Indole-3-glycerolphosphate synthase, a branchpoint for the biosynthesis of tryptophan, indole, and benzoxazinoids in maize

The maize (Zea mays) genome encodes three indole-3-glycerolphosphate synthase enzymes (IGPS1, 2, and 3) catalyzing the conversion of 1-(2-carboxyphenylamino)-l-deoxyribulose-5-phosphate to indole-3-glycerolphosphate. Three further maize enzymes (BX1, benzoxazinoneless 1; TSA, tryptophan synthase α subunit; and IGL, indole glycerolphosphate lyase) convert indole-3-glycerolphosphate to indole, which is released as a volatile defense signaling compound and also serves as a precursor for the biosynthesis of tryptophan and defense-related benzoxazinoids. Phylogenetic analyses showed that IGPS2 is similar to enzymes found in both monocots and dicots, whereas maize IGPS1 and IGPS3 are in monocot-specific clades. Fusions of yellow fluorescent protein (YFP) with maize IGPS enzymes and indole-3-glycerolphosphate lyases were all localized in chloroplasts. In bimolecular fluorescence complementation assays, IGPS1 interacted strongly with BX1 and IGL, IGPS2 interacted primarily with TSA, and IGPS3 interacted equally with all three indole-3-glycerolphosphate lyases. Whereas IGPS1 and IGPS3 expression was induced by insect feeding, IGPS2 expression was not. Transposon insertions in IGPS1 and IGPS3 reduced the abundance of both benzoxazinoids and free indole. Spodoptera exigua (beet armyworm) larvae show improved growth on igps1 mutant maize plants. Together, these results suggest that IGPS1 and IGPS3 function mainly in the biosynthesis of defensive metabolites, whereas IGPS2 may be involved in the biosynthesis of tryptophan. This metabolic channeling is similar, though less exclusive than that proposed for the three maize indole-3-glycerolphosphate lyases.

