Cytochrome P450 CYP78A9 is involved in Arabidopsis reproductive development.

Synchronized communication between gametophytic and sporophytic tissue is crucial for successful reproduction, and hormones seem to have a prominent role in it. Here, we studied the role of the Arabidopsis (Arabidopsis thaliana) cytochrome P450 CYP78A9 enzyme during reproductive development. First, controlled pollination experiments indicate that CYP78A9 responds to fertilization. Second, while CYP78A9 overexpression can uncouple fruit development from fertilization, the cyp78a8 cyp78a9 loss-of-function mutant has reduced seed set due to outer ovule integument development arrest, leading to female sterility. Moreover, CYP78A9 has a specific expression pattern in inner integuments in early steps of ovule development as well as in the funiculus, embryo, and integuments of developing seeds. CYP78A9 overexpression did not change the response to the known hormones involved in flower development and fruit set, and it did not seem to have much effect on the major known hormonal pathways. Furthermore, according to previous predictions, perturbations in the flavonol biosynthesis pathway were detected in cyp78a9, cyp78a8 cyp78a9, and empty siliques (es1-D) mutants. However, it appeared that they do not cause the observed phenotypes. In summary, these results add new insights into the role of CYP78A9 in plant reproduction and present, to our knowledge, the first characterization of metabolite differences between mutants in this gene family.


Introduction
Angiosperms have evolved the processes of double fertilization and fruit development as pivotal steps of their survival and dispersal strategies. Pollination and fertilization are essential for fruit initiation, considering that the angiosperm flower initiates terminal senescence and abscission programs if pollination has not taken place (Vivian-Smith et al., 2001;Fuentes and Vivian-Smith, 2009). For centuries, humans endeavored to prevent this association in order to develop seedless fruits. Most research has concentrated on the role of endogenous phytohormones as triggers for fruit initiation after fertilization and different strategies as exogenous application or artificial overproduction of plant hormones (Fuentes and Vivian-Smith, 2009), mutation or misexpression of specific genes has been tested. The principal lines of evidence suggest that increased auxin and gibberellin content in ovules and ovary leads to parthenocarpic fruits in Arabidopsis (Varoquaux et al., 2000;Goetz et al., 2007;Alabadi et al., 2009;Dorcey et al., 2009;Fuentes and Vivian-Smith, 2009;Pandolfini et al., 2009;Carbonell-Bejerano et al., 2010). But even with this increasing insight, the available knowledge of the mechanisms that underlie parthenocarpy is still limited.
The Arabidopsis transposon activation tagged mutant empty siliques (es1-D) overexpresses the Cytochrome P450 CYP78A9 gene and is characterized by the development of parthenocarpic fruits (Marsch-Martinez et al., 2002;de Folter et al., 2004). When crossed to wild type plants, it results in the formation of viable seed.
However, its fertility is reduced compared to wild type plants. Moreover, siliques that result from the pollination of mutant es1-D plants with wild type pollen grow larger than unpollinated mutant siliques and wider than wild type siliques (Marsch-Martinez et al., 2002). A T-DNA activation tagging screen also identified a mutant overexpressing CYP78A9 with a similar phenotype (Ito and Meyerowitz, 2000). P450 Cytochromes represent a family of heme-containing enzymes belonging to the monooxygenase group, which are found in all kingdoms and show extraordinary diversity in their chemical reactions (Schuler et al., 2006;Mizutani and Ohta, 2010). In plants, they are involved in the metabolism of most phytohormones including auxins, gibberellins (GA), cytokinins (CK), brassinosteroids (BR), absicic acid (ABA), and have an overlapping function in the control of seed size, as noted by the synergistic enhancement of the cyp78a6 seed size phenotype produced in cyp78a9 background (Fang et al., 2012).
Although the catalytic function of the Arabidopsis CYP78A enzymes remains unknown, the expression pattern and in vivo effect of the related genes bring up their possible engagement in the biosynthesis of some unknown type of plant growth regulator (Zondlo and Irish, 1999;Ito and Meyerowitz, 2000;Anastasiou et al., 2007;Kai et al., 2009). Hormones seem to have a prominent role in synchronizing fertilization and fruit growth. (2009)  In this work, we provide a detailed description of the CYP78A9 function during reproductive development. The genetic relation of CYP78A9 with its closest paralogs, CYP78A6 and CYP78A8, shows a redundant function controlling floral organ growth and integument development, which in turn affect fertility. All the defects observed in the cyp78a8 cyp78a9 double homozygous mutant are related to the sporophyte before fertilization, which supports the idea that this family of genes acts maternally during reproductive development. Moreover, the expression pattern of CYP78A9 studied in detail added a new insight into the possible function of this gene during reproductive development. CYP78A9 expression was observed in integuments during ovule development and highlights the possible communication role of the CYP78A9-produced signal between sporophytic and gametophytic tissue. During seed development CYP78A9 expression was also present in embryo and seed coat. In addition, the pattern of pCYP78A9::GUS visible in the funiculus 12 hours after hand-pollination, together with the fact that CYP78A9 overexpression uncoupled fruit development from fertilization, supports the possible role of a CYP78A9-produced signal in synchronizing fertilization and fruit development. Furthermore, metabolic profiling of mutants compared to wild type showed perturbation in flavonoid content. However, the differences in kaempferol and quercetin content did not explain the observed phenotypes of the mutants. In summary, the results add new insights to the role of CYP78A9 in plant reproduction and present the first metabolic characterization of mutants in this family of genes.

