Solanum nigrum: A model ecological expression system and its tools


Ian T. Baldwin. Fax: 03641 571102, E-mail:


Plants respond to environmental stresses through a series of complicated phenotypic responses, which can be understood only with field studies because other organisms must be recruited for their function. If ecologists are to fully participate in the genomics revolution and if molecular biologists are to understand adaptive phenotypic responses, native plant ecological expression systems that offer both molecular tools and interesting natural histories are needed. Here, we present Solanum nigrum L., a Solanaceous relative of potato and tomato for which many genomic tools are being developed, as a model plant ecological expression system. To facilitate manipulative ecological studies with S. nigrum, we describe: (i) an Agrobacterium-based transformation system and illustrate its utility with an example of the antisense expression of RuBPCase, as verified by Southern gel blot analysis and real-time quantitative PCR; (ii) a 789-oligonucleotide microarray and illustrate its utility with hybridizations of herbivore-elicited plants, and verify responses with RNA gel blot analysis and real-time quantitative PCR; (iii) analyses of secondary metabolites that function as direct (proteinase inhibitor activity) and indirect (herbivore-induced volatile organic compounds) defences; and (iv) growth and fitness-estimates for plants grown under field conditions. Using these tools, we demonstrate that attack from flea beetles elicits: (i) a large transcriptional change consistent with elicitation of both jasmonate and salicylate signalling; and (ii) increases in proteinase inhibitor transcripts and activity, and volatile organic compound release. Both flea beetle attack and jasmonate elicitation increased proteinase inhibitors and jasmonate elicitation decreased fitness in field-grown plants. Hence, proteinase inhibitors and jasmonate-signalling are targets for manipulative studies.


Molecular techniques are widely used in all biology, but their incorporation into ecological studies has largely been confined to the characterization of population structure and species distributions. As a consequence, the extraordinary advances that molecular techniques have permitted most biological disciplines, namely the ability to identify the genetic basis of a biological phenomenon and manipulate it, have yet to be realized in ecology. Many reasons underlie the nonparticipation of ecologists in the genomics revolution, but the limited availability of appropriate model systems has played an important role. Molecular tools developed for one model system can be difficult to transfer to other systems without substantial investment in technique development. Most techniques have been developed for agronomically and economically important organisms and cannot be readily applied to wild relatives. Transformation systems, in particular, can be difficult to use with near relatives. Agrobacterium-mediated gene transfer protocols are available for a number of higher plants and fungi, and allow the manipulation of the expression of genes mediating ecological interactions if these transformation systems have been adapted for native species. Plants and herbivores account for the majority of all higher species (Strong et al. 1984), and their interactions structure many of the planet's biological processes. The transformation of autotrophs allows for the ‘bottom-up’ manipulation of ecological interactions and thereby provides a powerful tool for studying community and ecosystem processes.

Once the ability to manipulate the expression of individual genes in a native species is available, the next task is to decide which genes to manipulate and how to interpret the responses to the manipulations. Microarrays allow biologists to examine the expression of hundreds of genes simultaneously, and their use in combination with elicitation studies provides a powerful means of identifying ‘suspect’ genes relevant for ecological interactions (Hui et al. 2003; Korth 2003). Alternatively, various differential display procedures (Voelckel & Baldwin 2003) allow researchers to ‘ask the organism’ to identify transcripts relevant in a given ecological interaction. Once a gene or a suite of genes has been selected for manipulation, the next challenge lies in interpreting the fitness consequences of the manipulation. Laboratory bioassays are not likely to provide a full functional understanding of many traits elicited by biotic interactions as is illustrated by the traits mediating plant–herbivore interactions.

Plants are known to recruit components of their community, as graphically illustrated by the elicitation of indirect defences (volatile organic compound [VOC] emissions, extrafloral nectar production, etc.) that plants use to enlist the natural enemies of herbivores in their defence against herbivores. A functional understanding of these complex responses is possible only in the context of the selective forces under which these responses evolved, namely their natural environments. Moreover, ecologists have long known that competition from other plants and herbivore pressure represent the two most important selective forces determining relative plant fitness in natural habitats (Begon et al. 1996). As a consequence, traits mediating competitive ability and herbivore resistance are likely to be intertwined, and understanding the genetic basis of these responses will require experimental manipulations in natural environments. Lastly, because environmental performance is a whole-plant trait best measured by various surrogates of Darwinian fitness (seed set, male reproductive success), an understanding of how the expression of a gene product contributes to a plant's Darwinian fitness is required.

The choice of an ecologically relevant organism is crucial for the study of plant–environment interactions. Adaptive responses are mediated by complex polygenic traits (Simms & Rausher 1992), and because agricultural plants have long been under intense selection for particular yield-enhancing traits, genetic associations mediating adaptive traits are likely to have been altered during agricultural selection and hence are difficult to interpret in these plants. We introduce a suite of molecular tools for the native plant Solanum nigrum that should facilitate the identification and manipulation of the genes that mediate these complex environmental responses. S. nigrum was selected not only because of its phylogenetic proximity to the agricultural species tomato and potato, for which substantial genetic tools are available, but also because its particular natural history makes it ideal for studying the interaction of competition and herbivore resistance. S. nigrum is attacked by various herbivores from different feeding guilds and grows in association with many other species. As an annual, it colonizes nitrogen-rich agricultural and disturbed habitats at a wide range of altitudes throughout its pan-arctic distribution (Edmonds & Chweya 1997).

