A Set of Regioselective O-Methyltransferases Gives Rise to the Complex Pattern of Methoxylated Flavones in Sweet Basil

Choice of Basil Lines for an Investigation of Flavone Biosynthesis

Extracts from four standard sweet basil lines (EMX-1, Sweet Dani [SD], SW, and MC; Gang et al., 2001; Iijima et al., 2004; Kapteyn et al., 2007; Xie et al., 2008) contained the major flavonoids known in this species (Grayer et al., 1996, 2001), with SALV and nevadensin (NEV; Fig. 1) being most abundant. Total flavonoid amounts and abundances of putative intermediates were significantly higher in lines SD and EMX-1 compared with SW and MC, while the overall profiles were not identical (Fig. 2A). Lines EMX-1 and SD, therefore, were chosen for further study.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Overview of flavones used as substrates and their positions in the metabolic network in basil, with a focus on steps elucidated in this paper. Color code: green (CIRL, EUP, ACA, cirsilineol [CIRLN], 8-OH-SALV, GARD B, NEV, PIL), accumulated but less relevant in this study; red (GENK, API-diMe, LAD, CIRM, SALV), accumulated in appreciable amounts and relevant to this study; blue (API, SCU7Me), key intermediates suggested by this work. Flavonoid backbone numbering and ring nomenclature are indicated on the API structure. For arrow codes, see key at top right. 7-O-Methylation of luteolin is not indicated for space reasons, but it is likely to occur.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Flavones in sweet basil lines and their localization. A, Representative UV₃₃₅ traces of extracts from four basil lines. Peak labels refer to the flavone designations in Figure 1. Peaks labeled RA, NA, and NB are rosmarinic acid and nepetoidins A and B, respectively. The key on the right indicates basil lines and fresh weight of ground tissue (sixth leaf pair from seven-leaf-pair stage plants) used for extraction using an identical procedure for all lines described in “Materials and Methods.” mAU, Milliabsorbance units. B and C, Relative abundances of major flavones in directly collected essential oil versus leaf extracts from lines SD (B) and EMX-1 (C). Total mass signal for the compounds shown was set as 100%. 8-OH-SALV and NEV were not separated and were quantified as one compound. Results are means ± se from five biological replicates.

A previous investigation indicated that flavonoids are stored in the subcuticular space of peltate glands in peppermint (Voirin et al., 1993). Therefore, we used stretched glass capillaries to collect the “oil” from the subcuticular space of sets of 150 randomly chosen peltate glands of an approximately 3-cm-long SD or EMX-1 leaf and compared the nonvolatiles in this oil with the flavone profile in the intact second leaf of the leaf pair (Fig. 2, B and C). Major flavonoids found in leaf extracts, such as the flavones SALV, NEV, gardenin B (GARD B), 8-hydroxysalvigenin (8-OH-SALV), cirsimaritin (CIRM), ladanein (LAD), and apigenin-7,4′-dimethyl ether (API-diMe; for structures, see Fig. 1), were easily detectable in these microextracts from trichomes. Given the small number of trichomes collected, the abundance of flavones in the oil seems very high, implying directed accumulation in this compartment (Supplemental Fig. S1, A–D). In comparison, rosmarinic acid, a relatively hydrophilic phenolic compound abundant in leaf extracts, was barely detectable in the collected oil and thus is likely to be equally distributed in the leaf tissue (Supplemental Fig. S1, E–H). The relative amounts of individual flavones in trichome extracts do not exactly mirror the leaf PMF profile (Fig. 2, B and C). Several consistent trends, such as lower relative abundances of pilosin (PIL; Fig. 1), CIRM, and LAD, could indicate a partitioning into a different location (e.g. secretory cells, leaf surface, or mesophyll) for specific PMFs. Alternatively, these discrepancies could be due to variable flavone blends in individual trichomes (e.g. according to their location on the leaf, as was observed for terpenes in peppermint; Maffei et al., 1989). However, random collection of five separate replicate sets of trichomes for each plant yielded highly consistent results and indicates that such variations are likely to be minor. Overall, our results suggest that the majority of the PMFs are primarily if not exclusively stored in the subcuticular oil of the trichomes. These findings show that the glandular trichome system is an appropriate tool for the investigation of the PMF metabolic network.

