It happened again: Convergent evolution of acylglucose specialized metabolism in black nightshade and wild tomato

Analysis of plant protective surface hair chemistry revealed evolutionary mechanisms leading to metabolic innovation.

Abstract

Plants synthesize myriad phylogenetically restricted specialized (aka “secondary”) metabolites with diverse structures. Metabolism of acylated sugar esters in epidermal glandular secreting trichomes across the Solanaceae (nightshade) family is ideal for investigating the mechanisms of evolutionary metabolic diversification. We developed methods to structurally analyze acylhexose mixtures by 2D NMR, which led to the insight that the Old World species black nightshade (Solanum nigrum) accumulates acylglucoses and acylinositols in the same tissue. Detailed in vitro biochemistry, cross-validated by in vivo virus-induced gene silencing, revealed two unique features of the four-step acylglucose biosynthetic pathway: A trichome-expressed, neofunctionalized invertase-like enzyme, SnASFF1, converts BAHD-produced acylsucroses to acylglucoses, which, in turn, are substrates for the acylglucose acyltransferase, SnAGAT1. This biosynthetic pathway evolved independently from that recently described in the wild tomato Solanum pennellii, reinforcing that acylsugar biosynthesis is evolutionarily dynamic with independent examples of primary metabolic enzyme cooption and additional variation in BAHD acyltransferases.

INTRODUCTION

Plants synthesize hundreds of thousands of structurally diverse specialized (historically referred to as “secondary”) metabolites, which provide humans with medicines, food additives, and natural insecticides. In recent years, an increasing number of these compounds were documented to serve ecological functions. Specialized metabolites group into taxonomically restricted classes, such as glucosinolates in Brassicales and benzoxazinoid alkaloids in Poaceae (1–4). Because the full diversity of specialized metabolites is distributed across the Kingdom, the vast majority of plant specialized metabolites are uncharacterized. In contrast to the typically conserved enzymes and pathways of primary metabolism, variation of the lineage-specific specialized metabolic pathways is observed within and across species (2, 5–7). Specialized metabolic pathway enzymes have characteristics that promote product diversity. These include promiscuity, seen, for example, in BAHD (BEAT, AHCT, HCBT, and DAT) acyltransferases, large amounts of gene duplication (e.g., cytochromes P450), and neo- and subfunctionalization of gene expression and enzyme activity.

Glandular-secreted trichome acylsugars are a structurally diverse group of specialized metabolites produced in the nightshade (Solanaceae) family that have developed into an exemplary system for investigating the emergence and diversification of novel plant metabolites (8–16). These compounds have documented roles in plant defense, including antimicrobial and antiherbivory activities. In species that accumulate and secrete large quantities, the stickiness of acylsugars can interfere with insect feeding and cause flypaper-style entrapment of small, soft-bodied insects (17–21). These characteristics make them of interest in understanding their ecological roles and as breeding targets for crop protection. Most characterized acylsugars consist of a sucrose, glucose, or inositol sugar core esterified at various positions with acyl chains that are typically aliphatic, differing in length, typically C2 to C12 (8, 14, 16, 22–33). A single species in Solanaceae can produce dozens of distinct acylsugars consisting of these simple biosynthetic building blocks (14, 26, 29, 32).

Sugar esters with a sucrose core (acylsucroses) are predominant in Solanaceae (8, 14). The well-characterized biosynthetic pathway producing cultivated tomato (Solanum lycopersicum) acylsucroses involves four trichome-expressed BAHD family acylsugar acyltransferases (ASAT1, ASAT2, ASAT3, and ASAT4) (9–11). Evolutionarily related ASATs were characterized in acylsugar biosynthetic pathways in species across the family, with all but one using sucrose or acylsucroses as acceptor substrate (12–14, 16). These enzymes sequentially transfer acyl chains from acyl-coenzyme A substrates (acyl-CoAs) onto specific positions of the sugar core, with individual ASATs varying in both acyl donor and acyl acceptor substrates. Substrate specificity and promiscuity of these enzymes were demonstrated to influence inter- and intraspecific diversification of acyl chain length and position of acylation on the sucrose backbone, as reported in Petunia axillaris, Salpiglossis sinuata, and wild tomato species (12–14).

