In vitro reconstruction and analysis of evolutionary variation of the tomato acylsucrose metabolic network

Significance

Throughout the course of human history, plant-derived natural products have been used in medicines, in cooking, as pest control agents, and in rituals of cultural importance. Plants produce rapidly diversifying specialized metabolites as protective agents and to mediate interactions with beneficial organisms. In vitro reconstruction of the cultivated tomato insect protective acylsucrose biosynthetic network showed that four acyltransferase enzymes are sufficient to produce the full set of naturally occurring compounds. This system enabled identification of simple changes in enzyme structure leading to much of the acylsucrose diversity produced in epidermal trichomes of wild tomato. These findings will enable analysis of trichome specialized metabolites throughout the Solanaceae and demonstrate the feasibility of engineering these metabolites in plants and microorganisms.

Abstract

Plant glandular secreting trichomes are epidermal protuberances that produce structurally diverse specialized metabolites, including medically important compounds. Trichomes of many plants in the nightshade family (Solanaceae) produce O-acylsugars, and in cultivated and wild tomatoes these are mixtures of aliphatic esters of sucrose and glucose of varying structures and quantities documented to contribute to insect defense. We characterized the first two enzymes of acylsucrose biosynthesis in the cultivated tomato Solanum lycopersicum. These are type I/IV trichome-expressed BAHD acyltransferases encoded by Solyc12g006330─or S. lycopersicum acylsucrose acyltransferase 1 (Sl-ASAT1)─and Solyc04g012020 (Sl-ASAT2). These enzymes were used—in concert with two previously identified BAHD acyltransferases—to reconstruct the entire cultivated tomato acylsucrose biosynthetic pathway in vitro using sucrose and acyl-CoA substrates. Comparative genomics and biochemical analysis of ASAT enzymes were combined with in vitro mutagenesis to identify amino acids that influence CoA ester substrate specificity and contribute to differences in types of acylsucroses that accumulate in cultivated and wild tomato species. This work demonstrates the feasibility of the metabolic engineering of these insecticidal metabolites in plants and microbes.

Plants are masters of metabolism, producing hundreds of thousands of small molecules known as specialized metabolites, which vary widely in structure, abundance, and physical and biological properties. These metabolites tend to be produced by enzymes that evolve faster than those that produce “central” metabolites such as amino acids, nucleotides, sugars, and cofactors (1–3), and the pathways and metabolic intermediates involved in biosynthesis of many specialized metabolites remain mysterious. Despite the growing availability of genomic DNA sequences, understanding the genetic and biochemical mechanisms that contribute to this phenotypic diversity and plasticity presents enduring and major challenges in plant biochemistry. It is of great interest to understand and manipulate the biosynthesis of these biologically active molecules.

Specialized metabolites typically are produced in a cell- or tissue-specific manner and are generally limited in their taxonomic distribution. Glandular secreting trichomes provide an example of such a differentiated structure; these epidermal “hairs” produce a variety of metabolites of importance to humans, including aromatic flavor components (e.g., in hops for beer and Mediterranean herbs for cooking), psychoactive cannabinoids in Cannabis, and the antimalarial drug artemisinin in Artemisia annua (4, 5).

Some trichome-produced metabolites have documented direct and indirect antiherbivore activities (4, 6–8). For example, acylsugars are a group of structurally related specialized metabolites produced in plants of the nightshade family—the Solanaceae (9, 10). Characterized examples in the tomato group of Solanum consist of either a glucose or a sucrose backbone with three to four aliphatic acyl groups of varying carbon numbers ranging from 2 to 12 esterified to the sugar hydroxyl groups (11–15). Nicotiana attenuata acylsucroses are at the center of a multitrophic defense interaction where they are metabolized to volatile fatty acids by Manduca sexta larvae, and these airborne products attract predatory ants (6). Protective properties against herbivores have made increasing total acylsugars or altering acyl chain types a target for breeding insect-resistant cultivated tomatoes (Solanum lycopersicum) (16–18). In addition, synthetic sucrose esters mimicking natural acylsugars have been applied as safe, biodegradable insecticidal compounds (19–21) and also have commercial value in the food, cosmetic, and pharmaceutical industries (22, 23).

Recent work on acylsucrose biosynthesis in trichomes of the tomato clade revealed that this relatively closely related group of plants produces a surprisingly diverse group of acylsucroses in tip cells of the long hair-like type I/IV trichomes (14, 24–26). The genetic and biochemical mechanisms underlying some of this phenotypic diversity have begun to be revealed. For example, 2-methylpropanoic acid (iC4) and 3-methylbutanoic acid (iC5) acyl chain variation in Solanum pennellii accessions is influenced by variation in the activity of a set of truncated isopropylmalate synthase 3 (IPMS3) enzymes (26). Differences in patterns of Solanum habrochaites acylsucrose acetylation (24) and acyl chain length and position variation (25) are due to genetic variation in two BAHD [BEAT, AHCT, HCBT, DAT (27, 28)] acyltransferases. Solanum lycopersicum AcylSugar AcylTransferase 3 (Sl-ASAT3) catalyzes acylation of diacylsucroses on the five-membered (furanose) ring to make triacylsucroses (25). Variant forms of ASAT3 were described in wild tomato accessions: these use different chain length acyl-CoA esters with diacylsucrose, or acylate the six-membered (pyranose) ring of monoacylated sucrose, demonstrating recent evolutionary changes leading to diversification of enzymatic activity. The phylogenetically related ASAT4 enzyme (formerly AT2) is responsible for making tetraacylsucroses in S. lycopersicum and S. habrochaites trichomes by acetylating triacylsucroses using acetyl-CoA (C2-CoA) (29). A variety of loss-of-function alleles of this enzyme are found in populations of S. habrochaites from northern Peru and Ecuador (24), reinforcing the idea that ASAT diversification plays an important role in shaping the strong phenotypic diversity in trichomes of wild tomato. Although variation in these enzymes influences the acylsugar phenotypes seen in cultivated and wild tomatoes, much of the diversity remains to be explained at a molecular level, including the order of acysugar assembly. In addition, identification of the enzymes that produce the ASAT3 substrates is needed to understand cultivated and wild tomato acylsucrose biosynthesis in greater detail.

