Substrate channeling in oxylipin biosynthesis through a protein complex in the plastid envelope of Arabidopsis thaliana

Together with the interaction of miR319-regulated TCP transcription factors and the COI1 receptor, the formation of LOX2-AOS-AOC2 complexes controls key steps in jasmonic acid precursor biosynthesis, facilitating fine-tuning of the plant’s defense machinery.

Here, we report on the identification of a protein complex comprising LOX2, AOS, and AOC2 in the plastid envelope of Arabidopsis chloroplasts that channels α-LeA into JA biosynthesis. Because the expression of LOX2, AOS, and AOC2 is under the control of JA-responsive microRNAs and COI1, a mechanism is suggested that boosts JA production over aldehyde production for rapid local and systemic activation of defense genes in plants.

Generation of transgenic lines expressing AOS or HPL with C-terminal (His) 6 or FLAG tags
Transgenic plants were generated expressing C-terminal hexa-histidine (His) 6 , FLAG-tagged AOS or FLAG-tagged HPL as described in Supplementary Protocol S1 at JXB online and used for affinity purification of proteins interacting with AOS or HPL in planta.

Production of proteins and import into chloroplasts
cDNA clones for AOS (Laudert et al., 1996) and HPL (Froehlich et al., 2001) were cloned into appropriate vectors, allowing for their purification as (His) 6 -or FLAG-tagged proteins. For routine chloroplast import assays, 35 S-methionine-or 14 C-leucine-labeled proteins were produced by coupled in vitro transcription/translation in wheat germ extracts. Radiolabeled proteins were added to 50 µl import assays consisting of 25 µl of doubleconcentrated import buffer, 10 µl of a plastid suspension containing 5 × 10 7 Arabidopsis chloroplasts, and 2.5 mM Mg-ATP. All import reactions were performed at 23 °C for 15 min in darkness. Post-import protease treatment of plastids with thermolysin or trypsin and extraction of membranes with sodium carbonate (pH 11) or 1 M NaCl were carried out as described (Cline et al., 1984;Kessler and Blobel, 1996). Trypsin quenching was monitored by western blotting using TIC110 as a diagnostic marker and employing a trypsin inhibitor (Jackson et al., 1998). Plastid subfractionation into envelopes, stroma, and thylakoids was done according to Li et al. (1991). Protein was extracted and precipitated with trichloroacetic acid [5% (w/v) final concentration], resolved by SDS-PAGE on 10-20% (w/v) polyacrylamide gradients (Laemmli, 1970), and detected by autoradiography. Because the presence of the tags only marginally affected the molecular masses of the tagged proteins [peptide mass of 1013 g mol −1 for the (His) 6 tag and 841 g mol −1 for the FLAG tag], molecular masses are reported uniformly throughout this paper and refer to the untagged proteins.
Isolation of AOS-containing and HPL-containing higher molecular mass complexes AOS-(His) 6 and HPL-(His) 6 were imported into isolated Arabidopsis chloroplasts, and proteins interacting with AOS-(His) 6 -or HPL-(His) 6 were purified from detergent-solubilized envelopes by Ni-NTA affinity chromatography (Qiagen). For comparison, AOS-or HPL-containing plastid envelope protein complexes were purified from plants expressing FLAG-tagged AOS (AOS-FLAG) or HPL (HPL-FLAG) (see Supplementary Protocol S1). Both types of complexes were subjected to size exclusion chromatography on Superose 6 (column model HR10/10, GE Healthcare) and individual fractions were harvested and traced for the presence of AOS or HPL by western blotting using AOS-, HPL-, or FLAG-specific antibodies (Laudert et al., 1996;see also Supplementary Fig. S1) and an enhanced chemiluminescence kit (ECL, GE Healthcare).

