Methyl Transfer in Glucosinolate Biosynthesis Mediated by Indole Glucosinolate O-Methyltransferase 5

Indole glucosinolate O-methyltransferase 5 catalyzes the methyl transfer reaction in the biosynthesis of Trp-derived glucosinolates that are modified at position 1 of the indole ring.

Abstract

Indole glucosinolates (IGs) are plant secondary metabolites that are derived from the amino acid tryptophan. The product of Arabidopsis (Arabidopsis thaliana) IG core biosynthesis, indol-3-ylmethyl glucosinolate (I3M), can be modified by hydroxylation and subsequent methoxylation of the indole ring in position 1 (1-IG modification) or 4 (4-IG modification). Products of the 4-IG modification pathway mediate plant-enemy interactions and are particularly important for Arabidopsis innate immunity. While CYP81Fs encoding cytochrome P450 monooxygenases and IGMTs encoding indole glucosinolate O-methyltransferases have been identified as key genes for IG modification, our knowledge about the IG modification pathways is not complete. In particular, it is unknown which enzyme is responsible for methyl transfer in the 1-IG modification pathway and whether this pathway plays a role in defense, similar to 4-IG modification. Here, we analyze two Arabidopsis transfer DNA insertion lines with targeted metabolomics. We show that biosynthesis of 1-methoxyindol-3-ylmethyl glucosinolate (1MOI3M) from I3M involves the predicted unstable intermediate 1-hydroxyindol-3-ylmethyl glucosinolate (1OHI3M) and that IGMT5, a gene with moderate similarity to previously characterized IGMTs, encodes the methyltransferase that is responsible for the conversion of 1OHI3M to 1MOI3M. Disruption of IGMT5 function increases resistance against the root-knot nematode Meloidogyne javanica and suggests a potential role for the 1-IG modification pathway in Arabidopsis belowground defense.

Glucosinolates are nitrogen- and sulfur-containing secondary metabolites found in Arabidopsis (Arabidopsis thaliana) and related plants. About 132 different glucosinolate structures were scientifically documented by 2011 (Agerbirk and Olsen, 2012). Glucosinolate biosynthesis from amino acids involves conversion of the amino acid to an oxime, addition of a thiol group and subsequent formation of a β-thioglycosidic bond with Glc, and finally sulfation of the oxime group (Sønderby et al., 2010). This reaction sequence results in the formation of the glucosinolate core structure, which then can undergo further modification reactions (for review, see Halkier and Gershenzon, 2006). In Arabidopsis, three classes of these β-d-thioglucoside-N-hydroxysulfates are present: Met-derived aliphatic, Phe-derived benzenic, and Trp-derived indole glucosinolates (IGs).

IGs have multiple functions in defense against a wide range of enemies, including leaf-clipping and phloem-feeding arthropods (Kim and Jander, 2007; de Vos et al., 2008; Kim et al., 2008; Pfalz et al., 2009; Sun et al., 2009), nonhost fungal and nonhost bacterial pathogens (Bednarek et al., 2009; Clay et al., 2009), and fungal pathogens via conversion to phytoalexins (Pedras et al., 2010, 2011). From a nutritional perspective, IGs also are of interest, but very little is known of the effect of those substitutions of the indole moiety that are commonly found in vegetables such as broccoli (Brassica oleracea) and several other cabbages (Brassica spp.; Agerbirk et al., 2009; Latté et al., 2011).

Four IGs differing in the substitutions of the indole moiety are commonly present in Arabidopsis and cabbages. Those are indol-3-ylmethyl (I3M), 1-methoxyindol-3-ylmethyl (1MOI3M), 4-hydroxyindol-3-ylmethyl (4OHI3M), and 4-methoxyindol-3-ylmethyl (4MOI3M) glucosinolates. Biosynthesis of the glucosinolate core structure from Trp results in the formation of I3M, which subsequently can be modified at position 1 or 4 of the indole ring. I3M is the prevailing IG in Arabidopsis leaves and seeds, while 1MOI3M dominates root IG profiles (Kliebenstein et al., 2001a; Petersen et al., 2002; Brown et al., 2003; Pfalz et al., 2011). Data supporting the existence of an additional, supposedly unstable, Arabidopsis IG, 1-hydroxyindol-3-ylmethyl glucosinolate (1OHI3M), was presented recently (Pfalz et al., 2011), but direct evidence for the presence of this metabolite in Arabidopsis is still required.

Key enzymes for the modification of I3M are cytochrome P450 monooxygenases of the CYP81F subfamily and indole glucosinolate O-methyltransferases (IGMTs) of plant O-methyltransferase family 2. CYP81Fs introduce hydroxy groups at specific positions of the indole ring of I3M, and IGMTs convert hydroxy to methoxy groups (Pfalz et al., 2009, 2011). The Arabidopsis genome contains four CYP81F genes. CYP81F1, CYP81F3, and CYP81F4 form a cluster on chromosome 4, while CYP81F2 is located on chromosome 5. In a transient Nicotiana benthamiana expression system, CYP81F1 to CYP81F3 were capable of catalyzing hydroxylation at position 1 and 4 of the indole ring, whereas CYP81F4 hydroxylated only the nitrogen (N) at position 1. Noteworthy, the 1-hydroxylated product, 1OHI3M, was not detected directly in these experiments but could be deduced from the presence of 1MOI3M when CYP81Fs were coexpressed with IGMTs. Analyses of Arabidopsis cyp81f mutants, however, revealed a more restricted enzymatic specificity of CYP81F1 to CYP81F3 and supported only their participation in the position 4 (4-IG) modification pathway (Pfalz et al., 2011).

Four tandemly arranged IGMT genes (IGMT1–IGMT4) cluster on the left arm of chromosome 1. Their gene products share more than 94% identity. IGMT1 and IGMT2 were shown to be capable of converting 4OHI3M and 1OHI3M to 4MOI3M and 1MOI3M, respectively, when coexpressed with the IG core pathway genes and with CYP81Fs in N. benthamiana (Pfalz et al., 2011). However, the Arabidopsis genome contains an additional putative IGMT. This gene, tentatively named IGMT5, is located separate from the other four IGMTs on the right arm of chromosome 1, and its gene product shares only about 70% identity with IGMT1 to IGMT4. Accordingly, IGMT1 to IGMT4 and IGMT5 branched as sister clades in a genealogy of Arabidopsis O-methyltransferases (Kim et al., 2005). Because IGMT5 expression correlated strongly with CYP81F4 expression across tissues and treatments (Pfalz et al., 2011), we speculated that this O-methyltransferase may play a role in 1-IG modification.