In this study, two trichome-specific S. lycopersicum BAHD acyltransferases that catalyze consecutive reactions to produce these diacylsucrose intermediates were identified. Sl-ASAT1 was found to use sucrose and various acyl-CoAs to make monoacylsucroses with pyranose R₄ acylation. Sl-ASAT2 adds a second acyl chain to the R₃ position of the ASAT1 S1:5 (iC5^R4) product (note that “S” refers to a sucrose backbone, “1:5” indicates the presence of a single iC5 ester decoration on sucrose, and the superscript “R4” describes the acylation position) to make the R_3,4 diacylsucrose Sl-ASAT3 substrates. With the four ASAT enzymes in hand, we reconstructed the cultivated tomato acylsucrose biosynthetic network using sucrose and acyl-CoA substrates. Comparative functional analysis of ASAT2 variants from nine wild tomato relatives led to identification of two residues affecting ASAT2 iso-C5-CoA (iC5-CoA) and anteiso-C5-CoA (aiC5-CoA; 2-methylbutyryl-CoA) substrate preference. Using a similar approach, a residue controlling the ability of ASAT3 to use long chain acyl-CoAs was identified. The in vitro reconstruction system allowed us to test the impact of these variant enzymes and changes in acyl-CoA substrate concentrations on the S. lycopersicum acylsucrose biosynthetic network, demonstrating the value of using the in vitro pathway to understand acylsugar evolution. These results provide a model for understanding how small changes in enzyme sequence lead to large changes in metabolic diversity.

Results

Identification of Two Trichome-Specific BAHD Acyltransferases Involved in Tomato Acylsugar Biosynthesis.

Cultivated tomato produces primarily tri- and tetraacylated sucrose esters in the tip cell of the long multicellular “type I/IV” trichomes (14, 29). Although two trichome apical cell-expressed BAHD acyltransferases that produce tri- and tetraacylsucroses from diacylsucroses were recently described (25, 29), the enzymes that convert sucrose to diacylated sucrose have not been previously reported. We used functional genomics approaches to identify candidate enzymes for these earlier steps in the pathway. Bioinformatic analysis revealed 92 genes predicted to encode BAHD acyltransferase sequences in the S. lycopersicum genome (SI Appendix, Fig. S1A). Twenty-two genes were selected as targets for RNAi suppression in M82; these were predicted to encode full-length proteins and had evidence of trichome expression based on a S. lycopersicum trichome EST database (30) (see the proteins indicated with arrows in SI Appendix, Fig. S1A). Acylsugars in RNAi T₀ primary transgenic plants were analyzed, and lines targeting Solyc12g006330, which we renamed Sl-ASAT1, showed reduction of total acylsugar levels compared with the M82 parental line (Fig. 1A) and no detectable acylsucrose intermediates, pointing to a role for Sl-ASAT1 in acylsugar biosynthesis in M82.

Fig. 1.

The S. lycopersicum M82 trichome acylsugar profile is affected by changes in two BAHD acyltransferases. (A) RNAi suppression of Sl-ASAT1 causes reduction of M82 tomato total acylsugar peak areas. Summed extracted ion chromatogram peak areas divided by internal standard peak areas for all detectable acylsugars are shown for each independent T₀ primary transformant. M82 data are from five biological replicates ± SD. (B) Introgression of S. pennellii chromosome 4 region IL4-1 causes accumulation of acylsucroses S3:15-P (5, 5, 5) and S3:22-P (5, 5, 12), which are not seen in M82 extracts. Negative-ion-mode base-peak intensity LC/MS chromatograms are shown for IL4-1 and M82. Fragment ion masses in positive-ion-mode mass spectra (SI Appendix, Fig. S2A) indicate that these metabolites contain all three acyl chains on one ring. “-P” means that three acyl chains are presumably on the pyranose ring. (C) Sl-ASAT2 transgenic expression causes reversal of the IL4-1 mutant phenotype. LC/MS peak area ratios for the IL4-1–specific S3:22-P and S3:22 acylsugars are shown for each independent T₀ primary transgenic line. Data for IL4-1 and M82 are each from five biological replicates ± SD.

Published studies revealed major differences in trichome metabolite accumulation (15, 25, 26, 29) in several introgression lines (ILs), which have regions of the S. pennellii LA0716 genome substituted in place of the S. lycopersicum M82 genome (31). Rescreening the ILs by liquid chromatography–time of flight mass spectrometry (LC–ToF MS) identified a more subtle phenotype: accumulation of two low-abundance species in IL4-1 not seen in the M82 parent (Fig. 1B). Positive-ion mode MS fragment ion masses revealed that these triacylsucroses—S3:15-P (5, 5, 5) and S3:22-P (5, 5, 12)—presumably have all three acyl chains on the pyranose ring (SI Appendix, Fig. S2A). This is in contrast to the acylsucroses S3:15 (5, 5, 5) and S3:22 (5, 5, 12), which are typically seen in the M82 parent and which have two acyl chains on the pyranose ring and one on the furanose ring (SI Appendix, Fig. S2A). The locus controlling this acylsugar phenotype was narrowed down to a region containing 64 genes by screening selected backcross inbred lines (BILs) that have recombination breakpoints on chromosome 4 (SI Appendix, Fig. S2B). Among the 64 genes, a strong candidate gene—Solyc04g012020 (Sl-ASAT2)—and its putative orthologous gene Sopen04g006140 (Sp-ASAT2) in LA0716 were identified based on the prediction that they encode enzymes belonging to the BAHD acyltransferase family.

