Accumulation of the Auxin Precursor Indole-3-Acetamide Curtails Growth through the Repression of Cell Proliferation and Development-Related Transcriptional Networks

: The major auxin, indole-3-acetic acid (IAA), is associated with a plethora of growth and developmental processes including embryo development, expansion growth, cambial activity, and the induction of lateral root growth. Accumulation of the auxin precursor indole-3-acetamide (IAM) induces stress related processes by stimulating abscisic acid (ABA) biosynthesis. How IAM signaling is controlled is, at present, unclear. Here, we characterize an ami1/rooty double mutant, that we initially generated to study the metabolic and phenotypic consequences of a genetic blockade of the indole glucosinolate and IAM pathways in Arabidopsis thaliana . Our mass spectrometric analyses of the mutant revealed that the combination of the two mutations is not sufficient to fully prevent the conversion of IAM to IAA. The detected strong accumulation of IAM was, however, recognized to substantially impair seed development. We further show by genome-wide expression studies that the double mutant is broadly affected in its translational capacity, and that a small number of cell proliferation and plant growth regulating transcriptional circuits are repressed by the high IAM content in the seed. In accordance with the previously described growth reduction in response to elevated IAM levels, our data support the hypothesis that IAM is a growth repressing counterpart to IAA.


Introduction
The auxin indole-3-acetic acid (IAA) is a major endogenous growth factor in plants and linked with a great variety of different developmental processes and adaptive responses, including elongation growth, polar development, cambial activity, gravitropism and phototropism, respectively. Thus, it is not surprising that IAA is recognized as an essential phytohormone necessary to ensure optimal plant growth and development [1,2]. The main source of auxin in plants is the indole-3-pyruvate (IPyA) pathway, encompassing tryptophan aminotransferases (TAA1/TAR2) and flavin containing monooxygenases (YUC1-11) that convert L-Trp to IAA via the intermediate IPyA [3,4]. The spatio-temporal expression of these key players in auxin biosynthesis is tightly developmentally controlled [5]. Alongside the main IPyA pathway, plants are assumed to involve a small number of additional pathways in the formation of IAA. These pathways act either redundantly or in a parallel manner with the main route [6][7][8]. The Brassicaceae family, including the model plant Arabidopsis thaliana, possesses the additional indole-3-acetaldoxime (IAOx) pathway ( Figure 1). IAOx is considered to be an important metabolic branching point, involved in connecting primary and secondary metabolism [9][10][11]. Remarkably, IAOx serves as the common biochemical source for the formation of L-Trp derived glucosinolates [12], such as glucobrassicin, and camalexin [13], two important plant defense compounds in Arabidopsis. The formation of both metabolites is induced in response to biotic stresses and involves the transcription factors MYB34, MYB51, MYB122 and WRKY33 for the control of indole glucosinolate and camalexin biosynthesis, respectively [14,15]. The cyp79b2/cyp79b3 double mutant, deficient in the formation of IAOx, shows a wild type-like phenotype and unaltered IAA levels under normal growth conditions [9,10,16], but demonstrates a significantly increased susceptibility towards pathogens [17], which underscores the importance of the IAOx pathway for biotic stress responses. Furthermore, cyp79b2/cyp79b3 has been reported to contain drastically reduced indole-3-acetamide (IAM) contents, suggesting that a large proportion of IAM in Arabidopsis originates from IAOx [10]. IAM has, however, been found in several non-Brassica plant species and, therefore, do not possess the IAOx pathway [18]. Hence, it must be concluded that there is another biosynthetic route leading to IAM, possibly through yet unidentified tryptophan 2monooxygenases in plants that share homology with known from enzymes from bacteria, such as iaaM or tms1 [19,20].
Indole glucosinolate biosynthesis mutants such as sur1/rty and sur2 are, on the other hand, characterized by strongly increased auxin contents and high-auxin phenotypes [21][22][23]. Mass spectrometric analyses demonstrated a significant increase of IAOx, IAM, and IAA in sur1-1, while indole-3-acetonitrile (IAN) levels appeared unchanged [10], which is suggestive for a redirection of the metabolic flux into the IAM pathway.
Increased IAM level in ami1 mutants have recently been shown to trigger abscisic acid (ABA) biosynthesis through the induction of NCED3, encoding for a key enzyme in the formation of ABA [24]. Along with the observation that IAM represses the expression of the K + transporters HAK/KT12 and KUP4 that are assumed to contribute to elongation growth [25], this led to the hypothesis that the conversion of IAM to IAA catalyzed by IAM hydrolases represents a molecular nexus involved in the crosstalk between auxin and ABA. Although the transcriptomics analysis provided conclusive evidence for an intimate connection of AMI1-mediated IAM conversion and abiotic stress responses, the molecular mechanism by which changes in IAM levels are perceived and integrated remain largely elusive.
Here, we characterize the impact of the simultaneous genetic blockade of the indole glucosinolate and IAM pathways in A. thaliana. By phenotypically and mass spectrometrically analyzing an ami1/rty double mutant, we demonstrate that accumulating IAM exerts a negative impact on embryo development. In addition, we report the comprehensive transcriptomics analysis of the mutant, which provided evidence for a number of physiological processes that are affected by the elevated IAM level in the mutant.

