Functional Characterization of a Flavone Synthase That Participates in a Kumquat Flavone Metabolon

Flavones predominantly accumulate as O- and C-glycosides in kumquat plants. Two catalytic mechanisms of flavone synthase II (FNSII) support the biosynthesis of glycosyl flavones, one involving flavanone 2-hydroxylase (which generates 2-hydroxyflavanones for C-glycosylation) and another involving the direct catalysis of flavanones to flavones for O-glycosylation. However, FNSII has not yet been characterized in kumquats. In this study, we identified two kumquat FNSII genes (FcFNSII-1 and FcFNSII-2), based on transcriptome and bioinformatics analysis. Data from in vivo and in vitro assays showed that FcFNSII-2 directly synthesized apigenin and acacetin from naringenin and isosakuranetin, respectively, whereas FcFNSII-1 showed no detectable catalytic activities with flavanones. In agreement, transient overexpression of FcFNSII-2 in kumquat peels significantly enhanced the transcription of structural genes of the flavonoid-biosynthesis pathway and the accumulation of several O-glycosyl flavones. Moreover, studying the subcellular localizations of FcFNSII-1 and FcFNSII-2 demonstrated that N-terminal membrane-spanning domains were necessary to ensure endoplasmic reticulum localization and anchoring. Protein–protein interaction analyses, using the split-ubiquitin yeast two-hybrid system and bimolecular fluorescence-complementation assays, revealed that FcFNSII-2 interacted with chalcone synthase 1, chalcone synthase 2, and chalcone isomerase-like proteins. The results provide strong evidence that FcFNSII-2 serves as a nucleation site for an O-glycosyl flavone metabolon that channels flavanones for O-glycosyl flavone biosynthesis in kumquat fruits. They have implications for guiding genetic engineering efforts aimed at enhancing the composition of bioactive flavonoids in kumquat fruits.


INTRODUCTION
Flavones, one of the largest subclasses of flavonoids, perform various physiological roles in plants, such as participating in responses to biotic and abiotic stresses (Morimoto et al., 1998;Du et al., 2010a,b;Kong et al., 2010;Jiang et al., 2016;Righini et al., 2019). In addition, these metabolites contribute to the internal and external qualities of fruits, herbs, and vegetables by improving their appearance, taste, and nutritional value. Kumquats look like miniature oranges, are consumed worldwide, and contain diverse and abundant bioactive flavonoids. Kumquat extract possesses several health-promoting effects, and these effects are associated with the compositions and quantities of these flavonoids (Sadek et al., 2009;Barreca et al., 2014;Lou et al., 2015;Nagahama et al., 2015). In contrast to the many other fruits that tend to biosynthesize anthocyanins and flavonols, kumquats and other citrus fruits predominantly accumulate large amounts of flavanones and flavone derivatives (Montanari et al., 1998;Lou et al., 2015;Butelli et al., 2017;Zhao et al., 2020).
In O-glycosides, the sugar moieties are bonded to the hydroxyl groups of flavonoid aglycones. C-glycosylation can occur before the formation of the flavone backbone, whereas O-glycosylation usually occurs after backbone formation (Martens and Mithofer, 2005;Wu et al., 2016;. A kumquat Cglycosyltransferase (CGT) utilizes 2-hydroxyflavanones, rather than flavones, as sugar acceptors and produces the corresponding C-glucosides (Ito et al., 2017). Therefore, FNSII enzymes might operate through multiple catalytic mechanisms.
The enzymes necessary for flavonoid biosynthesis could form flavonoid metabolons (ordered protein complexes), which are believed to occur in diverse plant species (Hrazdina and Wagner, 1985;Winkel, 2004;Nakayama et al., 2019). Specific protein-protein interactions could regulate the biosynthetic efficiency and spatiotemporal distributions of flavonoids. Metabolons are anchored to the endoplasmic reticulum (ER) membrane; therefore, ER-resident P450 enzymes could serve as anchors for metabolons. The formation of flavonoid metabolons on ER-resident P450 occurs in multiple plants, including rice (Shih et al., 2008), snapdragon, torenia (Fujino et al., 2018), and soybean (Mameda et al., 2018). In soybean roots, 2-hydroxyisoflavanone synthase (a P450 enzyme) can interact with chalcone reductase (CHR) and CHS to form a flavonoid metabolon, in which GmCHR5 is located near CHS to reduce the transit time of the CHR substrate from CHS to CHR (Mameda et al., 2018). In snapdragon and torenia plants, enzymes involved in late-stage anthocyanin biosynthesis interact with FNSII, although FNSII activity was not necessary for anthocyanin biosynthesis (Fujino et al., 2018). Therefore, ER-bound P450s (such as FNSII) are posited as components of flavonoid metabolons.
Considering that FNSII (also including F2H) are the only P450 enzymes involved in the biosynthetic pathway for O/C-glycosyl flavones in kumquat fruits (Figure 1), we hypothesized that they likely play roles in anchoring dynamic flavone metabolons to the ER. According to our model, specific protein-protein interactions modulate the diversity in terms of the accumulation patterns and chemical structures of flavone derivatives. However, no FNSII enzymes have been characterized in kumquat plants; this has limited further study on the regulatory and biosynthetic mechanisms of flavones and their derivatives. Therefore, in this study, we isolated 2 FNSII genes (FcFNSII-1 and FcFNSII-2) through RNA-seq analysis. The results of in vivo, in vitro, and in planta experiments indicate that FcFNSII-2 directly converted flavanones into corresponding flavones, whereas FcFNSII-1 had no detectable catalytic activity against flavanones. Both FcFNSII-1 and FcFNSII-2 localized to the ER; however, only the latter interacted with upstream enzymes in the flavonoidbiosynthesis pathway. These findings imply that FcFNSII-2 might serve as a nucleation site for flavone-metabolon formation, mediating the biosynthesis of O-glycosyl flavone and its derivatives in kumquats.

