Arabidopsis IQD1, a novel calmodulin-binding nuclear protein, stimulates glucosinolate accumulation and plant defense

Authors


(fax 001 530 752 9659; e-mail sabel@ucdavis.edu).

Summary

Glucosinolates are a class of secondary metabolites with important roles in plant defense and human nutrition. To uncover regulatory mechanisms of glucosinolate production, we screened Arabidopsis thaliana T-DNA activation-tagged lines and identified a high-glucosinolate mutant caused by overexpression of IQD1 (At3g09710). A series of gain- and loss-of-function IQD1 alleles in different accessions correlates with increased and decreased glucosinolate levels, respectively. IQD1 encodes a novel protein that contains putative nuclear localization signals and several motifs known to mediate calmodulin binding, which are arranged in a plant-specific segment of 67 amino acids, called the IQ67 domain. We demonstrate that an IQD1-GFP fusion protein is targeted to the cell nucleus and that recombinant IQD1 binds to calmodulin in a Ca2+-dependent fashion. Analysis of steady-state messenger RNA levels of glucosinolate pathway genes indicates that IQD1 affects expression of multiple genes with roles in glucosinolate metabolism. Histochemical analysis of tissue-specific IQD1::GUS expression reveals IQD1 promoter activity mainly in vascular tissues of all organs, consistent with the expression patterns of several glucosinolate-related genes. Interestingly, overexpression of IQD1 reduces insect herbivory, which we demonstrated in dual-choice assays with the generalist phloem-feeding green peach aphid (Myzus persicae), and in weight-gain assays with the cabbage looper (Trichoplusia ni), a generalist-chewing lepidopteran. As IQD1 is induced by mechanical stimuli, we propose IQD1 to be novel nuclear factor that integrates intracellular Ca2+ signals to fine-tune glucosinolate accumulation in response to biotic challenge.

Introduction

Glucosinolates are a small but diverse class of defense-related secondary metabolites that are synthesized mainly by cruciferous species such as nutritionally important Brassica crops and the reference plant Arabidopsis thaliana (Fahey et al., 2001; Wittstock and Halkier, 2002). The common, glucosinolate-defining glycone structure is derived from select protein amino acids and is composed of a sulfonated oxime and a β-thioglucose residue. Extensive modification of the amino acid side chain is responsible for the chemical diversity of glucosinolates, which are stable and hydrophilic molecules and stored in vacuoles of most plant tissues. Upon tissue damage, glucosinolates are rapidly hydrolyzed by β-thioglucosidases (myrosinases) to glucose and unstable intermediates that spontaneously rearrange to various reactive products, including isothiocyanates, thiocyanates, nitriles, oxazolidine-2-thiones, or epithioalkanes (Fahey et al., 2001; Wittstock and Halkier, 2002). Many of these derivative compounds are biologically active and have been implicated in plant defense against pathogens and herbivores (Kliebenstein et al., 2002b; Lambrix et al., 2001; Tierens et al., 2001), as allelochemicals in mediating plant–insect interactions (Ratzka et al., 2002), or as dietary inducers of detoxification enzymes that favorably modify carcinogen metabolism in mammals (Mithen et al., 2000; Talalay and Fahey, 2001). The wide range of biological activities of glucosinolate breakdown products as well as the intricate intersection of indole glucosinolate metabolism and auxin homeostasis (Ljung et al., 2002) has raised interest in glucosinolate biosynthesis and its regulation.

Biosynthesis of glucosinolates proceeds in three phases via (i) incremental amino acid side chain elongation; (ii) formation of the common glycone moiety to produce primary glucosinolates; and (iii) secondary modifications of the side chain to generate the known spectrum of glucosinolate compounds. More than 35 glucosinolates have been identified in Arabidopsis, which are largely derived from methionine, tryptophan, and phenylalanine (Kliebenstein et al., 2001b; Reichelt et al., 2002). Identification of glucosinolate pathway genes in Arabidopsis followed by biochemical studies of recombinant enzymes confirmed the tripartite biosynthetic concept (Mikkelsen et al., 2002; Wittstock and Halkier, 2002). The principle enzymes of the core pathway that converts amino acids or chain-elongated homoamino acids to primary glucosinolates have been cloned and biochemically characterized (Grubb et al., 2004; Mikkelsen et al., 2004; Piotrowski et al., 2004; Wittstock and Halkier, 2002). Although several genes involved in amino acid side chain elongation and modification have been identified (Kliebenstein et al., 2001c; Kroymann et al., 2001; de Quiros et al., 2000), a number of peripheral steps in glucosinolate biosynthesis remain to be studied at the genetic and biochemical levels.

While significant progress has been made in understanding the biochemistry and genetics of glucosinolate metabolism, little is known about the regulation of glucosinolate accumulation during plant development and in response to environmental stimuli. During the life cycle of A. thaliana, glucosinolate content and composition change significantly among organs and tissues. Developing roots, leaves, inflorescences, and siliques are major sites of glucosinolate synthesis or accumulation, which is consistent with a proposed function of these compounds in plant defense (Brown et al., 2003; Petersen et al., 2002). A number of studies showed that a range of biotic and abiotic factors modulate leaf and seed glucosinolate profiles. These include pathogen challenge, herbivore damage and mechanical wounding, mineral nutrition, drought stress, and light conditions (Bennett and Wallsgrove, 1994; Bones and Rossiter, 1996; Hirai et al., 2004). Treatment of Arabidopsis with pathogen elicitors or jasmonate (JA) induces differential accumulation of glucosinolates, and the analysis of signaling mutants indicates that JA- and salicylate (SA)-mediated defense pathways regulate the production of various glucosinolates in Arabidopsis (Brader et al., 2001; Kliebenstein et al., 2002a; Mikkelsen et al., 2003). The underlying regulatory mechanisms responsible for altered glucosinolate production are largely unknown and have only recently become a target of investigation. While some of the natural variation in glucosinolate profiles among various Arabidopsis accessions can be explained by the allelic status of certain biosynthetic genes, other differences in glucosinolate content and composition are likely to be caused by the activity of regulatory loci (Kliebenstein et al., 2001a,b,c, 2002a; Celenza et al., 2005; Kim et al., 2004).

In an attempt to elucidate glucosinolate biosynthesis and its regulation in Arabidopsis, we undertook a molecular genetic approach (Gross et al., 2000; Grubb et al., 2002) and developed a novel screening procedure for T-DNA-induced mutations that cause altered glucosinolate accumulation (Wang et al., 2002). Here, we describe the isolation of a T-DNA activation-tagged mutant with a high-glucosinolate chemotype followed by the functional characterization of IQD1 as a modifier of glucosinolate metabolism. We demonstrate that IQD1 encodes a novel nuclear-localized protein that binds to calmodulin in a Ca2+-dependent fashion and positively regulates glucosinolate accumulation. Interestingly, we observed that IQD1 overexpression leads to increased resistance against herbivory by generalist chewing and phloem-feeding insects. As IQD1 is induced by mechanical stimuli, the prospect arises that IQD1 integrates Ca2+-dependent signaling to augment and fine-tune glucosinolate accumulation in response to biotic challenge.

Results

Screen for activation-tagged loci conferring a high-glucosinolate chemotype

We previously developed a bioassay for leaf glucosinolate content in Arabidopsis that is based on the induction of quinone reductase (QR) activity in cultured murine hepatoma cells by isothiocyanates produced from glucosinolates during plant tissue destruction (Gross et al., 2000). We observed a positive correlation between leaf QR inducer potency and leaf glucosinolate content for A. thaliana and demonstrated the feasibility to identify chemically induced mutants with altered glucosinolate profiles (Grubb et al., 2002). We further showed that excised leaf disks can substitute for cell-free leaf extracts as a source of glucosinolate-derived isothiocyanates, which facilitates high through-put screening of insertional mutant collections (Wang et al., 2002). Here, we screened duplicate leaf disks of 16 500 T4 progeny derived from 5500 independent T-DNA activation-tagged lines (Weigel et al., 2000) for altered leaf QR inducer potency. We identified 404 putative mutants that were retested in the T5 generation. Leaf material of T5 progeny that inherited changes in QR inducer potency were analyzed by high-performance liquid chromatography (HPLC) for glucosinolate profiles. We identified 12 glucosinolate content and composition (gcc) lines with an altered glucosinolate chemotype that passed the three-stage screening procedure. Line gcc7, the focus of this study, accumulates about twice as much methionine- and tryptophan-derived glucosinolates as wild type (Col-0) control seedlings (Table 1; Figure S1).