While the majority of characterized acylsugars are based on sucrose, acylhexoses were also described. Specifically, acylglucoses were reported in species within the genera Solanum, Datura, and Nicotiana (22–26, 28), while acylinositols were characterized from the South American native fruit crop naranjilla (Solanum quitoense) and the Central American native orangeberry nightshade (Solanum lanceolatum) (16, 27). Relatively little is reported about acylhexose biosynthesis, with one published example of a BAHD acylinositol acetyltransferase in S. quitoense (16) and no reported glucose or acylglucose acyltransferases. We recently showed that wild tomato Solanum pennellii acylglucoses are produced by acylations of sucrose followed by conversion to acylglucoses by a trichome-expressed invertase-like enzyme, acylsucrose fructofuranosidase 1 (SpASFF1) (15). This enzyme uses the acylsucrose substrates with all acyl chains on the pyranose ring (P-type acylsucroses) found in S. pennellii but not those with furanose ring acylation (F-type) found in cultivated tomato or other wild tomato relatives. Hence, this ASFF1-dependent pathway evolved independently in the lineage of S. pennellii. This leaves open the question of how acylglucose biosynthesis proceeds outside the tomato subclade of the genus Solanum.

To address this problem, we investigated trichome acylsugar specialized metabolism in the predominantly Eurasian black nightshade species, Solanum nigrum. In contrast to the New World species S. pennellii, which produces both acylglucoses and acylsucroses, our previous work suggested that the trichomes on S. nigrum young leaves, stems, and fruits exclusively accumulate acylhexoses (14). Moreover, while other characterized species accumulate acylsugars with three or more acyl chains, S. nigrum accumulates acylhexoses with as few as two acyl chains.

Understanding the enzymology behind acylhexose biosynthesis requires accurate assessment of acylation sites on sugar cores, for which mass spectrometry (MS) yields limited information, but nuclear magnetic resonance (NMR) spectroscopy is well suited. Despite the power of modern methodologies, obtaining NMR-resolved chemical structures typically requires samples of sufficient abundance and purity. As a result, while we have liquid chromatography–MS (LC-MS) data for hundreds of structurally diverse acylsugars, only a subset have unambiguous structure annotations. For example, of more than 38 surveyed Solanaceae species that accumulate acylsugars, fewer than half have NMR structures elucidated for at least one acylsugar (16, 22–27, 30–33).

Here, we report the structural characterization of S. nigrum acylhexoses, consisting of triacylinositols, diacylglucoses, and triacylglucoses, and describe the characterization of the S. nigrum acylglucose biosynthetic pathway. Our accelerated structure elucidation pipeline provided NMR-resolved structures for S. nigrum acylsugars from leaf surface extracts without purifying individual compounds. A combination of in vitro biochemical pathway reconstruction and in vivo virus-induced gene silencing (VIGS) validated the acylglucose biosynthetic pathway. This pathway consists of four trichome-expressed enzymes. The two acylsucrose acyltransferases SnASAT1 and SnASAT2 produce diacylsucroses, which serve as substrates for an acylsucrose fructofuranosidase 1 (SnASFF1), which synthesizes diacylglucoses. Triacylglucoses are then produced by SnAGAT1 (S. nigrum acylglucose acyltransferase 1). While it is analogous to the S. pennellii acylglucose pathway, the S. nigrum pathway evolved independently, with co-option of the neofunctionalized SnASFF1 from a distinct lineage of the invertase gene phylogeny and evolution of a BAHD acyltransferase that acts on acylglucoses. In addition, the identification of both acylglucoses and acylinositols but not acylsucroses distinguishes S. nigrum from other acylglucose-accumulating species. This work demonstrates the value of leveraging analytical chemistry, phylogenetics, enzymology, reverse genetics, and knowledge gained from model organisms to facilitate chemical structure assignment and biochemical pathway elucidation in a nonmodel organism.

RESULTS

S. nigrum accumulates di- and triacylated acylglucoses and triacylated acylinositols

Understanding the atomic connections of products and metabolic intermediates is critical for biochemical pathway elucidation. NMR-resolved structures of purified compounds are the gold standard for small molecule metabolites, but obtaining these data can be time-consuming. We sought to speed up the process by subjecting total extracts or partially purified mixtures to two-dimensional (2D) NMR analyses, taking advantage of previously developed methods (Fig. 1C) (26, 32, 34).

Fig. 1. Integrated analysis of S. nigrum acylhexose structures.

