Functional identification of genes responsible for the biosynthesis of 1-methoxy-indol-3-ylmethyl-glucosinolate in Brassica rapa ssp. chinensis

Background Brassica vegetables contain a class of secondary metabolites, the glucosinolates (GS), whose specific degradation products determine the characteristic flavor and smell. While some of the respective degradation products of particular GS are recognized as health promoting substances for humans, recent studies also show evidence that namely the 1-methoxy-indol-3-ylmethyl GS might be deleterious by forming characteristic DNA adducts. Therefore, a deeper knowledge of aspects involved in the biosynthesis of indole GS is crucial to design vegetables with an improved secondary metabolite profile. Results Initially the leafy Brassica vegetable pak choi (Brassica rapa ssp. chinensis) was established as suitable tool to elicit very high concentrations of 1-methoxy-indol-3-ylmethyl GS by application of methyl jasmonate. Differentially expressed candidate genes were discovered in a comparative microarray analysis using the 2 × 104 K format Brassica Array and compared to available gene expression data from the Arabidopsis AtGenExpress effort. Arabidopsis knock out mutants of the respective candidate gene homologs were subjected to a comprehensive examination of their GS profiles and confirmed the exclusive involvement of polypeptide 4 of the cytochrome P450 monooxygenase subfamily CYP81F in 1-methoxy-indol-3-ylmethyl GS biosynthesis. Functional characterization of the two identified isoforms coding for CYP81F4 in the Brassica rapa genome was performed using expression analysis and heterologous complementation of the respective Arabidopsis mutant. Conclusions Specific differences discovered in a comparative microarray and glucosinolate profiling analysis enables the functional attribution of Brassica rapa ssp. chinensis genes coding for polypeptide 4 of the cytochrome P450 monooxygenase subfamily CYP81F to their metabolic role in indole glucosinolate biosynthesis. These new identified Brassica genes will enable the development of genetic tools for breeding vegetables with improved GS composition in the near future.

members of the CYP81F family of Arabidopsis thaliana as being involved in 4-hydroxylation of indol-3-ylmethyl GS and/or 1-methoxy-indol-3-ylmethyl GS biosynthesis [12]. When tissue is damaged, the thioglucoside linkage of GS is hydrolyzed by myrosinases, enzymes that are spatially separated from GS in intact tissue. In the presence or absence of specifier proteins the degradation results in the formation of a variety of hydrolysis products [13].
The different groups of GS and their various degradation products are extensively studied metabolites. It has been shown that genes encoding enzymes of the  Figure 1 Biosynthesis pathway of indole glucosinolates as known in Arabidopsis thaliana. Enzymes catalyzing each reaction are given with the respective gene name. Identified putative Brassica rapa homologues [14] are indicated with underscores.
specific glucosinolate biosynthesis pathways form stable co-expression clusters [15], and group together with tryptophan biosynthetic genes in response to stress conditions [16]. With respect to plant fitness they play important roles in plant defence against herbivores [17] and pathogens [9], and also abiotic stresses like UV-B irradiation specifically changes the GS profile [18]. In addition, there is increasing evidence that evolutionary and ecological forces shape polymorphism at loci involved in the GS-myrosinase defence system [19].
Brassica vegetables are cultivated and consumed worldwide and represent a highly important component in the human diet [20]. Their content of GS is varying dependent on genotype, development and environmental conditions [21] while the composition of GS and their respective degradation products is a major determinant of the characteristic flavor and smell of Brassica vegetables [22]. In addition, the secondary metabolites and their respective degradation products are believed to have protective cancer-preventing activity in higher animals and humans [23,24]. However, recent studies also provide evidence that juices of Brassicaceae might also be mutagenic because they form characteristic DNA adducts in bacteria and mammalian cells [25]. It is namely the 1-methoxy-indol-3ylmethyl GS and its degradation products that have been shown to exert these negative effects [26,27].
With this study new genes where identified that are involved in the biosynthesis of indole GS, namely the synthesis of 1-methoxy-indol-3-ylmethyl GS with focus on Brassica vegetables. After establishing the leafy Brassica vegetable pak choi (Brassica rapa ssp. chinensis) as suitable tool to elicit very high concentrations of 1-methoxy-indol-3-ylmethyl GS by application of methyl jasmonate (MeJA) [28] the identification of genes involved in this process was possible by comparing expression pattern in pak choi using the 2 × 104 K format Brassica Array with publicly available gene expression data from the Arabidopsis AtGenExpress effort [29]. With the functional characterization of the identified genes new genetic tools for breeding healthy vegetables with improved GS composition will be possible in the near future.