Here, we show that IGMT5 indeed participates in the 1-IG modification pathway, based on targeted metabolomics of two Arabidopsis lines with a T-DNA inserted into the IGMT5 promoter. We demonstrate that the proposed, but to date rather elusive, substrate of IGMT5, 1OHI3M, exists. We show that disruption of the IGMT5 function in roots leads to a strong reduction of 1MOI3M abundance, a concomitant accumulation of 1OHI3M, and the formation of previously undetected IG derivatives. Finally, we uncover an unexpected role of IGMT5 in defense against the plant-parasitic root-knot nematode Meloidogyne javanica.

RESULTS

T-DNA Insertions in the IGMT5 Promoter Alter IGMT5 Expression

To test the hypothesis that IGMT5 participates in IG modification, we investigated two independent Arabidopsis T-DNA insertion lines, SALK_151470 (Alonso et al., 2003) and GK_390B09 (Kleinboelting et al., 2012). Both lines are in the Columbia-0 (Col-0) background and have a T-DNA inserted into the promoter region of IGMT5, approximately 800 bp upstream of the IGMT5 start codon in SALK_151470 and approximately 160 bp upstream of the start codon in GK_390B09 (Fig. 1). Mutant plants displayed no visible phenotypic alterations from the Col-0 wild type, and mutant growth rate did not differ from that of the wild type. To test whether the T-DNA insertions influenced IGMT5 expression, we analyzed leaf and root transcripts of mutant and wild-type plants with real-time quantitative PCR (RT-qPCR) in two independent experiments, each with at least three biological replicates per genotype and tissue. Wild-type roots accumulated significantly more IGMT5 transcript than leaves (Figs. 2 and 3). Leaf IGMT5 transcript levels in both mutants were significantly higher than in the wild type, by more than 2 orders of magnitude in SALK_151470 (Fig. 2A) and by about 1 order of magnitude in GK_390B09 (Fig. 3A). Hence, both T-DNA insertion lines overexpressed IGMT5 in their leaves (IGMT5_l-OE). By contrast, the effect of the T-DNAs on root IGMT5 transcript levels differed between insertion lines. While SALK_151470 roots accumulated about 5-fold more IGMT5 transcript than the wild type (Fig. 2A), IGMT5 expression in GK_390B09 roots was strongly reduced, to about 2% of the wild-type level (Fig. 3A). Thus, SALK_151470 overexpressed IGMT5 in roots (IGMT5_R-OE) while GK_390B09 had IGMT5 knocked out in this organ (igmt5_R-KO). Leaf or root transcript levels of the other IGMT genes, IGMT1 to IGMT4, were not substantially affected by the T-DNA insertions in the IGMT5 promoter. There was also no effect on CYP81F4 expression; transcript analysis by RT-qPCR confirmed the previously reported expression pattern (Pfalz et al., 2011), with low CYP81F4 transcript abundance in leaves and high abundance in roots (Figs. 2A and 3A).

Figure 1.

The IGMT5 gene in Arabidopsis. Shown are exons (gray bars), introns (black bars), and the promoter region (black line). Start and stop codons are indicated by vertical black lines. The promoter region is not drawn to scale. The locations of T-DNA insertions in SALK_151470 and GK_390B09 are indicated by Y-shaped lines. Primer positions are indicated by small black triangles.

Figure 2.

Altered expression of IGMT5 in SALK_151470 increases 1MOI3M concentration in leaves and roots. A, RT-qPCR analysis of IGMT and CYP81F4 transcripts using RNA from leaves (L) and roots (R) of wild-type (W; white bars) and SALK_151470 (M; gray bars) plants. Transcript levels were normalized relative to the mRNA levels of UBIQUITIN10. For each genotype and tissue, at least three biological replicates were used (3 ≤ n ≤ 8). Expression values for wild-type leaves were set as 1; error bars show se. Data were analyzed with ANOVA: expression = constant + genotype + tissue + genotype × tissue. Statistical significance for differences in transcript levels is indicated for genotype (G), tissue (T), and the interaction between genotype and tissue (G×T): *, P < 0.05; and ***, P < 0.001. B, IG profiles in the Arabidopsis wild type and SALK_151470. IG profiles were assessed in leaves and roots of Arabidopsis wild-type (white bars) and SALK_151470 mutant (gray bars) plants. *, P < 0.05. Per genotype and tissue, at least seven biological replicates were used (7 ≤ n ≤ 32); error bars show se. Note that we used a different scale for 1MOI3M. FW, Fresh weight.

Figure 3.

Altered expression of IGMT5 in GK_390B09 has tissue-specific effects on leaf and root IG profiles. A, RT-qPCR analysis of IGMT and CYP81F4 transcripts using RNA from leaves (L) and roots (R) of wild-type (W; white bars) and GK_390B09 (M; gray bars) plants. Transcript levels were normalized relative to the mRNA levels of UBIQUITIN10. For each genotype and tissue, three biological replicates were used (n = 3). Expression values for wild-type leaves were set as 1; error bars show se. Data were analyzed with ANOVA: expression = constant + genotype + tissue + genotype × tissue. Statistical significance for differences in transcript levels is indicated for genotype (G), tissue (T), and the interaction between genotype and tissue (G×T): **, P < 0.01; and ***, P < 0.001. B, IG profiles in the Arabidopsis wild type and GK_390B09. IG profiles were assessed in leaves and roots of Arabidopsis wild-type (white bars) and GK_390B09 mutant (gray bars) plants. ***, P < 0.001; and n.d., not detected. Per genotype and tissue, 23 or 24 biological replicates were used; error bars show se. Note that we used a different scale for 1MOI3M root concentration. Note also that the bars underestimate the true quantity of the unstable compound c1. FW, Fresh weight.