(A) LC-MS profile of S. nigrum leaf dip extracts reveals acylsugar-like compounds. Metabolites a, b, and c represent the most abundant acylinositol, diacylglucose, and triacylglucose, respectively. H indicates a hexose core, X:Y(A,B,C,…) indicates a total of Y carbons distributed on X acyl chains with A, B, C… carbons on each chain. See figs. S1, S2, and S3; table S1; and Supplementary Text for details. (B) Fatty acid ethyl ester profile of saponified acylsugars analyzed using gas chromatography–MS (GC-MS) reveal abundant iC4, iC5, aiC5, as well as straight and branched C8, C9, and C10 acyl chains. The abbreviations “i” and “ai” refer to iso- and anteiso-branched isomers (terminally and subterminally branched, respectively). (C) Schematic of pipeline used for acylsugar structure determination. (D) NMR analysis of total acylsugars revealed that the most abundant acylsugar contains a glucose core and acyl chains on C2, C3, and C4 ring protons. BPI, base peak intensity; ES−, negative electrospray ionization. All ES− mode acylsugars are identified as formate adducts. The corresponding mass/charge ratios (m/z) of these compounds are listed in table S1 and Supplementary Text. Photo credit: Yann-Ru Lou, Michigan State University.

As a first step, we profiled metabolites in S. nigrum leaf dip extracts using LC-MS (Fig. 1A). We identified 99 metabolite features with characteristics similar to known acylsugars based on molecular ion adduct masses, mass spectra at elevated collision energies, and fractional hydrogen content as quantified by relative mass defect (Fig. 1A, figs. S1 and S2, table S1, and Supplementary Text) (34, 35). These include 45 metabolite features annotated as acylsugars with high confidence based on criteria discussed in the Supplementary Text. In contrast to the majority of solanaceous plants, we only detected acylhexoses in S. nigrum extracts (14). LC-MS fragmentation patterns revealed that the acylsugars are typically composed of hexose cores and two or three short-chain (C2, C4, and C5) or medium-chain length (C8, C9, and C10) acyl esters (table S1 and Supplementary Text). Thirty-nine of the 45 high-confidence putative acylsugars showed fragmentation patterns reminiscent of previously analyzed S. pennellii acylglucoses (26). However, we observed six triacylhexoses with unusually low fragment ion signals in negative mode (fig. S2); this is a characteristic of S. quitoense acylinositols (16, 36).

The acylglucose- and acylinositol-like peaks in S. nigrum appear as groups of two or more chromatographic peaks sharing identical molecular masses and indistinguishable mass spectra at high collision energy, as consistent with multiple isomeric forms (fig. S3). Extracted ion chromatograms for fragment ion masses characteristic of acyl groups (carboxylate anions) revealed that many individual chromatographic peaks result from multiple co-eluting metabolite features, some of which present indistinguishable mass spectra. For example, the second-eluting peak annotated as H3:16 (fig. S3; H indicates a hexose core, and 3:16 indicates a total of 16 carbons distributed on three acyl chains; full nomenclature is described below) exhibits asymmetry suggestive of four isomers, two of which overlap in chromatographic retention. All four components share the same acyl chain complements, C2, C4, and C10 acyl chains on a hexose core, as evidenced from the extracted fragment ion chromatograms that confirm C4 and C10 acyl groups (fig. S3). We expect each acylglucose to resolve as two peaks corresponding to α and β anomers that interconvert, accounting for the presence of two peaks for each (26). The observation of additional peaks led to the hypothesis that four isomeric forms reflected different acyl chain branching patterns rather than differences in acyl group lengths or positions of esterification. This is supported by the ethyl ester profile revealed by gas chromatography–MS (GC-MS) after saponification and transesterification (Fig. 1B). The ethyl ester profile demonstrated both abundant straight (nC8, nC9, and nC10) and branched (iC4, iC5, and aiC5 and branched C8, branched C9, and branched C10) aliphatic chains (Fig. 1B), supporting the conclusion that acyl chain branching variation contributes to the number of structural isomers.

The combined information from LC-MS and GC-MS led us to conclude that S. nigrum accumulates at least 45 different types of di- and triacylhexoses including structural isomers with acyl chains differing in branching patterns. While unequivocal structural information of these acylhexoses is essential for a complete understanding of their biosynthesis, chromatography-based analytical techniques do not provide needed resolution or selectivity. For example, LC-MS technologies have not yet been developed that can identify acylation positions or unambiguously differentiate isomeric hexose cores (e.g., glucose or inositol). This information is central to understanding functions of key metabolic enzymes that shape acylsugar chemical diversity. We therefore used NMR spectroscopy, which provides important information about molecular topology, to obtain these structural details.

