The cytochrome P450 CYP77A4 is involved in auxin-mediated patterning of the Arabidopsis thaliana embryo

Development and metabolism are intertwined with one another during organogenesis. This interaction is essential to maintain the metabolic state in a developmental context-dependent manner. There is also a growing awareness that metabolism plays instructive roles in developmental processes (Bulusu et al., 2017; Carey et al., 2015; Miyazawa et al., 2017; Oginuma et al., 2017; Shang et al., 2016; Toyokura et al., 2011; Tsukagoshi et al., 2010; Sciacovelli et al., 2016). The emerging picture depicts metabolism as a crucial system not only for sustaining physiological conditions, but also for regulating developmental patterning by coordinating various cellular processes (Teleman, 2016).

Forward genetic screenings of Arabidopsis thaliana (hereafter, Arabidopsis) mutants with developmental defects have led to the discovery of genes encoding metabolic enzymes, which have crucial roles in controlling plant development. Cytochrome P450 proteins are haem-dependent enzymes that catalyse a wide variety of mono-oxygenation and hydroxylation reactions (Mizutani and Ohta, 2010; Schuler and Werck-Reichhart, 2003). The cytochrome P450 genes constitute one of the largest families in the plant genome; for example, they comprise ∼250 genes in Arabidopsis. It is widely accepted that the biosynthesis and/or catabolism of phytohormones by cytochrome P450 proteins play pivotal roles in plant development (Mizutani and Ohta, 2010; Nelson and Werck-Reichhart, 2011). In addition, KLUH/CYP78A5 (also known as AT1G13710) has been shown to promote cell proliferation in petals by producing an as-yet-unknown signalling molecule, which is thought to be independent from any classical phytohormone (Anastasiou et al., 2007). The homologous gene most closely related to KLUH/CYP78A5 is CYP78A1 (also known as GRMZM2G167986), which is expressed preferentially in developing inflorescences of Zea mays (Larkin, 1994). Because CYP78A1 has been shown to have a catalytic activity for ω-hydroxylation of lauric acid in vitro (Imaishi et al., 2000), the enzymatic function of KLUH/CYP78A5 is speculated to be fatty acid modification (Anastasiou et al., 2007). Other members of the CYP78A family also contribute to regulating developmental processes, e.g. in seed growth [CYP78A6 (also known as AT2G46660), Fang et al., 2012], fruit morphogenesis [CYP78A9 (also known as AT3G61880), Ito and Meyerowitz, 2000] and determination of the time intervals between leaf initiations (plastochron) [PLASTOCHRON1/CYP78A11 (also known as Os10g0403000), Miyoshi et al., 2004].

In addition to the members of the CYP78 family, CYP77A6 (also known as AT3G10570) also links metabolism to a specific developmental process. This enzyme catalyses in-chain fatty acid hydroxylation to produce cutin polyester. Although the cyp77a6 mutant lacks the cuticle matrix on the petal surface and the nano-ridge structure beneath the matrix, some electron-dense material, which is a constituent of the nano-ridge structure in the wild type (WT), is detectable (Li-Beisson et al., 2009). This suggests that CYP77A6-mediated metabolism is required for the production of cuticle matrix on the petal surface; thereafter, it helps to form the correct nano-ridge structure. These previous studies indicate that various cytochrome P450 proteins have specific roles in plant development. However, the developmental function of many of these proteins remains unclear (Mizutani and Ohta, 2010). It is thus important to identify the functions of cytochrome P450-encoding genes in developmental regulation through metabolite modification. This poses a crucial challenge in modern functional genomic studies of cytochrome P450 genes in plant development and metabolism.

Advances in metabolome, transcriptome and proteome analyses, in combination with bioinformatics techniques, allow us to speculate on the metabolic functions of genes (Hirai et al., 2004; Ideker et al., 2001; Keurentjes, 2009; Saito et al., 2008). These approaches facilitate the identification and functional characterisation of genes of interest in metabolism using reverse genetics (Ginglinger et al., 2013; Hirai et al., 2007; Malitsky et al., 2008; Morohashi et al., 2012; Yonekura-Sakakibara et al., 2008). However, developmental traits result from concerted actions across different levels of biological hierarchy, from molecule to tissue and organ, and cannot be fully characterised by molecular-level ‘omics’ data alone. Because of this limitation, reverse genetics studies focused on the interaction between development and metabolism are limited. Systematic characterisation of developmental phenotypes (the so-called ‘phenome’) (Kuromori et al., 2006; Toyoda and Wada, 2004; Wilson-Sánchez et al., 2014) in metabolism-related mutants would help to connect metabolic regulation with developmental patterning.