If S. nigrum is to become a model ecological system, a minimum number of molecular tools are necessary. Most important, it must be readily transformable so that hypotheses about the ecological function of particular genes can be falsified. Here, we present such a transformation system for S. nigrum using an antisense (as) construct of the key photosynthetic gene RuBPCase. cDNA libraries of environmentally elicited plant tissues provide a means of cloning genes and microarrays allow biologists to examine the expression of hundreds of genes simultaneously. Here we present such a library and a 789-oligonucleotide microarray, representing 558 genes of ecological interest. We analyse traits thought to be important for plant performance that are quantifiable in complex environments and offer a means of measuring their correlation with S. nigrum's Darwinian fitness. We have selected a direct defence and an indirect defence for the analysis. Proteinase inhibitors (PIs) are among the best-studied induced direct defence chemicals in plants (Ryan 1990; Jongsma et al. 1994; Heath et al. 1997; Koiwa et al. 1997), which function by inhibiting particular digestive proteinases of herbivores, such as chymotrypsin and trypsin. High PI content has been found to reduce herbivore growth in plants that were transformed with heterologous PI genes (Rahbe et al. 2003; Xu et al. 1996). Xu et al. (2001) characterized two PIs of the pin2 family (SaPIN2a and SaPIN2b) in S. americanum, a species belonging to the taxonomically diverse S. nigrum group (Edmonds & Chweya 1997). We isolated the SaPIN2b homologue from S. nigrum (SnPIN2b) and used it to verify the responses observed on the array with a RNA-gel blot and real-time quantitative polymerase chain reaction (qPCR). We present a technique for quantifying the herbivore-induced VOC emissions from plants and illustrate its use with measurements from field-grown plants. VOCs are known to attract predators to the herbivore-damaged plant and therefore play a key role in plant–insect interactions (Kessler & Baldwin 2001; Baldwin et al. 2002; Dicke et al. 2003).

Materials and methods

Plant growth

The hexaploid Solanum nigrum L. inbred line, Sn30, of seeds collected from the field site in Jena, Germany (voucher specimens of the Sn30 line are deposited in the Max Planck Institute for Chemical Ecology branch of the Herbarium Haussknecht [JE], Jena) was used in all experiments. Seeds were incubated in 3.5 mm Ca(NO3)2 overnight at 4 °C and germinated in a peat-based substrate with clay additions (Tonsubstrat). Plants were grown in individual 400-mL pots for 3 weeks in the greenhouse (26 °C/16 h light; 25 °C/8 h dark). After 5 days of acclimatization to outside conditions, plants were randomly planted into monoculture plots at the experimental field site. The site is a former agricultural field with alluvial loam as substrate and located north of Jena, Germany. For the greenhouse experiments, plants were grown in 2 L pots with supplemental lighting from 400 W Na-vapour HID lamps and watered once a day.

Agrobacterium-based transformation

A fragment of the Nicotiana attenuata gene for RuBPCase (Hermsmeier et al. 2001) was PCR amplified. After digestion with XhoI and BstEII the resulting fragment (323 bp) was cloned in pRESC20 (Zavala et al. 2004), yielding the transformation vector pRESC2RUB (10.0 kb) which was used for the transformation of S. nigrum.

S. nigrum seeds (inbred line Sn30) were used for transformation. Seeds were sterilized for 5 min in a 5-mL aqueous solution of 0.1 g dichloroisocyanuric acid (Sigma) with 50 µL of 0.5% (v/v) Tween-20 (Merck). Seeds were washed three times with sterile water and incubated for 3 days at 4 °C in an aqueous solution of 3.5 mm Ca(NO3)2·4H2O (Merck). Subsequently, seeds were washed three times with sterile water and transferred onto a germination medium containing Gamborg's B5 with minimal organics (Sigma) and 0.6% (w/v) phytagel (Sigma). The plates were maintained in a growth chamber (Percival) at 26 °C/16 h light with 155 µm/m2/s PAR at shelf height and 24 °C/8 h dark. Agrobacterium tumefaciens strain 4404 was maintained and cultivated for transformation as described in Krügel et al. (2002). Hypocotyls of sterile 1-week-old seedlings were excised with a scalpel dipped into the Agrobacterium suspension, and co-cultivated as described in Krügel et al. (2002). Within one month, explants on callus induction media developed calli followed by green shoot primordia and shoots, so that subculture onto maturation media (described in Krügel et al. 2002) was necessary only until plantlets formed. Plantlets were subsequently subcultured onto maturation media, which also served as rooting media in this case, until roots appeared. Further selection of the putatively transformed plants, including segregation analysis of T1 plants using germination bioassays on hygromycin-containing media, is described in Krügel et al. (2002).

Microarray hybridization and analysis

Pooled leaf samples were ground under liquid nitrogen and total RNA was extracted with TRI REAGENT™ (Sigma) according to the manufacturer's instructions. The herbivore-infested test samples were labelled with Cy3 and the corresponding control (reference) samples with Cy5 according to the procedure described in Halitschke et al. (2003). The labelled samples were hybridized to the microarray (789 50-mer oligonucleotides spotted onto an epoxy-coated glass slide; Quantifoil Microtools) according to the published procedure (Halitschke et al. 2003).

An Affymetrix 428™ Array Scanner (Affymetrix) was used to scan the hybridized microarrays with sequential scanning for Cy5 cDNA and then for Cy3-labelled cDNA at a maximum resolution of 10 µm/pixel with a 16-bit depth. The images were evaluated with the aida image analyser (Raytest Isotopenmeßgeräte GmbH) software. Each image was overlaid with a grid to assess the signal strength (quantum level = QL) for both dyes from each spot. The background correction was calculated with the ‘nonspot’ mode of the aida software package.

The microarray-specific normalization factor was calculated based on the Cy5/Cy3 total fluorescence ratio (Halitschke et al. 2003). The ratios of normalized fluorescence values for Cy3 and Cy5 of each individual spot (expression ratio = ER) and the mean of the four replicate spots for each cDNA were calculated. A transcript was defined as being differentially regulated, if the following three criteria were fulfilled: (i) the average expression ratio for the four spots exceeded the thresholds (0.67 and 1.5); (ii) the individual expression ratios were significantly different from 1, as determined by a t-test; and (iii) the combined signal fluorescent intensity from both Cy3 and Cy5 averaged over the four spots was > 1000 QL. A complete list of all signal ratios (± SE) and details on all spotted genes is available from the authors. To evaluate these criteria, we hybridized two microarrays with the same cDNA pools and found that 84% of the genes had the same regulation (Heidel & Baldwin 2004).