To date, the only plant IGPS enzymes that have been characterized are AT2G04400 and AT5G48220 in Arabidopsis thaliana (Arabidopsis; Li et al., 1995a). AT2G04400 enzymatic activity was verified by complementing an Escherichia coli mutant strain containing a missense mutation in the trpC gene, which encodes a bifunctional IGPS/N-(5´-phosphoribosyl)anthranilate isomerase (PRAI) enzyme (Yanofsky et al., 1971, Li et al., 1995a. Based on the complementation assays using E. coli mutants deficient in either the IGPS or the PRAI activity of TrpC, it appears that plants only have monofunctional IGPS enzymes (Li et al., 1995a, Li et al., 1995b. Indole-3-glycerolphosphate synthase activity was also verified in a wheat tissue extract, though the actual enzyme was not functionally characterized (Singh and Widholm, 1974).
Thus, the pathways for the biosynthesis of tryptophan and tryptophan-independent indolecontaining metabolites may be separated in Arabidopsis subcellular compartments.
Three maize genes, BX1 (benzoxazinoneless 1), TSA (tryptophan synthase α subunit) and IGL (indole glycerolphosphate lyase), encode enzymes that convert IGP to indole (Figure 1; Kriechbaumer et al., 2008). TSA-GFP fusion proteins were chloroplast-localized in maize protoplast expression expriements (Kriechbaumer et al., 2008) and assays with spinach chloroplasts expressing maize BX1 also indicated a chloroplastic location for this protein (Stettner, 1998). A fourth protein, designated as TSA-like, was localized to the cytosol and did not have indole biosynthesis activity in vitro (Kriechbaumer et al., 2008). BX1, TSA, and IGL genes can be differentiated by both their expression levels and their relative contributions to the biosynthesis of tryptophan, benzoxazinoids, or free indole. Although BX1 is constitutively expressed in young maize seedlings, its expression also is induced by caterpillar feeding (Maag et al., 2016. Similarly, IGL expression is upregulated in response to herbivory, consistent with the observed herbivory-induced increase in maize indole emission (Frey et al., 2000). Mutations in IGL, but not BX1, led to an indole-deficient phenotype (Ahmad et al., 2011).
Conversely, only very low amounts of benzoxazinoids can be detected in bx1 mutant maize lines (Hamilton, 1964, Melanson et al., 1997, Betsiashvili et al., 2015, indicating that the indole produced by IGL and/or TSA is not easily incorporated into benzoxazinoids.
Nevertheless, exogenous indole application to bx1 mutant maize plants leads to complementation of the benzoxazinoid-deficient phenotype (Frey et al., 1997, Ahmad et al., 2011. Maize TSA forms a complex with TSB1 in vitro. Partial purification and proteomic analysis of the 160 kD tryptophan synthase complex from maize leaves identified TSA and TSB proteins, but not BX1, IGL, or TSA-like (Kriechbaumer et al., 2008). To date, no maize TSA mutation has been reported, but double mutations in the tryptophan synthase ß-subunit genes TSB1 and TSB2 led to a seedling-lethal phenotype with high indole accumulation (Wright et al., 1992). Maize plants with double mutations in BX1 and IGL1 are viable, indicating that neither of these enzymes makes an essential contribution to tryptophan biosynthesis (Ahmad et al., 2011) Similar to IGP catabolism by the BX1, TSA, and IGL protein products, IGP synthesis was predicted to be encoded by three maize genes, IGPS1, IGPS2, and IGPS3. Here we report experiments to confirm the function of maize IGPS enzymes and determine whether they represent an earlier branch point in the indole biosynthesis pathway. Bimolecular fluorescence complementation (BIFC) assays show differences in the interaction profiles. For instance, whereas IGPS1 strongly interacts with BX1 and IGL but not with TSA, IGPS2 has a stronger interaction with TSA. However, the analysis of IGPS1 and IGPS3 mutant plants suggests that the metabolic channeling of IGP is not as strong as that of indole produced by BX1, TSA, and IGL.
Figure 2). Maize IGPS2 was in a cluster with both monocot and dicot enzymes, including the two Arabidopsis enzymes with confirmed IGP synthesis activity (Zhang et al., 2008). By contrast, IGPS1 and IGPS3 were in separate clades that contain only monocot enzymes.
Phylogenetically, the maize IGPS enzymes were most similar to those of a close relative, S.
Aligning the B73 alleles of the three IGPS genes with the genomes of the 26 founder lines of the maize Nested Association Mapping (NAM) population (McMullen et al., 2009b; www.gramene.org) showed high sequence conservation among members of this enzyme family, with 96-100% identity in the predicted protein sequence (Supplementary Table S2). The only exception was inbred line M162W, where IGPS2 was partially aligned to two adjacent genes, which may result from an incorrect annotation of the M162W genome. The chromosome positions of the three IGPS genes were conserved in the 26 maize genome assemblies, with the inbred line Tzi8 being the exception, having a second, almost-identical, IGPS1 gene in an assembled scaffold that was not associated with a chromosome. This is possibly an assembly error, as none of the other NAM founder lines genomes had any additional predicted IGPS genes.
The IGPS enzyme family is known for catalyzing a carbon-carbon ring-forming reaction in which the carbon atoms become covalently bonded (Wierenga, 2001). The three maize IGPS proteins have the characteristic domains of this enzyme class, including a TIM (triosephosphate isomerase)-like beta/alpha barrel domain, which is one of the most common enzyme folds (Supplemental Figure S1; Chan et al., 2017;Nagano et al., 2002). The TIM barrel motif includes a phosphate-binding domain, which is a conserved protein fold consisting of eight α-helices and eight parallel β-strands that alternate along the peptide backbone (Farber and Petsko, 1990, Brändén, 1991, Wilmanns et al., 1991. The well-organized scaffold contains predicted binding sites for indole, phosphate, ribulose/trilose, and 1-(2-carboxyphenylamino)-1-deoxy-D-ribulose-5-phosphate (Supplemental Figure S1).
To confirm the enzymatic activity of IGPS1, IGPS2, and IGPS3, we complemented the tryptophan-auxotrophic phenotype of the E. coli trpC9800strain (Yanofsky et al., 1971), which is deficient in the IGPS but not the PRAI enzymatic activity of TrpC. The coding regions of the three maize IGPS genes, including the signal sequences, were cloned into E. coli plasmid vectors for expression in E. coli trpC9800. Whereas the empty vector control strain grew only when the minimal medium was supplemented with tryptophan, all three maize IGPS genes complemented the tryptophan-auxotrophic phenotype ( Figure 3).