The introgression of the ami1-2 mutation into rty1-1 is not sufficient to restore the wild type phenotype
Previous reports highlighted significant alterations in metabolite fluxes in indole glucosinolate biosynthesis mutants, including sur1/rty and sur2. The experiments suggested an accumulation of IAOx and its further channeling into IAA [22,23,[26][27][28]. While the cyp79b2/cyp79b3 double mutant that is practically devoid of IAOx shows wild type IAA levels, downstream mutations in sur1 translate into significantly increased IAOx, IAM, and IAA contents, suggesting the channeling of IAOx into IAA to proceed via IAM ( Figure 1) [10]. The significantly elevated IAA levels in those mutants provoke strong auxin overproduction-related phenotypes, including epinasty of cotyledons and true leaves, significant amplification of lateral root numbers, and an increased root hair number and density [21,23].
Picking up on these lines of evidence, which suggest a purely auxin-mediated phenotype, we reasoned that a genetic block of the IAM pathway (IAOx→IAM→IAA) in indole glucosinolate mutants may restore a wild type (wt) phenotype, due to the impaired conversion of IAM to IAA. To this end, we introgressed the ami1-2 mutant allele into the rty1-1/+ genetic background. After two rounds of selfing, we selected for plants homozygous for the ami1-2 mutation using PCR genotyping. Subsequently, three lines homozygous for ami1-2 were selected and subjected to cDNA sequencing of the RTY gene (At2g20610), in order to genotype the zygosity state of the rty1-1 mutation in these lines. Since rty1-1 originates from an ethylmethane sulfonate (EMS) mutagenesis approach, sequencing was necessary to genotype the state of the rty1-1 mutation in the selected lines. As can be taken from Figure 2, two of the selected lines, line 5 and line 15, appeared to be homozygous ami1/rty double mutants. At the protein level (Figure 2b), the RTY1-1 point mutation translates into a P213S amino acid exchange [29], which is likely to cause major structural changes in the RTY enzyme. Protein modelling of RTY and RTY1-1 against the 1.7 Å crystal structure of a closely related prephenate aminotransferase from Arabidopsis [30] (PDB: 6F5V) and subsequent structure predictions suggested the loss of one -helix motif in RTY1-1 (13 -helices) relative to the wild-type protein (14 -helices) in the C-terminal part of the protein.

Figure 2.
Genotyping of ami1/rty mutants. (a) PCR analysis comparing three selected double mutant lines with wt (Col-0) and ami1-2 control (Ctrl) seedlings. The figure shows the results of three individual PCR reactions using either a T-DNA specific primer (Lb) and a 5' AMI1-specific primer (Pr1), a T-DNA specific primer (Lb) and a 3' AMI1-specific primer (Pr2) or the two AMI1-specific primers, Pr1 and Pr2 (upper panel). Section of the alignment of the annotated RTY cDNA with the sequences obtained from the sequencing of three selected mutants homozygous for the ami1-2 mutation. The figure shows the point mutations at position 637 in lines 5 and 15 (grey box). The analysis of the trace files of line 17 revealed the heterozygosity of this line. Our results confirm the previously described point mutation for rty1-1 (C→T)637 [29]. Nucleic acid positions in the cDNA section are indicated (lower panel). (b) Overlay of the RTY (red) and RTY1-1 (blue) 3D models (upper panel). The loop section comprising the P213S mutation (white arrow) and the altered -helical structure (yellow arrows) are indicated. The lower panel additionally highlights the observed secondary structure changes.
Against our initial expectations, the ami1/rty double mutants still showed an auxin overproduction-related phenotype (Supplementary Figure S1), suggestive of a remaining conversion of IAM to IAA. This observation is in line with the recent identification of two additional acetamidase/formamidase family proteins described to act as IAM amidohydrolases in Arabidopsis [31]. Hence, it is plausible that the two IAM amidohydrolases, IAMH1 and IAMH2, are responsible for the remaining conversion of accumulating IAM to IAA. Interestingly, transcriptomics data from the presumably IAM accumulating sur2 mutant and iaaM overexpressing tomato plants provide no indication for transcriptional alterations of the IAM amidohydrolase genes [32,33]. Along with the findings that IAM application triggers strong auxin-related root phenotypes in wild-type Arabidopsis and conditional AMI1 overexpressor lines [24] and combined with the observation that Arabidopsis plants overexpressing the bacterial tryptophan 2-monooxygenase iaaM gene, producing IAM from L-Trp, also show considerable IAA and IAM overproduction [34,35], it must be concluded that Arabidopsis plants maintain a basal level of IAM amidohydrolases that readily convert accumulating IAM to IAA.