Enzyme activity measurements
Activity measurements using the isolated native and reconstituted complexes or free enzymes were carried out as described in Supplementary Protocol S1. HPLC and GC-MS analyses used to identify and quantify substrates and products of the LOX2, AOS, and AOC2 reactions were performed according to Holtman et al. (1997) and Zerbe et al. (2007). Capillary chiral GC analysis was used to demonstrate the optical purity of the cis-(+)-enantiomer (>95%), which was reconfirmed by TLC, HPLC, and GC analyses with a synthetic standard (Zerbe et al., 2007).

Yeast two-hybrid screens
Split-ubiquitin yeast two-hybrid screens for membrane-bound proteins were carried out according to the manufacturer's instructions using a commercial system (Dualsystems Biotech AG, Switzerland). Vector construction is described in Supplementary Protocol S1.

Molecular modelling
Molecular modeling methods and tools are described in Supplementary Protocol S1.

Miscellaneous
Western blotting was carried out according to Towbin et al. (1979), using the indicated antisera and enhanced chemiluminescence (ECL, GE Healthcare) or anti-rabbit, anti-goat alkaline phosphatase systems. Molecular mass standards used in Fig. 3B were 14 C-Leu-labeled heat shock proteins from tomato (Nover and Scharf, 1984).

Results
In vitro import into chloroplasts and membrane targeting of AOS and HPL AOS and HPL from A. thaliana and other plant species belong to the family of atypical cytochrome P450s, designated CYP74. AOS (CYP74A) and HPL (CYP74B) require neither oxygen nor NADPH-dependent cytochrome P450 reductase for activity and thus are non-canonical P450s (Song and Brash, 1991;Laudert et al., 1996;Froehlich et al., 2001). Previously described cDNAs for AtAOS and AtHPL (Laudert et al., 1996;Froehlich et al., 2001) were used for coupled in vitro transcription/translation in wheat germ extracts. The 35 S-methionineor 14 C-leucine-labeled proteins were then added to Arabidopsis chloroplasts that had been isolated by Percoll/sucrose density gradient centrifugation. Import experiments revealed that both 35 S-AOS and 35 S-HPL were taken up by isolated chloroplasts and processed by cleavage of their transit peptides ( Fig. 2A). The resulting mature enzymes were targeted to different locations ( Fig. 2B-D): whereas 35 S-AOS was targeted to the inner envelope of chloroplasts, 35 S-HPL was localized in the outer envelope fraction ( Fig. 2A, B). Western blotting using monospecific antibodies to AOS and HPL ( Supplementary Fig.  S1) as well as GFP tagging confirmed the differential localization of the endogenous AOS and HPL in planta ( Fig. 2C; Supplementary Fig. S2). The in vitro-imported 35 S-AOS and 35 S-HPL were degraded by added trypsin, but only 35 S-HPL was sensitive to thermolysin ( Fig. 2A, B). Thermolysin is a protease that degrades surface-exposed proteins of the outer plastid envelope, whereas trypsin is known to access the intermembrane space between the outer and inner envelope (Cline et al., 1984;Kessler and Blobel, 1996). Thus, AOS is likely to face the intermembrane space, as it was not degraded by thermolysin, whereas HPL is exposed to the cytosolic side of the outer plastid envelope and was, thus, sensitive to thermolysin. Western blot experiments revealed the co-localization of both the imported 35 S-labeled AOS and endogenous AOS with two translocon proteins of the inner chloroplast envelope, TIC55 and TIC110, and that of the imported 35 S-labeled HPL and endogenous HPL with the translocon protein of the outer chloroplast envelope, TOC75 (Fig. 2C). Using TIC110 as a (Try)-treated (+Try) and untreated (-Try) mixed outer and inner plastid envelopes (ME), inner plastid envelope (IM), outer plastid envelope (OM), thylakoids (Th), and stroma (St). CP, Chloroplast reference fraction prior to import and protease treatment. The western blot in panel c shows the levels of the inner chloroplast envelope translocon protein TIC110 and the outer chloroplast envelope protein TOC75 in non-trypsin-treated chloroplasts (CP) versus trypsintreated chloroplasts containing imported 35 S-AOS/HPL and respective subfractions. Signal detection was made with an enhanced chemiluminescence (ECL) system. (C) Western blot analysis of the endogenous AOS (AOS endo ) and HPL (HPL endo ) (predicted molecular masses 54.5 kDa and 51.3 kDa, respectively; panel a) compared with TOC75, the translocon at the inner chloroplast envelope membrane protein TIC55, the chlorophyll a/b binding protein LHCII (CAB), and the small subunit of ribulose-1,5-bisphosphate carboxylase/oxygenase (SSU) (panel b) in the indicated fractions of non-trypsintreated chloroplasts. Signal detection was made with either ECL or alkaline phosphatase-5-bromo-4-chloro-3-indolyl phosphate/nitro blue tetrazolium (NBT)-based systems, as indicated. (D) Membrane binding of imported 35 S-AOS (a) and 35 S-HPL (b), as assessed by their extractability by 1 M NaCl and 0.1 M Na 2 CO 3 , pH 11. Both pellet (P) and supernatant (S) fractions, obtained after sedimentation of the membranes, were tested by SDS-PAGE and autoradiography for the two labeled proteins. (This figure is available in colour at JXB online.) diagnostic marker, any post-import degradation of 35 S-AOS or 35 S-HPL could be excluded (Fig. 2B). TIC110 is a very sensitive marker for monitoring the trypsin digestion procedure and is easily degraded if no precautions are taken, such as the inclusion of a trypsin inhibitor during all steps of chloroplast fractionation (Jackson et al., 1998). Extraction of outer and inner plastid envelopes with 1 M NaCl or 0.1 M Na 2 CO 3 , pH 11, revealed that 35 S-HPL and 35 S-AOS were tightly bound to their respective target membranes (Fig. 2D). The localizations of HPL and AOS confirmed previous data for chloroplasts from Arabidopsis (Joyard et al., 2010;Mwenda et al., 2015) and tomato (Froehlich et al., 2001), respectively, while not precluding divergent locations in other plant species (Farmaki et al., 2007).

