges et al., 2002; Pontoppidan et al., 2001). Given the very rapid myrosinase-mediated glucosinolate breakdown in wild-type plants (Figure 5), this would indicate that there is little if any contact of the enzyme with its substrate during aphid feeding.


Together, TGG1 and TGG2 provide most or all of the myrosinase activity in Arabidopsis above-ground tissues. Total myrosinase activity varies widely among plant tissues and in the course of leaf development (Figure 4b,c), with the highest activity in mature rosette leaves and in flowers, which correlates with high glucosinolate levels in these tissues (Brown et al., 2003). Neither TGG1 nor TGG2 transcripts were detectable in roots, but in vitro enzyme activity measurements revealed very low, but comparable myrosinase activity in root extracts from wild-type and myrosinase mutants (Figure 4c). This activity may derive from TGG4 and/or TGG5, which are specifically expressed in roots (Toufighi et al., 2005; Zimmermann et al., 2004). Similarly, the very low level of myrosinase activity remaining in tgg1-3tgg2-1 flowers may result from the expression of TGG6 in stamens (Figure 4b; Toufighi et al., 2005; Zimmermann et al., 2004).

In all above-ground tissues that were tested, TGG1 provided the vast majority of myrosinase activity (Figure 4). Recent analysis of whole 22-day-old tgg1-1 plants also indicated that TGG1 accounts for most of the myrosinase activity in above-ground tissue (Ueda et al., 2006), although developmental and tissue-specific regulation was not considered in this study. TGG1 may provide more myrosinase activity than TGG2 due to its additional expression in stomatal guard cells (Figure 1). This expression in guard cells may provide a first line of defense against fungal and bacterial pathogens that enter plant tissue through stomata (Howard and Valent, 1996; Plotnikova et al., 2000). Different roles of the two myrosinases in pathogen defense are indicated by the fact that some isolates of Botrytis cinerea show improved growth on tgg1-2 but not tgg2-1 mutants (D. Kliebenstein, University of California, Davis, CA, USA, personal communication).

To confirm that TGG1 in fact contributes the majority of myrosinase activity and that TGG2 activity is not membrane-bound (Matsushima et al., 2003), in vitro activity was measured in individual fractions of leaf extracts. Most of the activity was present in the soluble fraction, showing that both TGG1 and TGG2 are soluble proteins. Ueda et al. (2006), using antibodies to detect the TGG2 protein, showed that only 50% of TGG2 is soluble, perhaps an indication that TGG2 localizes or aggregates differently in the absence of TGG1 in our experiments. Complex interactions between TGG1 and TGG2 are also indicated by the fact that TGG1 transcript levels are elevated when TGG2 is mutated (Figure 2), resulting in greater than wild-type myrosinase activity in tgg2 mutants (Figure 4).

Despite the distinct expression patterns and the different in vitro myrosinase activities of TGG1 and TGG2, the two myrosinases display comparable activity in crushed leaves (Figure 5). Although TGG2 contributes only approximately 5% to total in vitro myrosinase activity (Figure 4), this small portion is sufficient to break down glucosinolates at almost the same rate as wild-type. This suggests that TGG1 and TGG2 are redundant in crushed leaves and may also be redundant in defense against leaf-macerating herbivores. Enzymatic redundancy is further indicated by the fact that TGG1 and TGG2 do not show specificity for particular glucosinolates in crushed tissue (Figure S1).

Surprisingly, tgg1-3tgg2-1 double mutants, which lack break down of aliphatic glucosinolates and show no myrosinase activity in vitro, exhibited slow breakdown of indole glucosinolates in disrupted leaf material (Figure 5b; Figure S1). Myrosinase-independent breakdown of indole glucosinolates may occur in the crushed leaves. Chemical breakdown of glucosinolates by acid hydrolysis in plant extracts has been reported (Tiedink et al., 1991). Non-enzymatic breakdown by heat (Chevolleau et al., 1997), UV light or sunlight (Michajlovskij, 1968) is also possible. However, glucosinolate breakdown under these conditions occurs at a much slower rate and results in smaller amounts of breakdown products (Chevolleau et al., 1997). Another possibility is that an additional, perhaps indole-specific, β-thioglucoside glucohydrolase or some other enzyme is causing the observed breakdown.

