Complete Biosynthesis of the Anti-Diabetic Plant Metabolite Montbretin A.

Diabetes and obesity are affecting human health worldwide. Their occurrence is increasing with lifestyle choices, globalization of food systems, and economic development. The specialized plant metabolite montbretin A (MbA) is being developed as an antidiabetes and antiobesity treatment due to its potent and specific inhibition of the human pancreatic α-amylase. MbA is a complex acylated flavonol glycoside formed in small amounts in montbretia (Crocosmia × crocosmiiflora) corms during the early summer. The spatial and temporal patterns of MbA accumulation limit its supply for drug development and application. We are exploring MbA biosynthesis to enable metabolic engineering of this rare and valuable compound. Genes and enzymes for the first four steps of MbA biosynthesis, starting from the flavonol precursor myricetin, have recently been identified. Here, we describe the gene discovery and functional characterization of the final two enzymes of MbA biosynthesis. The UDP-glycosyltransferases, CcUGT4 and CcUGT5, catalyze consecutive reactions in the formation of the disaccharide moiety at the 4'-hydroxy position of the MbA flavonol core. CcUGT4 is a flavonol glycoside 4'-O-xylosyltransferase that acts on the second to last intermediate (MbA-XR2) in the pathway. CcUGT5 is a flavonol glycoside 1,4-rhamnosyltransferase that converts the final intermediate (MbA-R2) to complete the MbA molecule. Both enzymes belong to the UGT family d-clade and are specific for flavonol glycosides and their respective sugar donors. This study concludes the discovery of the MbA biosynthetic pathway and provides the complete set of genes to engineer MbA biosynthesis. We demonstrate successful reconstruction of MbA biosynthesis in Nicotiana benthamiana.


Introduction
The formation of the disaccharide moiety at the 4'-hydroxy of myricetin defines  (Ross et al., 2001;Caputi et al., 2012;Wilson and Tian, 2019). 131 The UGT family contains members in several clades that catalyze the glycosylation 132 of flavonoids at various hydroxyl groups (Caputi et al., 2012), as well as UGTs that 133 catalyze glycosylation of the sugar moiety of glycosides (Richman et al., 2005;134 Frydman et al., 2013;Yonekura-Sakakibara et al., 2014). The vast majority of 135 previously characterized UGTs use UDP-glucose, but some UGTs accepting UDP-136 galactose, UDP-rhamnose or UDP-xylose have also been reported (Jones et al., 137 2003;McCue et al., 2007;Shibuya et al., 2010;Yonekura-Sakakibara et al., 2012; 138 Sayama et al., 2012;Itkin et al., 2013). peak during the spring and early summer in developing yC followed by a decline 157 towards autumn, and are low in old corms (oC) of the previous growing season 158 (Irmisch et al., 2018;Irmisch et al., 2019a;Irmisch et al., 2019b). To identify the 159 remaining UGTs required for the completion of MbA biosynthesis, we explored the  Figure 1B). 182 We validated these six UGT transcripts for presence and differential 183 expression (DE) in the previously described contrasting transcriptomes that 184 compared yC and oC gene expression (Irmisch et al., 2018). Three of these six 185 candidates, UGT703G1, UGT703F1 and UGT709R1, were present as full-length 186 sequences with 36.8-fold, 9.9-fold and 10.2-fold higher transcript abundance, 187 respectively, in yC compared to oC. Two UGTs, UGT729A2 and UGT703E4, were non-full length. UGT729A2 was missing a sequence for 84 aa at the predicted UGT 189 C-terminus, and UGT703E4 was only detected as a short fragment encoding for 250 190 aa. UGT729A2 and UGT703E4, showed 90.7-fold and 2.8-fold higher transcript 191 levels in yC compared to oC, respectively. Unexpectedly, we were initially not able to 192 detect the sixth candidate, UGT703H1, in the yC/oC transcriptome dataset. However,193 upon closer inspection of the yC/oC transcriptome data, we identified a contig of ORFs on one contig may be due to in silico misassembly. The shorter of the two 197 ORFs matched UGT703H1. We had initially missed this contig in our data analysis, 198 which was trained to only select for the longest ORF on any given contig.  Table S1). All six proteins possess a glutamine in the C-terminal 208 position of the PSPG-motif. The amino acid in this position has been described to 209 affect sugar donor specificity (Supplemental Figure S1A) (Kubo et al., 2004;Ono et 210 al., 2010).