Plant Materials
"Huapi" kumquat trees (Fortunella crassifolia) were grown at the Citrus Research Institute (Chongqing, China). The flowers, young leaves, and young shoots were collected in the spring, and the fruits were collected at different growth stages, including 30, 90, and 150 days after full blooming (DAB). The samples were immediately frozen in liquid nitrogen and stored at −80 • C until use.

Isolation and Bioinformatics Analysis of FNSIIs
Total RNA from the fruit peels, flowers, young leaves, and young shoots of "Huapi" trees were extracted using the MiniBEST Plant RNA Extraction Kit (TaKaRa Bio). cDNA was synthesized  Table 1). A phylogenetic tree of the deduced amino acid (AA) sequences was constructed using the MEGA 6.0 program (Tamura et al., 2013), using the neighborjoining method.

Quantitative Reverse Transcription-Polymerase Chain Reaction Analysis
Following cDNA synthesis, quantitative PCR was performed using the TB Green R Fast qPCR Mix (TaKaRa, Beijing, China) on a LightCycler 480 real-time system (Roche, Basel, Switzerland). The PCR mix (10 µL) contained 5 µL of 2 × TB Green Premix Ex Taq II (Tli RNaseH Plus), 50 ng of cDNA, and 0.4 µM of each primer (Supplementary Table 1). The reaction conditions were: 95 • C for 30 s, followed by 35 cycles at 95 • C for 5 s, 56-58 • C for 10 s, and 72 • C for 25 s. Actin was used as the internal reference gene for normalizing the Quantitative PCR data. The relative transcript levels were calculated using the 2 − Ct method. Three biological replicates were analyzed for each gene. The sequences of the primers are included in Supplementary Table 1.