Results and discussion
Increased indole GS biosynthesis in pak choi treated with methyl jasmonate In a previous study it was shown that different cultivars of the leafy vegetable pak choi (Brassica rapa ssp. chinensis) contain a certain amount of indole GS in their green leaf tissue [30]. The different cultivars can be classified in distinct groups depending on their GS profiles, which are partly linked to the expression of specific genes involved in the aliphatic GS biosynthetic pathway. In a related study it was further demonstrated that a small set of elicitors known to induce GS biosynthesis in various organism is also functional in pak choi [28]. Amongst others it was namely methyl jasmonate (MeJA) that led to an increase of indole GS biosynthesis. In order to further characterize this induction of GS biosynthesis in pak choi seedlings in more detail a concentration series ranging from 100 μM to 3 mM was applied and GS accumulation was measured 48 hours after application (Additional file 1: Table S1). As shown in Figure 2A a doubling of specific aliphatic GS could be achieved when applying concentrations of more than 750 μM MeJA, and also the amount of the aromatic 2-phenylethyl GS was increased up to 3fold at such high concentrations applied. As expected, indole GS accumulation was more sensitive to the MeJA application, and the indole GS level was elevated even when the lowest concentration of 100 μM was used ( Figure 2B). With the application of higher concentrations of MeJA up to 2 mM a further increase of indole GS levels could be achieved until no additional elevation was detected. Notably it was mainly the 1-methoxy-indol-3-ylmethyl GS that was increased up to 30fold in pak choi seedlings after treatment with MeJA.
It is known for a long time that jasmonate, ethylene and salicylic acid upregulate the expression of scores of defense-related genes [31], and our knowledge of the complex network of jasmonate signaling in stress responses and development including hormone cross-talk is continuously increasing [32,33]. With respect to plant resistance GS present classical examples of compounds affecting insect-plant interactions [17] in which the GSmyrosinase defence system is also evolutionary and ecological modulated [19]. In terms of plants defense against pathogens it is further suggested that tryptophan-derived metabolites may act as active antifungal compounds [9,34]. Against this background the induced GS biosynthesis was strongly expected in pak choi after treatment with MeJA.

Specific induction of 1-methoxy-indol-3-ylmethyl GS in pak choi seedlings
In order to analyze the specificity of the increased indole GS biosynthesis in more detail a similar experiment with Arabidopsis seedlings was performed using MeJA concentrations ranging from 200 μM up to 5 mM. As evident from Figure 3 MeJA application also increased indole GS content in Arabidopsis (Additional file 1: Table S2). However, the increase was much lower in this plant species, and the major elevation was found in the non-methoxylated indol-3-ylmethyl GS. Further experiments demonstrated that pak choi seedlings exert stronger rise of indole GS levels upon MeJA application than adult plants [28], while in Arabidopsis no differences in the elevation between seedlings and adult rosette leaves were detectable (data not shown). This comparison with Arabidopsis thaliana Col-0 ecotype clearly revealed that Values represent the mean of three independent samples. Significant differences to the respective control treatment (P < 0.05) as determined using unpaired two-tailed t-test, are marked with an asterisk. For absolute concentrations of glucosinolates please see supporting Additional file 1: Table S1.  Figure 2; I3M, indol-3-ylmethyl GS; 4OHI3M, 4-hydroxy-indol-3-ylmethyl GS; 4MOI3M, 4-methoxy-indol-3-ylmethyl GS; 1MOI3M, 1-methoxy-indol-3-ylmethyl GS. 4OHI4M was undetectable in A.th. seedlings. Values represent the mean of three independent samples. Significant differences to the respective control treatment (P < 0.05) as determined using unpaired two-tailed t-test, are marked with an asterisk. For absolute concentrations of glucosinolates please see supporting Additional file 1: Table S2.
a very strong raise of 1-methoxy-indol-3-ylmethyl GS is specific to pak choi. The unambiguous difference between seedlings of pak choi and Arabidopsis discovered in this glucosinolate profiling analyses was used in further experiments to identify related genes involved in 1methoxy-indol-3-ylmethyl GS biosynthesis of Brassica rapa ssp. chinensis.