Maize IGPS1 and IGPS3 expression is induced by herbivory
Previous experiments showed that 24 h of S. exigua caterpillar feeding on maize inbred line B73 resulted in a 100-fold induction of IGPS1 expression (p = < 0.001, Dunnett's test) and a 5-fold induction of IGPS3 (p = < 0.001, Dunnett's test), but no significant changes in IGPS2 , Wisecaver et al., 2017Supplemental Figure S2A). To determine whether the gene expression response is similar in the W22 inbred line, we infested maize seedlings with S. exigua for 24 h and measured the transcript levels by quantitative real-time PCR (qRT-PCR). Similar to the previous results with inbred line B73 (Supplemental Figure S2A), this showed significant induction of IGPS1 (p < 0.001, t-test) and IGPS3 (p < 0.001, t-test) transcripts ( Figure 4A) and no significant transcriptional changes for IGPS2. Methyl jasmonate treatment, which regulates many insect defense responses in plants (Howe and Jander 2008;Martin et al., 2002;, also increased IGPS1 and IGPS3 but not IGPS2 expression ( Figure 4B). We measured IGPS expression by 3'RNAseq (Kremling et al., 2018) in a time course with 0 h, 2 h, 8 h, and 24 h of R. maidis feeding. IGPS1 expression showed a significant increase after 8 h of aphid feeding W22 (p < 0.05, t-test; Figure 4C), which is comparable to the significant induction at 4 h after R. maidis infestation in inbred line B73 (Tzin et al., 2015; Figure S2B). Although IGPS3 expression was not significantly changed in response to R. maidis feeding, the Illumina 3'RNAseq read counts of IGPS3 (in average: 188) in W22 plants were higher than those of IGPS1 (average: 2.5) and IGPS2 (average: 21), even in the absence of insect feeding. These results are different from the 3'RNAseq results with inbred line B73, which showed an average of 30 read counts for IGPS1, 18 read counts for IGPS3, and 12 read counts for IGPS2 (Tzin et al., 2015). Given that benzoxazinoids are both constitutively synthesized and induced by insect feeding, both IGPS1 and IGPS3 could function in W22 benzoxazinoid biosynthesis. In contrast to IGPS1 and IGPS3, IGPS2 expression was unchanged or even slightly decreased by insect feeding in both W22 and B73 (Figure 4; Figure S2; Tzin et al., 2015.

Transposon insertions in IGPS1 and IGPS3 reduce benzoxazinoid levels
A search of publicly available transposon insertions (McCarty et al., 2005, Settles et al., 2007, Williams-Carrier et al., 2010 identified mu-02540, a likely IGPS1 knockout line with a Mu insertion in the second exon (Chr.7: 82883253 bp to 82883261 bp; maize W22 genome v2). We used quantitative reverse transcriptase-PCR (QRT-PCR) to compare homozygous, heterozygous, and wildtype plants after methyl jasmonate induction ( Figure 5A). Heterozygous plants showed lower expression levels than wildtype, and no IGPS1 transcripts above background levels could be amplified from the homozygous mutant plants. The expression of IGPS3, but not IGPS2, was significantly increased in the igps1 mutant line ( Figure 5B).
We identified a likely IGPS3 knockout mutation in the Mu-Illumina collection (mutant line 228894.6; chr10: 69445669 -69446382; Williams-Carrier et al., 2010). Plants were induced with methyl jasmonate to activate IGPS3 transcription, and qRT-PCR was used to verify homozygous mutants and wildtype plants. IGPS3 expression was significantly reduced (p < 0.001) in the mutant line, but the other two IGPS genes showed no significant changes in expression levels ( Figure 5C). and igps3 mutant lines ( Figure 6C). By contrast, there were no significant changes in the tryptophan content in either of the mutant lines relative to wildtype W22 (p > 0.05, t-test; Figure   6D,E).
Spodoptera exigua is a generalist lepidopteran herbivore that grows better on bx1::Ds and bx2::Ds mutant plants, which are almost completely devoid of benzoxazinoids, than on wildtype W22 maize . We conducted bioassays to determine whether the benzoxazinoid decrease in the homozygous igps1 mutant line ( Figure 6A) also affects S. exigua caterpillar growth. There was a significant increase in caterpillar body mass, with an average weight of 20 mg on homozygous igps1 mutant plants compared to 13.6 mg average weight on wildtype plants after nine days of feeding (p < 0.05, t-test; Figure 6F). We did not expect to see a measureable effect on caterpillar growth due to the relatively small decrease in benzoxazinoids in the igps3 mutant ( Figure 6B) and therefore did not conduct experiments with this line.

IGPSs enzymes are localized in maize chloroplasts
In addition to gene expression levels, another factor that can influence the function of IGPS enzymes is their sub-cellular localization. The TargetP 1.1 software program (http://www.cbs.dtu.dk/services/TargetP/), which predicts subcellular localization and cleavage sites for signal peptides, indicated a chloroplast location for all three maize IGPS proteins. To protoplasts showed that YFP by itself was located in the cytosol (Richter et al., 2016).