Table 1.  Glucosinolate content of wild-type accessions, T-DNA insertion lines, and transgenic IQD1 overexpression and antisense lines
LineaS3S4S8T43-OHIM4IM1IMTotal
  1. S3, 3-methylsulfinylpropyl; S4, 4-methylsulfinylbutyl; S8, 8-methylsulfinyloctyl; T4, 4-methylthiobutyl; 3-OH, 3-hydroxypropyl; IM, indole-3-ylmethyl; 1IM, 1-methoxyindole-3-ylmethyl; 4IM, 4-methoxyindole-3-ylmethyl; ND, not detectable.

  2. aThe desulfoglucosinolate fraction extracted from shoots of 14-day-old seedlings was analyzed by HPLC.

  3. bGlucosinolate concentrations are given in nmol mg−1 FW ± SD (n = 3–4 extractions of individual plants).

WT (Col-0)0.11 ± 0.02b0.62 ± 0.180.06 ± 0.060.05 ± 0.02ND0.05 ± 0.020.04 ± 0.020.05 ± 0.010.98 ± 0.33
gcc70.16 ± 0.011.26 ± 0.150.10 ± 0.020.11 ± 0.03ND0.13 ± 0.020.08 ± 0.020.07 ± 0.011.91 ± 0.26
Sense0.18 ± 0.021.42 ± 0.150.06 ± 0.050.12 ± 0.01ND0.05 ± 0.010.14 ± 0.030.09 ± 0.042.06 ± 0.31
Salk_280.11 ± 0.020.58 ± 0.160.05 ± 0.030.05 ± 0.02ND0.06 ± 0.020.06 ± 0.020.06 ± 0.020.97 ± 0.27
WT (Ws-0)1.25 ± 0.13ND0.11 ± 0.01NDND0.12 ± 0.020.14 ± 0.010.03 ± 0.011.65 ± 0.18
ABB80.64 ± 0.12ND0.07 ± 0.02NDND0.03 ± 0.010.10 ± 0.010.01 ± 0.000.85 ± 0.16
Antisense0.76 ± 0.19ND0.06 ± 0.03NDND0.06 ± 0.020.04 ± 0.01ND0.92 ± 0.25
WT (Ler)0.26 ± 0.08NDNDND2.86 ± 0.600.32 ± 0.110.11 ± 0.030.09 ± 0.043.64 ± 0.86
GT69350.16 ± 0.07NDNDND0.60 ± 0.370.12 ± 0.030.08 ± 0.040.01 ± 0.020.97 ± 0.53
SGT168200.11 ± 0.05NDNDND1.34 ± 0.150.11 ± 0.020.09 ± 0.060.01 ± 0.031.77 ± 0.31

Genetic analysis of line gcc7

The T-DNA vector pSKI015 used for activation mutagenesis contains four repeats of the enhancer region of the constitutively active 35S promoter of cauliflower mosaic virus (CaMV) and the BAR gene for phosphinotricin (glufosinate) resistance in the T-DNA insert. Introduction of this T-DNA into the genome can cause increased gene expression near the integration site in an orientation-independent manner (Weigel et al., 2000). Analysis of the T5 and T6 generation of line gcc7 demonstrated glufosinate resistance of all progeny and a heritable high-glucosinolate chemotype. To test whether the high-glucosinolate chemotype is linked to the T-DNA insert, we backcrossed line gcc7 to wild type and analyzed the F1 and selfed F2 progeny for glufosinate resistance and glucosinolate accumulation. All analyzed F1 progeny (12 plants) were glufosinate-resistant and accumulated about 50% more total glucosinolates than wild-type plants (data not shown). The backcross F2 generation segregated in a 3:1 ratio for both glufosinate resistance (96 resistant plants:32 susceptible plants; chi-square test, P = 1.0) and the high-glucosinolate chemotype (10 plants with high glucosinolate content: two plants with wild-type levels; one-tailed binominal test, P = 0.75, P = 0.39). The T-DNA was present in all mutant plants but absent in the wild-type segregants. These data are consistent with the presence of a single T-DNA insert conferring glufosinate resistance and a semi-dominant high-glucosinolate chemotype. We further performed Southern analysis on genomic DNA prepared from gcc7 plants and digested with the restriction enzymes available for plasmid rescue in the activation-tagging vector. Hybridization of a T-DNA-specific gene probe to only one fragment of each DNA digest revealed a single T-DNA insertion event for line gcc7; there was no indication for rearrangement of the T-DNA insert (data not shown). The dominance of the high-glucosinolate chemotype indicated that the mutation most likely resulted from increased gene expression driven by multiple 35S enhancers in the single T-DNA insert.

Identification and cloning of IQD1

We recovered genomic DNA fragments flanking the T-DNA in line gcc7 by plasmid rescue, which we analyzed by restriction mapping and DNA sequencing of 1.4 kb across the left T-DNA border. Figure 1(a) shows the structure of the T-DNA insertion site in the genomic region covered by BAC F11F8 of chromosome 3. The T-DNA inserted into putative exon 4 of gene At3g09720, which is predicted to encode an RNA helicase. Several gene candidates encoding proteins of unknown or hypothetical function are located adjacent to the left and right border of the T-DNA insert. Analysis of messenger RNA (mRNA) expression by semiquantitative RT-PCR demonstrated a robust overexpression of At3g09710, located 3.8 kb next to the left T-DNA border, in line gcc7 relative to the wild-type control (Figure 1a). On the contrary, steady-state mRNA expression levels of the distal genes, At3g09700 and At3g09730, were not affected in line gcc7 and indistinguishable from wild-type levels, while expression of At3g09720 was only modestly reduced. These data suggest that the high-glucosinolate chemotype of line gcc7 is caused by constitutive overexpression of At3g09710, which is hereafter referred to as IQD1 (IQ-DOMAIN 1, see below).

Figure 1.

Identification of IQD1 (At3g09710) and verification of its role in glucosinolate accumulation.
(a) T-DNA insertion site of line gcc7 and mRNA expression analysis. The upper panel depicts the original T-DNA insertion in gene At3g09710 and the organization of neighboring genes. The lower panel shows results from semiquantitative RT-PCR performed with gene-specific primers on RNA extracted from 4-week-old wild type (WT) and gcc7 plants. ACTIN1 primers were used to control for cDNA input (ACT). Numbers of PCR cycles are given above the gels. The region corresponding to the amplified cDNA is depicted as a solid line above each gene model.
(b) Glucosinolate content of wild type, T-DNA insertion lines, and transgenic IQD1 overexpression (sense) and antisense suppression lines. The T-DNA insertion sites of lines ABB8 (Ws-0), GT6935 (GT, Ler), SGT16820 (SGT, Ler), SALK_042728 (SALK28, Col-0), SALK_008887 (SALK87, Col-0), and gcc7 (Col-0) are indicated above the gene models (upper panel). Glucosinolates were extracted from shoots of 14-day-old plants grown in agar medium and analyzed by HPLC. Mean contents (±SD, n = 3–4 individual plants) of methionine-derived (black bars) and tryptophan-derived (gray bars) glucosinolates are given for each line and its respective wild-type accession (lower panel). Insets show analysis of IQD1 and ACTIN (ACT) mRNA expression of select lines by semiquantitative RT-PCR.

Verification of IQD1 function in glucosinolate accumulation

To evaluate a proposed function of IQD1 in glucosinolate accumulation, we analyzed five publicly available insertional mutations and generated transgenic recapitulation as well as antisense suppression lines. As expected, T-DNA insertions into At3g09720 (line SALK_042728) does not affect glucosinolate accumulation (Figure 1b; Table 1), which is also true for line SALK_008887 (data not shown). By contrast, glucosinolate concentrations are significantly reduced for IQD1 promoter insertion lines SGT16820 and GT6935 (Ler accession), as well as for line ABB8 (Ws-0 accession), which harbors a T-DNA insert in the fourth exon of the predicted IQD1 open reading frame, and for an antisense suppression line (Figure 1b; Table 1). When compared with the respective wild-type accessions, methionine-derived glucosinolates are decreased by as much as 75%, whereas tryptophan-derived glucosinolates are reduced to a lesser extent (up to 40%). The reduction in glucosinolate content correlated with reduced (GT6935) or undetectable (ABB8) levels of IQD1 transcripts. Finally, we generated recapitulation lines (Pro35S:IQD1) that accumulate 65–125% more total glucosinolates (Figure 1b; Table 1). This observation supports genetic data indicating that the high-glucosinolate chemotype of line gcc7 is caused by overexpression of IQD1 (At3g09710) and not by gene disruption of At3g09720. Thus, the analysis of loss- and gain-of-function alleles suggests that IQD1 encodes a modifier of glucosinolate accumulation.