The in vivo function of Sl-ASAT2 was tested by transgenic plant experiments. F1 plants generated by crossing IL4-1 to M82 showed the M82 acylsugar phenotype, consistent with the hypothesis that Sp-ASAT2 is recessive to Sl-ASAT2. Thus, we predicted that transformation of IL4-1 with an Sl-ASAT2 transgene driven by its own promoter would restore IL4-1 acylsugar profiles to the wild-type M82 phenotype. Indeed, 15 of 23 independent IL4-1 T₀ transformant plants showed varying levels of complementation with the peak area ratio of S3:22-P to S3:22 restored from 0.23 in IL4-1 to less than 0.05, which is close to the ratio observed in M82 (Fig. 1C). Sl-ASAT2 RNAi lines were generated in M82, and reduced total acylsugar levels were observed for 24 of 29 independent transgenic lines (SI Appendix, Fig. S2C); this is consistent with the hypothesis that Sl-ASAT2 plays an in vivo role in wild-type M82 acylsugar production. The T₀ transgenic RNAi (SI Appendix, Fig. S3) and complementation results (SI Appendix, Fig. S4) were confirmed in T₁ progeny lines generated by self-crossing.

Deep RNA-seq analysis revealed that both Sl-ASAT1 and Sl-ASAT2 mRNAs are highly enriched in M82 tomato trichomes (26), and trichome-enriched expression of Sl-ASAT1 and Sl-ASAT2 in M82 stem trichomes and Sp-ASAT2 in S. pennellii LA0716 leaf trichomes were confirmed by RT-PCR (SI Appendix, Figs. S1C and S2D). These results were further validated and refined by producing M82 transgenic lines expressing a green fluorescent protein–β-glucuronidase (GFP–GUS) reporter driven by the promoter of Sl-ASAT1 or Sl-ASAT2. Each promoter drove GFP expression in the tip cell of type I/IV trichomes in stably transformed M82 plants (SI Appendix, Figs. S1D and S2E). This pattern is identical to that of Sl-ASAT3 (25), Sl-ASAT4 (29), and Sl-IPMS3 (26) and supports the hypothesis that Sl-ASAT1 and Sl-ASAT2 have functions in acylsugar biosynthesis.

Sl-ASAT1 and Sl-ASAT2 Work Sequentially to Produce Diacylsucroses in Vitro.

The combination of lack of accumulation of partially acylated acylsugar pathway intermediates in RNAi plants and type I/IV apical cell expression led to the hypothesis that Sl-ASAT1 catalyzes the first step of acylsucrose biosynthesis. Indeed, recombinant Sl-ASAT1 protein expressed in Escherichia coli converted sucrose and iC5-CoA to make the monoacylsucrose S1:5 (Fig. 2A). Sl-ASAT1 can also use other short chain (iC4-CoA, aiC5-CoA) or long-chain (nC10-CoA, nC12-CoA) acyl-CoAs as in vitro acyl donors to make the respective monoacylsucroses (Fig. 2B). To determine the Sl-ASAT1 reaction product structures, S1:5 and S1:12 were purified and analyzed using NMR spectroscopy. Acylation at the sucrose R₄ position was observed for both, with the NMR chemical shift data for S1:12 (nC12^R4) shown in SI Appendix, Table S1 and for S1:5 (iC5^R4) in table S4 in Schilmiller et al. (25). A chromatographically separable S1:5 isomer, seen as a minor later eluting peak (Fig. 2A), was purified from the original S1:5 product and found to be acylated at the R₆ position; this was demonstrated to be an in vitro artifact caused by acyl chain rearrangement that is promoted by high pH. The putative ortholog of ASAT1 in S. pennellii LA0716, Sopen12g002290, encodes a protein that shares 97.9% amino acid identity with Sl-ASAT1. This protein also produces monoacylsucroses (Fig. 2B), indicating that ASAT1 has a conserved function in the tomato branch of Solanum. Taken together, our results indicate that ASAT1 catalyzes the first step of sucrose acylation and produces an R₄ monoacylated sucrose product.

Fig. 2.

Consecutive in vitro reactions with Sl-ASAT1 and Sl-ASAT2 proteins produce diacylsucroses from sucrose. (A) Result of Sl-ASAT1 enzyme activity assay using sucrose and iC5-CoA as substrates. Negative-ion-mode LC/MS extracted ion chromatograms for m/z 387.1 (sucrose; [M+formate]⁻) and the corresponding ion for reaction product m/z 461.1 (S1:5; [M+Cl]⁻) are shown. “S” represents an acylsucrose backbone, “1” indicates the number of acyl chains, and “5” corresponds to the total number of carbons in the acyl chain. The minor peak is a S1:5 (iC5^R6) isomer produced by nonenzymatic rearrangement. (B) Summary of the reactions catalyzed by Sl-ASAT1 or Sp-ASAT1 (Sopen12g002290) with sucrose and acyl-CoA substrates of different chain lengths. The R₄ acylation of sucrose by Sl-ASAT1 was verified by NMR for the monoacylsucroses containing an iC5 acyl chain as shown in table S4 of Schilmiller et al. (25) or the nC12 chain (SI Appendix, Table S1). (C) In vitro production of R₃-acylated diacylsucroses by Sl-ASAT2 using S1:5 (iC5^R4) and different acyl-CoAs (iC4-, iC5-, aiC5-, nC10-, and nC12-CoA) as substrates. Negative-ion-mode LC/MS extracted ion chromatograms for S1:5 and different diacylsucrose products are shown. The S2:10 (iC5^R4, aiC5^R3) structure was verified by NMR as shown in table S3 of Schilmiller et al. (25).