Characterization of the Arabidopsis activation tagging line es1-D
In the activation tagged es1-D mutant ( Figure S1; Marsch-Martinez et al., 2002;de Folter et al., 2004), androecia and gynoecia developed in an uncoordinated manner; anthers dehisce later than in wild type Arabidopsis plants, resulting in lack of pollination ( Figure 1). Interestingly, fruits developed even without pollination, which never happens in wild type plants (Vivian-Smith et al., 2001). Despite the apparently normal carpel morphology, the majority of the mature fruits were empty ( Figure 1A). These parthenocarpic fruits were wider and shorter than the wild type ( Figure 1G), and when developing flowers were emasculated the produced empty fruits reached a longer and wider size ( Figure 1B). At the end of the es1-D life cycle, pollination occurs but seed yield and fruit length are still reduced ( Figure 1G and H). Reciprocal crosses were made to analyze whether the male and/or the female reproductive side was affected. The reduced fertility could not be rescued with wild type pollen ( Figure 1H), suggesting that es1-D has defects during ovule development. Ovule number was not changed in es1-D, but had larger ovules and displayed outer integuments with more and larger cells than wild type (45 to 50 compared to 29 to 35 cells in wild type Ws-3) (Figure 2A). The ovule perimeter was significantly (P< 0.001 of Student's t test) larger in es1-D, compared to wild type Ws-3 ( Figure 2B). However, es1-D pollen also failed to produce normal fruitset when crossed to wild type, only 17 to 20 seeds per fruit were observed ( Figure 1H), compared to normally 40 to 50 seeds found in wild type fruits. Although the fruits of this cross (♀ Ws-3 x ♂ es1-D) had more seeds, they did not reach the length of a wild type fruit ( Figure 1G). These results indicate that both male and female reproductive development is affected in the es1-D mutant.
Furthermore, we observed that es1-D sepals, petals, pistils, and seeds were larger in comparison to wild type ( Figure 1D,E,I). However, es1-D stamen filaments failed to elongate ( Figure 1E). Moreover, basipetal floral buds were slow to open, and stamen filament elongation more severely affected. The more acropetal flowers were less severely affected, although the sepals and petals of those flowers did not abscise normally. In addition to the flower phenotypes, the es1-D plants showed larger dark green leaves and stout stems with a zig-zag pattern ( Figure 1C), a phenotype that has also been observed when KLUH/CYP78A5 and EOD3/CYP78A6, CYP78A subfamily members, are overexpressed (Zondlo and Irish, 1999;Ito and Meyerowitz, 2000;Marsch-Martinez et al., 2002;Fang et al., 2012).
The es1-D mutant flowered later (21 days) than the wild type plants (16 days), but no difference in the number of leaves produced at the moment of flowering was observed ( Figure S2). Furthermore, the period of flower production and total number of flowers produced in the main inflorescence of mature es1-D and wild type plants was compared. At week 5 after flower initiation, wild type plants stopped producing flowers, while at week 8 the main inflorescence of a still growing es1-D plant was producing flowers ( Figure 3B). Data presented in Figure 3 shows that es1-D clearly produces flowers over a longer period and that the growth of its main inflorescence is extended by several weeks ( Figure 3A) and therefore produced more flowers, but not because of a faster growth. This phenotype was also described by Hensel et al. (1994) for mutants that have reduced fertility, whereby mutations that reduced the number of seeds per silique by more than 50% were associated with an increased proliferative capacity of the main inflorescence stem.
To confirm that the observed phenotypes in es1-D were due to CYP78A9 overexpression, the entire genomic region was cloned and expressed constitutively under the 35S promoter (35S::gCYP78A9). We obtained 21 independent lines and 17 of them displayed the es1-D phenotype ( Figure S3).

0
The phenotypes observed in the activation tagging mutant es1-D as well as in the transgenic lines constitutively overexpressing CYP78A9, suggest that CYP78A9 has a function controlling floral organ size and, moreover, that it is involved in producing a signal that can overcome the pistil senescence program that initiates when in wild type plants fertilization does not occur (Vivian-Smith et al., 2001;Fuentes and Vivian-Smith, 2009).