The ami1/rty double mutant is compromised in germiation
When further analyzing the offspring of the two identified ami1/rty double mutants, we found a severe impairment of the germination process for both lines. The obtained seeds did either not germinate at all or aborted germination after a very short period of time, ranging between 48 and 72 h, shortly after radicle protrusion through the seed coat ( Figure 3a). The ami1/rty seeds abort germination shortly after the radicle broke through the seed coat, while wt seeds (Col-0) were able to germinate. (b) Analysis of IAM and IAA levels in wild-type Arabidopsis, iaaMox, and ami1/rty. Imbibed seeds from wild-type Arabidopsis (Col-0), iaaMox, and the ami1/rty double mutant were analyzed for their endogenous IAM (upper panel) and IAA levels (lower panel). The plots show mean values with their corresponding standard errors (SE), n = 5. To assess significant differences between the endogenous levels of each compound an analysis of variance (ANOVA) of the scores followed by Tukey's post-hoc test was performed. The different letters indicate significant differences, p  0.001.
The mass spectrometric analysis corroborated the high-auxin phenotype, providing evidence for a significant 10-fold increase of IAA levels in imbibed ami1/rty seeds compared to wt seeds. This underscores the notion that IAMH1 and IAMH2, respectively, contribute to the conversion of IAM to IAA in Arabidopsis. The iaaMox seeds, on the other hand, displayed an only 2-fold increase of endogenous IAA levels relative to wt. With respect to the IAM contents, it became evident that ami1/rty seeds accumulate over 45-times the amount of IAM of wt seeds. In agreement with previously published data of IAM contents in conditionally iaaM overexpressing seedlings [35], the iaaMox seeds showed a significant accumulation of IAM (14-times). With respect to the data published for sur1-1 [10] that reported a 34-and 21-fold increase of endogenous IAM and IAA levels, respectively, the presented data underscore the role of AMI1 in the conversion of IAM to IAA in Arabidopsis. Apparently, the introgression of ami1-2 into the rty1 indole glucosinolate biosynthesis mutant background considerably altered the conversion of IAM into IAA and led to an enhanced accumulation of IAM.
Auxin has been described to control seed dormancy in Arabidopsis through the interaction with ABA signaling in an ABI3-dependent manner [36]. Elevated IAA contents were shown to increase the dormancy in transgenic iaaMox seeds. In order to exclude dormancy effects through the significantly increased IAA level in the double mutant, we tried different stratification periods, ranging between 2 to 6 days, but under none of the tested conditions, we were able to rescue ami1/rty mutant plants. However, given that most seeds were able to initiate germination, it must be concluded that seed dormancy is not the decisive factor responsible for the observed abortion of the germination process in the double mutants.

Alterations in cellular IAA and IAM contents impact embryo size in Arabidopsis
It is widely accepted that auxin synthesized by fertilized achenes drives fleshy fruit development and ripening, e.g. in strawberries [37][38][39], and that genetic boosting of auxin formation in ovules through the tissue-specific expression of the iaaM gene promotes parthenocarpy in tobacco, eggplant, and tomato [40][41][42]. The role of auxin in regulating seed development in dry dehiscent fruits, such as in the model plant Arabidopsis, is still less well understood, although great progress has been made over the past years [43][44][45]. In previous studies, we observed that the overproduction of IAA in Arabidopsis significantly increased seed size, while the accumulation of IAM was shown to reduce it [24,25]. Auxin has also been demonstrated to significantly impact seed size and starch accumulation in pea [46]. In Arabidopsis, the embryo takes up large part of the space in the seed and, thus, greatly determines seed size, as the endosperm is reduced to a single cell layer [47]. In order to quantify the effect of altered IAA and IAM contents on seed development, we compared the embryo sizes of a number of auxin biosynthesis-related mutants in comparison to wt embryos ( Figure 4). As displayed in Figure 4, embryos of auxin overproducing YUC8ox and iaaMox lines [34,48] were significantly bigger than wt embryos, whereas the IAM accumulating ami1 embryos [24] appeared to be considerably smaller. This confirmed our previous findings and highlighted the already well-recognized growth promoting effect of auxin in fruit development. The effect on seed development in YUC8ox was so strong that the siliques were not able to follow the increased seed growth which, in consequence, resulted in the premature opening of the valves of non-dehiscent siliques along the replum (Figure 4c). At the same time, our results suggested that either the excessive accumulation of IAM or the partially blocked conversion of IAM to IAA in the ami1 mutant have growth inhibiting effects. Interestingly, the rty mutant [23,49], which is allelic to superroot1 (sur1) [21], showed a slightly bigger embryo size than wild-type Arabidopsis. The rty seedlings are known to have elevated IAA contents [23], and can be expected to also have strongly elevated IAM levels, like sur1 [10]. Taking the heterozygosity of the tested rty1-1 seeds into account, the rty1-1 embryos resemble those of iaaMox, that also contain elevated IAM and IAA level [19,34]. The largely IAM deficient cyp79b2/cyp79b3 double mutant displayed no significant alteration of embryo size, which can likely be attributed to the wt IAA levels in this mutant [9,10,16]. Most remarkably, however, was the strong embryo phenotype of the ami1/rty double mutant. Potentiating the ami1-2 embryo phenotype previously described [24], the ami1/rty double mutant embryos showed to be significantly smaller than wt and ami1-2 embryos (Figure 4a,b).