In this study, we analysed cytochrome P450 mutants in Arabidopsis and obtained quantitative phenome data to identify a novel mutant that sheds light on the interaction between development and metabolism. We found that the cytochrome P450 mutant cyp77a4 (at5g04660) exhibits a developmental defect in embryonic patterning. By using auxin-related reporters, we showed that CYP77A4 is involved in shaping the auxin response pattern in embryos by regulating the distribution of an auxin efflux carrier, PIN-FORMED1 (PIN1) (Benková et al., 2003; Friml et al., 2003; Xu et al., 2006). As CYP77A4 has fatty-acid epoxidation activity in the microsomal fraction (Sauveplane et al., 2009), our study highlights the importance of P450-mediated fatty-acid modification in embryogenesis. More generally, our study indicates that quantitative phenome analysis of metabolism-related mutants is a novel method for functional identification of metabolic genes in plant development.

We first collected 35 Arabidopsis T-DNA insertion mutants of cytochrome P450 genes (Table S1) and quantified 12 non-biased developmental traits to identify potential cytochrome P450 genes involved in developmental regulation (Fig. 1A). We validated this approach based on the detection of previously reported smaller rosettes in cyp79f1 (at1g16410) (Reintanz et al., 2001) and shorter petioles in cyp85a2 (at3g30180) (Kim et al., 2005) as positive controls (Fig. 1A). We found that the cotyledon in a cyp77a4 mutant (SALK_046395) was smaller than in the WT, whereas other developmental traits were largely comparable with the WT (Fig. 1A). This obvious reduction in size was unique among the 35 mutants. A public microarray database (eFP-browser, bar.utoronto.ca/efp/cgi-bin/efpWeb.cgi) indicated that CYP77A4 is expressed primarily in early-stage embryos (Winter et al., 2007; Fig. S1), suggesting that it is involved in embryogenesis. The enzymatic function of CYP77A4 is proposed to be as an epoxidase of unsaturated fatty acids, including linoleic acid, to produce vernolic acid, which is a precursor of antifungal compounds (Hou and Forman, 2000; Pinot and Beisson, 2011; Sauveplane et al., 2009). However, the developmental role of CYP77A4 is still unclear. We therefore selected CYP77A4 for further functional investigation.

We confirmed that the SALK_046395 allele contains a T-DNA insertion in the exon of the AT5G04660 locus (Fig. 1B). T-DNA insertions in the 82 bp upstream region and in the exon were also verified in two independent available alleles (SALK_076143 and SALK_112368, respectively) (Fig. 1B). These three alleles are hereafter referred to as cyp77a4-1 (SALK_046395), cyp77a4-2 (SALK_076143) and cyp77a4-3 (SALK_112368). The CYP77A4 transcript was undetectable in the cyp77a4-1 and cyp77a4-3 alleles, and present at a low level in the cyp77a4-2 allele (Fig. 1C). All three alleles exhibited the smaller cotyledon phenotype compared with the WT with 30%, 6% or 23% penetrance in cyp77a4-1, cyp77a4-2 or cyp77a4-3, respectively (Fig. 1D). We detected a significant decrease in cell number, but not in cell area, in the null alleles cyp77a4-1 and cyp77a4-3 (Fig. S2). This indicated that a cell proliferation defect is a primary cause for the decrease in cotyledon size in the mutants.