T-DNA Insertions in the IGMT5 Promoter Alter IG Composition

The composition of aliphatic glucosinolates in both T-DNA insertion mutants was typical for a Col-0 genetic background, determined largely by a MAM1 allele at the GS-Elong locus and an AOP1 allele at the GS-AOP locus (Magrath et al., 1994; Giamoustaris and Mithen, 1996; Kliebenstein et al., 2001a, 2001b; Kroymann et al., 2001, 2003). Compared with the Col-0 wild type, methylsulfinylalkyl glucosinolates seemed slightly elevated in GK_390B09 roots, but these effects were not statistically significant, except for 6-methylsulfinylhexyl glucosinolate (F_1,30 = 10.51, P < 0.01), an aliphatic glucosinolate present in minor amounts. By contrast, IG profiles displayed substantial alterations compared with the wild type (Figs. 2B and 3B). In SALK_151470 (with the IGMT5_R-OE genotype), there was a 50% increase in 1MOI3M concentration (F_1,19 = 5.24, P < 0.05), while the levels of other IGs were unaltered. In GK_390B09 (igmt5_R-KO), 1MOI3M root concentration was reduced to less than 2% of the wild-type level (F_1,30 = 390.7, P < 10⁻⁸), whereas I3M and 4MOI3M concentrations were elevated significantly (both F_1,30 > 19, P < 10⁻³). Furthermore, igmt5 roots accumulated three additional compounds, termed c1, c2, and c3, which had not been detected previously in glucosinolate extracts. Compounds c1 and c2 had UV light absorption spectra typical for IGs, with a peak around 220 nm and a broad shoulder from 260 to 300 nm (Supplemental Fig. S1). These features also were present in the UV light profile of c3 but were accompanied by a dominant peak at 200 nm, suggesting that an additional, aromatic moiety was present in the c3 molecule. Furthermore, c1 was unstable; its concentration decreased over the time of HPLC analysis (Supplemental Fig. S2), suggesting that intact mutant roots accumulated more of this compound than we were able to detect.

Surprisingly, leaves of GK_390B09 (IGMT5_l-OE) accumulated only approximately 50% of wild-type 1MOI3M (F_1,31 = 70.2, P < 10⁻⁸; Fig. 3B), whereas the 1MOI3M concentration increased in leaves of SALK_151470 (also IGMT5_l-OE). This increase was moderate, approximately 25%, but nonetheless statistically significant (F_1,37 = 4.71, P < 0.05; Fig. 2B). None of the three newly discovered compounds was detectable in SALK_151470 (Fig. 2B), but c3 was present in the leaves of GK_390B09 (Fig. 3B). Leaf quantities of I3M, 4OHI3M, and 4MOI3M did not differ statistically between the mutants and the wild type. Likewise, aliphatic glucosinolate profiles did not differ between mutant and wild-type leaves.

The Newly Discovered Metabolites in igmt5 Roots Are N-Hydroxy I3M and Two Derivatives

The postulated pathway for 1-IG modification, with CYP81F4 converting I3M to 1OHI3M and an IGMT converting 1OHI3M to 1MOI3M (Pfalz et al., 2011), predicts that a block of the responsible IGMT function should result in accumulation of the methyltransferase substrate 1OHI3M and depletion of the reaction product 1MOI3M. We used mass spectrometry (MS), including ion-trap HPLC-MS and high-resolution mass spectrometry (HR-MS), to test these predictions and also to characterize the newly discovered metabolites.

Ion-trap HPLC-MS and tandem mass spectrometry (MS/MS) of desulfated glucosinolates confirmed that the strongly reduced IG in mutant roots was 1MOI3M (Supplemental Fig. S3). Furthermore, HR-MS of intact glucosinolates showed that one of the three newly discovered root compounds, c1, had the m/zof 463.0471 expected for [1OHI3M − H]⁻, within the usual experimental error (Fig. 4; Table I). Importantly, 4OHI3M, whose anion has the same m/z value, eluted significantly earlier than c1 (Supplemental Figs. S1 and S3). Analysis of desulfated glucosinolates by HPLC with high-resolution quadrupole-time of flight MS and with ion-trap MS/MS detection (Supplemental Fig. S3) confirmed this result and provided further evidence: while the desulfo derivative of 4OHI3M did not lose a hydroxyl (OH) radical in ion-trap MS/MS, this was the major loss observed for desulfated c1, consistent with the hydroxyl group being attached to the indole N and, therefore, more susceptible to fragmentation. Therefore, we conclude that c1 corresponded to 1OHI3M.

Figure 4.

HR-MS of root extracts from the wild type and GK_390B09 (igmt5_R-KO). A, Ultra-performance liquid chromatography separation of root extracts from the wild type (Col-0) and mutant. Shown are HSO₄⁻ ion currents, which are characteristic for glucosinolates. Numbers next to peaks indicate retention times in minutes. Vertical arrows indicate peaks that differ between the wild type and mutant; gray, present in the mutant (c2, c1, and c3); black, strongly reduced in the mutant (1MOI3M). B, HR-MS and MS² spectra of c2, c1, and c3, and proposed chemical structures. Numbers indicate observed m/z values. Stars indicate diagnostic ions for glucosinolates present in all three MS² spectra. Horizontal arrows indicate major shifts in m/z caused by substituent loss.

Table I. HR-MS of newly discovered compounds cl, c2, and c3.

Method	Compound			Experimental m/z	Elemental Composition	Theoretical m/z	Error	Error	Interpretation/Remark
	c1	c2	c3
							ppm	D
MS			x	659.1231	[C26H31N2O14S2]−	659.1217	2.2	0.0014	[(OHI3M + C10H12O4) − H]−
		x		625.1000	[C22H29N2O15S2]−	625.1009	1.4	0.0009	[(I3M + Glc) − H]−
	x	x		463.0471	[C16H19N2O10S2]−	463.0481	2.2	0.0010	[OHI3M − H]−
		x		447.0567	[C16H19N2O9S2]−	447.0532	7.8	0.0035	[I3M − H]−
		x		207.0280
	x		x	160.0382	[C9H6O2N]−	160.0399	10.6	0.0017
MS2		x	x	447.0524	[C16H19N2O9S2]−	447.0532	1.8	0.0008	[I3M − H]−
	x		x	311.9240
	x			274.9908	[C6H11O8S2]−	274.9895	4.7	0.0013	Diagnostic for glucosinolatesa
		x		270.0072
	x	x	x	259.0136	[C6H11O9S]−	259.0124	4.6	0.0012	Diagnostic for glucosinolatesa
	x			240.9996	[C6H9O8S]−	241.0018	9.1	0.0022	Diagnostic for glucosinolatesa
	x	x	x	209.8955
		x		199.0284
			x	179.8847
		x		172.0977	[C8H14NO3]−	172.0974	1.7	0.0003
	x			157.0405
		x		130.0878	[C6H12NO2]−	130.0868	7.7	0.0010
	x			116.0480
	x	x	x	96.9600	HSO4−	96.9596	4.1	0.0004	Diagnostic for glucosinolatesa
	x		x	74.9903	[HOC2H2S]−	74.9904	1.3	0.0003	Diagnostic for glucosinolatesb

^aFrom Fabre et al. (2007).

^bFrom Bialecki et al. (2010).