The ami1/rty mutant shows major impairments in gene transcription and protein synthesis
Considering the drastic curtailing of embryo growth in the ami1/rty double mutant and the detected significant accumulation of IAM in the mutant seeds, we concluded that the canalization of IAM into IAA during seed filling must be an important factor for proper seed development. It seems as if the excessive accumulation of IAM triggers limited embryo development, while the observed growth arrest of the ami1/rty mutants during the germination process is possibly attributable to the limitation of storage compounds in the seed. To address the question what molecular processes are affected and responsible for the strong phenotype, we decided to take a comprehensive transcriptomics approach, comparing the transcriptional profile of ami1/rty mutant seeds with that of wt seeds by RNAseq. First, we analyzed the expression levels of 128 selected auxin metabolism-, signaling-, and transport-related genes in the ami1/rty mutant relative to wt (Table 1). Genes that exhibit a significant difference in their expression are marked in bold letters.
The directed assessment of target gene expression levels revealed that genes associated with auxin de novo-biosynthesis were not significantly altered in their transcription. This underpins that the high auxin contents in the mutant must derive from the genetic block of indole glucosinolate biosynthesis and the remaining flux of IAM to IAA, and that alteration of de novo auxin biosynthesis is not involved. At the same time, we found a significant induction of the Gretchen Hagen gene GH3.5. GH3.5 encodes an acyl amino synthetase that preferentially conjugates IAA to aspartate if IAA and Asp contents are high [50]. This may be interpreted as a response to counteract the high IAA contents in the mutant. Inconsistent with this notion of buffering high IAA contents through conjugation, we determined the significant repression of an UDP-glucose:indole-3-acetate ß-Dglucosyltransferase (IAGLU). IAGLU is assumed to be capable of glycosylating IAA [51], as it shares homology with known IAGLUs from maize [52]. However, the in vivo-function of IAGLU from Arabidopsis is yet to be elucidated, particularly because another publication described an enzymatic activity of IAGLU towards kaempferol [53]. Thus, IAA may not be the endogenous substrate for IAGLU in vivo. With respect to the biosynthesis of L-Trp derived defense compounds, i.e. camalexin and indole glucosinolate, we were not able to find conclusive evidence for the redirection of the metabolic flux into the biosynthesis of camalexin. Consistent with the role of the significantly induced transcription factor WRKY33 as negative regulator of camalexin biosynthesis [15], we registered no induction of camalexin biosynthesis-related genes, such as CYP71A13 and PAD3. Hence, it must be concluded that there is no alternative metabolic bypass and the entire metabolic flux normally directed into indole glucosinolate biosynthesis now passes through the IAM shunt. Additionally, we detected no significant changes in the expression of auxin transport-related genes, and only the IAA9, IAA12/BODENLOS (BDL), and PLETHORA5 (PLT5) transcriptional regulators appeared to be differentially expressed in the mutant. While the reduced expression of IAA9 and PLT5 may be neglected because of their only minor differential expression, the repression of IAA12/BDL was far more pronounced. IAA12/BDL is known to affect primary root formation and apical-basal patterning in embryos [54]. Aux/IAA auxin signaling repressor proteins act in response pairs with their particular auxin response factor (ARF) transcription factors. IAA12/BDL is described to closely interact with MONOPTEROS (ARF5) [55,56]. Hence, the transcriptional repression of IAA12/BDL may result in an increased amount of free ARF5, because of a reduced recruitment of the transcription factor by its transcriptional repressor into inactive heterocomplexes. However, apical-basal pattern formation is apparently no issue in ami1/rty, as the embryos showed a clear polar organization ( Figure  4b).
Since our targeted approach did not disclose auxin related processes that could explain the embryo and germination phenotype of ami1/rty, we next undertook an in-depth transcript profiling approach to obtain a broader picture of differentially expressed genes (DEGs) and their functional relationships. Employing an adjusted p-value (FDR) of < 0.05 and an arbitrarily chosen differential expression value of log2FC ≥ +1 for induced genes and log2FC ≤ -1 for repressed genes, respectively, 101 induced and 564 repressed DEGs were identified (see Supplementary table S1). To functionally score the DEGs, we used the MapMan software. The application returned a small number of key processes that are affected in the double mutant. These processes include RNA processing, ribosome assembly, nucleotide metabolism, translation, as well as protein modification and degradation. In addition, the results pointed towards the transcriptional alteration of stress response-, signaling-, and transport-related genes (see Supplementary table S1).
With the aim to gain more detailed insight into the relationships of the selected DEGs, we performed a functional association network analysis. For the induced DEGs, we obtained a network with 94 nodes and 62 edges. Further analysis of the network, focusing on the node with the highest degree of connectivity, WRKY40, provided a subnet with 8 nodes and 15 edges. Along with the genes ZAT7, CML37, AOX1A, and ACS6, the central WRKY40 hub could be associated with oxidative stress responses, while its association with SYP121 and the glycolipid transfer protein gene At4g39670 pointed towards an additional role in cell membrane organization. When searching for the nodes with the highest betweenness centrality, we identified the transcription factor FUS3, which appeared to be connected in a small sub-network with the seed storage albumin genes SESA2 and SESA4, as well as with the fatty acid desaturase FAD2. From this it may be concluded that seed storage protein biosynthesis is sought to be induced in the double mutant, but the comprised translational machinery in this mutant inhibits the process. For the downregulated DEGs, we inferred a network with 554 nodes and 7128 edges. The gene with the highest degree of connectivity was the ribosomal protein gene REDUCED POLLEN NUMBER 1 (RDP1) [57] with 140 connections, followed by the ribosome biogenesis related FIBRILLARIN 2 (FIB2) gene [58] and the RIBOSOMAL LARGE SUBUNIT 4 (RPL4) gene with 138 and 133 connections, respectively. RPL4 is a particularly interesting candidate, because it has been associated with auxin-regulated developmental processes, endomembrane trafficking, and lipid metabolism [59]. However, from a more general perspective, the network analysis referred to an impairment of ribosome-dependent processes, which confirms the previously discussed classification of the DEGs. The Kyoto Encyclopedia of Genes and Genomes (KEGG) [60] pathway enrichment analysis of the three mentioned sub-networks also highlighted ribosome-and ribosome biogenesis-related processes to be underrepresented in the ami1/rty mutant (Supplementary Figure  S2). Among the downregulated DEGs, the genes NITRATE REDUCTASE 1 (NIA1), PHOSPHOGLUCOSE ISOMERASE 1 (PGI1), and the uridine 5'-monophosphate synthase gene At3g54470 showed the highest betweenness centrality. On the basis of the topology of the extracted network for the three central hubs, we performed a KEGG pathway enrichment analysis. As can be taken from Figure 5, the network inferred from the betweenness centrality in combination with a projection onto the KEGG pathway maps identified a series of metabolic routes, including amino acid and carbon metabolism pathways, that are seemingly affected in the ami1/rty double mutant. Most remarkable, however was the high betweenness centrality of NIA1. NIA1 encodes the cytosolic minor isoform of nitrate reductase NIA2 and accounts for about 15% of the nitrate reductase activity in plants [61]. Nitrate reductases and nitrite reductases catalyze the first two consecutive steps of nitrate assimilation in plants, a process that is closely related with the biosynthesis of amino acids.
Gene Ontology (GO) biological process terms represent a rich resource for the functional characterization of large 'omics' data sets. GO annotations include a mix of manually curated and electronically inferred sources [62]. Taking advantage of this resource, we next subjected the previously derived functional association networks for both up-and downregulated DEGs to a GO enrichment analysis. As demonstrated in Figure 6, the inferred GO enrichment map largely confirmed the previously obtained results, highlighting a significant underrepresentation of biological processes related with translation, including ribosome biogenesis and assembly, rRNA processing, and ribonucleotide biosynthesis. However, in addition to these already discussed processes, the GO enrichment analysis provided additional evidence for the underrepresentation of some interconnected seed development-and germination-related terms, while a small number of GO terms associated with seed maturation appeared to be overrepresented. Most notably, the study brought additional insight into the enrichment of GO classifications associated with drought and osmotic stress responses to light. Moreover, it highlighted the impairment of the temperature stress response in ami1/rty. Figure 6. GO term enrichment map for up-and downregulated DEGs. GO terms that share members are shown in connected clusters. Cluster labels were retrieved using the AutoAnnotate v1.3.3 application in Cytoscape. The node color reflects the normalized enrichment score with underrepresented terms represented in blue and overrepresented terms given in red. The node size gives account on the number of members in the particular GO term.
Among the overrepresented abiotic stress related genes, the cell wall remodeling associated EARLY RESPONSE TO DEHYDRATION (ERD7) [63,64] gene, the transcription factor ZAT10 [65], and the galactinol synthase gene GOLS2 [66] are noteworthy. All three genes have been demonstrated to play central roles in responses to drought and salt stress in Arabidopsis. Their induction may suggest that the mutant plant is prone to drought stress, as it is possibly not able to efficiently control its water balance over the course of seed development. This argument is further strengthened by the observed induction of FUSCA 3 (FUS3) in the mutant. FUS3 is an ABA-responsive master regulator of seed development and essential for the establishment of desiccation tolerance in seeds [67], including the substantial induction of LATE EMBRYOGENESIS ABUNDANT PROTEIN (LEA) genes [68]. Intriguingly, an intimate relationship between IAM accumulation in the ami1 mutants and the induction of ABA biosynthesis and downstream regulatory processes has just recently been reported [24].
With respect to the repressed genes falling into the seed development and germination clusters, the ABC transporter gene ABCG11, which contributes to cutin and suberin metabolism as well as to reproductive organ development [69], and the WDxR motif-containing protein gene WDR55 that is involved in the control of seed development [70] have to be mentioned. In addition, we registered the significant downregulation of the S-adenosyl-L-homocysteine hydrolase HOG1. HOG1 is known to interact with the likewise repressed RECEPTOR FOR ACTIVATED C KINASE 1A product RACK1 [71], and hog1 knockout mutants are described to have an embryo growth arrest at the globular stage [72]. The downregulation of the HOG1 hub may contribute to the curtailed embryo development of the ami1/rty mutant. Moreover, we identified SLEEPY1 (SLY1) as a potentially interesting candidate. SLY1 encodes a F-box protein that targets gibberellin signaling repressors and, thus, interferes with the gibberellin response [73]. Intriguingly, sly1-10 mutants exhibit a dwarfish phenotype, and many sly1-10 plants were infertile, which leaves room for the speculation of the contribution of SLY1 and downstream alterations in the crosstalk with gibberellins to the ami1/rty phenotype.