In addition to the abnormality in size, we found that cotyledon arrangement was perturbed in the cyp77a4 mutants. Whereas bilateral arrangement of cotyledons was present in the WT and was predominant among cyp77a4 mutants (Fig. 2A,B; Table 1), irregular arrangement of cotyledons was also observed among mutants (Fig. 2C; Table 1). More-severe defects in the cotyledon phenotype of the cyp77a4 mutants included single or cup-shaped cotyledons (Fig. 2D-F; Table 1). A single central mid-vein with some lateral veins was formed in the morphologically normal cotyledons of the WT and the cyp77a4 mutants (Fig. 2G). We observed two sets of this vascular pattern within the single or cup-shaped mutant cotyledons (Fig. 2G-I). This suggested that bilateral symmetry is, at least partially, maintained within an individual mutant despite the severe cotyledon phenotype. These phenotypes could be rescued by introducing the genomic fragment containing AT5G04660, from 1.58 kb upstream of the initiation codon to its stop codon, into the cyp77a4-3 mutant (cyp77a4-3/pCYP77A4::CYP77A4) (Table 2), showing that CYP77A4 is responsible for the developmental defect in cotyledons.

Our complementation test using the cyp77a4-3/pCYP77A4::CYP77A4 transgenic lines showed that the upstream 1.58 kb fragment was sufficient to induce the expression of CYP77A4 for proper cotyledon patterning. We established β-glucuronidase (GUS)-reporter lines using the fragment as a CYP77A4 promoter (pCYP77A4::GUS) to investigate the expression profile of CYP77A4. GUS staining was observed in the embryos, but was undetectable in other tissues in immature seeds from 5-week-old plants (Fig. 3A-C). We confirmed the accumulation of CYP77A4 transcripts in embryos by whole-mount in situ hybridisation (Fig. S3). CYP77A4 promoter activity seemed to be absent after germination, except in trichomes (Fig. 3D-H). Quantitative real-time reverse-transcription PCR (qRT-PCR) yielded results consistent with those of the expression analyses (Fig. 3I), indicating that CYP77A4 is expressed primarily in embryos.

Given the expression pattern of CYP77A4, we observed the cyp77a4 embryos to characterise developmental defects during embryogenesis. We found that the mutants exhibited irregular cell division, resulting in uncoordinated tissue patterning from early embryogenesis (Fig. 4A,E; Figs S4-S5). Similar defects were observed in subsequent developmental stages in the mutants, as evidenced by the lack of characteristic globular and heart shapes for embryos (Fig. 4B,C,F,G; Fig. S5). Whereas two cotyledon primordia were separately grown at the torpedo stage in the WT, the boundary between cotyledon primordia was obscure at the comparable stage in the mutant (Fig. 4D,H; Fig. S5). The penetrance of the mis-shapen embryo at the torpedo stage is relatively milder than that at the heart stage (Fig. 4G,H; Fig. S5), suggesting that morphological defects might be, at least partially, fixed toward the torpedo and/or later developmental stages.

These results strongly suggested that the irregular arrangement of cotyledons, and single or cup-shaped cotyledon morphology, are formed by a failure of embryonic patterning, owing to the absence of the separation boundary between cotyledon primordia, rather than by the fusion of two emerging primordia. In contrast to mutant embryos, we detected no morphological defect in mutant trichomes (Fig. S6).

Cotyledon development in embryonic patterning requires the auxin response maximum at cotyledon primordia (Aida et al., 2002; Benková et al., 2003; Friml et al., 2003; Furutani et al., 2004; Treml et al., 2005). Morphological defects in the cyp77a4 embryos suggested an abnormal pattern of auxin response. To examine this possibility, we crossed a transgenic line expressing a gene that encodes GREEN FLUORESCENT PROTEIN (GFP) driven by the auxin-responsive promoter DR5_rev (DR5_rev::GFP; Friml et al., 2003) with the cyp77a4-3 mutant (cyp77a4-3/DR5_rev::GFP) and performed confocal microscopy. Local auxin response, as represented by GFP fluorescence, was observed in the hypophysis at first (100%, n=288), and gradually became evident in the tips of emerging cotyledon primordia against the WT background (100%, n=312) (Fig. 5A-I). In contrast, the GFP fluorescence was often absent in the entirety of the early embryo (11.5%, n=364) (Fig. 5J-L), or it was expanded with very weak intensity or undetectable in the apical domain of the embryos without a clear separation boundary between cotyledon primordia against the cyp77a4-3 background (13.3%, n=339) (Fig. 5M-R). These penetrances in the cyp77a4-3 mutants were significantly higher than in the WT (P<0.01, two-tailed Fisher's exact test). The pattern of DR5_rev::GFP expression was not perturbed in the cyp77a4-3 roots (Fig. S7), which is consistent with the finding that CYP77A4 expression was hardly detectable in this tissue, except for during the embryonic stage (Fig. 3A-C,I; Fig. S3).