In Silico Analysis of Candidate Basil FOMTs

While both cation-dependent and non-cation-dependent OMTs (Noel et al., 2003) are represented by numerous contigs in our sweet basil EST database (Gang et al., 2001), the more abundant and diverse group, the non-Mg²⁺-dependent OMTs, was previously reported to encode FOMTs (Willits et al., 2004). BLAST searches using a peppermint F7OMT protein sequence revealed several candidate F7OMTs from basil. One of the best matches was subsequently used to perform a BLAST search against the EST database. The resulting group is composed of 15 contigs, three of them representing full-length transcripts. The two most abundant putative F7OMT contigs (termed ObFOMT1 and ObFOMT2 and shown in Fig. 3, which will be discussed further below) contain ESTs from all four basil lines, are among the most highly expressed genes in our libraries, and are 94% identical. The remaining 13 polypeptides encoded by minor contigs share 98% to 100% identity with one of the major F7OMTs. The identity of these proteins with the peppermint F7OMT amounted to less than 68%. For comparison, many methyltransferases with conserved functions, such as caffeic acid OMTs, often share more than 80% identity across the plant kingdom.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Similarity tree of plant OMTs. Relationships of basil class II OMTs (Ibrahim et al., 1998) to other published OMTs of this class were inferred using the neighbor-joining method (Saitou and Nei, 1987). Only bootstrap values greater than 70% are presented. Bootstrap values next to branch labels refer to the node at the base of a compressed subtree. Scale is in substitutions per amino acid. Branch labels of the type b xxxx (where xxxx indicates designations containing four digits) are the contig numbers in our basil EST database. The compressed COMTs (caffeic acid, or catechol OMTs) branch contains numerous published flavonoid 3′- and 3′/5′-OMTs. For protein designations and accession numbers, including those in selected compressed branches, see “Materials and Methods.” MEGA5 (Tamura et al., 2011) was used for tree construction. Outgroup JQ1393ODPO is an O-demethylpuromycin OMT from S. anulatus.

We followed an analogous strategy to identify the proteins engendering the 4′-O-methylations, yielding 20 contigs, with one of them full length and three missing just two to three amino acids at the N terminus. The putative F4′OMT subfamily is more diverse than the F7OMTs, with identities between individual proteins ranging from 87% to 100%. Again, the two most abundant contigs, termed ObFOMT3 and ObFOMT4 (Fig. 3), seemed to represent two clades of the F4′OMT subfamily. Whereas numerous minor contigs were more than 96% identical to one of the major contigs, two contigs stood out by having a lower level of identity to ObFOMT3 or ObFOMT4. The protein termed ObFOMT5 and encoded by a contig with 28 ESTs is 92% identical to ObFOMT3, whereas the protein ObFOMT6, encoded by a very minor (two ESTs) contig, is 92.6% identical to ObFOMT4 (Fig. 3; Supplemental Table S1). Because the activities and substrate preferences of non-Mg²⁺-dependent O-methyltransferases have been shown to switch dramatically by a change in just a few amino acids in several instances (Frick and Kutchan, 1999; Wang and Pichersky, 1999; Frick et al., 2001; Gang et al., 2002; Scalliet et al., 2008; Joe et al., 2010), we chose to study these six putative FOMTs in detail.

Functional Expression and Biochemical Characterization of F7OMTs

The recombinant proteins of all six FOMTs chosen for further characterization were produced in Escherichia coli. An N-terminally fused 6× His tag facilitated their purification to near homogeneity (Supplemental Fig. S2). Both basil F7OMTs accepted a broad range of substrates including flavonols such as kaempferol and quercetin, thus being similar to peppermint F7OMTs (Willits et al., 2004). Because basil is only known to accumulate lipophilic flavonoids lacking the 3-hydroxyl moiety (Grayer et al., 1996, 2001; Vieira et al., 2003), we collected physiologically relevant compounds (flavanones and flavones; for a list of trivial, abbreviated, and International Union of Pure and Applied Chemistry names of the flavonoids used in this study, see Supplemental Table S2), including some produced using our FOMTs, and offered them to both F7OMTs. As judged by relative activities at high flavone concentration (100 µm), both ObFOMT1 and ObFOMT2 prefer apigenin (API) and luteolin (LUT) to scutellarein (SCU; for structures, see Fig. 1; Table I). However, kinetic parameters revealed much higher catalytic efficiency (k_cat/K_m) of both F7OMTs with API than with LUT and even more so with SCU (Table II).

Within our substrate screen, we noted a striking difference between the relative activities with SCU, a fairly good substrate, and its 6-O-methylated derivative hispidulin (HISP; Fig. 1), which is barely methylated by both F7OMTs (Table I). These data are important for our mapping of metabolic routes in vivo, as they suggest that, at least with these two F7OMTs, 7-O-methylation has to precede the 6-O-methylation, and if the latter occurs before the former, the products are potentially diverted into a branch leading to compounds like NEV carrying a free 7-hydroxyl group.

The affinities of ObFOMT1 and ObFOMT2 were quite high for API and for the methyl donor, S-adenosylmethionine (SAM), as reflected by the low apparent K_m (Table II). The respective demethylated product, S-adenosylhomocysteine (SAH), inhibits the reaction by 32.2% (ObFOMT1) and 27.7% (ObFOMT2) when supplied at one-half concentration of SAM (100 µm). Competitive inhibition by SAH is a common feature of O-methyltransferases (Clarke and Banfield, 2001). Some product inhibition is also exerted by 10 μm genkwanin (GENK), which reduced the O-methylation of 10 μm API by 22.5% (ObFOMT1) and 7.1% (ObFOMT2). Both flavonoid 7-O-methyltransferases studied possessed very similar properties. The possibility of their temporal/spatial complementarity is addressed below. Because the intermediate occurrence of LUT, SCU, and above all API is very likely, it appears that the latter may serve as the natural substrate for both basil F7OMTs.

