Endogenous indole-3-acetamide levels contribute to the crosstalk between auxin and abscisic acid, and trigger plant stress responses in Arabidopsis

Arabidopsis AMIDASE 1, which catalyses the conversion of indole-acetamide into the auxin indole-3-acetic acid, is involved in coordinating the trade-off between plant growth and stress responses, balancing auxin and ABA homeostasis.


Introduction
A constantly changing and often adverse environment represents a steady challenge to plants. These challenges include various biotic and abiotic stresses, such as pathogen infection, herbivory, high salinity, drought, or temperature changes (Spoel and Dong, 2008;Verma et al., 2016). In order to survive and secure their reproduction, plants often restrict their growth and development under adverse conditions, because energy resources are limited, and their allocation must be tightly balanced to meet the requirements of both growth and adaptive stress responses. To keep track with specific environmental demands, plants have evolved a complex hormonebased network to control their development in response to given changes in their surroundings. Using this system, they can shape their body plan and optimize their metabolism in accordance with prevailing environmental conditions. Key determinants of this regulatory network are limited number of signaling molecules referred to as plant hormones. They orchestrate plant growth and development mainly by influencing the transcriptome of responding cells and organs, respectively. Plant hormones act in a combinatorial manner to produce a large number of different responses that are dependent not only on the perceived stimulus, but also on the specific properties of the responding tissue (Depuydt and Hardtke, 2011;Voß et al., 2014). There is mounting evidence for the involvement of auxin in the trade-off between growth and defense. For example, the transcription factor MYC2, which controls the expression of jasmonic acid (JA) responsive genes (Kazan and Manners, 2013), has been reported to negatively regulate the expression of PLETHORA (PLT1 and PLT2) transcription factor genes, which control stem cell development and auxin biosynthesis in roots (Pinon et al., 2013). Other studies also revealed that auxin production is increased by JA through the induction of ANTHANILATE SYNTHASE 1 (ASA1) and a small number of YUCCA genes in certain plant tissues (Sun et al., 2009;Hentrich et al., 2013).
Plants also balance their growth with responses to abiotic stresses (Zhu, 2002;Verma et al., 2016). When exposed to drought or high salinity, two major abiotic stress factors in the field, plants reduce their growth rate, while stress-resistance mechanisms are initiated. Through the flexible reprogramming of their gene regulatory networks, plants strive for the best suited phenotype for the prevailing stress condition (Claeys and Inzé, 2013;Julkowska and Testerink, 2015). Brassinosteroid signaling pathways are reported to contribute to the control of plant growth under drought or starvation stress by regulating autophagy (Nolan et al., 2017). However, other plant hormones including auxin and abscisic acid (ABA) are also assumed to contribute to the coordination of the growth-stress response trade-off. Auxin is known to be the crucial determinant for plant growth (Davies, 2010), whereas ABA is reported to operate as a stress hormone in responses to abiotic stimuli (Gomez-Cadenas et al., 2015;Ma et al., 2018).
Under optimal conditions, plants usually grow rapidly with their energy resources mainly dedicated to primary growth, including root system and leaf expansion, shoot elongation, and reproduction. In this scenario, auxin is pivotal because it governs virtually all aspects of plant growth and development through the promotion of cell elongation, expansion, and differentiation (Vanneste and Friml, 2009). To ensure optimal plant growth, auxin homeostasis needs to be tightly regulated (Ljung, 2013). In order to control cellular auxin contents, plants possess a series of different biochemical and biological tools, including de novo biosynthesis, inactivation through conjugation, sequestration, and degradation (Ruiz Rosquete et al., 2012;Brumos et al., 2018). In contrast, when plants are subjected to biotic and abiotic stresses, they respond with the adjustment of their growth program, which in many cases results in reduced growth rates and premature phase transition from vegetative to reproductive growth (Silva et al., 2013;Suzuki et al., 2014;Perdomo et al., 2015). JA and ABA, two well-characterized plant stress hormones, play crucial roles in plant responses to biotic and abiotic stress factors. However, an underlying mechanism by which these two phytohormones are connected with cellular auxin levels remains largely elusive.
Mainly based on in vitro evidence, AMIDASE 1 (AMI1) has been suggested to act in a side pathway of indole-3-acetic acid (IAA) biosynthesis ( Fig. 1), converting indole-3-acetamide (IAM) into IAA (Pollmann et al., 2003;Neu et al., 2007;Nemoto et al., 2009;Sánchez-Parra et al., 2014). In Arabidopsis, the majority of IAM (95%) has been reported to derive from the precursor indole-3-acetaldoxime (IAOx) (Sugawara et al., 2009). The enzyme responsible for the biochemical conversion, however, has yet to be identified. IAOx depleted mutant plants, i.e. cyp79b2 cyp79b3, show neither significant alterations of free auxin contents under standard conditions nor transcriptional induction of other auxin biosynthesis pathway components to compensate the loss of IAOx-dependent pathway(s) (Zhao et al., 2002;Lehmann et al., 2017). Together, these findings argue against a contribution of AMI1 in general auxin biosynthesis. Here, we report the functional characterization of AMIDASE 1 (AMI1) in vivo. The comprehensive analysis of AMI1 gain-and loss-of-function mutants provided indication for an involvement of AMI1 in auxin homeostasis. A loss of AMI1 results in increased IAM levels and moderately reduced IAA contents, whereas conditional overexpression of AMI1 (AMI1ind) had the opposite effect. The examined ami1 mutant alleles, ami1-1 and ami1-2, show moderate growth reductions, while independent AMI1ind lines are characterized by auxin overproduction-related phenotypes. Comprehensive microarray analyses comparing AMI1 mutants with wild-type (WT) plants revealed a tight connection of AMI1 with various biotic and abiotic stresses, as numerous stress marker genes appeared to be differentially expressed. Abiotic stress treatments confirmed that ami1 mutants are more susceptible towards osmotic stress. In summary, our results suggest that AMI1 is involved in the coordination of the trade-off between growth and stress resistance, rather than in the de novo biosynthesis of auxin.