RNA gel blot analysis

RNA samples (20 µg) were size-fractionated by 1.2% (w/v) agarose formaldehyde gel electrophoresis and capillary blotted onto a nylon membrane (GeneScreenPlus; NEN-DuPont) as described in the manufacturer's instructions. Ethidium bromide staining of the gel prior to blotting revealed rRNA bands, which served as the loading control. After blotting and UV cross-linking, 32P-labelled probes specific for PIN2b were used for detection. The probe for PIN2b was obtained by PCR of S. nigrum cDNA with primers specific for SaPIN2b from S. americanum (Xu et al. 2001). This fragment was used to screen a S. nigrum leaf cDNA library (Lambda ZAP II kit, Stratagene), and the longest resulting PIN2b sequence (679 bp; SnPIN2b; AY422686) was used as a probe to detect PIN2b transcripts.

Real-time qPCR

Total RNA was reverse transcribed into cDNA using SuperScript™ II RNaseH-Reverse Transcriptase (Invitrogen) according to the manufacturer's instructions. The amount of cDNA template used per well was reverse transcribed from 10 ng total RNA; each sample was replicated three times. The following sequences were used for the design of primers specific for PR-1, SnPIN2b, psbA and RuBPCase: Lycopersicon esculentum PR-1 (Tornero et al. 1997), S. nigrum PIN2b (see above), S. nigrum photosystem II D1 protein (Zhu et al. 1989) and L. esculentum RuBPCase LESS17 (McKnight et al. 1986). 18S RNA (template for primers: S. tuberosum gene for 18S RNA, GenBank Accession no. X67238) was used for quantitative normalization. The ABI PRISM® 7700 Sequence Detection System (Applied Biosystems) was used for the SYBR Green I-based assay. The qPCR™ Core Kit for SYBR® Green I (Eurogentec) was used according to the manufacturer's instructions with the following cycler conditions: 10 min at 95 °C; 40 cycles: 30 s at 95 °C and 30 s at 60 °C. To ensure the specificity of the PCR, a melting curve analysis was conducted using the abi prism® 7700 Dissociation Curve Software. To detect asRuBPCase transcripts amplicons specific for the as-construct were designed (Halitschke & Baldwin 2003). The assay using a double dye-labelled probe was performed on an ABI PRISM® 7700 Sequence Detection System (qPCR™ Core Kit, Eurogentec) with 18S RNA for normalization (TaqMan® Ribosomal RNA Control Reagents, Applied Biosystems). The relative expression of the target genes was determined by using standard curves (Applied Biosystems 1997).

Isolation and blotting of genomic DNA

Plant genomic DNA was prepared from leaves of S. nigrum using CTAB (Reichhardt & Rogers 1994). DNA samples were restriction digested with EcoRV, size-fractionated by 0.8% agarose gel electrophoresis, and Southern blotted onto a nylon membrane with high-salt buffer (Brown 1995). The blot was analysed with 32P-labelled probe specific for the hygromycin resistance gene (hph).

Plant treatments

MeJA induction.  We applied 250 µg of methyl jasmonate (MeJA; Sigma) in 20 µL lanolin (Sigma) to the stem of the 5-week-old plants (n = 62) above the third leaf node. To exclude possible lanolin effects, we treated control plants (n = 55) with 20 µL lanolin.

Flea beetle damage.  Flea beetles (Epitrix pubescens) were abundant at our field site. To compare uninfested control plants with flea beetle-infested plants, we sprayed the controls (n = 28) with the pyrethroid-based insecticide Spruzit (0.1%, Neudorff) and the infested plants (n = 31) with water directly after planting each day until tissue was harvested for microarray and PI analysis. The average flea beetle load of the water-sprayed plants for the duration of the experiment was ~40 adults per plant, compared with 1–5 adults per plant for pyrethroid-sprayed plants. In a comparable field experiment, Baldwin (1998) detected no influence of pyrethrin treatment on inducible defences (nicotine) in the Solanaceous plant, Nicotiana attenuata.

For the microarray analysis, we harvested and pooled fully expanded leaves of eight individual plants 48 h after exposure to flea beetles, flash-froze them in liquid nitrogen, and stored the leaf samples at −80 °C until RNA extraction. For the PI analysis, a systemic leaf near the youngest node of all control and elicited plants was harvested 3 days after induction (MeJA or flea beetle), flash-frozen in liquid nitrogen, and stored at −80 °C until protein extraction.

Herbivore comparison. Leptinotarsa decemlineata originated from wild populations on S. tuberosum, Acherontia atropos, from a laboratory population reared on S. nigrum. To achieve approximately the same leaf area damage among herbivore treatments, we placed two L. decemlineata adults or two second-instar A. atropos larvae on individual 5-week-old plants (five replicates per treatment). After 24 h of continuous feeding we collected VOCs from the differentially treated plants (see VOC analysis).

Trypsin–PI analysis

Harvested leaves were ground in liquid nitrogen. Proteins were extracted according to the protocol used for Nicotiana attenuata (Van Dam et al. 2001) and protein content was measured by the method of Bradford (1976) with IgG (Sigma) as standard. We determined the activity of trypsin inhibitors by the radial immunodiffusion assay (Jongsma et al. 1993). A series of soybean trypsin inhibitor (STI, Sigma) solutions was used to obtain a reference curve. Trypsin–PI activity is expressed as nmol/mg of total protein.

VOC analysis

In an open-flow trapping system, VOCs of one fully expanded stem leaf were collected (see Fig. 3A). To confine insects to a single leaf and to trap volatiles from the same leaf, leaf and insects were enclosed in 400-mL polystyrene chambers fitted with holes at both ends. Air was pulled through the chamber at 450–500 mL/min (measured by a mass flow meter: Aalborg Instruments) and subsequently through a charcoal air-sampling trap (ORBO™-32; Supelco) using a portable vacuum pump. Each charcoal trap was spiked with 300 ng tetraline as an internal standard (ISTD) for quantification, eluted with 750 µL dichloromethane, and analysed by GC-MS according to Halitschke et al. (2000).

