Characterization of a Citrus R2R3-MYB Transcription Factor that Regulates the Flavonol and Hydroxycinnamic Acid Biosynthesis

Flavonols and hydroxycinnamic acids are important phenylpropanoid metabolites in plants. In this study, we isolated and characterized a citrus R2R3-MYB transcription factor CsMYBF1, encoding a protein belonging to the flavonol-specific MYB subgroup. Ectopic expression of CsMYBF1 in tomato led to an up-regulation of a series of genes involved in primary metabolism and the phenylpropanoid pathway, and induced a strong accumulation of hydroxycinnamic acid compounds but not the flavonols. The RNAi suppression of CsMYBF1 in citrus callus caused a down-regulation of many phenylpropanoid pathway genes and reduced the contents of hydroxycinnamic acids and flavonols. Transactivation assays indicated that CsMYBF1 activated several promoters of phenylpropanoid pathway genes in tomato and citrus. Interestingly, CsMYBF1 could activate the CHS gene promoter in citrus, but not in tomato. Further examinations revealed that the MYBPLANT cis-elements were essential for CsMYBF1 in activating phenylpropanoid pathway genes. In summary, our data indicated that CsMYBF1 possessed the function in controlling the flavonol and hydroxycinnamic acid biosynthesis, and the regulatory differences in the target metabolite accumulation between two species may be due to the differential activation of CHS promoters by CsMYBF1. Therefore, CsMYBF1 constitutes an important gene source for the engineering of specific phenylpropanoid components.

identified. In grapevine, the expression pattern of VvMYBF1 directly correlates with that of VvFLS1 and subsequent flavonol accumulation 9 . In Japanese gentian (Gentiana triflora), GtMYBP3 and GtMYBP4 activate the expression of flavonol biosynthesis genes and increase flavonol accumulation when heterologously expressed in tobacco and Arabidopsis 10 . It is noteworthy that the characterized flavonol-related MYB proteins are usually involved in the regulation of the other branches of phenylpropanoid pathway, particularly the hydroxycinnamic acid metabolism. AtMYB12, the flavonol-specific transcriptional activator in Arabidopsis, induces the accumulation of high levels of caffeoyl quinic acids and flavonols when overexpressed in tomato and tobacco 11,12 . AtMYB11 and AtMYB111, the closest homologs of AtMYB12 in Arabidopsis, also show similar functions when overexpressed in heterologous systems 13,14 . In maize, the flavonol regulator P1, could affect the levels of specific phenylpropanoids such as ferulic acid when overexpressed in cultured maize cells, and the corresponding p1 locus is also implicated in controlling the levels of chlorogenic acid in maize silk tissue [15][16][17][18] .
Some MYB transcription factors involved in the regulation of hydroxycinnamic acid biosynthesis have also been characterized. For instance, AtMYB4 in Arabidopsis, VvMYB4a and VvMYB4b in grapevine, are characterized as negative regulators of hydroxycinnamic acid metabolism 19,20 . Several MYB regulators of hydroxycinnamic acid biosynthesis are also involved in the regulation of other classes of phenylpropanoid pathway, such as the VvMYBC2-L1 and VvMYBC2-L3 that additionally control the flavonoid levels in grapevine 20 . Increasing evidence reveal that many MYB transcription factors could impact on various branches of phenylpropanoid metabolism simultaneously, and even regulate nearly the whole pathway in some cases 21 , which may reflect the elaborate and complex transcriptional regulation mechanisms in plants.
Citrus is one of the most important fruit crops worldwide and citrus plants are abundant with phenolic compounds. Citrus plants contain various kinds of esters and glycosides of hydroxycinnamic acids, which carried out important functions in plant defense and fruit maturation processes 22,23 . The spatiotemporal accumulation of flavonols is divergent in citrus organs. The flavonol compounds were preferentially accumulated in leaves of most citrus varieties, while only a very small amount of flavonols was found in fruit tissues of a limited number of varieties such as lemon and lime 24,25 . The tissue-specific accumulation pattern implicates the special regulation mechanism of flavonol biosynthesis in citrus.
Recently, increasing insights on the biochemical properties of the phenolic compounds have been gained in citrus. However, our knowledge about the molecular mechanism underlying the biosynthesis and regulation of the phenylpropanoid metabolism is limited. Some structural genes of the phenylpropanoid pathway in citrus, like CHS and FLS genes, were isolated and their expression patterns have been investigated. The high transcript levels of citrus FLS gene in young leaves and the early developmental stage of the pulp tissues indicated that this gene was differentially regulated in the developmental stage and in a tissue-specific manner 26 . For the regulation of the structural genes, only the CsRuby, a R2R3-MYB transcription factor, was characterized as the anthocyanin regulator in blood orange, implicating the important roles of citrus MYB transcription factors in phenylpropanoid metabolism regulation 27 .