Characterization of Arabidopsis cyp78a9, cyp78a6, and cyp78a8 single and double mutants
To better understand the function of CYP78A9, T-DNA insertion loss-of-function mutants for CYP78A9 and for its closest paralogs, CYP78A6 and CYP78A8, were analyzed (Figure S1B-E). Homozygous plants were identified by PCR genotyping and RT-PCR experiments confirmed that they were true knock-out mutants ( Figure S4).
Regarding fruit development, no evident loss-of-function phenotype was observed in cyp78a9, cyp78a6, and cyp78a8 single mutants compared to wild type Col-0. All single mutants showed no difference in fruit-length and seed number per fruit compared to each other and to wild type (p< 0.001) ( Figure 4A,B and Figure S5A,B). To test whether CYP78A6 or CYP78A8 act redundantly with CYP78A9 regulating fruit-length and seedset we crossed cyp78a9 to cyp78a6 and to cyp78a8. A reduction in both fruit-length and seed-set was already evident in F1 plants cyp78a8 (+/-) cyp78a9 (+/-) compared to wild type and single mutants (p≤0.001). These results were also confirmed in F2 double heterozygous plants ( Figure 4A,B).
This reduction in seed-set was related with defects in ovule development. From a total of 194 cyp78a8 (+/-) cyp78a9 (+/-) ovules analyzed, 109 (56%) presented different levels of integument development arrest. The most abundant phenotype, 105 of 194 ovules (54%), were those that still conserved at some extent the asymmetry of a wild type ovule, but presented short integuments that failed to accommodate the developing embryo sac resulting in physical restriction of the gametophyte leading to female sterility ( Figure 5B). The affected ovules (56%) were smaller, the reduction of the cyp78a8 (+/-) 1 1 cyp78a9 (+/-) ovule perimeter was significantly different (at P< 0.001 of Student's t test) compared to wild type Col-0 ( Figure 5H). Furthermore, in severe cases (around 1% of ovules), the outer integument interrupted its growth and the ovule developed a somehow orthotropic morphology ( Figure 5C). In other cases (around 1% of ovules) tracheid-like cells appeared at the position of the embryo sac ( Figure 5D).
Interestingly, the cyp78a8 (-/-) cyp78a9 (-/-) double homozygous mutant showed alterations in silique length and seed-set, but to a lesser extent than in the cyp78a8 (+/-) cyp78a9 (+/-) double mutant ( Figure 4A,B). In the double homozygous mutant, ovules were also affected in outer integument development. From 213 analyzed ovules, 74 (35%) were significantly (P<0.001 of Student's t test) reduced in size ( Figure 5H) due to less cells in the outer integument (15 to 20 compared to 29 to 33 cells in wild type Col-0), but still presented a normal embryo sac ( Figure 5E). Seventy seven (36%) developed short integuments that failed to accommodate the developing embryo sac (Figure 5F), and 3 (1.4%) presented tracheid-like cells at the position of the embryo sac ( Figure 5G). The frequency of the phenotype severity correlates well with the observed seed-set.
The analysis of the cross between cyp78a6 and cyp78a9 showed no reduction in silique length although the number of seeds produced per silique was slightly diminished in F1 generation plants ( Figure S5A,B). However, Fang and coworkers (2012) showed that cyp78a6 cyp78a9 had no reduced seed-set but reduced seed-size due to smaller cells in the integuments of the developing seeds and they proposed that cyp78a9 synergistically enhanced the seed-size phenotype of cyp78a6.

CYP78A9 expression during reproductive development
In order to learn more about the CYP78A9 gene, we first made a transcriptional fusion with the GUS reporter under the control of the putative 3 kb CYP78A9 promoter region (CYP78A9::GFP:GUS) ( Figure 6A-G). Transgenic plants were analyzed during reproductive development and GUS staining was observed first at the floral developmental stage 11 according to (Smyth et al., 1990). At stage 11 the GUS staining was localized in anthers and around stage 12 staining was also observed at the gynoecium ( Figure 6A) and the inner integuments of the ovules ( Figure 6B). At stage 13, when flowers open and fertilization occurs, the staining was localized at the placenta and the funiculi ( Figure 6C). During seed development, strong staining was observed in the embryo, from the globular stage till the curly leaf stage of the embryo ( Figure 6C-G, arrows), with a peak at the torpedo stage ( Figure 6F, arrow). Furthermore, signal was also evident at the maternal chalazal region, the endosperm, and the developing testa ( Figure 6D-F, arrowheads).
A more detailed expression analysis by in situ hybridization confirmed the observed GUS patterns and gave additional information on CYP78A9 expression ( Figure 6H-O).
Floral buds at stage 10 showed that CYP78A9 was expressed in the tapetum cells of the anthers ( Figure 6I). Moreover, at this stage, CYP78A9 mRNA was localized in ovules, placenta, and also in the valves and stigma. After fertilization, signal in the developing seeds was stronger compared to mature ovules and was localized to the chalazal and micropilar regions ( Figure 6J-K). During seed development, mRNA localization was observed in the embryo in the same manner as seen in transgenic 1 3 plants with CYP78A9::GFP:GUS, however, the highest in situ hybridization signal was localized at the epidermis of developing embryos ( Figure 6L-O). Furthermore, signal was detected in endosperm, in the testa of the seed, and in valves of developing fruits.
Notably, mRNA expression analysis in es1-D mutant floral buds presented the same expression pattern compared to wild type, though with around twofold increase in signal intensity, confirming that the es1-D phenotype is due to the altered amount of mRNA and not to ectopic expression of CYP78A9 ( Figure S6). Furthermore, we analyzed the protein localization for CYP78A9. For this, a transient expression assay was performed by infecting tobacco leaves with Agrobacterium tumefaciens containing a 35S::gCYP78A9:GFP translational GFP fusion construct. GFP fluorescence signal was detected by confocal microscopy at both sides of the celluar wall of consecutive cells, suggesting that CYP78A9 could be a plasma membraneassociated protein (Figure 7).