Figure 3.

Expression of Solanum nigrum genes in response to attack from flea beetles, Epitrix pubescens, by microarray, RNA gel-blot, and real-time qPCR analysis (C = control, FB = flea beetle infested). Microarray analysis revealed gene regulation in different categories and independent gel-blot and real-time qPCR analyses verified expression patterns of individual genes from the array. The expression ratios (normalized mean Cy3/Cy5 ratio) for three categories (defence, photosynthesis, signalling) from an array hybridized with fluorescently labelled cDNA from attacked and unattacked plants are depicted. Inset Table (A) summarizes the numbers of up- and downregulated genes in the categories. Arrows identify genes used for verification of the expression data (B–E). (B) RNA gel blot analysis hybridized with an SnPIN2b probe (18S RNA is shown as a control for equal loading). (C–E) Real-time qPCR for the genes (C) SnPIN2b, (D) PR1 and (E) psbA. The expression level in the control sample = 1.


Herbivore community

During the 2002 and 2003 growing seasons, we sampled different native and planted populations of Solanum nigrum near Jena for insects. In 2002, the main collection sites were a planted population in Jena-Isserstedt (~100 plants) and native and planted populations in Jena-Nord (~500 plants); 2003 specimens were collected mainly from a planted population (~140 plants) on the Jena/Beutenberg-Campus. Populations were sampled at least once a week from May to September of each year. We classified insects as being herbivores on S. nigrum only if the adult insects or their larvae were repeatedly observed to feed on S. nigrum in both years (Table 1). Many additional insect species were observed on the plants, but only a small subset were observed to consistently feed on this plant. For example, only 3 of at least 24 Coleopteran species repeatedly found on the plants were classified as S. nigrum herbivores. The phytophagous insects belong to different groups according to their feeding behaviour and host-plant specialization. Two Solanaceous specialists, both leaf-chewing beetles, namely the Colorado potato beetle Leptinotarsa decemlineata and the flea beetle Epitrix pubescens, were repeatedly found on S. nigrum (Fig. 1). The flea beetles infested the plantation and native S. nigrum plants heavily throughout 2002 in Jena-Nord. L. decemlineata appeared in August, having dispersed from small potato fields. In addition to the leaf-chewing species, we found a variety of piercing-sucking insects, including bugs, cicadas, and aphids. Aphis fabae colonized S. nigrum plants from May to September. Later in the season (July/August) mirid bugs of the genus Lygus were the most abundant herbivores. The rich herbivore fauna attracted a large variety of parasitoids and predatory species [syrphids (Fig. 1F) chysopids, coccinellids, etc.].

Table 1.  List of herbivore species observed feeding on Solanum nigrum in experimental field plots in Jena, Germany, during two growing seasons. Species are classified by feeding guild (C- leaf chewing, PP- piercing sucking on phloem, PM- piercing sucking on mesophyll, PX- piercing sucking on xylem) and whether they are specialized on Solanaceous plants
SpeciesOrderFamilyFeeding guildSpecialization
Epitrix pubescensKochColeopteraChrysomelidaeCSpecialist
Leptinotarsa decemlineataSayColeopteraChrysomelidaeCSpecialist
Barypeithes pellucidusBoh.ColeopteraCurculionidaeCGeneralist
Plutella xylostella LinnèLepidopteraPlutellidaeCGeneralist?
Lygus pratensisLinnèHeteropteraMiridaePMGeneralist
Lygus rugulipennisPoppiusHeteropteraMiridaePMGeneralist
Lygus wagneriRemaneHeteropteraMiridaePMGeneralist
Stenodema sericansFieberHeteropteraMiridaePMGeneralist
Dolycoris baccarumLinnèHeteropteraPentatomidaePMGeneralist
Holocostethus vernalis WolffHeteropteraPentatomidaePMGeneralist
Corizus hyoscyamiLinnèHeteropteraRhopalidaePMGeneralist
Philaenus spumariusLinnèAuchenorrhynchaCercopidaePPGeneralist
Balclutha punctataFabriciusAuchenorrhynchaCicadellidaePPGeneralist
Evacanthus interruptusLinnèAuchenorrhynchaCicadellidaePPGeneralist
Empoasca sp.AuchenorrhynchaCicadellidaePPGeneralist
Eupteryx aurataLinnèAuchenorrhynchaCicadellidaePPGeneralist
Macrosteles sexnotatusFallènAuchenorrhynchaCicadellidaePPGeneralist
Aulacorthum solani langeiBörnerSternorrhynchaAphididaePPGeneralist
Aphis fabaeScopoliSternorrhynchaAphididaePPGeneralist
Figure 1.

Herbivores on Solanum nigrum: (A) Colorado potato beetle (Leptinotarsa decemlineataSay); (B) flea beetle (Epitrix pubescensKoch); (C) cicada (Macrosteles sexnotatusFallèn); (D) mirid bug (Lygus wagneriRemane); (E) pentatomid bug (Dolycoris baccarumLinnè); (F) bean aphids (Aphis fabaeScopoli) and syrphid fly.