Seed glucosinolate levels in the tgg1-3tgg2-1 mutant are significantly different from wild-type, but reach wild-type levels during early seedling growth (Figure 6b; Figure S2). Perhaps there is a ‘target’ glucosinolate profile in seedlings that the plants approach, independent of seed glucosinolate levels and seedling myrosinase activity. Other catabolic mechanisms, perhaps an as yet unknown plant sulfatase converting glucosinolates to inactive desulfoglucosinolates, must be responsible for the developmental decline. Similarly, glucosinolates in maturing leaves of myrosinase mutants are elevated, but drop to wild-type levels as leaves senesce (Figure 6c,d; Figures S2 and S3). If TGG1 and TGG2 played a major role in this non-defensive glucosinolate decrease during development, we would predict that the glucosinolate concentration would not change in senescing tgg1tgg2 mutant leaves, or at least would be always higher than in the wild-type. Therefore, TGG1 and TGG2 are unlikely to be involved in this process.

There are several possible explanations for the elevated glucosinolates observed in expanding leaves of myrosinase mutants (Figures 5a,b and 6c,d). Firstly, if myrosinase catalyzes the formation of isothiocyanates that have been detected in surface washes (M. Haribal and J. Renwick, Boyce Thompson Institute for Plant Research, Ithaca, NY, USA, personal communication) and headspace samples (Finch, 1978; Tollsten and Bergström, 1988) of undamaged crucifers, myrosinase mutants might have higher glucosinolate levels. Secondly, glucosinolate levels are increased by treatment with defense-related signaling molecules (Mikkelsen et al., 2003), and tgg1tgg2 double mutants may have been more afflicted by pathogens or herbivores in our growth chambers than single mutants and wild-type plants. Finally, glucosinolate production at some growth stages may be upregulated as a direct response to reduced myrosinase, similar to how TGG1 transcription is upregulated in response to a tgg2 mutation (Figure 2). Further experiments will need to be performed to determine which of these hypotheses is correct.

TGG1 and TGG2 myrosinase activity has a deterrent effect on T. ni and Ma. sexta, two Lepidoptera that are not crucifer-feeding specialists (Figure 7). Interestingly, the absence of either TGG1 or TGG2 alone slightly decreased T. ni weight gain (Figure 7a). One possible explanation is that the elevated glucosinolate levels in expanding leaves of tgg1 and tgg2 mutants (Figures 5a,b and 6c,d), combined with non-limiting myrosinase activity (Figure 5), produce more deterrent glucosinolate breakdown products than in the wild-type. In contrast, greatly reduced glucosinolate breakdown makes double mutants more susceptible to T. ni (Figures 4 and 7c). Weight gain experiments (Figure 7d) and leaf plug choice tests with Ma. sexta show that glucosinolate breakdown products are the major feeding deterrent for this insect on Arabidopsis.

The two crucifer-specialist lepidopteran larvae, Pl. xylostella and Pi. rapae, differ in their response to myrosinase mutations. Pl. xylostella growth was unaffected by single or double myrosinase mutations (Figure 7a), perhaps because the glucosinolate-inactivating sulfatase of Pl. xylostella (Ratzka et al., 2002) is so efficient that the presence or absence of myrosinase is irrelevant for this herbivore. In contrast, Pi. rapae shows greater weight gain on wild-type plants than on ones that are devoid of myrosinase (Figure 7b), and also preferred wild-type leaf plugs in choice tests. High levels of isothiocyanates are repellent to Pi. rapae in vitro (Agrawal and Kurashige, 2003), and it might be expected that Pi. rapae would perform better on tgg1tgg2 double mutants, because fewer toxic glucosinolate breakdown products are produced in these plants. However, some specialist herbivores, including Pi. rapae, use glucosinolates or their breakdown products as host recognition cues (Renwick, 2002). We suggest that the absence of feeding stimulants reduces Pi. rapae weight gain on the tgg1tgg2 mutant, but we cannot rule out the possibility that the double mutant may produce another Pi. rapae deterrent to compensate for the loss of glucosinolate breakdown products.