Interactions of IGPS and indole-3-glycerolphosphate lyases
To determine whether the three IGPS enzymes interact with BX1, TSA, and IGL, we performed BiFC interaction studies for all nine possible two-protein combinations. IGPS1, IGPS2, and IGPS3 were cloned into p2YC (Yang et al., 2007), such that their C-termini were fused to a Cterminal YFP fragment (amino acids 158-238). The open reading frames for BX1, TSA, and IGL were cloned into p2YN (Yang et al., 2007), fusing an N-terminal YFP fragment (amino acids 1-159) at the C-termini. All constructs were transformed into Agrobacterium tumefaciens and infiltrated into Nicotiana benthamiana plants in pairs. Two days after infiltration, protoplasts from the infiltrated leaf area were isolated and analyzed by confocal microscopy. Strong YFP signals were localized in the chloroplasts for the combinations of IGPS1-C/Bx1-N and IGPS1-C/IGL-N ( Figure 8A). On the other hand, coexpression of IGPS1-C and TSA-N did not result in reconstitution of YFP fluorescence. By contrast, coexpression of IGPS2 with TSA resulted in a strong fluorescence signal, and weaker signals were observed when IGPS2 was coexpressed with BX1 or IGL ( Figure 8B). IGPS3 consistently showed reconstitution of the fluorescence when coexpressed with all three indole-3-glycerolphosphate lyase partners (BX1, IGL, and TSA; Figure   8C). Additionally, by cloning IGPS2 and IGPS3 cDNA sequences into p2YN, we were able to test whether IGPS1 could interact with the other two IGPS enzymes. The co-expression of IGPS1 and IGPS3 could be detected, suggesting a possible complex that could bind to the BX1 and IGL proteins ( Figure 8D). However, no interaction signal was detected for the IGPS1 -IGPS2 combination.

Discussion
Both complementation of an E. coli trpC mutation and analysis of maize transposon insertion lines indicated that maize IGPS1, IGPS2, and IGPS3 encode functional indole-3glycerolphosphate synthases. In a phylogenetic analysis, IGPS2 clustered with previously characterized Arabidopsis IGPS proteins (Li et al., 1995a; Figure 2), suggesting a role in primary metabolism. By contrast, IGPS1 and IGPS3 are in separate phylogenetic clades, consistent with their functions in defensive benzoxazinoid metabolism that is not found in Arabidopsis and other dicots that are included in the dendrogram. It is possible that the IGPS1 and IGPS3 enzyme clades in other monocots (rice, wheat, sorghum, barley and S. viridis; Figure 2) have additional functions in the production of defense-related metabolites in these species. A similar phylogenetic analysis of maize indole-synthesizing enzymes showed that BX1 and IGL, which contribute to defensive metabolism, cluster separately from the Arabidopsis and maize TSA proteins (Frey et al., 2000, Dutartre et al., 2012. Together, these observations are consistent with the hypothesis that pathways for the production of defense-related benzoxazinoids and indole arose through the duplication and repurposing of ancestral IGPS and TSA enzymes that contribute to primary metabolism in maize and other plants. Our results are consistent with the observations of Wisecaver et al. (2017), who used global co-expression network analysis to propose that IGPS1 is part of the benzoxazinoid biosynthesis pathway in maize. Both B73 and W22 inbred lines show a greater than a 100-fold increase of IGPS1 transcripts in response to caterpillar feeding ( Figure 4A, Figure S2A, . Consistent with this induction, igps1 maize mutants have lower levels of benzoxazinoids ( Figure 6A), and S. exigua caterpillars grow better on igps1 mutant than on wildtype W22 plants ( Figure 6F). IGPS3 did now show up in co-expression studies with BX genes (Wisecaver et al., 2017), perhaps due to the lower level of defense-related induction of gene expression, which may have been a factor in the integrative comparison of multiple gene expression studies.
Unlike in the case of IGPS1, IGPS2, IGPS3, BX1, IGL, and TSA, which localize to the chloroplasts when overexpressed in maize protoplasts, we were not able to verify a chloroplastic location for BX2, a cytochrome P450 that catalyzes the first oxygenation reaction in the pathway from indole to DIBOA-Glc (Frey et al., 1995). Consistent with localization of the BX2 enzymatic activity in microsomes after cell separation (Glawischnig et al., 1999), our protein fusion studies indicated that BX2 is localized in the endoplasmic reticulum ( Figure 7B). Given that both bx1 and bx2 mutations result in an almost complete loss of benzoxazinoids in maize inbred line W22 (Betsiashvili et al., 2015, Tzin et al., 2015, there must be a mechanism by which the indole that is produced by BX1 in the chloroplast is passed to BX2 in the endoplasmic reticulum. Indole produced by TSA or IGL appears to be in a separate metabolic pool and is not used as a substrate by BX2. There are direct interactions between maize TSA and TSB enzymes (Kriechbaumer et al., 2008), which may prevent the release of free indole in this pathway.
However, as BX1 and BX2 are in different cellular compartments (chloroplast and endoplasmic reticulum, respectively), it is harder to imagine a direct interaction that allows passage of indole from one protein to the other, without including free indole that is produced by IGL.
Indole production highly induced in maize after herbivore feeding (Frey et al., 2000, Erb et al., 2015. As in the case of benzoxazinoids, free indole was less abundant in both igps1 and igps3 mutant lines. Whereas bx1 mutations cause an almost complete loss of benzoxazinoids , and IGL is primarily involved in indole biosynthesis (Frey et al., 2000), IGPS1 and IGPS3 seem to be involved in both metabolic pathways. Consistent with a function in both benzoxazinoid and free indole biosynthesis, IGPS1 and IGPS3 appear to interact with both BX1 and IGL ( Figure 8A,C), and interactions between IGPS1 and IGPS3 ( Figure 8D) may also contribute to the overlapping functions of these enzymes. An igps1 igps3 double mutant would be required to determine whether benzoxazinoids and free indole in maize are derived from these two enzymes alone or whether IGPS2 also functions in the production of defensive maize metabolites.
Among the three maize indole-3-glycerolphosphate lyases, IGPS2 had the strongest interaction with TSA ( Figure 8B), which is thought to function in tryptophan biosynthesis ( Figure 1). Consistent with the hypothesis that tryptophan is synthesized primarily by an IGPS2dependent pathway, neither igps1 nor igps3 mutations significantly decreased the tryptophan content of maize leaves ( Figure 6D,E). Maize tsb1 tsb2 double mutations caused a seedling lethal phenotype (Wright et al., 1992), and it has been hypothesized that TSA, the only maize indole-3glycerolphosphate lyase that interacts with TSB1, is therefore also essential (Kriechbaumer et al., 2008). By extension, if IGPS2 has a primary function in providing IGP to TSA, it may also be essential. When conducting our experiments, we were not able to find any IGPS2 transposon knockout mutations of IGPS2 in public collections. However, there are predicted IGPS2 codingregion insertions in the recently published BonnMu collection (Marcon et al., 2020). If these insertions can be confirmed, it will be possible to determine whether IGPS2 has an essential function in maize metabolism. BX1, IGL, and TSA channel their metabolic products to particular pathways. The free indole-producing function of maize IGL (Frey et al., 2000, Ahmad et al., 2011 is not provided by the BX1 or TSA enzyme activity. Similarly, a bx1 knockout phenotype cannot be complemented by the activity of IGL or TSA, only by supplementing free indole (Frey et al., 1997, Melanson et al., 1997. To this metabolic channeling, we can now add the effects of IGPS1, IGPS2, and IGPS3, which have distinct, though partially overlapping, functions in channeling IGP to specific indole pools that are formed by BX1, IGL, and TSA. The protein interactions that we have observed (Figure 8) provide new insight into the regulation of indole metabolism and how, within the chloroplast, separate pools of this metabolite can be used for primary metabolism (tryptophan) and specialized metabolism (volatile indole and benzoxazinoids).