Although IQD1-overexpressing plants are smaller than wild type at an early growth stage (Figure 2a) they are very similar to wild type in appearance when mature (data not shown). Loss-of-function IQD1 alleles in the Ler background (GT6935 and SGT16820) are slightly stunted (Figure 2c–f), which is not observed for line ABB8 in the Ws-0 background (Figure 2b; data not shown). When compared with wild-type seeds, total glucosinolate content of gcc7 seeds is about twofold higher, indicating that the high-glucosinolate chemotype of gcc7 is independent of the developmental stage and caused by increased glucosinolate production (data not shown).

Figure 2.

Overall morphology of plants expressing IQD1 gain- and loss-of-function alleles.
Growth of lines gcc7 (a), ABB8 (b), SGT16820 (c–e), and GT6935 (f) was compared with growth of wild type (plants on the right in each panel): Col (a), Ws-0 (b), and Ler (c–f). The arrow in (f) points to rosette leaves seriously damaged by herbivory. Bar = 1 cm.
(a) 28 days (long-day),
(b) 21 days (short-day),
(c) 12 days (short-day),
(d) 18 days (long-day, MS agar), and
(e, f) 35 days (long-day).

Predicted properties of the encoded IQD1 protein

Conceptual translation of the IQD1 open reading frame predicts expression of a basic protein (pI of 10.4) with a molecular mass of 50.5 kDa (454 amino acid residues). We constructed a full-length IQD1 cDNA (AY827468) and confirmed the predicted IQD1 (At3g09710) gene model (see Experimental procedures). A blast search with the primary structure of IQD1 as the query identified 11 closely related (E-value: <e−20) hypothetical proteins: five in Arabidopsis, one in sunflower (SF16; Dudareva et al., 1994), and five in rice. These proteins share similar molecular masses (33.6–59.6 kDa) and isoelectric points (9.2–10.9). In addition to their basic nature (Arg/Lys content of 13–20%), IQD1-related proteins are Ser/Thr-rich (14–21%). The amino acid and nucleotide sequence identity among the 12 genes varies from 22 to 60% and 40 to 67%, respectively, indicating that IQD1-like genes are quite divergent. blast searches with lower stringency suggest the presence of 33 and at least 28 related genes in the genomes of A. thaliana and Oryza sativa, respectively (M. Levy and S. Abel, unpublished data).

As shown in Figure 3, sequence similarity among IQD1-related polypeptides is primarily limited to their central portion, which contains three segments (I–III) of high amino acid sequence conservation. Secondary structure predictions suggest that the three segments contribute to α-helical folds. The 12 IQD1-like proteins are predicted to consist primarily of random coil structures (46–63%), α-helices (30–45%), and to a lesser extent of β strands (5–12%). Multiple repeats of three different classes of recognition motifs for calmodulin interaction are present in segment I and separated by invariant spacing. This segment of 67 amino acid residues contains two to three copies of the IQ motif, IQxxxRGxxxR, or of its more relaxed version [I,L,V]QxxxRxxxx[R,K]. The IQ motif was initially thought to mediate Ca2+-independent retention of calmodulin, although the motif was subsequently found to be present in some proteins that interact with calmodulin in a Ca2+-dependent manner (Bähler and Rhoads, 2002). Segment I also contains two repeats of the 1-8-14 motif as well as three repeats of the 1-5-10 motif at conserved positions, which partially overlap with the IQ motifs (Figure 3). The 1-5-10 and 1-8-14 motifs mediate Ca2+-dependent calmodulin binding and were named based on the conserved spacing of hydrophobic residues (Choi et al., 2002; Rhoads and Friedberg, 1997). Another hallmark of IQD1-related proteins is the presence of clusters of basic amino acid residues that satisfy structural characteristics of three classes of nuclear localization signals. As indicated in Figure 3, the N- and C-terminal regions of IQD1-related proteins contain several basic clusters that conform to the SV40-type, MATα2-type, and bipartite type of nuclear localization signals (Abel and Theologis, 1995). Thus, the prospect arises that IQD1 binds to calmodulin and functions in the cell nucleus.

Figure 3.

Amino acid sequence alignment of IQD1 and related proteins.
Aligned are closely related proteins from Arabidopsis thaliana (AtIQD1–AtIQD6), Oryza sativa (OsIQD1–OsIQD5), and Helianthus annuus (SF16). The three segments of high amino acid sequence conservation are underlined and indicated by roman numerals. Amino acid residues conserved in at least 11 proteins are highlighted in black, and residues conserved in at least nine polypeptides are shaded (gray). Segment I contains repeats of three motifs with roles in calmodulin interaction, the IQ motif (amino acid residues highlighted in red), the 1-8-14 motif (green asterisks), and the 1-5-10 motifs (brown asterisks). Clusters of basic amino acid residues that conform to the SV40-type, MATα2-type, and bipartite type of nuclear localization signals are shaded in blue. AtIQD1 (At3g09710), AtIQD2 (At5g03040), AtIQD3 (At3g52290), AtIQD4 (At2g26410), AtIQD5 (At3g22190), AtIQD6 (At2g26180), OsIQD1 (Os05m00240), OsIQD2 (Os01m06082), OsIQD3 (Os03m04309), OsIQD4 (Os03m05627), OsIQD5 (Os01m00929), and SF16 (CAA52782).

Recombinant IQD1 binds to calmodulin

To test whether recombinant IQD1 binds to calmodulin, we expressed an epitope tagged T7-IQD1 fusion protein in Escherichia coli. Crude extract from induced bacterial cells expressing T7-IQD1 was used in pull-down assays with bovine calmodulin-agarose beads in the presence of Ca2+ or in the absence (+ EGTA) of Ca2+. After co-incubation of the calmodulin-agarose beads with bacterial extract, the beads were repeatedly washed, and bound proteins were eluted by suspension in sample loading buffer. Proteins of all fractions were separated by electrophoresis, transferred to a membrane, and probed with T7-Tag antibody to detect the T7-IQD1 fusion protein. As shown in Figure 4, T7-IQD1 co-sedimented with calmodulin-agarose beads only in the presence of Ca2+ but did not co-sediment when the incubation mix and wash buffer were supplemented with EGTA. We used recombinant T7-IAA3 as a negative control (Abel and Theologis, 1995), which did not bind to calmodulin-agarose beads (data not shown). Thus, our data suggest Ca2+-dependent, but T7 epitope-independent, calmodulin binding of T7-IQD1 in vitro.

Figure 4.

Ca2+-dependent in vitro interaction of IQD1 and calmodulin.
Calmodulin-agarose beads were incubated in the presence of Ca2+ or absence of Ca2+ (+EGTA) with soluble proteins prepared from induced bacterial cultures expressing a T7-tagged IQD1 protein and treated as described in Experimental procedures. Proteins of the total bacterial extract, the supernatant fraction, the entire pellet (beads) fraction, and of the last wash were resolved by SDS-PAGE, transferred to a membrane, and probed with a HRP-conjugated T7-Tag monoclonal antibody.

IQD1-GFP is targeted to the cell nucleus

To study the subcellular localization of IQD1 in vivo, we generated transgenic Arabidopsis lines that express GFP and an IQD1-GFP fusion protein under control of the constitutive 35S CaMV promoter. As expected, histochemical analysis revealed localization of the authentic GFP protein in the cytosol and cell nucleus, whereas the chimeric IQD1-GFP fusion protein accumulated mainly in the nucleus and was largely excluded from the cytosol (Figure 5). Thus, our data demonstrate the potential for nuclear localization of an IQD1-GFP fusion protein when stably expressed in Arabidopsis and therefore suggest the cell nucleus as a cellular compartment of IQD1 function.

Figure 5.