Isolation of plastid envelope proteins interacting with AOS and HPL
We next attempted to identify proteins interacting with AOS in the envelope of Arabidopsis chloroplasts. C-terminal, (His) 6tagged AOS [AOS-(His) 6 ] was expressed in bacteria and purified to apparent homogeneity by Ni-NTA chromatography. The chemically pure protein was incubated with isolated chloroplasts in standard import reactions containing 2.5 mM Mg-ATP (Froehlich et al., 2001). After incubation, mixed outer and inner envelopes were isolated from ruptured chloroplasts (Li et al., 1991;Schnell et al., 1994) and solubilized with 1.3% decyl maltoside (Caliebe et al., 1997). Proteins adhering to AOS-(His) 6 were subjected to size exclusion chromatography and AOS-(His) 6 was detected with AOS-specific or (His) 6 antibodies (Laudert et al., 1996;Froehlich et al., 2001). For comparison, AOS-containing plastid envelope complexes were isolated from transgenic plants overexpressing AOS-FLAG under the control of the 35S cauliflower mosaic virus promoter. Figure 3A and B depict the elution profiles of AOS after its purification from either isolated chloroplasts following the in vitro import reaction of AOS-(His) 6 or transgenic plants overexpressing AOS-FLAG. Whereas AOS-FLAG and AOS-(His) 6 were present in higher molecular mass complexes of ~250 kDa (peak I), a second, lower molecular mass complex was detectable only for AOS-FLAG (peak II). The lack of a corresponding AOS-(His) 6 peak may be due to the lower abundance of this protein in the in vitro uptake assays with isolated chloroplasts. In control experiments using protein extracts from transgenic plants overexpressing HPL-FLAG, a single, low molecular mass peak was seen, the elution of which was virtually the same as that observed for AOS-FLAG in peak II ( Supplementary Fig.  S3). We assumed that the AOS-FLAG and HPL-FLAG contained in the low molecular mass peak represent enzyme monomers not interacting with other plastid envelope proteins. When the total pattern of proteins that co-purified with AOS-(His) 6 in peak I was analyzed by SDS-PAGE, at least 10 bands were found (Fig. 3C, panel a, lane 1); of these, besides the bait protein, the ~100 kDa band was identified by protein sequencing as LOX2 and the ~72 kDa and ~24 kDa bands were identified as AOC2, forming SDSresistant trimers and monomers, respectively. Western blotting using monospecific antibodies to Arabidopsis LOX2 ( Supplementary Fig. S4) confirmed the identity of the ~100 kDa band as LOX2 (Fig. 3C, panel b). When similar experiments were carried out with HPL-(His) 6 , a completely different protein pattern was obtained, which, most notably, did not contain any of the proteins detected with AOS-(His) 6 (Fig. 3C, panel a, lane 2).