Plant material and methyl jasmonate treatment
Maize (Zea mays L.) seeds from the inbred lines B73, W22, and transposon insertion lines induction experiments using a previously described method , Tamiru et al., 2017, the third leaf of each plant was cut off and incubated in 2 ml tap water with 250 µM methyl jasmonate (Sigma-Aldrich; 392707) for 24 h, after which the leaves were snap-frozen in liquid nitrogen and stored at -80 °C until they were used for assays. Control leaves were harvested directly into liquid nitrogen without methyl jasmonate treatment.

Gene expression induction by insect feeding
Maize inbred line W22 gene expression assays after R. maidis infestation were performed as described for inbred line B73 by Tzin et al. (2015), but we used 10-day-old seedlings and samples were harvested at different time points: 0, 2, 8, and 24 h. The control plants received empty cages without aphids for 24 h and were harvested and frozen in liquid nitrogen at the same time as the experimental samples. To induce gene expression changes in maize line W22 with S. exigua caterpillars, we followed the protocol of Tzin et al. (2017). Leaf tissue for analysis was harvested after 24 h of caterpillar feeding.

Total RNA extraction and gene expression analysis
Harvested leaf material was frozen in liquid nitrogen, and ground to a fine powder for RNA isolation with on-column DNA digestion, using the SV Total RNA isolation kit (Promega, Madison, WI, USA). The total RNA concentration was measured with a NanoDrop instrument (2000c; ThermoFisher Scientific Inc., Waltham, MA, USA), and 500 ng total RNA was incorporated for first-strand cDNA synthesis using the M-MLV reverse transcriptase kit (TaKaRa Bio USA, Mountain View, CA, USA), and the library was used as a template for qRT-PCR analysis. Thereafter, 5 µl SYBR Green Mix (Universal SYBR Green Supermix, Bio-Rad Laboratories, Hercules, CA), 1 µl gene-specific forward primer, 1 µl gene-specific reverse primer, 0.5 µl template, and 2.5 µl PCR-grade water were mixed. The PCR reaction was performed in a QuantStudio TM 6 Flex Real-Time PCR system instrument (384-well, Applied Biosystems, Foster City, CA) with an initial incubation at 95°C for 10 min. The following cycle was repeated 40 times: 95°C for 30 s, 60°C for 15 s, and 72°C for 30 s. Maize adenine phosphate transferase 1 (APT1) was used as a reference gene. For each analyzed gene, a cDNA pool from all plants was diluted from 1 times to 80 times to generate a standard curve. With Applied Biosystems software, the ∆CT (∆-threshold cycle) for each gene was calculated relative to the reference gene. For the CT values an average from at least three biological replicates was calculated. The primer sets used to amplify the genes are given in Supplementary Table S1.