Sl-ASAT2 in vitro enzyme activity was tested using protein expressed in E. coli. We found that Sl-ASAT2 uses S1:5 (iC5^R4)—the product of Sl-ASAT1—and the structurally diverse acyl-CoA donor substrates iC4-CoA, aiC5-CoA, nC10-CoA, and nC12-CoA to make the corresponding diacylsucroses (Fig. 2C). The enzyme produced only a small amount of product with iC5-CoA compared with other substrates, which indicates that iC5-CoA is not a preferred substrate for Sl-ASAT2. NMR analysis was performed on the purified compound S2:10 (iC5, aiC5) made by sequential reaction of sucrose with iC5-CoA catalyzed by Sl-ASAT1 and the subsequent reaction of this product with aiC5-CoA catalyzed by Sl-ASAT2. NMR chemical shift data indicate that Sl-ASAT2 added the aiC5 group to the R₃ position of the sucrose backbone as shown in table S3 of Schilmiller et al. (25). The diacylsucrose S2:17 (iC5, nC12), made by sequential reaction of sucrose with iC5-CoA (catalyzed by Sl-ASAT1) and its product with nC12-CoA (catalyzed by Sl-ASAT2), has the same chromatographic retention time as S2:17 (iC5^R4, nC12^R3) purified from IL11-3 (SI Appendix, Fig. S5), which has NMR-resolved structural information showing the iC5 group at the R₄ position and the nC12 group at the R₃ position (25). This result suggests that Sl-ASAT2 added the long acyl chain to the R₃ position of the S1:5 monoacylsucrose substrate to make R₃, R₄ substituted diacylsucroses. In contrast, Sl-ASAT2 does not efficiently use purified S1:12 (nC12^R4) as the acyl acceptor using any of the acyl-CoA donors tested.

Kinetic analyses were performed to obtain more detailed information regarding the in vitro properties of the acyltransferases. ASAT1 and ASAT2 used short chain acyl-CoA esters with apparent K_m values in the 20- to 50-µM range, and the longer chain nC12-CoA exhibited substrate inhibition of both enzymes (SI Appendix, Table S2 and Fig. S6), as was also previously reported for ASAT3 (25). An apparent K_m of 2.3 mM was measured for ASAT1 and the acceptor substrate sucrose (SI Appendix, Table S2 and Fig. S6). In addition, the ASAT1 enzyme showed evidence of substrate promiscuity as was previously documented for other BAHD enzymes (32). The aromatic benzoyl-CoA was efficiently used as a donor substrate with sucrose, whereas the negatively charged malonyl-CoA did not yield detectable product (SI Appendix, Table S3 and Fig. S7). Although ASAT1 had no activity with the monosaccharide glucose, it used the glucose-containing disaccharides cellobiose, lactose, maltose, and trehalose as acceptor substrates when acyl-CoA substrates were used, albeit much less efficiently than with sucrose (less than 3% of the monoacylsucrose products peak areas) (SI Appendix, Fig. S7). Finally, all four ASAT enzymes were found to be readily reversible when incubated with their usual products and 100 µM CoA (SI Appendix, Fig. S8), as reported for other BAHD enzymes (33, 34). In addition, S1:5 (iC5^R4) tended to hydrolyze to sucrose incubated with CoA in the absence of enzyme.

In Vitro Reconstruction of M82 Acylsucrose Biosynthesis.

In vitro reconstruction serves as an excellent approach to validate biosynthetic pathways and provides the opportunity to explore the feasibility of metabolic engineering approaches. We asked whether the four recombinant enzymes—ASAT1 through ASAT4—could use acyl-CoA substrates to produce the acylsucroses extracted from cultivated tomato. Reconstruction of the M82 acylsucrose biosynthetic network starting with sucrose was performed by sequentially adding the four enzymes and appropriate acyl-CoA substrates in a single tube (Fig. 3A). Because M82 tomato trichomes produce acylsucroses with long or short chains at the R₃ position (14), parallel Sl-ASAT2 reactions were performed using either aiC5-CoA or nC12-CoA as the acyl donor substrate and S1:5 (iC5^R4), produced by Sl-ASAT1, as the acyl acceptor substrate. The diacylsucrose products from the second step were then used as substrates for Sl-ASAT3, followed by reaction of these triacylsucrose products with C2-CoA and Sl-ASAT4 (chromatograms “S” and “L,” respectively, in Fig. 3B). The sequential assays produced mono, di-, tri-, and tetraacylsucroses, and the resultant S3:15 (5, 5, 5), S4:17 (2, 5, 5, 5), S3:22 (5, 5, 12), and S4:24 (2, 5, 5, 12) had chromatographic retention times identical to the M82 trichome acylsucroses (Fig. 3B). The high-collision-energy negative-ion mode mass spectra of these four compounds revealed fragment ion spectra indistinguishable from the tri- and tetraacylsucroses extracted from M82 that were previously documented by Schilmiller et al. (29).

Fig. 3.

In vitro reconstruction of production of the four major M82 acylsugars by sequential addition of ASAT1, ASAT2, ASAT3, and ASAT4 using sucrose and acyl-CoA substrates. (A) Schematic representation of sequential enzyme assays. The first reaction used sucrose and iC5-CoA to make the monoacylsucrose product S1:5 (iC5^R4) catalyzed by Sl-ASAT1. After enzyme heat inactivation, the acyl-CoA short-chain aiC5-CoA (S) or long-chain nC12-CoA (L) was added with Sl-ASAT2, and the corresponding diacylsucroses were produced. Next, Sl-ASAT3 and iC5-CoA were added followed by Sl-ASAT4 and C2-CoA to produce tri- and tetraacylsucroses, respectively. (B) LC/MS-extracted ion chromatogram analysis of the products of the sequential reactions. “S” and “L” represent the sequential assay for which the short-chain aiC5-CoA or long-chain nC12-CoA, respectively, were used with Sl-ASAT2. Color coding of the product peaks and names corresponds to enzyme names and acyl chains in A. Relative abundance for each chromatogram is based on setting the major peak to 100%; M82: S4:17 (2, 5, 5, 5); L: S3:22 (5, 5, 12); S: S4:17 (2, 5, 5, 5).