HR-MS² spectra of compounds c2 and c3 displayed several ions diagnostic for glucosinolates (Fig. 4; Table I; Supplemental Fig. S3), confirming that c2 and c3 were glucosinolate derivatives. Compound c2 had an m/z of 625, but ions of m/z 463 and m/z 447 also were present in simple MS. These ions with m/z 463 and m/z 447 matched the expected masses of [1OHI3M − H]⁻ and [I3M − H]⁻. The m/z differences of 162 and 178 corresponded to anhydro-Glc and Glc (or an isomeric hexose), indicating that fragmentation of c2 occurred on both sides of an oxygen involved in a glycoside bond to Glc or another hexose. These data suggested that compound c2 is a glycosylated derivative of 1OHI3M, with the additional hexoside attached to the indole N via an oxygen atom and thus substituting the unstable N-hydroxy group. In the following, we refer to this compound as N-Hex-I3M.

Finally, c3 had a mass of m/z 659, and HR-MS suggested a molecular formula of C₂₆H₃₁N₂O₁₄S₂. HR-MS² spectra of this compound contained the fragment of m/z 447 typical for IGs, in addition to other ions diagnostic for glucosinolates. HR-MS of the corresponding desulfoglucosinolate agreed with the deduced molecular formula (found, 581.1795; calculated for C₂₆H₃₂N₂O₁₁S, 581.1800; Supplemental Fig. S3). From the biochemical context, molecular formula, and UV light spectrum, our working hypothesis for c3 was a structure based on I3M but with an O-C₁₀H₁₃O₄ radical at the indole N. Fragmentation of desulfo-c3 in MS² supported the presence of this substituent at the indole N. Losses corresponding to Glc and anhydro-Glc were observed (Supplemental Fig. S3), meaning that the substituent could not be attached at the thio-Glc moiety (Agerbirk and Olsen, 2011). A loss corresponding to 212 (substituent radical – hydrogen) was observed from the sodium adduct, meaning that the additional mass compared with that of I3M was in the form of a single substituent. Indeed, a known N derivative of I3M, 1MOI3M, likewise fragmented at the indole N in the ion trap, as opposed to 4MOI3M (Supplemental Fig. S3). While 1MOI3M lost an odd electron CH₃O radical, c3 lost the radical minus hydrogen (loss of OC₁₀H₁₂O₄, with mass 212). This MS² result is compatible with an N-O-C connected radical with the ability to form an aldehyde, allowing hydrogen transfer to the indole N. The deduced radical with a mass of 213 and predicted formula O-C₁₀H₁₃O₄ from both MS² and HR-MS of the intact and desulfated glucosinolate c3 would fit a trihydroxy-methoxy-phenylpropane-O moiety as well as a number of other isomers. In conclusion, MS² of desulfo-c3 supported the deduced molecular formula from HR-MS of intact c3, [M − H]⁻ = C₂₆H₃₁O₁₄N₂S₂, and a structure as I3M substituted with an O-C₁₀H₁₃O₄ radical at the indole N. We refer to this compound as N-X-I3M.

The desulfated derivatives of N-Hex-I3M (=c2) and N-X-I3M (=c3) were isolated successfully by preparative HPLC, but in three independent attempts, both were lost during lyophilization for solvent change before NMR. Residual NMR signals did not allow any conclusions regarding the compound identities. As a known desulfoglucosinolate (desulfo 4-methylsulfinylbutyl glucosinolate) used as positive control was stable during the procedure, and as a number of desulfated IG derivatives had been isolated previously at essentially identical conditions (Agerbirk et al., 2001; Agerbirk and Olsen, 2011), losses were attributed to a particular instability of desulfo N-Hex-I3M and desulfo N-X-I3M. Attempted isolation as intact glucosinolates N-Hex-I3M and N-X-I3M, in order to avoid the apparently destabilizing sulfatase treatment, also was unsuccessful despite using a gentle, well-established procedure (Agerbirk et al., 2014).

Mutations in IGMT5 But Not CYP81F4 Affect Resistance against M. javanica

We challenged Arabidopsis with the root-knot nematode M. javanica to test whether the 1-IG modification pathway had a potential impact on plant-parasite interactions. M. javanica is a soil-borne generalist plant parasite capable of completing its life cycle on Arabidopsis (Wu et al., 1998). Successful invasion of root tissue by juvenile root-knot nematodes induces the formation of root galls, in which the nematodes mature and reproduce.

We inoculated igmt5_R-KO (GK_390B09), cyp81f4 (SALK_024438), cyp81f2 (SALK_005882), and Col-0, the common genetic background of the three mutants, with juvenile nematodes (n = 10 replicates per genotype). The cyp81f4 mutant was defective in the conversion of I3M to 1OHI3M in the 1-IG modification pathway, indicated by the nearly complete absence of 1MOI3M (Pfalz et al., 2011). Similar to the igmt5_R-KO genotype, cyp81f4 roots had higher levels of 4OHI3M, 4MOI3M, and I3M than the wild type (Pfalz et al., 2011). The cyp81f2 mutant was impaired in the conversion of I3M to 4OHI3M in the 4-IG modification pathway (Pfalz et al., 2009). Roots of this mutant displayed an approximately 50% reduction in the concentration of 4OHI3M and 4MOI3M and a moderate increase of I3M, but 1MOI3M content was indistinguishable from that of the wild type (Pfalz et al., 2011).

One month after inoculation, we counted the total number of nematode-induced root galls. On the Col-0 wild type, we found that, on average, approximately 0.9 galls had developed per nematode in the inoculum, indicating a high rate of infection and nematode establishment. Root gall numbers differed significantly among genotypes (F_3,36 = 8.62, P < 0.0002). The igmt5_R-KO mutant had the lowest number of root galls, whereas cyp81f2 and cyp81f4 were indistinguishable from the wild type (Fig. 5). We confirmed these results (except for cyp81f2, which had germination problems) in an independent experiment (n = 10–30 replicates per genotype). We analyzed the combined results from both experiments with a mixed-model ANOVA, treating genotype as a fixed effect and experiment and genotype × experiment interaction as random effects to obtain mean squares (MS) for genotype and genotype × experiment in the model: gall number = constant + genotype + experiment + genotype × experiment. We calculated F ratios as MS_genotype/MS_{genotype × experiment}. The genotype effect was significant (F_2,2 = 25.02, P < 0.05), and igmt5_R-KO had, on average, approximately 20% fewer galls than the wild type or cyp81f4, while gall numbers were indistinguishable between the wild type and cyp81f4. Hence, disruption of IGMT5 function (igmt5_R-KO), but not the absence of the entire 1-IG modification pathway (cyp81f4), attenuated nematode performance, suggesting that 1OHI3M (or its in planta derivatives N-Hex-I3M and N-X-I3M) had toxic and/or deterrent effects on nematodes.