A total of 100 R2R3-MYB transcription factors have been identified in the genome of sweet orange, the principal variety of citrus species. Further phylogenetic analysis identified several putative MYB functional clades that were involved in phenylpropanoid metabolism including lignins, anthocyanins and flavonols biosynthesis 28 . Due to the importance of flavonols in plant physiology and human health and their differential accumulation pattern across citrus tissues, the flavonol-related MYB subgroup was selected for further investigation. In the present study, the citrus R2R3-MYB transcription factor CsMYBF1 was isolated. Functional analysis including ectopic expression in tomato and the RNAi suppression in citrus callus indicated that CsMYBF1 was involved in the regulation of the flavonol and hydroxycinnamic acid biosynthesis. Interestingly, overexpression of CsMYBF1 in tomato fruits reduced the contents of the major sugars and specially increased the amount of hydroxycinnamic acid compounds but not the flavonols. The unexpected phenotype changes were found to be due to the differential activation of CHS promoters by CsMYBF1, some potential factors including variation in MYBPLANT cis-elements may affect such regulatory differences.

Results
Sequence features of CsMYBF1. Three putative flavonol-related MYB transcription factors were identified from the genome-wide analysis of the sweet orange genome 28 , and these proteins were named as CsMYBF1 (Cs9g02640), CsMYBF2 (Cs5g33870) and CsMYBF3 (Cs5g33880), respectively. CsMYBF1 was selected for further analysis in this study, based on its expression feature and transactivation ability. The full-length of CsMYBF1 cDNA from sweet orange was obtained by 5′/3′ RACE experiments (GeneBank accession number KT727073). Analysis of the deduced amino acid sequence revealed that CsMYBF1 contained the R2R3 imperfect repeats involved in binding to target DNA sequences and highly conserved among MYB proteins. Apart from the significant sequence similarity among the MYB proteins within the N-terminus, the redefined SG7 motif ( 7 , which were characteristic of flavonol regulators, were partially conserved in CsMYBF1 (Supplementary Figure S1). A phylogenetic analysis of 48 MYB domains by the neighbor-joining method indicated that CsMYBF1 belonged to the flavonol clade which consisting of the MYB flavonol regulators from various plant species (Fig. 2a). The protein sequence of CsMYBF1 showed relative high similarity with the flavonol-related MYB proteins, and was most closely related to AtMYB111, with 89% identity in the R2R3 MYB domain, and 43% identity in the overall protein. The R2R3 MYB domain of CsMYBF1 was 84.3%, 81.5%, 88%, 80.6%, 84.3% identical to Arabidopsis AtMYB12, AtMYB11, grape VvMYBF1, maize ZmP1 and gentian GtMYBP3, respectively.
Expression pattern and subcellular localization of CsMYBF1. To reveal the spatiotemporal expression of CsMYBF1 in citrus, qRT-PCR analyses were performed using the total RNAs extracted from various organs of citrus and the pulp tissues at different development stages. As shown in Fig. 2b, transcript levels of CsMYBF1 were high in the flower and young leaf but relative low in the mature leaf, seed and callus. In the pulp tissues, the mRNA level of CsMYBF1 was the highest in 80 days after flowering (DAF), then decreased along with the fruit development and remained constant up to 170DAF. The expression features of CsMYBF1 coincided with the accumulation profiles of citrus flavonoids and the transcript levels of the citrus flavonol-related structural genes [29][30][31] .
The subcellular distribution of CsMYBF1 were analysed using transient expression in a citrus protoplast system. OsGhd7-CFP was used as a positive control since OsGhd7 was localized in the nucleus 32 . The green fluorescence generated by CsMYBF1-GFP was distributed in the same area as that of the cyan fluorescence generated by OsGhd7-CFP, indicating that CsMYBF1 was a nuclear protein (Fig. 2c). Additionally, CsMYBF1-GFP was stably transformed into tomato, and root tip cells of positive transgenic tomato lines possessed strong fluorescence signals in the nucleus (Supplementary Figure S2). Both these results provided evidence to show that CsMYBF1 was a nucleus-localized protein and may function as a transcription factor.

Effects of overexpression of CsMYBF1 on the metabolite and gene expression profiles in tomato fruits.