Hyperaccumulation of IAM provokes repression of cell proliferation and elongation growth regulatory pathways
In order to explore gene regulatory networks that are possibly involved in causing the ami1/rty embryo phenotype and to experimentally validate the conclusions drawn from the RNAseq experiment, we centered our interest on transcription factors with significantly altered expression levels. A compilation of these factors can be found in Supplementary Table S1. Overall, 34 transcription factors were identified, from which eight were significantly induced, while 26 appeared to be significantly repressed. As presented in Figure 7, ten of these genes were selected and their transcriptional response towards a short-term treatment with IAM was tested by qRT-PCR. The most pronounced induction was registered for the MYB4 gene, followed by ZAT11. MYB4 is a R2R3-subfamily MYB domain protein associated with radiation and thermal stress responses in Arabidopsis [74,75]. The induced Cis2/His2-type zinc finger proteins ZAT7 and ZAT11 are also associated with abiotic stress, particularly with responses to salinity stress and nickel tolerance, respectively [76,77]. Along with these genes we found a pronounced induction of the WRKY-domain protein gene WRKY40, which is reported to be involved in biotic stress responses and the control of indole glucosinolate biosynthesis [78,79]. Most interesting, however, is the link of WRKY40 expression with ABA signaling in Arabidopsis and its control by the transcriptional regulators GOLDEN2-LIKE 1 and -2 (GLK1/2) [80], because increased IAM contents are known to trigger ABA biosynthesis [24] and GLK2 was registered among the repressed factors.
Among the repressed factors, we also found several stress related target genes, such as the heat shock factors HSFA2 and -3, implicated in thermotolerance and osmotic stress tolerance [81,82], as well as the cytokinin response factors CRF3, CRF4, involved in cold stress responses and freezing tolerance in plants [83,84]. This further substantiates the tight connection between IAM accumulation and plant stress responses. In addition, we identified the significantly repressed TEOSINTE BRANCHED 1/CYCLOIDEA/PROLIFERATING CELL FACTOR genes TCP4, TCP10, and TCP23 as potential targets. TCPs are reported to play pivotal roles in the control of morphogenesis of shoot organs and developmental processes, such as flowering [85,86]. Remarkably, however, is the direct connection of TCP4 and TCP10 with auxin homeostasis related processes, especially the induction of YUC5 by TCP4 [87] and the downregulation of TCP4 and TCP10 in the iamt1-D gain-of-function mutant [88], which implies the existence of regulatory loops that connect developmental processes with endogenous auxin levels that possibly also involve the HD-ZIP protein HAT5.
However, two other groups of transcription factors, including the gibberellin signaling repressors GNC, GNL/CGA1 [89], and the growth-regulating factors GRF3 and GRF5 that are recognized as mediators of organ size establishment and cell proliferation [90,91], attracted our interest, as they promised mechanistic insight into the crosstalk between IAM and gibberellin, as well as with general growth control.
In order to investigate as to whether IAM can trigger similar responses in wild-type Arabidopsis, we tested the gene regulatory effect of exogenously applied IAM in wild-type seedlings. Apart from the induced ZAT11, for which the induction could not be confirmed, we studied nine further genes repressed in ami1/rty. Except for GNC, the repression of all other selected genes by IAM could be confirmed by qRT-PCR analysis. However, it has to be remarked that the impact of the short-term IAM treatment was generally less pronounced than the effect observed in the RNAseq analysis of the ami1/rty double mutant.