During embryogenesis, the establishment of an auxin response maximum is facilitated by the characteristic distribution of the PIN1 efflux carrier, which transports auxin downward to the root pole through vascular tissue and upward to the tips of emerging cotyledon primordia through protodermal tissue (Benková et al., 2003; Friml et al., 2003). We next investigated whether the inappropriate distribution of an auxin response maximum in the cyp77a4 embryos was due to a defect in the PIN1 expression pattern. PIN1 localisation was visualised using transgenic lines that express a gene that encodes PIN1 fused with ENHANCED GFP (EGFP) driven by the PIN1 promoter against the WT background (pPIN1::PIN1-EGFP, Xu et al., 2006) and the cyp77a4-3 background (cyp77a4-3/pPIN1::PIN1-EGFP). The overall distribution of PIN1-EGFP was largely similar between the WT and the cyp77a4-3 mutant embryos with relatively normal shape (Fig. 6A-C,E-G,I-K). However, at a cellular level we found that PIN1-EGFP fluorescence was often absent in the tips of cotyledon primordia in the cyp77a4-3 embryos, in contrast to the WT (Fig. 6D,H). Although protodermal cell division is predominantly anticlinal to the surface of embryos in the WT for correct auxin distribution, we observed the periclinal or oblique division plane in the cyp77a4-3 embryos, in which PIN1-EGFP was accumulated (Fig. 6L; Fig. S4). In the cyp77a4-3 embryo without protrusions of cotyledon primordia, PIN1-EGFP fluorescence was very weak and its localisation was often missing in the tip of cotyledon primordia (Fig. 6P). These results suggested that CYP77A4 is involved in the correct distribution of PIN1, and thus proper shaping of an auxin response maximum for a coordinated cell-division pattern during embryogenesis.

The epoxide derivative of linoleic acid (vernolic acid) is produced by CYP77A4 and is subsequently metabolised by epoxide hydrolases in Arabidopsis microsomes (Sauveplane et al., 2009). To examine whether subcellular localisation of CYP77A4 is spatially consistent with this metabolic pathway, we transiently expressed CYP77A4-GFP or GFP-CYP77A4 in mesophyll protoplast cells using the cauliflower mosaic virus 35S promoter. We prepared protoplasts from WT as a negative control of background fluorescence (Fig. 7A,B) or from a transgenic line that expresses monomeric RED FLUORESCENCE PROTEIN (mRFP) fused with the functional soluble N-ethylmaleimide sensitive factor attachment protein receptor (SNARE) VAM3/SYP22 (AT5G46860) (mRFP-VAM3) to visualise vacuole-related membranes (Ebine et al., 2008) (Fig. 7C,D). In contrast to the GFP (Fig. 7E,F), CYP77A4-GFP fluorescence was observed in the cell periphery, transvacuolar cytoplasmic strands and perinuclear region, which overlapped with mRFP-VAM3 fluorescence (Fig. 7G-R). We detected no GFP-CYP77A4 fluorescence, likely because N-terminal GFP fusion may perturb CYP77A4 localisation by preventing its N-terminal hydrophobic domain from anchoring to the membrane (Fig. 7S,T). We then established cyp77a4-3 transgenic lines expressing CYP77A4-GFP using the CYP77A4 promoter (cyp77a4-3/pCYP77A4::CYP77A4-GFP) to confirm the intracellular localisation of CYP77A4. The mutant phenotype in cotyledon patterning was fully rescued by the expression of CYP77A4-GFP (Table 2), indicating its functionality. Confocal microscopy showed that CYP77A4-GFP fluorescence was most evident in the protodermal cell layer of the embryos (Fig. 7U). Intracellularly, we detected CYP77A4-GFP fluorescence around the nucleus, cell periphery and cytoplasmic strands, similar to our transient assay with protoplasts (Fig. 7V,W), in contrast to the negative control of GFP fluorescence (Fig. 7X). Based on these observations, we concluded that CYP77A4 localises to the endoplasmic reticulum membrane.