Plant material and plant growth conditions
The experiments were carried out using Arabidopsis ecotype Col-0 (from Nottingham Arabidopsis Stock Centre (NASC), stock N1092). Seeds for 35S::tms2 (N6265), and the T-DNA insertion lines for AMI1 (At1g08980; ami1-2 (SALK_019823C)) were obtained from the Nottingham Arabidopsis Stock Centre, or kindly provided by Dr Henrik Aronsson, ami1-1 (Aronsson et al., 2007). T-DNA mutants were genotyped according to Alonso et al. (2003). Primers for genotyping are given in Supplementary Table S1 at JXB online. If not stated otherwise, seedlings were raised under sterile conditions on solidified ½ MS medium containing 1% (w/v) sucrose in Petri dishes (Murashige and Skoog, 1962). Plantlets were kept under constant environmental short day (SD) conditions (8 h light at 24 °C, 16 h darkness at 20 °C, 105 µmol photons m −2 s −1 photosynthetically active radiation) for 2-3 weeks. If older plant material was used, the plant organs were harvested from plants grown in a greenhouse on a mixture of soil and sand (2:1) for 4-6 weeks in SD conditions. Thereafter, plants were transferred to long day (LD) conditions (16 h photoperiod). The greenhouse was maintained under constant climatic conditions, 22-24 °C during the daytime and 18-20 °C overnight. The photosynthetically active radiation was no less than 150 µmol photons m −2 s −1 . Expression of the transgene in AMI1ind lines was induced by either 10 µM β-estradiol added to the growth medium or by the administration of 50 µM β-estradiol through irrigation. In all experiments that employed AMI1ind lines, corresponding WT control plants were also grown on medium containing β-estradiol.
We next added a 5′ myc-tag extension to the AMI1 cDNA by PCR amplification. The PCR product was integrated into pGEM-T and the sequence integrity was verified. In order to construct an inducible AMI1 plant expression system, the myc:AMI1 fragment was integrated into pSP-ENTR1, yielding the vector pENTR-AMI1ind. Using the Gateway LR Clonase II mix, recombination of pENTR-AMI1ind with the destination vector pMDC7, a Gateway-compatible pER8 derivate (Zuo et al., 2000), created the binary vector pER-AMI1ind in which conditional AMI1 expression is facilitated by the β-estradiol inducible XVE transactivator. Correct integration of the target sequences in all resulting binary vectors was verified by PCR. Sequence information on the employed primer can be found in Supplementary Table S1. All constructs were transformed into Arabidopsis WT (Col-0) by Agrobacterium-mediated transformation (Clough and Bent, 1998). T 1 seeds were harvested, sown on half-strength MS medium (0.8% phytoagar and 1% sucrose) adding either 5 mg l −1 BASTA (Hoechst) (pSP-AMI1-PromA) or 15 µg ml −1 hygromycin (pER-AMI1ind) to select transformants. In all subsequent experiments only selected T 3 seeds were used.

Transcriptomics
Total RNA was extracted from 100 mg plant tissue as previously described (Oñate-Sánchez and Vicente-Carbajosa, 2008). Labeling and hybridization of cDNA libraries from ami1-2, AMI1ind-2, and Col-0 to ATH1 DNA chips (Affymetrix) was performed by the CNB Genomics Service (Madrid, Spain). For each genotype three biological replicates were processed. Variation between replicates was accounted for by using the LIMMA model (Ritchie et al., 2015). Differentially expressed genes were identified by a modified t-test implemented in the LIMMA package. A Benjamini-Hochberg correction was used to adjust for multiple testing. An adjusted P-value <0.05 and fold-change ≥1.5 were arbitrarily chosen to select 378 and 878 differentially expressed genes (DEGs) in ami1-2 and AMI1ind-2, respectively, relative to WT (see Supplementary Table S2). The Gene Set Enrichment Analysis (GSEA) used a hypergeometric test with a Benjamini-Hochberg false discovery rate (FDR) correction and a post-correction selection significance level of P<0.05 (Maere et al., 2005). Complementing the GSEA, parametric analysis of gene set enrichment (PAGE) analyses were carried out to determine GO terms enriched in the de-regulated gene groups. The PAGE method is statistically more sensitive and accounts for both the number of genes and their respective expression patterns, employing a Benjamini-Hochberg (FDR) multi-test adjustment and a significance of P<0.05 (Kim and Volsky, 2005).
Selected transcripts were validated in independent experiments by qRT-PCR. For each genotype and condition, total RNA from three independent biological replicates was harvested and analysed in triplicate (technical replicates). First-strand synthesis was performed according to the supplier's instructions, using M-MLV reverse transcriptase and oligo(dT) 15 primer (Promega). Two nanograms of cDNA was used as template for qPCR, which was performed according to the manufacturer's instructions using the FastStart SYBR Green Master solution (Roche Diagnostics) on a Lightcycler 480 Real-Time PCR system (Roche Diagnostics). Relative quantification of expression was calculated after data analysis by the Lightcycler 480 software (Roche Diagnostics), using the comparative 2 −ΔΔCT method (Livak and Schmittgen, 2001) with APT1 (At1g27450) and UBQ10 (At4g05320) as reference genes (Czechowski et al., 2005;Jost et al., 2007). See Supplementary Table S1 for primer sequences.
GUS activity was calculated as follows: GUS activity pmol / min −1 = ∆F · 10 −1 t with ΔF giving the fluorescence intensity T 60 −T 0 , 10 referring to the fluorescence units corresponding to 1 pmol of hydrolysed 4-MUG, and t representing the incubation time in minutes. GUS activity expressed in units was activity per mg of protein, where the protein amount was quantified using the Bradford method (Bradford, 1976).

Amidase activity assay
The IAM hydrolase activity in plants was determined as previously described (Neu et al., 2007). In brief, crude extracts from 2-week-old seedlings were prepared. One hundred milligrams of plant tissue was shock-frozen in liquid nitrogen. The tissue was disrupted using micropestles. As the material started to thaw, buffer (50 mM HEPES, pH 8.5, 200 mM sucrose, 3 mM EDTA, 3 mM DTT, 5% (w/v) insoluble polyvinylpyrrolidone) was added in a ratio of 1:3 (w/v). The suspension was homogenized until no more tissue fragments were visible. Cell debris and insoluble matter were collected by centrifugation (16 000 g, 15 min, 4 °C) and the supernatant was transferred to fresh tubes. Protein concentrations were determined according to Bradford (1976), using serum albumin as a protein standard. Aliquots containing 200 µg of protein were incubated with 5 mM IAM in a total volume of 300 µl of 50 mM potassium phosphate buffer (pH 7.5) for 3 h at 30 °C. Finally, the amount of IAA produced was quantified by reverse phase HPLC. The data were analysed using Student's t-test to compare two means. Statistical analyses were conducted using Prism version 7.0a (GraphPad Software).