CYP78A9 responds to the fertilization event
Because of the fertility related phenotypes observed, we investigated what happened with the expression of CYP78A9 during fertilization. In situ and CYP78A9::GFP:GUS analyses ( Figure 6) demonstrated signal in mature ovules and in the septum before fertilization. After fertilization, signal was observed in developing seed and septum, but also in the funiculus and placental tissue. The funiculus is the connection between the mother plant, via the placenta, and the developing offspring and is important for nutrition. Stadler and coworkers (2005) demonstrated that the release of photoassimilates and macromolecules to the developing seed is mediated by the outer integument that represents a symplastic extension of the funicular phloem (Stadler et al., 2005). afterwards hand-pollinated pistils, GUS signal was observed also in the funiculus and in the placental tissue ( Figure 8C,D). These results indicate that the CYP78A9 promoter responds to the fertilization event.
The CYP78A9-produced signal seems to be different from the known hormones that trigger fruit growth after fertilization To determine whether CYP78A9 was involved in known pathways regulating fruit growth, the response pattern of two of the major hormones involved in this process, auxins and gibberellins, were analyzed in the es1-D mutant.
A cross between es1-D and the auxin-response marker DR5:: GUS (Ulmasov et al., 1997) showed a similar expression pattern to wild type DR5::GUS plants throughout flower development ( Figure S7). The only noted difference was observed in stage 14 flowers. In wild type DR5::GUS plants, where normal fertilization occurs at floral stage 13, GUS signal was seen in fertilized ovules (developing seeds) and in funiculi, while there was no signal observed in this tissue in es1-D DR5::GUS plants due to the absence of pollination ( Figure S7). However, it is possible that the auxin increase required to produce parthenocarpic fruit development occurs earlier in the es1-D mutant.
We analyzed the expression of the GA biosynthesis marker line GA20ox1::GUS (Desgagné-Penix et al., 2005) in the es1-D background. The same expression pattern was observed during flower and fruit development in the wild type background as well as in es1-D background ( Figure S7). This data suggests that GAs cannot be attributed as the cause for the uncoupled growth of es1-D fruit from fertilization.
On the other hand, we evaluated the transcriptional effect of the application of different hormones, as well as hormone inhibitors in seedlings, on the expression of CYP78A9.
For this, publicly available microarray data present in the eFP browser (Winter et al., 2007) was used. We observed that CYP78A9 responded transcriptionally to the application of many hormones ( Figure S8) as well as hormone inhibitors ( Figure S9).

5
CYP78A9 expression changes observed did not allow us to identify if CYP78A9 might be involved in one of these hormone pathways, however, a hypothesis about CYP78A9 expression being regulated by inputs from multiple hormone signals could be made.

Microarray expression analysis in the es1-D mutant
In order to increase our knowledge about the transcriptional responses to CYP78A9 activity, a microarray experiment comparing es1-D and wild type closed floral buds from young primary inflorescences was performed. CYP78A9 responsive genes were defined as genes that were significantly differentially expressed in es1-D (z-score > 2SD) compared to wild type, which resulted in 409 upregulated and 825 downregulated genes (Table S1). Interestingly, CYP78A8 was downregulated in response to the overexpression of CYP78A9 as well as chalcone isomerase (CHI), responsable of converting chalcones into flavonones (Buer et al., 2010), and CYP71B16 reported as co-expressed with the phenylpropanoid pathway (Ehlting et al., 2006). A relationship between the phenylpropanoid pathway and parthenocarpy was suggested also by the gene expression profiles of highly and weakly parthenocarpic pear cultivars. The highly 1 6 parthenocarpic group showed changes in genes located in the junctions of the metabolic routes leading to the production of flavonoids or monolignols and sinapolymalate (Nishitani et al., 2012).
Several genes related to ovule and stamen development were changed. Concerning to ovule integument development, we found HUELLENLOS (HLL) upregulated (Skinner et al., 2001) and AINTEGUMENTA (ANT) downregulated (Elliott et al., 1996). Furthermore, a comparison was made between the data obtained in es1-D and the described sets of hormone-regulated genes (Nemhauser et al., 2006), but no clear overlap was found (Table S2).
Previous information on transcriptional responses to CYP78A9 was obtained by de Folter et al. (2004) in fruits, using a macro-array containing probes to analyze expression of over 1100 transcription factors. By means of comparison between the micro-(this study) and the macro-array data (de Folter et al., 2004), 28 downregulated and 16 upregulated genes were down-and upregulated in both datasets, respectively (Table S3). Among these, we found a series of transcription factors that could be related to the altered phenotypes seen in the es1-D mutant. For instance, ANT and BEL1-like (SAWTOOTH 1 (BLH2), SAWTOOTH 2 (BLH4), and BEL1-LIKE HOMEODOMAIN 5 (BLH5)) transcription factors related to integument development (Elliott et al., 1996). In summary, the microarray experiment showed an altered transcriptional pattern of genes associated to male and female reproductive development. Related to hormones, some genes were found to be downregulated, though, no clear trends could be observed. A set of sixteen P450 enzymes changed its level of expression in the es1-D mutant. Some of them were not annotated as being part of a specific biosynthetic pathway and the ones that were annotated participate each one in a different pathway (CYP79F1: glucosinolate biosynthesis; CYP710A2 brassinosteroid biosynthesis; CYP86A7: fatty acid metabolic process). Interestingly, CYP78A8, one of CYP78A9 closest paralogs, was downregulated in response to CYP78A9 overexpression.
Furthermore, enzymes involved in phenylpropanoid pathway as chalcone isomerase (TT5) and other P450 enzymes as CYP71B16 and CYP716A1 reported as coexpressed with this pathway responded to the overexpression of CYP78A9.

CYP78A9 function prediction using public database analyses
Another strategy undertaken for the identification of the role of CYP78A9 was the analysis of genes co-expressed with CYP78A9 based on floral tissue public microarray data. 88 genes (Table S4)  All these data, together with bioinformatics predictions reported in CYPedia (Ehlting et al., 2006), suggest that CYP78A9 could be an enzyme in the core phenylpropanoid pathway ( Figure 9). Although, we cannot discard its function in other predicted pathways like cell wall carbohydrates metabolism, lipid, fatty acid, and isoprenoid biosynthesis.