Functional Expression and Biochemical Characterization of F4′OMTs (F6/4′OMTs)

All common flavones were methylated by the putative F4′OMTs under long incubation times with excess protein (“max” reactions). However, analysis of products from O-methylation of SCU, the simplest 6-hydroxylated flavone, exposed distinct behavior between the four proteins. ObFOMT5 carried out both 6- and 4′-O-methylations, whereas ObFOMT3 mostly yielded HISP and smaller amounts of scutellarein-4′-methyl ether (SCU4′Me; Fig. 1). ObFOMT4 and ObFOMT6 only acted as 6-OMTs. In max reactions, three of the four proteins converted the monomethyl ethers of SCU into its 6,4′-dimethyl ether, pectolinarigenin (PECTOLI; Fig. 1). Only with ObFOMT4 did the reaction virtually stop after the first step (Supplemental Fig. S3A, traces 1, 3, 6, and 7).

Another substrate that the four F6/4′OMTs acted differentially upon was scutellarein-7-methyl ether (SCU7Me). Whereas ObFOMT5 yielded almost exclusively the 4′-O-methylation product LAD next to minuscule amounts of CIRM, ObFOMT3 almost immediately converted what little it made of CIRM into SALV. Therefore, the regioselectivity of these two FOMTs was altered compared with the products obtained with SCU. Longer incubation with SCU7Me resulted in complete 6- and 4′-O-methylation and thus SALV as a final product with ObFOMT3, ObFOMT5, and ObFOMT6. Time-course monitoring indicated that LAD was a better substrate for ObFOMT3 than for ObFOMT5, as the latter only started methylating LAD when very little SCU7Me was left. This is in line with the relative turnover rates of ObFOMT3 and ObFOMT5 with LAD (Table I). Only ObFOMT4 acted strictly as a 6-OMT and seemed unable to carry out the subsequent 4′-methylation of CIRM (Supplemental Fig. S3B). This distinct activity of highly similar proteins with SCU and SCU7Me clearly called for an investigation of structural elements causing the differences (see below).

Unexpectedly, all studied F6/4′OMTs are capable of methylating scutellarein-7-O-glucuronide (SCIN). This implies that their substrate-binding sites must have a way to accommodate the bulky sugar acid residue at position 7 rather than the smaller hydroxyl or a more lipophilic methoxyl group. All four F6/4′OMTs prefer a 6-OH residue as methyl group acceptor with SCIN (Supplemental Fig. S4). Subsequent 4′-O-methylation was detectable only with ObFOMT3 as catalyst, but at a negligible rate. If the 7-glucuronide (or similar glycoconjugate) indeed occurs as an intermediate in planta, the sugar moiety would have to be cleaved off after the 6-O-methylation in order for the 4′-methyl group to be introduced.

The final overview of relative activities with relevant collected substrates revealed that all four F6/4′OMTs performed best with partially methylated flavones (Table I). Strong and distinct preference of one acceptor moiety allowed the grouping together of ObFOMT4 and ObFOMT6 as predominantly 6-O-methyltransferases. On the other hand, ObFOMT3 and ObFOMT5 could be designated 4′-OMTs. Of the four proteins, ObFOMT4 has the most stringent regioselectivity, its 4′-O-methylating activity being negligible with all offered flavonoids except chrysoeriol (CHRYS; Fig. 1), an unlikely natural substrate.

Kinetic properties indicate LAD to be a better substrate than SCU7Me for ObFOMT4 and ObFOMT6 (Table II). The dimethylated substrate CIRM is also strongly preferred by ObFOMT3 over the monomethylated SCU7Me. Only ObFOMT5 kinetically prefers SCU7Me over CIRM. With sufficient SAM supply, the reactions easily proceed to full turnover, ruling out product inhibition. These properties imply that the abundance of SCU7Me might affect the relative abundances of three methylation products: CIRM, LAD, and SALV. As illustrated by Supplemental Figure S5, high amounts of SCU7Me favor higher relative abundances of CIRM and LAD when incubated with both ObFOMT3 and ObFOMT4. In the presence of lower concentrations of SCU7Me, CIRM and LAD outcompete it more easily and serve as preferred substrates, resulting in SALV as a major product.

An important regulatory mechanism to consider was inhibition of F6/4′OMTs by flavones that are poor or not substrates but are likely to occur in the cell. SALV was not a promising candidate because, as mentioned above, both ObFOMT4 and ObFOMT3 easily convert all supplied LAD and CIRM, respectively, into this common final product. Instead, we incubated ObFOMT4 with 5 μm LAD, its best substrate, and 5 μm CIRM, the flavone it produces from SCU7Me but does not methylate further. The yields of SALV only reached 73.7% of the uninhibited reaction, suggesting product inhibition by CIRM. This effect was not exerted by LAD in analogous experiments for ObFOMT3.

