Brassinosteroids interact negatively with jasmonates in the formation of anti-herbivory traits in tomato

Abstract

Given the susceptibility of tomato plants to pests, the aim of the present study was to understand how hormones are involved in the formation of tomato natural defences against insect herbivory. Tomato hormone mutants, previously introgressed into the same genetic background of reference, were screened for alterations in trichome densities and allelochemical content. Ethylene, gibberellin, and auxin mutants indirectly showed alteration in trichome density, through effects on epidermal cell area. However, brassinosteroids (BRs) and jasmonates (JAs) directly affected trichome density and allelochemical content, and in an opposite fashion. The BR-deficient mutant dpy showed enhanced pubescence, zingiberene biosynthesis, and proteinase inhibitor expression; the opposite was observed for the JA-insensitive jai1-1 mutant. The dpy×jai1-1 double mutant showed that jai1-1 is epistatic to dpy, indicating that BR acts upstream of the JA signalling pathway. Herbivory tests with the poliphagous insect Spodoptera frugiperda and the tomato pest Tuta absoluta clearly confirmed the importance of the JA–BR interaction in defence against herbivory. The study underscores the importance of hormonal interactions on relevant agricultural traits and raises a novel biological mechanism in tomato that may differ from the BR and JA interaction already suggested for Arabidopsis.

Herbivory, hormones, Solanum lycopersicum, trichomes, zingiberene

Introduction

Insect herbivory is one of the main causes of agricultural losses. Insects consume significant amounts of plant biomass and also cause indirect losses due to their role as vectors for pathogens (Schröder et al., 2005). It is estimated that 14% of all world agricultural output is lost to insect pests (Oerke et al., 1994). However, plants have developed several effective anti-herbivory traits for self-defence after millions of years of co-evolution with insects (Rausher, 2001). Trichomes, secondary metabolites, toxic amino acids, and systemic resistance, among many other so-called ‘plants’ solutions to plants’ problems’ (Hilder and Boulder, 1999), are being extensively investigated aiming at natural methods of pest control.

The tomato (Solanum lycopersicum L.) is one of the most susceptible crops to insect herbivory, demanding a heavy load of insecticides for pest management in commercial production (Zalom, 2003). However, tomato plants have evolved numerous defence mechanisms that are effective against insect pests, such as glandular trichomes (Simmons and Gurr, 2005), synthesis of allelochemicals (Carter et al., 1989a), and expression of enzymes such as proteinase inhibitors (Howe et al., 1996) and polyphenol oxidase (Thaler, 1999).

Plant hormones play a central role in the regulation of developmental processes and signalling networks involved in plant responses to a wide range of biotic and abiotic stresses (Robert-Seilaniantz et al., 2007), including herbivory (Howe and Jander, 2008). Concerning insect herbivory resistance in tomato plants, a great deal of focus has been given to jasmonates (JAs) (Lin et al., 1987; Glazebrook, 2005; Chen et al., 2006), although increasing evidence shows the relevance of multiple hormones in plant–pest interactions (Robert-Seilaniantz et al., 2007). Examples are ethylene and salicylic acid, which are being unveiled as regulators of many pest defence responses (Lorenzo et al., 2003; von Dahl et al., 2007; Zarate et al., 2007). Yet, little is known about the role of other hormones in plant defence against insect attack and nothing has been reported on hormonal interactions regarding herbivory.

Mutant organisms are useful tools in biological studies to gain insights into functions of genes and organic compounds involved in specific aspects of an organism's life. Plants impaired in hormonal metabolism or signalling are commonly used to study the complex network of factors controlling plant development and its interaction with the environment (Wikinson et al., 1995; Howe et al., 1996; Koornneef et al., 1997; Koka et al., 2000; Li et al., 2004). However, only few works have focused on the role of plant hormones in anti-herbivory traits in tomato (Li et al., 2004). The fact that characterized tomato hormone mutants are available in different genetic backgrounds makes it difficult to study comprehensively the role of various hormones in a given process as well as the relationship between them through double mutant analyses due to the resulting genetic background noise.

In a previous study, a collection of tomato hormonal mutants introgressed into a single genetic background, the Micro-Tom (MT) cultivar, were established (Carvalho, 2008). MT was initially described for ornamental purposes (Scott and Harbaugh, 1989), but its short life cycle and small size have made it a suitable genetic model system (Meissner et al., 1997). For this reason MT is being widely used in plant genetics and physiology studies (Tieman et al., 2001; Isaacson et al., 2002; Serrani et al., 2007), including hormone responses to herbivory (Li et al., 2004). Here, advantage was taken of this MT hormonal collection (Table 1) for the study of the role of various hormone classes in the formation of anti-herbivory traits, such as trichomes and secondary metabolites. It was found that gibberellins, auxins, and ethylene are indirectly involved in altering trichome density, but brassinosteroids (BRs) and jasmonic acid (JA) act directly in the formation of this trait in addition to the formation of secondary metabolites and proteinase inhibitor. Surprisingly, the action of BR antagonizes JA in all traits analysed, and data obtained from double mutants evidenced that BR is upstream of the JA pathway despite the formation of such characteristics. Herbivory tests showed that this interaction is important in the defence against insect pests.

Table 1.