Kaempferol and quercetin content in leaves and inflorescences of cyp78a9, cyp78a8, cyp78a8 cyp78a9, and es1-D mutants
The response pattern of DR5::GUS and GA20ox::GUS as well as the transcriptional responses to the overexpression of CYP78A9 indicates that the CYP78A9-produced metabolite seems to be different from the known hormones that trigger fruit growth after fertilization. The group of genes found to be co-expressed with CYP78A9 in public flower microarray data showed that this gene could be involved in many processes including secondary metabolism and signaling pathways. Moreover, Aracyc predicted the function of CYP78A9 to the same point in phenylpropanoid pathway as F3'H. All these data together with the changes observed in the testa color of the double mutant that are similar to that observed in the tt7 mutant (mutated in F3'H), motivated us to explore whether CYP78A9 could act in the flavonol biosynthesis pathway.
Metabolic profiling was performed using LC-MS and multiple fragmentation. Acetone crude extracts from leaves and inflorescences were screened by LC-MS (ACQUITY UPLC-LCT Premier TM XE equipment, Waters). The analyses of the spectra obtained both from the mutant and its respective wild type ecotype was performed using MarkerLynx TM (Waters) and an integrated package comparing and discriminating data sets using multivariable statistics such as Principal Component Analysis (PCA).
Selected metabolites were further fragmented using a SYNAPT HDMS system (Waters) and fragmentation spectra of highlighted metabolites were compared to authentic kaempferol and quercetin standards.
In order to find out if metabolic fluxes were guided towards a specific direction in the phenylpropanoid pathway, we analyzed hydrolyzed acetone extracts, which reveals the total pool of kaempferol and quercetin in the samples. cyp78a9 and cyp78a8 (-/-) cyp78a9(-/-) leaves have a 40 and 50% kaempferol content reduction, respectively, compared to wild type Col-0 (p<0.05), in contrast to the 120% increase of kaempferol content that presented es1-D leaves with respect to wild type Ws-3 (p<0.05) ( Figure   10A). Even though cyp78a8 leaves showed a tendency of kaempferol content reduction compared to wild type, this difference was not statistically significant ( Figure 10A). It has been reported that Arabidopsis leaves have a reduced flux through the flavonoid biosynthetic pathway, and accumulate a higher ratio of kaempferol to quercetin derivatives than flowers and seedlings (Sheahan and Cheong, 1998). In line with this, in our study, hardly any quercetin intensities in leaves were detected and also no difference by PCA analysis was found.
Inflorescences presented less difference than leaves with respect to kaempferol content. However, cyp78a9 inflorescences showed a significant reduction in quercetin content with respect to wild type ( Figure 10B).
In summary, although we observed alterations in the kaempferol and quercetin content in cyp78a9, cyp78a8(-/-) cyp78a9(-/-), and es1-D mutants, which supports the predictions of CYP78A9 having a certain function in this pathway, the phenotypes of the mutants in terms of seed-set are not in accordance with the ones observed for the 0 known tt7 and tt4 mutants that do not show these levels of seed-set impairment (Ylstra et al., 1996).

Genetic interaction between CYP78A9 and chalcone synthase (tt4)?
Given the prediction of ARAcyc that positioned CYP78A9 in the flavonol biosynthesis pathway, at the conversion of dihydrokaempferol to dihydroquercetin step, and the perturbations in kaempferol levels observed in both es1-D (CYP78A9 overexpression), cyp78a9, and cyp78a8 (-/-) cyp78a9 (-/-) mutant, we wanted to further clarify the relationship between CYP78A9 and the flavonoid branch of the phenylpropanoid pathway. To test this, we examined the cross between es1-D (CYP78A9 overexpression) and tt4-1 (null mutant that has a lesion in chalcone synthase (CHS) resulting in no flavonoid production and known for its yellowish seed color) (Peer et al., 2001). The phenotype of the tt4-1 es1-D double mutant plants, found in the F2 generation, presented the characteristic yellow testa seed color of the tt4-1 mutant, however, also the reduction in fertility, the larger seed, the zig-zag stem, and the floral organ size phenotypes of the es1-D mutant ( Figure 11). This means that there is no genetic interaction between CYP78A9 and tt4-1, because the double mutant showed a purely additive phenotype. Metabolic analysis of this double mutant showed no kaempferol neither quercetin accumulation (Figure 10), suggesting that the alterations in kaempferol observed in es1-D cannot be responsible for the CYP78A9 overexpression phenotypes. Moreover, previous studies showed that the Arabidopsis tt4-1 mutant, which lacks all downstream compounds of the flavonoid pathway, has no impairment in seed-set (Burbulis et al., 1996;Ylstra et al., 1996).

Preliminary metabolic screen
A preliminary metabolic screen was performed in es1-D mutant using LC-MS and multiple fragmentations (Table 1). Acetone crude extracts from leaves and inflorescences were screened by LC-MS (ACQUITY UPLC-LCT Premier TM XE equipment, Waters). Following the same procedure as described before, after PCA analysis selected metabolites were further fragmented using a SYNAPT HDMS system (Waters) and fragmentation spectra of highlighted metabolites were compared to authentic standards (kaempferol and quercetin) and public databases, and recently published data (von Roepenack-Lahaye et al., 2004;Beekwilder et al., 2008;Kachlicki et al., 2008;Bollinger et al., 2009;McNab et al., 2009;Gouveia and Castilho, 2010). A distinct pattern of metabolite markers was observed by PCA for the mutants when compared to wild type plants (Table 1 (Table 1). Also, the presence of glucohirsutin m/z 492.10, a glucosinolate, was confirmed by its fragmentation pattern m/z 428 and 311 and reported as being more concentrated in es1-D inflorescences (Table 1).
In summary, this work provides the first characterization of metabolite differences between mutants in this gene family, giving interesting indicators for future investigation of the reaction(s) they catalyze.