ObFOMT3 could also be the enzyme that forms API-diMe in vivo by 4′-methylation of GENK, even though it is not clear whether API-diMe originates from 4′- or 7-methylation as the first step (Table I). Kinetic analysis of this reaction revealed that GENK inhibits turnover at concentrations above 2.5 μm (substrate inhibition) and reaches half-maximal velocity at approximately 130 nm (Table II). ObFOMT5 behaves similarly, with substrate inhibition setting in at GENK concentrations above 10 μm.

A feature of all characterized F6/4′OMTs that distinguishes them from the F7OMTs and might prove important to physiological processes is the significantly lower affinity (relative to F7OMTs) for the cosubstrate, SAM, reflected by apparent K_m values between 21 and 78 µm (Table II). We tested whether the SAM/SAH balance would affect the product profile by monitoring reactions with SCU7Me as substrate for ObFOMT3 and ObFOMT4 at 0, 25, or 50 μm SAH in the presence of 400 μm SAM. The relative abundances of the three products do change substantially as SAH concentrations change (Supplemental Fig. S6). Not surprisingly, the second methylation step, which relies on products formed in the first, is affected more than the first methylation of directly supplied SCU7Me. While EST and proteomics analyses of glandular trichomes from young leaves (Xie et al., 2008) clearly showed that the SAM cycle enzymes are among the most abundant in basil glandular trichome secretory cells, the SAM/SAH ratio may well be different in trichomes from the older leaves used in this analysis.

As a result of substrate screening, SCU7Me emerged as a good substrate for all FOMTs we characterized that were not F7OMTs. SCU7Me is also the latest possible common precursor for CIRM and LAD. Based on kinetic properties of the putatively involved OMTs, the abundance of SCU7Me might affect the relative abundances of CIRM, LAD, and SALV (see above), making it a key intermediate in the metabolism of 5,6,7,4′-hydroxylated flavones in basil. Metabolic routes passing through SCU7Me as a precursor of all higher methylated flavones are highly favored (Table I). Conversely, the turnover rates are much lower in the reaction sequence beginning with the 4′-O-methylation of API or SCU and including 7-O-methylation as the second step. The 7-O-methylation activities with 6-methoxylated substrates represented by HISP and nepetin (NEP; Fig. 1) are close to negligible. Therefore, it seems likely that the formation of both LAD and CIRM should proceed via a 7-O-methylation as the first step, which can occur either on SCU or on API, which is the best substrate of the F7OMTs.

Flavone Methylation in Basil Trichome Protein Extracts

Only non-cation-dependent FOMTs have been reported to methylate positions 7 and 4′ (Schröder et al., 2004; Willits et al., 2004), whereas earlier published studies of flavonoid 6-OMTs identified responsible proteins as SAM- and Mg²⁺-dependent enzymes (De Luca and Ibrahim, 1985a; Ibdah et al., 2003). Such putative phenylpropanoid-flavonoid Mg²⁺-dependent OMTs (Ibdah et al., 2003) are quite abundant in our EST database. However, our results revealed that several recombinant non-Mg²⁺-dependent FOMTs can efficiently and selectively catalyze 6-O-methylation. To estimate the respective importance of these OMT classes in flavone metabolism, we compared the methylating activities of crude desalted trichome protein from lines SD and EMX-1 with and without Mg²⁺. Such protein extracts were expected to comprise all or most of the FOMT activities present in the glands.

In these assays (Fig. 4A), SD protein extracts showed higher specific activities with all tested substrates. Highest turnover was measured with SCU7Me, consistent with our finding that it serves as an excellent substrate both for 6- and 4′-OMTs. Only low activity was detectable with NEV and virtually none with PECTOLI (Fig. 4, A and B). These two flavones have free 7-OH but methylated 6- and 4′-OH groups and are not readily methylated by the two F7OMTs we characterized (Table I). These results demonstrated that we have not failed to identify a divergent F7OMT that is specialized to accept these flavones. Instead, there is indeed only very low methylating activity toward both NEV and PECTOLI present in trichomes. This explains why NEV can accumulate in the trichomes and excludes PECTOLI and NEV as precursors of SALV and GARD B, respectively (Fig. 1).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

Flavone O-methylation activities in crude desalted protein extracts from isolated trichomes. A, Overview of activities with crude trichome protein as catalyst. Results are means ± sd (n = 3). B, Activities in the presence of Mg²⁺ relative to activities without cation addition. Assays with four potential substrates (SCU, SCU4′Me, LAD, and SCU7Me) for flavone 6-O-methyltransferase activity were conducted to check for Mg²⁺-dependent OMTs and those with PECTOLI/NEV were conducted to demonstrate that F7OMT activities with these substrates are negligible with and without cation. Error bars indicate sd (n = 3). C and D, Methylation products in crude desalted protein extracts in the presence of Mg²⁺ relative to their abundances without cation addition and SCU (C) or SCU7Me (D) as substrate. Error bars indicate sd (n = 2). *P < 0.05 (Student’s unpaired two-tailed t test).