CYP77A6 is the gene exhibiting the closest homology to CYP77A4, and it regulates fatty acid modification in floral organs (Li-Beisson et al., 2009). To examine whether CYP77A6 is also involved in cotyledon development during embryogenesis, we observed the cotyledon phenotype in the cyp77a6 mutants. We determined that these mutants produced normal cotyledons (Table S2). The frequency of the cotyledon phenotype in the cyp77a4-3 cyp77a6-2 double mutant is similar to that in the cyp77a4-3 (Table S3). Overall, the growth of the double mutant was comparable with that of the WT in mature plants (Fig. S8). These data indicate that the developmental function of CYP77A4 is independent of CYP77A6.

The cyp77a6 petal is readily stained with Toluidine Blue, owing to the lack of cuticle matrix on the petal surface (Li-Beisson et al., 2009). We tested whether dye permeability was affected by the cyp77a4 mutation, to further investigate the functional relationship between CYP77A4 and CYP77A6. In contrast to the petals on the cyp77a6-2 mutant, petals on the cyp77a4-3 mutant were not permeable to dye, similar to the WT (Fig. S9A-C). Petals on the cyp77a4-3 cyp77a6-2 double mutant showed similar dye permeability to cyp77a6-2 petals (Fig. S9C,D). Although CYP77A4 is expressed in trichomes (Fig. 3D,F), the dye did not permeate into the trichomes of the cyp77a4-3 mutant, similar to results seen in the WT (Fig. S9E,F,I,J). This impermeability was also observed in the trichomes of cyp77a6-2 and the cyp77a4-3 cyp77a6-2 double mutant (Fig. S9E-L). Interestingly, dye permeability slightly increased in the germinating embryos of cyp77a4-3 and cyp77a4-3 cyp77a6-2 double mutant, compared with WT and cyp77a6-2 (Fig. S9M-P). These data suggest that CYP77A4 is involved in the production of cuticle matrix, but it is limited within embryos and can be spatially separable from the CYP77A6 function in petals.

We performed quantitative phenome screening of cytochrome P450 mutants to identify a novel metabolic gene involved in plant development. This analysis revealed that CYP77A4, which is able to catalyse the fatty-acid epoxidation (Sauveplane et al., 2009), participates in cotyledon development during embryogenesis. We showed that CYP77A4 facilitates proper PIN1 distribution within embryos, thereby promoting the establishment of an auxin response maximum for polarity formation and organised cell-division patterning. Although auxin action in embryos is well established (Aida et al., 2002; Benková et al., 2003; Friml et al., 2003; Furutani et al., 2004; Treml et al., 2005), our study provides new molecular insight into this process. Interestingly, we found that CYP77A4 has specific developmental functions in this embryonic process, indicating that the CYP77A4-dependent metabolite modification is involved in polarity establishment in a developmental context-dependent manner. In addition, our study shows that quantitative phenome screening could determine as-yet-unknown developmental functions of the cytochrome P450 genes. Given that transcriptome and metabolome techniques are widely applied to elucidate the molecular function of metabolism-related genes (Ehlting et al., 2008; Ginglinger et al., 2013; Hirai et al., 2004, 2007), we propose that quantitative phenome-based analysis further links molecular function with developmental control.