We sought to avoid the time-consuming process of purifying metabolites to homogeneity by analyzing unfractionated and partially purified fractions (Fig. 1C). This took advantage of the 2D NMR spectroscopy methods heteronuclear single quantum coherence (HSQC), which provides information about hydrogen-carbon attachments, and heteronuclear multiple bond correlation (HMBC), which yields information about atoms separated by two to three bonds. Analysis of the total, unfractionated acylsugar extracts from young plants revealed the dominant presence of a glucose backbone with acyl chains at positions 2, 3, and 4 (Fig. 1D). However, abundant triacylglucoses obscured the signals of diacylglucoses and acylinositols in the total extract; therefore, we used partial purification to obtain similar information about these less abundant acylsugars.

Silica gel chromatography was used to separate triacylglucoses (fraction 1) from diacylglucoses and triacylinositols (fraction 2) based largely on the number of hydroxyl groups: There are two hydroxyls on triacylglucoses and three on diacylglucoses and triacylinositols (Fig. 2, A to C). Analysis of the NMR spectra from fraction 1 confirmed that it predominantly contains triacylglucoses with one medium-length (8 to 10 carbon) acyl chain esterified to position 4, a short (4 to 5 carbon) acyl chain at position 3, and an acetyl (C2) group at position 2 (Fig. 2D and table S2). In agreement with LC-MS data, NMR spectra of fraction 2 exhibited signals of two distinct groups of acylhexoses in the sugar region [δ_H = 3 to 6 parts per million (ppm); δ_C = 60 to 100 ppm] (fig. S4). The anomeric positions in hexoses have larger (downfield) ¹³C chemical shifts (δ_C ~ 90 to 100 ppm) than other acylsugar carbon atoms (fig. S4). We used 2D HSQC–total correlation spectroscopy (HSQC-TOCSY), which generates spectra at long mixing times to provide connectivity information between coupled nuclei in a molecule (37). This information aids in associating sugar core hydrogen and carbon signals within a molecule and distinguishing signals from other molecules in a mixture. This approach provided a useful strategy for distinguishing acylation positions on acylglucoses and acylinositols in fraction 2 where signals often overlap. Our analysis revealed acylations at positions 3 and 4 on a sugar core with an anomeric center (glucose) and acylations at positions 2, 3, and 4 on a hexose core with no anomeric center (inositol) (Fig. 2E and fig. S4). HMBC data analysis confirmed the presence of abundant diacylglucoses with a medium-length acyl chain as R₄ and a short acyl chain as R₃ (Fig. 2F and table S3). This approach also revealed the presence of triacylinositols with two short-length and one medium-length acyl chain as R₂, R₃, and R₄ (Fig. 2G and table S4).

Fig. 2. S. nigrum accumulates triacylinositols and di- and triacylated acylglucoses.

(A) Silica flash column separation of leaf dip extracts into less polar fraction 1 and more polar fraction 2, as analyzed by thin layer chromatography. (B) LC-MS profile of fraction 1 reveals mostly triacylglucoses. The peaks of metabolite c (cyan dashed rectangle) with identical m/z and mass spectra represent the two anomers each of two structural isomers of triacylglucose G3:16(2,4,10). (C) LC-MS profile of fraction 2 reveals abundant diacylglucoses and triacylinositols. The two peaks with m/z and mass spectra consistent with metabolite a (orange dashed rectangle) represent two structural isomers of I3:17(4,5,8), whereas metabolite b (magenta dashed rectangle) resolves into multiple peaks representing the two anomers of the two structural isomers of G2:14(4,10). (D) Example of an NMR-resolved structure of metabolite c, the most abundant type of triacylglucoses in fraction 1. See table S2 for details. (E) HSQC-TOCSY separated spin systems between diacylglucoses and triacylinositols in fraction 2. See fig. S4 for details. (F) Examples of NMR-resolved structures of metabolite b, the most abundant diacylglucose, and (G) metabolite a, the most abundant triacylinositol in fraction 2. See tables S3 and S4 for details. All chromatograms are showing telmisartan as internal standard (I.S.). All ES− mode acylsugars are identified as formate adducts. The corresponding m/z’s of these compounds are listed in table S1.