Transcriptome sequencing and RNAseq data analysis
The transcriptome of the W22 aphid time course experiment was sequenced using the 3´RNAseq method (Kremling et al., 2018). RNA was isolated as described above with a total of 8 biological replicates for each treatment. The purity of all RNA samples was proofed with a NanoDrop2000 instrument (Thermo Scientific  (Dobin et al., 2013) was used to align the reads against the W22 genome (Springer et al., 2018), with a maximum of one location to map to (-outFilterMultimapNmax 1). Another step in the pre-processing was increasing the region size of each gene by 500 bp to decrease the number of genes mapped to a region with no feature from > 120000 to < 60000 on average across the libraries. For this, we used the slop command from bedtools (https://bedtools.readthedocs.io/en/latest/ ) with -r 500 (-r the number of base pairs to add to the end coordinate). The program Tablet (https://ics.hutton.ac.uk/tablet/ ) was used to visualize the annotated genes on each chromosome. HTSeq32 v.0.6.1 were used to obtain genelevel counts from the resulting BAM files (htseq-count -r name -s yes -a 10 -f bam). To normalize the read counts, the CPM (count per million) was calculated by dividing the read counts for each gene by the total counts of the RNAseq sample. Illumina sequencing data were submitted to the Sequence Read Archive (https://www.ncbi.nlm.nih.gov/sra) as data set SUB8291142 (Submission ID), PRJNA669140 (Bioproject ID).

Isolation and cloning of cDNA
To isolate the open reading frames of IGPS1, IGPS2, IGPS3, TSA, IGL, BX1, and BX2 from B73 cDNA, specific primers (Supplementary Table S1) were created based on the B73_v4 maize genome. iproof high fidelity polymerase (Bio-Rad) was used for PCR amplification. The fragments were cloned into the specific vectors mentioned for each experiment and sequenced for their accuracy.

Complementation assays in E. coli
The open reading frames of IGPS1, IGPS2 and IGPS3, including their respective signal sequences, were cloned into the bacterial expression vector pASKIBA37+ (IBAGmbH, Göttingen, Germany). To amplify these genes, special primers with a BsaI restriction site were created with the ''Primer D ´ Signer'' (IBA GmbH) (Supplementary Table S1). After sequencing the constructs, plasmids were used for complementation assays in the E. coli trpC9800tryptophan-auxotrophic mutant background. The plasmids were transformed into trpC9800competent cells and plated out on 100 µg/ml ampicillin LB-agar plates. To test their activity in the mutant background, a starter overnight culture of 2 ml LB medium with 100 µg/ml ampicillin was prepared for each construct, including the empty vector as a negative control. The next morning, the starter culture was used to inoculate 5 ml of tryptophan-deficient Vogel and Bronner (Vogel and Bonner, 1956) minimal medium containing 2 % dextrose and 100 µg/ml ampicillin. For positive controls, 20 µg/ml L-tryptophan was added. All cultures started with an optical density at 600 nm (OD600) = 0.2 in the time-course experiment. OD600 was measured six times over 11 hours, until the bacteria had reached stationary phase.

Indol-3-glycerolphosphate lyase phylogenetic tree
A dendrogram analysis was conducted using MegaX software (Kumar et al., 2018). For the alignment we used the MUSCLE Codon algorithm (gap open −2.9, gap extend 0, hydrophobicity multiplier 1.2, clustering method, upgma). The evolutionary history was inferred by using the Maximum Likelihood method and JTT matrix-based model (Jones et al., 1992). The percentage of trees in which the associated taxa clustered together is shown next to the branches. Initial tree(s) for the heuristic search were obtained automatically by applying Neighbor-Join and BioNJ algorithms to a matrix of pairwise distances estimated using a JTT model, and then selecting the topology with superior log likelihood value. A discrete Gamma distribution was used to model evolutionary rate differences among sites (5 categories (+G, parameter = 0.7732)).
In an analysis involved 27 amino acid sequences, a bootstrap resampling analysis with 1,000 replicates was performed to evaluate the tree topology. All positions with less than 80% site coverage were eliminated, i.e., fewer than 20% alignment gaps, missing data, and ambiguous bases were allowed at any position (partial deletion option). There were a total of 358 positions in the final dataset. The maize TSA protein was used as outgroup for the phylogeny.