Genetic evidence for the existence of a chloroplast envelope complex comprising AOS, LOX2, and AOC2
Because the expression of LOX2, AOS, and AOC2 in Arabidopsis is under the control of the microRNA 319 (miR319)-regulated clade of TCP (TEOSINTE BRANCHED/CYCLOIDEA/ PCF) transcription factor genes (Schommer et al., 2008), we used the jaw-D mutant with strongly reduced mi319dependent expression of JA biosynthetic genes to investigate the interaction of LOX2, AOS, and AOC2 genetically. When jaw-D plants, in which TCP4 expression is strongly down-regulated, and primary transformants expressing a miR319-resistant version of TCP4 (referred to herein as TCP4 plants) that dominantly regulates LOX2 expression (Palatnik et al., 2003;Schommer et al., 2008) were grown for 14 days in a greenhouse, not much difference in phenotype was seen. However, phenotypic differences were more pronounced at later stages of plant development and especially when plants matured and entered senescence. In agreement with previous results (Schommer et al., 2008), TCP4 plants displayed a marked acceleration of senescence, whereas jaw-D plants exhibited a considerable (2-week) delay in leaf senescence. These differential changes correlated with altered patterns of expression of the plastid envelope proteins LOX2, AOS, and AOC2 ( Fig.  4; see also Supplementary Fig. S13, which shows respective densitometric analyses using ImageJ software, https://imagej. net/). TCP4 plants more rapidly accumulated all three JA biosynthesis enzymes than wild-type plants (Fig. 4A).
JA biosynthesis is under the control of a feed-forward loop and consequently can be boosted by exogenously added JA (Wasternack and Hause, 2013;Yan et al., 2013). To pinpoint the role of LOX2, AOS, and AOC2 in this loop, JA-deficient seedlings of the aos mutant (Park et al., 2002;von Malek et al., 2002) and JA-Ile-insensitive seedlings of the coi1 mutant (Feys et al., 1994) were used. For comparison, we used Arabidopsis mutants defective in the LOX2 and AOC2 genes, respectively (Seltmann et al., 2010;Stenzel et al., 2012). When the patterns of expression of plastid envelope proteins were compared, some differential effects were observed. As displayed in Fig. 4B, the aoc2 knockout mutants contained wild-type levels of AOC1, AOC3, and AOC4, as well as wild-type levels of AOS and LOX2. By contrast, aos and lox2 mutants expressed reduced amounts of LOX2 and AOC2 trimers, while retaining the wild-type levels of AOC1-4 monomers. These findings were suggestive of partially overlapping roles and expression patterns of the four AOC family members (see Supplementary  Figs S5 and S6), with AOC1, AOC3, and AOC4 presumably being capable of replacing AOC2 and forming homo-and heterocomplexes. It has recently been demonstrated that all four AOCs display identical catalytic activities (Otto et al., 2016).
To further dissect the feed-forward loop operating in JA biosynthesis, seedlings of the JA-insensitive coi1 mutant (Feys et al., 1994) and the JA-resistant jar1 mutant (Staswick et al., 1992) were used. JAR1 encodes an enzyme that provides the physiologically active JA-Ile conjugate for CORONATINE INSENSITIVE 1 (COI1)-dependent signal transduction (Staswick et al., 1992;Feys et al., 1994). coi1 and jar1 seedlings were grown for 14 days in continuous white light and then sprayed with MeJA. MeJA treatment triggered the expression of LOX2, AOS, and AOC1-4 in wild-type plants but not in jaw-D, coi1, and jar1 plants (Fig.  4C). Hence, JA production was required and relied on an intact JA signal transduction chain comprising JA-Ile and COI1, linking JA signal with miR319-regulated TCPs.