Identification of candidate genes using gene expression analysis with the Brassica microarray As strong induction of 1-methoxy-indol-3-ylmethyl GS was found 48 hours after application of 2 mM MeJA to pak choi seedlings gene expression differences to control treatments were analyzed in these samples using the Brassica microarray. In order to get maximum amount of information the 2 × 104 K array was chosen in the investigation. The elements on the Brassica array were identified by their homology to known genes of Arabidopsis thaliana and were classified to respective bins using MapMan [35] and Mercator [36]. As expected when MeJA was applied to plant seedlings, defense related genes showed the most significantly changed transcript levels ( Table 1). With respect to a putative function in GS metabolism [37] the genes with highest expression differences are listed in Table 2. Mainly the transcripts of genes putatively involved in GS degradation were induced, but also genes involved in indole GS core structure formation were strongly elevated and among the most significantly changed. The increased expression of genes specifically involved in indole GS core structure biosynthesis reflects the elevation of indole GS levels.
Among the most significantly altered transcripts candidates were selected that are putatively involved in side chain modification of indole GS biosynthesis, namely those that show typical structures of the large gene families of cytochrome P450 monooxygenases or O-methyltransferases (Table 2). These selected candidates were further evaluated regarding respective expression differences of the related homologs in available Arabidopsis thaliana microarray hybridization experiments using the Genevestigator database [38]. As shown in Table 3 the Arabidopsis homologs of the selected genes involved in GS metabolism were found responsive to MeJA treatments with log2-ratios being 1 or greater. This is in good agreement with the reported modulation of the GS profile in Arabidopsis by defense signaling pathways [39] and is also reflected in results presented in Figure 3. The Arabidopsis homologs of the selected candidate genes show strong variation in their responsiveness to MeJA. While At3g28740 (CYP81D11) and At5g36220 (CYP81D1) were strongly induced by MeJA application, At4g37410 (CYP81F4), At4g37430 (CYP81F1) and At5g42590 (CYP71A16) were only weakly influenced, while At4g35160 (OMT) and At1g13080 (CYP71B2) showed unchanged expression. As At1g13080, At5g42590 and At3g28740 were already expected to be involved in other metabolic pathways we concentrate in further experiments on At4g37410 and At4g37430 as genes putatively involved in GS metabolism, and on At4g35160 and At5g36220 without any linked pathway identified so far.
GS profiling in Arabidopsis mutants with knock out of the respective candidate gene homologs In order to verify a putative involvement of the selected candidate genes in indole GS biosynthesis respective Arabidopsis knock out mutants were profiled for their GS accumulation. Since there are tissue specific differences in the proportional distribution of individual GS with indole GS being mainly present in either roots or old leaves [40] plants were grown in tissue culture and leaves and roots analyzed separately, or GS profiles of leaves of flowering plants grown in the greenhouse were measured (Additional file 1: Table S3). As evident from Table 4 there is one of the four selected Arabidopsis knock out mutants that did not produce 1-methoxy-indol-3ylmethyl GS in any of the tissues analyzed. This confirms the expectation that the Arabidopsis gene product of At4g37410 (CYP81F4) is needed in leaves and roots to synthesize 1-methoxy-indol-3-ylmethyl GS [12,41]. The absence of a metabolic phenotype on GS level in the selected Arabidopsis mutant with knock out in the selected O-methyltransferase (Atomt) further shows that at least in Arabidopsis there are other O-methyltransferases present which could contribute to the synthesis of 1-methoxyindol-3-ylmethyl GS in leaves. Consequently, it needs to be analyzed whether the O-methyltransferase activity is provided through IGMT5 (At1g76790) an O-methyltransferase family protein that is strongly co-expressed with At4g37410 (CYP81F4) as determined using the ATTED-II coexpression database [42]. In addition, in Arabidopsis there are further members of the O-methyltransferase family, IGMT1 (At1g21100), IGMT2 (At1g21120) and IGMT4 (At1g21130), that are coexpressed with At5g57220 (AtCYP81F2). At least in an artificial expression system using Nicotiana benthamiana it has been shown that IGMT1 and IGMT2 can be employed for O-methylation of indole GS [12].