A total of 24 independent CsMYBF1-overexpressing tomato transformants were regenerated, with indistinguishable growth phenotype from that of the wild type (WT) plants. The chlorogenic acid contents in transgenic tomato flesh tissues detected by HPLC were associated with the transcript levels of CsMYBF1 (Supplementary Figure S3). Together with the WT plants, progeny of four transgenic lines (lines D7, C2, E2 and D4) were selected for further analysis, according to their phenolic compound contents and high transcript levels of CsMYBF1. As indicated in Fig. 3a,b, principal component analysis (PCA) of both the LC-MS and GC-MS data sets could distinguish metabolite profiles of WT and transgenic lines.
Results of the metabolic profile found by LC-MS indicated that specific phenylpropanoid compounds were markedly altered by CsMYBF1 expression (Table 1 and Fig. 3c). The caffeoyl quinic acids and dicaffeoyl quinic acids and the common hydroxycinnamic acid compounds in tomato, caffeic acid, ferulic acid and sinapic acid, except coumaric acid, were all significantly increased in all overexpression lines (Table 1). Additionally, their corresponding derivatives also showed enhanced levels in these transgenic lines. However, the detected flavonoids exhibited no significant differences between the WT and transgenic lines (Supplementary Table S2). The results were further confirmed by quantitative analyses of fruit methanol extracts using HPLC (Fig. 3d). The aromatic amino acid levels of phenylalanine, tryptophan and tyrosine were significantly reduced in some of the transgenic lines (Table 1).
GC-MS profiles revealed that the CsMYBF1-overexpressing tomato fruits displayed substantial changes in the level of primary metabolites ( Fig. 3e and Supplementary Table S3). Sugars were the predominant compounds in total detected primary metabolites. Of the 6 sugars measured, fructose, glucose, mannose, galactose and sucrose, with the exception of xylose, all showed significant reduced levels in the overexpression tomato fruits. For the organic acids, the increased amount of isocitric acid and glucuronic acid and decreased levels of 2-ketoglutaric acid were observed in transgenic fruits.
A large number of differentially expressed genes in the overexpression lines were identified from the Digital Gene Expression (DGE) data (Supplementary Table S4). Nearly all the genes encoding enzymes of the shikimate pathway and the phenylpropanoid pathway were all significantly induced (Table 2). However, genes encoding chalcone synthase, the key rate-limiting enzyme in tomato flavonoid biosynthesis, were not observed in the differential genes list. Additionally, expression levels of large amount genes involved in primary metabolism were also significantly induced in transgenic lines. For example, a series of genes involved in glycolysis, pentose phosphate pathway and sucrose biosynthesis, like genes encoding enolase, glucose-6-phosphate 1-dehydrogenase and sucrose synthase, were expressed more highly in transgenic fruits than in WT fruits ( Table 2). Expression levels of 17 selected phenylpropanoid related genes were further validated by quantitative RT-PCR and exhibited similar results (Fig. 3f). Taken together, ectopic expression of CsMYBF1 in tomato significantly increased transcript levels of a series of genes involved in primary and secondary metabolism, reduced the contents of the major sugars and selectively increased the amount of hydroxycinnamic acid compounds. Inconsistent with expectations, the flavonol contents were not significantly altered in transgenic fruits, which could be due to the constant transcript levels of the key CHS genes.
Effects of suppression of CsMYBF1 on the phenylpropanoid metabolism in citrus callus. To test the function of CsMYBF1 in a homologous system, RNA interference (RNAi) was used to suppress the expression of CsMYBF1 in citrus callus cultures. We preliminary examined citrus callus cultures of different varieties, and confirmed that the 'Guoqing' NO. 1 satsuma mandarin (G1) callus line, accumulates large amount of phenolic acids and flavonol compounds and expresses a quantity of CsMYBF1. Therefore, the G1 callus line was selected for the RNAi experiments.
Three callus lines (RNAi-12, 39, 57) with greatest reduction in CsMYBF1 expression and phenolic compound contents were further selected for PCA, indicating that the metabolic profiles of the WT and RNAi lines distinguished significantly (Fig. 4d). Flavonol accumulation in calli was visualized with diphenylboric acid 2-aminoethyl ester (DPBA), a flavonol-specific dye, which generates yellow-green fluorescence with kaempferol and yellow-gold with quercetin 33 . The yellow-green fluorescence was observed in the WT, indicating the mass accumulation of kaempferol compounds, which was consistent with the metabolite data identified by LC-MS. The CsMYBF1 RNAi callus lines displayed only small amount of fluorescence spots, similar with that in negative control (callus does not accumulate detectable flavonols), indicating the sharp decrease of flavonol contents in the RNAi lines. Representative pictures were shown in Fig. 4e. The differences of the flavonol contents between genotypes were further validated by HPLC analysis (Fig. 4f).