Discussion
Auxins are well-characterized phytohormones that control a huge variety of different growthand development-related processes. The role of auxin in pattern formation and embryogenesis has been studied in great detail [92,93]. Several recent publications ascribe auxin also an important role in later stages of seed development, including seed filling, dormancy control, and germination. An example is the pivotal role of local auxin biosynthesis in the endosperm for proper seed coat formation [94], the increased dormancy of IAA overproducing mutant plants [36], or the observed impact of reduced auxin formation on seed size growth and starch formation in legumes [46]. A similar positive regulatory relationship of auxin formation and starch production in rice has also been suggested [95]. On the other hand, nutrient allocation and plant hormone crosstalk are assumed to play essential roles in seed development as well. A number of K + channels and transporters are downstream targets of auxin [25], and auxin biosynthesis is reported to be controlled by sugars [96,97]. Most recent results give reason to speculate on an involvement of auxin-ABA crosstalk in this process [98,99]. However, overall, the role of auxin in orchestrating seed development has received only little attention, despite is undeniable role as key regulator of plant development. A more detailed analysis may entail biotechnological advances that could be harnessed to improve agricultural productivity in an environmentally friendly manner.
Our previous work shed some light on the role of IAA and its direct precursor, IAM, in seed development and a connection with phytohormone crosstalk [24,25,48,100]. Our findings led to the hypothesis that IAM is a negative plant growth regulator and that the enzymatic conversion of IAM to IAA by AMI1 terminates its growth inhibitory action. A similar function of another amidase signature family member, FAAH (FATTY ACID AMIDE HYDROLASE), has already been demonstrated. FAAH catalyzes the hydrolysis of N-acylethanolamines, which represent lipid signaling molecules, thereby controlling their action [101,102]. In this study, we aimed at addressing the question on the molecular and physiological consequences of endogenously accumulating IAM contents in Arabidopsis. For this reason, we decided to cross the indole glucosinolate biosynthesis mutant rty1-1, which has significantly elevated IAM and IAA levels, with the ami1-2 mutant, to block the enzymatic conversion of IAM to IAA. Against our initial expectation, homozygous double mutants showed no wild-type phenotype and IAA levels, but resembled a mild rooty phenotype (Supplementary Figure S1) and were characterized by elevated IAA contents (Figure 3). However, the observed remaining IAM hydrolase activity confirmed our previous observation of a remaining amidase activity in ami1 mutants [24]. Moreover, it is in line with the recent identification of two additional enzymes, IAMH1 and IAMH2, that also contribute to the conversion of IAM to IAA in Arabidopsis [31]. Taking the reported remaining 40 to 45% IAM hydrolase activity of ami1 mutants into account, the obtained data neatly reflect the loss of AMI1 activity in the mutant, particularly because only one of the two other amidases, IAMH1, shows expression during seed development [103].
To our surprise, seeds from the offspring of homozygous ami1/rty mutant plants showed to be nonviable, aborting germination shortly after the radicle broke through the seed coat. Given that the homozygous parent plants germinated normally and set a small number of siliques, it may be concluded that they derived from ancestors heterozygous for rty, and that, e.g., the nutrient acquisition during seed filling is compromised in the homozygous parents, which may explain the impairment of germination in the offspring. To gain closer insight into the embryo phenotype, we inspected the ami1/rty embryos and compared them to a series of other auxin biosynthesis-related mutant embryos. The quantitative assessment of the embryos disclosed a striking reduction of the size of ami1/rty mutant embryos. The phenotypic inspection was accompanied by the mass spectrometric analysis of IAM and IAA contents in the seeds. In comparison to the reported 34-and 21-fold increase of endogenous IAM and IAA levels in sur1-1 mutant [10] that is allelic to rty, we found a 45-and 10-fold increase of endogenous IAM and IAA levels in double mutant seeds. From this, we conclude that the embryo phenotype of ami1/rty is likely attributable to the further elevated IAM contents. To obtain further insight into the role of IAM catabolism on seed development, it will be highly interesting to generate an ami1/iamh1/iamh2 triple mutant in the future, although such mutants might be prone to nonviability.
With the objective to further our understanding about the molecular consequences of the observed IAM accumulation in the dwarfish ami1/rty mutant embryos, we subjected double mutant seeds to whole-genome transcript sequencing (RNAseq) and compared the obtained transcriptional profile with that of wild-type seeds. First, we checked the expression of a subgroup of 128 auxinrelated genes. The targeted analysis provided no evidence for the misregulation of auxin metabolism-, transport-, or signaling-related genes. In particular, the missing activation of genes of the camalexin biosynthesis pathway is noteworthy, because we initially speculated that accumulation of IAOx or IAM may trigger a metabolic redirection into the camalexin pathway. On the contrary, we found the induction of WRKY33, a transcriptional repressor of camalexin biosynthesis ( Table 1). The mRNA sequencing revealed, however, a significant downregulation of RNA processing-and translationrelated genes. The mainly affected processes included ribosome biogenesis and assembly, as well as rRNA processing ( Figure 6, Supplementary Table S1). A directed search for the differential expression of master regulators of ribosome biogenesis, such as SMO4 (SMALL ORGAN 4) or MAS2 (MORPHOLOGY OF ARGONAUTE 1-52 SUPPRESSED 2) [104,105], was however not successful. In addition, our network analysis of the expression data revealed that the ami1/rty mutant seeds fall significantly short in the expression of sugar and amino acid metabolism-associated transcripts. Considering the dwarfish embryo phenotype and impaired germination, we conclude that ami1/rty mutants are most likely compromised in seed filling. The major storage organ in Arabidopsis seeds are the cotyledons. The substantially reduced cotyledon size in the double mutant is suggested to prevent the deposition of adequate storage reserves (oil, protein, starch) in the seed, which ultimately leaves the germinating seedlings with insufficient energy resources to cover the initial growth phase until the seedlings establish an autotroph lifestyle.
The transcriptomics analysis provided additional evidence for the differential regulation of thermal, drought, and osmotic stress related genes. This observation matches well with our previous finding of an intimate connection between abiotic stress responses and increased IAM contents [24]. The misregulation of a substantial number of LEA protein genes and the involvement of drought stress related processes additionally point towards an important misregulation of the establishment of desiccation tolerance related processes. Of particular interest for our future work was, however, the identification of a number of plant growth regulating processes that were significantly repressed in the ami1/rty double mutant. The repression of the growth-regulating factors GRF3 and GRF5 in IAM treated wild-type Arabidopsis seedlings confirmed a direct connection between IAM and the regulation of plant growth. In addition, our bioinformatics analyses brought a small number of plant hormone-regulatory circuits to light, which suggest an even more important role of IAM in plant hormone crosstalk, connecting with gibberellin signaling through the repression of SLY1, a F-box protein that targets gibberellin signaling repressors (DELLA proteins) [73], as well as GNC and CGA1, two transcription factors that act downstream of the DELLAs on gibberellin signaling [89], although the transcriptional repression by IAM was only confirmed for CGA1. Interesting was also the identification of the repression of the three TCP family members TCP4, TCP10, and TCP23 that are assumed to play pivotal roles in the control of shoot morphogenesis and developmental transitions, such as flowering [85,86]. A significant repression of TCP23 in wild type seedlings by IAM could be confirmed by transcript quantification. Particularly, the further investigation and elucidation of IAM perception and signaling will be a thrilling task for our future work.