Yeast Expression of Recombinant Proteins and Enzyme Assays
Yeast expression and in vivo yeast assays was performed according to previous studies (Lam et al., 2014;Wu et al., 2016;Righini et al., 2019), with several minor modifications. The ORF of each FcFNSIIs were ligated directly into the yeast expression vector, pESC-HIS (to replace the FLAG epitope) and expressed under the control of the GAL10 promoter. Unless otherwise specified, all recombinant vectors were constructed following a seamless cloning strategy (ClonExpress Ultra One Step Cloning Kit, Vazyme, China). All recombinant plastids and the empty vector (negative control) were separately introduced into the Saccharomyces cerevisiae strain WAT11. Briefly, yeast cultures were grown overnight at 30 • C in liquid synthetic dextrose minimal medium lacking histidine; they were collected through centrifugation and diluted to an optical density (at 600 nm: OD 600 ) of 1.0 in induction medium (synthetic galactose minimal medium). After inducing protein expression for 6 h, the substrate naringenin or isosakuranetin was added to a final concentration of 0.2 mM. Prior to addition, the flavanone substrates were dissolved in dimethyl sulfoxide (DMSO). After a 24 h incubation, the reactions were terminated by extraction with ethyl acetate, evaporated under nitrogen gas, and dissolved in 200 µL of 80% methanol for ultra-high performance liquid chromatographyquadrupole time-of-flight mass spectrometry (UPLC-Q-TOF-MS) analysis.

Transient Overexpression Assay in Kumquat Peel Tissue
The transient overexpression assay was performed, as described previously (Wang F. et al., 2017;Gong et al., 2021;Zhao et al., 2021), with minor modifications. The ORF of each FcFNSII gene was inserted into the pBI121 vector using a seamless cloning strategy to generate the constructs. All constructs and the empty vector were separately introduced into the Agrobacterium strain, EHA105. The EHA105 strain was cultivated in liquid LB medium at 28 • C; the cells were collected through centrifugation, and diluted to an OD 600 of 0.8 in infiltration medium containing 0.05 M MES, 2 mM Na 3 PO 4 , 0.5% D-glucose, and 0.1 mM acetosyringone. The suspensions were injected into the peels of kumquat fruits (150 DAB). The injected fruits were maintained in the dark for 24 h and then under a long photoperiod (16 h light and 8 h dark) for 4 days. Flavonoid components were extracted, as described previously (Liu et al., 2018(Liu et al., , 2020.

Flavonoid Extraction and Ultra-High Performance Liquid Chromatography-Quadrupole Time-of-Flight-Mass Spectrometry Conditions
Filtered samples from enzyme assays and tissue sample extracts were separated on an ACQUITY UPLC BEH C18 column (2.1 × 100 mm, 1.7 µm, United Kingdom) connected to an ACQUITY UPLC I-Class PLUS System (Waters, Milford, MA). The mobile phases, consisting of 0.01% formic acid solution (A) and methanol (B) were used at a flow rate of 0.4 mL·min −1 , with the following gradient program: 0-0.6 min, 90-80% A; 0.6-5 min, 80-30% A; 5-7 min, 30-10% A; and 7-8 min, 10-90% A. A photodiode-array detector was used to scan from 240 nm to 400 nm. A Xevo G2-S Q-TOF instrument (Waters MS Technologies, Manchester, United Kingdom) was used with an ESI source that was set from a mass: charge (m/z) ratio of 100-1,000. Data were collected in real time (scan time, 0.5 s; interval, 15 s) and processed using Waters UNIFI 1.7 software.

Subcellular-Localization Assay
The full-length ORFs of FcFNSIIs and their partial segments were used for subcellular-localization assays. These sequences (which lacked a stop codon) were separately ligated into the pBWA(V)HS-GLosgfp vector (BioRun, Wuhan, China) in frame with the 5 end of the coding sequence of green fluorescent protein (GFP) to create different fusion constructs (i.e., 35S::FcFNSII-1/2-GFP, 35S::nFcFNSII-1/2-GFP, and 35S::delFcFNSII-1/2-GFP). The red fluorescent protein (RFP) gene was fused with the ER localization sequence (MKTNLFLFLFLIFSLLLSLSSAEF) and used as an ERlocalization marker. The fusion constructs were co-transformed into tobacco leaves (Nicotiana benthamiana), along with the ERlocalization marker, through Agrobacteria-mediated infiltration. Following transformation, the tobacco leaves were maintained at 25 • C under a long photoperiod (16 h light and 8 h dark) for at least 48 h. The fluorescence signals were detected using an LSM 700 confocal microscope (Zeiss).

