Isochorismate-derived biosynthesis of the plant stress hormone salicylic acid

See allHide authors and affiliations

Science  02 Aug 2019:
Vol. 365, Issue 6452, pp. 498-502
DOI: 10.1126/science.aaw1720

Spontaneous degradation in hormone synthesis

The phytohormone salicylic acid (SA) helps plants respond to biological and physical stresses. Rekhter et al. identified the biosynthetic pathway that produces SA in response to pathogens. A precursor, isochorismic acid, is formed in the chloroplast and then exported to the cytosol. There, enzymatically produced isochorismate-9-glutamate spontaneously decomposes to release SA plus a by-product. The results clarify key steps in the mechanisms involved in synthesizing this important regulator of plant immunity.

Science, this issue p. 498


The phytohormone salicylic acid (SA) controls biotic and abiotic plant stress responses. Plastid-produced chorismate is a branch-point metabolite for SA biosynthesis. Most pathogen-induced SA derives from isochorismate, which is generated from chorismate by the catalytic activity of ISOCHORISMATE SYNTHASE1. Here, we ask how and in which cellular compartment isochorismate is converted to SA. We show that in Arabidopsis, the pathway downstream of isochorismate requires only two additional proteins: ENHANCED DISEASE SUSCEPTIBILITY5, which exports isochorismate from the plastid to the cytosol, and the cytosolic amidotransferase avrPphB SUSCEPTIBLE3 (PBS3). PBS3 catalyzes the conjugation of glutamate to isochorismate to produce isochorismate-9-glutamate, which spontaneously decomposes into SA and 2-hydroxy-acryloyl-N-glutamate. The minimal requirement of three compartmentalized proteins controlling unidirectional forward flux may protect the pathway against evolutionary forces and pathogen perturbations.

The plant hormone salicylic acid (SA) is required for adaptive responses to biotic and abiotic stresses (1). The plant-specific SA-biosynthesis pathway(s) remain unclear (1). Plants can produce SA via two metabolic processes, which diverge in plastids after the metabolite chorismic acid. To execute signaling functions, either SA or one of its precursors needs to be transported out of the plastids. The multidrug and toxin extrusion (MATE) family transporter ENHANCED DISEASE SUSCEPTIBILITY5 (EDS5) localized in the chloroplast envelope is required in this process (2). In the model plant Arabidopsis thaliana, ~10% of defense-related SA is produced by the cytosolic PHENYLALANINE AMMONIA LYASE pathway, whereas ~90% is derived from isochorismate generated by the plastid-localized ISOCHORISMATE SYNTHASE1 (ICS1) (fig. S1) (3). SA-producing bacteria use an isochorismate pyruvate lyase (IPL) to convert isochorismate to SA (4). How plants convert plastidial isochorismate to SA remains unclear, as no plant IPL homologs are evident from genomic analyses (5).

Similar to plants with defects in ICS1 (6) and EDS5 (2), plants carrying mutations in the avrPphB SUSCEPTIBLE3 (PBS3) gene also show reduced accumulation of SA in response to pathogens (79). PBS3 belongs to the GH3 acyl-adenylase gene family, whose founding member JASMONIC ACID RESPONSE1 (JAR1; AtGH3.11) catalyzes conjugation of isoleucine to jasmonic acid, yielding the active hormone conjugate jasmonoyl-isoleucine (10). Although SA is a poor substrate of PBS3 (11), transcriptional co-regulation of PBS3 with ICS1 and EDS5 (fig. S2) suggests that all three gene products contribute coordinately to modification, transport, or biosynthesis of SA or its precursors. Key to SA signaling is NONEXPRESSOR OF PATHOGENESIS-RELATED GENES1 (NPR1), which encodes an SA-dependent regulator of defense gene expression (12). Consequently, npr1 mutant plants show compromised defense against diverse pathogens (12). A suppressor screen in the npr1-1 mutant background identified a gain-of-function mutation in the receptor-like protein SNC2. This snc2-1D (suppressor of npr1-1, constitutive 2) mutant shows an autoimmune phenotype accompanied by a constitutive excess of SA and stunted plant growth (13) (Fig. 1, A and B). We used this mutant as a tool to investigate the contribution of PBS3 and EDS5 to SA accumulation in the snc2-1D npr1-1 background (Fig. 1, B to D).