Although the sequential reconstruction experiments provided strong support that these four enzymes are sufficient to produce the major products in leaf-surface extracts from sucrose and CoA esters, simultaneous presence of enzymes and substrates presumably more closely reflects the in vivo reaction conditions. The four enzymes, sucrose, and acyl-CoA substrates were added simultaneously for these mixed assays: iC5-CoA, which is the acyl donor for Sl-ASAT1 and Sl-ASAT3, and C2-CoA, the acyl donor for Sl-ASAT4, were added to the mixed reactions, together with varied types of acyl-CoA substrates for Sl-ASAT2, either with short (iC4-CoA, aiC5-CoA) or with long acyl chains (nC10-CoA, nC12-CoA). Collectively, the reactions produced the full set of tri- and tetraacylsucroses that accumulate in vivo, including the iC4- and nC10-containing compounds that are relatively minor components in tomato plants (SI Appendix, Fig. S9 A and B). All in vitro-produced tri- and tetraacylsucroses shared the same m/z and chromatographic retention times with the corresponding acylsucroses extracted from M82 trichomes (SI Appendix, Fig. S9 A and B). The in vitro production of the full set of M82 acylsucroses provides strong evidence that Sl-ASAT1, Sl-ASAT2, Sl-ASAT3, and Sl-ASAT4 are the major enzymes in the acylsugar metabolic network in the apical cell of cultivated tomato type I/IV trichomes.

In Vitro System Responds to Changes in Acyl-CoA Precursor Availability.

Acylsugar phenotypic diversity was observed for various wild tomato species and for accessions within S. pennellii and S. habrochaites (14, 24–26). For example, we recently demonstrated that introgression of a region of S. pennellii LA0716 from the top of chromosome 8 (IL8-1-1) causes increased accumulation of iC4-containing acylsucroses due to introduction of a gene encoding a truncated Leu feedback-insensitive isopropylmalate synthase (IPMS)-like enzyme (Sp-IPMS3) (15, 26). Unlike the enzymatically active Sl-IPMS3, this S. pennellii isoform has a defect in in vitro IPMS activity, and we hypothesized that this defect blocks production of iC5-CoA and diverts its precursor to iC4-CoA, leading to accumulation of iC4 acyl chain-containing acylsugars (26). We used the in vitro reconstructed S. lycopersicum acylsucrose pathway to test the hypothesis that the IL8-1-1 high-iC4 acylsucrose phenotype is due to an increase in the ratio of iC4-CoA to iC5-CoA substrates. We varied the relative amounts of iC4-CoA and iC5-CoA added to in vitro reactions that contained all substrates and enzymes added simultaneously (SI Appendix, Fig. S9C). As shown in Fig. 4, both the absolute and relative amounts of the four C4-containing metabolites increased as the percentage of total iC4-CoA [100 × iC4-CoA/(iC4-CoA + iC5-CoA)] went from 12.5% to 50%. Increases in S3:20 (4, 4, 12) and S4:22 (2, 4, 4, 12) were especially sensitive to iC4-CoA substrate availability, presumably because their synthesis is completely dependent on this substrate. These in vitro results are consistent with the hypothesis that provision of acyl-CoA esters influences the overall composition of acylsucroses in trichomes of S. lycopersicum. They validate the idea that differences in iC4-containing acylsugars in the IL8-1-1 introgression line and S. pennellii accessions are due to changes in iC5- and iC4-CoA availability caused by differences in IPMS3 enzyme structure and function (26).

Fig. 4.

Varying the ratio of iC4-CoA and iC5-CoA causes changes in accumulation of iC4-containing acylsucroses in vitro. Quantification of LC/MS-extracted ion chromatogram peak areas divided by internal standard peak areas for the in vitro-produced acylsucroses S3:20 (4, 4, 12), S4:22 (2, 4, 4, 12), S3:21 (4, 5, 12), and S4:23 (2, 4, 5, 12) showed that the accumulation of iC4-containing acylsucroses increased as the percentage of total iC4-CoA [100 × iC4-CoA/(iC4-CoA + iC5-CoA)] went from 12.5% to 50%. The fold change for each acylsucrose is shown with replicates ± SE.

Comparative Analysis of Natural Variant Enzymes Reveals Amino Acid Residues Affecting ASAT2 Acyl-CoA Substrate Specificity.

Differences in ASAT3 and ASAT4 enzyme activities were previously demonstrated to contribute to metabolite diversity in S. habrochaites and S. pennellii (24, 25). Sp-ASAT2, which shares 95.1% protein identity with Sl-ASAT2, showed barely detectable activity with the Sl-ASAT2 acyl acceptor substrates S1:5 (iC5^R4) and different acyl-CoAs. This observation suggests that diversification of ASAT2 substrate selectivity could also influence the acylsugar diversity observed in various wild tomato species (25). To test this prediction, nine putative ASAT2 orthologs were cloned from wild tomato relatives that are phylogenetically positioned between cultivated tomato and S. pennellii, using primers based on published genomic resequencing data (35). Enzyme activities were then tested using S1:5 (iC5^R4) and different acyl-CoAs as substrates (Fig. 5A). The ASAT2 isoforms of the closest relatives of cultivated tomato showed similar acyl-CoA substrate preference to that of Sl-ASAT2 when using S1:5 (iC5^R4) as the acyl acceptor substrate: Solanum pimpinellifolium LA1578 and Solanum galapagense LA1401 had barely detectable activities with iC5-CoA as a donor substrate, but used a variety of other acyl-CoAs (Fig. 5A). This in vitro activity is consistent with the lack of iC5 acylation at the R₃ position of cultivated tomato acylsucroses (14). In contrast, ASAT2 isoforms from the remaining species used iC5-CoA more efficiently (Fig. 5A). An interesting exception is that aiC5-CoA was a poor substrate for the S. habrochaites LA1718-ASAT2 isoform (Fig. 5A). The relatively high protein sequence identity of ASAT2 variants allowed us to recognize a small number of amino acid residues that correlate with differences in ASAT2 variant substrate specificity (Fig. 5A; SI Appendix, Fig. S10). This analysis led to identification of four candidate residues associated with the preference for iC5-CoA and six candidate residues that correlate with aiC5-CoA utilization.

Fig. 5.