Parthenocarpic fruit development in CYP78A9 overexpression line, and the hormonal context
Overexpression of the cytochrome P450 CYP78A9 gene allows fruit growth to be uncoupled from fertilization. In the es1-D activation tagging mutant, where the CYP78A9 gene is overexpressed, androecia and gynoecia develop in an uncoordinated manner.
Anthers present a dehiscence delay and the filaments never reach the stigma so pollination does not occur. In wild type plants during the maturation and receptive periods, specific molecular pathways restrict the growth of the pistil and accessory tissues and stop them from developing into fruits (Vivian-Smith et al., 2001;Goetz et al., 2006). Apparently though, CYP78A9 overexpression can overcome this restriction allowing the pistil to grow before the androecium is mature. Dorcey and collaborators (2009)

CYP78A6, CYP78A8, and CYP78A9 are involved in reproductive development
In the more acropetal flowers of es1-D plants, at the end of its life cycle, pollination and fertilization occur, although fertility is still reduced. Cross-pollination experiments showed that both pollen and ovules of es1-D were affected. The most relevant altered feature in es1-D ovules was integument size, with more and larger cells than wild type.
This fact also has been recently reported for eod31-D, the activation tagging mutant for CYP78A6 (the closest paralog of CYP78A9) in which the ovules have more and larger integument cells (Fang et al., 2012), and has also been reported for plants with directed expression of KLUH/CYP78A5 to the outer integuments (Adamski et al., 2009). In contrast, when cyp78a6 cyp78a9 and cyp78a8 cyp78a9 double mutants were analyzed, a reduction in fertility was detected and was more severe in the latter. The reduction of fertility in cyp78a8 cyp78a9 double mutant was correlated with the growth arrest of outer integuments that fail to accommodate the embryo sac.
Arabidopsis mutants with impaired integument initiation and outgrowth such as ant, ino (INNER NO OUTER; a YABBY transcription factor), and bel1 are associated with aborted embryo sac development and reduced fertility (Ray et al., 1994;Elliott et al., 1996;Baker et al., 1997;Villanueva et al., 1999). Other mutants with altered integuments have been proven to present abnormalities during seed development as is the case of the abs stk double mutant, that completely lacks endothelium (Mizzotti et al., 2011) or the ttg2 mutation that affects primarily cell elongation in the integuments by double mutants are linked to the sporophyte before fertilization, which supports the idea that this family of genes act maternally during reproductive development. Although, the defects detected in cyp78a8 cyp78a9 mutant were linked to the sporophyte before fertilization, the expression pattern detected during seed and embryo development suggest that CYP78A9 could have a possible communication role between the embryo and the seed coat while they develop. Furthermore, the fact that CYP78A9::GFP:GUS responds to the fertilization event, indicates the possible communication role of CYP78A9 between the placenta, funiculus, and ovule during the fertilization process.