Plant material and growth conditions
The presented experiments used the Arabidopsis thaliana Col-0 background (N1092) as reference.
Seedlings were grown under sterile conditions on solidified ½ MS-medium containing 1 % (w/v) sucrose in Petri dishes [107]; plantlets were kept under constant environmental short day conditions (8 h light at 24 °C, 16 h darkness at 20 °C, photosynthetically active radiation 105 µmol photons m -2 s -1 from standard white fluorescent tubes) for two to three weeks. Thereafter, plants were transferred to a mixture of soil and sand (2:1) and kept under long day conditions (16 h photoperiod) in a greenhouse, which was maintained under constant climatic conditions, 22 to 24 °C during daytime and 18 to 20 °C overnight. The photosynthetically active radiation was no less than 150 µmol photons m -2 s -1 (supplementary light, if required, from sodium-vapor lamps).

Genotyping of the ami1/rty double mutant
To genotype ami1/rty mutants, a two-step approach has been taken. First, the zygosity state of the ami1-2 T-DNA integration was analyzed by PCR [108]. Instead of running a multiplex PCR with the two AMI1-specific primers Pr1/Pr2 and the T-DNA-specific primer Lb, we performed three independent reactions (Pr1/Pr2, Pr1/Lb, Pr2/Lb) for each selected line. After identifying lines homozygous for the ami1-2 mutation, we extracted total RNA [109] and prepared cDNA libraries from those lines using M-MLV reverse transcriptase and oligo(dT)15 primer according to the manufacturer's instructions. Next, we amplified the RTY gene using the primers RTY-fwd and RTYrev by PCR. See Supplementary Table S2 for primer sequences. The obtained RTY PCR fragments of the three lines have then been sequenced on ABI 3730 xl sequencer by the company Stabvida. Sequence and trace file analysis have been carried out using the CLC Main Workbench 7.