ASAT2 amino acid polymorphisms and variation in acyl-CoA substrate specificities. (A) Alignment of amino acid polymorphisms of ASAT2 putative orthologs with acyl-CoA substrate specificities. ASAT2 phylogenetic tree obtained using protein sequences. “+” means good enzyme activities with detectable product peaks; “−” means no or barely detectable enzyme activity. Polymorphisms at amino acid residues 30, 104, 106, and 408 correspond with different ASAT2 activity using iC5-CoA as the substrate. LA1718 ASAT2, which is the only enzyme that does not efficiently use aiC5-CoA, has residues at positions 44, 185, 257, 295, 298, and 320 that are unique compared with other ASAT2. (B, Left) The alignment of homology-modeled cartoon structure of Sl-ASAT2 (orange) superimposed upon the crystallographic cartoon structure of a trichothecene 3-O-acetyltransferase−acyl CoA complex (3B2S) (gray) (36), which has acyl-CoA (red) and protein crystallized in a complex. (Right) Highlights of Sl-ASAT2 residues that correlate with differences in acyl-CoA substrate specificity and model near the 3B2S acyl-CoA.

We reasoned that variant amino acids near to the predicted acyl-CoA–binding pocket were the strongest candidates for affecting acyl donor specificity. Protein structure homology modeling was performed for Sl-ASAT2 using a trichothecene 3-O-acetyltransferase−acyl CoA complex (3B2S) (36). This analysis revealed that—of the candidate residues possibly affecting iC5-CoA and aiC5-CoA substrate specificity identified in the comparative sequence analysis (Fig. 5A)—Phe⁴⁰⁸ and Ile⁴⁴ were closest to the putative acyl-CoA–binding pocket (Fig. 5B).

Site-directed mutagenesis and in vitro enzyme assays were used to test the hypothesis that these candidate amino acids influence acyl-CoA substrate preference. Consistent with expectation, mutagenesis of Sl-ASAT2 from Phe⁴⁰⁸ to Val⁴⁰⁸ increased the ability of the enzyme to use iC5-CoA to produce S2:10 (5, 5) (Fig. 6): the variant enzyme has an apparent K_m value for iC5-CoA of 26.3 ± 3.2 μM. This is similar to the apparent K_m value 27.1 ± 5.2 μM for iC5-CoA for Solanum arcanum LA2172 ASAT2, an isoform that efficiently uses iC5-CoA as a substrate (SI Appendix, Fig. S11). The reciprocal mutagenesis change—with S. arcanum LA2172 ASAT2 mutated from Val⁴⁰⁸ to Phe⁴⁰⁸—led to an enzyme with undetectable activity for iC5-CoA (Fig. 6) while retaining the ability to use aiC5-CoA and nC12-CoA as substrates. In contrast, mutating position 43 of the S. habrochaites LA1718 ASAT2 from Leu to Ile conferred the ability to use aiC5-CoA as a substrate (Fig. 6) with the apparent K_m value observed of 159 ± 23 μM (SI Appendix, Fig. S11). The conversion of Ile to Leu at amino acid 44 of Sl-ASAT2 abolishes its ability to use aiC5-CoA as a substrate (Fig. 6; SI Appendix, Fig. S11) without affecting the ability to accept the long-chain nC12-CoA as a substrate. Taken together, this comparative biochemical analysis identified two residues affecting ASAT2 acyl-CoA substrate specificity.

Fig. 6.

Single-residue substitutions affect ASAT2 substrate preference for iC5-CoA or aiC5-CoA. (Left) The amount of S2:10 (iC5,aiC5) produced by ASAT2 of varying structures using S1:5 (iC5^R4) and aiC5-CoA as substrates. The ASAT2 variants with Ile⁴⁴ have higher activity using aiC5-CoA as substrate than those with Leu⁴⁴. (Right) The amount of S2:10 (iC5,iC5) produced by different ASAT2s using S1:5 (iC5^R4) and iC5-CoA as substrates. The Val⁴⁰⁸ ASAT2 variants have higher activity using iC5-CoA as substrate than those with Phe⁴⁰⁸. Average LC/MS peak areas divided by internal standard peak areas of the S2:10 products with ±SD were calculated from three replicates.

Identification of an ASAT3 Residue Associated with Long-Chain Acylation of the Acylsucrose Furanose Ring.

In contrast to acylsucroses extracted from S. lycopersicum M82, which have an iC5 acyl chain at the R_3′ position of the furanose ring, some S. habrochaites accessions produce acylsucroses with long chains (C10–C12) at this position (14, 25). These differences correlate with variation in in vitro ASAT3 activities from those accessions (25). We used the comparative biochemical approach to identify amino acids responsible for these differences in ASAT3 acyl-CoA substrate specificities. As shown in SI Appendix, Fig. S12, 19 residues correlated with activity differences between Sl-ASAT3 and the long-chain acyl-CoA−using ASAT3 variants from S. habrochaites LA1777 and LA1731. Homology modeling of Sl-ASAT3 with the trichothecene 3-O-acetyltransferase−acyl CoA complex 3B2S structure suggested that amino acids 35 and 41 are close to the acyl-CoA–binding pocket (Fig. 7A). Indeed, mutation of Tyr⁴¹—found in short acyl chain-using enzymes—to the Cys⁴¹ found in the S. habrochaites Sh-ASAT3-F enzymes converted Sl-ASAT3 into an enzyme that adds nC12 to the furanose ring of S2:10 (5, 5) to produce S3:22 (5, 5, 12), a product not seen in M82 trichome extracts (Fig. 7 B and C). Sl-ASAT3_Y41C had an apparent K_m value of 1.1 ± 0.5 μM for nC12-CoA and also exhibited substrate inhibition by nC12-CoA with an apparent K_i value of 4.5 ± 2.0 μM (SI Appendix, Fig. S11). Again, these results are similar to values observed for LA1777 Sh-ASAT3-F (25).

Fig. 7.

ASAT3 position 41 polymorphisms control acyl chain length preference. (A) The predicted homology-modeled structure of Sl-ASAT3 superimposed on the 3B2S structure. The Sl-ASAT3 structure (green ribbon) and acyl-CoA (red) are shown. The amino acids highlighted as blue or orange are the residues in Sl-ASAT3 that are not found in two S. habrochaites ASAT3 isoforms that can use long-chain acyl-CoAs as substrates. The residue Y⁴¹ is shown in orange. (B) Sl-ASAT3_Y41C uses purified S2:10 (iC5^R4, aiC5^R3) and nC12-CoA as substrates to produce a triacylsucrose S3:22 (5, 5, 12). LC/MS-extracted ion chromatograms are shown for m/z 555.2 (S2:10) and m/z 737.4 (S3:22). (C) Positive-ion-mode mass spectrum with a collision potential of 40 V is shown for S3:22 (5, 5, 12). Presence of the fragment ion with m/z 345 indicates that the long acyl chain was added to the furanose ring of sucrose.