Subcellular localization of GFP and IQD1-GFP in root cells of transgenic Arabidopsis plants.
Root tips of transgenic seedlings were incubated with propidium iodide and imaged as described in Experimental procedures. GFP-generated fluorescence (GFP), propidium iodide-generated fluorescence (PI), and merged propidium iodide/GFP images of GFP and IQD1-GFP transgenic seedlings are shown.

IQD1 modulates expression of glucosinolate pathway genes

We used reverse transcriptase (RT)-mediated PCR to monitor steady-state mRNA levels of several genes encoding enzymes involved in glucosinolate metabolism (Figure 6). Several of these genes are members of closely related families and are expressed at low levels (e.g., CYP79s), which precluded RNA blot analysis. Both semiquantitative and real-time RT-PCR were performed on RNA isolated from shoots of 14-day-old seedlings, and relative mRNA levels were compared between Columbia wild type, line gcc7, and the Pro35S:IQD1 recapitulation line. Relative to wild type, steady-state mRNA levels for genes with roles in Trp-derived glucosinolate biosynthesis (CYP79B2, CYP79B3, and CYP83B1) were elevated in line gcc7 (2- to 16-fold), most notably for CYP79B3 and CYP83B1 (Figure 6a,c). However, expression levels of genes encoding enzymes related to the biosynthesis of Met-derived glucosinolates (CYP79F1 and CYP79F2) and glucosinolate degradation (myrosinase-encoding TGG1; Husebye et al., 2002) were appreciably decreased (10–25% of wild-type levels), whereas expression of UGT74B1 was only twofold reduced (Figure 6a,c). The same tendency of altered gene expression was observed in the Pro35S:IQD1 recapitulation line (Figure 6b,c). Thus, our data suggest that overexpression of IQD1 deregulates at least a subset of glucosinolate pathway genes.

Figure 6.

Relative steady-state mRNA levels of select glucosinolate pathway genes.
Semiquantitative and real-time RT-PCR was performed with total RNA prepared from shoots of 2-week-old plants and with gene-specific primers for glucosinolate pathway genes. Each PCR method was conducted twice with two independently grown sets of plants, and similar results were obtained in both duplicate analyses. For the semiquantitative RT-PCR, 25 cycles and 27 cycles were used to monitor expression of ACTIN1 and glucosinolate pathway genes, respectively.
(a) Comparison of gene expression by semiquantitative RT-PCR between wild type (Col-0) and gcc7 plants.
(b) Comparison of gene expression by semiquantitative RT-PCR between wild type (Col-0) and the Pro35S:IQD1 recapitulation line (sense).
(c) Analysis of transcript levels by real-time RT-PCR in line gcc7 (gray bars) and the Pro35S:IQD1 recapitulation line (black bars) relative to wild type. The log2 ratio of mutant (gcc7 or sense)/wild-type (Col) transcript levels are given. The difference between duplicates was within 15%.

Tissue-specific expression of IQD1

Tissue-specific expression of IQD1 was tested by semiquantitative RT-PCR with RNA extracted from 21 to 28-day-old plants and was found to be present in the all major organs tested (Figure 7a). In our search for insertional loss-of-function IQD1 alleles, we identified one gene trap line of the CSH collection, GT6935, in which the bacterial β-glucuronidase gene uidA under control of a minimal promoter is inserted 101 bp upstream of the predicted translational start codon of IQD1. Therefore, although the transcriptional start site of IQD1 has not been determined, line GT6935 likely reports authentic IQD1 promoter activity. Histochemical analysis of GUS activity in up to 4-week-old GT6935 seedlings revealed reporter gene expression exclusively in the vascular bundles of hypocotyls, leaves, stems, flowers, and roots (Figure 7b–v). In 6-day-old germinating seedlings, GUS expression is detected in the shoot apical meristem, at the branching zone of the vascular tissue beneath the meristem (Figure 7b,c), and in the vasculature of the hypocotyls and roots (Figure 7d). At day 9 (Figure 7e–g) and day 14 (Figure 7h,i), GUS expression is clearly detectable in the vascular tissues of rosette leaves and of the root system; weaker expression is observed in the cotyledons. The highest level of IQD1 promoter activity was observed in fully expanded rosette leaves and shoot meristems of non-flowering seedlings. GUS expression in roots was excluded from the elongation zone (Figure 7g). ProIQD1:GUS is also expressed at the receptacle of the flower and silique, in developing seeds, seed funicles, the pistil and stamen filaments (Figure 7j–n). During the transition from vegetative to reproductive development, ProIQD1:GUS expression gradually diminishes in rosette leaves but is clearly detectable in the developing inflorescence and flower buds; expression in the root system appears unaffected (Figure 7o–v). It is interesting to note that the observed tissue-specific pattern of IQD1 promoter activity is strikingly similar to the expression patterns reported for several genes that encode enzymes of the glucosinolate synthesis pathway, CYP79F1, CYP79F2 (Reintanz et al., 2001; Tantikanjana et al., 2001), CYP79B2 (Mikkelsen et al., 2000), or UGT74B1 (Grubb et al., 2004).

Figure 7.

Analysis of tissue-specific IQD1 expression.
(a) Analysis of steady-state IQD1 mRNA levels in wild-type plants by semiquantitative RT-PCR using gene-specific primers for IQD1 (27 cycles) or ACTIN1 (ACT, 25 cycles) and total RNA from roots (Ro), flowers (Fl), stems (St), siliques (Si), inflorescence stem (In), and whole shoots (Sh).
(b–v) Histochemical GUS staining of gene trap line GT6935 in seedlings at 6 days (b–d), 9 days (e–g), 14 days (h, i), in the inflorescence of seedlings between 21 and 32 days (j–n), and during the transition of seedlings to flowering between 12 and 18 days (o–v).

IQD1 overexpression deters insect herbivory

We occasionally observed that loss-of-function iqd1 mutants are more susceptible to insect attack when grown in soil (see Figure 2f), which prompted us to analyze the effect of altered IQD1 gene expression on insect herbivory in more detail. Herbivory-induced tissue damage causes glucosinolate degradation and release of bioactive products that may act as attractants and oviposition stimulants for specialist or as repellents against generalist insects (Giamoustaris and Mithen, 1995; Kliebenstein et al., 2002b; Lambrix et al., 2001). In a first set of experiments, we examined and compared the weight gain of newly hatched larvae of the cabbage looper (Trichoplusia ni), a generalist lepidopteran whose larvae feed on cruciferous and many other plant species (Shorey et al., 1962). We observed that T. ni larvae developing on line gcc7 were significantly smaller and gained 25% less fresh weight than larvae on wild-type (Col-0) plants, which were appreciably more damaged by herbivory than gcc7 plants (Figure 8a). We did not observe a significant weight difference between T. ni larvae grown on ABB8 and control plants (Ws-0). However, the fresh weight of larvae developing on both Ws lines was about 35% higher than the weight of larvae developing on Col-0 wild-type plants (Figure 8a), which may be explained by the different glucosinolate composition and the different types of glucosinolate hydrolysis products formed in both accessions. It has been reported that nitrile-producing A. thaliana accessions such as Ws-0 are more susceptible to T. ni herbivory than isothiocyanate-producing lines such as Col-0 (Lambrix et al., 2001).

Figure 8.

IQD1 overexpression reduces insect herbivory.
(a) Weight gain assays of Trichoplusia ni larvae on 5-week-old Arabidopsis plants grown in short days (8 h light). The left panel shows the mean and SE of fresh weight gain of newly hatched cabbage looper larvae feeding for 10 days on Arabidopsis lines Col-0 and gcc7 (Col-0, n = 39; gcc7, n = 40; t-test, P < 0.0001), as well as on Ws-0 and ABB8 (n = 24 for each line; Mann–Whitney rank-sum test, P = 0.2725). The right panel shows representative samples of larvae that were left for 10 days on wild type (Col-0) and gcc7 plants, and the lower panel shows a representative sample of both plant lines after 10 days of T. ni herbivory.
(b) Dual-choice assays of Myzus persicae nymph deposition. The graphs show the preference of green peach aphids (represented by percentage) for either wild-type Col-0 or line gcc7 (upper graph; n = 85 choice pairs, P = 0.0044), and for either wild-type Ws-0 or line ABB8 (lower graph; n = 67 choice pairs, P = 0.025) in dual-choice assays (see Experimental procedures). The calculated P-values of the one-tailed binominal tests showed significant evidence for repellence by the high-glucosinolate line in both choice pairs, i.e., gcc7 (Col-0) and Ws-0 wild type.