Further qRT-PCR expression analysis revealed that the down-regulation of CsMYBF1 accompanied with proportionally reduced expression of most phenylpropanoid pathway genes: i.e. CsPAL, CsC4H, Cs4CL, CsCHS, CsCHI and CsFLS in the transgenic lines except CsF3H (Fig. 4g). Different from the unchanged transcript levels of SlCHS genes in transgenic tomato fruits, the expression levels of CsCHS (CitCHS2) were significantly reduced in the RNAi callus lines. These expression changes of the structural genes could clearly account for the alteration of the phenolic compound composition in RNAi lines. The above results indicated the essential roles of CsMYBF1 in regulating the flavonol and hydroxycinnamic acid biosynthesis in citrus.

CsMYBF1 activates promoters of the phenylpropanoid pathway genes in tomato and citrus.
To evaluate the ability of CsMYBF1 to transcriptionally activate phenylpropanoid pathway genes, tobacco leaves were coinfiltrated with Agrobacterium harboring the CsMYBF1 overexpression construct together with a series of constructs containing a LUC reporter gene driven by the promoters of phenylpropanoid pathway genes. As illustrated in Fig. 5, CsMYBF1 activated the promoters of SlPAL (~approximately 4 fold compared with the control), SlC4H (~7 fold) and Sl4CL (~8 fold) in tomato. CsMYBF1 also induced the promoters of the CsPAL (~4 fold), CsC4H (~4 fold) and Cs4CL (~18 fold) in citrus, indicating its ability to activate the expression of the general phenylpropanoid pathway genes in both species. The promoters of the flavonol-specific FLS genes from two species were both greatly activated, whereas the activities of F3H promoters from two species were neither substantially altered by CsMYBF1. Interestingly, CsMYBF1 activated the CsCHS promoter approximately 30 fold, but the promoter activities of two SlCHS genes were not significantly activated. These results showed that CsMYBF1 activated the heterologous and homologous promoters of the general phenylpropanoid pathway genes and the flavonol-specific genes, but exhibited different specificity on the activities of CHS promoters from citrus and tomato.    34 . Sequence comparison revealed that cis-elements similar to MYBPLANT were presented in all these three selected promoters, indicating that these elements might be potential target motifs of CsMYBF1. As depicted in Fig. 6 (details in Supplementary Figure S4), a set of deletion and block mutation constructs were generated to test the functional roles of these motifs. Six putative MYBPLANT cis-elements were identified in the promoter regions of Cs4CL, and deletions of the four elements in the distal region did not significantly affect the activation, whereas the deletions and mutations for the two elements in the proximal region greatly reduced the transactivation (Fig. 6a). Three putative MYBPLANT cis-elements located upstream of SlFLS. Transactivation of the SlFLS promoter by CsMYBF1 was reduced by each deletion of the MYBPLANT-like cis-elements. Comparably, deletions in regions without MYBPLANT-like elements did not cause significant reduction (Fig. 6b). The mutation assays results for the last two elements also supported the above observations. The transactivation assays for the two target elements in CsCHS promoter exhibited similar results (Fig. 6c). The G-box cis-element was characterized as an important motif in CHS promoter from Arabidopsis 34 . Mutations in G-box sequences were also introduced in truncated CsCHS promoter, and the results indicated that CsMYBF1 could activate CsCHS promoter via the MYBPLANT cis-element but not the G-box cis-element (Fig. 6c).
To further ascertain whether the MYBPLANT cis-elements could be specially activated by CsMYBF1, we constructed chimeric promoters containing four copies of wild type (in CsCHS promoter) or mutant MYBPLANT motifs followed by the minimal CaMV35S promoter (Pro35Smin), and fused these promoters to the LUC reporter. As shown in Fig. 6d, chimeric promoter containing wild type MYBPLANT motif was strongly activated by CsMYBF1, while mutation of the MYBPLANT motif completely abolished this activation. The above results indicated that the MYBPLANT cis-element was an essential motif for CsMYBF1 in activating the phenylpropanoid biosynthetic genes.