CYP78A8 and CYP78A9 in metabolic processes
Many studies suggest that members of the CYP78A family produce a novel kind of signal. Anastasiou and colleagues (2007)  However, no rescue of klu/cyp78a5 phenotype was obtained with the application of 12hydroxyl-lauric acid, pointing out that this might not be the only substrate in planta. Su and coworkers (2010) showed that transient expression in petals in Phalaenopsis of CYP78A2 cloned from the same species increased anthocyanin content in that organ.
CYP78A2 boosts the pathway without the biosynthesis of any new anthocyanin and the effect is not limited to Phalaenopsis, but also occurs in rose and carnation (Su and Hsu, 2010). However, they could not prove if CYP78A2 acts directly in this pathway or indirectly via the production of another hormone or secondary metabolite (Su and Hsu, 2010).
In an effort to confirm the prediction that CYP78A9 has an overlapping function with F3'H -in the conversion of dihydrokaempferol to dihydroquercetin-we performed a metabolic exploration of the flavonoid pathway. Here, one would expect that a partial block of the pathway results in an increase of dihydrokaempferol or kaempferol. Still, quercetin can be present, as the double mutant is in a F3'H-intact background.
Interestingly, hydrolyzed extracts showed that cyp78a9 and cyp78a8 (-/-) cyp78a9(-/-) leaves have a 40 and 50% kaempferol content reduction, respectively, compared to wild type Col-0 (p<0.05), in contrast to the 120% increase of kaempferol content that presented es1-D leaves with respect to wild type Ws-3 (p<0.05). However, cyp78a9 inflorescences showed a significant reduction in quercetin content with respect to wild type. Flavonoids are plant secondary metabolites that comprise both pigments such as chalcones and anthocyanins as well as colorless molecules such as flavonones, Alterations in seed size and development were observed in mutants defective in proanthocyanidin synthesis or accumulation, as is the case for ttg2 which is affected in integument cell elongation (Garcia et al., 2005) and for the abs stk double mutant that has severely reduced its fertility and completely lacks endothelium (Mizzotti et al., 2011). A link between parthenocarpy and the flavonoid pathway has been established in tomato by Schijlen and coworkers (2007). Downregulation of the flavonoid pathway using RNA interference (RNAi)-mediated suppression of chalcone synthase (CHS) rendered plants with impaired pollen tube growth and pollination dependent parthenocarpic fruit development (Schijlen et al., 2007). Parthenocarpy can also be achieved by overexpression of the grape stilbene synthase (STS) gene (Ingrosso et al., 2011). This enzyme competes for the same substrates as CHS, and its overexpression produced a decrease in flavonoid content that leads to male sterile pollen in tobacco and a decreased seed-set in tomato fruits. This connection could not be made in Arabidopsis as the tt4 (CHS) mutant did not present male sterility neither parthenocarpic development, though, a slight reduction in seed-set was observed (Burbulis et al., 1996;Ylstra et al., 1996). Recently, Mahajan and colleagues (2011) have proposed a new strategy to generate fruits with reduced seed-set in tobacco. By mean of posttranscriptional gene silencing of FLS (Flavonol synthase) they obtained plants reduced in quercetin and anthocyanidins content, but increased in catechin, epi-catechin, and epi-gallocatechin. Interestingly, FLS silenced lines were significantly reduced in seed number (Mahajan et al., 2011). In summary, flavonoids are important metabolites that have diverse biological functions. Within the past few years, increasing interest in these metabolites has been reported because their possible role during reproductive development and the implied potential of biotechnologically control seed-set by manipulating this pathway (Taylor and Grotewold, 2005;Falcone Ferreyra et al., 2012).
Although, the metabolic data showed differential accumulation of flavonoids between wild type and mutants, some contradictions did not support these alterations to be in line with the CYP78A9 catalytic bioinformatics prediction or CYP78A9 as being the direct cause of the phenotypes observed in the mutants. First, kaempferol levels of cyp78a9 and cyp78a8 cyp78a9 mutants were not in accordance with the data reported for the tt7 mutant, deficient in F3'H activity, which overaccumulates kaempferol and does not produce quercetin (Shirley et al., 1995;Peer et al., 2001). In this sense, cyp78a8 cyp78a9 kaempferol profile resembles more the tt5 mutant that shows a drastically reduced flux through the biosynthetic pathway in relation to wild type (Sheahan and Cheong, 1998), however, has not the same change in testa color, cyp78a8 cyp78a9 has pale brown testa and tt5 has yellowish testa. The profiles reported for tt3 and tt6 mutants did not resemble the situation of our mutant, as tt3 has been reported to accumulate excess of quercetin and kaempferol (Peer et al., 2001), and tt6 has reduced kaempferol content and has no quercetin (Shirley et al., 1995).
Moreover, the tt7 (kaempferol overaccumulator) and tt4 (devoided of kaempferol) mutants did not show the levels of seed-set impairment that showed the cyp78a8 cyp78a9 double mutant (Ylstra et al., 1996). Second, the tt4-1 es1-D double mutant showed a purely additive phenotype and did neither accumulate kaempferol nor quercetin, suggesting that the alterations in flavonoids present in the overexpression mutant are not responsible for the observed phenotypes. Third, variations in other metabolites like glucohirsutin (a glucosinolate), lysophosphatidic acids (LPAs), unknown flavonoid aglycones, and unidentified compounds were found in the analyses. And fourth, Kai and coworkers (2009) showed that CYP78A5/KLUH, CYP78A7, and CYP78A10 catalyze the ω -hydroxylation of short chain fatty acids, and proposed that this family of enzymes modifies a fatty-acid related molecule, which could participate in another biosynthetic pathway.
Although CYP78A9 was predicted to possibly be chloroplast localized (Schuler et al., 2006), our work suggest its localization to the plasma membrane. Based on this information, together with experimental evidence that CYP78A9 forms a protein-protein interaction with CALMODULIN 7 (CAM7; At3g43810) and with the protein kinase superfamily protein (At1g48210), localized also in the plasma membrane (Popescu et al., 2007), and predictions made by ARACNE that this protein could be implicated with calcium dependent lipid binding signaling involved in cell death, makes this an interesting case to explore.
Furthermore, future studies should elucidate whether a specific metabolite causes all the observed morphological phenotypes or different metabolites cause particular phenotypes.

Conclusions
Our findings suggest that CYP78A9 has a function during reproductive development.
The genetic evidence supports the idea that CYP78A9 and its closest paralogs participate in a pathway that control floral organ size and ovule integuments development as denoted by the phenotypes of es1-D overexpression and cyp78a8 cyp78a9 double mutants.
The CYP78A9 specific expression pattern suggests that the produced signal coordinates growth between sporophytic and gametophytic tissue, and between the structures that protect the ovules and the seed while they develop. Studies with the CYP78A9 promoter line suggest an interesting function of the gene in communication between the placenta, funiculus, and ovule during the fertilization process. Metabolic analyses of the mutants showed the existence of alterations in flavonoid content with respect to wild type. However, these alterations seem not to cause the observed phenotypes, as the tt4-1 es1-D double mutant presents purely additive phenotypes without having kaempferol and quercetin contents. Despite this, the work presented here contributes to the first characterization of metabolite differences between mutants in this gene family, giving interesting indicators for future investigation of the reaction(s) they catalyze.