In a second experiment, we investigated the nymph deposition preference of the green peach aphid (Myzus persicae), a generalist phloem-feeding pest with hosts in over 40 plant families (Pollard, 1972). In dual-choice assays, we placed individual late instars of alate aphids, which mature into winged adults, onto the surface of agar plates containing two seedlings of different glucosinolate chemotypes (either wild type Col-0 and gcc7, or wild type Ws-0 and ABB8) and scored host plant preference through an analysis of nymph deposition. Our data show that green peach aphids consistently avoided the line with the higher glucosinolate content. For each of the two choice arrangements, about two-thirds of the aphids preferred the lower glucosinolate line for reproduction, i.e., wild type Col-0 and IQD1 loss-of-function line ABB8 (Figure 8b). In summary, our data demonstrate that overexpression of IQD1 significantly reduces herbivory by chewing and phloem-feeding insects.

IQD1 is induced by mechanical stimuli and aphid infestation

Treatment of Arabidopsis plants with JA, SA, and ethylene leads to the accumulation of specific glucosinolates (Brader et al., 2001; Cipollini et al., 2004; Mikkelsen et al., 2003). As overexpression of IQD1 causes elevated glucosinolate levels, we examined whether IQD1 mRNA expression is responsive to JA, SA, the ethylene precursor ACC, or IAA. None of the plant hormones tested appreciably altered steady-state IQD1 mRNA levels when externally applied (Figure 9a), which was confirmed by quantitative RT-PCR for plants treated at different developmental stages (data not shown). Induction of PR1 by SA and of PDF1.2 by JA and ethylene confirmed effective hormone treatment (Figure 9a). As shown in Figure 9(b), IQD1 mRNA levels are also unaffected in various mutants defective in hormone synthesis or signaling (JA –jar1, fad3-2 fad7-2 fad8; SA –NahG, npr1, cpr1; and ethylene –ein2), suggesting that IQD1 expression is independent of these hormones. However, we noticed that mock spraying of seedlings with the solvent control or water alone caused a modest, about twofold elevation of IQD1 mRNA levels, which was also observed after mechanical wounding of leaves, or after infestation of seedlings with green peach aphids (Figure 9c). Increased IQD1 expression as a result of aphid herbivory is likely a consequence of both mechanical stimulation and responses to chemical cues generated during phloem feeding.

Figure 9.

Analysis of steady-state IQD1 mRNA expression levels in response to hormones and mechanical stimuli.
Semiquantitative and quantitative RT-PCR reactions were performed with total RNA prepared from shoots and gene-specific primers. Each PCR method was conducted at least twice with two independently grown sets of plants, and similar results were obtained in both duplicate analyses. Numbers of cycles are indicated above the gels for semiquantitative RT-PCR, and the results of quantitative RT-PCR are shown only when significant differences in IQD1 mRNA expression were measured.
(a) Hormone treatments. Soil-grown plants (3-week old, short-days) were mock-treated (spraying) or treated for 4 and 8 h with the indicated hormones as described in Experimental procedures. Relative mRNA levels of IQD1, PR1 (SA-inducible), PDF1.2 (JA and ethylene-inducible), and ACTIN (ACT) gene expression are shown for semiquantitative RT-PCR analyses.
(b) Hormone mutants. Comparison of IQD1 mRNA expression between wild type (Col) and various hormone-related mutants (2-week old, MS agar) by semiquantitative RT-PCR.
(c) Mechanical and biotic stimuli. Comparison of IQD1 mRNA expression between untreated, wounded (4 h), and aphid-infested (3 days), agar-grown plants (2-week old) by semiquantitative and quantitative (-fold change) RT-PCR. The difference between qRT-PCR duplicates was within 15%.

Discussion

In contrast to the enzymatic reactions in glucosinolate metabolism, very little is known about the regulation of the glucosinolate pathway and its underlying mechanisms during plant development and in response to environmental cues (Mikkelsen et al., 2002; Wittstock and Halkier, 2002). For example, the myb transcription factor ATR1 activates genes with roles in Trp biosynthesis and Trp secondary metabolism, including indolic glucosinolate homeostasis (Celenza et al., 2005). Likewise, the TU8 mutation that affects accumulation of various glucosinolates (Haughn et al., 1991) has recently been cloned and shown to be a new allele of TERMINAL FLOWER2, which encodes a chromodomain protein with broad regulatory functions (Kim et al., 2004). We previously developed a high-throughput bioassay for glucosinolate-derived isothiocyanates to identify genetic modifiers of glucosinolate production (Grubb et al., 2002; Wang et al., 2002). Here, we screened a collection of T-DNA activation-tagged lines (Weigel et al., 2000) for plants with altered leaf glucosinolate content, isolated several Arabidopsis lines with a heritable high-glucosinolate chemotype and functionally characterized IQD1 (At3g09710) as a positive regulator of glucosinolate production and plant defense against herbivores.

IQD1 is a modifier of glucosinolate accumulation

We used a series of constructed and publicly available loss- and gain-of-function IQD1 alleles in three different accessions (Col-0, Ler, and Ws-0) to probe IQD1 gene function in glucosinolate metabolism. Our data show that IQD1 mRNA expression correlates positively with glucosinolate accumulation in A. thaliana (Table 1; Figure 1). The original T-DNA activation-tagged line, gcc7, and generated recapitulation lines that constitutively express a Pro35S:IQD1 transgene in the Col-0 background, accumulate about twice as much total glucosinolates as the wild-type control. In contrast, several lines with T-DNA insertions in the predicted promoter and coding region of IQD1/At3g09710 (Ler, Ws-0) as well as Pro35S:IQD1 antisense suppression lines (Ws-0) showed a significant reduction of total glucosinolate content (by 45–70%). Two insertional alleles of At3g09720, the predicted RNA helicase-encoding gene that is disrupted by the T-DNA activation tag in line gcc7, do not affect glucosinolate levels, which is consistent with the observed dominance of the gcc7 glucosinolate chemotype. Thus, our genetic analysis demonstrates that IQD1 functions as a positive regulator of glucosinolate accumulation.

Such a role for IQD1 is further indicated by its expression patterns during plant development (Figure 7). The observed tissue-specific expression of IQD1 conspicuously overlaps with the expression patterns of glucosinolate pathway genes for which promoter activity has been histochemically analyzed in Arabidopsis: CYP79F1, CYP79F2 (Reintanz et al., 2001; Tantikanjana et al., 2001), CYP79B2 (Mikkelsen et al., 2000), and UGT74B1 (Grubb et al., 2004). Furthermore, the temporal shift of tissue-specific IQD1 expression during the transition to flowering and the pronounced IQD1 promoter activity in vascular tissues and flower stalks correspond to major sites of glucosinolate biosynthesis or accumulation in Arabidopsis. Consistent with a proposed function of glucosinolates in plant defense against herbivores and pathogens, the reproductive organs, including developing inflorescences, flowers, siliques and seeds, accumulate the highest concentrations of glucosinolates, followed by young leaves, the root system, and fully expanded leaves (Brown et al., 2003; Du and Halkier, 1998; Koroleva et al., 2000; Petersen et al., 2002). Tracer studies demonstrated de novo synthesis of glucosinolates in reproductive organs (Du and Halkier, 1998), phloem transport from mature leaves to inflorescences and fruits (Brudenell et al., 1999; Chen et al., 2001), and glucosinolate turnover during seed germination and early seedling development (Petersen et al., 2002). In summary, IQD1 gene expression correlates with glucosinolate accumulation levels and with prominent sites of glucosinolate metabolism in Arabidopsis.