IAA and IAM quantification
The LC-MS analysis of endogenous IAA and IAM was carried out according to Novák et al. (2012). For each sample 100 mg of plant material was harvested and immediately frozen in liquid nitrogen. Each independent experiment used at least three biological replicates. Sample handling times were kept as short as possible to prevent distortion of the IAA pool through the autocatalytic decay of precursor molecules (Gélinas-Marion et al., 2020). The plant material was transferred into 2 ml screw-cap tubes filled with ceramic beads (MagNA Lyser Green Beads; Roche Diagnostics) and endogenous IAA and IAM was extracted into 1 ml of cold potassium phosphate buffer (50 mM, pH 7.0) containing 1% diethyldithiocarbamic acid sodium salt and 50 pmol of the internal standards [ 2 H 5 ]IAA and [ 2 H 5 , 15 N] IAM (OlchemIm). The plant material was crushed in a MagNA Lyser Instrument (Roche Diagnostics) at 8000 g for 1 min. After centrifugation (5 min, 16 000 g at 4 °C), 750 µl of the supernatants was transferred into fresh tubes and the pH adjusted to 2.7 with 1 M hydrochloric acid. Thereafter, samples were loaded onto Oasis TM HLB columns (1 ml, 30 mg; Waters Corp.) conditioned with 1 ml methanol and 1 ml water and equilibrated with 1 ml acidified sodium phosphate buffer (50 mM, pH 2.7). The columns were washed twice with 1.5 ml 5% methanol and subsequently eluted with 2 ml 80% methanol. The eluates were taken to dryness in vacuo. Prior to the mass spectrometric analysis, the evaporated samples were dissolved in 45 µl methanol with 0.1% formic acid (v/v). Ten microliters of the extract was separated using an Ultimate3000 RSLC system (Dionex), and analysed using a microTOF-Q II mass spectrometer (Bruker Daltonics). Analyte separation utilized a 50 mm×2.1 mm i.d., 1.7 μm, Acquity UPLC BEH C18 column with a 5 mm×2.1 mm i.d. Acquity UPLC BEH C18 VanGuard pre-column, and a 0.2 μm×2.1 mm i.d. in-line filter (Waters Corp.). The following binary gradient was applied: 0-2 min isocratic 98% solvent A (water with 0.1% (v/v) formic acid), 2% B (acetonitrile with 0.1% (v/v) formic acid); 2-30 min linear gradient to 5% A, 95% B; at 30 min step gradient to 100% B; isocratic for 1 min. Thereafter, the column was set to 98% A, 2% B and conditioned for 2 min before the next injection. The flow rate was 400 μl min −1 . Eluted compounds were analysed using a microTOF-Q II mass spectrometer (Bruker Daltonics) operated in positive electrospray mode. Typical settings were as follows: capillary voltage, −4500 V; dry gas temperature, 200 °C; dry gas flow, 10 l min −1 ; funnel, RF 200 Vpp. Ions were detected from m/z 50 to 1000 at a repetition rate of 2 Hz. The instrument was operated in multiple reaction monitoring mode. Mass calibration was performed using sodium formate clusters (10 mM solution of NaOH in 50/50% v/v isopropanol/water containing 0.2% formic acid). For data processing the DataAnalysis 4.0 software (Bruker Daltonics) was used. The following transitions were recorded to determine analyte contents: IAM, m/z=175. 2→130 . Quantification was achieved by comparing peak areas of the endogenous compounds with those of the internal standards. The data were analysed using Student's t-test to compare the mutant mean with the mean of the control. Statistical analyses were conducted using Prism version 7.0a.

ABA quantification
The ABA contents in 50 mg of 10-day-old sterilely grown Col-0 and ami1 seedlings were determined using GC-MS/MS (Bruker Daltonics, Scion-TQ) analysis as previously described (Ramos et al., 2018). Quantification was carried out using stable isotope-labelled [ 2 H 6 ]ABA as internal standard. Endogenous hormone contents were calculated from the signal ratio of the unlabeled over the stable isotope-containing mass fragment observed in measurements that have been performed in quintuplicate. The data were analysed with one-way ANOVA followed by a Tukey-Kramer post hoc test to allow for comparison among all means. Statistical analyses were conducted using Prism version 7.0a

Western blotting
Both SDS-PAGE and amidase immunoblots were carried out as described by Neu et al. (2007). A monoclonal α-myc antibody, clone 9.E10 (Evan et al., 1985), was used as first antibody at a final dilution of 65 mg l −1 . As second antibody was used, a rabbit-anti-mouse-IgG (Promega) coupled with an alkaline phosphatase.

Biomass and root morphology analysis
For the analysis of the root morphology, five replicate Petri dishes with ½ MS agar medium (0.5% sucrose, w/v), containing 21-26 seedlings, were used for each line. The root system morphology of the seedlings was investigated using the WinRhizo TM image analysis software (Regent Instruments Inc.). The roots were scanned using a 600 dpi resolution STD1600+ scanner (Regent Instruments Inc.) while still attached to the dish. The images were analysed to identify various root system morphology parameters (Björk et al., 2007), namely root length (cm), surface area (cm 2 ), number of branches (forks), and number of root tips. The shoot and the root systems were then separated and their dry weight determined using a Sartorius ultra-micro scale (Sartorius GmbH, Göttingen, Germany) after drying the individual parts at 70 °C for at least 48 h. Root-to-shoot ratio (R/S ratio), specific root length (SRL; mg DW −1 root), specific root area (SRA; m 2 kg DW −1 root), specific root tip density (tips g DW −1 root), root tip density (tips cm −1 root), specific branching density (forks g DW −1 root), and branching density (forks cm −1 root) were calculated for each genotype (Björk et al., 2007). Differences in the root architecture were investigated using a MANOVA, which was followed by Tukey's HSD post hoc test for the genotype. Statistical analyses were conducted using SPSS 18.0 (SPSS Inc.).

Root growth response assay
The root growth response was analysed according to the protocol of Zimmerman and Hitchcock (1942). Plants were grown on ½ MS plates for 7 d before they were transferred onto medium supplemented with either 5 nM to 10 µM IAA (from 100 mM stock in methanol) or 10 nM to 200 µM IAM (from 100 mM stock in methanol). Plates were wrapped in gas-permeable surgical tape (BSN medical GmbH) and grown vertically under constant conditions for 2 weeks. Subsequently, the primary root length increase of at least 12 individual plants per treatment was determined. Differences in primary root elongation were analysed using Student's t-test to compare two means. Statistical analyses were conducted using Prism version 7.0a.