Identification and Sequence Analysis of Flavone Synthase II Homologs
TBLASTN (E-value < 1e −5 ), using functionally characterized FNSIIs or F2Hs as queries against Fortunella hindsii genome 1 and transcriptome (Zhu et al., 2019), identified many candidate genes. The AA sequence identity between the candidates and functionally characterized FSNIIs and F2Hs was low; therefore, it was difficult to identify the genuine F2Hs or FNSIIs among the candidates. We used transcriptome co-expression analyses to screen target genes. Based on transcriptome analysis of "Hongkong" kumquat (F. hindsii), generated from 13 different tissues from five organs (Zhu et al., 2019), two FNSII homologs were identified. Their expression was positively correlated with the expression levels of upstream genes in the flavonoid-biosynthesis pathway, such as CHS1, CHS2, CHI, and chalcone isomerase-like (CHIL) (Pearson's correlation coefficient; between FcFNSII-1 and flavonoid biosynthetic genes was >0.7 and between FcFNSII-2 and flavonoid biosynthetic genes >0.6) (Figure 2A). The ORFs of the two FNSII homologs (FcFNSII-1 and FcFNSII-2) were cloned through RT-PCR-based amplification from "Huapi" kumquat, a close relative of the "Hongkong" kumquat cultivar. The ORFs of FcFNSII-1 and 1 http://citrus.hzau.edu.cn/index.php/ FcFNSII-2 were 1542 and 1557 base pairs in length, encoding 513 and 518 amino acids with molecular weights (MWs) of 58.31 and 58.48 kDa, respectively. Sequence alignment revealed that both the FcFNSII-1 and FcFNSII-2 proteins harbor signature P450 motifs, such as a heme-binding motif and a Pro hinge region (Supplementary Figure 1). The phylogenetic analysis revealed that the proteins could be divided into two major groups (FSNIIs and F2Hs; Figure 2B). FcFNSII-1 and FcFNSII-2 clustered with members of the FNSII clade, which commonly catalyze the direct conversion of flavanones to flavones.
We analyzed the correlations between the accumulation of flavone derivatives and the expression patterns of FcFNSII-1 and FcFNSII-2 in different tissues including flowers, young leaves, young shoots, and fruit peels at 30, 90, and 150 DAB. Five flavone derivatives could be quantified in the different tissues of "Huapi" kumquats, including two O-glycosyl flavones (RHO and FOR) (Supplementary Figures 2A, 3) and three C-glycosyl flavones (APN, VIC, and MAR) (Supplementary Figures 2B, 3). The contents of O-glycosyl flavones in the peels at 30 DAB were higher than those in the other tissues and stages; C-glycosyl flavones showed preferential accumulation in flowers and in young leaves. Quantitative PCR was used to compare the spatiotemporal expression patterns of FcFNSIIs (Supplementary Figure 2C) with the spatiotemporal-accumulation patterns of different flavone derivatives. The FcFNSII-1 and FcFNSII-2 transcripts were preferentially expressed in peels of fruits at 30 DAB, where the levels of O-glycosyl flavones were relatively high.