We developed an Agrobacterium-based transformation procedure for S. nigrum, which requires ~5–6 months from the transformation to the production of T0 plants bearing mature fruit. Transformation and regeneration of plantlets (2 cm in size) that can be transferred into soil requires 2 months. Callus generation in S. nigrum occurred at a lower rate than it does in Nicotiana attenuata, but the calli that did generate, did so at a higher rate (M. Lim unpublished results). Short tissue culture times are helpful in reducing somaclonal variation, which limits the production and utility of transgenic plants (Beaujean et al. 1998). In comparison to the published S. nigrum protoplast transformation procedure using Agrobacterium rhizogenes (Wei et al. 1986), our protocol proved to be more efficient in the regeneration of plants. Also, the use of Ri-based vectors increases the frequency of phenotypically abnormal and infertile plants in comparison with the T-DNA based vectors used here (Davey et al. 1987). Hygromycin resistance (hph) proved to be a reliable selectable marker that could be incorporated into the germination media and subsequently allowed seedlings to be rapidly selected. The efficiency of the procedure is high (Fig. 2B inset): 19 of the 22 haphazardly selected asRuBPCase lines (86%) could be verified as harbouring the transgene by means of antibiotic selection and PCR. Remarkably, the transgenes in the hexploid S. nigrum line segregated as in a diploid, in accordance with findings for transgenes in the tetraploid Arabidopsis suecica (Lawrence & Pikaard 2003), indicating that chromosome pairing occurs among homologues. For further characterization, we selected 10 T1 lines and examined them by DNA gel blot analysis (four lines are shown in Fig. 2A). All lines contained the transgene, and 2 of the 10 lines tested contained the transgene as a single copy insertion. Real-time qPCR supported the successful incorporation of the transgene into S. nigrum’s genome (Fig. 2C). As expected, no antisense (asRuBPCase) transcripts were detected in wildtype (wt) plants, whereas these transcripts were abundant in the transformed lines. The quantity of asRuBPCase-transcripts in transformed lines correlated negatively with the quantity of RuBPCase-transcripts (Pearson's correlation coefficient r = −0.549, P = 0.0327), demonstrating that the transformation had successfully reduced the expression of this important photosynthetic gene.

Figure 2.

Characterization of asRuBPCase Solanum nigrum lines. (A) DNA gel-blot analysis of three asRuBPCase T1 lines (as1, as2, as3, as4; three replicates each) and one wildtype, untransformed line (wt). Genomic DNA was digested with EcoRV and the blot was hybridized with a probe specific for the hygromycin resistance gene (hph). The lines as1 and as2 have single copy insertions of the transgene, whereas as3 has two copies and as4 has multiple copies. The wt DNA is shown as a negative control. (B) Summary of transformation efficiency. Twenty-two independent lines were examined in detail, and 19 were found to be transformed as verified by antibiotic selection and PCR. (C) Relative expression of asRuBPCase (lower) and RuBPCase (upper) in as-lines (as1–as4) and wt, as determined by real-time qPCR. The as-lines showed expression of the as-transcript in different amounts and had reduced amounts of RuBPCase transcripts in comparison with the amounts measured in wt plants (r = −0.549, P = 0.0326). Transformation clearly resulted in differential silencing of the endogenous RuBPCase transcripts.

Microarray analysis

Results of the oligonucleotide microarray analysis are summarized in Fig. 3 and the complete list of all regulated genes, their expression ratios, and annotations can be obtained from the authors. The hybridization compared E. pubescens-infested plants with plants treated with insecticide to protect from damage by flea beetles. We found a total of 155 genes to be significantly regulated (27% of 568 genes on the microarray). The regulated genes were assigned to different putative functional categories (Fig. 3A). Several genes involved in the biosynthesis of defence-related secondary metabolites were downregulated (tropinone reductase, TRII; phenylalanine ammonia lyase, PAL). A suite of PI genes was strongly upregulated (SaPIN2a, SaPIN2b, pin2, PI-WuSP), which correlated with the observed increase of trypsin–PI activity in flea beetle-infested plants (Fig. 4C). Among defence-related genes, we found an α-dioxygenase (PIOX; cv57.4) to be upregulated, in addition to several pathogenesis-related proteins (PR-1, PR-2, PRP-4, PRP-5, PRP-6, PRp27). Photosynthesis-related genes were generally downregulated (e.g. different subunits of RuBPCase). Genes involved in defence-signalling processes were strongly upregulated in flea beetle-infested plants (octadecanoid pathway: lipoxygenases LOX; alleneoxide synthase AOS; 12-oxophytodienoate reductase opr). The microarray results were verified by RNA gel blot analysis with an SnPIN2b probe and real-time qPCR for the genes SnPIN2b, PR-1 and psbA (Fig. 3B–E).

Figure 4.

Trypsin–PI protein levels (mean ± SE) in MeJA-treated (250 µg/plant) and flea beetle-infested plants (*P < 0.001, ***P < 0.05). (A) Greenhouse-cultivated Solanum nigrum plants, MeJA-treated; (B) field-grown S. nigrum, MeJA-treated; (C) field-grown S. nigrum, flea beetle infested. All treatments significantly increased trypsin–PIs in the leaves of S. nigrum. Trypsin–PI levels correlated with expression of trypsin–PI genes (Fig. 3A–C).

Proteinase inhibitors

We assayed systemic leaves of MeJA- and flea beetle (E. pubescens)-infested plants for their trypsin–PI content. Greenhouse-grown plants showed a significant increase in trypsin–PI activity (Student's t-test, t = −15.239, P = 0.001) after treatment with 250 µg MeJA (Fig. 3A). A similar increase was found in field-grown plants (Fig. 3B; Student's t-test, t = 5.301, P < 0.001), but the constitutive trypsin–PI levels of field plants were higher than those of greenhouse plants at a similar developmental stage. Trypsin–PI levels in field-grown S. nigrum plants were significantly higher in response to attack from the naturally occurring herbivore, E. pubescens (Fig. 3B; Student's t-test, t = 2.397, P = 0.0199). The increase of PIs was similar in MeJA-elicited and flea beetle-infested plants.

Volatile organic compounds

In a field experiment, we collected VOCs emitted from plants in response to attack from phytophagous insects. We used an open-flow trapping system (Fig. 5A), which allowed us to synchronously sample all experimental replicates and to maintain the sampled leaves under physiological conditions. We allowed two herbivore species (L. decemlineata and A. atropos) to feed separately on S. nigrum for 24 h and compared the composition of VOCs trapped with those trapped from uninfested control plants. Emissions of several compounds, ranging from monoterpenes (e.g. 3-carene, β-myrcene, ± limonene) to green-leaf volatiles (cis-3-hexenyl acetate, cis-3-hexen-1-ol) and sesquiterpenes (longifolene, trans-β-caryophyllene), were increased in insect-attacked plants (Fig. 5B).