IQD1 encodes a novel calmodulin-binding nuclear protein

We demonstrated Ca2+-dependent binding of recombinant IQD1 to bovine calmodulin in vitro (Figure 4) and nuclear localization of an IQD1-GFP fusion protein in transgenic Arabidopsis plants (Figure 5). These experiments were prompted by the presence of putative calmodulin interaction motifs as well as nuclear localization signals in IQD1 and closely related proteins, which are the only structural features to suggest a potential biochemical role for IQD1 (Figure 3). Three major classes of calmodulin recruitment motifs are currently known: two related motifs, termed 1-5-10 and 1-8-14, are typified by their spacing of hydrophobic and basic amino acid residues and bind calmodulin in a Ca2+-dependent fashion, whereas the IQ-motif mediates association with calmodulin in a Ca2+-independent manner. However, not all characterized calmodulin-binding domains contain these features (Bähler and Rhoads, 2002; Choi et al., 2002; Hoeflich and Ikura, 2002). Conserved segment I of 67 amino acid residues, which is shared by IQD1 and closely related proteins, contains two or three copies of each recruitment motif at invariant positions (Figure 3). This compact arrangement of multiple calmodulin-binding motifs, referred to as the IQ67 domain, is the defining feature of more distantly related proteins of unknown biochemical functions, which are members of relatively large plant-specific families (e.g., 33 proteins in Arabidopsis; M. Levy and S. Abel, unpublished data). The majority of IQD1-related proteins contain putative nuclear localization signals, although not at conserved positions, which may be explained by their high frequency of basic amino acid residues. In addition to IQD1-like proteins, the IQ motif occurs in several plant protein families, including myosins, cyclic NMP-gated ion channels, and a class of calmodulin-binding transcription factors involved in ethylene responses. However, the structural context of the respective IQ motif(s) is different from the IQ67 domain (Bähler and Rhoads, 2002; Bouche et al., 2002; Köhler et al., 1999; Reddy and Day, 2001; Yang and Poovaiah, 2002). Binding of recombinant IQD1 to calmodulin, possibly via multiple calmodulin recruitment motifs, suggests that the biochemical activity of IQD1 is regulated by Ca2+-calmodulin in vivo. However, the precise Ca2+ sensors that interact with IQD1 remain to be identified.

Calcium signaling in plants is complex and utilizes a large repertoire of sensor and regulatory target proteins. Several classes of sensor proteins bind to Ca2+ via a helix-loop-helix fold known as the EF-hand motif. This motif is found in more than 250 Arabidopsis proteins (Day et al., 2002; Yang and Poovaiah, 2003), including six typical calmodulins and 50 calmodulin-like proteins that differ significantly in sequence and number of EF-hand motifs (McCormack and Braam, 2003). About 200 calmodulin-binding target proteins are currently known in plants, a number that is expected to rise (Reddy and Reddy, 2004). Plant calmodulins have been identified in different subcellular locations, including the cytosol, nucleus, peroxisome, or extracellular matrix, and a diverse set of calmodulin-binding proteins is involved in a variety of cellular processes such as cytoskeleton organization, regulation of gene expression, ion transport, disease resistance, or stress responses (Yang and Poovaiah, 2003). For example, a subset of Arabidopsis TCH (touch) genes that are induced by a variety of mechanical stimuli encode calmodulin and calmodulin-related proteins, which likely mediate some plant responses to the environment (Braam et al., 1996, 1997; Sistrunk et al., 1994). There is increasing evidence for the generation of nucleus-specific Ca2+-signatures in plant cells (Pauly et al., 2000; Xiong et al., 2004) and for a potential regulatory role of calmodulin and other Ca2+ sensor proteins in nuclear processes such as transcription or gene silencing (Anandalakshmi et al., 2000; Bouche et al., 2002; Perruc et al., 2004; Szymanski et al., 1996; Yang and Poovaiah, 2002; Yoo et al., 2005). Although IQD1 and related proteins do not contain known DNA- or RNA-binding motifs, the high isoelectric point and Ser/Thr content of IQD1-like proteins, which are reminiscent of certain splicing factors (Chaudhary et al., 1991), suggest that they may associate with nucleic acids and regulate transcriptional or post-transcriptional processes of gene expression.

A regulatory role of IQD1 in glucosinolate metabolism is further supported by altered steady-state mRNA levels of multiple glucosinolate pathway genes in IQD1 gain-of-function and loss-of-function mutants. Transcript levels of genes involved in indole glucosinolate biosynthesis, CYP79B2, CYP79B3, and CYP83B1 are elevated in T-DNA activation-tagged and Pro35S:IQD1 recapitulation lines (Figure 6). This observation is consistent with the accumulation of indole glucosinolates to higher than wild-type levels in IQD1-overexpressing lines. Although overexpression of IQD1 causes enhanced accumulation of aliphatic glucosinolates, genes with roles in methionine-derived glucosinolate biosynthesis (CYP79F1 and CYP79F2) and glucosinolate degradation (TGG1) are significantly repressed. This unexpected result may be explained by negative feedback regulation in aliphatic glucosinolate biosynthesis. Such a feedback control is suggested by an unusually high expression of ProCYP79F1:GUS in homozygous versus heterozygous cyp79F1 gene trap lines (Tantikanjana et al., 2001). Elevated accumulation of methionine-derived glucosinolates may also result from metabolic cross-talk between the aliphatic and indolyl branch of glucosinolate biosynthesis, which is indicated by compensatory glucosinolate production in cyp79F1/bus1 and cyp83A1/ref2 mutants to maintain some degree of glucosinolate homeostasis (Hemm et al., 2003; Reintanz et al., 2001).

Role of IQD1 in plant defense

Increased IQD1 gene expression and glucosinolate accumulation correlate with enhanced resistance to generalist chewing and phloem-feeding insects, as demonstrated by cabbage looper (T. ni) herbivory and green peach aphid (M. persicae) reproduction (Figure 8). In crucifers, glucosinolates and their breakdown products are thought to be important components of the defense arsenal against herbivores and pathogens (Rask et al., 2000). Recent studies showed that T. ni herbivory is profoundly deterred by high glucosinolate levels and that glucosinolate-derived isothiocyanates are stronger feeding repellants than the corresponding nitrile products (Jander et al., 2001; Kliebenstein et al., 2002b; Lambrix et al., 2001). Our data are consistent with these observations. Wild-type Col-0, an isothiocyante-producing accession (Lambrix et al., 2001), is less susceptible to T. ni herbivory than the Ws-0 accession, which presumably produces nitriles based on our previous observation that Ws-0 leaf extracts were nearly inactive in a bioassay for leaf glucosinolate-derived isothiocyanates (Grubb et al., 2002). Furthermore, as evidenced by 25% reduced larvae weight gain, the high-glucosinolate line, gcc7, is significantly more resistant to T. ni herbivory than the wild type (Col-0). It is interesting to note that about a similar degree of resistance to caterpillars of the generalist herbivore Spodoptera exigua was reported (approximately 25% reduced insect growth) when Arabidopsis plants (Col-0) were treated with JA, which led to the accumulation of twofold higher total glucosinolate levels (Cipollini et al., 2004). Using a second Arabidopsis–herbivore interaction system, we observed that the generalist green peach aphid avoids Arabidopsis lines with higher glucosinolate content in dual-choice assays of nymph deposition. Consistent with our data, Ellis et al. (2002) recently reported that JA treatment of Arabidopsis plants (Col-0) retarded growth of M. persicae populations, which was likely caused by elevated glucosinolate content as specific glucosinolates are known to inhibit the reproduction of green peach aphids (Fraybould and Moyes, 2001).

We are left to speculate on the biochemical function of IQD1. Inducible plant defenses against chewing and phloem-feeding insects involve pathogenesis-related responses that recruit Ca2+-calmodulin, JA, SA, and ethylene signaling modules (Moran and Thompson, 2001; Moran et al., 2002; Reymond et al., 2004). Although IQD1 mRNA expression appears to be independent of plant hormone signaling, we noticed that mechanical stimuli such as touch (mock spraying), wounding, and aphid infestation cause a moderate increase of IQD1 transcripts (Figure 9). Enhanced IQD1 expression due to aphid feeding likely reflects a combination of mechanical stimuli and more complex responses to salivary components (Moran et al., 2002). There is increasing evidence for integrated and temporally controlled cross-talk between plant response pathways involved in wounding, herbivory, and necrotic pathogen infection (Maleck and Dietrich, 1999; Moran and Thompson, 2001; Reymond et al., 2000, 2004). The IQ67 domain and its arrangement of multiple calmodulin-interacting motifs are shared by members of relatively large protein families in Arabidopsis, rice and likely all vascular plants (M. Levy and S. Abel, unpublished data). In view of the ubiquitous but unknown functions of IQD1-like proteins and the restriction of glucosinolate biosynthesis to select species, the prospect arises that Arabidopsis IQD1 has broader roles in plant defense. We propose that IQD1 and possibly related calmodulin-interacting proteins participate in the decoding of Ca2+-signatures elicited by biotic and abiotic challenges. IQD1 may integrate early wound- and pathogen/elicitor-induced changes in cytoplasmic Ca2+ concentrations (Blume et al., 2000; Grant et al., 2000) to stimulate and fine-tune a wide array of coordinated defense responses, including the upregulation of glucosinolate biosynthesis (Brader et al., 2001; Mikkelsen et al., 2003; Tierens et al., 2001). Elucidation of the biochemical activities of IQD1 may provide an important impetus for understanding the roles of Ca2+-calmodulin signaling in plant defense.