Fig. 1 Analysis of SA-related metabolites in autoimmune mutants.

(A) Morphology of 4-week-old plants. Col0, wild type. Absolute amounts of (B) SA and (C) SAG and relative amounts of (D) isochorismate (IC). The highest mean value was set 100%. Bars represent mean ± SD of three biological replicates. Statistical differences among replicates are labeled with different letters [P < 0.05, one-way analysis of variance (ANOVA) and post-hoc Tukey’s test; n = 3 independent pools of 10 individual plants].

Both triple mutants snc2-1D npr1-1 pbs3-1 and snc2-1D npr1-1 eds5-3 retained the dwarf phenotype (Fig. 1A), which is indicative of constitutive activation of SA-dependent and SA-independent resistance pathways in the snc2-1D npr1-1 mutant (13). However, SA and its glycosylated storage form SA glycoside (SAG), which over-accumulated at 20-fold and 120-fold excess in the snc2-1D npr1-1 double mutant, respectively, regained wild-type quantities in both triple mutants (Fig. 1, B and C). In contrast, the snc2-1D npr1-1 pbs3-1 triple mutant showed elevated levels of isochorismate (the ICS1 reaction product), accumulating threefold more than wild-type and 20-fold more than the snc2-1D npr1-1 eds5-3 triple mutant (Fig. 1D). Thus, we hypothesized that isochorismate is the substrate of PBS3. Isochorismate did not accumulate in the snc2-1D npr1-1 eds5-3 mutant, most likely because ICS1 operates near equilibrium and isochorismate cannot be exported from the plastid in the mutant (14).

To test our hypothesis, we purified recombinant PBS3 and ICS1 to homogeneity (figs. S3 and S4). As isochorismate is not commercially available, we used ICS1 to produce isochorismate from chorismate (fig. S5). This isochorismate preparation, which included residual amounts of chorismate, was incubated with PBS3 (Fig. 2D). The reaction monitored by liquid chromatography coupled to high-resolution mass spectrometry yielded four signals for the extracted ion chromatogram of mass-to-charge ratio (m/z) 354.083 (Fig. 2A). Three minor signals represented chorismate conjugates and isochorismate-7-glutamate (Fig. 2, A and B, and fig. S6, A to F). The signal at 2.85 min was identified as isochorismate-9-glutamate by fragmentation analyses and 15N labeling (Fig. 2, A and B, and fig. S7, B to F). We concluded that isochorismate-9-glutamate was the preferred product of PBS3. The incubation of PBS3 with chorismate confirmed the structure of two minor products (Fig. 2, A and B) (11). Isochorismate decomposes in aqueous solution into SA and pyruvate (15). We confirmed this observation by detecting small amounts of SA in the absence of glutamate, adenosine triphosphate (ATP), or PBS3 (Fig. 2C). However, in the complete PBS3 assay, we detected four times more SA within 1 hour (Fig. 2C). This difference is most likely higher in planta, because we detected nonenzymatic decomposition of isochorismate into SA before adding PBS3 to the reaction (Fig. 2, C, D, and F). Lyase activity has not been reported for GH3 proteins so far, suggesting that formation of isochorismate-9-glutamate accelerates the nonenzymatic decomposition of isochorismate by elimination, leading to increased accumulation of SA.

Fig. 2 PBS3 catalyzes the formation of isochorismate-9-glutamate, which nonenzymatically decomposes into SA.