Probing the interaction of LOX2, AOS, and AOC2 through cross-linking
The presence of higher molecular mass complexes containing LOX2, AOS, and AOC2 in the envelope of Arabidopsis chloroplasts suggested the possibility that the enzymes may interact physically and functionally in JA precursor biosynthesis. As a first step to test this hypothesis, cross-linking experiments were conducted using bacterially expressed, (His) 6 -tagged, purified precursors that had been derivatized with 125 I-APDP, which is a heterobifunctional, photoactivatable, and cleavable label-transfer cross-linker frequently used in the chloroplast protein import field (e.g. Ma et al., 1996). 125 I-APDP contains a 21 Å spacer arm that, depending on its location on the bait protein, can penetrate to different extents into the respective target membrane and label proteins and lipids. 125 I-APDP-labeled LOX2, AOS, and AOC2 were imported into isolated Arabidopsis chloroplasts in darkness (Springer et al, 2016;Fig. 2, and Supplementary Fig. S7). After incubation, labeltransfer cross-linking was induced by exposure to ultraviolet light with the chloroplasts on ice. Envelopes were then isolated from ruptured chloroplasts and solubilized with 3% SDS (Ma et al., 1996), and 125 I-APDP-labeled proteins were detected by SDS-PAGE and autoradiography. As shown in Fig. 5A, different banding patterns were revealed for the three labeled proteins. LOX2 and AOC2 (which was present as a trimer) both interacted with AOS, with little or no direct interaction between the two of them. For 125 I-APDP-AOS, label transfer occurred on to LOX2 and AOC2 ( Fig. 5A; see also the densitometric scans in Supplementary Fig. S13). Interestingly, 125 I-APDP-AOS gave rise to both 125 I-labeled AOC2 trimers and monomers (Fig. 5A). AOC2 monomers were also seen in assay mixtures containing 125 I-APDP-LOX2. This observation could suggest a shuffling of monomers and trimers of AOC2 into the envelope complex. On the other hand, the nature of the chosen cross-linker and its topology on the bait protein may explain this result. Last but not least, it is also possible that substrate binding and conversion affected the interaction of LOX2, AOS, and AOC2, and thereby had an impact on the cross-linking results.