Figure 5.

Volatile organic compounds (VOCs) released in response to attack from different herbivore species. (A) VOCs were collected from 15 field-grown plants, using an open-flow trapping system. (B) Representative total ion chromatograms of the headspace volatiles eluting from a GC column of an undamaged leaf from an undamaged Solanum nigrum plant (control), a leaf damaged by Leptinotarsa decemlineata, and a leaf damaged by an Acherontia atropos hornworm (from separate plants). The labels represent an unknown monoterpene (1), 3-carene (2), β-myrcene (3), ± limonene (4), unknown compound (5), cis-3-hexenyl acetate (6) cis-3-hexen-1-ol (7), longifolene (8), trans-β-caryophyllene (9), unknown sesquiterpene 1 (10), unknown sesquiterpene 2 (11).


We evaluated the fitness consequences of MeJA-elicitation under field conditions in the plants previously assayed for trypsin-PIs. S. nigrum is extremely plastic in its growth form and is able to adjust its morphology to diverse conditions. In botanically precise terminology, the growth shape varies from decurrent with plagiotropic to fastigiate branching to excurrent with orthotropic branching. Plant size at reproductive maturity varies from 5 cm to > 1 m. We manipulated plant size in a greenhouse experiment by planting S. nigrum into different sized pots and found that the above-ground dry biomass correlated strongly with the number of fruits produced (R2 = 0.665). Pot size was used to manipulate above-ground biomass, but also likely influenced the root architecture and biomass and the available rooting space may fundamentally influence plant fitness. In the field, we did not find morphologically detectable changes in response to a single elicitation with MeJA, nor was the number of flowers produced (data not shown) significantly influenced by elicitation. However, fruit number and seed production differs significantly between induced and uninduced plants. MeJA-treated plants produced ~40% fewer fruits than did control plants (Fig. 6), suggesting that MeJA-elicited responses result in large fitness costs for an individual plant.

Figure 6.

Fitness estimates of Solanum nigrum. (A) Fruit number: Comparison of MeJA-elicited vs. control plants. Fruit numbers were significantly lower in plants that received a single treatment of 250 µg MeJA (Student's t-test, t = 3.256, P = 0.0038, **P < 0.01). One fruit contained 77 ± 1.45 (SEM, n = 33) seeds; hence control plants produced ~130 000 seeds, while elicited plants produced 80 000 seeds per plant. (B) Fruit number is linearly correlated with the above-ground dry mass of greenhouse-grown plants. Different plant sizes were obtained by growing plants in (1) 1-L pots, (2) 2-L pots or (3) 3-L pots.


To facilitate the identification of genes mediating responses to ecological interactions, and to allow for the manipulation of their expression, we present the following tools for the Solanum nigrum expression system: (i) an Agrobacterium-based transformation system; (ii) an oligonucleotide microarray, enriched with ecologically relevant genes; (iii) measures of direct (PIs) and indirect defences (VOC emission) under both field and laboratory conditions; and (iv) measures of Darwinian fitness. These tools have been optimized to analyse responses that are rapidly elicited by ecological interactions. We illustrate their utility by analysing responses to attack by a native herbivore of S. nigrum and compare the elicited responses with those elicited by MeJA treatment. The analysis links changes in transcript abundance with phenotypic changes, which, in turn, are correlated with changes in the fitness of plants grown under field conditions.

An initial survey of the differential gene expression of S. nigrum revealed a large-scale change in the plants’ transcriptome in response to flea beetle attack: 27% of the monitored genes showed a significant change in expression pattern. Genes involved in important primary processes, such as carbon fixation and metabolism, were largely downregulated, whereas defence genes, including genes involved in defence-related signalling, were largely upregulated. The octadecanoid signalling cascade with its key compound, jasmonic acid (JA), is known to play an important role in triggering many of the insect-induced responses of a plant (Blee 2002; Farmer et al. 2003). Consistent with a regulatory role for oxylipins in the elicitation of defence-related genes in S. nigrum, oxylipin biosynthetic enzymes (AOS, allene oxide synthase; LOX, lipoxygenase; OPR, 12-oxophytodienoate reductase) were strongly upregulated, as were the transcripts of the JA-elicited PI defence genes (e.g. SaPIN2a, SaPIN2b, pin2). PIs are known to adversely affect the performance of herbivorous insects (Xu et al. 1996; Rahbe et al. 2003), and the causal associations among JA signalling, PI elicitation, and insect resistance were recently established with Nicotiana attenuata plants in which JA signalling was silenced by expressing LOX-H3 in an antisense orientation (Halitschke & Baldwin 2003). In this study herbivore performance was enhanced on asLOX plants most likely due to attenuated PI and nicotine defence responses.

Several pathogenesis-related (PR) proteins (PR-1; PR-2, β-1,3-glucanase; PR-3, PR-4 both chitinases; PRP6; Prp27), which are known to be elicited by pathogen attack or salicylic acid (SA) treatment, were among the strongest upregulated transcripts. From this, we deduce that herbivore attack to S. nigrum may elicit both SA- and JA-related signalling and the commonly invoked trade-off between systemic resistance of plants to microorganisms and resistance to insect herbivores (Felton et al. 1999; Thaler et al. 2002) may not apply to this plant species. Some transcripts may be co-regulated by both insects and microorganisms. For example, α-dioxygenase (α-DOX; formerly PIOX, pathogen-induced oxygenase), an enzyme that catalyses the conversion of linolenic acid to its 17-hydroperoxy-derivative (Hamberg et al. 1999), is upregulated in S. nigrum in response to flea beetle feeding. In N. tabacum,α-DOX can be induced by bacterial elicitors (Sanz et al. 1998; de Leon et al. 2002), and in N. attenuata, by attack from Manduca sexta larvae (Hermsmeier et al. 2001). Although the biological function of α-DOX remains unclear, its transcriptional behaviour illustrates the likely interactions of pathogen and herbivore signalling under natural conditions.