Experimental procedures

Plant material and growth conditions

Wild-type A. thaliana (L.) Heynh. accessions Columbia (Col-0), Landsberg erecta (Ler) and Wassilewskija (Ws-0), pools of T-DNA activation-tagged lines (Col-0) (Weigel et al., 2000), T-DNA insertion lines SALK_008887 and SALK_042728 (Col-0), and mutant lines npr1, ein2, jar1, and nahG (Col-0) were obtained from the ABRC (Columbus, OH, USA). Gene trap line GT6935 (Ler) was provided by J. Simorowski (CSHL, Cold Spring Harbor, NY, USA) (Sundaresan et al., 1995), T-DNA insertion line ABB8 (Ws) by INRA (Versailles, France), Ds transposon-tagged line SGT16820 (Ler) (Parinov et al., 1999) by V. Sundaresan (University of California, Davis, CA, USA), the fad3-2 fad7-2 fad8-1 (Col-0) triple mutant (McConn and Browse, 1996) by J. Browse (Washington State University, Pullman, WA, USA), and the cpr1 (Col-0) mutant (Bowling et al., 1994) by X. Dong (Duke University, Durham, NC, USA). All insertion lines were backcrossed once to the respective wild-type accessions prior to phenotypic analysis. Plants were grown in soil at 22°C and 70% relative humidity under illumination with fluorescent and incandescent light at a photo fluence rate of approximately 120 μmol m−2 sec−1. For plant growth in sterile conditions, seeds were surface-sterilized for 10 min in 1.5% (w/v) sodium hypochlorite and placed on solid medium containing 8 g l−1 phytagar, 30 g l−1 sucrose, 2.15 g l−1 (0.5×) Murashige-Skoog (MS) salts, pH 5.6 (Murashige and Skoog, 1962), and 1× MS vitamin mix (Sigma, St Louis, MO, USA). After 2–3 days of stratification at 4°C, plants were grown at 24°C at a photon fluence rate of approximately 160 μmol m−2 sec−1 for 18 h (long days) or 8 h (short days) per day. For selection of transgenic plants, the agar medium was supplemented with 25 mg ml−1 kanamycin (Sigma), or seedlings grown in soil were sprayed with 1% (v/v) Basta herbicide (Finale; Farnam Companies, Phoenix, AZ, USA). For hormone treatments, soil-grown or agar-grown plants were sprayed with 250 μm MeJA, 5 mm SA, 2.5 mm ACC (Sigma) and with 0.025% ethanol as the solvent control. For external mechanical stimuli, plants were sprayed with water or wounded with forceps a few times in each leaf. For herbivory challenge, five green peach aphid nymphs (M. persicae) were placed onto each plant grown in MS agar medium.

Screen for glucosinolate accumulation mutants

T4 progeny of T-DNA activation-tagged lines (Weigel et al., 2000) were grown in sterile conditions as previously described (Grubb et al., 2002). After 14–15 days of growth, two single leaf disks (3 mm diameter) were excised from opposite leaves per seedling. Leaf disks were individually tested for leaf QR inducer potency using a colorimetric bioassay of QR activity in cultured murine hepatoma cells, which was optimized for direct analysis of leaf disks (Wang et al., 2002). Fifty-five pools of 100 lines each were screened by analyzing 300 T4 progeny per pool. Putative mutants were re-evaluated by assaying eight leaf disks obtained from four T5 progeny. Glucosinolate profiles of T5 lines that showed heritable changes in leaf QR inducer potency were analyzed by HPLC.

HPLC analysis of desulfoglucosinolates

For glucosinolate analysis of plant tissues, desulfoglucosinolate extracts were prepared according to Wang et al. (2002) with minor modifications. Leaves of 2-week-old plants grown on agar medium were harvested and extracted by boiling for 10 min in 1 ml H2O supplemented with 1 μl of 100 μm benzylglucosinolate (Merck, Darmstadt, Germany) as internal standard. The supernatant was collected and the plant residue washed with 1 ml H2O. The combined extract was applied to a DEAE Sephadex A-25 (60 mg) column that was equilibrated with 2 ml of 0.5 m pyridine acetate and washed with 1 ml H2O. Subsequently, the column was washed with 3 ml of 0.02 m pyridine acetate and 2 ml H2O. The glucosinolates were converted into desulfoglucosinolates by overnight treatment with 75 μl of 0.1% (1.4 U) aryl sulfatase (Sigma). Desulfoglucosinolates were eluted with 1 ml H2O, and aliquots of 100 μl were analyzed by HPLC using a Shimadzu VP Liquid Chromatograph with a dual wavelength spectrophotometer (Shimadzu, Tokyo, Japan). Samples were separated at 45°C on a Waters Spherisorb C18 column (Waters, Milford, MA, USA; 150 × 4.6 μm i.d.; 5 μm particle size) using acetonitrile and water at a flow rate of 1 ml min−1. The column was developed by isocratic elution with 1.5% acetonitrile (5 min) followed by a linear gradient to 20% acetonitrile (15 min) and isocratic elution with 20% acetonitrile (10 min). Absorbance was detected at 226 and 280 nm. Desulfoglucosinolate concentrations were calculated based on published response factors (Haughn et al., 1991; Petersen et al., 2002) and the internal benzylglucosinolate standard.

Plasmid rescue and identification of the T-DNA insertion site

Genomic DNA of T5 progeny was isolated using the Nucleon Phytopure Plant DNA Extraction Kit (Amersham Pharmacia, Piscataway, NJ, USA), restricted with BamHI, ligated (T4 DNA ligase), and transformed into E. coli DH5α for plasmid rescue of the T-DNA insert (Weigel et al., 2000). Recovered plant genomic DNA fragments adjacent to the left border of the T-DNA insert were analyzed by DNA sequencing using the T-DNA left border primer JL202 (5′-CATTTTATAATAACGCTGCGGACATCTAC-3′) and primer T7 (5′-TAATACGACTCACTATAGGGCGAAT-3′) complementary to the T7 promoter (Weigel et al., 2000).

mRNA detection by reverse transcriptase-mediated PCR

Total RNA was extracted from rosette leaves of 14- to 28-day-old seedlings (as indicated) using the RNeasy Kit (Qiagen, Valencia, CA, USA) followed by treatment with RNAse-free DNAse (Qiagen) to remove genomic DNA contamination. Five micrograms of total RNA was reverse-transcribed with the First-Strand cDNA Synthesis SSII Kit (Invitrogen, Carlsbad, CA, USA) according to the manufacturer's instructions. The resulting cDNA was diluted fivefold, and 1 μl was used for semiquantitative RT-PCR reactions, which were performed in a total volume of 25 μl in PCR buffer (Eppendorf, Hamburg, Germany) containing 250 μm dNTPs, 1.5 mm MgCl2, 10 pmol of forward and reverse primers and 2.5 U of Taq DNA polymerase (Eppendorf). PCR conditions were 2 min at 94°C followed by 15–35 cycles (as indicated) of 94°C for 15 sec, 60°C for 15 sec, and 72°C for 1 min. Ten microliters of the PCR was analyzed by gel electrophoresis on a 1% agarose gel and visualized by ethidium bromide staining. ACTIN1-specific primers were used as a control for the amount of template cDNA. RT-PCR reactions were performed at least twice with independent RNA preparations.

For real-time RT-PCR analyses, the resulting cDNA was diluted 500-fold and 5 μl was used as template. Reactions were performed as previously described (Grubb et al., 2004) using an ABI 7300 thermocycler and the SYBR Green PCR Master Mix (Applied Biosystems, Foster City, CA, USA) according to the manufacturer's instructions. The results were analyzed with SDS software (version 2.1) and default settings were used for baseline and threshold cycle values (Applied Biosystems). All mRNA levels were calculated from threshold cycle values and as relative to wild-type controls and normalized with respect to ACTIN 1 transcript levels (Applied Biosystems user bulletin no. 2, ABI Prism 7700 Sequence Detection System).