Hormonal mutants introgressed into the MT cultivar and used in this work

Mutanta	Chromosomeb	Class	Genetic function	Morphology	Reference
diageotropica (dgt)	1 (152)	Auxin	Low sensibility. Defective for a cyclophilin.	Agravitropic and lack of lateral roots. Hyponastic leaves.	Oh et al. (2006)
Never ripe (Nr)	9	Ethylene	Low sensibility. Defective for an ethylene receptor.	Ripening-impaired fruits.	Wikinson et al. (1995)
epinastic (epi)	4	Ethylene	Ethylene high producer. Unknown gene function.	Severely epinastic leaves.	Fujino et al. (1988)
notabilis (not)	7 (40)	ABA	Low levels of ABA. Defective in carotenoid cleavage (NCED).	Severe loss of water under high temperatures.	Burbidge et al. (1999)
gibberellin deficient3 (gib3)	7 (64)	GA	Low levels of GA. Disrupted in an ent-kaurene producer cyclase.	Dwarf phenotype. Small and dark-green leaves.	Koornneef et al. (1990)
procera (pro)	11	GA	GA constitutive. Loss of function in the repression domain of the tomato DELLA protein.	Increased height. Reduced lobbing of the main leaflets.	Bassel et al. (2008)
dumpy (dpy)	2 (57)	BR	Low levels of BR. Probably defective in the conversion of 6-deoxocatasterone to 6-deoxoteasterone.	Short stature. Altered leaf morphology.	Koka et al. (2000)
jasmonic acid insensitive1-1 (jai1-1)	5	JA	Insensitive to JA. Tomato homologue of Arabidopsis COI1.	Low density of trichomes in leaves and fruits. Female sterile.	Li et al. (2004)

a

An initial upper case letter in the names of genotypes denotes a dominant allele.

b

Numbers correspond to chromosomal localization and mutation position (in parentheses), in each of the 12 tomato chromosomes.

ABA, abscisic acid; BR, brassinosteroid; GA, gibberellin; JA, jasmonic acid.

Materials and methods

Plant material

Hormonal mutants were previously backcrossed for six generations and self-pollination (BC₆F₆) with the tomato MT cultivar (Carvalho, 2008; Zsögön et al., 2008; Table 1). The mutant jai1-1 was already available in the MT background (Li et al., 2004) and was kindly provided by Dr Gregg A Howe (Michigan State University). Plants were grown in a greenhouse under automatic irrigation (four times per day to field capacity), 28 °C, 11 h/13 h (winter/summer) photoperiod, and 250–350 μmol m⁻² s⁻¹ PAR (photosynthetically active radiation) irradiance. Mutant seeds were grown in trays containing a 1:1 mixture of commercial mix (Plantmax HT, Eucatex, Brazil) and expanded vermiculite mixture, supplemented with 1 g l⁻¹ 10:10:10 NPK and 4 g l⁻¹ lime. The gib3 mutant was germinated according to Koornneef et al. (1990). jai1-1 homozygous plants were selected using methyljasmonate (MeJA) medium (Li et al., 2004). Ten days after germination seedlings were transferred to 150-ml pots containing the aforementioned soil mix. All experiments were carried out using fully expanded terminal leaflets from the middle section of 40- to 60-d-old (flowering) plants.

Trichome quantification

Trichomes were classified according to Luckwill (1943), based on trichome stalk length/format and presence/absence of glands. Two terminal leaflets obtained from 20 plants (n=40) of each genotype were analysed by marking a 0.25 cm² area in the leaflet and counting every trichome found under a stereoscope. Since trichome density in tomato is dependent on leaflet age (Li et al., 2004), measurements were made using fully expanded leaflets taken from the middle section of the plants.

Scanning electron microscopy

Leaflets samples were fixed using a modified Karnowsky solution (Karnowsky, 1965), composed of 2.0% (v/v) glutaraldehyde in 0.05 M sodium cacodylate buffer at pH 7.2. Samples were then rinsed three times in distilled water and dehydrated in an ethanol series (30, 50, 70, 80%), followed by three changes in 100%. Samples were critical point dried through liquid CO₂, mounted in metal stubs, sputter coated with 20 nm gold, and examined under a LEO 435 (VP) scanning electron microscope at 20 kV.

Epidermal cell surface area

Quantification of epidermal cell surface area was made using the imprint technique (Weyers and Johansen, 1985). Speedex-Vigodent^® (Rio de Janeiro, Brazil) dental resins (Speedex Light Body + Speedex Universal Activator) was mixed as described by the manufacturer and applied to the leaflets’ surfaces. After complete drying of the resins (imprint), reverse imprints were obtained with nail polish (L'Oréal, São Paulo, Brazil) and photographed on a light microscope at ×100. Epidermal cell surface area was measured using the ImageJ software (http://rsbweb.nih.gov/ij/). Twenty plants of each genotype were used.

Allelochemical content

Tomato plants synthesize a wide range of allelochemicals capable of acting in defence against herbivory. Two of those metabolites, zingiberene (zgb) and acyl sugars (ASs), were determined in the hormone mutants due to their well-established capability of interacting negatively with herbivore insects (Carter et al., 1989a; Hartman and StClair, 1999; Nombela et al., 2000). Zgb and ASs were quantified using high precision spectrophotometric methods described for tomato (Freitas et al., 2002; Resende et al., 2002). The tomato-related wild species Solanum habrochaites S. Knapp & D.M. Spooner (PI127826), which is known to accumulate high zgb levels (Carter et al., 1989a), and Solanum pennellii Correll (LA716), also known to have an elevated AS content (Nombela et al., 2000), were used for comparison. They were germinated and grown in the same conditions described for MT and hormone mutants. For allelochemical quantification, 20 leaflets of each genotype were used.