In vivo and vitro Enzyme-Activity Assays for FcFNSIIs
To examine the catalytic functions of both the FcFNSII genes, the ORF of each sequence was subcloned into the pESC-HIS vector and transformed into the WAT11 yeast strain. WAT11 cells express an Arabidopsis ATR1 P450 reductase that provides reducing equivalents to plant P450s, such as the FNSII enzymes. The transformed yeast cultures were incubated with naringenin or isosakuranetin. These two flavanones are the predominant precursors for the biosynthesis of flavone derivatives, such as RHO, FOR, isovitexin, vitexin, VIC, and MAR that are naturally present in kumquat fruits (Figure 1). After incubation for 24 h, yeast cells expressing FcFNSII-2 metabolized naringenin and isosakuranetin to apigenin and acacetin, respectively (Figures 3A,B), whereas FcFNSII-1 showed no detectable catalytic activities against naringenin and isosakuranetin. Control cells harboring the empty vector did not produce any detectable apigenin and acacetin. The reaction products were detected using UPLC-Q-TOF-MS; their retention times and mass-fragmentation patterns were compared with those of authentic standards.
To gain further insight into the enzymatic properties of FcFNSII-2, microsomes were extracted from the transformed yeast cells and assayed for their enzymatic activities on flavanone substrates in Tris/HCl buffer supplemented with NADPH (which supplied reducing equivalents for P450). After incubation for 1 h, the reaction products were extracted and analyzed using UPLC-Q-TOF-MS. Changing the pH or temperature strongly affected the activities of the recombinant enzymes; they exhibited maximum activity at pH 7.0∼7.5 and a temperature of 30∼35 • C (Supplementary Figure 4). Under these optimized conditions, the kinetic parameters of FcFNSII-2 were determined. The Km and Vmax for naringenin were 3.77 µM and 12.33 fkat mg −1 , and those for isosakuranetin were 2.10 µM and 9.63 fkat mg −1 , respectively ( Figure 3C). The relatively higher Vmax: Km ratio for isosakuranetin suggested that it was slightly more preferred than naringenin as an in vitro substrate for FcFNSII-2.

In planta Enzyme Assays of Recombinant FcFNSIIs
To further understand the metabolic functions of FcFNSIIs in planta, we transiently overexpressed them in the peels of kumquat fruits at 150 DAB (Figure 4A), as done previously with success in citrus plants (Wang F. et al., 2017;Gong et al., 2021;Zhao et al., 2021). Gene-transcript levels at the injected sites were determined using quantitative PCR ( Figure 4B). Compared to that in the control, the transcript levels of FcFNSII-1 and FcFNSII-2 were significantly higher at the corresponding injection sites. FcFNSII-2 overexpression significantly enhanced the expression of FcCHS1, FcCHS2, FcCHI, and FcCHIL genes; however, FcFNSII-1 overexpression did not significant affect the expression of these flavonoid-related genes. CHS, CHI, and CHIL are key enzymes in the early biosynthetic pathway of flavonoids, and their expression levels are closely related to flavonoid accumulation.
To confirm the flavonoid accumulation-enhancing function of FcFNSIIs in kumquats, the main flavonoids were analyzed and quantified using UPLC-Q-TOF-MS at the injected sites. The three flavonoids, 3 ,5 -di-C-β-glucopyranosylphloretin (DCGP), RHO, and FOR, showed differential accumulation between the control and pBI121-FcFNSII-2 vector-injected sites ( Figure 4C  and Supplementary Figure 3). The amounts of DCGP decreased significantly at the pBI121-FcFNSII-2 injected sites; however, those of RHO and FOR increased significantly. These changes were associated with the elevated levels of FcFNSII-2 and other flavonoid-related genes. FcFNSII-1 overexpression did not influence the flavonoid contents, as expected. Taken together, the results of in vivo, in vitro, and in planta experiments demonstrated that FcFNSII-2 can directly convert flavanones into the corresponding flavones.

Subcellular Localization
The two FcFNSIIs were predicted to harbor an N-terminal membrane-spanning domain (NMSD) spanning AAs 1-27 (Supplementary Figure 5), which anchors the P450s to the ER. The subcellular localizations of FcFNSII-1 and FcFNSII-2 were investigated by overexpressing C-terminally tagged GFP fusion proteins (FcFNSII-1-GFP and FcFNSII-2-GFP) to avoid interference with the NMSD. In tobacco leaf epidermal cells, the fluorescent signals of the fusion proteins were in the pattern of a network, which is consistent with the fluorescent signals specific to the ER (red color; Figure 5), suggesting that both FcFNSII-1 and FcFNSII-2 are localized to the FIGURE 3 | UPLC-Q-TOF-MS analysis of the reaction products of FcFNSIIs, as determined by performing in vivo enzyme-activity assays. (A) The schematic diagram represents the catalytic mechanism of FcFNSII-2. (B) In vivo enzymatic assays performed with recombinant FcFNSII-2. The assays were conducted using naringenin and isosakuranetin as substrates. The mass spectra of apigenin and acacetin standards and the reaction products produced by incubation with FcFNSII-2. (C) In vitro kinetic parameters of recombinant FcFNSII-2 were determined using yeast microsome extracts.