Several lines of evidence indicate that lipid modification is a major requirement for correct auxin distribution, facilitated by polar positioning of auxin efflux carriers, including PIN1. STEROL METHYLTRANSFERASE1 (SMT1, also known as AT5G13710) is involved in sterol biosynthesis through methylation (Diener et al., 2000; Schrick et al., 2002). Various defects are observed in the embryo, vasculature and root patterning of the smt1 mutants, owing to aberrant cell polarity caused by PIN1 mislocalisation. Because the level of a major membrane sterol, sitosterol, is decreased in the smt1 mutant, it has been suggested that a sterol-dependent microdomain (or lipid raft) contributes to the polarised PIN1 localisation (Willemsen et al., 2003). An alternative explanation is that PIN1 transportation is associated with sterol-dependent polar trafficking (Grebe et al., 2003). A recent study focusing on other types of sterol methyltransferases [SMT2 (AT1G20330) and SMT3 (AT1G76090)] showed that sterol modification by these enzymes is involved in the endocytic recycling required for PIN2 polar localisation in roots (Nakamoto et al., 2015). Similar pleiotropic defects in development were also observed in mutants of the fatty-acid biosynthesis pasticcino genes (Baud et al., 2004; Bellec et al., 2002; Haberer et al., 2002; Kajiwara et al., 2004; Roudier et al., 2010). Our observed phenotype in the cyp77a4 mutants may be attributable to missing or perturbed PIN1 polar localisation, and the resultant aberrant distribution of auxin, in the embryos. Therefore, based on these results, in combination with the proposed enzymatic function of CYP77A4 in fatty-acid epoxidation (Sauveplane et al., 2009), we assume that the molecular mechanism by which CYP77A4 contributes to embryonic patterning is related to membrane docking and/or trafficking of PIN1 for its polar localisation by adjusting fatty-acid composition. Besides auxin transport, local auxin production is also important for embryonic patterning (Robert et al., 2013; Stepanova et al., 2008). It would be informative to test whether auxin biosynthesis is affected or not in the cyp77a4 mutant embryos.

It appears that the CYP77A4-mediated formation of an auxin response maximum is absent in other tissues, such as roots, as our phenome analysis detected little difference in developmental traits other than cotyledon size in the cyp77a4 mutant. This is consistent with the expression pattern of CYP77A4 in only embryos and trichomes, and normal DR5_rev::GFP expression in roots. One attractive hypothesis is that the optimal lipid composition for polarity establishment differs between tissues. However, we cannot exclude the possibility that other enzymes can compensate for the deficit of CYP77A4 in other tissues.

CYP77A6 is phylogenetically close to CYP77A4 among CYP77 family members (Li-Beisson et al., 2009). Whereas CYP77A4 is expressed in the embryos to establish auxin distribution, CYP77A6 action for cutin polyester biosynthesis is limited to the petal surface (Li-Beisson et al., 2009). The cyp77a4 cyp77a6 double mutant does not exhibit ubiquitous growth defects, as are observed in biosynthesis mutants of sterols or very-long-chain fatty acids (Kajiwara et al., 2004; Nakamoto et al., 2015; Roudier et al., 2010; Willemsen et al., 2003). These data suggest that the auxin-related developmental function of CYP77A4 is independent of CYP77A6. Identifying and characterising other epoxidases in Arabidopsis will facilitate further investigation of this issue.

CYP77A4 is the first cytochrome P450 that has been reported as being able to catalyse the epoxidation of fatty acids in Arabidopsis (Sauveplane et al., 2009). A previous study demonstrated that vernolic acid is converted to the corresponding diol in the Arabidopsis microsomal fraction. Because the resulting diol has antifungal properties (Hou and Forman, 2000), the function of CYP77A4 is speculated to be as a defence response (Pinot and Beisson, 2011; Sauveplane et al., 2009). Our GUS-reporter assay using the CYP77A4 promoter showed that the promoter is active in trichomes, as well as in embryos. A publicly available database of genes expressed in trichomes (TrichOME, www.planttrichome.org/) (Dai et al., 2010) also indicated that the CYP77A4 transcript is detectable in Arabidopsis trichomes. Based on this expression pattern, the synthesis of a defensive compound by CYP77A4 through the vernolic acid pathway appears to be viable in trichomes. In addition to this function, we propose that the CYP77A4-dependent metabolic pathway is related to embryonic patterning. Given that CYP77A4 widely catalyses the epoxidation of unsaturated fatty acids, including linoleic acid (Sauveplane et al., 2009), this enzyme can affect metabolic pathways during embryogenesis other than those that are linoleic acid related. We indeed detected that CYP77A4 knockout or overexpression has little effect on linoleic acid content in immature seeds compared with the WT, as well as on compounds chiefly related to central carbon metabolism and amino acids (Y.L., Y.S., M.O. and M.Y.H., unpublished). We found that CYP77A4 also participates in cuticle matrix production in embryos, based on the Toluidine Blue permeability test. This might be the key to identifying relevant substrates of CYP77A4 that are required for auxin-mediated embryonic patterning.