Double mutant analysis

Double mutants were developed according to Weigel and Glazebrook (1989). Plants were selected for MeJA insensibility and confirmed by PCR-based analysis as described by Li et al. (2004) for the jai1-1 mutation and for the short stature and leaf curled morphology of the dpy mutant (Koka et al., 2000).

Quantitative real-time RT-PCR (qRT-PCR)

Total RNA extraction from leaflets was carried out using TRIzol reagent (Invitrogen). Reverse transcription and quantitative real-time amplifications were developed according to Leal et al. (2007) in triplicate.

Evaluation of expression was performed for two genes: (i) sesquiterpene synthase (SST), which encodes a key enzyme involved in the formation of several sesquiterpenes, the chemical class of zgb, present in tomato trichomes (Besser et al., 2009). Primers for a trichome-expressed SST were designed based on sequence information available at SOL Genomics Network (SGN ID=SGN-U314320) (5′-AAGTTCACCGATGACCAAGG-3′; 5′-CTGAATTGTGCTGCCTCGTA-3′). (ii) Proteinase inhibitor I (PI-I), which encodes a serine proteinase inhibitor involved in defence against insect herbivory in tomato (Graham et al., 1986). Primers for PI-I were designed based on the sequence available at GenBank (Young et al., 1994, accession number: L21194) (5′-TTGCTCTCCTCCTTTTATTTGG-3′; 5′-GCAAGCCTTGGCATGTTC-3′). Primers were designed using Primer3 (http://frodo.wi.mit.edu/) and evaluated for stability and mispairing with NetPrimer (http://www.premierbiosoft.com/netprimer/index.html).

Since wounding can alter the expression pattern of anti-herbivory genes in tomato (Graham et al., 1986; Li et al., 2001), wound responses were induced by causing a lesion with a scalpel through the middle vein of fully expanded leaflets. RNA was extracted 24 h after wounding and further processed for qRT-PCR analyses. For all gene expression experiments, gene transcript levels were normalized to tomato glyceraldehyde phosphate dehydrogenase (GAPDH; GenBank accession number: U97257) transcript levels (Zhong and Simons; 1999).

Herbivory test

Herbivory tests were carried out using the fall armyworm Spodoptera frugiperda J.E. Smith, a polyphagous insect, and the tomato leafminer Tuta absoluta Meyrick, an oligophagous insect and one of the main pests of tomato culture (Picanço et al., 1998). Although utilization of excised leaflets can sometimes lead to different results when compared with a whole-plant approach (Schmelz et al., 2001), the use of the detached leaflets was chosen in the herbivory tests as it offers a better experimental control especially when using small plants such as the dpy mutant, or very small larvae such as the young T. absoluta.

For S. frugiperda, two experiments were conducted. The first consisted of a non-choice assay, where one neonate larva was placed over a detached leaflet of the genotype tested. Experiments were conducted inside a glass vial containing moistened cotton for humidity. Completely eaten leaflets were replaced. Initially, and after 3, 6, and 9 days, all larvae were weighed. The second experiment was a two-choice assay, where three neonate larvae were placed within a 1 cm² circle located equidistantly between squares of 4.64 cm² obtained from leaflets from the wild type and a selected mutant. The assay was conducted inside Petri dishes containing a moistened filter paper for humidity. After 5 days, the final area of each leaflet square was measured.

Because of its small size, leading to low mobility and low weight, only the non-choice assay was performed for T. absoluta. Four neonate larvae of T. absoluta were placed over one detached leaflet of the genotype tested. The experiment was conducted inside a Petri dish containing a moistened filter paper for humidity. Mortality was counted every 2 days, during 8 days. All assays (S. frugiperda and T. absoluta) were carried out using neonate larvae instead of eggs to avoid the use of non-viable eggs. Experiments consisted of 20 repetitions and were performed at 27±2 °C with a 12 h light/dark photoperiod.

Statistical analysis

Each mutant line was independently tested, and data obtained for every genotype were compared with the MT treatment (control). Statistical inferences were made using the Student t-test at the 5% level of significance. For the sake of simplicity, just one MT average result is shown in the figures/graphics (except in Fig. 7A).

Results

High-density trichomes in hormonal mutants

Trichome types I, III, and VII (Fig. 1A, B, and D, respectively) appear at low density in tomato leaves (<50 trichomes cm⁻²; Supplementary Table S1, Supplementary Data available at JXB online), providing insufficient statistical accuracy to estimate differences between genotypes. High-density type V (Fig. 1C) and VI (Fig. 1D) trichomes, which are, respectively, the main non-glandular (NG) and glandular (G) trichomes present in tomato, were estimated at the adaxial (Fig. 2A, C) and abaxial (Fig. 2B, D) leaflet sides in eight hormonal mutants (Table 1).

Fig. 1.

Scanning electron micrographs of trichome types found in S. lycopersicum. (A) Glandular trichome type I (bar=50 μm); (B) non-glandular trichome type III (bar=50 μm); (C) non-glandular trichome type V (bar=20 μm); (D) glandular trichomes types VI and VII (bar=20 μm). Trichome distribution on the adaxial side of tomato leaflets for the genotypes MT (E); epi (F); dgt (G); dpy (H), and jai1-1 (I). From E to I: bars=200 μm.

Fig. 2.