Figure 5.

Root gall numbers in the wild type and IG mutants. Shown are means, variances, and outliers in the distribution of root gall numbers on the Col-0 wild type and cyp81f2, cyp81f4, and igmt5_R-KO genotypes 30 d post inoculation of 1-month-old plants with juvenile nematodes. Ten plants were analyzed per genotype. Different letters indicate significant differences (P < 0.05) between genotypes according to posthoc two-tailed Student’s t test. The igmt5_R-KO genotype has significantly fewer root galls than the other genotypes (all t > 4.6, P < 0.001).

DISCUSSION

Biosynthesis of 1MOI3M in Arabidopsis Requires IGMT5

In GK_390B09 (igmt5_R-KO) roots, the accumulation of 1OHI3M and the near absence of 1MOI3M corresponded with the highly reduced quantity of IGMT5 transcripts. Vice versa, increased IGMT5 transcript abundance in SALK_151470 (IGMT5_R-OE) roots resulted in an increase in 1MOI3M concentration. Hence, IGMT5 converts 1OHI3M to 1MOI3M. Moreover, disruption of the IGMT5 function in GK_390B09 roots was not compensated for by other IGMTs, even though these enzymes have the potential to convert 1OHI3M to 1MOI3M in a transient N. benthamiana expression system (Pfalz et al., 2011). Thus, IGMT5 is the only IGMT that targets 1OHI3M in Arabidopsis. When IGMT5 is nonfunctional, then 1OHI3M and I3M, the metabolic precursors of 1MOI3M, are piling up. Surplus I3M is partly shuttled into the 4-IG modification pathway, resulting in increased 4MOI3M concentration in roots. Surplus 1OHI3M is partially removed by conjugation, resulting in the formation of two conjugates, N-Hex-I3M (=c2) and N-X-I3M (=c3). Thus, two parallel pathways for IG modification exist, with different but related sets of enzymes being responsible for modification either at position 1 or position 4 of the indole ring. CYP81F4 and IGMT5 carry out hydroxylation and methoxylation at the indole N, whereas CYP81F1 to CYP81F3 and IGMT1 and IGMT2 (and probably also the highly similar IGMT3 and IGMT4) modify the indole ring at position 4 (Fig. 6).

Figure 6.

The 4-IG and 1-IG modification pathways in Arabidopsis. Shown are the 4-IG modification pathway (top), the 1-IG modification pathway (middle), and the newly discovered 1OHI3M conjugates (bottom). Inferred enzyme functions are shown next to arrows. The proposed function of IGMT3 and IGMT4 has not yet been demonstrated biochemically. Question marks indicate unknown enzymes. I3M is the product of the IG core pathway. Members of the CYP81F subfamily of cytochrome P450s carry out hydroxylation reactions either in position 4 or 1 of the indole ring. IGMTs belonging to family 2 plant O-methyltransferases convert hydroxy intermediates to methoxylated products. IGMT5 is specific for 1OHI3M.

CYP81F4 Expression and IG Transport Processes between Roots and Rosette May Account for Different Leaf Phenotypes of IGMT5 Promoter Mutants

Both T-DNA insertion mutants had an IGMT5_l-OE genotype; they strongly overexpressed IGMT5 in their leaves (Figs. 2A and 3A). However, in relation to the strong increase in leaf IGMT5 transcript levels, alterations in leaf 1MOI3M quantity were fairly moderate (but nonetheless significant; Figs. 2B and 3B). Furthermore, compared with the wild type, 1MOI3M leaf concentration was elevated in SALK_151470 but reduced in GK_390B09. How can these observations be explained?

Previous work had shown that 1MOI3M accumulation strongly depended on CYP81F4, the gene coding for the entry enzyme into the 1-IG modification pathway (Pfalz et al., 2011). CYP81F4 activity controls the availability of the IGMT5 substrate 1OHI3M and, in turn, also the abundance of the corresponding reaction product 1MOI3M. Therefore, low expression of CYP81F4, as observed in leaves (Figs. 2A and 3A), results in low amounts of 1MOI3M even when IGMT5 expression is high.

Leaves of GK_390B09 overexpressed IGMT5 and CYP81F4 expression was indistinguishable from that of the wild type. Nonetheless, 1MOI3M concentration was reduced to 50% of the wild-type level (Fig. 3A). Thus, it appears that on-site biosynthesis accounts only partially for leaf 1MOI3M content; the remainder is synthesized in another organ (i.e. the roots) and transported into the leaves. Because GK_390B09 roots are virtually devoid of 1MOI3M (Fig. 3B), they cannot contribute to leaf 1MOI3M accumulation, and the corresponding fraction of 1MOI3M is missing in the leaves. Like GK_390B09, leaves of SALK_151470 overexpressed IGMT5 and CYP81F4 transcript levels were unaltered (Fig. 2A), but in contrast to GK_390B09, SALK_151470 overexpressed IGMT5 also in its roots. Combined with the high, wild-type-like root expression of CYP81F4 (Fig. 2A), this results in an increase in root and, in turn, via transport to the leaves, also in leaf 1MOI3M concentration (Fig. 2B). Accordingly, the increased 1MOI3M level in SALK_151470 is caused by root (and not leaf) IGMT5 overexpression.

IG transport between plant organs has been reported previously, and two transporters, GTR1 and GTR2, have been identified as being essential for the transport of Met- and Trp-derived glucosinolates into seeds (Nour-Eldin et al., 2012; Andersen et al., 2013). Noteworthy, these transporters were not required for IG transport between roots and rosette, implying that other, unknown transporters for IG transport exist (Andersen et al., 2013; Jørgensen et al., 2015).

Transport also could explain why N-X-I3M is present in the leaves of GK_390B09. As we discuss below, N-X-I3M and N-Hex-I3M are very likely 1OHI3M detoxification products that are formed when IGMT5 function is disrupted, suggesting that they both originate in igmt5_R-KO roots. However, N-Hex-I3M was present only in roots but not in leaves, indicating that IG transport processes may discriminate between different IG structures.