Interaction of LOX2, AOS, and AOC2 probed in the split-ubiquitin system
As an independent approach to confirm the interaction of LOX2, AOS, and AOC2, we carried out split-ubiquitin yeast two-hybrid screens for membrane-bound proteins. Yeast cells Fig. 6. Genetic dissection of AOS-LOX2-AOX2 interactions in the split-ubiquitin system. Experiments were performed using fusions to the C-terminal (Cub) and N-terminal (NubG) halves of ubiquitin. Alg5-NubI, the fusion of the unrelated endoplasmic reticulum membrane protein Alg5 to NubI, was used as positive control. Negative controls were fusions of Alg5 to Cub (Alg5-Cub) or NubG (Alg5-NubG). Qualitative estimation of interactions was deduced from the growth behavior of co-transformed yeast cells on solidified minimal synthetic defined (SD) base medium supplemented with -Leu/-Trp/-His dropout supplements (Clonetech). Yeast strains were plated to an OD 600 of 0.8 (left colony) and at 1:1 dilution (right colony). For quantitative assays the β-galactosidase activity of cells grown in SD liquid culture containing -Leu/-Trp dropout supplements was analyzed.
(strain DSY1) were first transformed with AOS-and LOX2containing bait vectors (AOS-Cub, LOX2-Cub), which express the bait proteins fused to the C-terminal part of ubiquitin. For each bait protein, we constructed and tested three different vectors (pAMBV4, pCMBV4, and pTMBV4), providing promoters of different strength to drive bait protein expression. In our experiment, the pTMBV4 vector, which contains a highly potent TEF1 promoter, gave the clearest results. Co-transformation was accomplished in a second, separate step in which AOS-and AOC2-containing prey vectors were introduced into the selected yeast cells (NubG-AOS, NubG-AOC2). As prey vectors, we used constructs that added the N-terminal part of ubiquitin to the N-terminal extremity of the prey proteins (pADSL-Nx). In addition to the specific constructs used to study potential LOX2-AOS-AOC2-interactions, we used control vectors provided with the system to test for non-specific activation and autoactivation (NubG-Alg5), as well as positive (Alg5-Cub, NubI-Alg5) and negative (Alg5-Cub, NubG-Alg5) system controls. The co-transformed yeast cells were subjected to qualitative Hiscomplementation growth tests carried out on SD plates supplemented with 5 mM 3-amino-1,2,4-triazole and lacking the amino acids Leu, Trp, and His. Additionally, the relative β-galactosidase activities of the investigated co-transformed yeast strains were determined to provide quantitative data on the strength of interaction of the analyzed protein pairs. According to the results summarized in Fig. 6, interactions were observed between AOS and AOC2, with a fraction of AOS presumably forming dimers, as found for AOS from Parthenium argentatum Li et al., 2008). Some interactions could also be detected for LOX2 with AOC2 and AOS, although the strengths of these interactions were dependent on the respective bait and prey protein and presumably reflected the different membrane compositions of yeast cells versus higher plant chloroplast envelopes.

Sequestration of jasmonate precursor biosynthesis through a protein complex in the plastid envelope