Results
The AMI1 gene is generally expressed in proliferating tissues, but repressed during early stages of germination In our previous work, we performed microscopic experiments using AMI1-GFP fusion proteins and semi-quantitative RT-PCR to investigate the subcellular localization and tissue specific expression of AMI1. The experiments revealed a mainly cytoplasmic localization and the expression of AMI1 in proliferating tissues, such as young leaves and flowers (Pollmann et al., 2006b). Additional qPCR experiments confirmed the expression pattern for AMI1 on a tissue-specific basis. AMI1 expression was detected in all analysed samples. The expression in roots was, however, low compared with the levels in all other tissues. In adult plants, AMI1 expression was highest in juvenile leaves and petioles; lower transcript levels were detected in stems and in flower buds (see Supplementary Fig. S1A).
To enable further examination of AMI1 expression during plant development, we generated pAMI1::GUS promoter reporter lines. Since qPCR data pointed towards high AMI1 expression in young seedlings, we monitored β-glucuronidase (GUS) activity in pAMI1::GUS lines between 26 h and 72 h after imbibition ( Fig. 2A). The GUS staining showed an increasing AMI1 promoter activity with the ongoing of seedling development. Starting after about 48 h in the cotyledons, GUS activity peaked 64 h after germination. The observed promoter activity matches considerably well with the previously determined IAM profile during seed germination and early seedling development (Pollmann et al., 2002;Hoffmann et al., 2010), pointing towards a rapid decline of IAM levels during early seedling development. At later stages, especially in the primary root, the expression slowly decreased, while in root tips a clear GUS activity remained detectable. The qPCR experiment provided further evidence for a substantial AMI1 expression in leaves, petioles, and flower buds. For this reason, we analysed AMI1 promoter activity during later stages of development. In cotyledons, expression was mainly detected in the vascular tissue and on the very tip of the leaf (Fig. 2B). In primary leaves, GUS staining was visible in the vascular tissue, at the tip of the leaf, and also in trichomes (Fig. 2C). During flower development, at developmental stage 15 (Smyth et al., 1990), strong activity was detected in the petioles, pedicels and receptacles, in the vasculature of the sepals, and at the end of the pistil including the stigma (Fig. 2D). Only low activity was visible in the filaments and anthers, whereas the petals showed no staining (Fig. 2D, E). In developing siliques, at stage 17, strong promoter activity was observed at the end of the pedicels and on both ends of the silique, whereas nearly no activity was found in the developing seeds and funiculi (Fig. 2F). Taken together, these results confirm the expression of AMI1 in proliferating tissues, including young seedlings and developing flowers.

High levels of IAM induce AMI1 gene expression
Numerous previous studies already analysed the transcriptomic response towards IAA in Arabidopsis. The inspection of publicly available datasets, e.g. GSE631 and GSE42007 (Okushima et al., 2005;Lewis et al., 2013), demonstrated that most auxin homeostasis-related genes show no significant response to IAA (see Supplementary Table S2). Consistent over both datasets, only a small number of Gretchen Hagen 3 (GH3) genes appear to be significantly induced by IAA. GH3 genes encode IAA amidosynthases that catalyse the conjugation of free IAA to amino acids, thereby physiologically inactivating the plant hormone (Staswick et al., 2005). To investigate the response of AMI1 towards the exogenous application of IAA and IAM in closer detail, we performed qPCR experiments on 2-week-old seedlings, which were treated for different periods of time with the indoles. Generally confirming the transcriptomic data, the application of IAA (20 µM) triggered only a weak repression (log 2 =−1.3±0.1) of AMI1 transcription. In contrast, the treatment with 20 µM IAM significantly induced the expression of AMI1. The transcriptional level of AMI1 temporarily increased to reach a log 2 value of 2.9±0.3 at 2 h, only to return to a lower level after 3 h of IAM treatment (Fig. 3). The induction of AMI1 gene expression by its putative substrate, IAM, suggests a function of AMI1 in the control of cellular IAM levels.