High-density trichome (types V and VI) quantification in tomato genotypes (±SE, n=20). The densities of the non-glandular type V trichome (A and B) and the glandular type VI trichome (C and D) are shown for the adaxial (A and C) and abaxial (B and D) sides. Squares represent a statistically significant difference compared with MT according to Student t-test (P <0.05). Filled and open squares represent values statistically significantly higher or lower than MT, respectively. A description of hormone alterations of each genotype is given in Table 1.

A striking reduction in trichome density was observed for the ethylene-overproducer mutant epi (Fig. 2). The reduction in trichome density might also be accounted for indirectly through the increase in individual epidermal cell surface area (Traw and Bergelson, 2003), thus decreasing the number of cells (and trichomes) in a given area. For this reason, the epidermal cell surface area was evaluated in every mutant genotype under study. Type V trichome density in epi was >10-fold lower on the adaxial and almost 8-fold lower on the abaxial sides when compared with MT (P <0.001). The density of type VI trichomes was also significantly lower, being just 26% and 38% of that observed in the MT adaxial and abaxial sides, respectively. Compared with MT, epi leaflets presented a semi-glabrous phenotype (Fig. 1E, F). As shown in Fig. 3A–D, the epidermal cell surface area of leaflets in epi was larger than in MT on both leaflets sides, suggesting that epi may have a modified trichome density indirectly through a reduced number of epidermal cells per area. For the ethylene-low-sensitive mutant Nr, only a significant decrease (P <0.05) in type VI trichome density in the adaxial side in comparison with MT was observed. No significant differences either in leaflet epidermal cell area or in any other trichome density was observed in this mutant.

Fig. 3.

Epidermal cell surface area of tomato genotypes. Mean cell area (±SE, n=20) is shown for the adaxial (A) and abaxial (B) sides of MT and hormonal mutants. Squares represent a statistically significant difference compared with MT according to Student t-test (P <0.05). Filled and open squares represent values statistically significantly higher or lower than MT, respectively. Imprints of the abaxial side of leaflets from MT (C); epi (D); gib3 (E); dgt (F), and dpy (G). Bar=100 μm.

Two opposite gibberellin (GA) mutants where evaluated: gib3, defective in GA production, and pro, a DELLA mutant which exhibits a constitutive GA response phenotype (Table 1). Estimation of epidermal cell surface area for these mutants (Fig. 3A, B) underscored the importance of GA for cell elongation: gib3 cells were significantly smaller than those from MT for both sides (Fig. 3E), whereas pro showed a significantly larger (P <0.05) epidermal cell surface area on the abaxial face than MT. This might be the main cause of the increased trichome density in gib3 compared with MT for both types of trichomes and sides, except for type V on the adaxial side. For instance, type VI trichome density on the gib3 adaxial side was almost twice as high as in MT. However, the results obtained for pro are unclear, as a low density of trichomes V and VI was observed on the adaxial side (Fig. 2A), with no significant changes in epidermal cell area (Fig. 3A), whereas for the abaxial side (Fig. 2B), which presents a significantly higher cell area (Fig. 3B), only type VI trichome density was lower than that of MT.

The effects of the smaller cell area on the number of trichomes were also observed for the auxin less sensitive mutant dgt. Cell surface area in this mutant was ∼72% and 63% of that of MT for the adaxial and abaxial sides, respectively (P <0.001, Fig. 3A, B, F). This alteration is probably the main cause of the significantly higher density of type V and VI trichomes on both leaflet sides of this mutant, except type VI on the adaxial side (Fig. 2A, B). The hairy phenotype of dgt is also evident in micrographs (Fig. 1G).

The only significant trichome density alteration in the ABA-deficient mutant not was a reduction in type VI trichome density on the adaxial side (Fig. 2A). Although this mutant did not show any significant effect on trichome density, a reduced area of epidermal cell surface was observed (Fig. 3A, B).

dpy, a mutant defective in BR biosynthesis, showed a remarkable increase in trichome density (Fig. 2A, B, D) with no significant alteration in adaxial and abaxial epidermal cell surface area (Fig. 3A, B, and G). Indeed, the dpy leaflets surfaces showed a very hairy phenotype (Fig. 1H). The density of type V trichomes on both leaf sides of dpy was significantly higher than that of MT (P <0.001). For type VI trichomes, the density was higher only on the abaxial side. These results suggested that BRs may act as negative regulators on the formation of trichomes.

In contrast to the dpy phenotype, the JA-insensitive jai1-1 revealed a greatly reduced trichome density (Figs 1I, 2A, B), which, as observed for dpy, was not an effect due to changes in epidermal cell surface area (Fig. 3A, B). Quantification of G trichomes in jai1-1 showed a similar pattern to that described by Li et al. (2004) who proved that jai1-1 is disrupted in the formation of G trichomes. In the present experiments, type VI trichome density on the adaxial side of jai1-1 was only ∼11% of that observed for MT (P <0.001).

Allelochemical content in hormone mutants

Zgb, a sesquiterpene associated with tomato resistance against many insects (Carter et al., 1989a; Freitas et al., 2002), was quantified in hormone mutants and in the tomato wild relative species S. habrochaites, which is known to accumulate high zgb levels (Carter et al., 1989a).

As expected, the quantities of zgb in S. habrochaites were higher than in any tomato mutant analysed (Fig. 4A). Conversely, despite the fact that tomato usually presents low zgb levels, as observed for MT, the dpy mutant showed a 4.5-fold increase of the levels observed for MT, close to those of S. habrochaites. This result also suggests an important role for BRs in the regulation of the synthesis of a secondary metabolite involved in herbivory defence.

Fig. 4.