Adaptive responses are likely to be the result of complex interacting signal networks rather than single signal cascades (Reymond & Farmer 1998; Genoud et al. 2001), and transcription factors may play an important role in coordinating the responses from these many signal cascades. In plants attacked by flea beetles, a transcription factor of the WRKY family (WRKY3) was strongly upregulated, as it was shown to be in N. attenuata attacked by M. sexta larvae (Hui et al. 2003). WRKY transcription factors occur in large gene families and are known to regulate numerous stress-related genes, including those responsive to pathogens and wounding (Eulgem et al. 2000). Transcription factors may coordinate large-scale patterns of transcriptional changes and deserve more attention in the regulation of environmental responses.

The coordination of transcriptional responses to flea beetle attack extends beyond the upregulation of stress-responsive transcripts to include a coordinated downregulation of growth-related transcripts, which is most clearly seen in the negative correlation between the expression of photosynthetic and defence genes (Hermsmeier et al. 2001; Schittko et al. 2001). The herbivore-induced suppression of RuBPCase transcripts (pDH64.7, RUB inas, rbcL) and of additional elements of the photosynthetic machinery (e.g. photosystem proteins: N. tabacum PSI, N. t. PSII precursor, S. nigrum PSII D1) might benefit the plant by redirecting carbon flux toward the production of defences. RuBPCase activase (rca), a stromal protein catalysing the dissociation of inhibitory sugar bisphosphates from uncarbamylated and carbamylated RuBPCase in an ATP-requiring process (Robinson & Portis 1989; Portis 1995), may play an important role in regulating RuBPCase transcripts and perhaps other photosynthetic proteins, and was downregulated in attacked plants. It has been shown that the light-dependent regulation of RuBPCase is controlled by rca (Zhang et al. 2002) but recently, Voelckel & Baldwin (2003) demonstrated that the expression of rca in N. attenuata increased in response to attack by the mirid bug, Tupiocoris notatus. Hence it is possible that rca participates in the herbivore-induced downregulation of photosynthetic metabolism. In addition to the downregulation of photosynthetic genes, genes involved in cell wall metabolism (XTH4, xyloglucan endotransglycosylase), glycolysis (DH63, homologous to Petunia hybrida triosephosphate isomerase; DH123 and cGap, both homologous to N. tabacum glyceraldehyde-3-phosphate dehydrogenase), and nitrogen metabolism (nir, nitrate reductase; GOGAT, glutamine oxoglutarate aminotransferase, etc.) were also downregulated. These alterations suggest many hypotheses about the regulation of primary metabolism in response to herbivore attack and require additional work.

Even though only a few of the sequences on this ‘Solanaceous’ microarray were designed from S. nigrum-specific sequences, it is clear from the verifications of selected array responses by RNA gel blot analysis and real-time qPCR (Fig. 3B–E) that the microarray provided valuable information about differential gene expression in S. nigrum. Several studies have established the utility of using sequence information of related species for monitoring gene expression. Kane et al. (2000) demonstrated that expression patterns derived from oligonucleotide (50-mer) microarrays reflected those from cDNA (~300–400 bp PCR products) microarrays and that 50-mer oligonucleotides are specific, if the target sequences shared 80% or more with the oligonucleotide. Izaguirre et al. (2003) analysed the transcriptome of N. longiflora with a cDNA microarray consisting of N. attenuata sequences and Held et al. (2004) used the same microarray to characterize transcriptional responses in N. clevelandii and N. quadrivalvis. Girke et al. (2000) compared gene expression in developing seeds of Arabidopsis thaliana and Brassica napus with an Arabidopsis-specific microarray. Regardless of whether the array is designed from homologous or heterologous sequences, responses should always be verified before a hypothesis about the functional significance of a gene is pursued. The need for verification is particularly acute when microarray studies suggest a lack of response, as negative results can be caused by small sequence differences between the oligonucleotides and the targeted transcripts. When a microarray produces signals, the spotted oligonucleotides will likely function as probes for screening S. nigrum cDNA libraries, thus facilitating the verification procedure. Once a transcriptional response is verified, the next step in an ecological analysis is to determine if the response correlates with a change in phenotype, not only in the laboratory, but also in the field.

The hypothesis from the microarray analysis that flea beetle attack elicited JA signalling and thereby increased PI production was supported by the observations with both field- and greenhouse-grown plants that beetle attack and MeJA treatment significantly increased PI levels. The absolute levels of PI activity were higher in field-grown plants (Fig. 4), suggesting that plants growing in the rough-and-tumble of the natural world may be partially induced in comparison to the coddled plants grown in the glasshouse, which excludes normal solar UV-B radiation. PIs are probably elicited by exposure to UV-B, as shown by Izaguirre et al. (2003), who used phenotypic measures and microarrays to compare transcriptional patterns in N. longiflora elicited by UV-B with the pattern elicited by M. sexta herbivory. In this study, UV-B exposure not only increased PI transcripts and activity, but also downregulated photosynthesis-related transcripts in a manner similar to herbivore-elicited responses, suggesting that common regulatory elements had been recruited by this pair of abiotic and biotic stressors.