211
To assess candidate UGTs for glycosyltransferase activity, we expressed the  When incubated with MbA-XR 2 (peak 1) and UDP-Xyl, protein extracts 221 containing UGT703H1 showed the formation of a single product (peak 2) with m/z 1081.5 identified as MbA-R 2 based on matching retention time and fragmentation 223 pattern with an authentic standard (Figure 3A and C; Supplemental Figure S2A). 224 The fragmentation pattern of MbA-R 2 shows the initial loss of Xyl (loss of 132, m/z 225 949), indicative for the attachment of Xyl to the 4'-hydroxy-position of the flavonol B-226 ring. Protein extract containing UGT729A2 converted MbA-R 2 (peak 2) and UDP-Rha  Figure S1). Taken together, these activity screens identified UGT703H1 (CcUGT4) 238 and UGT729A2 (CcUGT5) as the enzymes that catalyze the final two glycosylation 239 steps in MbA biosynthesis.

240
For further characterization, CcUGT4 (UGT703H1) and CcUGT5 (UGT729A2) 241 proteins were Ni-purified. We tested CcUGT4 for substrate specificity with UDP-Xyl 242 as the sugar donor and different sugar acceptors, specifically the MbA pathway 243 precursor and intermediates (myricetin, MR, MRG, mini-MbA, MbA-XR 2 ) as well as 244 quercetin, kaempferol and caffeic acid (Supplemental Table S2). Among 245 intermediates in MbA biosynthesis, in addition to MbA-XR 2 , CcUGT4 was also active 246 with MRG and mini-MbA but not with myricetin or MR. However, product formation 247 with MRG and mini-MbA was only 0.6% and 5.2%, respectively, relative to product 248 formation with MbA-XR 2 as substrate (Supplemental Figure S3B and C). No activity 249 was detected with any of the other acceptor substrates tested (Supplemental Table   250 S2). CcUGT4 was specific for UDP-Xyl as the sugar donor and did not accept UDP-251 Rha or UDP-Glc when MbA-XR 2 was used as the acceptor (Supplemental Figure   252 S3A). We tested CcUGT5 for substrate specificity with UDP-Rha as the sugar donor  Table S2). In addition to MbA-R 2 , CcUGT5 was active with MbA-CR 2 (which is MbA-R 2 missing the caffeoyl moiety) as a substrate. Here, a 257 single product peak m/z 1065 was observed and identified as MbA-C (MbA missing 258 the caffeoyl moiety) (Supplemental Figure 4A and C). Quercetin 4'-O-glucoside 259 (spiraeoside) did not serve as a substrate for CcUGT5 (Supplemental Table S2). In 260 addition to UDP-Rha, CcUGT5 also accepted UDP-Xyl but not UDP-Glc as a sugar 261 donor with MbA-R 2 as acceptor (Supplemental Figure S4B). However, product 262 formation using UDP-Xyl and MbA-R 2 was only 0.7% relative to product formation 263 with UDP-Rha, suggesting that UDP-Rha was the preferred sugar donor substrate for 264 CcUGT5 (Supplemental Figure S4B and C).    Figure S5).

301
In agreement with our previous work, leaves expressing myricetin and MbA-302 XR 2 biosynthesis genes produced MbA-XR 2 (m/z 949, peak 4 in Figure 5) and MbB-  Table S4). Peak areas corresponding to these products decreased 305 when CcUGT4 alone or CcUGT4 in combination with CcUGT5 were co-expressed 306 with genes for myricetin and MbA-XR 2 biosynthesis, indicating substrate conversion.

585
To test for UGT activity, initial enzyme assays were performed with 100 μL of the MbA-XR 2 , MbA-R and MbA (Williams et al., 2015).  Table S6). Primer  To test for significant differences in CcUGT4 and CcUGT5 transcript abundance in 701 yC and oC at different time points, data were log transformed to meet statistical 702 requirements and a two-way analysis of variance (ANOVA) was performed followed 703 by a Tukey-Test using SigmaPlot 11.0 for Windows (Systat Software Inc. 2008)