Conjugation of 1OHI3M May Avoid Self-Toxicity Caused by the Disruption of IGMT5 Function

Plants need to contain potential deleterious, self-toxic effects that may arise from the biosynthesis of defensive metabolites. Among other mechanisms, self-toxicity can be avoided by chemical modification to obtain less reactive metabolites, sequestration of toxic molecules in vacuoles, or excretion into extracellular compartments (for review, see Sirikantaramas et al., 2008). In the case of glucosinolates, self-toxicity is usually avoided by the incorporation of a stabilizing S-linked β-d-Glc moiety in the glucosinolate structure and by the compartmentalization of glucosinolates separate from myrosinases, which are specialized thioglucoside glucohydrolases. Myrosinase-catalyzed removal of the d-Glc moiety generates an unstable intermediate that decomposes rapidly into a variety of bioactive products, such as isothiocyanates, thiocyanates, nitriles, epithionitriles, or oxazolidine-2-thiones (Halkier and Gershenzon, 2006). S-Glycosylation and compartmentalization, however, may be insufficient to prevent the potential self-toxicity caused by 1OHI3M. Unsubstituted N-hydroxyindoles tend to be highly reactive (Acheson et al., 1974; Acheson, 1990; Somei, 1999). To stabilize these molecules, chemical groups withdrawing electrons from the electron-rich nitrogen of the N-hydroxy group or resonance-stabilizing substituents on the indole nucleus are required (Rani and Granchi, 2015). The conversion of the N-hydroxy group in 1OHI3M into an N-methoxy group, catalyzed by IGMT5, is such a stabilizing reaction. Hence, IGMT5 activity limits 1OHI3M quantity.

Strict control of 1OHI3M abundance in the Arabidopsis wild type is reflected in the tight correlation between CYP81F4 and IGMT5 expression across a wide range of tissues, developmental stages, and environmental conditions (Pfalz et al., 2011). But surprisingly, disruption of IGMT5 expression in GK_390B09 (igmt5_R-KO) did not entail a concomitant decrease in CYP81F4 expression: CYP81F4 transcript abundance was indistinguishable between GK_390B09 and the wild type (Fig. 3A). Likewise, accumulating 1OHI3M did not provoke a reduction in the accumulation of I3M, 4OHI3M, or 4MOI3M (Fig. 3B). Thus, there is no evidence for negative feedback (Wentzell et al., 2007; Clay et al., 2009) that would reduce flux through the IG core pathway or block entry into the 1-IG modification pathway upon 1OHI3M accumulation. Rather, IG conjugates are formed in GK_390B09 (igmt5_R-KO), indicating that this mutant relies on a different mechanism, conjugation, to limit potential hazardous effects that could be caused by accumulating 1OHI3M. Even though attempts to fully elucidate the structure of the I3M conjugates were unsuccessful, evidence from MS analyses supports the conclusion that the substituents are attached to the indole N, thus substituting the reactive N-hydroxy group and stabilizing the molecule.

The 1-IG Modification Pathway Can Generate Defense-Related Metabolites

The importance of IGs for aboveground defense against a broad range of enemies is well documented. In vitro and in vivo studies have shown that IGs, in particular those that are modified at position 4 of the indole ring, negatively impacted the performance of the phloem-feeding green peach aphid Myzus persicae (Kim and Jander, 2007; Kim et al., 2008; Pfalz et al., 2009). Moreover, IGs influenced the oviposition preferences of crucifer-specialist insects. Intact IGs served as host recognition cues for the diamondback moth Plutella xylostella (Sun et al., 2009), and both intact IGs and IG hydrolysis products determined oviposition choices of the butterfly Pieris rapae (Huang and Renwick, 1994; de Vos et al., 2008). Likewise, IGs played an important role in defense against fungal, oomycete, and bacterial pathogens, largely attributed to the biosynthesis of CYP81F2-dependent 4-IG modification products and their activation by PEN2 myrosinase-catalyzed hydrolysis (Bednarek et al., 2009; Clay et al., 2009; Hiruma et al., 2010; Sanchez-Vallet et al., 2010; Schlaeppi et al., 2010; Buxdorf et al., 2013; Frerigmann et al., 2016). The CYP81F2/PEN2 pathway is active in leaves but also may contribute to defense against root pathogens (Millet et al., 2010; Iven et al., 2012).

The potential significance of the 1-IG modification pathway for plant-enemy interactions has remained unexplored, even though one of its products, 1MOI3M, is the dominant IG in Arabidopsis roots (Petersen et al., 2002; Brown et al., 2003; Pfalz et al., 2011). We used a soil-borne parasite, the root-knot nematode M. javanica, to test whether metabolites from the 1-IG modification pathway contributed to belowground defense. Mutant genotypes suitable for dissecting a potential influence of the 1-IG modification pathway on nematodes, cyp81f4 and igmt5_R-KO, were in the Col-0 background, which had low basal resistance against M. javanica. The success of nematode infection and establishment was high in Col-0, indicating that the quantity of 1MOI3M present in roots of this accession posed no major obstacle to M. javanica. Consequently, the complete absence of metabolites from the 1-IG modification pathway did not alter nematode success; root gall number on cyp81f4, which is defective in the entry enzyme for the 1-IG modification pathway (Fig. 6), was statistically indistinguishable from that of the wild type. By contrast, the number of nematode-induced root galls was reduced significantly in the igmt5_R-KO genotype, indicating that bioactive molecules with nematode-repellent and/or nematode-toxic effects were generated when the 1-IG modification pathway was interrupted at IGMT5. Furthermore, the nematicidal principle required intact CYP81F4, because nematode performance on cyp81f4 did not differ from that on the wild type. Since 4MOI3M accumulated to similar levels in both mutants, igmt5_R-KO and cyp81f4, we can rule out that nematicidal effects were caused by this metabolite. Based on the comparison of these two mutants, we also can rule out that the reduction in gall numbers was caused indirectly by altered auxin levels upon disruption of the 1-IG modification pathway (Zhao et al., 2002; Glawischnig et al., 2004). Therefore, we conclude that increased nematode resistance in the igmt5_R-KO genotype was caused by 1OHI3M and/or the partially characterized conjugates N-Hex-I3M and N-X-I3M, all of which accumulated as a consequence of CYP81F4 activity in the absence of IGMT5 function.