Perturbations of AMI1 gene expression alter auxin contents and trigger auxin-related phenotypes
To explore the role of AMI1 in auxin homeostasis, we took a complementary reverse genetics approach. First, we studied two independent mutant alleles for AMI1, carrying T-DNA insertions either in the fourth intron (ami1-1) (Aronsson et al., 2007) or in the sixth exon (ami1-2). The T-DNA insertion sites of ami1-1 and ami1-2 were confirmed by sequencing. In the case of ami1-2, the sequencing of the genomic DNA revealed a tandem T-DNA insertion (Fig. 4A). Homozygous plants were isolated and genotyped (Fig. 4B). Both null alleles were devoid of detectable amounts of full-length AMI1 mRNA (see Supplementary Fig. S1B), which could also be validated by qPCR, demonstrating a lack of expression of the second half of the AMI1 gene, containing the T-DNA insertions ( Supplementary Fig. S1C). The results provided evidence that the T-DNA insertions in ami1-1 and ami1-2 result in the formation of truncated and likely functionally impaired transcripts.
The phenotypic alterations observed in the ami1 alleles were only moderate. Areal plant parts showed a slight growth reduction, particularly when grown on soil for a longer period of time (Fig. 4C). A more detailed analysis of the root system morphology of AMI1 mutants employing the WinRhizo TM image analysis software confirmed this observation and revealed a significantly reduced root branching and total root length and area, respectively, for both ami1 alleles (see Supplementary Fig. S2). Although the phenotypic alterations were not pronounced, we asked whether the altered AMI1 expression translates into detectable changes in the chemotype. To this end, we examined IAM hydrolase activity in the ami1 alleles (Fig. 4D). In comparison with WT, the amidase activity in ami1-1 and ami1-2 was not null but rather reduced by approximately 45-50%, pointing towards the existence of an additional IAM hydrolase in Arabidopsis. Next, we analysed changes in IAA and IAM contents by mass spectrometry. We found a correlation between the decreased IAM hydrolase activities in the ami1 mutants and endogenous IAA and IAM levels. Both mutant alleles showed significantly reduced auxin contents, ranging from 15 to 30% under the WT level alongside with significantly higher IAM contents (Fig. 4E, F).
Secondly, we generated mutants conditionally overexpressing N-terminally c-myc tagged AMI1 cDNA. In these lines, the c-myc:AMI1 construct was under the control of the β-estradiol  inducible XVE transactivator (Zuo et al., 2000). After screening numerous homozygous AMI1ind T 3 -lines, two independent lines, AMI1ind-1 and AMI1ind-2, with considerable transgene expression (see Supplementary Fig. S3A) and significantly increased amidase activity (Fig. 4D) were selected for all following experiments. In contrast to the ami1 alleles, 2-weekold AMI1ind lines displayed significantly higher IAA levels and nearly 60% less IAM relative to WT plants (Fig. 4E, F). However, the root system architecture of the AMI1ind lines resembled the WT (Supplementary Fig. S2A-C). To survey the influence of high-level AMI1 expression on later stages of plant development, the AMI1ind mutants were grown on sterilized soil. Expression of the transgene was induced by irrigating the plants with water containing 50 µM β-estradiol ( Supplementary Fig. S3A). In comparison with non-induced plants, the induced AMI1ind plants exhibited an auxin-related phenotype, showing retarded growth ( Supplementary Fig.  S3B) as well as curled leaf shapes ( Supplementary Fig. S3C), reminiscent of other plants showing auxin overproduction (Zhao et al., 2001;Zhang et al., 2007;Hentrich et al., 2013). Furthermore, the AMI1ind lines showed considerably accelerated flowering (Supplementary Fig. S3D).
To further characterize the impact of AMI1 overexpression, we monitored the root growth in response to the suggested substrate of AMI1, IAM. AMI1ind lines and WT plants were grown on plates containing rising amounts of IAM (see Supplementary Fig. S4A). With rising concentrations, the growth inhibiting effect of IAM became more pronounced in AMI1ind plants relative to the WT. At 200 µM IAM, the primary root growth was nearly fully suppressed in AMI1ind. Under these conditions, the overexpressors carried predominantly adventitious roots on their hypocotyls. In contrast, WT seedlings grown in parallel were at least able to produce short primary roots with an increased number of lateral and adventitious roots. The described adventitious root phenotype at high IAM levels is in line with the phenotype previously observed for tobacco plants constitutively overexpressing the Agrobacterium tumefaciens iaaM and iaaH genes (Sitbon et al., 1992).
To quantify the observed differences, primary root elongation responses towards IAM of AMI1ind, ami1, and 35S::tms2 were determined and compared with WT Arabidopsis (Fig. 5). The 35S::tms2 mutant served as a control to assess the effect of AMI1 overexpression. The line contains the 35S-driven INDOLEACETAMIDE HYDROLASE (tms2) gene from Agrobacterium tumefaciens (Karlin-Neumann et al., 1991). Seedlings of the different genotypes were grown on ½ MS plates for 7 d, before they were transferred to plates containing IAM. While ami1-1 and ami1-2 seemed to be less sensitive towards IAM in the medium (Fig. 5A), AMI1ind-1 and AMI1ind-2 primary roots exhibited an increased sensitivity towards IAM, manifesting in a significantly stronger repression of primary root elongation growth relative to WT plants.
Resembling the root elongation response of 35S::tms2, both AMI1ind lines exhibited reduced root growth with rising IAM contents. At very high IAM concentrations (100 µM), however, root elongation was equally suppressed in all tested seedlings (Fig. 5B). The auxin response of the AMI1 gain-of-function mutants was also examined in a similar manner employing plates containing different amounts of IAA (see Supplementary  Fig. S4B, C). Only at low IAA contents (0.01 µM), AMI1ind and 35S::tms2 seedlings showed significantly less primary root elongation, which might be attributable to the higher intrinsic IAA level in those lines. With higher concentrations of IAA (0.1-10 µM) the growth response of the mutant roots was indistinguishable from that of the WT control. Hence, it can be concluded that IAA perception is not affected by the conditional overexpression of AMI1. In summary, the metabolomic data underscore a contribution of AMI1 to auxin homeostasis. The only mild phenotype of the ami1 mutants, however, implies that the role of the enzyme is likely not to be to control the general supply of auxin, but possibly to control the cellular IAM pool.