In the present study, evidence is provided for the existence of a protein complex involved in JA precursor biosynthesis in Arabidopsis chloroplasts. We show that LOX2, AOS, and AOC2, the enzymes that catalyze consecutive steps in JA precursor biosynthesis (Fig. 1), are co-localized in the inner envelope of Arabidopsis chloroplasts ( Fig. 2 and Supplementary  Fig. S7; see also Springer et al., 2016), and form complexes operating in cis-(+)-12-OPDA synthesis from α-LeA (Figs 3 and 5; Supplementary Fig. S8). It was possible to reconstitute a similar complex in vitro from soluble, catalytically active LOX2, AOS, and AOC2 enzymes and isolated plastid envelope lipids ( Supplementary Fig. S8). In either case, LOX2, AOS, and AOC2 were present in a 1:1:4 stoichiometry, indicating that these complexes may contain both AOC2 monomers and trimers. The significance of this observation and the possibility of activity regulation through heterotrimerization of AOC2 with AOC1, 3, and 4 remains to be established. It must be noted in  Youn et al. (2006), for AOS2 on Lee et al. (2008), and for AOC2 on Hofmann et al. (2006), and transmembrane domain predications are based on TMpred for LOX2 and (AOC2) 3 . this context, however, that Zerbe et al. (2007) also used AOS and AOC2 at a 1:4 molar ratio. Interaction studies in the splitubiquitin yeast two-hybrid system confirmed the interaction of AOS and AOC2 (Fig. 6). When isolated plastid envelope complexes were supplied with α-LeA, only cis-(+)-12-OPDA accumulated in the reaction medium (Fig. 5). No evidence was obtained for the release of significant amounts of HPOT or EOT and its short-lived disintegration products (α-ketols and γ-ketols), as well as of racemic OPDA, into the reaction mixture. These findings are suggestive of a tight channeling of metabolites through both the isolated and the reconstituted protein complexes. By virtue of the observed channeling of metabolites, the dilution of the reaction intermediates was kept low. On the other hand, the observed channeling of metabolites greatly increased the rate of formation and, thus, the yield of cis-(+)-12-OPDA from α-LeA. Compared with the free-enzyme assay, 120-fold higher activities were measured in the isolated envelope complex. Our data confirm and extend previous findings by Zerbe et al. (2007), who reconstituted cis-(+)-12-OPDA synthesis from 13-HPOT by combining purified recombinant AOS and AOC2 in vitro. The authors found that both soluble and matrix-bound enzymes are active and that their co-fixation on a solid matrix increased the yield of cis-(+)-OPDA from 13-HPOT by ~50%. In contrast to those studies, however, in which AOS and AOC2 were randomly bound to the matrix, thus excluding tight substrate channeling, in our experiments the enzymes were associated in an orderly fashion, thereby allowing the channeling of α-LeA. Because vernolic acid failed to inhibit cis-(+)-12-OPDA synthesis from α-LeA or 13-HPOT in the native and reconstituted complexes ( Fig. 5; Supplementary Fig. S8), we conclude that the active site of AOC2 was largely inaccessible to the inhibitor. On the other hand, exogenously administered 13-HPOT was not accepted as substrate for cis-(+)-12-OPDA synthesis by the native and reconstituted complexes, but it was accepted in the free-enzyme assay (Fig. 5C). On the basis of these results we conclude that synthesis of cis-(+)-12-OPDA from α-LeA is strictly compartmentalized, presumably to prevent competing side reactions, such as that from 13-HPOT catalyzed by HPL or by chemical decay, which give rise to αand γ-ketols. By this means, plants avoid the costly formation of n-hexenal and 12-oxo acids implicated in direct and indirect defenses against herbivores (Wu and Baldwin, 2010;Hoffmann et al., 2011), while maintaining the capacity to respond to biotic challenges and abiotic stresses (Wasternack and Hause, 2013;Yan et al., 2013).