Gene-specific amplimers for semiquantitative and quantitative RT-PCR are listed in Table S1.

Expression of recombinant IQD1 and calmodulin binding assay

A full-length cDNA fragment encoding the predicted IQD1 protein was generated by RT-PCR using gene-specific primers cIQD1-F (5′-ATGGTTAAAAAAGCGAAATG-3′) and cIQD1-R (5′-GACTTACACTGCACAAATGAATCAA-3′). The PCR fragment was subcloned into the TA cloning vector pGEMT (Promega, Madison, WI, USA), creating plasmid pGEMT(IQD1). After verifying the authenticity of the insert by DNA sequencing, the IQD1 cDNA fragment was mobilized into the EcoRI site of vector pET21 (Novagen, Madison, WI, USA), which provides an N-terminal T7-epitope tag, MASMTGGQQMG. The orientation of the insert in plasmid pET21(T7-IQD1) was verified by restriction enzyme analysis. The recombinant T7-IQD1 protein was expressed in E. coli M15[pREP4] (Qiagen) at 37°C for 4 h by induction with 1 mm isopropyl β-d-thiogalactopyranoside (IPTG). Bacterial cells were harvested and sonicated in 1x T7-Tag washing buffer (T7-Tag Affinity Purification Kit; Novagen). After centrifugation, the supernatant was aliquoted and stored at −20°C.

Aliquots of 100 μl of calmodulin-agarose beads (phosphodiesterase -3′:5′-cyclic nucleotide activator from bovine brain; Sigma), pre-equilibrated with 1x T7 washing buffer, were mixed with 500 μl of bacterial supernatant (2 μμl−1 protein) supplemented with 1 mm CaCl2 or 5 mm EGTA and incubated for 1 h at 4°C under gentle shaking. Calmodulin-agarose beads were sedimented by centrifugation and washed four times with 500 μl of 1x T7 washing buffer, followed by a final wash with 100 μl of the same solution. The bound proteins were eluted by boiling the beads for 2 min in 100 μl of 4x SDS sample buffer. Proteins of the total extract, the initial supernatant, the last wash, and the pellet fraction were resolved on 10% (w/v) SDS-polyacrylamide gels (Laemmli, 1970). Proteins were transferred to polyvinylidene difluoride membranes (Osmonics, Westborough, MA, USA) in 15.6 mm Tris, 120 mm glycine, 20% (v/v) methanol and 0.1% (w/v) SDS) for 60 min at constant 60 V. Membranes were washed with 1x TTBS buffer [50 mm Tris-HCl, pH 8.0, 150 mm NaCl, 0.05% (v/v) Tween-20] and blocked with 5% (w/v) non-fat milk in 1x TTBS buffer for 1 h at 25°C. Blots were probed with a 5000-fold dilution of T7-Tag monoclonal antibody conjugated with horse radish peroxidase (Novagen) for 16 h at 4°C. After washing the membranes with 1x TTBS buffer for 3 × 15 min, signals were detected using an enhanced chemiluminescence system (SuperSignal-West Femto Maximum Kit; Pierce, Rockford, IL, USA).

Construction of transgenic Arabidopsis plants

To generate IQD1 overexpression (recapitulation) and antisense transgenic lines, the XbaI-NotI fragment of plasmid pET21(T7-IQD1) was mobilized into pATC940, a PBI101-derived binary plant transformation vector, to drive expression of T7-IQD1 cDNA, or of its antisense orientation, under control of the CaMV 35S promoter. For transgenic expression of translational IQD1-GFP fusions, the EcoRI-SmaI fragment of pGEMT(IQD1) was subloned into the EcoRI-SpeI site of vector pEGAD (Cutler et al., 2000) to create pEGAD(IQD1-GFP). Plasmids pATC940(senseT7-IQD1), pATC940(antisenseT7-IQD1), pEGAD(IQD1-GFP), and pEGAD were transformed into Agrobacterium tumefaciens (strain GV3101) by electroporation. Transgenic Arabidopsis plants were generated by A. tumefaciens-mediated transformation as described previously (Clough and Bent, 1998). T2 lines showing a segregation ratio of 3:1 for resistance to kanamycin were selected for subsequent analysis.

Histochemical analysis of transgenic plants

Gene trap line GT6935 expresses a promoter-less bacterial β-glucuronidase gene uidA as a transcriptional IQD1 promoter fusion. The uidA insertion site is 101 bp upstream of the predicted translational start codon of IQD1. Histochemical staining for GUS activity was performed essentially according to Jefferson et al. (1987). Sample tissues were fixed in 80% cold acetone for 20 min and incubated for 11–16 h at 37°C in reaction buffer [50 mm Na2HPO4-NaH2PO4, pH 7.0, 0.5 mm K3Fe(CN)6, 0.5 mm K4Fe(CN)6] containing 2 mm 5-bromo-4-chloro-3-indolyl-β-d-glucuronic acid (X-Gluc) as the substrate. Plant pigments were destained with ethanol, and the GUS staining patterns were recorded under a dissecting microscope (Nikon SMZ800, Tokyo, Japan).

Roots of transgenic plants expressing the green fluorescence protein (GFP) reporter protein, pEGAD, or translational fusions of GFP and IQD1, pEGAD(IQD1-GFP), under control of the CaMV 35S promoter were stained with 1 μg ml−1 propidium iodide for 5 min. After transfer to 50 mm Na2HPO4-NaH2PO4, pH 7.0, green (GFP) and red (propidium iodide) fluorescence of root tips was analyzed using a Leica TCS 4D confocal laser microscope (Leica, Wetzlar, Germany). Specimens were viewed using wavelengths specific for each stain, producing two images that were merged using Leica confocal software.

Aphid dual-choice assays

Green peach aphid (M. persicae) colonies were maintained on cabbage seedlings (Brassica oleracea var. capitata) in laboratory conditions (25 ± 5°C, 50 ± 20% relative humidity, 16 h light). Dual-choice assays were performed to study the preference of aphids for transgenic (high-glucosinolate line gcc7; low-glucosinolate line ABB8) or wild type (Col-0; Ws-0) control Arabidopsis plants. For this purpose, aphid nymphs (late instars of alate aphids) were transferred using a fine hair brush (no. 00) and released into the center of a petri dish (100 × 25 mm; Nunc, Rochester, NY, USA) between two 15-day-old test plants of different genotypes that were grown on each half of the dish in 0.8% phytagar medium. After aphid transfer, the petri dishes were returned to the controlled growth chamber, incubated at 20°C (16 h light, 70 ± 5% relative humidity), and examined every 24 h for three successive days. The aphids could easily walk or fly toward the test plants, and the location of the first new-born nymphs was recorded. One-tailed binomial tests were performed to test the hypothesis that the aphids will choose the low- glucosinolate line over the high-glucosinolate line for nymph deposition (Zar, 1999).

Cabbage looper weight gain assay

Eggs of the cabbage looper (T. ni) from a highly inbred population were purchased from Enthopath, Inc. (Easton, PA, USA). One newly hatched larva was transferred with a fine hair brush (no. 00) to a 5-week-old plant of the specified genotype that was grown in soil and short-day condition (8 h light). Individual plants were confined in a thrips-screen cage and returned to the environmentally controlled growth chamber (20°C, 45 ± 5% relative humidity, 8 h light). After 10 days of feeding, the fresh weights of looper larvae were individually determined. t-Tests were performed to compare larvae weight when justified, and Mann–Whitney rank sum tests when the assumptions for parametric tests were violated (Zar, 1999).

Acknowledgements

We thank John Harada and Judy Callis for critical reading of the manuscript, Fred Ausubel and Dan Kliebenstein for discussions, and V. Sundaresan, J. Simorowski, J. Browse, X. Dong, the ABRC, and INRA for Arabidopsis seed stocks. This project was supported by grants to S.A. from the National Research Initiative of the United States Department of Agriculture Cooperative State Research, Education and Extension Service (grant no. 2002-35318-12672) and the UC Davis Cancer Center. Q.W. was supported by a fellowship from the Chinese Scholarship Council and by the National Natural Science Foundation of China (30370974).

Accession numbers: The accession numbers for the genes mentioned in this article are as follows: AY827468 (IQD1/At3g09710 mRNA).

Ancillary