As shown previously there is a certain increase of indole GS biosynthesis in Arabidopsis after application of MeJA ( Figure 3). Therefore, the selected knocks out mutants of genes responsive to MeJA treatment (Table 3) were also analyzed after application of this elicitor. While mutants in AtCYP81F1 and AtCYP81D1 showed a comparable increase of indole GS biosynthesis as the treated control plants (Table 5), the mutant in AtCYP81F4 did not accumulate any 1-methoxy-indol-3-ylmethyl GS while an expected increase of the precursor indol-3-ylmethyl GS   could be observed 48 hours after MeJA application in this mutant. This finally confirms that the gene product of At4g37410, the cytochrome P450 monooxygenase 81F4 is utterly necessary to synthesize 1-methoxy-indole-3ylmethyl GS in Arabidopsis at standard growth conditions. It additionally demonstrates that there is none of the other P450 monooxygenase 81F family proteins involved in 1-methoxy-indole-3-ylmethyl GS synthesis even under conditions of increased biosynthesis when defense related pathways are induced. The Genevestigator database [37] was used to evaluate expression differences of the selected genes. Arabidopsis genes homologous to the identified MeJA responsive Brassica genes are in the same order as in Table 2. Grading of changes of the log2-ratios is as follows: 0, unchanged with log2-ratio smaller 0.5; +, log2-ratio between 0.5 and 1; ++, log2-ratio between 1 and 2.5; +++, log2-ratio larger than 2.5.  Arabidopsis ecotype Wu-0 without 1-methoxy-indol-3ylmethyl GS accumulation Further evidence of the importance of At4g37410 (CYP81F4) for 1-methoxy-indol-3-ylmethyl GS biosynthesis is coming from a survey of the GS content in leaves and roots of the 19 key accessions [43] used to develop the MAGIC lines [44]. A total of 20 distinct GS could be identified and quantified by Witzel and co-workers, with most of the aliphatic GS showing accession-specific distribution while the indole GS were present in almost all 19 accessions [43] with one exception: ecotype Wu-0 did not contain 1-methoxy-indol-3-ylmethyl GS in any tissue analyzed. Since the corresponding whole genome sequences of all 19 accessions are available [45] the respective sequence variants at locus At4g37410 (http://mus.well. ox.ac.uk/19genomes/variants.tables/) were inspected for the presence of relevant polymorphisms. Indeed, at bp coordinate 18595917 in the pseudo genome and bp coordinate 17592444 of the Col-0 reference genome on chromosome 4 the insertion of one C nucleotide could be found solely in the accession Wu-0. This insertion produces a frame shift in the coding sequence thus disrupting CYP81F4 and leading to an altered protein sequence from amino acid 390 with a premature stop at amino acid 395. In contrast, the putative functional protein is composed of 501 amino acids in all other accessions that produce 1-methoxy-indole-3-ylmethyl GS. In summary this is an excellent example were publicly available sequence data together with comprehensive metabolite profiling enables the identification of a gene that is putatively involved in the respective metabolic pathway at question. In addition, since the ecotype Wu-0 is an Arabidopsis accession collected from Germany the presence of 1-methoxy-indol-3-ylmethyl GS does not seem to be essential for survival of this ecotype in its natural habitat. As shown previously defense related co-expression networks in Arabidopsis thaliana group together with tryptophan and GS biosynthesis genes in response to stress conditions [16]. Thus, the increase of indole GS biosynthesis in Arabidopsis and the relatively small accumulation of 1-methoxy-indol-3-ylmethyl GS when compared to Brassica rapa ssp. chinensis revealed that this specific indole GS might not play a pivotal role in stress response in Arabidopsis thaliana.

Characterization of the CYP81F4 genes identified in the Brassica rapa genome
It was already shown that genes involved in the GS biosynthesis exist in more than one copy in the Brassica rapa genome accession Chiifu-401-42 [37]. Besides this there is also a high co-linearity when compared to Arabidopsis thaliana. This co-linearity is similarly found for AtCYP81F4 (At4g37410) surrounded by AtCYP81F3 (At4g37400) and AtCYP81F1 (At4g37430) on Arabidopsis chromosome 4. When compared to Arabidopsis At4g37410 two different orthologues of the Brassica rapa accession Chiifu-401-42 on BAC clones KBrB006J12 and KBrH064I20 could be identified: While KBrB006J12 corresponds to a region on chromosome A01, no match for KBrH064I20 has been found so far. On KBrB006J12 the orthologue to AtCYP81F4 was identified as Bra011759 (BrCYP81F4-1) on the reverse strand on chromosome A01, and is preceded by Bra011758 orthologous to AtCYP81F3 and followed by Bra011761 orthologous to AtCYPF1. On KBrH064I20 the orthologue to AtCYP81F4 was named BrCYP81F4-2, and is preceded by another orthologue to AtCYP81F3 while the following sequence orthologous to AtCYPF1 is corrupted.