The ability of a single amino acid substitution to change Sl-ASAT3 into a Sh-ASAT3–type activity suggested that we could transform the S. lycopersicum metabolic network into that seen in S. habrochaites LA1777 by substitution of this variant mutant enzyme. Indeed, use of this Sl-ASAT3_Y41C variant enzyme in the sequential in vitro metabolic network reconstruction system led to accumulation of four different acylsucroses containing C10 acylations on the furanose ring (SI Appendix, Fig. S13). These four in vitro-synthesized acylsucroses shared the same chromatographic retention time with the corresponding acylsugars extracted from LA1777 leaf trichomes (SI Appendix, Fig. S13A), and positive-mode high-collision-energy mass spectra revealed long-chain acylation on the furanose ring (SI Appendix, Fig. S13B). These results are consistent with the hypothesis that a single amino acid change explains some of the major differences in acylsucrose structures observed among accessions of S. habrochaites and between S. lycopersicum and S. habrochaites.

Discussion

During the past 10 y the glandular-secreting trichomes of cultivated and wild tomatoes have emerged as a model for studying the evolution of previously uncharacterized specialized metabolic networks, including terpenes in type VI glands (37–41), methylated flavonoids (42), and acylsugars in type I/IV trichomes (24–26, 29). These studies revealed strong metabolic diversity both within tomato species and across the tomato clade of the genus Solanum, leading us to study the genetic and biochemical mechanisms associated with this evolutionarily rapid diversification. In this work we describe the characterization of ASAT1 and ASAT2 BAHD acyltransferases that catalyze the first two steps of acylsucrose biosynthesis from sucrose in S. lycopersicum. These enzymes were used in combination with the previously described Sl-ASAT3 and Sl-ASAT4 to reconstitute the synthesis of the major S. lycopersicum tri- and tetraacylsucroses in vitro. When combined with phenotypic analyses and demonstration that these enzymes are expressed in the acylsucrose-producing type I/IV trichome apical cells, these in vitro reconstruction experiments confirm the roles of these enzymes in acylsugar synthesis.

ASAT1, which catalyzes the first acylation step in tomato acylsucrose biosynthesis, has several features that distinguish it from other characterized ASAT enzymes. First, its RNAi lines do not accumulate detectable acylsucrose intermediates, consistent with its role in catalyzing the first step of the biosynthetic pathway. Second, in contrast to ASAT2-ASAT4, we found no in vivo phenotypic or in vitro enzyme activity evidence for ASAT1 genetic variation leading to altered enzyme activities across the tomato clade, which shares a last common ancestor several million years ago (43). As acylsugar metabolic networks are characterized in Solanaceae species outside of the tomato group, it will be interesting to learn whether this enzyme was also conserved as the committing step over tens of millions of years of evolution.

As seen for other BAHD acyltransferases (for example, see ref. 32), Sl-ASAT1 showed evidence of promiscuity in vitro, acylating sucrose at the R₄ position using acyl-CoAs with different acyl chain lengths (iC4, iC5, nC10, iC12) or branching patterns (subterminally branched aiC5 and terminally branched iC5). Despite the ability of the enzyme to use longer chain CoAs, only iC4 and iC5 chains thus far have been reported at the R₄ position of acylsucroses of S. lycopersicum M82 and three different accessions of S. habrochaites (14). This seeming conflict likely arises because the next enzyme—Sl-ASAT2—is unable to use the S1:12 (nC12^R4) ASAT1 products as substrates to produce diacylsucroses. This presumably causes production of dead-end monoacylated products that may be degraded by an acylsucrose acylhydrolase that cleaves the R₄ position of acylsucroses—or converted to compounds of sufficiently different structure that they are not detected by our analytical methods.

In contrast to the uniform composition of the R₄ acyl groups of acylsucroses extracted from S. lycopersicum, the R₃ position acyl chain length and branching pattern is quite variable, with iC4, aiC5, iC10, nC10, and nC12 acyl chains observed (14, 25). This in vivo diversity correlates well with the ability of Sl-ASAT2 to acylate S1:5 (iC5^R4) at the R₃ position using diverse acyl-CoAs. The next two enzymes—Sl-ASAT3 and Sl-ASAT4—can use acyl acceptor substrates with varied R₃ substitutions in vitro to produce the major acylsucroses found in M82 (Fig. 3 and SI Appendix, Fig. S9).

Although we have reconstituted a four-enzyme pathway in cultivated tomato, open questions remain regarding the pathway in different tomato species. For example, what enzyme(s) are responsible for production of the acylsugars containing all acylchains on the pyranose ring in IL4-1 (SI Appendix, Fig. S2A)? How does IL4-1 produce most of the same major acylsugars as M82 despite the introgression of the putative SpASAT2 ortholog, which does not use S1:5 (iC5^R4) as substrate. These suggest that unknown enzymatic activities remain to be discovered to add to the current acylsugar metabolic network.

Understanding the genetic mechanisms leading to changes in the types of specialized metabolites that accumulate over evolutionary time and the biochemical basis for substrate specificity and enzymatic promiscuity is central to plant improvement efforts. Enzyme structure and function analysis depends upon identification of amino acids that influence activity, but even using directed evolution and “semi-rational” approaches to protein engineering by in vitro mutagenesis can be quite time-consuming, requiring construction and testing of large numbers of single or multiple amino acid variants (44–47) and the availability of suitable substrates.