(A) Analysis of the in vitro products of PBS3. The assays were performed with chorismate (CA) and glutamate (Glu) (red); with CA, Glu, and ICS1, which converts CA to isochorismate (IC) (black); or without substrates (blue). The extracted ion chromatogram for the amido conjugates (m/z 354.083) is shown. (B) Formula of amido conjugates formed by PBS3 with Glu (green) and CA (labeled a and d) or IC (labeled b and c). Structures were solved by tandem mass spectrometry (figs. S6, S7, and S11). (C) SA (m/z 137.024) accumulates in the assay with PBS3 in the presence of IC, Glu, and ATP (black) but only in minor amounts in the absence of Glu (red), ATP (blue), or PBS3 (brown). SA in the control assays results from the chemical decomposition of IC in solution during experimental procedure (D). (E) Kinetic parameters of PBS3 for 4-hydroxy-benzoate (4HBA) or IC. Data were obtained spectrophotometrically in triplicates. The turnover number (kcat) was calculated from the maximal reaction rate (V) for the total enzyme concentration (Et). The Michaelis constant (Km) was measured by using different concentrations of 4HBA and IC. (F) Time course for the formation of SA from the chemical decomposition of IC (red) and IC-9-glutamate (IC-9-Glu) (black). (G and H) Accumulation of IC-9-Glu and 2-hydroxy-acryloyl-N-glutamate (2HNG) in double and triple mutants. Bars represent mean ± SD of three biological replicates. The highest mean value was set 100%. Statistical differences among replicates are labeled with different letters (P < 0.05, one-way ANOVA and post-hoc Tukey’s test; n = 3).

GH3 enzymes are bifunctional (10) and catalyze the adenylation of carboxyl groups, which react with an amino group to form an amide bond. The kinetic parameters for the reaction of PBS3 with isochorismate and glutamate were determined spectrophotometrically to compare the data with those described for 4-hydroxy-benzoic acid (Fig. 2E) (11). The catalytic efficiency of PBS3 was 737 times as high when isochorismate was used as the carboxyl substrate instead of 4-hydroxy-benzoic acid. This higher substrate affinity and enhanced turnover rate are required to produce SA in sufficient amounts, because isochorismate rearranges chemically into isoprephenate eight times as fast as it decomposes into SA (15). To confirm that isochorismate is the preferred substrate for PBS3, we used structural modeling to assess the fit of isochorismate into the binding pocket of PBS3 (fig. S8B) using a published crystal structure in complex with SA and adenosine monophosphate (fig. S8A) (16). Superimposition of the ring structure of SA with that of isochorismate revealed that (i) there is space to accommodate isochorismate and isochorismate-9-glutamate in the binding pocket (figs. S8, C and D, and S9); and (ii) the two reaction partners (the phosphate group of adenosine monophosphate and the C9-carboxyl group of isochorismate) are in close proximity (fig. S8B). We monitored the nonenzymatic decomposition of isochorismate-9-glutamate or isochorismate into SA over a 24-hour time course. During the first 6 hours, formation of SA from isochorismate-9-glutamate was 10 times as high as SA formation from isochorismate (Fig. 2F). To explain this accelerated decomposition, we used molecular modeling. This suggested the formation of two additional hydrogen bonds between the two glutamate carboxyl groups of isochorismate-9-glutamate and the C2-hydroxyl and C7-carboxyl group of isochorismate (fig. S10A). The closer proximity of the amide hydrogen of the peptide bond and the oxygen of the ether bridge likely enhances the reaction rate of the hydrogen transfer. We propose that this protonation followed by base-initiated aromatization and subsequent elimination yield SA and 2-hydroxy-acryloyl-N-glutamate as final products (fig. S10B).