Our results clearly showed that the 1-IG modification pathway had the potential to generate bioactive molecules. Release of this defensive potential required a disruption of the IGMT5 function, which could be triggered by soil-borne parasites in the wild type. Why, then, was this potential not exploited in the Arabidopsis wild type for defense against M. javanica? Sedentary endoparasitic nematodes like M. javanica have the ability to suppress their host plant’s defenses (Hamamouch et al., 2012; Jaouannet et al., 2013; Xue et al., 2013; for review, see Hewezi and Baum, 2013). Root-knot nematodes require this ability for root entry, migration within the root tissue, and establishment and maintenance of their feeding sites. Feeding site establishment results in the formation of giant cells and, together with a concomitant response of adjacent plant tissue, root gall formation. Indeed, transcript profiles of laser-dissected root galls showed that the IG biosynthesis genes CYP79B2, CYP83B1, GGP1, SUR1, AKN1, and IGMT5 and the IG regulator gene MYB34 were down-regulated in giant cells (Barcala et al., 2010), indicating that IG biosynthesis was suppressed at nematode feeding sites. Vice versa, CYP79B2, IGMT5, and MYB34 were up-regulated in intact root galls (Barcala et al., 2010), possibly to limit secondary damage that could arise from nematode-associated pathogens. Hence, root-knot nematode-triggered repression of IG biosynthesis could explain why nematode performance on the Col-0 wild type did not differ from that of cyp81f4. However, this would not explain why the igmt5_R-KO genotype was more resistant against nematodes, unless 1OHI3M and/or N-Hex-I3M and N-X-I3M exerted their effects in the rhizosphere, before or while juvenile nematodes entered the plant roots. Root exudates from cruciferous plants indeed contain glucosinolates and glucosinolate hydrolysis products (Schreiner et al., 2011; Strehmel et al., 2014), and these metabolites have the potential to alter soil microbial communities (Bressan et al., 2009).

Even though 1MOI3M did not alter the performance of M. javanica, this does not preclude a function of this metabolite in defense against other soil-borne parasites in Arabidopsis or other crucifers. Indeed, IGMT5 has been suggested as a candidate gene for a Sclerotinia spp. stem rot resistance quantitative trait locus in Brassica napus (Wu et al., 2013). In terms of chemical reactivity, the methoxy group at the indole N serves to stabilize the corresponding isothiocyanate and the downstream alcohol originating from spontaneous hydrolysis (Hanley et al., 1990). By this effect, the N-methoxy group could influence the quantitative balance of the complex mixture of hydrolysis products that originate from myrosinase-catalyzed hydrolysis as well as the actual resulting structures and bioactivities (Agerbirk et al., 1996, 1998; Stephensen et al., 2000).

Further biodiversity has been discovered as 1- and 4-substituted indole phytoalexins in various crucifers (Pedras et al., 2010, 2011) and as the rare disubstituted 1,4-dimethoxy I3M in Barbarea vulgaris (Agerbirk et al., 2001). From the growing amount of bioinformatics data in chemical ecology model systems (Liu et al., 2016) and the biochemical data from Arabidopsis, structural variants such as these can be tested for complementary resistance properties. Similarly, translational research with Brassica spp. vegetables will allow precise testing of the nutritional effects of IG nitrogen substitution (Schumacher et al., 2014; Wiesner et al., 2014). With the characterization of mutants for CYP81F4 (Pfalz et al., 2011) and IGMT5 (this work), the tools necessary to determine the significance of the 1-IG modification pathway for plant-enemy and vegetable-human interactions are now at hand.

MATERIALS AND METHODS

Plant Growth Conditions

Arabidopsis (Arabidopsis thaliana) lines SALK_151470 (accession no. N437365; Alonso et al., 2003) and GK_390B09 (accession no. N651470; Kleinboelting et al., 2012) were obtained from the Nottingham Arabidopsis Stock Centre. For the analysis of glucosinolates and for RNA extraction prior to RT-qPCR, Arabidopsis seeds were sown into damp potting medium, cold treated at 6°C in the dark for 3 to 4 d, and germinated in the light. Plants were grown in an environmentally controlled growth chamber equipped with T8 LUXLINE PLUS F58W/865 bulbs (Sylvania) with 200 µmol s⁻¹ m⁻² as a light source, with an 11.5-h-day/12.5-h-night cycle, and 22°C and 60% relative humidity (day) and 16°C and 75% relative humidity (night). Seedlings were transferred to sand (particle size of 2 mm or less; Sacamat sable fin 0/2) 5 to 7 d post germination. Every 7 to 10 d, plants were fertilized with 30 mL of 0.1% (v/v) Hydrocani C2 liquid fertilizer (Hydro Agri). For nematode bioassays, Arabidopsis seeds were surface sterilized and sown onto petri dishes with one-half-strength Murashige and Skoog medium + B5 vitamin Gamborg, 1% (w/v) plant agar, 1% (w/v) Suc (all Duchefa Biochimie), and 0.1% (w/v) MES (Carl Roth), adjusted to pH 5.7. Following cold treatment and germination in the light, seedlings were transferred to sand 10 to 12 d post germination.

DNA Extraction and Genotyping

Isolation of DNA from freeze-dried Arabidopsis leaves was described previously (Kroymann et al., 2001). Seed material for GK_390B09 and SALK_151470 provided by the Nottingham Arabidopsis Stock Centre was segregating for the T-DNA insertion. T3 plants were screened with PCR-based markers to identify lines with either homozygous mutant or homozygous wild-type alleles. Primers are listed in Supplemental Table S1. For GK_390B09, primer combination GABI_390F + GABI_390R was specific for the wild type and primer combination GABI_390R + GABI-Lb1 was specific for mutant alleles. For SALK_151470, primer combination 151470F + 151470R was specific for the wild type and primer combination 151470R + LB1 was specific for mutant alleles. PCR-based genotyping was carried out as described by Pfalz et al. (2011).

RNA Extraction and RT-qPCR

Leaves and roots were ground to a fine powder. Tissue samples were cooled with liquid N₂ during processing to prevent RNA degradation. Total RNA was extracted with TRIsure reagent (Bioline) following the manufacturer’s recommendations, treated with DNase (Turbo DNase; Ambion), and purified with RNeasy MinElute columns (Qiagen). RNA integrity was verified with RNA Nano chips (Agilent Technologies) on an Agilent 2100 Bioanalyzer. RNA quantity was determined with a Nanodrop ND-1000 spectrophotometer (Nanodrop Technologies). First-strand cDNA synthesis from a maximum of 500 ng of DNA-free total RNA was carried out with the Maxima First Strand cDNA Synthesis Kit (ThermoFisher) according to the manufacturer’s recommendations. RT-qPCR was performed with Maxima SYBR Green qPCR Master Mix (ThermoFisher) on the StepOnePlus Real Time PCR System (Applied Biosystems) with an initial denaturation for 10 min at 94°C, followed by 40 cycles of 15 s at 94°C, 30 s at 58°C, and 30 s at 72°C. RT-qPCR data were analyzed with qbase⁺ (Biogazelle).