We took advantage of existing acylsucrose variation within accessions of S. habrochaites and other tomato species to seek ASAT2 and ASAT3 amino acids that are responsible for these phenotypic differences. Comparisons of primary sequence variation with in vitro assays performed using a variety of acyl-CoA substrates revealed relatively small numbers of amino acids as candidates for influencing substrate specificity. These candidate residues were screened further for those that might be positioned to interact with the CoA substrates using homology modeling. In all three cases the approach led to identification of amino acid positions that impacted substrate utilization. ASAT2 Val/Phe⁴⁰⁸ strongly influenced utilization of terminally branched iC5-CoA as a donor substrate. The second ASAT2 residue identified in this analysis, Leu⁴³ found in S. habrochaites LA1718 (Ile⁴⁴ in Sl-ASAT2), influences utilization of aiC5-CoA as donor substrate. The ability to facilely identify a single amino acid residue that influences discrimination between such structurally similar substrates (terminally and subterminally branched C5-CoAs) speaks to the power of this approach. Finally, comparison of Sh-ASAT3 variants that can add a longer chain to the R_3′ position to those with a preference for short-chain acyl-CoAs revealed that substitution of Cys in place of Tyr at position 41 is sufficient to transform substrate specificity. The comparative genomic/biochemical approach coupled with in vitro analysis has also been applied to identify key residues of enzymes involved in terpene (48) and artemisinin biosynthesis (49). This approach has the potential to be generally applicable to structure–function studies of any enzymes that make products that are variable across related populations or related species of plants.

The results from this study have implications for engineering of acylsugars and related compounds. The ability to test the impact of combinations of BAHD acyltransferase isoforms on product types should inform breeding and genome-editing approaches to modify biotic stress tolerance of tomato and other Solanaceae plants. It will also permit regiospecific synthesis of compounds for activity screening or large-scale production by synthetic biology approaches—for example, novel pesticides, antimicrobials, pharmaceutical excipients, and emulsifiers.

Materials and Methods

Tomato Transformation.

Transformation of tomato M82 and IL4-1 was performed using Agrobacterium tumefaciens strain AGL0 (50). BAHD acyltransferase gene suppression was performed by cloning a fragment of each gene from M82 genomic DNA into the pHELLSGATE12 binary vector (51) and transforming M82 plants. Sl-ASAT2 under the control of its native promoter was cloned into pK7WG (52) and transformed into IL4-1. For in planta reporter gene analysis, the promoter regions of Sl-ASAT1 and Sl-ASAT2 were cloned into pKGWFS7 (52) and transformed into M82. Detailed information is in SI Appendix, Materials and Methods.

Plant Trichome Acylsugar Extraction.

Trichome acylsugars were extracted from the youngest expanded leaves of 3- to 4-week-old plants (15). A single leaflet was dipped in 1 mL of extraction solvent, which contained acetonitrile/isopropanol/water (3:3:2) with 0.1% formic acid and 10 μM propyl 4-hydroxybenzoate as internal standard, and the mixture was gently agitated for 2 min.

Mapping the IL4-1 Acylsugar Locus Using Backcross Inbred Lines.

The BIL population was constructed and genotyped as previously described (26). BILs that have chromosome introgression regions covering the IL4-1 and IL4-2 overlap region were selected for acylsugar profile screening using LC/MS. A set of BILs with recombination breakpoints in the overlap region were tested for their acylsugar phenotypes. Two SNP markers and one self-designed insertion/deletion marker (forward primer 5′-TAAAACCTTAGAATCGTTCTCGT-3′ and reverse primer 5′-AAATGATCACTGAAGAATTTCCA-3′) were used for further mapping analysis.

Amplification of ASAT2 Putative Orthologs from Wild Tomatoes.

Accessions of S. pimpinellifolium (LA1578), S. galapagense (LA1401), Solanum neorickii (LA2133), S. arcanum (LA2172), Solanum chilense (LA1969), Solanum peruvianum (LA1278), Solanum corneliomulleri (LA0107), and S. habrochaites (LA1718, LA1777) were obtained from the C. M. Rick Tomato Genetic Resource Center (tgrc.ucdavis.edu). ASAT2-coding sequences were amplified using cDNA transcribed from RNA extracted from 5-wk-old plant leaf tissues as the templates and using primers that were designed based on the wild tomato whole-genome resequencing data (35). Details of ASAT2 sequencing and determination of orthology and GenBank accession numbers for different ASAT2 genes are described in SI Appendix, Materials and Methods.

Protein Expression and Enzyme Assay.

Recombinant proteins were generated using E. coli as the host for enzyme assays. The full-length ORF sequence was cloned into pET28b. ASAT1 and ASAT2 enzyme assays were performed by incubating purified recombinant proteins in 30 μL of 50 mM ammonium acetate (pH 6.0) buffer with 100 μM acyl-CoA and an acylsucrose acceptor. Methods used to determine the apparent K_m value for different acyl-CoA substrates were performed as previously described (25). The detailed steps for protein expression and enzyme activity assays are in SI Appendix, Materials and Methods.

LC/MS Analysis of Acylsugars.

The trichome acylsugars extracted from leaves were analyzed using a Shimadzu LC-20AD HPLC system connected to a Waters LCT Premier ToF MS. Enzyme assay samples were analyzed using a Waters Acquity UPLC system connected to Waters Xevo G2-S QToF LC/MS. Detailed LC/MS methods are in SI Appendix, Materials and Methods.

Homology Structural Modeling of ASAT2.

The Phyre web-based protein homology/analogy recognition engine (53) was used to predict the tertiary structures of Sl-ASAT2 and Sl-ASAT3. The trichothecene 3-O-acetyltransferase structure (Protein Data Bank ID: 3B2S) was used as a template to overlay with Sl-ASAT2 and Sl-ASAT3 modeled structure and displayed using PyMOL (Version 1.7.4 Schrödinger).

Site-Directed Mutagenesis.

Site-directed mutations were created by PCR-based plasmid amplification using the Q5 Site-Directed Mutagenesis Kit (NEB). The primers used to introduce mutations were designed based on the web-based software NEBaseChanger (Version 1.2.2, NEB) and are listed in SI Appendix, Materials and Methods. The presence of the mutations was confirmed by DNA sequencing.