Altered abundance of salt-, osmotic-, and woundingrelated transcripts in ami1-2
To obtain further information on the molecular alterations triggered by perturbations of AMI1 gene expression, we conducted transcriptome profiling of sterilely grown 12-day-old WT, ami1-2, and AMI1ind-2 seedlings. Herein, we were especially interested in transcriptional changes of genes previously associated with auxin-related processes. Hence, we first analysed the differential expression of a sub-group of 128 genes, related with auxin biosynthesis, conjugation/deconjugation, degradation, transport, and signaling. The study revealed that only a very few of the tested genes were differentially expressed in the mutants (see Supplementary Table S2). Except for the induction of YUC8 and ILL5/IAR3, no other auxin homeostasis-related genes appeared to be differentially expressed in ami1-2. YUC8 is involved in auxin biosynthesis (Hentrich et al., 2013), and ILL5/IAR3 are two highly similar IAA-amino acid hydrolases, specific for IAA-Leu and IAA-Phe (Davies et al., 1999;LeClere et al., 2002). The induction of YUC8 and ILL5/IAR3 could possibly be conducive to compensating the loss of AMI1 in auxin biosynthesis. On the other hand, YUC8 and ILL5/IAR3 have been associated with biotic stress responses (Truman et al., 2010;Hentrich et al., 2013), which could also suggest a link of AMI1 with plant stress-related processes. For AMI1ind-2, we found the auxin conjugationrelated genes UGT75D1 and GH3.17 to be significantly induced. Together with the de-regulation of a number of auxin signaling and transport components, including LAX2, PIN4, PIN5, IAA1, IAA12, IAA14, ARF6, ARF7, and ARF16, this likely represents an answer to the overproduction of IAA in the conditional mutant.  (At1g08980) showing the exon/intron structure and the T-DNA insertion sites in intron 4 and exon 6. The primer pairs used for either RT-PCR (1/2, 3/4) or qRT-PCR (5/6) are shown as arrows. The resulting cDNA with the fused exons, but without the 5′ and 3′ untranslated regions, is depicted below. Also given is the amidase signature (AS) domain of the derived protein. Insertion sites of the T-DNA in the alleles ami1-1 and ami1-2 were analysed by sequencing. Genomic sequence and cDNA, but not the T-DNA, are drawn to scale. Lb, left boarder; ATG, start codon; TGA, stop codon. (B) PCR zygosity analysis of the studied genotypes. Total genomic DNA was extracted from wild-type and homozygous T-DNA insertion lines. One T-DNA and two gene specific primers (Supplementary Table S1) were used in three combinations for genotyping: the first reaction (1) contained two gene-specific primers, flanking the T-DNA insertion; the second (2) comprised the T-DNA left border primer and the upstream gene specific primer; the third reaction (3) contained the T-DNA left border primer and the downstream gene specific primer. (C) Phenotype of Arabidopsis wild-type (Col-0) and ami1-2 plants grown for either 2 weeks on ½ MS plates or 4 weeks on soil. (D) IAM hydrolase activity in ami1 T-DNA insertion mutants and AMI1 overexpressing AMI1ind lines. Crude extracts were incubated for 3 h at 30 °C with 5 mM IAM as substrate. The IAA produced was detected after separation by reverse phase HPLC. Means ±SE are given (n=5). Similar results were obtained in two independent experiments. (E, F) Analysis of IAM (E) and free IAA (F) levels. Two-week-old seedlings were used for the determination of IAA and IAM by LC-MS analysis. Three independent biological replicates were assessed. Quantification of free IAA levels in ami1-1, ami1-2, AMI1ind-1, and AMI1ind-2 relative to the wild-type (Col-0). The standard error of the mean is given (n=9). Student's t-test: *P≤0.05, **P≤0.01, ***P≤0.0001.
Next, we undertook a detailed profiling approach to obtain a broader overview of the transcriptomic alterations in the AMI1 mutants. Based on an adjusted P-value of <0.05 and an arbitrarily chosen fold-change value of ≥1.5, 391 and 969 DEGs were selected in ami1-2 and AMI1ind-2, respectively (see Supplementary Table S2). From these genes, 287 and 347 were up-and 88 and 622 were down-regulated in ami1-2 and AMI1ind-2, respectively (Fig. 6A). From the DEGs, 289 and 775 genes, respectively, could be assigned to 178 and 273 significant GO terms for ami1-2 and AMI1ind-2 (Supplementary Table S3).
Parametric analysis of gene set enrichment (PAGE) suggested a prominent de-regulation of plant stress responserelated genes in the AMI1 mutants. Among the up-regulated genes in ami1-2, we observed the enrichment of genes related with salt stress and osmotic stress, as well as oxylipin biosynthesis and signaling. The corresponding genes included, for instance, the AP2/ERF transcription factors ERF53 and RAP2.6, the MYB transcription factors MYB34, MYB47, MYB74, MYB102, and MYB108, the JA signaling repressors JAZ3 and JAZ9, ALLENE OXIDE SYNTHASE (AOS), HYDROPEROXIDE LYASE (HPL1), JASMONIC ACID RESPONSE 1 (JAR1), the dehydration responsive genes RD29A, ERD7, and EDR9, as well as the 9-cis-epoxycarotenoid dioxygenase (NCED) gene NCED3 ( Fig. 6B; Supplementary Fig. S5). Among the down-regulated genes in AMI1ind-2, we found an over-representation of biotic and drought stress-related genes. This second group of genes included, among others, the transcription factors WRKY33, ORA47, ZAT10, DREB1B, and DREB26, as well as PHENYL ALANINE AMMONIA LYASE 3 (PAL3), MITOGEN ACTIVATED PROTEIN KINASE 3 (MAPK3), and the RESPIRATORY BURST OXIDASE HOMOLOGUE D (RBOHD) ( Fig. 6C; Supplementary Fig.  S5). Altogether, the transcriptome profiling implied a tight relationship of AMI1 with plant stress responses. Intriguingly, the genetic perturbations of AMI1 expression had strong impact on gene expression regulatory processes. Pathway analysis using the MapMan tool (Thimm et al., 2004) highlighted that AP2/ERF-, MYB-, and WRKY-class transcription factors were among the most affected molecular components associated with the regulation of transcription (Supplementary Table S3). This likely suggests a role for IAM, or one of its derivatives, as a signaling molecule.
The ami1 mutants are hypersensitive to osmotic stress and accumulate ABA Transcriptome profiling of ami1-2 implicated a possible connection of IAM accumulation with abiotic stress responses. To gain additional insights into how AMI1 could be involved in plant stress responses, we examined the transcriptional response of AMI1 towards different abiotic stress conditions (Fig. 7). While drought stress and salinity had either no or very little impact on AMI1 expression, the osmotic stress treatment strongly repressed the expression of AMI1 (Fig. 7B). To further characterize salinity and osmotic stress responses in ami1, we analysed the growth behavior of mutant and WT seedlings under stress conditions. The survival rate of the seedlings was assessed on the basis of first leaf emergence. When grown on 100 mM NaCl, ami1 seedlings showed no considerable difference from WT. However, under osmotic stress conditions the survival rate of ami1 seedlings was clearly reduced (Fig. 7A). The hypersensitivity of ami1 seedlings towards osmotic stress conditions implies that the controlled repression of AMI1 expression under osmotic stress conditions is an important mechanism in the stress adaptation process.
Osmotic stress responses are mediated by ABA-dependent and ABA-independent pathways (Agarwal et al., 2006;Yoshida et al., 2014). The microarray data underpinned the activation and IAM amidohydrolase overexpressing (B) mutants with wild-type (WT) seedlings. Seeds were germinated and grown for 7 d on ½ MS plates, before they were transferred onto plates containing indicated amounts of either IAM or IAA. In order to examine the sensitivity toward IAM and IAA in the medium without the bias of initially shorter primary roots, the impact of the two compounds was expressed in relative terms. Therefore, the length of the longest primary root of each genotype grown under control conditions was set to a value of 1 and all other roots of the corresponding genotype were expressed relative to this value. The primary root elongation of seedlings was quantified. The expression of the transgene in the AMI1ind lines was induced by adding 10 µM β-estradiol to the plates. At least 16 seedlings of each genotype were measured for each condition. The data represent means ±SE. Asterisks indicate significant differences between the corresponding WT control and the tested genotypes under the described conditions. Student's t-test: *P≤0.05, **P≤0.01. of both pathways through the transcriptional regulation of key genes such as NCED3 as well as several MYB and DREB class transcription factor genes, respectively. Dioxygenases of the NCED family catalyse a rate-limiting step of ABA biosynthesis. Mass spectrometric assessment of ABA levels in WT and ami1 mutants revealed significantly increased hormone levels in the mutants (Fig. 7C). To investigate whether NECD3 expression is responsive to ABA or to IAM, the transcriptional response of the gene was analysed by qPCR. Both substances induced NCED3 expression (Fig. 7D). However, the observed effects were stronger in WT compared with mutant seedlings. Most interestingly, IAM showed a stronger effect in WT seedlings relative to ABA, which suggests that IAM acts either in parallel to ABA or upstream of the hormone in the transcriptional regulation of NCED3. Since AMI1 gene expression was shown to be regulated by osmotic stress, we asked whether ABA could be involved in controlling AMI1 expression in a feedback loop. On the basis of qPCR experiments and the quantification of promoter activity in pAMI1::GUS lines, we were not able to detect any regulatory effect of ABA on the expression of AMI1 (see Supplementary Fig. S6).