Involvement of FcFNSIIs in the Flavone Metabolon
The flavonoid-synthesizing enzymes in plants associate with the cytoplasmic surface of the ER to form metabolons, which is commonly mediated by ER-bound P450 (Shih et al., 2008;Waki et al., 2016;Fujino et al., 2018;Mameda et al., 2018). To examine whether FcFNSIIs interact with the upstream enzymes in the flavonoid pathway to form a complex, interactions between FcFNSII-1/FcFNSII-2 and CHS1, CHS2, CHI, and CHIL were analyzed using SU-YTH. Considering that FcFNSII enzymes contain an NMSD that locates them to the ER membrane, we designed SUC FcFNSII-Cub-LexA-VP16 constructs to ensure the NMSDs of both P450 proteins could be anchored to membranes. The yeast-growth results indicated that FcFNSII-2 interacted with FcCHS1, FcCHS2, and CHIL. No appreciable yeast growth was observed when interactions between FcFNSII-2 and CHI enzymes were assayed (Figure 6). In addition, no apparent interaction occurred between FcFNSII-1 and these enzymes (Supplementary Figure 6).

DISCUSSION
The stepwise catalytic reactions of CHS and CHI drive carbon flow from the phenylpropanoid pathway toward the flavanonebiosynthesis pathways; these drive the formation of different subclasses of flavonoids. Functional loss of Ruby gene (which encodes a MYB transcription factor) caused by a singlenucleotide insertion that resulted in a frame-shift mutation abolished anthocyanin biosynthesis in kumquat plants (Butelli et al., 2017;Huang et al., 2018). Therefore, flavone-derived metabolites are one of the predominant flavonoids synthesized in kumquats. Flavone biosynthesis from flavanones was catalyzed by either FNSI or FNSII, which resulted in a C2-C3 double bond in the C-ring (Martens and Mithofer, 2005;Lee et al., 2008;Han et al., 2014;Lam et al., 2017).
Here, we investigated the activities of two kumquat FNSII homologs by transiently overexpressing the recombinant  enzymes in kumquat peels and performing enzyme assays (Figures 3, 4). The catalytic activity of FcFNSII-2 was similar to that of gentian and soybean FNSII; it was capable of directly producing flavones from flavanones in yeast assays. FcFNSII-1 did not lead to flavone accumulation. Consistently, overexpressing FcFNSII-2 in kumquats enhanced accumulation of the O-glycosyl flavones, RHO and FOR, which are naturally present in kumquat fruits. In licorice, flavones are synthesized from 2-hydroxyflavanones generated by GeFNSII, which functions as an F2H (Akashi et al., 1998). However, this catalytic activity could not be assigned to FcFNSII-2 due to the lack of 2-hydroxyflavanones accumulation, both in vivo and in vitro.
Further studies are required to determine the involvement of FcFNSII-1 in flavonoid biosynthesis. In addition, prior to functional characterization, the expression of FcFNSII-1 in yeast and plants should be verified. The success of P450 expression depends on factors such as the expression vectors, posttranslational modifications, compatibility with the host, and coupling efficiency with CPR.
Most flavones in kumquat fruits are predominantly present as C-or O-glycosides. Flavone O-glycosylation occurs after the flavone backbones are generated by FNSI and/or FNSII. To synthesize C-glycosyl flavones, flavanones are first converted into 2-hydroxyflavanones by F2Hs, the open-circular forms of which are subsequently C-glycosylated by CGTs, followed by dehydration to generate the corresponding C-glycosyl flavones (Brazier-Hicks et al., 2009;Du et al., 2010b;Falcone Ferreyra et al., 2013;Ito et al., 2017). Thus, two catalytic mechanisms for FNSII enzymes might exist in kumquat plants. F2H (CYP93G2) and FNSII (CYP93G1) enzymes are present in rice (Du et al., 2010a;Lam et al., 2014). The former channels flavanones to 2hydroxyflavanones, which is a substrate for C-glycosylation by OsCGT (Brazier-Hicks et al., 2009;Du et al., 2010a). The latter converts flavanones to flavones for the biosynthesis of different tricin O-linked conjugates. CYP93G1 and CYP93G2 are key branch-point enzymes that compete for channeling flavanone substrates to flavones or 2-hydroxyflavanones, respectively. In this study, we observed such competition; the transient FcFNSII-2 overexpression significantly elevated the levels of O-glycosyl flavones (RHO and FOR) and reduced the level of flavonoid C-glycoside (DCGP) (Figure 4). An intriguing question elicited by our findings is why DCGP has a significant reduction, but not C-glycosyl flavones. We speculated that the DCGP content is much higher than that of C-glycosyl flavones in kumquat peels. Consequently, When FcFNSII-2 was overexpressed, the change in the amount of DCGP was easily discernable. However, owing to the low basal levels of C-glycosyl flavones, the changes in their levels were not obvious.
The protein-protein interaction studies (based on the SU-YTH system and BiFC assays, Figures 6, 7A and Supplementary Figure 6) demonstrated that ER-bound FcFNSII-2 was a component of the flavone metabolon and that it closely interacted with three upstream enzymes in flavonoid pathways, i.e., FcCHS1, FcCHS2, and FcCHIL ( Figure 7B). Additionally, considering that FcFNSII-2 was preferentially expressed in young fruit peels (Supplementary Figure 2C), we speculate that this ER-bound metabolon likely plays a key role in biosynthesizing fruit-specific O-glycosyl flavones.