Metabolite analysis via LC-MS and data processing
Frozen, powdered leaf material was weighed (~100 mg), three volumes of 100% methanol were added to each sample, and samples were incubated for 45 min shaking at 4 °C. After a 10 min centrifugation step (15,000 x g) the samples were filtered in a 96-well MultiScreen filter plate (0.65 µm, Millipore, Burlington, MA). Reverse-phase liquid chromatography was performed using a DionexUltimate 3000 Series LC system (HPG-3400 RS High-Pressure pump, TCC-3000RScolumn compartment, WPS-3000TRS autosampler) controlled by ChromeleonSoftware (Thermo Fisher Scientific, Waltham, MA) and coupled to Orbitrap Q-Exactive mass spectrometer controlled by the Xcalibur software (Thermo Fisher Scientific). A Titan™ C18 UHPLC Column, 1.9 μm 100 mm × 2.1 mm at 40C and the flow rate of 0.5 ml/min of mobile phases A (H2O:0.1% formic acid) and B (AcN:0.1% formic acid) was used for the separation of target molecular features. The gradient starting condition was 0% B at 0 min, rising to 95% B in 15 min, then was held for 0.5 min, and followed by 0.5 min re-equilibration to the starting condition. A heated electrospray ionization source (HESI-II) in negative mode was used for the ionization, with the parameters of spray voltage at 3.5 kV, the capillary temperature at 380 C, sheath gas and auxiliary gas flow at 60 and 20 arbitrary units, respectively, S-Lens RF Level 50, and probe heater temperature 400°C. The data were acquired in the m/z range of 100-900, 140,000 FWHM resolution (at m/z 200), AGC target 3e6, maximum injection time of 200ms, in profile mode. Metabolite quantification was estimated with ThermoScientific Xcalibur™ Version 4.1.31.9 (Quan/Processing). For data analysis, the raw mass spectrometry files were converted to mzxml formats using the MSConvert tool (Chambers et al., 2012). The raw mzxml files were further processed using the XCMS (http://metlin.scripps.edu/download/, Tautenhahn et al., 2012) and the CAMERA software packages for R (http://www.bioconductor.org/packages/release/bioc/html/CAMERA.html). Data were analyzed using the XCMS-CAMERA mass scan data processing pipeline (Tautenhahn et al., 2008, Benton et al., 2010, Kuhl et al., 2012. To quantify tryptophan content of maize leaf samples, a stock solution was made by diluting 1 M solution of L-tryptophan (Sigma-Aldrich, St. Louis, MO) in water 1:10 with methanol. Further dilutions were made using 100% methanol and a standard curve was prepared by LC-MS analysis, as described above, of samples containing to 0.3 to 10 ng/µl tryptophan (Supplemental Figure S3A). The concentration of tryptophan in maize tissue samples (µg/g) was calculated based on comparisons to this standard curve.

Indole quantification by GC-MS
For the quantification of indole, leaf material was harvested, ground with a mortar and pestle in liquid nitrogen, and stored at -80°C until sample preparation. The volatile indole released from 50 mg powdered plant material was collected on a solid-phase microextraction (SPME) fiber (100 µm polydimethylsiloxane, fused silica 24 Ga; Supelco, Bellefonte, PA) for 45 min. Samples were analyzed using an Agilent 7890B gas chromatograph coupled to an Agilent 7000D triple quadrupole mass spectrometer (Agilent Technologies, Santa Clara, CA). Collection of volatiles from frozen and macerated tissue in this manner produces similar results to volatile collection from intact maize leaves (Köllner et al., 2004). To release volatiles from the fiber, an injection temperature of 220°C was used. Helium served as a carrier gas with a flow rate of 2.3 ml min -1 and nitrogen as collision gas with a flow rate of 1.5 ml min -1 . To separate the volatiles, an EC5-MS column (30 m length, 0.25 mm diameter., and 0.25-µm film; Agilent J&W GC columns) was used under the following conditions: 60°C for 3 min, first ramp of 10°C min -1 to 150°C, second ramp of 100°C min -1 to 300°C, and final 2-min hold. Indole was identified and quantified by running purified indole (Supelco) as a standard. Indole was dissolved in dimethylsulfoxide (DMSO), diluted in water 1:10 and added to 50 mg powdered leaf tissue from uninduced maize inbred line W22, which contains no detectable indole. A standard curve of indole concentrations ranging from 0.58 µg to 5.8 µg indole added to 50 mg powdered leaf material (Supplemental Figure S3B), was used to quantify indole abundance in methyl jasmonate-induced leaf samples.