Impaired AMI1 expression translates into decreased seed and embryo size
ABA is a major determinant of seed dormancy and germination (Vishal and Kumar, 2018). For this reason, we tested the effect of IAM on seed germination in comparison with ABA. Wild-type seeds were germinated on ½ MS plates containing methanol (mock), 10 µM IAM or ABA (Fig. 7E). As expected, ABA blocked seed germination nearly completely. Relative to the mock control, IAM also retarded the emergence of the radicle, most likely through the previously described induction of NCED3 expression and the therewith coupled increase in ABA levels. A recent study reported the specific role of the KUP/HAK/KT-family K + transporter KUP9 in controlling cellular auxin homeostasis through the symport of IAA from the endoplasmic reticulum into the cytoplasm (Zhang et al., 2020). Work from our lab related IAA and IAM with potassium influx in Arabidopsis embryos through the transcriptional control of K + transporter gene expression. The work also associated the two indolic compounds with the control of the seed size (Tenorio-Berrio et al., 2018). In order to gain detailed information on the ami1 mutant seed phenotype, we inspected seed and embryo size of ami1 loss-of-function mutants in comparison with WT (Fig. 7F, G). The seed size of both ami1 mutant alleles was significantly reduced. Consistent with this finding, the embryos of the ami1 mutants were also observed to be considerably smaller. Overall, the presented data suggest that IAM triggers ABA production through the stimulation of its biosynthesis. Furthermore, it can be concluded that the accumulation of IAM in the ami1 mutants exerts a negative effect on seed maturation.

AMI1 contributes to auxin homeostasis in vivo
Auxins are well-characterized plant growth regulators ubiquitously distributed throughout the plant kingdom. Despite considerable scientific interest in this substance class, our knowledge of IAA biosynthesis and cellular homeostasis in plants is still incomplete. The formation of IAA is assumed to proceed via several independent pathways that act either redundantly or in parallel to each other (Woodward and Bartel, 2005;Pollmann et al., 2006a;Zhao, 2010). However, only the major auxin biosynthesis pathway via indole-3-pyruvate is yet fully uncovered with respect to the enzymes and intermediates involved (Zhao, 2014;Kasahara, 2016). Among the discussed additional pathways, one is suggested to proceed via IAM, which is further converted by IAM hydrolases to give rise to the active plant hormone, IAA. In recent years, evidence for the wide distribution of corresponding IAM hydrolases in the plant kingdom has been provided (Pollmann et al., 2003;Nemoto et al., 2009;Sánchez-Parra et al., 2014), suggesting a conserved function of AMI1-like IAM hydrolases in plants.
Here, we report the detailed characterization of AMI1 lossand gain-of-function mutants and show that AMI1 contributes to the conversion of IAM into IAA in planta . Remarkably, IAM hydrolase activity was not null in ami1-1 and ami1-2 (Fig. 4D), which suggested the existence of other enzymes that might be involved in the conversion of IAM into IAA, but the analysis of the other members of the amidase signature (AS) family in Arabidopsis provided no indication for a second specific IAM hydrolase in this group (Pollmann et al., 2006b;Neu et al., 2007). A recent Fig. 7. Relationship between AMI1 expression levels and ABA. (A) Abiotic stress assay growing control plants (Col-0) and ami1 mutants for 12 d under control, salt (100 mM NaCl), and osmotic stress (300 mM mannitol) conditions. Stress resistance was assayed on the basis of first true leaf establishment. Qualitative data are given in the figure. (B) Transcriptional changes in AMI1 expression in response to different stress treatments in comparison to control conditions. Gene expression levels were normalized to the house-keeping gene APT1 using the 2 −ΔΔCT method. Given are the means with their SE (n=9). (C) GC-MS/MS quantification of ABA contents in wild-type (WT) Arabidopsis and the two ami1 mutant alleles. The box plots display the median, quartiles and extremes of the compared data sets (n=5). (D) Induction of NCED3 gene expression by 10 µM IAM and ABA in wildtype (WT) and ami1-2 seedlings, respectively. Gene expression levels were normalized to the house-keeping gene APT1 using the 2 −ΔΔCT method. Means with their SE are given (n=9). (E) Effect of either 10 µM ABA or IAM on the germination of WT Arabidopsis seeds. The germination rate was analysed on the basis of radicle emergence. Means with their SE are given (n=20). (F) Relative seed size of WT Arabidopsis and ami1 mutant seeds. To analyse the seed size, seeds were photographed and the perimeter of at least 20 seeds determined using the Fiji software (Schindelin et al., 2012). Different letters indicate significant differences between the means of the compared datasets analysed by ANOVA and a Tukey-Kramer post hoc test (P<0.05). (G) Size difference of representative WT (Col-0) and ami1 mutant embryos. The photographs are scaled to the same size and scale bars (1 mm) are given. study, however, associated two additional formamidase-like proteins, IAMH1 and IAMH2, with this reaction (Gao et al., 2020), which is likely the reason for the remaining enzymatic activity. It will be highly interesting to introgress the ami1 mutation into the iamh1 iamh2 CRISPR/Cas line in a future project, in order to fully understand the role of IAM in plants.