In response to herbivore attack, plants frequently activate indirect defence responses that complement the function of the direct defences (Kessler & Baldwin 2002). By releasing VOCs in response to herbivore attack, plants attract predators and parasitoids of herbivorous insects (Turlings et al. 1990; Kessler & Baldwin 2001; Dicke et al. 2003). When S. nigrum plants were attacked by herbivores from two different Orders of insects in the field (Fig. 6), the composition and the quantities of the VOCs trapped from the headspace of leaves differed significantly from those of unattacked plants. The fact that an induced response was observed in field-grown plants is significant, particularly in light of recent reports on Zea mays demonstrating just how significantly abiotic factors such as soil nutrition, air humidity, temperature, and light can influence the herbivore-induced VOC response (Gouinguene & Turlings 2002; Schmelz et al. 2003). The large and diverse predator community (syrphids, chrysopids, coccinellids and braconid wasps) we observed during our field studies of S. nigrum could respond to these VOC emissions. Whether any of these potential predators actually responds to increased VOCs can be readily tested by adding components of or the entire herbivore-induced volatile blend to plants containing a ‘predator-monitor’ (an herbivorous insect or egg used to score predation events: Kessler & Baldwin 2001).

Whether or not a trait can be formally considered to be a defence depends on whether it increases a plant's Darwinian fitness in environments with aggressors. The best surrogate measures for Darwinian fitness are determined by a plant's life history, but for selfing annual plants such as S. nigrum, lifetime seed production is likely to be an adequate measure. Fruit and seed production were found to be strongly correlated with plant above-ground dry mass (Fig. 6B), and these are important measures for the resource partitioning to growth, reproduction and defence in response to biotic interactions (Givnish 1986; Bazzaz et al. 1987; Mauricio et al. 1993). The reproductive output of plants that have been under strong artificial selection for particular yield components will frequently be strongly buffered from variations in canopy size. For example, tomato plants are strongly buffered from leaf area loss from herbivores and do not decrease fruit number in response to herbivore attack (Thaler 1999). We found that jasmonate elicitation of S. nigrum did not reduce plant size, a conclusion Thaler et al. (1996) had reached with regard to jasmonate treatment of tomato. Moreover, we found that jasmonate elicitation in S. nigrum significantly reduced lifetime fruit and seed production, although this is not observed in tomato (Thaler 1999), which likely reflects differences between agronomic and native species in their selective history. Jasmonate elicitation is known to significantly decrease lifetime seed production in N. attenuata (Baldwin 1998), and a large fraction of these fitness costs can be attributed to the induced production of PIs (Zavala et al. 2004). The underlying mechanisms of S. nigrum's fitness costs remain to be explored.

Direct manipulation of the genetic basis of an observed response is the most powerful means of falsifying functional hypotheses available to biologists. Agrobacterium-based transformation systems provide the means to produce stably transformed lines in which particular genes are silenced or over-expressed. Independently transformed lines are typically variable in their phenotypes due to the random insertion of the transgene into the genome, which, in turn, leads to differences in transcriptional activity (‘positional’ effects). Krügel et al. (2002) demonstrated that N. attenuata lines, transformed with an asLOX construct exhibited up to 71% reduction in wound-induced JA accumulation. This genetically determined phenotypic variation is enormously useful, because it allows the fitness consequences of a trait to be quantitatively analysed. Also, possible pleiotropic effects resulting from single genetic changes must to be considered in the analyses of transgenic lines, and such effects require that multiple independently transformed lines with the same transgene must be examined before the phenotype can be attributed to the expression of the transgene. The S. nigrum asRuBPCase lines showed a clear reduction in RuBPCase transcripts (Fig. 2C), yet they did not reveal differences in their growth phenotype in comparison to wildtype plants in the glasshouse. Whether this lack of growth phenotype persists when these plants are grown under field conditions will be interesting to determine. Experiments in realistic environments may reveal why plants appear to be ‘over-engineered’ with respect to their RuBPCase pools (Quick et al. 1991; Matt et al. 2002). The rapid development of new transformation vectors that allow for more efficient silencing of endogenous genes (RNAi with inverted repeat elements: Waterhouse & Helliwell 2003), will make the process of producing transformants with fully silenced genes more efficient, thereby facilitating the search for phenotypes of plants that are grown in complex environments.

Most of the tools presented in this study have been used only in controlled laboratory experiments. The results obtained from such experiments can be different from those seen in plants growing in nature. For example, although constitutive PI levels were higher in field-grown plants than in greenhouse-grown plants, field-grown plants (Fig. 4) were still inducible. The combination of several stresses often provokes a potentiation of a response, leading to either an increase or a decrease of subsequent responses to the same or other stresses (Zimmerli et al. 2000). Therefore, the plant may recruit similar fundamental cellular reactions in response to various stresses, which may be what is happening in response to UV-B irradiation and herbivory (Izaguirre et al. 2003) or cold and drought stress in comparison to disease resistance (Singh et al. 2002). Experimentation with field-grown plants in which a variety of stresses are factorially manipulated will elucidate both the amount of cross-talk that occurs among environmental responses ans also the fitness consequences of the different selective forces for plants whose ability to respond is selectively impaired.

While molecular biology has provided the ability to manipulate the expression of individual genes, understanding the functional consequences of these manipulations will require additional ecological tools to dissect the complex interplay of selective forces that all organisms face in nature. Multivariate and path analyses provide a means to examine correlations among the different levels of analysis that occur from gene expression to the formation of a phenotype with a given Darwinian fitness. Ecologists have been successful in using such approaches to evaluate complex correlations and to recognize interrelationships among the biotic and abiotic factors that structure ecosystems. The challenge remains to find a way to incorporate the powerful manipulative and descriptive molecular methods into this ‘big picture’ analysis so as to harvest the fruits of the molecular revolution.


We thank Susan Kutschbach and Thomas Hahn for invaluable assistance in microarray processing and data analysis, Dr Jörg Perner for assistance in determining beetles, Dr Wolfgang Weisser for assistance in determining aphids, Dr Tamara Krügel for valuable support and advice and the Max-Planck-Gesellschaft and the Deutsche Forschungsgemeinschaft (project BA2138/1–1) for financial support.

The authors’ interests are in understanding the chemical and molecular basis of plant–insect interactions The authors have differing and complementary expertise spanning plant molecular biology, biochemistry, analytical chemistry, insect and plant ecology.