Analysis of the combined LC-MS, GC-MS, and NMR results led to the conclusion that S. nigrum accumulates diacylglucoses esterified at positions 3 and 4 with triacylglucoses and triacylinositols esterified at positions 2, 3, and 4. For the remainder of this report, we follow the nomenclature established by Schilmiller et al. (38) to describe these compounds: SX:Y(A^a,B^b,C^c), in which S indicates the type of sugar core (G for glucose and I for inositol), X is the total number of substituent acyl chains, Y signifies the sum of acyl chain carbons, and (A^a,B^b,C^c) represents the number of carbons of each chains with the superscript documenting the site of substitution. On the basis of this nomenclature, G2:14(4^R3,10^R4), G3:16(2^R2,4^R3,10^R4), and I3:17(4^R4,5^R2,8^R3) represent some of the most abundant acylsugars in each class (Fig. 2, D, F, and G). In cases where the identity of the hexose core remains ambiguous, the letter H for hexose is used.

Identification of a trichome-expressed β-fructofuranosidase involved in S. nigrum acylsugar biosynthesis

The observation that acylglucoses are predominant in S. nigrum trichome extracts led us to posit that, as previously shown for S. pennellii LA0716 acylglucose biosynthesis, acylsucroses are intermediates in acylglucose production. To test this hypothesis, we identified trichome-expressed transcripts in S. nigrum predicted to encode proteins of the glycoside hydrolase family 32, which includes SpASFF1. Twenty invertase-like homologs were identified by BLAST from S. nigrum RNA sequencing data (fig. S5A). We reasoned that, as acylsugars appear to accumulate on S. nigrum trichome tip cells (fig. S5B), the acylglucose biosynthetic genes in S. nigrum would be highly expressed and enriched in trichomes. Among the 20 homologs, 6 demonstrate trichome-enriched expression (fig. S5C). Five of the six have sequence lengths comparable to functional invertases and have WXNDPNG, RDP, and EC amino acid sequences characteristic of β-fructofuranosidases (fig. S5D). Because we previously observed a PSTP noncanonical substrate binding site in SpASFF1 (15), we sought invertase-like sequences also lacking the canonical DXXK in S. nigrum trichome expressed transcripts. Assembly c70979_g1 has amino acid sequence PLTY in place of that conserved substrate binding site (fig. S5D). On the basis of the in vivo transient silencing and in vitro enzyme activities results described below, we designated this protein SnASFF1.

We developed and deployed S. nigrum VIGS to test the hypothesis that SnASFF1 is involved in S. nigrum acylglucose biosynthesis in vivo. Plants with silenced SnASFF1 accumulated diacylsucroses (referred to as “in vivo diacylsucroses” below) that were not detected in the control plants (Fig. 3; fig. S5, E and F; and table S5). NMR spectra of the early- and late-eluting S2:14(4,10) isomers revealed that acylation on positions 3 and 4 (fig. S5G) mirrors those of S. nigrum acylglucoses (Fig. 2, D and F), supporting their role as intermediates in the acylglucose biosynthetic pathway. As characterized ASATs show acylation position specificity, this result is consistent with the hypothesis that there are two or more BAHD acyltransferases that produce diacylsucrose substrates for this SnASFF1 invertase-like enzyme.

Fig. 3. Diacylsucroses accumulate in SnASFF1 VIGS lines.

LC-MS analysis of acylsugars extracted from SnASFF1-targeted (purple trace) and empty vector VIGS plants (black trace) show that SnASFF1-silenced lines accumulate diacylsucroses that are undetectable in control plants. The acylation at positions 3 and 4 on S2:14(4,10) was verified by NMR. as shown in fig. S5. Extracted ion chromatogram (EIC) values indicate telmisartan as internal standard (I.S.) (m/z 513.23) and the formic adducts of S2:12 (m/z 583.26), S2:13 (m/z 597.28), S2:14 (m/z 611.29), and S2:15 (m/z 625.31).

Identification of trichome-specific BAHD acyltransferases

On the basis of published results with acylsucrose- and acylinositol-producing plants in Solanaceae, we hypothesized that acyltransferase enzymes of S. nigrum acylsugar biosynthesis would be trichome enriched and evolutionarily related to characterized ASATs (13, 14, 16). The combined tissue-specific expression and homology-guided approach led us to consider six highly trichome-enriched BAHD family members in the S. nigrum transcriptome (Fig. 4A and fig. S6). These candidates are predicted to encode proteins with characteristics of active BAHD family enzymes (39). First, they have the conserved HXXXD catalytic motif and the DFGWG-like structural motif (fig. S7). Second, the predicted proteins all have lengths comparable to characterized functional BAHDs (fig. S6B) (8–14, 16).