As the fecundity of My. persicae and B. brassicae is unaffected by the presence or absence of myrosinase (Figure 7e,f), other Arabidopsis defense mechanisms must be more important for combating aphids. Mutants affected in signaling pathways, including the jasmonic acid-insensitive mutant coi1 and the salicylic acid-insensitive mutant npr1, are more susceptible to My. persicae (Ellis et al., 2002; Moran and Thompson, 2001). Furthermore, expression of both salicylic acid- and jasmonic acid-dependent defense genes is altered in response to aphid feeding (Moran et al., 2002; de Vos et al., 2005). Additional experiments are necessary to elucidate how aphid feeding is perceived by plants, how defensive signals are transmitted, and what the proximal causes of aphid resistance and sensitivity are.

Our analysis of tgg1, tgg2 and tgg1tgg2 Arabidopsis mutants has allowed us to answer several long-standing questions regarding the glucosinolate–myrosinase defense system of Brassicaceae. The availability of myrosinase single and double mutant lines will permit future research on myrosinase-independent glucosinolate breakdown, interactions of myrosinases with other plant proteins, and the role of myrosinase in plant defense. Furthermore, our newly developed method for generating the tgg1tgg2 double mutants will have broad applicability for generating double mutants of the many other tandem-duplicated genes in Arabidopsis.

Experimental procedures

Plant material and growth conditions

Seeds of wild-type Arabidopsis (ecotype Col-0) are from the Arabidopsis Biological Resource Center (ABRC, Lines SALK_130474 (tgg1-1), SALK_069615 (tgg1-2) and SALK_038730 (tgg2-1) are from the SIGnAL collection (Alonso et al., 2003; Lines SAIL_786_B08 (tgg1-3) and SAIL_237_G11 (tgg2-2) are from the Syngenta Arabidopsis Insertion Library (SAIL; Sessions et al., 2002; Col-0 AP3/ap3-6 seeds were kindly supplied by T. Jack (Dartmouth College, Hanover, NH, USA). Homozygous SALK_130474 (CS6566), SALK_069615 (CS6565) and SALK_038730 (CS6567) insertion lines and segregating Col-0 AP3/ap3-6 seeds (CS6568) have been submitted to the ABRC. Lines containing SAIL T-DNA insertions will be made available by the authors, but will require obtaining a material transfer agreement from Syngenta Biotechnology, Inc. (Research Triangle Park, NC, USA,

For T-DNA confirmation experiments, plants were sown on Metromix 200 (Scotts, Marysville, OH, USA) without additional fertilizer. Later experiments were performed with plants grown on Cornell Mix (Landry et al., 1995) with Osmocoat fertilizer (Scotts, Marysville, OH, USA). Plants were grown in Conviron (Winnipeg, Canada) growth chambers in 20 × 40 cm nursery flats at a photosynthetic photon flux density of 200 μmol m−2 sec−1 and a 16 h photoperiod. The temperature in the chambers was 23°C and the relative humidity was 50%. For growth on agar, 1× Murashige and Skoog (MS) medium without sugar was used (Weigel and Glazebrook, 2002).

Construction of transgenic plants with TGG1 and TGG2 promoter:GUS fusions

TGG1 and TGG2 promoter fragments, 1800 and 1500 bp of DNA for TGG1 and TGG2, respectively, were generated by PCR amplification with the primers in Table S1 and Taq DNA polymerase (Perkin Elmer, Norwalk, CT, USA). Amplified products were digested with HindIII and NcoI (New England Biolabs, Beverly, MA, USA), cloned into the binary vector pCAMBIA3301 ( and transformed into Agrobacterium tumefaciens strain GV3101 (Koncz and Schell, 1986). Col-0 wild-type plants were transformed by vacuum infiltration (Bechtold et al., 1993). T0 transformants were selected by spraying seedlings at the four-leaf stage with 0.1 mg ml−1 glufosinate-ammonium (Basta; Chem Service, West Chester, PA, USA) and 0.005% Silwet (Lehle Seeds, Round Rock, TX, USA).