In order to analyze the tissue specific expression of the selected genes in more detail isoform specific primer pairs were developed using the respective sequences of the Brassica rapa accession Chiifu-401-42 BAC clones KBrB006J12 and KBrH064I20. Semi-quantitative realtime RT-PCR analysis was performed with cDNA synthesized from RNA isolated from 12 days old seedlings, and leaves and roots of six weeks old Brassica rapa ssp. chinensis plants. As evident from Table 6 expression of all selected genes could be detected in pak choi. In most cases a higher expression was found in leaves than in seedlings and only BrCYP81F4-1 is expressed at a higher level in roots than in leaves. The highest expression level in leaves was detected for BrCYP81F4-2 while BrCYP81F4-1 was the main expressed isoform in roots. This already indicates that the BrCYP81F4 isoforms may play an important role on a tissue-specific level and during development at standard growth conditions. Further expression analysis was performed with different tissues of pak choi treated with 500 μM MeJA. Expression analysis confirmed induction of mainly the two Total aliphatic GS, 3-methylsulfinyl-propyl GS; 4-methylsulfinyl-butyl GS; 4-methylthio-butyl GS; 5-methylsulfinyl-pentyl GS; 8-methylsulfinyl-octyl GS. Total indol GS, I3M, indol-3-ylmethyl GS; 4MOI3M, 4-methoxy-indol-3-ylmethyl GS; 1MOI3M, 1-methoxy-indol-3-ylmethyl GS. Values represent the mean ± standard deviation of three individual plants homozygous for the respective T-DNA insertion. Significant differences to the respective control treatment (P < 0.05) as determined using unpaired two-tailed t-test, are marked with an asterisk. Values are given in% on dry matter basis of the respective treatment of control plants. -, below detection limit in mutant.
Since there was some increased expression also detectable for other isoforms seedlings of pak choi were treated with a series of different concentrations of MeJA and expression differences to control treatment were analyzed for all BrCYP81F (Figure 4). This unequivocally confirms that both BrCYP81F4 isoforms were most responsive to the elicitor treatment while the others did not show comparable sensitivity to this elicitor. Application of 100 μM MeJA already elevated the expression of BrCYP81F4-1 and BrCYP81F4-2 4fold with highest increase of BrCYP 81F4-2 of more than 64fold after application of 2 mM MeJA. This confirms that the two isoforms of BrCYP81F4 are the candidate genes from Brassica rapa ssp. chinensis that are crucial for 1-methoxy-indol-3-ylmethyl GS biosynthesis. Jasmonic acid signaling is a central component of inducible plant defense and the expression of jasmonateinduced responses are tightly regulated by the ecological background of the plant [46] and also by the plant species itself. While in Arabidopsis thaliana tryptophan and GS biosynthesis genes respond to stress conditions [16] there is only relatively small accumulation of 1-methoxyindol-3-ylmethyl GS when compared to Brassica rapa ssp. chinensis. The role of this distinct response to the elicitor and differences in accumulation of a specific defense compound will be the subject of future analysis in an ecological context.

Functional identification of BrCYP81F4 isoforms for biosynthesis of 1-methoxy-indol-3-ylmethyl GS
In order to finally assess BrCYP81F4 isoform function full length cDNAs of both genes were amplified and heterologously expressed in the Arabidopsis thaliana mutant Atcyp81f4, which does not produce 1-methoxy-indol-3ylmethyl GS. Using oligonucleotide primers developed with the Brassica A genome sequence from Brassica rapa accession Chiifu-401-42 [37] two full length cDNA sequences from Brassica rapa ssp. chinensis coding for putative BrCYP81F4 isoforms were amplified. Both sequences show 90.7% pair-wise identities and code for proteins of 501 amino acids with 93% similarity. Compared to the Arabidopsis protein similarities of 85.4% and 90.2% could be calculated. The sequences of interest (BrCYP81F4-1  and BrCYP81F4-2) were recombined into the plant expression vector pK7WG2 [47] and Agrobacterium mediated gene transfer was performed using the knock out mutant Atcyp81f4 as the host. Kanamycin resistant seedlings of the T2 generation were selected and analyzed for heterologous gene expression and GS accumulation. As shown in Table 7 expression of both cDNAs from pak choi in the Atcyp81f4 mutant background led to metabolic complementation with accumulation of 1-methoxy-indol-3-ylmethyl GS in leaves and a reduced level of I3M when compared to the mutant without expression of the Brassica rapa ssp. chinensis genes. Although the identical heterologous expression system was used, BrCYP81F4-2 led to much higher accumulation of 1-methoxy-indol-3ylmethyl GS. Whether this difference is caused by a higher protein level of the heterologous enzyme in the mutant plant background or is linked to advanced enzyme activity will be the topic of further studies. Another interesting point here is the significant decrease of I3M in the Atcyp81f4 mutant background when the highly active BrCYP81F4-2 is expressed. In summary the level of indole GS stayed constant in these plants demonstrating unaltered total flux into the indole GS pathway thus indicating no further metabolic regulation by the end products.