Fig. 4. SnASAT1 and SnASAT2 sequential reaction produces diacylsucroses.

(A) Maximum likelihood phylogeny analysis of amino acid sequences of ASAT candidates showed that SnASAT1 and SnASAT2 cluster with characterized ASAT1 and ASAT2 in tomato species, while SnAGAT1 clusters with the Solanaceae acylsugar acetyltransferases, SlASAT4, SsASAT5, and SqTAIAT. (B) LC-MS analysis of in vitro assay products showing that SnASAT1 produces monoacylsucrose S1:10(10^R4) from sucrose and nC10-CoA substrates. The acylation at position 4 was verified by NMR as shown in fig. S9. Blue trace represents full enzyme assays, whereas the black trace has heat-denatured SnASAT1. (C) LC-MS analysis of in vitro assay products showing that SnASAT2 produces diacylsucrose S2:14(4,10) from monoacylsucrose S1:10(10^R4) and iC4-CoA substrates. Green trace represents full enzyme assays, whereas the black trace has heat-denatured SnASAT2. (D) In vitro–generated S2:14(iC4, nC10) and S2:14(iC4, iC10) co-elute (top and middle, respectively) with the in vivo S2:14(4,10) isomers from SnASFF1-silenced lines (bottom). S2:14(iC4, nC10) products in (C) and (D) are from two independent technical replicates. (E) Summary of the reactions catalyzed by SnASAT1 and SnASAT2 with sucrose and monoacylsucroses, respectively, and acyl-CoA substrates of different chain lengths. For more details, see fig. S12. *Acylation at position 3 verified from SnASAT2-product co-elution with NMR characterized S2:14(4,10) with acyl chains as R₃ and R₄ from S. nigrum. All chromatographs are combined extracted ion chromatogram (ES−) with telmisartan internal standard (I.S.) (m/z 513.23) and the formic adducts of S1:10(10) (m/z 541.25) or S2:12(4,10) (m/z 611.29).

We tested the hypothesis that these genes encode proteins of acylsugar biosynthesis by examining the activities of Escherichia coli–expressed and His-tag affinity-purified enzymes. C10, C8, iC5, iC4, and C2-CoAs were chosen as acyl chain donors in our in vitro assays based on the abundance of these ester groups in our structural annotations (Fig. 1B). Using LC-MS, activity was detected with sucrose and the closest S. nigrum homolog of tomato ASAT1 (SlASAT1), which is encoded by the transcript c63608_g1 (Fig. 4B and fig. S8). This enzyme catalyzes the formation of monoacylsucroses, S1:8 and S1:10, from sucrose and nC8, iC10, and nC10 acyl-CoAs (Fig. 4B and fig. S8). The apparent K_m values for SnASAT1 and nC8- and nC10-CoAs are 114 ± 21 and 11.7 ± 2.1 μM (SEM), respectively (fig. S9A). An apparent K_m of 5 mM was measured for SnASAT1 and the acceptor substrate sucrose (fig. S9A). On the basis of its in vitro activity and the in planta results presented below, we named this enzyme SnASAT1. SnASAT1-produced S1:10(nC10) co-eluted with SlASAT1-produced S1:10(nC10^R4) but not with SsASAT1-produced S1:10(nC10^R2) (fig. S9B) (11, 13, 14). Positive-mode MS fragmentation supports localization of the C10 chain on the six-member pyranose ring (fig. S9C). This result was extended by NMR analysis, revealing S1:10(nC10^R4) as the major product of SnASAT1 in vitro enzyme assays (Fig. 4E and fig. S9D).

A second, chromatographically separable S1:10 isomer, S1:10(10^R6), accumulated in SnASAT1 enzyme assays (figs. S10A and S9D). This 6-position acylated isomer appeared as a minor product in assays carried out at pH 6.0 with short incubation periods (5 to 30 min) (fig. S10A). The concentration of this second isomer increases after extended incubation time, especially under neutral-to-alkaline pH conditions (fig. S10, B to D). For example, comparable amounts of S1:10(10^R4) and S1:10(10^R6) accumulated after 30 min at 30°C under pH 8.0 (fig. S10C). We also observed non-enzymatic conversion of S1:10(10^R4) to S1:10(10^R6) after brief exposure to elevated temperature (65°C) (fig. S10, B and D) or extended incubation of purified S1:10(10^R4) in unbuffered distilled water at room temperature. Non-enzymatic chain migration was previously documented for a R₄ monoacylsucrose produced by SlASAT1, where the short acyl chain moved from R₄ to R₆ (11).