CsMYBF1 specially binds to the MYBPLANT cis-elements. To further confirm whether CsMYBF1
interacted with the target promoters directly, a yeast one-hybrid system was used. As shown in Fig. 7a,b, the interaction tests indicated that CsMYBF1 interacted directly with both promoters of Cs4CL and SlFLS. To test whether CsMYBF1 specially bound to the region of MYBPLANT motif in the CsCHS promoter, region of about 60 bp in length surrounding the MYBPLANT cis-element, and the same region with the mutant MYBPLANT sequences were used as baits for binding assays. It was evident that yeast co-transformed with the CsMYBF1 and the natural promoter fragment, but not with the CsMYBF1 and the corresponding mutant region, grew well in the selective medium. Additionally, quadruple repetitions of the wild type/mutant MYBPLANT fragments in the CsCHS promoter were separately cloned into the reporter construct for binding assays, and the interaction tests exhibited identical results (Fig. 7b).
We next performed electrophoretic mobility shift assays (EMSA) to further verify the physical interaction between CsMYBF1 protein and MYBPLANT cis-elements. The purified recombinant CsMYBF1 protein was incubated with Cy3-labeled promoter fragments of the three selected genes (Fig. 8a). As shown in Fig. 8b, a specific DNA complex was detected with all three wild type probes, which was not formed when the respective MYBPLANT cis-element was mutated. Above all, the results showed that CsMYBF1 can specifically bind to the promoters of Cs4CL, SlFLS and CsCHS through the MYBPLANT cis-elements.

Discussion
In this study, a R2R3-MYB transcription factor gene, CsMYBF1, was isolated from citrus. The expression features of CsMYBF1 coincided with the accumulation profiles of flavonoids in citrus, which were mainly biosynthesized  Figure S4. (d) Diagrams showing four copies of wild type/mutant MYBPLANT ciselements linked to the minimal 35S promoter and the firefly luciferase reporter. The corresponding relative ratio of LUC/REN was shown on the right. The ratio of LUC/REN of the empty vector (EV) plus promoter was used as a calibrator (set as 1). Error bars indicate SE from six replicates. Asterisks indicate significant differences as determined by t-test analysis (P < 0.05).
Scientific RepoRts | 6:25352 | DOI: 10.1038/srep25352 in immature fruit and in young, rapidly growing leaves 29,30 . Accordingly, the mRNA levels of the flavonol-related structural genes were also high in these young active citrus tissues 31 . Additionally, the sugar contents in citrus pulp tissues were low in young fruits and increased along with the citrus fruit development 35 . The correlation between expression levels of CsMYBF1 and the metabolites contents in citrus fruits may implicate the important role of CsMYBF1 in the regulation of related metabolites.
The gene function was characterized through analysis of CsMYBF1 overexpression lines in tomato and the RNAi lines in citrus callus. The functional role of CsMYBF1 that acted as the flavonol regulator was validated in the RNAi lines. In transgenic tomato fruits, the general phenylpropanoid pathway genes were coordinately unregulated and the contents of the hydroxycinnamic acids were significantly induced. On the contrary, the hydroxycinnamic acid and its derivatives were significantly reduced in the RNAi lines. The opposite metabolite changes between the overexpression and RNAi lines provided strong evidence to support the function of CsMYBF1 in regulating the hydroxycinnamic acid biosynthesis. It was speculated that the regulation of hydroxycinnamic acid biosynthesis may be an additional and perhaps ancient function of the flavonol-related MYB proteins 12 .
It was thought that the flavonol-related MYB proteins have a preference for early biosynthesis genes in the flavonoid pathways as their target genes 36 . In this study, the CsCHS and CsFLS genes were both significantly induced by CsMYBF1. It was notable that the CsF3H gene, which belongs to early flavonoid biosynthetic genes in citrus, was not activated by CsMYBF1 in transactivation assays. The results could account for the unaltered expression level of CsF3H in the transgenic callus lines. F3H was responsible for the production of dihydroflavonol, the intermediate for the flavonol biosynthesis. In vitro experiments revealed that the citrus FLS enzyme harboured an intrinsic F3H-like activity as well 37 . Therefore we speculated that the regulation of the flavonol biosynthesis by CsMYBF1 may not involve CsF3H. In addition to flavonols, CsMYBF1 was also involved in controlling the accumulation of flavones derivatives in callus, suggesting a wide role of this R2R3-MYB regulator in the control of phenylpropanoid compounds. Above all, CsMYBF1 was an essential transcription factor for the regulation of the phenylpropanoid biosynthesis in citrus, and could serve as effective molecular tool for the breeding of citrus with optimized phenylpropanoid levels and compositions.