Subcellular localization
Maize mesophyll protoplasts were isolated as described by Richter et al. (2016) tumefaciens cultures of both constructs with the same OD600, they were co-infiltrated into N. benthamiana leaves. Two days after infiltration, protoplasts were isolated from the infiltrated leaf tissue area and used for microscopic analysis. CFP fluorophore excitation was performed using a 458 nm argon laser and detected at 468 to 496 nm. For co-localization, CFP and mCherry were scanned sequentially.
For interaction studies, the open reading frames of IGPS1, IGPS2, IGPS3, TSA, IGL, and BX1 were cloned without stop codons into the BIFC vectors p2YC and p2YN (Kong et al., 2014). In addition, IGPS1 was cloned into p2YN to study the interaction with IGPS2 and IGPS3.
The constructs used for protein-protein interactions studies were co-infiltrated into N. benthamiana leaves, as described previously (DeBlasio et al., 2018). To detect the YFP signal and chloroplast autofluorescence, we used the microscope settings described above.

Caterpillar growth assays
Eleven-day-old homozygous igps1 mutant and wildtype W22 plants were used for caterpillar feeding assays. Spodoptera exigua eggs were purchased from Benzon Research (Carlisle, PA) and kept for 52 h in a 28°C incubator on artificial diet (Beet Armyworm Diet, Southland Products Inc., Lake Village, AR). One 10 x 15 cm organza mesh bag (amazon.com, item B073J4RS9C) containing one second instar S. exigua caterpillar was placed on each 11-day-old maize plant. The bags were gently closed with bobby pins without wounding the plant. After one week of feeding on maize plants, the caterpillars were weighed. Figure 1. Biosynthetic pathways from indole-3-glycerolphosphate to DIBOA-Glc, tryptophan, and free indole in maize. Protein localization and confirmation of IGPS1, IPGS2, and IGPS3 enzymatic activity is described in this manuscript. IGPS = indole-3-glycerolphosphate synthase, BX = benzoxazinoneless, TSA = tryptophan synthase alpha subunit, TSB = tryptophan synthase beta subunit, IGL = indole-3-glycerolphosphate lyase, DIBOA-Glc = 2,4-dihydroxy-1,4benzoxazin-3-one glucoside, ER = endoplasmic reticulum. . The enzymes cluster into phylogenetic groups that include the three maize enzymes, IGPS1 (orange cluster); IGPS2 (grey cluster); and IGPS3 (blue cluster). The MUSCLE Codon algorithm was used to perform a sequence alignment, which was used to reconstruct the tree using the Maximum Likelihood method and JTT matrix-based model. The percentage of trees in which the associated taxa clustered together is shown next to the branches (1,000 bootstrap trials). Maize tryptophan synthase alpha subunit (TSA) was used as an outgroup for the analysis.

Figure 3. Identification of maize indole-3-glycerol-phosphate synthase (IGPS) activity.
Maize IGPS enzyme activity was verified by cloning the cDNA and complementing an E. coli trpC9800mutation, which causes tryptophan auxotrophy through loss of indole-3glycerolphosphate synthase activity. Empty vector controls were grown with or without the addition of 20 µg/ml tryptophan to the minimal medium. Optical density at 600 nm (OD600) was measured over 11 h. Values are fold change relative to uninfested controls (n = 5). *P < 0.05, ***P < 0.001 by Student´s t-test (panels A and B) or Dunnett's test relative to 0 h control (panel C).    The enzymes cluster into phylogenetic groups that include the three maize enzymes, IGPS1 (orange cluster); IGPS2 (grey cluster); and IGPS3 (blue cluster). The MUSCLE Codon algorithm was used to perform a sequence alignment, which was used to reconstruct the tree using the Maximum Likelihood method and JTT matrix-based model. The percentage of trees in which the associated taxa clustered together is shown next to the branches (bootstrap trials 1,000). Maize tryptophan synthase alpha subunit (TSA) was used as an outgroup for the analysis. All protein sequences were retrieved from GenBank (https://www.ncbi.nlm.nih.gov/genbank/) using the indicated identifiers.  Figure 3. Identification of maize indole-3-glycerol-phosphate synthase (IGPS) activity. Maize IGPS enzyme activity was verified by cloning the cDNA and complementing an E. coli trpC9800mutation, which causes tryptophan auxotrophy through loss of indole-3-glycerolphosphate synthase activity. Empty vector controls were grown with or without the addition of 20 µg/mL tryptophan to the minimal medium. Optical density at 600 nm (OD600) was measured over 11 h. Values are fold change relative to uninfested controls (n = 5). *P < 0.05, ***P < 0.001 by Student´s t-test (panels A and B) or Dunnett's test relative to 0 h control (panel C). (F) Growth of Spodoptera exigua larvae on homozygous igps1 mutant and wildtype W22. *P < 0.05, **P < 0.01, ***P < 0.001 by Student´s t-test. Different letters on bars indicate P < 0.05, ANOVA and Fischer´s LSD test. No significant differences were detected for panels D and E (P > 0.05). Samples in panels A-E were treated methyl jasmonate to induce production of defensive metabolites.