The IAM pathway is not essential for general IAA supply
The analysis of the root system architecture of ami1-1 and ami1-2 revealed significant differences in the specific root branch density, as well as the specific root area and length (see Supplementary Fig. S2). AMI1 expression in primary roots declined shortly after germination (Fig. 2), but apparently a tight regulation of the spatio-temporal distribution of AMI1 is important to facilitate proper root growth. On the contrary, the conditional overexpression of AMI1 had no significant effect on root growth ( Supplementary Fig. S2). Taking the increased response of AMI1ind lines towards exogenously applied IAM into account (Fig. 5), this implies that the IAM pool size in roots is low or not accessible for recombinant AMI1. Our observation is consistent with the lack of any considerable root phenotype when the iaaM and iaaH genes from Agrobacterium tumefaciens are expressed under the control of their natural promotors in Arabidopsis (van der Graaff et al., 2003). Apparently, the IAM pool in areal plant parts must be bigger or more accessible to recombinant AMI1, as the leaves of AMI1ind plants showed clear phenotypic alterations (Supplementary Fig. S3) that can be associated with auxin overproduction, resembling the phenotypes of mutants such as yuc1-D (Zhao et al., 2001) or the sur2 mutant (Delarue et al., 1998), although the effects in the AMI1ind lines are somewhat weaker.
However, on a whole-plant scale, the phenotypic alterations caused by mutations in AMI1 are only moderate ( Fig. 4; Supplementary Fig. S2), and the transcriptomics analysis of ami1-2 did not provide evidence for a consistent induction of genes involved in other auxin biosynthesis pathways, and nor are IAMH1 or IAMH2 induced to compensate the lack of AMI1 in the mutant (Supplementary Table S2). Hence, it must be concluded that the IAM pathway is not strictly necessary for the general supply of IAA in Arabidopsis, although some particular developmental processes, such as lateral root growth or seed maturation (Fig. 7), may depend on AMI1 action to achieve full effectiveness.

Transcriptomics analyses revealed a link between AMI1 and plant stress responses
Our expression analyses demonstrated that AMI1 transcription is differentially regulated by IAM and IAA, even though IAA appeared to be less effective controlling AMI1 expression (Fig. 3). This further suggested a tight transcriptional connection between AMI1 and both its substrate and reaction product. Most interesting, however, was the observation that the genetic reduction of AMI1 levels caused no major alterations of auxin homeostasis-related gene expression. This contrasts with the results obtained for the conditional AMI1 overexpression line. Here, the induction of a few auxin conjugation-related genes suggested the activation of compensatory effects, likely triggered to counteract auxin overproduction in the line (see Supplementary Table S2).
The comprehensive profiling of transcriptional alterations in AMI1 mutants revealed a strong impact on genes related with biotic and abiotic stress responses, including key enzymes for the biosynthesis of JA and ABA ( Fig. 6; see Supplementary  Tables S2, S3). Along with numerous marker genes for ABA signaling, e.g. RD26A, COR47, ERD4, ERD9, and ERD10, the gene for a key-enzyme of ABA synthesis, NCED3, appeared strongly induced in ami1-2. The overexpression of NCED3 has recently been associated with an improvement of drought tolerance in soybean (Molinari et al., 2020). Both the induction of NCED3 through IAM application and the accumulation of ABA in the ami1 alleles could be confirmed (Fig. 7C, D), which provided strong arguments for an intimate relationship between AMI1 and IAM contents, respectively, and ABA. The considerable induction of AP2/ERF transcription factors in ami1-2 (Supplementary Table S3) has to be particularly emphasized, as it underpins our observations of a close connection between AMI1 action and abiotic stress responses. The particular role of AP2/ERF transcription factor networks in orchestrating hormone and abiotic stress responses is well documented (Xie et al., 2019). Quantitative gene expression analysis revealed a significant suppression of AMI1 transcription by osmotic stress (Fig. 7B), and ami1 mutants grown under osmotic stress conditions show a considerably increased sensitivity to the stress conditions (Fig. 7A). In consequence, we conclude that the suppression of AMI1 expression by osmotic stress and the therewith linked accumulation of IAM are important variables that contribute to the fine-tuning of ABAgoverned stress responses by triggering NCED3-mediated ABA biosynthesis in Arabidopsis.
ABA and gibberellins are essential determinants of seed development and dormancy (Carrera-Castaño et al., 2020). A recent publication also highlighted a direct role of auxin in seed dormancy (Matilla, 2020). In addition, crosstalk between IAA and ABA in seed development and germination has been reported to occur on both the transcriptional and the metabolic level (Chauffour et al., 2019;He et al., 2020). With respect to our data on seed development and germination (Fig. 7E-G), it can be concluded that the crosstalk between AMI1/IAM contents and ABA-related processes is also involved in facilitating proper seedling development and germination. We were able to demonstrate that IAM application reduces the germination rate in Arabidopsis, likely through the induction of ABA biosynthesis. Moreover, mutations in AMI1 translated into aberrant embryo and seed size.
The Arabidopsis AS superfamily member FATTY ACID AMIDE HYDROLASE (FAAH) is associated with numerous developmental processes, including plant growth and early flowering, and interacts with ABA signaling (Wang et al., 2006;Teaster et al., 2007Teaster et al., , 2012. FAAH catalyses the hydrolysis of N-acylethanolamines, lipid signaling molecules that orchestrate a wide array of physiological processes in multicellular eukaryotes (Chapman, 2004), which terminates their action (Aziz and Chapman, 2020). Although there are marked differences between FAAH and AMI1 with respect to their substrate preferences (Pollmann et al., 2006b), the growth inhibiting properties of their preferred substrates led us to the hypothesis that the main role of AMI1 might be similar to that of FAAH, i.e. terminating the action of IAM through its conversion to free IAA and NH 4 + . Particularly in plant stress responses and seed development, this mechanism involving the IAM-induced biosynthesis of ABA (Fig. 8) seems to be important to drive proper adaptational responses. In consequence, the AMI1-dependent metabolic flux through the IAM shunt is seemingly an important regulatory variable that connects auxin-mediated plant growth processes with plant stress responses. It will be highly interesting to further investigate the regulatory role of IAM in plants and how the corresponding signal is perceived and translated into downstream responses. At the same time, it appears tempting to us to further investigate the transcriptional networks that are involved in the integration of the IAM signal.

Supplementary data
Supplementary data are available at JXB online. Fig. S1. Expression analysis of AMI1 in WT and genotyping of ami1 alleles. Fig. S2. Root phenotype of AMI1 mutant plants. Fig. S3. Phenotype of conditionally AMI1 overexpressing AMI1ind lines. Fig. S4. Root growth responses of conditional AMI1 overexpression lines towards IAM and IAA in the media. Fig. S5. Validation of microarray data by qRT-PCR, related to Fig. 6. Fig. S6. Effect of ABA on AMI1 gene expression. Table S1. Primers used for genotyping, cloning and expression analysis. Table S2. Microarray analysis. Table S3. GO and MapMan analysis of selected up-and down-regulated genes in the AMI1 mutants. and SP performed the experiments; MMPA, RGB, HA, MDW, and SP analysed and interpreted the data; SP wrote the paper with significant input from all other authors.

Data availability
All data supporting the findings of this study are available within the paper and within its supplementary data published online.