The constitutive overexpression of CsMYBF1 resulted in significant metabolite alterations in tomato fruits, including the elevated levels of a subset of hydroxycinnamic acid derivatives and reduced amounts of major sugars. The transcriptome data indicated that the expression of high number of genes in primary and secondary metabolism were significantly affected, such as genes involved in glycolysis, the pentose phosphate pathway and the phenylpropanoid pathway. Elevated expression of the genes involved in primary metabolism could increase the supply of precursors for secondary metabolism 38 . A series of significantly induced genes in primary metabolism implicated that there may be greater carbon flux through glycolysis and the pentose phosphate pathway to drive the biosynthesis of secondary metabolic pathways in CsMYBF1 overexpression fruits. As a result, the contents of the major sugars were significantly reduced, and the enhanced carbon flux could contribute to the increasement of phenylpropanoid contents in transgenic fruits.
In this study, overexpression of CsMYBF1 not only activate the genes of the phenylpropanoid pathway, but also reprogram the primary metabolism and change the carbon flux in tomato fruits. The homologous transcription factor of CsMYBF1 in Arabidopsis, AtMYB12, has also been reported to exhibit similar metabolic effects when overexpressed in tomato fruits, which can activate genes encoding enzyme of both primary metabolism and the phenylpropanoid metabolism, and control both the supply and demand of the carbon 38 . Similarly, homologous gene to CsMYBF1 in maize, P1, could also modify the expression of high number of genes including those involved in glycolysis 18 . So we speculate that the flavonol-related MYB transcription factors CsMYBF1 could also affect the genes involved in primary metabolism simultaneously, to ensure adequate supply of precursors for the phenylpropanoid biosynthesis.
Interestingly, the contents of flavonols and even all the flavonoids compounds in tomato fruits were not altered when overexpressing this flavonol regulator. This was beyond our expectations and quite different from other known flavonol-related MYB proteins when overexpressed in tomato 12,14 . Transcriptome data revealed that the CHS genes, encoding the rate-limiting enzyme of flavonoid biosynthesis, were not significantly induced in transgenic tomato, which may explain the constant flavonoid contents in CsMYBF1-overexpressing tomato fruits.
CHS play a pivotal role in flavonoid biosynthesis and the loss of CHS activity exerts significant metabolic effects in many plant species 8 . In CHS RNAi tomato plants, the contents of key flavonoids such as naringenin chalcone and rutin were dramatically reduced 39 . In Arabidopsis, the tt4 mutant plants, defective in CHS, showed the absence of downstream flavonols 40 . Ectopic concomitant expression of CHS and FLS induced the accumulation of flavonols in tomato flesh tissues, while either gene alone was insufficient. CHS and FLS have been regarded as key biosynthetic enzymes that determined accumulation of flavonols throughout the tomato fruit tissue 41 . The transcript levels of CHS genes were all significantly increased in transgenic tomato fruits, where induced flavonol accumulation was observed 12,14,42 . In our study, the expression of CHS genes was not altered in the transgenic tomato, albeit a set of other structural genes in flavonoid pathway were all significantly induced. Therefore, the CHS activity became the main limiting factor for flavonol accumulation in CsMYBF1-overexpressing tomato fruits.
The promoter activation assays indicated that CsMYBF1 significantly activated both promoters of FLS genes from tomato and citrus, consisted with the role of the flavonol regulator. However, CsMYBF1 only activated the promoter of CsCHS but not of SlCHS, which could lead to different effects on the alterations of phenylpropanoid metabolites. Such differential activation of CHS promoters by regulatory factors were also observed in previous studies. The combination of the anthocyanin-related regulators, C1 and R or AN2 and JAF13, can activate the promoter of CHS from maize, but not from petunia. The failure to active the petunia CHS promoter led to the anthocyanin pigmentation pattern that was not fundamentally altered in petunia plants when co-expressing AN2 and JAF13 43 . Similarly, the failure to activate the tomato CHS promoters could give rise to the constant flavonoid levels in CsMYBF1-overexpression tomato fruits. Some potential factors that affect the transactivation of CHS promoters may be responsible for such regulatory differences.
The CHS promoters have been well studied in many other plant species. The MRE was widely distributed in CHS promoters from different species and required for the activation by many flavonol-related MYB proteins 11,44 . MYBPLANT motif is a common MYB-binding cis-acting element and frequently resides in promoters of the genes involved in phenylpropanoid biosynthesis 34 . The MRE sequence 5′ -ACCTACC-3′ was identical with most regions of the conserved MYBPLANT cis-elements, and these two motifs were actually related to the same sequences in many cases. A series of deletions and mutation analyses indicated that CsMYBF1 could activate a set of target promoters through MYBPLANT cis-element, but not other motifs like G-box cis-element. Therefore, the MYBPLANT cis-element could also act to confer the CHS promoter activation in response to CsMYBF1. The cis-elements play an important role in determining the target gene specificity of MYB proteins, changes in the cis-elements sequences may affect the biochemical interactions with transcription factors and cause the phenotypic differences 36,45 . Therefore, the failure of CsMYBF1 to activate SlCHS promoters may be due to the corresponding elements in tomato have already diverged away to be recognized by CsMYBF1.