Allelochemical contents in tomato genotypes and two tomato-related wild species (±SE, n=20). (A) Relative zingiberene content per cm² of leaflets. Solanum habrochaites, a high zingiberene producer, is shown for comparison. (B) Acyl sugar content per cm² of leaflets. Solanum pennellii, a high acyl sugar producer, is shown for comparison. Squares represent a statistically significant difference compared with MT according to Student t-test (P <0.05). Filled and open squares represent values statistically significantly higher or lower than MT, respectively.

On the other hand, jai1-1, pro, and gib3 had a lower zgb content in the leaves. Since pro and gib3 are opposite GA mutants, it is unlikely that GA might be directly involved in zgb production. Li et al. (2004) already described low terpene content in jai1-1, but zgb levels were not measured in their work. In agreement with their results, it was shown here that JA is involved in secondary metabolite formation. No other mutant studied showed significant alteration in zgb content.

ASs are carbohydrate-based substances acting as natural glue on insects walking on leaves and are involved in defence against several tomato pests (Hartman and StClair, 1999; Nombela et al., 2000). Quantification of ASs was done in the hormone mutants and in the tomato wild relative species S. pennellii, known to have an elevated AS content (Nombela et al., 2000).

Small, yet significant reductions in AS levels were observed in jai1-1, Nr, and not (Fig. 4B). Although no hormonal mutant had an increased AS content, data from jai1-1, Nr, and not suggest that plants with opposite mutations in the same hormone classes would probably have increased AS levels. However, at least for ethylene, this assumption was proved to be wrong, since the ethylene-overproducer epi showed no alteration in AS content. As expected, S. pennellii showed high AS levels. Since S. pennellii possesses a type of glandular trichome not present in S. lycopersicum (type IV) (Luckwill, 1943), it is possible to speculate that this metabolite may be produced mainly in this trichome. This would explain the low levels of ASs and small alterations in S. lycopersicum hormonal mutants. On the other hand, the high zgb levels in dpy show that the production of this allelochemical is independent of type IV trichomes (Carter et al., 1989a).

Double mutant analysis

The results showed that ethylene, auxin, and GA are capable of contributing to the modification of trichome density, maybe indirectly, through the modification of the epidermal cell surface area. However, analysis of BR and JA mutants showed that these hormones are directly involved in trichome development and allelochemical biosyntheses. Since their responses proved to be opposite (JA promotes defence traits whereas BR reduces them), a possible hormonal interaction was further investigated by producing double mutants between dpy and jai1-1. F₁ progeny showed neither the dpy nor the jai1-1 phenotype, confirming the recessive nature of both mutations. F₂ plants showing MeJA insensibility (jai1-1 phenotype) and short stature with altered leaf morphology (dpy phenotype) were used for evatuation of the same anti-herbivory traits conducted for the single hormone mutants (Supplementary Fig. S1A at JXB online presents the phenotype of jai1-1, dpy, and their F₁ and F₂ generations of crossings). The homozygozity of the jai1-1 recessive allele in double mutants was also confirmed by a PCR-based method (Supplementary Fig. S1B) (Li et al., 2004). The double mutant appeared in a 1:15 (χ²=0.07) ratio, indicating independent segregation, which is consistent with the proposed chromosomal position of their corresponding loci (Table 1).

Trichome quantification in the dpy×jai1-1 double mutant

Besides the type V trichome density on the adaxial side of the double mutant demonstrating a similarity phenotype to dpy (Fig. 5A), all other trichome densities (Fig. 5B–D) presented no significant difference (P >0.1) between the double mutant and the jai1-1 mutant. This high phenotypic similarity suggests that jai1-1 is epistatic to dpy, which indicates that induction of trichome formation in dpy occurs through a jai1-1-dependent pathway. Scanning electron micrographs of the abaxial side showed that the trichome density in dpy×jai1-1 (Fig. 5H) resembles the jai1-1 phenotype (Fig. 5F) and differs from the dpy trichome density (Fig. 5G). An MT micrograph is displayed for comparison in Fig. 5E. The best interpretation of this epistatic hormone interaction is that BR negatively acts on trichome formation by controlling some point of the JA pathway upstream of COI1 (whose loss of function leads to the jai1-1 phenotype; Li et al., 2004).

Fig. 5.

High-density trichome quantification in tomato genotypes (±SE, n=20). The densities of type V (A and B) and type VI (C and D) trichomes are showed for the adaxial (A and C) and abaxial sides (B and D). Squares represent a statistically significant difference compared with MT according to Student t-test (P <0.05). Filled and open squares represent values statistically significantly higher or lower than MT, respectively. Scaning electron micrographs of the adaxial side of MT (E); jai1-1(F); dpy (G); and dpy×jai1-1 (H). Bars=200 μm.

Zgb quantification in double mutant

Zgb quantification in dpy×jai1-1 was not significantly different from that in jai1-1 (P >0.8, Fig. 6A). This result supports the conclusions drawn for trichome quantification, suggesting that indeed BR acts negatively in controlling zgb content through operating in the JA pathway. Interestingly, when the transcription activity of SST, an enzyme involved in sesquiterpene synthesis (Besser et al., 2009) and thus probably in zgb formation, was quantified no alteration was observed for dpy in comparison with MT (Fig. 6B).

Fig. 6.