To probe the effect of Mg²⁺ on 6-O-methylation by trichome proteins, we tested four potential substrates, SCU, SCU7Me, SCU4′Me, and LAD. Neither the total turnover rates (Fig. 4B) nor the formation rates of individual products with SCU and SCU7Me (Fig. 4, C and D) supported significant involvement of Mg²⁺-dependent OMTs in 6-O-methylation. In fact, several reactions were inhibited by Mg²⁺ when crude protein from trichomes was used in assays but not when recombinant individual FOMTs were tested in vitro (Supplemental Table S3). This effect is likely due to other FOMTs present in the protein preparations used. Importantly, a third flavone with mass-to-charge ratio 315 in positive ionization mode was formed by crude protein extracts with SCU7Me as substrate. Judging by a previous report (Greenham et al., 2003) and considering our knowledge about other dimethylated SCU derivatives, it is likely to be SCU5,7-dimethyl ether, a compound never reported to occur in basil. This product was barely detectable in assays without Mg²⁺ and increased by over 700% upon addition of the cation, implying that it was formed by an Mg²⁺-dependent OMT (Fig. 4D). The detection of this activity fortuitously provides an endogenous control for the inactivity of Mg²⁺-dependent OMTs in assays lacking externally added cation. It also implies that the involvement of the latter OMT class in 6-O-methylation is negligible in basil. This reaction thus presents another example of convergent evolution in plants, where the same regioselectivity is exhibited by proteins belonging to two different enzyme families.

The Basis for the Distinct Regioselectivities of F6/4′OMTs

The high identity of the four ObF6/4′OMT proteins (see above) dictates that their differential regioselectivity is determined by differences in just a few specific amino acid residues. Therefore, these proteins offer a unique opportunity to study structural elements that define the regiospecificity and substrate specificity within a group of closely related enzymes.

The crystal structures of several non-Mg²⁺-dependent OMTs have been solved (Zubieta et al., 2001, 2002; Liu et al., 2006; Louie et al., 2010), and all have similar overall fold and catalytic mechanisms (Noel et al., 2003). Basil FOMTs share the highest identity (30%–32%) and similarity (approximately 55%) to isoflavone 7-O-methyltransferase from alfalfa (Medicago sativa; IOMT [Protein Data Bank accession 1FP2]; Zubieta et al., 2001) compared with other OMTs with available structure data (Supplemental Fig. S7). Consequently, we used IOMT as a template for homology model construction.

The active sites of ObFOMT3 and ObFOMT4, as predicted by modeling, are slit shaped yet spacious enough to explain the acceptance of several different substrates, especially of planar molecules like the flavones (Fig. 5). As has been found for all OMTs of this class, the N terminus of the dyad-related monomer contributes to shaping the back wall of the substrate-binding site. In the case of basil FOMTs, the hydrophobic residues Trp-14 and Leu-18 will potentially accommodate the 7-OMe residue of SCU7Me when it is positioned for 4′-O-methylation. In addition, Leu-289 could contribute to the hydrophobic pocket. Tyr-20 can form an H-bond with the 5-OH of the flavone backbone in ObFOMT3. His-241 is very likely to act as the catalytic base (Zubieta et al., 2001, 2002). The overall shape of the active sites is identical in the three proteins (ObFOMT3–ObFOMT5) that we were most interested in.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

Active sites of ObFOMT3 (A) and ObFOMT4 (B) modeled with SAM and SCU7Me as ligands. Walleye views are presented. The SCU7Me ligands were placed in a position consistent with the observed activity before running Modeller. O4′ was positioned facing SAM in ObFOMT3, and O6 was positioned facing SAM in ObFOMT4. Backbone bonds for the chain in A are colored cyan, and backbone bonds for the chain in B are colored orange. Amino acid side chains are labeled where space allows.

One of our goals was to find a structural explanation for the stringent regioselectivity of ObFOMT4, whose activity is almost limited to 6-O-methylation, as opposed to ObFOMT3, which can act as either a 6- or 4′-OMT, depending on the substrate. The difference is most evident with SCU7Me, a potential key intermediate of flavone metabolism. ObFOMT3 preferentially acts as a 4′-OMT upon this substrate, while ObFOMT4 efficiently catalyzes only its 6-O-methylation. For the reactions to yield their respective products, SCU7Me has to be bound in opposite orientations. An analysis of the amino acid residues suggested by the models to line the substrate-binding site pointed out two residues of interest: Met-296 and Thr-292 in ObFOMT4, corresponding to Val-296 and Ile-292 in ObFOMT3 (Fig. 5). Met-296 of ObFOMT4 might extend into the binding cavity, leaving less space for the C-ring of flavones bound for 4′-O-methylation, thereby reducing the 4′-OMT activity. The mode of its interference with substrate binding could not be deduced from the model, and detailed crystallographic investigations will be required in order to determine just how Met-296 might be involved in restricting 4′-O-methylation. Thr-292 is positioned on the diagonally opposite side across from His-241, the probable catalytic base that deprotonates the acceptor hydroxyl group. This Thr can form a hydrogen bond with a 4′-OH group of SCU7Me or SCU, stabilizing its binding orientation for 6-O-methylation.