Besides SA, we detected 2-hydroxy-acryloyl-N-glutamate (m/z 216.051) as the second product of the nonenzymatic decomposition of isochorismate-9-glutamate in the in vitro assays and confirmed its structure by 15N labeling (figs. S7, E and F, and S11). This process was corroborated by detecting isochorismate-9-glutamate and 2-hydroxy-acryloyl-N-glutamate in snc2-1D npr1-1 double mutants but not in any of the triple-mutant combinations (Fig. 2, G and H, and fig. S7A). Neither any other amino acid conjugate of isochorismate nor isochorismate-7-glutamate were found in planta. We concluded that isochorismate-9-glutamate is the in planta precursor of SA. This was underpinned by the accumulation of isochorismate-9-glutamate, SA, and 2-hydroxy-acryloyl-N-glutamate in wild-type plants upon Pseudomonas syringae infection (fig. S12).

Previous evidence suggested that biosynthesis of SA occurred in plastids, where ICS1 is localized (3) (fig. S1). However, sequence analyses using the in silico online tools TargetP ( and Predotar ( predicted a cytosolic localization for PBS3. We used Arabidopsis efr mutant plant leaves, which allow higher rates of Agrobacterium-mediated transformation (17), for transient coexpression of PBS3 C-terminally tagged with YELLOW FLUORESCENT PROTEIN (PBS3-YFP) and ICS1 fused to CYAN FLUORESCENT PROTEIN (ICS1-CFP). Confocal laser scanning microscopy confirmed the presence of ICS1-CFP in plastids and PBS3-YFP in the cytosol (Fig. 3A). Fluorescence-lifetime imaging validated the identity of the corresponding fluorescence signals in the respective cellular compartments (fig. S13A). Immunoblot analyses demonstrated production of full-length fusion proteins (fig. S13C). Functionality of the PBS3-YFP fusion protein was confirmed by transient expression in the pbs3-1 mutant, which restored Agrobacterium-induced SA accumulation (fig. S14). Our data suggest a spatial separation of cytosolic PBS3 from its substrate isochorismate, which is produced by plastid-localized ICS1. We therefore hypothesized that EDS5, residing in the plastid envelope, exports isochorismate and not SA into the cytosol and that the exported isochorismate is metabolized into SA in the cytosol by PBS3 and spontaneous decomposition (Fig. 3D). Evidence for this pathway includes the observations that SA does not accumulate in eds5 mutants (2) and that expression of native PBS3 alone or in combination with ICS1 does not restore SA accumulation in eds5-3 mutants (Fig. 3C).

Fig. 3 PBS3 and ICS1 are localized in different compartments.

Confocal laser scanning microscopy 3 days after Agrobacterium infiltration for transient coexpression of (A) ICS1-CFP and PBS3-YFP or (B) ICS1-CFP and chloroPBS3-YFP in Arabidopsis efr leaves. Scale bars, 10 μm. (C) Transient expression of ICS1 and chloroPBS3 in Arabidopsis eds5-3 leaves restores Agrobacterium-induced SA accumulation. Twenty-four hours after infiltration, leaves were collected, and metabolites were extracted as described for the metabolite fingerprint analysis. Infiltration medium was used as mock treatment. The SA content was analyzed by liquid chromatography–mass spectrometry, and the highest mean value was set 100%. Bars represent mean ± SD of three biological replicates. Statistical differences among replicates are labeled with different letters (P < 0.05, one-way ANOVA and post-hoc Tukey’s test; n = 3). (D) Model summarizing pathogen-induced SA biosynthesis in plants.