Zingiberene content quantification (±SE, n = 20) and sesquiterpene synthase (SST) gene expression in tomato genotypes. (A) Relative zingiberene content per cm² of leaflets in MT, jai1-1, dpy, and the double mutant. Squares represent a statistically significant difference compared with MT according to Student t-test (P <0.05). Filled and open squares represent values statistically significantly higher or lower than MT, respectively. (B) qRT-PCR analysis of SST gene expression in wounded (+) and non-wounded (–) leaflets. RNA was extracted after 24 h of wounding. SST expression was normalized according to GAPDH transcript levels.

Fig. 7.

Herbivory tests and relative PI-I expression. Herbivory tests using S. frugiperda are displayed in A and B. (a) Non-eaten area (cm²) of the leaflet squares after the two-choice assay (±SE, n=20). An MT versus MT experiment was used as control. (B) Caterpillar weight in the non-choice assay (±SE, n=20). Larvae were weighed at the beginning of the experiment (day 0), and after 3, 6, and 9 d of the experiment. (C) Mortality of T. absoluta larvae 0, 2, 4, 6, and 8 d after inoculation in different genotypes (±SE, n=20). Squares represent a statistically significant difference compared with MT according to Student t-test (P <0.05). Filled and open squares represent values statistically significantly higher or lower than MT, respectively. (D) qRT-PCR analysis of PI-I gene expression in wounded (+) and unwounded (–) plants. RNA was extracted 24 h after wounding. PI-I expression was normalized according to GAPDH transcript levels.

Nevertheless, SST expression seems to be JA controlled, since low levels of SST transcripts were observed in jai1-1 and dpy×jai1-1. Although wounding is known to produce a rapid increase of JA endogenous levels in tomato (Howe et al., 1996), the present results showed that it was not enough to change the SST expression levels in any of the genotypes analysed (Fig. 6B).

Herbivory tests

To evaluate the relevance of the proposed interaction between BR and JA in defence against insect pests, dpy, jai1-1, and the double mutant were assayed in herbivory tests using a polyphagous insect, S. frugiperda, and one of the most important tomato pests, T. absoluta.

For S. frugiperda, when the caterpillars could choose a genotype to feed on (two-choice assay), no statistical differences were observed between dpy and MT (Fig.7A), suggesting that neither trichomes nor zgb are involved in defence against this insect. Support for this conclusion comes from results obtained with S. habrochaites, a pubescent and high zgb species that also did not differ from MT. The same pattern was obtained in the non-choice assay (Fig. 7B), where caterpillars were fed with leaflets of one pre-selected genotype only. No alterations in caterpillar weight were noticed when feeding with dpy, S. habrochaites, or MT.

On the other hand, jai1-1 and dpy×jai1-1 were significantly better for caterpillars than MT when those genotypes were confronted and also in the non-choice assay (Fig. 7A and B, respectively), showing that some additional anti-herbivory trait controlled by JA but not quantified in this work (e.g. polyphenol oxidase; Thaler, 1999) may be important in defence against S. frugiperda. Interestingly, although no alteration was observed for dpy, the double mutant dpy×jai1-1 was preferred (Fig. 7A) and produced heavier caterpillars (Fig. 7B) when compared with jai1-1. It is tempting to speculate that the curled and defenceless leaflets of the double mutant dpy×jai1-1 (see Supplementary Fig. S1A at JXB online) might be more nutritive or more easily digested by the larvae than the normally expanded leaflets of jai1-1.

Because of its small size, leading in low mobility and weight, only the non-choice assay was performed for the tomato pest T. absoluta, measuring mortality rates rather than weight changes. For this insect, trichomes and zgb appear to have an important deterrent effect. As demonstrated in Fig. 7C, mortality in S. habrochaites and dpy was significantly higher than in MT (P <0.001). For S. habrochaites all larvae were dead after 4 days. Mortality in dpy leaflets was >2-fold higher than in MT after 8 days. These results show the importance of tomato defensive traits in the control of specific pests and also point to the relevance of BR in defence against insect herbivory.

Results for the herbivory test with dpy×jai1-1 were similar to those for jai1-1 (Fig. 7C, P >0.85 for day 8), which is consistent with their similar patterns of defensive traits seen for trichome density (Fig. 5) and zgb content (Fig. 6A). These results once again indicate that jai1-1 is epistatic to dpy, supporting the idea of a hormonal interaction between BR and JA. Interestingly, in these two mutants, mortality was comparable with that of MT. Since T. absoluta is one of the most aggressive tomato pests, the levels of defensive traits present in MT might not be capable of generating perceptive defence against this insect.

The transcript levels of the serine proteinase inhibitor PI-I (Young et al., 1994) were also evaluated in the mutants. PI-I is highly correlated with defence against several Lepidoptera in tomato (Hilder and Boulder, 1999) and its expression is known to be promoted after wounding (Graham et al., 1986). Although levels of PI-I in unwounded plants in dpy were similar to those in MT, wounded dpy plants displayed an intensified PI-I expression compared with MT (Fig. 7D). High PI-I levels in dpy may be acting in concert with the elevated trichome number and zgb levels to generate the elevated mortality of T. absoluta observed in this mutant. As expected, no expression of PI-I was observed in jai1-1 and dpy×jai-1 mutants, even after wounding, which is in agreement with jai1-1 being epistatic to dpy. Taken together, the trichome density, zgb content, PI-I expression, and herbivory test strongly indicate an important function of BR in the control of anti-herbivory traits and that its action occurs through the JA pathway.

Discussion

Several hormone classes indirectly affect trichomes density

Tomato hormone mutants presented in the same genetic background (MT) have been evaluated to understand whether hormones were capable of affecting anti-herbivory traits. While JA has been the main focus of previous herbivory studies (Lin et al., 1987; Wasternack and Hause, 2002; Glazebrook, 2005), it was demonstrated in this study that other hormone classes are also relevant.