Interestingly, our EST database contains several minor contigs encoding the C-terminal OMT fragment that differs from ObFOMT3 by just this I292T mutation (contigs basil_0688, basil_0489, and basil_0260). To assess the effect of this single amino acid substitution, we generated a single mutant of ObFOMT3, I292T. Its relative activities with our flavonoid substrate set indicated that the mutant was very similar to ObFOMT3 (Table I). Only the turnover with naringenin-7-methyl ether (NAR7Me) was significantly higher. However, LAD’s portion of the total products decreases from 90% in ObFOMT3 to 70% in this single mutant (Fig. 6A, trace 2). Due to the high affinity of ObFOMT3 and presumably of its I292T mutant, the formed CIRM is rapidly converted into SALV as soon as its concentration in the assay is high enough to outcompete SCU7Me. Analogous behavior was displayed by the ObFOMT3 I292V mutant, confirming the effect of residue 292 on substrate orientation (76% LAD; Fig. 6A, trace 3).

• Download figure
• Open in new tab
• Download powerpoint

Figure 6.

Identification of residues involved in the modulation of regioselectivities and therefore functional differentiation of basil FOMTs. Proteins and mutations are indicated on the left, and trace numbers as used in the text are on the right above the baseline. A, Effects of mutations on the 6- or 4′-O-regioselectivity of ObFOMT3 and ObFOMT4 with SCU7Me as substrate. B, Effect of mutations on the 6- or 4′-O-regioselectivity of ObFOMT3 and ObFOMT5 with SCU as substrate. [See online article for color version of this figure.]

The single I292T mutation did not affect the high activity of the protein with CIRM while, nevertheless, affecting the regioselectivity (Table I). In contrast, the double mutant of ObFOMT3, I292T/V296M, methylated CIRM at a much lower rate, in agreement with the steric inhibitory effect ascribed to Met-296, and the shift toward SCU7Me 6-O-methylation was even more pronounced (Fig. 6A, trace 4). The two divergent Thr residues following Met-296 in ObFOMT4 were the next mutation target. The triple mutant of ObFOMT3, I292T/V296M/M297T, yielded almost exclusively CIRM as product (Fig. 6A). Because the reciprocal triple mutant of ObFOMT4 did not produce significantly more LAD, we also carried out the fourth mutation, T298A in ObFOMT4 and the reverse in ObFOMT3. The product profile did not change visibly for ObFOMT3, but detectable amounts of LAD were formed by the ObFOMT4 mutant (Fig. 6A, traces 6 and 7). The fact that the activity with LAD of ObFOMT4 mutants did not decrease but remained comparable to that with SCU7Me as substrate (Supplemental Table S4) suggests that SALV formation by ObFOMT4 triple and quadruple mutants (visible in traces 8 and 7 of Fig. 6A) is due to the methylation of LAD produced in the first step from SCU7Me. In contrast, the reaction velocities with LAD increase gradually relative to those with SCU7Me as more mutations are introduced in ObFOMT3, reaching a maximum in the quadruple mutant (Supplemental Table S4).

The introduced mutations reduced reaction rates of ObFOMT3 and ObFOMT4 with SCU7Me. The activity of ObFOMT3 and ObFOMT4 quadruple mutants amounted to 1.71 and 0.75 nkat mg⁻¹, respectively, thus being moderately or strongly decreased as compared with wild-type enzymes (Table I). Yet, the turnover rate of the ObFOMT3 quadruple mutant with LAD reached up to 3.08 nkat mg⁻¹, which exceeds the highest turnover rates found in ObFOMT3.

The second apparent difference between the F6/4′OMTs was illustrated by the product profiles with SCU (Supplemental Fig. S3A). It was most interesting to identify the structural elements that modulate the methylation of SCU by ObFOMT3 and ObFOMT5 toward dominance of HISP and SCU4′Me as respective products. The predicted flavone-binding sites of these two OMTs are conserved. This suggested that the different behaviors of these two proteins were caused by indirect effects on the proteins’ conformation. The importance of the subunit interface in determining the shape of the substrate-binding site is a known feature of Mg²⁺-independent OMTs (Noel et al., 2003). The models pointed out two residues (109 and 111) that are divergent in ObFOMT3 and ObFOMT5 and are also likely to affect the interaction between the subunits of the homodimer. Indeed, the product ratio of the C109S single mutant of ObFOMT5, and even more so of the C109S/T111A double mutant, shifted toward 6-methylation (Fig. 6B; Supplemental Table S5). Reciprocal mutations in corresponding positions of ObFOMT3 resulted in significantly increased formation of SCU4′Me.