We reasoned that forced mistargeting of PBS3 into plastids would allow restoration of SA accumulation in the transport-deficient eds5 mutant. To test this hypothesis, we fused the plastid transit peptide of ICS1 (3) to the N terminus of PBS3-YFP (chloroPBS3-YFP). Transient coexpression of ICS1-CFP and chloroPBS3-YFP in Arabidopsis efr mutant leaves confirmed colocalization of the full-length fluorescent fusion proteins in plastids (Fig. 3B and fig. S13, B and C). The subcellular localization of chloroPBS3-YFP in plastids (and PBS3-YFP in the cytosol) was independently validated in stable transgenic plants (fig. S15). Plastidial colocalization of both ICS1 and PBS3 restored Agrobacterium-induced SA synthesis in Arabidopsis eds5-3 mutants (Fig. 3C). Individual expression of ICS1 and PBS3 protein variants or combined expression of ICS1 with cytosolic PBS3 did not. The observation that plastid-targeted PBS3 alone does not restore SA production can be explained by feedback inhibition of SA on PBS3 activity (11). On the basis of the published three-dimensional structure of PBS3 (16), we postulate that the binding pocket can accommodate either SA or isochorismate, confirming the described feedback inhibition by SA (figs. S8 and S9). To overcome competitive inhibition, ectopic overexpression of ICS1 is required to produce sufficient isochorismate and to allow for quantitative displacement of SA from the active site of PBS3.

By studying SA formation in the autoimmune mutant snc2-1D npr1-1 pbs3-1, we were able to identify isochorismate as the substrate for PBS3 (Fig. 2E). We confirmed that recombinant PBS3 metabolizes isochorismate and glutamate to yield isochorismate-9-glutamate (Fig. 2A). Kinetic analyses (Fig. 2E) as well as in silico studies (figs. S8 and S9) corroborated the preference of PBS3 for isochorismate as its native substrate. SA formation via PBS3-derived isochorismate-9-glutamate is ~2 × 106 times as fast as its formation via direct chemical decomposition from its precursor isochorismate (15). The detection of isochorismate-9-glutamate as well as 2-hydroxy-acryloyl-N-glutamate in planta supported our in vitro findings (Fig. 2, G and H, and fig. S12). Our study suggests that EDS5 exports isochorismate from the plastid into the cytosol, where PBS3 metabolizes it to isochorismate-9-glutamate. Subsequent nonenzymatic decomposition results in SA formation and efficient metabolite channeling (Fig. 3D). The minimal requirement of only three proteins (ICS1, EDS5, and PBS3), their spatial separation, and the unidirectional flux may protect the pathway against evolutionary forces and pathogenic effectors.

Supplementary Materials

Materials and Methods

Figs. S1 to S15

Tables S1 and S2

References (1826)

References and Notes

Acknowledgments: We are grateful to E. Petutschnig (University of Goettingen) for confocal microscopy support, J. Parker (MPIPZ Cologne) for pXCSG-YFP/CFP destination vectors, E. Hornung (University of Goettingen) for pEntry-C-eYFR vector, A. Kelly for critical reading of the manuscript, and E. Fanghänel for advice on chemical decomposition. Funding: This research has been funded by the Deutsche Forschungsgemeinschaft (DFG) IRTG 2172 “PRoTECT” program of the Göttingen Graduate Center of Neurosciences, Biophysics, and Molecular Biosciences (to D.R., D.L., V.L., M.W., and I.F.). I.F. and V.L. were also supported by the DFG (ZUK 45/2010 and INST 186/822-1 to I.F. and INST 186/1277-1 to V.L.). Y.Z. was supported by Natural Sciences and Engineering Research Council of Canada (Discovery and CREATE-PRoTECT), Canada Foundation for Innovation, and British Columbia Knowledge Development Fund. Author contributions: D.R., D.L. Y.D., K.F., V.L., M.W., Y.Z., and I.F. conceived of and designed the experiments. D.R., D.L., K.Z., and Y.D. performed the experiments. D.R., D.L., K.F., Y.D., V.L., M.W., Y.Z., and I.F. analyzed and discussed the data. D.R., D.L., K.F., K.Z., Y.Z., M.W., V.L., and I.F. wrote the paper. Competing interests: The authors declare no competing interests. Data and materials availability: All data are available in the main text or the supplementary materials. I.F., M.W., and Y.Z. are responsible for distribution of materials integral to the findings presented in this paper.
View Abstract

Stay Connected to Science

Navigate This Article