Trichome density was the most altered attribute in hormonal mutants among the traits studied. This alteration may occur directly, through changes in the determination of epidermal cells, or indirectly, by modifying the epidermal cell surface area and thus cell number per area unit (Traw and Bergelson, 2003). Ethylene, GA, and auxin seem to act indirectly, at least in part, on trichome density. The ethylene-overproducer epi, the GA low producer gib3, and the auxin less sensitive dgt mutants showed alterations in epidermal cell number leading to increased trichome density. Although inconclusive data were obtained for the ethylene less sensitive Nr and GA constitutive mutant pro, the involvement of these hormones in cell expansion is well established (Lang et al., 1982; Baluska et al., 1993; Wenzel et al., 2000; Fry, 2006). Despite the fact that no expressive alterations on other traits were observed in these mutants (zgb and AS levels), further analysis on the action of these hormones in tomato defence mechanisms is suggested for two reasons: first, even though epi presented a low trichome density, it was observed that younger leaflets of this mutant showed normal trichome numbers (Supplementary Fig. S2 at JXB online). Given that trichome number is dependent on leaf age (Li et al., 2004), it is suspected that ethylene may be involved in a process of ‘trichome senescence’. Secondly, there was a methodological challenge with the GA mutant: by germinating MT seeds in 100 μM GA₃ (required treatment used for gib3 to induce seed germination; Koornneef et al., 1990), promotion of trichome formation without an alteration in cell surface area (unpublished data) was observed, suggesting that trichome promotion in gib3 may be a residual effect of seed contact with GA₃. Indeed, GA has been previously suggested to promote trichome formation in Arabidopsis (Perazza et al., 1998).

BRs antagonize JAs in the formation of anti-herbivory traits

In contrast to ethylene, GA, and auxin, it was observed that BR and JA directly affect trichome formation, i.e. without significantly modifying the epidermal cell surface area. BR and JA were also involved in accumulation of zgb and PI-I transcripts, indicating the importance of these two hormones in defence against herbivory in tomato.

Using the JA-insensitive jai1-1 and the BR biosynthesis-defective dpy mutants, opposite patterns were observed. The results from the defective mutants confirmed that JA promoted the development of anti-herbivory traits whereas BR prevented it. For instance, trichome number in jai1-1 was severely reduced in comparison with MT, emphasizing the importance of JA in trichome formation (Li et al., 2004). On the other hand, dpy presented a high trichome density, indicating that BR negatively controls its formation. Interestingly, Perazza et al. (1998) also reported promotion of some trichome types in the Arabidopsis BR-insensitive mutant bri1; although this model species is not fully comparable with tomato due to the lack of glandular trichomes, some aspects of trichome development may have been evolutionarily conserved.

The same antagonistic effect between BR and JA was confirmed for zgb contents, as low levels were observed in jai1-1 and high levels in dpy. The association of zgb with insect resistance is well documented (Carter et al., 1989b; Maluf et al., 2001; Freitas et al., 2002); however, the control mechanisms are unclear. Gianfagna et al. (1992) showed that zgb content varies according to temperature and photoperiod. It is proposed herein that JA and BR can also control zgb content, a useful result for future research regarding the production of natural insecticides. Evidence is also provided of the involvement in zgb formation of SST, an enzyme whose transcription activity seems to be controlled by JA. However, SST expression was unchanged in the high-zgb dpy mutant. This result suggests that the alteration in zgb biosynthesis in dpy occurs through one of these three hypothesized routes: (i) independently of SST (i.e. via a biosynthetic route that bypasses the reaction catalysed by this enzyme); (ii) at a point upstream of this enzyme in the zgb biosynthesis pathway; or (iii) given that tomato presents two related loci coding for SSTs nearly identical in their nucleotide sequences but with divergent substrates (van der Hoeven et al., 2000), it cannot be ruled out that the expression pattern obtained is being masked by the expression of the SST paralogue.

JA and BR were involved in PI-I accumulation. No detectable levels of PI-I transcripts were observed in jai1-1 even after wounding, which is expected since this mutant is impaired in lesion responses (Li et al., 2004). However, the intense increase in PI-I transcript accumulation after wounding in dpy suggests that BR may act not only in the control of constitutive defences but also in wound-induced responses in tomato. Holton et al. (2007) observed similar results, where the tomato BR-insensitive mutant curl3 overexpressed PI-II after wounding. However, the authors proposed an effect of leaf morphology rather than an alteration in wounding responsiveness.

The herbivory tests with dpy and jai1-1 support our hypothesis: since the polyphagous S. frugiperda can adapt its digestive physiology to bypass inhibitory effects of defensive traits (Paulillo et al., 2000; Brioschi et al., 2007), neither zgb, trichomes, nor PI-I overexpression were capable of establishing an effective defence against this insect, explaining the similar results for dpy and even S. habrochaites in comparison with MT. Conversely, S. frugiperda herbivory in the defenceless jai1-1 suggests that this easily accessible insect has potential for future works involving JA mutants.