However, the distribution of the target motifs alone cannot fully explain the observed target gene specificity of CsMYBF1. Other factors like the surrounding sequences of the cis-elements and the numbers of cis-elements, may also affect the transactivation. For example, the CsF3H was not significantly activated by CsMYBF1, albeit the conserved MYBPLANT motif was identified in its promoter region. Additionally, differences in protein sequences of CsMYBF1 to closely related MYB genes of other species, especially variation in particular amino acid domains, also possibly leading to different specificity to CHS promoters.
Taken together, we speculated that the flavonol regulator CsMYBF1 specially increased the hydroxycinnamic acid contents and was not involved in the regulation of flavonol biosynthesis could be due to the failure to activate CHS promoters in the heterologous expression system. Therefore, CsMYBF1 could be considered as a valuable candidate gene for plant metabolic engineering, which can significantly affect primary metabolism and increase flux to phenylpropanoid metabolites of interest without incurring effects on the flavonoid contents.

Methods
Isolation of CsMYBF1 cDNA and phylogenetic analysis. The full-length of the CsMYBF1 cDNA was obtained using the 5′ RACE kit and 3′ RACE kit (Takara) following the supplier's protocol. The primers were listed in Supplementary Table S6. Different protein sequences were aligned and edited using the Clustal W program and GeneDoc software. A neighbor-joining tree was built up using the MEGA 5.0 software with the bootstrap values from 1000 replicates.
Plant materials. Samples of leaves (young and mature stages), callus, flowers, seeds and pulp from different fruit development stages were all derived from the ' Anliu' sweet orange (Citrus sinensis L. Osbeck) in Institute of Citrus Research of Guangxi. All samples were frozen in liquid nitrogen immediately, and stored at − 80 °C until use. The G1 callus was maintained on solid MT basal medium and subcultured every 20 days. The transgenic tomato plants were grown under standard glasshouse conditions. The Nicotiana benthamiana was planted in a growth chamber controlled at 14 h light, 10 h dark, and 24 °C cycles. Subcellular protein localization. The 35S:CsMYBF1-GFP fusion construct was produced by inserting the full ORF of CsMYBF1 into the pM999-35 vector. Plasmids were extracted and purified using the Plasmid Midi Kit (QIAGEN) following manufacturer's protocol. The isolation of the mesophyll protoplast of HB pummelo (Citrus grandis (L.) Osbeck) was carried out according to Grosser and Gmitter 46 . The OsGhd7 in rice was used as a nuclear maker. The 35S:CsMYBF1-GFP and 35S:OsGhd7-CFP plasmids were co-transformed into citrus protoplasts according to the procedures described by Fang et al. 32 . The root tip cells of the positive stable overexpression tomato lines for 35S:CsMYBF1-GFP were also used. The florescence images were captured by using a confocal laser-scanning microscope (TCS SP2, Leica, Germany). Generation of constructs and transformation. The full length of CsMYBF1 was ligated into binary vector PBI121 driven by 35S promoter. The fragment of CsMYBF1 ORF (428 bp length, 562-989 bp) was recombined into the gene silencing vector pHGRV by BP reaction. These constructs were electroporated into Agrobacterium strain EHA105. The overexpression and RNAi vector was transformed into tomato and callus, respectively.
Tomato variety Ailsa Craig (AC) was transformed according to the method of Fillatti et al. 47 . The transformation for callus line was performed according to Duan et al. 48 . The putative transgenic lines were examined by semi-quantitative RT-PCR using target gene specific primer (GSP-CS) and the NPTII gene-specific primer (NPTII) (Supplementary Table S6).
RNA isolation and quantitative analysis of gene expression. Total RNA was isolated using a modified Trizol extraction protocol 49 . First-strand cDNA was synthesized using PrimScript ™ RT reagent Kit with gDNA Eraser (Takara). Quantitative real-time RT-PCR was performed on ABI Prism 7900HT (Applied Biosystems) with the SYBR Green system. The gene specific primers used in qRT-PCR analysis were listed in Supplementary Table S6. The Actin gene was used as internal control.