GUS staining

Tissue for histochemical GUS staining (Weigel and Glazebrook, 2002) was collected from roots, leaves, flower stalks, and flower organs of 4-week-old T2 transgenic individuals carrying TGG1 promoter:GUS and TGG2 promoter:GUS constructs. Samples were examined and photographed using an Axioscop microscope (Zeiss, Jena, Germany) equipped with a digital camera (Jenoptik, Jena, Germany).

Identification of homozygous tgg1 and tgg2 T-DNA mutant lines

For genomic DNA isolation, 20–40 mg fresh leaf tissue from individual plants was harvested on dry ice in 1.4 ml tubes (Matrix Technologies, Hudson, NH, USA) containing three 3 mm steel ball bearings. Tissue was ground in a Harbil model 5G-HD paint shaker (Harbil, Wheeling, IL, USA) for 1 min. DNA was extracted using the cetyltrimethylammonium bromide (CTAB; Sigma, St Louis, MO, USA) protocol (Weigel and Glazebrook, 2002). Three separate PCR reactions were carried out to identify the T-DNA orientation and position within the target gene. Forward and reverse primers were designed to amplify the wild-type genes (SALK_069615F, SALK_069615R; SALK_038730F, SALK_038730R; SAIL_786_B08F, SAIL_786_B08R; SAIL_237_G11F, SAIL_237_G11R; Table S1). These primers were used in combination with T-DNA left border primers (LBa1 for SALK lines or LB3 for SAIL lines; Table S1) to verify the presence and orientation of the T-DNAs. The orientation and position of the T-DNA insertions was confirmed by DNA sequencing.

Extraction of RNA and gene expression analysis by RT-PCR

Total RNA from 0.1 g leaves or roots was isolated using TRI Reagent (Molecular Research Center, Cincinnati, OH, USA) following the manufacturer's instructions. TGG1 and TGG2 cDNA fragments were generated by RT-PCR using the Access RT-PCR kit (Promega), oligonucleotide primers (TGG1RT-F6, TGG1RT-R6, TGG2RT-F2, TGG2RT-R2; Table S1), and a previously described protocol (Barth et al., 2004) with the modification that 17 amplification cycles were run. Amplification with these primers is in the linear range after 17 cycles, as determined by running increasing cycle numbers and analyzing the amount of cDNA fragments (loaded as 1:3 dilutions in distilled water) on 1% agarose gels containing ethidium bromide. Band intensities were quantified with ImageQuant 5.0 (GE Healthcare Biosciences, Piscataway, NJ, USA) and the linear range of amplification was determined using Microsoft Excel. A ubiquitin UBQ10 (At4g05320, primers in Table S1) cDNA served as an internal control.

Generation and identification of tgg1tgg2 double mutants

Figure 3 illustrates the crossing scheme used to identify genetic recombination between the tgg1-3 and tgg2-1 mutations. Progeny from the cross in step 1 were selected on MS plates containing both 40 μm glufosinate-ammonium (Chem Service) and 25 μg ml−1 kanamycin (Sigma). Double-resistant F1 plants were transplanted to soil and crossed to Col-0 ap3-6 (Yi and Jack, 1998; step 2). F1 progeny from this second cross were plated on MS agar with Kan and Basta, double-resistant plants were transferred to soil, and F2 progeny seeds were plated on MS with Kan and Basta (step 3). Double-resistant F2 plants were transplanted to soil and tissue was collected for genomic DNA extraction as described above. Per plant, four PCR reactions were carried out to identify homozygous tgg1-3tgg2-1 double mutants using the primer combinations SAIL_786_B08F + SAIL_786_B08R and SAIL_786_B08F + LB3 (for the tgg1-3 mutation) and SALK_038730F + SALK_038730R, SALK_038730R + LBa1 (for the tgg2-1 mutation). The absence of both TGG1 and TGG2 transcripts was confirmed by RT-PCR, as described above.