Constructs and transformation
For the promoter analysis, 3020 bp directly upstream of the CYP78A9 (At3g61880) translational start was amplified and cloned into the pENTR-D vector (Invitrogen) using the specific primers (forward 5'-GGTGGGATACCGGTCAAGTG-3' and reverse 5'-GGATGCAGAGGAACAAGAGAGAG-3'). The resulting construct was sequenced

RNA extraction and RT-PCR
RNA was isolated using LiCl (Verwoerd et al., 1989). Around 1 µg RNA was treated with DNAse I (Invitrogen) and 1/10 of the treated RNA was used for cDNA synthesis with M-MLV Reverse Transcriptase or Superscript II Rnase H-Reverse Transcriptase (both from Invitrogen), following the supplier's instructions. The obtained cDNA was used for gene expression analyses. PCR experiments were performed using cDNA from wild type and from mutant tissues. A PCR using ACTIN (forward 5'-GTGTTGGACTCTGGAGATGGTGTG-3' and reverse 5'-GCCAAAGCAGTGATCTCTTTGCTC-3') primers for all the samples was used as a control. The reactions were performed as follows: 95ºC 3min, (95ºC 30 sec, 55ºC 40 sec, 72ºC 2 min) 30 cycles, and 72ºC 5 min.

In situ hybridization
Arabidopsis tissue was collected and for siliques, transverse cuts were made in order to remove the apical and basal tips (to allow better infiltration), fixed, and embedded in paraplast. In summary, the samples were placed in 10 mL vials and fixation was performed with FAA solution (50% Ethanol, 5% Glacial Acetic Acid, 3.7% Formaldehyde), vacuum was applied twice for 15 min, followed by replacement of fixative and then the samples were left overnight at 4ºC. Before imbibing the tissue in paraplast the samples were dehydrated in a series of ethanol (50%, 60%, 70%, 80%, 90%, and 95%), 30 min between each step, followed by incubation overnight in 95% ethanol at 4ºC. Afterwards, samples were incubated twice in 100% ethanol, 30 min between steps, and then left overnight at 4ºC. Next, the tissue samples were taken through a histoclear series of 1 hour each of 25% histoclear: 75% ethanol, 50% histoclear: 50% ethanol, 75% histoclear:25% ethanol, and finally 100% histoclear.
Infiltration was made with 100% fresh histoclear and 10 to 15 chips of paraplast and incubated overnight at room temperature. The next day, the solution was replaced with 100% paraplast and repeated 5 times during a time span of 6 hours at 60ºC, followed by making the molds. After this, sections (10 μ m) were made on a rotary microtome (Leica RM 2025) and mounted on slides.
The sense and antisense probes were synthesized by an in vitro transcription reaction using T7 and SP6 polymerase (Metabion), respectively. Prior to hybridization, the slides were dewaxed in histoclear twice for 10 min, 100% ethanol 2 min and ethanol series (95% to 30%; 1 min each), transferred to saline solution (NaCl 8.5 g/L) for 15 min, and then 5 minutes in PBS 1x. To improve probe penetration into the tissue, slides were incubated with proteinase K (1 µg/mL) at 37ºC for 30 min. The proteinase K digestion was stopped by keeping the slides for 2 min in 2 mg/mL glycine-PBS 1x. For postfixation, the slides were transferred to fresh fixation solution and kept for 10 min at room temperature. Next, the slides were washed twice with 1x PBS for 5 min and transferred to 0.1 M triethanolamine plus 1 mL acetic anhydride for 10 min. Followed by 5 min wash in 1x PBS and a reverse ethanol series (from 30%, 50%, 85%, 95% to 100%) 2 min each, a step in 0.5% NaCl for 2 min and 2 min 1x PBS. Hybridization was done in a humidified box with a digoxigenin-labeled probe at 52ºC overnight. Immunological detection was performed with 1:1250 antibody final concentration (Anti-Digoxigenin-AP, Roche) in BSA (10 g/L) solution for 2 hours and detected with an overnight incubation in NBT/BCIP containing solution as described previously (Coen et al., 1990).

Expression Analysis (microarray)
RNA was extracted (Verwoerd et al., 1989)  background correction, lowess normalization, intensity filter, analysis of replicates, and selection of differentially expressed genes. The software identifies differentially expressed genes by calculating an intensity-dependent z-score using a sliding window algorithm to calculate the mean and standard deviation within a window surrounding each data point, and then defines a z-score where z measures the number of standard deviations a data point has from the mean. z i = (R i -mean(R)) / sd(R), where z i is the zscore for each element, R i is the log-ratio for each element, and sd(R) is the standard deviation of the log-ratio. Ratio calculations of significant changes in gene expression derived from globally normalized data are performed by simply computing the ratio of the average of all of the measurements from one condition or sample to another. With this criterion, the elements with a z-score > 2 standard deviation are considered significantly differentially expressed genes (Cheadle et al., 2003).

PCR based genotyping
Identification of the cyp78a9 mutant allele was performed by PCR analysis using the  Beekwilder et al., 2008;Kachlicki et al., 2008;Bollinger et al., 2009;McNab et al., 2009;Gouveia and Castilho, 2010).The collected peak lists with m/z, and peak area intensities were further processed with Excel software, and the statistically significant differences in individual markers between the wild type and the mutants were demonstrated by pairwise t-test (two-tailed, two-sample unequal variance).

Accession numbers
Sequence data for this article can be found in The Arabidopsis Genome Initiative or GenBank/EMBL databases under the following accession numbers: CYP78A9            Table S1. Microarray upregulated and downregulated genes 2SD. Table S2. Comparison of CYP78A9-regulated and phythormone responsive genes. Table S3. Gene list derived for the comparison between macro and microarray data. Table S4. CYP78A9 co-expressed genes in flower tissue. Table S5. Gene functional classification derived from the genes co-expressed with CYP78A9 in flower tissue.