Transcriptome analysis. Two independent transgenic tomato plants (line D7 and C2) and the WT plants were divided into two groups (overexpression/WT) and each with two independent biological replicates. Fruit samples from at least five plants of each genotype, which were harvested between 7 and 10 days after breaker stage, were pooled for total RNA isolation. The DGE experiments were performed by Novogene Bioinformatics Technology Co. Ltd (Beijing). Total reads were mapped to the tomato genome (http://solgenomics.net/) using Top Hat (2.0.9) software 50 . Differential expressed genes between the transgenic and WT plants were defined with the edgeR 51 , based on log 2 fold change >1 and adjusted P value < 0.05. DGE results have been submitted to NCBI with accession number GSE73300. Putative metabolite identification was achieved by searching the accurate m/z values of the peaks against the tomato fruit specific databases Komics (http://webs2.kazusa.or.jp/komics) and the public databases (http:// www.massbank.jp, https://metlin.scripps.edu), with a maximum deviation of observed mass to calculated mass of 5 ppm. For some other compounds, the identification was based on the information from published literatures. Candidate metabolites were further confirmed according to the corresponding in-source fragments and the retention times, if present, and using commercially available authentic standards. The putative compounds were quantified by calculating the area of each individual peak and comparing this to calibration curves obtained from the corresponding closest authentic standards. HPLC analysis was conducted according to the methods described in Chen et al. 52 with some modifications.
GC-MS analysis. The GC-MS analysis of the methanol extract of the fresh sample from the overexpression/ WT tomato fruit was carried out by using a Thermo Trace GC Ultra coupled with DSQ II mass spectrometer (Thermo Fisher), based on the methods described in Yun et al. 54 . The processed data were used for PCA using the SIMCA-P 11.5 program (Umetrics).

Dual luciferase assay of transiently transformed tobacco leaves. The putative promoters
(approximately 1 kb upstream of the ATG start codon) of a series of phenylpropanoid biosynthesis genes were amplified in citrus and tomato respectively; the oligonucleotide sequences containing four tandem repeats of wild type/mutant MYBPLANT cis-elements were directly synthesized; various deletions of the promoter fragments were PCR-amplified from tomato and citrus genomic DNA; the block mutations for the MYBPLAT cis-elements in the corresponding promoter sequences were directly synthesized or amplified by recombinant PCR, all these fragments were inserted into the cloning site of pGreen0800-LUC. The full length of CsMYBF1 mRNA was amplified and cloned into the effector vector (Modified PCAMBIA1380) driven by 35 S promoter. All the primers used were listed in Supplementary Table S6. Agrobacterium infiltration and measurement of the luciferase activities were conducted as described previously 55,56 . Yeast one-hybrid assays. The oligonucleotide sequences containing four tandem repeats of wild type/ mutant MYBPLANT cis-elements, and other corresponding promoter sequences were cloned into the pHIS2 to generate the reporter vector. The full-length ORF of CsMYBF1 was amplified by PCR and fused to the GAL4 activation domain in the vector pGADT7-Rec2 (Clonetech) to create a fusion protein pGADT7-CsMYBF1 (effector vector). All of the constructs were co-transformed into yeast strain Y187 according to the manufacturer's instructions. The positive clones were subsequently incubated in SD (-Leu, -Trp) liquid media to an OD 600 of 0.1(10 −1 ) and diluted in a 10 × dilution series. From each dilution, 5 μ L was spotted on SD (-His, -Leu, -Trp) media plates supplemented with 0 or 50 mM 3AT (Sigma-Aldrich).
EMSA. The full length of CsMYBF1 was ligated into pET28a vector, and the resultant pET28a-CsMYBF1 vector was introduced into Escherichia coli BL21(DE3). The His-tagged protein was purified with the nickel-nitrilotriacetic acid (Ni-NTA) agarose column. The Cy3-labeled oligonucleotide probes (in Fig. 8a) were directly synthesized and annealed. Binding reactions were performed using a solution containing 0.1% Nonidet P-40, 1 mM benzamidine, 0.5 mM phenylmethylsulfonyl fluoride (PMSF), 0.5 mM DTT, 50 μ g/mL BSA and 100 ng/μ L poly (dI-dC). A total of 5 μ M Cy3-labeled probes and 1 μ g purified His-CsMYBF1 protein were added to final volume of 20 μ L and incubated at 4 °C for 2 h. The resulting DNA-protein complexes were loaded onto a pre-run 6% polyacrylamide gel. Electrophoresis were performed at 4 °C using 0.5× TBE (Tris-Borate-EDTA) as the electrophoresis buffer. Gel images were acquired using Amersham TM Imager 600 (GE Healthcare).