Total myrosinase activity assay

Total myrosinase activity was measured using a modification of the method described by Palmieri et al. (1982). Except for flower stalks and siliques, tissue (20–40 mg fresh weight) was collected on dry ice in 1.4 ml tubes (Matrix) containing three 3 mm steel ball bearings. Frozen tissue was crushed in a paint shaker (Harbil) for 1 min. Siliques and flower stalks were collected in 1.5 ml Kontes microtubes (Fisher Scientific, Hampton, NH, USA) and ground with an electric drill using Kontes pellet pestles (Fisher Scientific). Cold 33 mm sodium phosphate buffer, pH 7.0 (0.3 ml) was added and tubes were shaken again. Plates and tubes were centrifuged at 2870 g and 10 000 g, respectively, at 4°C, and 200 μl of the supernatants were added to Sephadex G-50 (Sigma) columns. Columns were prepared by adding 45 μl (dry volume) resin to a deep-block 96-well plate with a 0.45 μm pore size filter membrane (Whatman, Maidstone, UK). The resin was swelled in 1.5 ml 33 mm sodium phosphate buffer, pH 7.0, overnight at 4°C. Buffer was passed through the membrane microtiter plate using a vacuum manifold (Qiagen, Valencia, CA, USA) and centrifugation at 720 g for 1 min at 4°C. After adding the sample supernatants, centrifugation was repeated and the filtrate was collected in a 96-well plate. The final desalted and purified extract was brought to a volume of 200 μl by adding 100 μl water to each well and centrifuging again. For activity measurements, extracts were pipetted into 96-well UV-permeable plates (Corning Inc., Corning, NY, USA) containing 0.34 mm sinigrin and 0.3 mm ascorbic acid in a total volume of 200 μl extraction buffer. Degradation of the glucosinolate sinigrin (Sigma) was measured spectrophotometrically at 227 nm at 25°C for 15 min, collecting a data point every 10 sec using a Spectramax Plus 384 plate reader (Molecular Devices, Sunnyvale, CA, USA). Myrosinase activity was based on gram fresh weight and calculated according to Beer's law using the extinction coefficient ɛ227 = 7273 m−1 cm−1 for sinigrin. The extinction coefficient, calculated from a sinigrin standard curve, is similar to previous reports (James and Rossiter, 1991; Palmieri et al., 1982). To determine whether myrosinase activity also derives from membrane-bound protein, unpurified total extracts (i.e. not passed through Sephadex columns), supernatants, and pellets resuspended in buffer were used for activity measurements as described above.

Glucosinolate analysis

For analysis of germinating seedlings, entire seedlings, minus the roots, were harvested from agar after 0 (ungerminated seeds), 5, 7 and 10 days. For each time point, seedlings from an entire plate were pooled to provide enough tissue. For analysis of leaf glucosinolates in the course of development, all seeds were planted at the same time and plants were grown together in the same growth chamber. Leaf material was harvested after 2, 3, 4 and 5 weeks. Due to the small size at 2 and 3 weeks, leaves from four individual plants were pooled. Seedling and leaf samples were dried overnight in a Savant SC 110 rotary evaporator (Savant Instruments, Farmingdale, NY, USA). Dry weight was recorded and plant samples were crushed in 1.4 ml tubes (Matrix) containing three 3 mm steel ball bearings in a Harbil model 5G-HD paint shaker; 600 μl of 60% methanol and 10 μl of 1.25 mm sinigrin were then added to each sample. Samples were shaken again and heated at 75°C for 10 min. After centrifugation at 10 000 g for 10 min, 480 μl of the supernatant were added to Sephadex A-25 (Amersham) columns. To elute desulfoglucosinolates, 20 μl arylsulfatase (0.77 mg ml−1; Sigma) and 80 μl HPLC-grade water (Fisher Scientific) were added to the columns and incubated overnight in the dark at room temperature. Finally, 200 μl 60% methanol and 200 μl HPLC water were added to bring up the volume to 500 μl. Samples were lyophilized for approximately 2 h and dissolved in 100 μl HPLC water. Desulfoglucosinolates were detected by HPLC (Kim et al., 2004) and quantified using response factors reported previously (Brown et al., 2003).