Glucosinolate Extraction and HPLC Analysis in a 96-Well Format

Wild-type (Col-0) and mutant plants were randomized and grown in sand. Roots and leaves from 4-week-old plants were harvested separately. Roots were rinsed briefly with water to remove attached sand and dried quickly with a paper towel. The fresh weight of leaves and roots was determined, and samples were snap frozen in liquid N₂. Tissue was lyophilized to dryness and homogenized with 2.3-mm ball bearings in a shaker. Extraction of glucosinolates in their desulfated form was carried out as described by Pfalz et al. (2011) but without adding an internal standard. Desulfoglucosinolates were separated on the Dionex UltiMate 3000 Rapid Separation LC System equipped with a LiChrospher 100 RP-18 end-capped LiChroCART column (4 × 250 mm; pore size, 100 A; particle size, 5 µm). HPLC separation, desulfoglucosinolate identification based on retention times, and diode array UV light detection were conducted as described by Kroymann et al. (2001). Integrated peak area at UV_229nm was corrected for dilution and sample weight. Data for wild-type and mutant root and leaf glucosinolate profiles were analyzed with Systat version 9 (SPSS) using a general linear model: glucosinolate = constant + genotype + row + column, with row and column as variables to control for position effects during plant growth and desulfoglucosinolate extraction in a 96-well format.

HPLC-MS/MS of Desulfated Glucosinolates

The identification of newly discovered IG derivatives in Arabidopsis igmt5 roots was attempted using an established, gentle method based on extraction in boiling aqueous methanol, binding to anion-exchange columns, enzymatic desulfation, and preparative HPLC with a Luna C18 (2) column (250 mm × 4.6 mm, 5 µm; Phenomenex) with diode array UV light detection followed by either ion-trap HPLC-MS/MS using NaCl-doped eluent or lyophilization plus ¹H-NMR at 400 MHz in deuterium oxide (Agerbirk and Olsen, 2011). The sulfatase enzyme was partially purified by fractional precipitation (Graser et al., 2001) from a commercially obtained Helix pomatia enzyme powder known to contain other hydrolytic activities, including glucuronidase (S9626; Sigma-Aldrich). Known glucosinolates in the samples were identified in parallel based on comparisons of UV light spectra, retention times, m/z values, and MS² results with those of authentic desulfated standards including known IG (Agerbirk et al., 2001) and, in one case, by NMR as a positive control (desulfo 4-methylsulfinylbutyl glucosinolate). Additional HR-MS of desulfoglucosinolates was carried out using a Bruker Compact electrospray ionization-quadrupole-time of flight instrument (Bruker Daltonics). The attempted isolation of IG derivatives as intact glucosinolates was carried out as reported previously (Agerbirk et al., 2014), except that the acidic wash was left out in an attempt to avoid degradation.

HR-MS of Intact Glucosinolates

Samples were homogenized for 5 min in 2.5 mL of 3:1 (v/v) acetonitrile:water using a Polytron homogenizer (Fisher Scientific) and centrifuged for 15 min and 20°C at 8,000g. Supernatants were collected. The pellets were reextracted with 1 mL of 3:1 (v/v) acetonitrile:water and sonicated (15 min, 25 Hz). After centrifugation, supernatants from the first and second extractions were pooled. Extracts were evaporated to dryness at ambient temperature in a Thermo Savant SpeedVac. Dried extracts were dissolved in 1:19 (v/v) acetonitrile:water and filtered. Liquid chromatography-HR-MS analysis was carried out on an I-class ultra-performance liquid chromatography system coupled with a Xevo G2-S QToF mass spectrometer (Waters). Liquid chromatography separation was conducted at 40°C on a Waters ACQUITY UPLC CSH C18 column (2.1 × 100 mm length, 1.7 μm) using an acetonitrile (+0.1% formic acid)/water (+0.1% formic acid) gradient as a mobile phase. MS data were acquired in negative electrospray, resolution mode (32,000 FWHM) and by MS and MS^E modes. Leu-enkephalin was used as lockspray. The electrospray capillary potential was 1.5 kV. The source block and desolvation gas had temperatures of 110°C and 450°C, respectively. N₂ was used as nebulization gas and desolvation gas (100 and 800 L h⁻¹, respectively). Collision-induced dissociation MS^E experiments were performed using a 20- to 55-eV collision energy ramp.

Nematode Assays

A Meloidogyne javanica population was maintained on tomato (Solanum lycopersicum) plants of the nematode-susceptible variety Roma, grown in 20-cm-square pots filled with 2:1 (v/v) sand (particle size, 2 mm; Sacamat):potting soil in climate chambers (Pol-EkoAparatura-type kk 1200 STD; Wodzisław Śląski-Pologne) with a 14-h-day/10-h-night cycle at 28°C and 65% relative humidity (day) and 19°C and 65% relative humidity (night). Nematode eggs were collected and placed on a small mesh (mesh size > 200 μm) floating on distilled water in a petri dish. Eggs were placed for 1 week in the dark at 28°C, after which nematode juvenile larvae had hatched. Nematode number was estimated with nematode-counting slides (Chalex) and a stereomicroscope, and the nematode concentration was adjusted to approximately 50 to 100 mL⁻¹. Arabidopsis plants were inoculated with 1 mL of nematode suspension per plant 1 month after plants had been transferred to sand. Thirty days after inoculation with nematodes, plant roots were carefully washed with water and dispersed in a flat dish, and the total number of galls formed on the roots was counted with the help of a stereomicroscope.

Accession Numbers

Sequence data from this article can be found in the Arabidopsis Genome Initiative database under accession numbers At1g76790 (IGMT5), At4g37410 (CYP81F4), At1g21100 (IGMT1), At1g21120 (IGMT2), At1g21110 (IGMT3), and At1g21130 (IGMT4).

Supplemental Data

The following supplemental materials are available.

• Supplemental Figure S1. HPLC profiles and UV light absorption spectra of desulfated glucosinolates extracted from roots of Col-0 and GK_390B09 (igmt5_R-KO).

• Supplemental Figure S2. Quantity and stability of desulfated IGs during HPLC analysis.

• Supplemental Figure S3. Desulfated glucosinolates from Arabidopsis igmt5_R-KO roots analyzed by HPLC-MS/MS.

• Supplemental Table S1. Primers used in this study.

Glossary

IG

indole glucosinolate

RT-qPCR

real-time quantitative PCR

MS

mass spectrometry

HR-MS

high-resolution mass spectrometry

MS/MS

tandem mass spectrometry