When using the oligophagous leafminer T. absoluta, the action of anti-herbivory traits promoted by dpy became clearer. Trichomes, zgb, and PI-I seem to act effectively in the defence against the pest, as elevated mortality rates were observed in S. habrochaites and dpy. Tuta absoluta is one of the main pests in tomato crops (Picanço et al., 1998) and its control generally occurs with utilization of insecticides, which are leading to population resistance (Consoli et al., 1998; Siqueira et al., 2000). For this reason, more effective control methods need to be developed and the present work suggests that BR may shed some light on this problem. It is important to note that MT itself represents a cultivar with reduced levels of BR, since it harbours the dwarf mutation (Lima et al., 2004; Martí et al., 2006), a locus involved in BR biosynthesis (Bishop et al., 1999). This implies that differences in anti-herbivory traits seen here comparing dpy and MT may be even more accentuated comparing others cultivars with their isogenic BR mutants. Such differences may foster the development of commercial cultivars or hybrids harbouring weak alleles of loci involved in BR physiology, which would combine pest resistance without the extreme detrimental dwarfism of the dpy mutant.

BRs act through the jasmonate pathway

Since the defensive traits analysed here showed opposite results for the BR and JA mutants, a double mutant was developed to evaluate a potential hormonal interaction. The dpy×jai1-1 double mutant showed the jai1-1 phenotype, which was also reflected in the herbivory results. The interpretation is that BR negatively controls anti-herbivory traits in tomato (an opposite effect to that of JA) through acting on the JA pathway, at some point upstream of jai1-1. Although BR action seemed similar to that of salicylic acid (SA) by antagonizing JA responses (Zarate et al., 2007), Nakashita et al. (2003) showed previously that SA is not essential for BR action.

A plausible mechanism to explain the opposite effects of BR and JA in tomato would rely on the so-called ‘plant dilemma’ of resource allocation to growth versus defence (Herms and Mattson, 1992). Since BRs are involved in plant growth (Koka et al., 2000) and JAs are clearly involved in defence (Li et al., 2004), the interaction between these hormones might represent an interesting way of modulating resource investment. The present results show that BR may be the hormone that adjusts the resource allocation by changing the sensitivity to JAs: when using a tomato mutant impaired in BR biosynthesis, dpy (thus, impaired in growth), an up-regulation of the jai1-1-dependent jasmonate pathway was observed, leading to an increased production of defensive traits. Indeed, the control of jasmonate sensitivity is now being suggested as a way plants found to solve this dilemma (Moreno et al., 2009).

However, this BR×JA presumption does not seem to be evolutionarily conserved in the entire plant kingdom: in Arabidopsis, BR stimulates the JA pathway through activation of 12-oxophytodienoate reductase 3 (OPR3) transcripts, an enzyme involved in JA biosynthesis (Mussig et al. 2000; Strassner et al., 2002). A preliminary analysis of tomato OPR3 transcripts in dpy showed, however, no gene expression alteration in comparison with MT (unpublished data). Thus, the antagonistic effect between BR and JA seen here in tomato also suggests that such cross-talk could be the opposite of that proposed for Arabidopsis (Mussig et al. 2000; Strassner et al., 2002). Interestingly, even though JA is also responsible for trichome formation in Arabidopsis (Traw and Bergelson, 2003), opposite roles for this hormone in male and female fertility have been reported when comparing this species with tomato (Li et al., 2001). Further experiments should be conducted to search for a possible cross-talk node between these two hormones, the possible evolutionary divergence in dicots, and the relevance of this interaction for other developmental traits in different species. The proposed model of BR and JA action in defence against herbivory in tomato (Fig. 8) confirms, as proposed by Robert-Seilaniantz et al. (2007), the relevance of multiple hormones in plant–pest interactions.

Fig. 8.

Brassinosteroid (BR) and jasmonic acid (JA) interaction in the defence against herbivory in tomato. In the proposed model, BR antagonizes JA somewhere upstream of jai1-1, in the JA biosynthesis or signalling pathway, acting in the formation of some anti-herbivory traits in tomato.

Concluding remarks

Herein evidence was presented for a novel BR function as a key hormone in defence against herbivory by negatively interacting with the JA pathway. Such novel function and cross-talk also have an interesting agronomic appeal. The contrasting mechanisms of defence against biotrophic and herbivory/necrotrophic pathogens (Glazebrook, 2005) indicate that the simple manipulation of the JA content would not be applicable for pest control in the complex agricultural environment. However, a fine modulation of two or more classes of hormones with opposite roles could supply an optimum level for multiresistance. Moreover, some BR defective mutations cause a semi-dwarf phenotype (Sakamoto et al., 2006), which resembles the GA-based semi-dwarfism used to increase crop yield (Silverstone and Sun, 2000) in the so-called ‘Green Revolution’ (Peng et al., 1999; Hedden, 2003). The search for weak alleles of BR-related genes or mutations that causes a not so severe phenotype as the dpy used here, such as the rice osdwarf4-1 (Sakamoto et al., 2006), may provide for the development of BR-based semi-dwarf crops combining high harvest index along with improved defence against pests.

We thank FAPESP (grant 02/00329-8 and fellowship 06/05911-8) and CNPq (grant 475494/03-2 and fellowship 308075/03-0) for financial support. We also thank Celso Omoto (ESALQ/USP) for providing S. frugiperda eggs, Fernando K Edagi (ESALQ/USP) for helping with the spectrophotometer, Elliot Kitajima (NAP/MEPA-ESALQ/USP) for the electronic microscopy facility, and Jason Wargent (Lancaster University, UK) for helping with the leaflet imprints.

Author notes

*

Present address: Genetics and Developmental Biology Program, Division of Plant and Soil Sciences, West Virginia University, 1090, Agricultural Science Building, Morgantown, WV 26506, USA