Whereas epoxidation of fatty acids in animals is well studied because of the biological activities of derivatives (Catella et al., 1990; Roman, 2002; Zeldin et al., 1996), less attention has been paid to P450-mediated epoxidation in plants. We shed light on this topic by uncovering the function of CYP77A4 in plant development. This finding could extend understanding of the P450-mediated metabolic regulation of development by linking epoxidation to the establishment of polarity. Future work to analyse the comprehensive lipid composition within embryos, focused in particular on epoxidation, with solid biochemical evidence for CYP77A4 enzymatic activity, and any links to the developmental function will uncover how this modification affects the metabolic regulation of plant development.

The Arabidopsis WT accession used in this study was Colombia-0. Cytochrome P450 T-DNA insertion mutants were included in an allele set of confirmed T-DNA insertion lines (CS27941) provided by the Arabidopsis Biological Resource Center. Homozygous mutation was confirmed by genotyping PCR using primers that are listed in Tables S1 and S4. Before analysis, all cyp77a4 alleles were backcrossed with the WT three times. The cyp77a4-1 or cyp77a4-3 mutation was introduced into the DR5_rev::GFP (Friml et al., 2003) and pPIN1::PIN1-EGFP (Xu et al., 2006) transgenic lines by crossing. The cyp77a6-1 and cyp77a6-2 mutations were introduced into cyp77a4-1 and cyp77a4-3, respectively, by crossing. Each genotype was confirmed by genotyping PCR using primers that are listed in Table S4. As essentially the same results were obtained irrespective of the cyp77a4 allelic background, we show representative results against the cyp77a4-3 background. Plants were grown on rockwool, as has been previously described (Kawade et al., 2010).

Dry seeds were observed using a stereoscopic microscope (M165FC; Leica Microsystems) and analysed in terms of size (Herridge et al., 2011) or colour (binary intensity) (n>25). Image processing was performed using ImageJ software (imagej.nih.gov/ij/). Primary root length was measured in 7-day-old seedlings that were grown on vertically placed plates (n=12). In vitro culture was conducted according to a previous study (Kawade et al., 2013). The first leaves from 3-week-old plants were fixed, cleared and observed (Kawade et al., 2013; Tsuge et al., 1996) for measurement of blade area and petiole length (n=8). Cotyledon area in 3-week-old plants (n=16), and flowering stalk width (n=9), pollen size (n>14) and fruit length (n=27) in 5-week-old plants were measured using a stereoscopic microscope. Rosette diameter was calculated as the average of the longest three leaves in 3-week-old plants (n=8). Flowering time was evaluated as has been described elsewhere (n=9) (Notaguchi et al., 2008). Petals were collected from fully opened flowers in 4-week-old plants and observed under the stereoscopic microscope (n=9).

Because CYP77A4 does not contain introns, genomic DNA that had been extracted from WT seedlings was used as the template for vector construction. The primers that were used are listed in Table S4. To create the GUS-reporter construct, a 1580 bp region extending from the first intron of the upstream gene (AT5G04650) to the 5′-UTR of CYP77A4 was selected as the promoter construct. The promoter was amplified using primer pair 17 and 19 (Table S4) and introduced into pGWB3 using the TOPO and Gateway system (Invitrogen). To create the overexpression construct, a 1539 bp region spanning the coding region of CYP77A4 was amplified using primer pair 18 and 20 and introduced into pGWB2. To create complementation and GFP constructs driven by the CYP77A4 promoter, a 3119 bp region spanning the promoter, 5′-UTR and coding region of CYP77A4 was amplified using primer pair 17 and 20, and introduced into pGWB1 and pGWB4. For GFP constructs driven by the cauliflower mosaic virus 35S promoter, the coding region of CYP77A4 was amplified using primer pairs 18 and 20, or 18 and 21, and introduced into pGWB6 and pGWB5 for the N- and C-terminal GFP fusion constructs, respectively. The pGWB binary vectors are described elsewhere (Nakagawa et al., 2007). These vectors were transformed into the GV3101 Agrobacterium strain and introduced into Arabidopsis plants. For the GUS-reporter, overexpression and 35S promoter GFP constructs, the WT was used as a transgenic background. For complementation and own promoter GFP constructs, cyp77a4-3 was used as the transgenic background. Transgenic lines were screened on agar-solidified 1/2 Murashige and Skoog medium with 25 mg/l hygromycin. T3 homozygous lines were subjected to further experiments.