Structural modeling of the LOX2-AOS-AOC2 envelope complex
We conclude from our results that AOS forms complexes with both LOX2 and AOC2 in planta as well as in vitro. Molecular modeling was carried out to obtain insight into the potential structure of this complex, using published X-ray structure data for soybean lipoxygenase L3 (PDB ID 1LNH; Skrzypczak-Jankun et al., 1997;Youn et al., 2006), AOS (D 3CLI; Lee et al., 2008), and AOC2 (PDB ID 2GIN; Hofmann et al., 2006), obtained at 2.60 Å, 1.80 Å, and 1.80 Å resolution, respectively, as templates (Supplementary Figs S9-S11). First, homology modeling was done for LOX2, based on the X-ray structure of soybean LOX L3 (Cho and Stahelin, 2006;Youn et al., 2006). The established LOX2 structure of Arabidopsis ( Supplementary Fig. S9) was then used in modelling of the whole LOX2-, AOS-, and AOC2-containing complex. The top rank model (depicted in Fig. 7) suggests that at least two amino acid residues, Ser92 and Gly94, of LOX2 are potentially involved in the LOX2-AOS interaction, forming hydrogen bonds with Ser272 of AOS ( Supplementary Fig. S12A). The interacting Ser and Gly residues of LOX2 are not conserved between LOX2 and soybean LOX3 and reside in a loop region between β-sheets 1 and 2 of LOX2. Similarly, one region of interacting amino acids could be identified for each of the hypothetical LOX2-AOC2 and AOS-AOC2 complexes. In these complexes, Asp96 of LOX2 is predicted to interact with Asn42 of AOC2, and Phe44 and Ser45 of AOC2 were found when analyzing these enzymes apart from LOX2, while they were not seen in the modeled LOX2-AOS-AOC2 complex ( Supplementary Fig. S12B).
Details of how the LOX2-AOS-AOC2 complex may be bound to the membrane remain elusive. The fact that a functional complex could be reconstituted from soluble LOX2, AOS, and AOC2 molecules at first glance suggests that the lipid bilayers might not be required for the formation of the whole complex. Nevertheless, all three enzymes attained saltand protease-resistant states after respective in vitro import reactions with chloroplasts, suggesting that they are tightly bound to the lipid bilayers of the inner plastid envelope. Hydrophobic transmembrane domains may anchor LOX2, AOS, and AOC2 in the membrane. Lee et al. (2008) identified several non-polar detergent-binding α-helices in AOS, of which those designated α-F and α-G were tentatively defined as putative transmembrane domains in our structural models ( Supplementary  Fig. S10A). Predictions made with TMpredn, however, suggest that a different part of the AOS polypeptide forms a single transmembrane domain and that this overlaps with some of the active site residues (Supplementary Fig. S10B). Such an overlap would resemble that in the AOC2 trimer, where some active site residues appear to be embedded into, or are part of, the predicted transmembrane domain ( Supplementary Fig. S11). However, it cannot be excluded that some of the hydrophobic β-sheets forming the characteristic β-barrel cavity are involved in membrane binding ( Supplementary Fig. S11). Lastly, the LOX structure of Skrzypczak-Jankun et al. (1997) used for modelling LOX2 from Arabidopsis ( Supplementary Fig. S9) suggests a hydrophobic environment of the catalytic pocket that could overlap with the predicted transmembrane domain. Further work is needed to resolve the structure of the LOX2-AOS-AOC2 complex and elucidate its membrane binding.
AOC2 and the other members of the AOC family easily form homo-and heterotrimers (Hofmann et al., 2006;Hofmann and Pollmann, 2008;Otto et al., 2016). As described by Otto et al. (2016), trimerization of AOC isoenzymes is likely to contribute to regulation of their activity. Several salt bridges between monomers and a hydrophobic core within the AOC2 trimer were identified and functionally proven by site-directed mutagenesis. While Lys152 of one monomer and Glu128 of the neighboring monomer established the observed salt bridges, the amino acids involved in building the hydrophobic core of the trimer comprised Leu40, Leu50, Leu53, and Ile79 of all three monomers (Otto et al., 2016). Notable are the conservation of interacting amino acids and the overlapping expression patterns of AOC1/2 during germination, leaf production, rosette growth, inflorescence emergence, and flowering, both with each other and with the respective expression patterns of LOX2 and AOS ( Supplementary Figs S4-6). By contrast, the expression pattern of AOC3 and AOC4 is more distinct (Supplementary Figs S4-6). On the other hand, HPL expression is comparably low in all of the developmental stages analyzed (Supplementary Fig. S6). Together, these data are suggestive of a co-evolution of mechanisms that allow JA precursor biosynthesis to be favored over the production of volatile compounds during plant development. This situation obviously changes when plants are challenged by chewing insects and both direct and indirect defenses are triggered through the HPL pathway (Bate and Rothstein, 1998;Blée, 1998;Croft et al., 1993;Wu and Baldwin, 2010). In functional terms, both AOS and HPL belong to the same superfamily of non-canonical cytochrome P450 enzymes, but their amino acid sequences are distinct enough to permit their unique reaction mechanisms (Lee et al., 2008). While AOS interacts with LOX2 and AOC2, no interacting partners could be identified for HPL (Fig. 3). Finally, the chloroplast localization of AOS and HPL is quite distinct and helps to assure that AOS operates in the channeling of α-LeA to OPDA, whereas HPL drives volatile production.
In agreement with previous publications (Spivey and Ovádi, 1999;Gilbert et al., 2008), we conclude that compartmentalization of enzymes as well as organization into multi-protein complexes provides a highly specific cellular mechanism for controlling the flow of metabolites through key regulatory pathways and preventing unfavorable competing reactions. Work is in progress to obtain X-ray structural data for the identified LOX2-AOS-AOC2 complex from higher plant chloroplasts.

Supplementary data
Supplementary data are available at JXB online.