Another divergent residue we tested was Ser-181 (Gly-181 in ObFOMT3), which, being part of the conserved signature SAM-binding motif (Ibrahim et al., 1998), could affect SAM binding and thus the shape of the substrate-binding pocket. We generated triple mutants for both ObFOMT5 and ObFOMT3. A comparison of the full substrate profiles for these triple mutants (positions 109, 111, and 181) revealed that their patterns became nearly identical to that of the reciprocal OMT (Table I). Most striking was the effect on the turnover with CIRM relative to SCU7Me, which was strongly increased in the ObFOMT5 C109S/T111A/S181G mutant and reduced in the reciprocal ObFOMT3 mutant. At the same time, the product profiles with SCU did not fully interchange. Both criteria (product profile with SCU and relative activities with SCU7Me/CIRM) indicate that residues 109 and 111 are primarily responsible for the achieved effects (Fig. 6B; Supplemental Table S5).

In summary, homology modeling-guided site-directed mutagenesis allowed us to identify important residues that strongly affect the 6- or 4′-regioselectivity by two distinct mechanisms: either by direct interaction in the substrate-binding site, as shown for the ObFOMT3 and ObFOMT4 comparison, or by indirectly modulating the shape of the active site, as in the case of the ObFOMT3 and ObFOMT5 comparison. In the latter case, the exact nature of the interactions cannot be predicted using modeling. Real crystallographic data will be necessary to reveal the actual substrate-binding modes in individual cases.

Flavone Profiling and FOMT Expression over a Developmental Time Course

To assess whether the FOMTs described above play the proposed roles in vivo, we evaluated the correlation between accumulated metabolites and individual gene expression. Analysis of directly collected oil suggested that the flavones are predominantly localized to the subcuticular space of the peltate glands. However, transcript profiling required RNA isolation, which is not feasible using the glass capillary sampling technique. Therefore, we used whole leaf tissue as the source for metabolite and nucleic acid extraction. To recognize existing trends, we collected true leaf pairs of five plants of each line when they reached three-leaf-pair, five-leaf-pair, and seven-leaf-pair developmental stages. The youngest leaf pair (fourth, sixth, and eighth, respectively) was not analyzed because it would not yield enough material for both analyses.

The total flavone patterns are fairly conserved over the monitored time periods but exhibit some differences between the two basil lines (Fig. 7; Supplemental Fig. S8). For example, the concentrations of individual compounds increase steadily toward the youngest leaf in SD plants, while in EMX-1, the youngest leaf contains less flavones per gram fresh weight than the second youngest. In both lines, SALV is the major 8-unsubstituted flavone. Importantly, the amounts of SALV’s precursors CIRM and LAD relative to total flavone accumulation were higher in EMX-1 (11%–18%) compared with SD (4%–11%) leaves; moreover, the content of LAD relative to CIRM was 7% to 12% higher in EMX-1. The relative concentration of API-diMe was nearly doubled in SD compared with EMX-1 but was generally less than 10% of the total flavone content. GENK is not shown in the diagrams but can be detected in most samples in slightly lower amounts than API-diMe. A peak with mass signal of mass-to-charge ratio 301 eluting at the retention time of SCU7Me, the preferred substrate for major basil F6/4′OMTs, is present in amounts of up to 1.5 μg g⁻¹ fresh weight in numerous but not all samples. Evaluation of relative flavone amounts in successive leaves revealed a consistent increase toward the youngest leaves in both basil lines in SALV’s proportion of the total flavone content (e.g. from 24.2% in the first leaf pair to 33.5% in the seventh leaf pair in line SD, and from 28.2% in the oldest leaf pair to 33.5% in the seventh leaf pair in EMX-1; Fig. 7). The older SD leaves contained more CIRM and LAD as a percentage of total flavone content (9.48% and 3.61%) than younger leaves (4.85% and 1.66%). The latter trend is not pronounced in EMX-1.

• Download figure
• Open in new tab
• Download powerpoint

Figure 7.

Accumulation of selected flavones in basil lines SD (A) and EMX-1 (B) at the seven-leaf-pair growth stage. The key indicates leaf number (1 is the oldest). Results are means of five biological replicates ± se. [See online article for color version of this figure.]

The 8-substituted flavones detected were GARD B, PIL, NEV, and 8-OH-SALV. The relative concentration of GARD B was highest in youngest leaves, whereas PIL accumulation varied strongly between replicates. NEV and 8-OH-SALV were quantified as one compound due to separation difficulties. Their relative concentration increased by 11% to 15% toward older leaves in both basil lines. The opposite shifts in relative abundances of SALV and NEV/8-OH-SALV occur as the leaf expands and thus depend on leaf age. These shifts will probably be explained when we elucidate the underlying biosynthetic steps.