We characterized the activity of trichome-expressed c65670_g1-encoded SnASAT2, the closest homolog of tomato SlASAT2 and SpASAT2 (Fig. 4A). SnASAT2 used iC4-CoA as acyl chain donor to decorate a second ring position on the pyranose ring of both S1:10(nC10^R4) and S1:10(iC10) (Fig. 4, C and D, and fig. S11A). MS fragmentation patterns and retention times of the produced S2:14(iC4,iC10) and S2:14(iC4,nC10) (referred to below as “in vitro diacylsucroses”) were indistinguishable from the early- and late-eluting in vivo S2:14(4,10) metabolites found in the SnASFF1 VIGS plant extracts, respectively (Fig. 4D and fig. S11). We observed broad acyl donor and acceptor substrate specificity with iC4-, iC5-, and nC8-CoAs serving as substrates with both S1:10(10^R4) and S1:8(8) acceptors, producing diacylsucroses chromatographically indistinguishable from those detected in invertase-like deficient VIGS-SnASFF1 lines (Fig. 4E and fig. S12A). In contrast, no products of SnASAT2 were detected with unacylated sucrose, glucose, the SsASAT1 product S1:10(10^R2), or the S1:10(10^R6) SnASAT1 rearrangement product as acceptor (fig. S12, B to D). As seen with SnASAT1-generated monoacylsucroses, a second diacylsucrose isomer accumulated in a pH- and time-dependent manner after the SnASAT1 + SnASAT2 reaction (fig. S13). Together with the identification of SnASFF1, these results are consistent with the hypothesis that S2:14(4,10) from SnASAT1 + SnASAT2 sequential reactions are precursors of the S. nigrum di- and triacylglucoses [e.g., G2:14(4^R3,10^R4) and G3:16(2^R2,4^R3,10^R4)].

SnASFF1 is an acylsucrose hydrolase that evolved independently from SpASFF1

We tested SnASFF1 protein activity on in vitro– and in vivo–produced diacylsucroses using recombinant His-tagged SnASFF1 protein expressed in and purified from the Nicotiana benthamiana transient expression system (15). SnASFF1 hydrolyzed all three tested diacylsucroses: The two in vivo pathway intermediates and the unrearranged in vitro diacylsucrose product generated by SnASAT1 and SnASAT2 sequential reactions all were converted to the corresponding diacylglucose products (Fig. 5A and fig. S14A). The reactions with in vitro and in vivo S2:14(4,10) isomers each yielded two major products that are the anomers of S. nigrum G2:14(4^R3,10^R4), as suggested by their identical retention times and mass spectra (Fig. 5A and fig. S14B). The G2:14(4,10) isomers produced from in vitro S2:14(iC4,nC10) and in vivo late-eluting S2:14(4,10) were indistinguishable from the late-eluting anomers of S. nigrum G2:14(4^R3,10^R4). In addition, SnASFF1 converted the in vivo early-eluting S2:14(4,10) into anomers co-eluting with those of the early-eluting S. nigrum G2:14(4^R3,10^R4) (Fig. 5A and fig. S14B). We hypothesize that these two sets of diacylglucoses, differing in an nC10 or iC10 acyl chain at R₄, collectively explain the overlapping signals of S. nigrum G2:14(4^R3,10^R4) (Fig. 5A). This result is consistent with the presence of abundant nC10 and iC10 in the acylsugar-derived ethyl ester profile (Fig. 1B). Together, these results provide strong evidence that SnASAT1, SnASAT2, and SnASFF1 with sucrose and acyl-CoAs are sufficient to reconstitute production of diacylglucoses produced in S. nigrum.

Fig. 5. SnASFF1 and SnAGAT1 sequential activities with both in vitro and in vivo diacylsucrose substrates produce triacylglucoses.