Conclusions
In conclusion this is an explicit example were elicitation of a specific metabolic difference and subsequent comparative microarray analysis together with focused metabolite profiling permits the targeted discovery of genes involved in the respective metabolic pathway.
Here this enables the functional attribution of new identified Brassica rapa ssp. chinensis genes to their metabolic role in indole glucosinolate biosynthesis that in the near future will contribute to develop new genetic tools for breeding vegetables with improved glucosinolate profile.

Elicitor treatment
Methyl jasmonate (Sigma Aldrich, Seelze, Germany) was resolved in water containing 0.01% (v/v) Tween20 to reduce surface tension and water containing 0.01% (v/v) Tween20 was sprayed as control treatment. The 10 days old pak choi seedlings were treated by spraying each bar of fleece with 15 ml of the respective solution. The 10 days old Arabidopsis seedlings were treated by spraying each petri dish with 2 ml of the respective solution. 48 hours after treatment the total aerial tissue was harvested. Samples were quickly frozen in liquid nitrogen, subsequently lyophilized, and blended to a fine powder. For each treatment, at least three samples were taken as replicates.
Sample preparation and desulfo-glucosinolate analysis by HPLC Glucosinolate concentration was determined as desulfoglucosinolates according to Wiesner et al. [30]. Briefly, 20 mg of powdered samples were extracted and analyzed by HPLC using a Merck HPLC system (Merck-Hitachi, Darmstadt, Germany) with a Spherisorb ODS2 column (Bischoff, Leonberg Germany; particle size 5 μm, 250 mm × 4 mm). Desulfo-glucosinolates were identified based on comparison of retention times and UV absorption spectra with those of known standards. Glucosinolate concentration was calculated by the peak area relative to the area of the internal standard. Each replicate sample was measured in duplicate. Results are given as μmol g -1 dry weight.

Microarray analysis
The microarray analysis was performed as described [18]. Briefly, frozen pak choi sprout material was ground in liquid nitrogen in an orbital ball mill for 2 min at a frequency of 30 Hz s −1 (MM400 Retsch GmbH, Haan, Germany). Total RNA was extracted using the RNeasy Plant Mini Kit (Qiagen GmbH, Hilden, Germany), including the on-column DNase digestion step with the RNase-free DNase Set (Qiagen). The microarray analysis was done with 1 mg of total RNA isolated from each of three replicates of methyl jasmonate treated and control treated seedlings. Agilent One-Color Gene Expression Microarray analysis following the recommendation of MIAME  [35]. The assignment of the different genes was done by comparison of the translated protein sequences of the 95 k Brassica unigene set with the Arabidopsis TAIR9 database using the Mercator pipeline for automated sequence annotation [36] (http://mapman. gabipd.org/web/guest/app/mercator). For each identifier the gene with the highest homology was provided with identifier and description. The respective bitscores were classified as follows: very weakly similar (bitscore smaller than 100); weakly similar (bitscore 101-200); moderately similar (bitscore 201-500); highly similar (bitscore greater than 500).

Isolation of mutants
Plants were obtained from the Salk collection [50]. Screening and selection within mutant populations was done following the Signal Salk instructions (http://signal. salk.edu). Genomic DNA was isolated by a standard procedure using NucleoSpin PlantII (Macherey-Nagel GmbH & Co. KG, Dueren, Germany). PCR genotyping was performed using the T-DNA LB-specific primer SALK LBb 5′-GCGTGGACCGCTTGCTGCAACT and the gene-specific primer pairs of Atcyp81f4l2 5′-AGG