Glucosinolate breakdown assay

The formation of glucosinolate breakdown products was assessed at five time points (0, 0.25, 1, 3 and 5 min). Leaves (70–100 mg fresh weight) of the third foliage pair of 3-week-old plants were harvested from individual plants and the fresh weight was recorded. Tissue was either frozen immediately in liquid nitrogen (0 min) or crushed in 400 μl HPLC-grade water (Fisher Scientific) for 15 sec using 1.5 ml Kontes tubes and pestles (Fisher Scientific). Immediately after grinding (0.25 min time point), or after 1, 3 or 5 min incubation at room temperature, 60 μl of 100% methanol (HPLC grade, Fisher Scientific) were added. The samples were vortexed briefly and immediately heated at 75°C for 15 min in a water bath to inactivate myrosinase and to extract glucosinolates. After heating, 10 μl of 1.25 mm sinigrin were added to each sample as an internal standard. For controls (0 min time point), leaf tissue was ground in a frozen state and 1 ml 60% methanol was added prior to heating at 75°C. HPLC analysis of desulphoglucosinolates was performed as described above.

Insect no-choice test experiments

The sources of insects for experiments were: T. ni eggs, Entopath Inc. (Easton, PA, USA); Ma. sexta eggs, kindly provided by M. del Campo (Cornell University, Ithaca, NY, USA) and C. Miles (Binghampton University, Binghampton, NY, USA); Pi. rapae eggs, kindly supplied by A. Agrawal (Cornell University, Ithaca, NY, USA) or purchased from Carolina Biological Supply (Burlington, NC, USA); Pl. xylostella eggs, Benzon Research (Carlisle, PA, USA); B. brassicae, kindly supplied by T. Shelton (New York State Agricultural Experiment Station, Geneva, NY, USA); My. persicae, colony maintained by the Boyce Thompson Institute greenhouse staff. For all experiments, paired mutant and wild-type plants were grown at the same time in the same pot, from seeds produced by plants grown previously in the same growth chamber. With the exception of T. ni, experiments with double mutants were run at a later date than experiments with single mutants. Insects were confined on the leaves of 2-week-old (Pl. xylostella, T. ni, My. persicae and B. brassicae) or 3-week-old (Pi. rapae and Ma. sexta) plants with mesh-covered cups. Neonate lepidopteran larvae were allowed to feed on plants for 6 days (Pl. xylostella and T. ni) or 9 days (Pi. rapae and Ma. sexta) before being harvested and lyophilized for 1 day or dried at 80°C for 3 days. Larval dry weight was determined using a precision balance. My. persicae and B. brassicae for experiments were reared on cabbage (var. Wisconsin Golden Acre; Seedway, Hall, NY, USA). One first instar aphid was placed in the middle of each Arabidopsis rosette. After 10 days, the number of progeny on each plant was counted.

Insect choice test experiments

Leaf plugs, 8 mm in diameter, were made from similarly sized leaves of 3-week-old paired Col-0 wild-type and Col-0 tgg1-3tgg2-1 plants growing together in the same pot. Mutant and wild-type leaf plugs were placed about 5 mm apart on moist paper towels in 100 mm diameter Petri plates. Single 4-day-old Pi. rapae or Ma. sexta larvae raised on artificial diet (Bell and Joachim, 1976; Webb and Shelton, 1988) were placed lengthwise between each pair of leaf plugs. After 6 h (Pi. rapae) or 24 h (Ma. sexta), the remaining leaf material was collected and scanned. The area of scanned leaf plugs was calculated with ImageJ (Abramoff et al., 2004), and, to account for leaf shrinkage during the experiment, the amount of leaf area consumed was calculated by comparison with the area of control leaf plugs.


The authors would like to thank Alyssa Whu, Teresa Rojas and Nedjie Exantus for assisting with insect experiments and Tom Jack for providing ap3-6 seeds. Anurag Agrawal, Marta del Campo, Carol Miles and Tony Shelton kindly provided insect cultures. We also thank Anurag Agrawal, Marta del Campo and Alan Renwick for useful comments on the manuscript. This research was supported by the Atlantic Philanthropies and the Boyce Thompson Institute.