Modelling of the RTY and RTY1-1 protein structure
The three-dimensional structures of RTY and RTY1-1 were modeled by using the SWISS-MODEL interface (http://swissmodel.expasy.org) [110], utilizing the 1.7 Å crystal structure of a bifunctional aspartate aminotransferase and glutamate/aspartate-prephenate aminotransferase (PAT, At2g22250) deposited in the RCSB Protein Data Base (PDB: 6F5V) [30]. In accordance with [29], we used the PAT structure as template, although RTY is structurally closely related with alanine aminotransferases, e.g. from Hordeum vulgare (PDB: 3TCM), because this crystal structure originates from the same species. Structural examination was performed using either PyMOL v1.47 or CLC Main Workbench 7.

Quantitative comparison of embryo sizes
Dry seeds from wild-type Arabidopsis and the different auxin mutant lines were surface sterilized using successive treatments with 70% ethanol (v/v, 5 min) and a 5-7% sodium hypochlorite solution (v/v, 5 min), before being rinsed three times with water. Afterwards, the seeds were left overnight at 4 ºC in DEPC water. Under a binocular (Leica MZ10 F), seeds were then transferred onto an object slide and mixed with glycerol. After removing the seed coat from the embryo, pictures were taken with 25fold magnification using a color CCD camera (Leica DFC 420C, 5 Mpixels). Subsequently, the embryo size was measured by determining the embryo area employing the imageJ software [111].

Mass spectrometric analysis of IAA and IAM
Endogenous IAM and IAA contents have been analyzed by liquid chromatography-mass spectrometry following a previously published protocol [24]. In brief, organic compounds have been extracted from 100 mg of imbibed seeds into 1 ml ice-cold potassium phosphate buffer (50 mM, pH 7.0) containing 1% diethyldithiocarbamic acid sodium salt and 50 pmol of [ 2 H5, 15 N]-IAM and [ 2 H5]-IAA as internal standards. The organic phase was transferred into fresh tubes and acidified (pH 2.7) by adding 1 M hydrochloric acid. After the samples were pre-purified using 1 ml HLB columns, the eluates were taken to dryness in vacuo and re-dissolved in 45 µl methanol with 0.1% formic acid (v/v). Subsequently, 10 µl of the samples were subjected LC-MS analysis employing an Ultimate3000 RSLC system (Dionex) and a microTOF-Q II mass spectrometer (Bruker Daltonics). In order to determine the analyte contents, the following ion transitions were monitored: IAM, m/z = 175.2→130.1; [ 2 H5, 15

Transcriptomics analysis of the ami1/rty double mutant
Total RNA from 100 mg imbibed Arabidopsis wt and ami1/rty mutant seeds was extracted using the RNasy Plant Mini Kit (Qiagen) according to the manufacturer's instructions. The quality and quantity of the extracted RNA was tested by absorbance analysis using a Nanodrop® ND-1000 spectrophotometer (ThermoFisher). Additionally, the RNA samples were tested on a Bioanalyzer 2100 (Agilent) by the CNB Genomics Service (Madrid). Library construction and RNA sequencing (SE50) was performed by the Beijing Genomics Institute (BGI) on Illumina HiSeq™ 2000 machines. Basic data analysis, including data filtering, sequence alignment [112], transcript quantification [113], and differential gene expression analysis [114] was also performed by BGI. For each genotype three biological replicates were processed. The resulting p-values were adjusted for multiple testing using the Benjamini-Hochberg correction [115]. An adjusted p-value (FDR) of < 0.05 and absolute differential expression of log2FC ≥ 1 were arbitrarily chosen to select differentially expressed genes.
The functional classification of DEGs was performed using the MapMan v3.6 software [116], paying special attention to DEGs related with transcriptional regulation and development. Furthermore, functional relationships between the DEGs were investigated employing the stringApp v1.3 [117], MCODE v2.0 [118], and EnrichmentMaps v3.3.1 [119] in Cytoscape v3.8.2 [120]. In order to analyse the importance of the nodes in the inferred networks, the nodes with the highest degree of connectivity (k) and betweenness centrality (BC) have been examined in closer detail.
Selected transcripts were validated in independent experiments by qRT-PCR. For this, 7-day-old wt seedlings were incubated for 2 h either in½ MS or ½ MS with 20 µM IAM. For each 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 were used as template for the qRT-PCR, 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 [121] with APT1 (At1g27450) as reference gene [122]. See Supplementary Table S2 for primer sequences.

Statistical analysis
We used JASP v0.14.1 for statistical data assessment and the generation of plots. The box plots display the median, quartiles, and extremes of the compared embryo sizes. One-way ANOVA and Tukey's post-hoc test or Student's t-test were performed to statistically analysis the data. Sample sizes (n) for each experiment are given in the respective figure legends.

Acknowledgments:
The authors appreciate the thoughtful feedback and highly valuable comments by all members of the CBGP laboratories 127 and 132, particularly the fruitful discussions with the Drs. Jesús Vicente-Carbajosa and Raquel Iglesias-Fernández. In addition, the authors thank Dr. Yunde Zhao (University of California San Diego) for kindly sharing the iaaMox gain-of-function line with us, and Víctor Carrasco-Loba and Leticia Martín-Torres for excellent technical support.

Conflicts of Interest:
The authors declare no conflict of interest.