A comparison of corresponding leaf positions over the developmental time course suggests that total flavone content in mature leaves reaches a certain maximum after which production ceases. For example, the oldest leaves of EMX-1 and SD contain approximately 47 and 30 μg flavones g⁻¹ fresh weight when the plants have five leaf pairs, which compares very well with 45 and 25 μg g⁻¹ found in plants with seven leaf pairs (Fig. 7; Supplemental Fig. S8). Except for the youngest EMX-1 leaf pair, the flavone concentrations decrease in expanding leaves that are gaining fresh weight. For example, the flavone content in the fourth leaf pair of EMX-1 and SD decreases from 173 and 209 μg g⁻¹ fresh weight at the five-leaf-pair stage to 144 and 149 μg g⁻¹ at the seven-leaf-pair stage, respectively. Exclusive biosynthesis and accumulation of flavones in peltate glands that stop metabolic activity after their maturation provide the best explanation for this observation.

In order to estimate whether the studied FOMTs may indeed contribute to the differences in flavone profiles between the two basil lines, as well as between leaves of different ages in agreement with their biochemical properties, we analyzed their relative expression levels using quantitative reverse transcription-PCR in all leaf pairs of seven-leaf-pair stage plants. We were looking for subtle differences, since the flavone profiles are overall similar. Clearly, even though the same tissue was used for metabolite and gene expression analyses, the results of these experiments should be interpreted with caution, as transcript levels for selected genes may not reflect the overall enzymatic activities present in the glands. ObFOMT6, which is encoded by only two ESTs and has essentially the same activity as ObFOMT4, was not analyzed. The trends in FOMT expression were compared with those of flavone synthase (FNS), a core flavone metabolism gene whose expression in nontrichome leaf tissue is expected to be low to negligible (Grayer et al., 2002). Chalcone synthase, the first committed enzyme of the flavonoid pathway, would be less suitable, as it is involved in leaf flavonoid glycoside biosynthesis. FNS transcript levels were highest in the youngest leaves and reached comparable levels in both lines (Fig. 8). The absence of correlation with the much lower flavone content of the youngest leaf pair of line EMX-1 (Fig. 7) suggests that FNS expression is not the key limiting factor for flavone accumulation. Indeed, the level of chalcone synthase expression/activity may exert control over the total flux (Knogge et al., 1986; Morreel et al., 2006). As mentioned above, monitoring the chalcone synthase transcripts in our experimental setup did not appear appropriate.

• Download figure
• Open in new tab
• Download powerpoint

Figure 8.

Transcript levels of ObFOMT1 to -5 and FNS in all leaf pairs of SD (A) and EMX-1 (B) plants at the seven-leaf-pair stage. Results are means of five biological replicates ± se. For normalization and scaling, see “Materials and Methods.” [See online article for color version of this figure.]

The expression levels of all five methyltransferases decreased as the leaves aged in both lines. Between the sixth and seventh leaf pair in line SD, the difference was 2- to 2.5-fold. This difference was even more pronounced in EMX-1 leaves, ranging between 3- and 5-fold. This result was not altogether surprising, since, as discussed above, the total amount of flavones is expected to be determined by the activity of core flavonoid pathway proteins. Overall, both FOMT and FNS expression levels drop dramatically as leaves get older. In agreement with metabolite accumulation data, this strongly implies that the localization of the external flavone production machinery is restricted to peltate glands, whose formation and maturation were earlier found to be completed long before the leaf reached its final size in basil (Werker et al., 1993).

The abundance of individual FOMT transcripts was generally higher in young SD leaves compared with EMX-1, except for ObFOMT5, whose expression levels were similar in both lines. The expression profiles of both F7OMTs that appear to be biochemically redundant did not provide any evidence of their specialization or complementary function. Interestingly, the differences between the F7OMT abundances in corresponding leaf pairs of the two lines were consistently lower (1.3- to 2.8-fold) than the differences in relative abundances of ObFOMT3 (2.4- to 6.4-fold) and ObFOMT4 (2.0- to 5.2-fold). Lower relative expression of the two major 6/4′-FOMTs in EMX-1 should favor higher relative amounts of LAD and CIRM in this line, agreeing with the metabolite data. The gene expression data also supported higher relative accumulation of LAD by line EMX-1 than by SD due to the already mentioned comparable expression level of ObFOMT5 and significantly lower abundances of ObFOMT3 and ObFOMT4. Whereas in SD, ObFOMT5 might be a less important player, in EMX-1, it is likely to contribute more noticeably to the conversion of SCU7Me into LAD. At the same time, the lower expression of ObFOMT4 in the same tissue results in slower turnover of LAD into SALV. In summary, keeping in mind the caveat stated above, the differences in metabolite profile appear to correlate with differential expression of appropriate FOMTs, which can be interpreted as strong evidence in support of their designated physiological functions.