CONCLUSION
We characterized a kumquat type II FNS (FcFNSII-2) that catalyzes the direct conversion of flavanones to flavones. In vivo and vitro assays showed that it could directly synthesize apigenin and acacetin from naringenin and isosakuranetin, respectively. Moreover, transient FcFNSII-2 overexpression enhanced the transcription of structural genes in the flavonoid-biosynthesis pathway and drove the accumulation of different O-glycosyl flavones that are naturally present in kumquats. Subcellularlocalization analyses and protein-protein interaction assays revealed that FcFNSII-2 was anchored to the ER and that it interacted with CHS1, CHS2, and CHIL. Our results provide strong evidence that FcFNSII-2 serves as a nucleation site for the O-glycosyl flavone metabolon that channels flavanones toward the biosynthesis of O-glycosyl flavones in kumquat fruits. These results would be useful for engineering the pathway to improve the composition of bioactive flavonoids in kumquat fruits.

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/Supplementary Material.

AUTHOR CONTRIBUTIONS
XGL and MZ designed the study. XGL, ST, and WX wrote the manuscript. ST, YY, CL, and XL performed the experiments. ST, YY, and XZ contributed to analyzing the results. All authors reviewed the manuscript.

ACKNOWLEDGMENTS
We would like to thank Editage (www.editage.cn) for English language editing.

SUPPLEMENTARY MATERIAL
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2022. 826780/full#supplementary-material The expression levels of FcFNSII-1 and FcFNSII-2 were calculated via the 2 − Ct method. FL, flowers; YL, young leaves; YS, young shoots; 30P, peels of fruits at 30 DAB; 90P, peels of fruits at 90 DAB; 150P, peels of fruits at 150 DAB. The data are presented as the mean ± SE of three independent replicates. Asterisks indicate significant differences ( * * P < 0.01).
Supplementary Figure 4 | Effects of pH and temperature on the enzyme activities of FcFNSII-2 proteins. (A) Effects of pH on enzyme activities at 30 • C for 1 h. Relative activities are presented as a percentage of the activity measured at pH 7.0 (100%). (B) Effects of temperature on enzyme activities at a pH of 7.0 for 1 h. Relative activities as a percentage of the activity measured at 30 • C. The data are presented as the mean ± SE of three independent replicates.