Immature seeds were harvested from fruits 80 h after flowering. Shoot and root samples were collected from 3-week-old plants. Total RNA was extracted using the RNeasy Plant Mini Kit (Qiagen) and treated with the DNase TURBO DNA-free kit (Thermo Fisher Scientific) according to the manufacturer's instructions. The first-strand cDNA was synthesised using SuperScript II (Invitrogen) with identical RNA concentration for each sample.

The expression levels of CYP77A4 in immature seeds of the WT and three cyp77a4 mutant alleles were analysed using semi-quantitative reverse transcription PCR (RT-PCR). ACTIN2 (AT3G18780) was used as an internal control (Han and Kim, 2006). The expression levels of CYP77A4 in different tissues was analysed by quantitative real-time reverse-transcription PCR (qRT-PCR). PCR amplification using Fast SYBR Green (Thermo Fisher Scientific) was monitored on the StepOnePlus real-time PCR system (Applied Biosystems). UBIQUITIN CONJUGATING ENZYME 9 (UBC9, also known as AT4G27960) was used as an internal control (Hirai et al., 2007). The primers that were used are listed in Table S4. Data were collected from three independent biological replicates with triplicates in each sample, and were subjected to statistical analysis, as described in the figure legends.

Cotyledons from 7-day-old seedlings were fixed in formalin/acetic acid/alcohol solution (FAA) and cleared in chloral hydrate solution (200 g chloral hydrate, 20 g glycerol, 50 ml water) (Tsuge et al., 1996) to observe the vascular pattern with the stereoscopic microscope. Isolated immature seeds from 5-week-old plants were cleared using chloral hydrate solution for observation of embryos with a Nomarski differential interference contrast microscope (DM2500; Leica Microsystems).

Fluorescence microscopy was performed using a confocal laser-scanning microscope (A1Rsi; Nikon). For staining with propidium iodide, root tissue was soaked in 1 mg/ml propidium iodide solution for 1 min, washed several times with water and then observed. Images were processed using NIS-Elements (Nikon), in addition to ImageJ software.

Tissue samples were infiltrated with a reaction buffer [100 mM NaPO₄, pH 7.0, 1.5 mM 5-bromo-4-chloro-3-indolyl-β-glucuronic acid, 4.5 mM K₃Fe(CN)₆, 4.5 mM K₄Fe(CN)₆, 10 mM EDTA, 0.1% Triton X-100], incubated overnight at 22°C or 37°C and fixed with FAA. Immature seed samples from 5-week-old plants were dehydrated in an ethanol series and embedded in Technovit 7100 (Heraeus). The resin block was sectioned at 5 µm with a rotary microtome (Leica Biosystems) and placed on glass slides for microscopy. We observed three independent pCYP77A4::GUS lines and obtained consistent results.

The first leaves from 2-week-old plants and flowers from 5-week-old plants were soaked in Toluidine Blue solution with 0.01% Silwet L-77 for 30 min, then washed several times with water before observation under a stereoscopic microscope.

Protoplasts isolated from the first leaves of 3-week-old plants were used for DNA transfection assays to characterise subcellular localisation, as has been previously described (Wu et al., 2009; Yoo et al., 2007).

We are grateful to Jiri Friml and Ben Scheres for allowing us to use DR5_rev::GFP and pPIN1::PIN1-EGFP, respectively; to Masahiko Furutani for providing DR5_rev::GFP and pPIN1::PIN1-EGFP seeds; and to Tomohiro Uemura and Takashi Ueda for the kind gift of vam3-1/pVAM3::mRFP-VAM3 seeds. We thank the Arabidopsis Biological Resource Center for providing mutant seeds; the Spectrography and Bioimaging Facility, NIBB Core Research Facilities, and the Model Plant Research Facility, NIBB BioResource Center, for technical support; and Akane Sakata (RIKEN CSRS) and Chinami Yamaguchi (ExCELLs) for technical support.