(A) SnASFF1 activity (purple traces) with the major in vitro S2:14(iC4, nC10) from consecutive SnASAT1 and SnASAT2 reactions (top) and the early- and late-eluting in vivo diacylsucroses S2:14(4,10) from VIGS-SnASFF1 lines (bottom and middle, respectively) produced two distinct sets of diacylglucoses that collectively explains the multiple peaks of S. nigrum G2:14(4^R3,10^R4) (magenta trace). The diacylglucose products from in vitro S2:14(iC4, nC10) and the late-eluting in vivo S2:14(4,10) are chromatographically indistinguishable from the late-eluting S. nigrum G2:14(4^R3,10^R4). Black traces represent reactions with heat-denatured enzymes. (B) SnAGAT1 activities (orange traces) with the diacylglucoses produced by SnASFF1 activities with in vitro and in vivo diacylsucrose S2:14(4,10) substrates produce two distinct sets of triacylglucose products that collectively explains the triplet-like peaks of G3:14(2^R2,4^R3,10^R4) in S. nigrum leaf extracts (cyan trace). Combined EIC under ES− showing formic adducts of S2:14(4,10) (m/z 611.29), G2:14(4,10) (m/z 449.24), and G3:16(2,4,10) (m/z 491.25). (C) Summary of results of SnASFF1 and SnAGAT1 activities in the reconstructed acylglucose biosynthetic pathway.

SnASFF1 hydrolytic activity was not detected using sucrose, in contrast to the canonical yeast invertase activity (fig. S15A). Conversely, yeast invertase failed to use any tested S2:14(4,10) as substrate (fig. S15B). These data are consistent with the hypothesis that SnASFF1 has acylsucrose-specific activity. We also observed SnASFF1 hydrolytic activity with in vitro–produced S1:10(10^R4) and the S. pennellii triacylsucrose S3:18(4^R2,4^R4,10^R3) as substrate (fig. S15B). However, the monoacylglucose product, G1:10(10), does not appear to be an acceptor substrate for SnASAT2 with nC10-, nC8-, iC5-, iC4- or acetyl-CoA (fig. S12A). Together, the results support the hypothesis that SnASFF1 is an acylsucrose hydrolase that contributes to S. nigrum acylglucose biosynthesis by converting diacylsucroses to diacylated glucoses.

To explore the evolution of acylglucoses in S. nigrum and wild tomato, we generated a gene tree with invertase-like genes and transcripts from S. lycopersicum, S. pennellii, and S. nigrum (Fig. 6). Phylogenetic analysis showed that the S. nigrum ASFF1 is a close homolog of the tomato LIN5 (Solyc09g010080) and LIN7 (Solyc09g010090) paralogs, tandem duplicates of sucrose-specific invertases. In contrast, the S. pennellii ASFF1 is in a clade distinct from SnASFF1, with greater similarity to tomato Solyc03g121680 and Solyc06g064620 and the two S. nigrum transcripts, c65240_g1 and c62944_g1. We observed no acylsugar phenotype upon in planta silencing c65240_g1, the closest BLAST hit of SpASFF1 in the S. nigrum transcriptome (fig. S15C). This result is consistent with the low S. nigrum trichome expression of c65240_g1. Together, these relationships indicate that the acylsugar hydrolyzing activity of SnASFF1 and SpASFF1 evolved independently in S. nigrum and the wild tomato, S. pennellii.

Fig. 6. Phylogenetic analysis of SnASFF1 and SpASFF1 reveals independent evolution.

Invertase-like proteins from S. lycopersicum, S. pennellii, and S. nigrum were aligned using MUSCLE with default parameters in MEGA X. SnASFF1 is marked bold in purple, whereas SpASFF1 is marked bold in black. Maximum likelihood tree constructed using the Jones-Taylor-Thornton algorithm with 1000 bootstrap support using the topology-only tree in MEGA X.

Identification of a trichome-specific BAHD acylglucose acetyltransferase

The in vitro production of diacylglucoses by sequential reaction of SnASAT1, SnASAT2, and SnASFF1 is consistent with the hypothesis that the acetylation step leading to S. nigrum triacylglucoses is carried out on diacylglucose substrates. The search among the remaining trichome-enriched ASAT candidates led to identification of acetylation by SnAGAT1 (encoded by c65306_01), as the strongest candidate for the last step of triacylglucose biosynthesis in S. nigrum.

As hypothesized, SnAGAT1 acetylated the diacylglucoses produced by SnASFF1 activities with in vitro and in vivo diacylsucrose S2:14(4,10) substrates (Fig. 5B). The G3:16(2,4,10) anomers obtained from both in vitro S2:14(iC4,nC10) and in vivo late-eluting S2:14(4,10) have retention times and mass spectra identical to the late-eluting peaks of S. nigrum G3:16(2^R2,4^R3,10^R4) (Fig. 5B and fig. S16A). SnASFF1 and SnAGAT1 sequential activities with in vivo early-eluting S2:14(4,10), which shares identical retention times and mass spectra with in vitro–generated S2:14(iC4,iC10), generated early-eluting triacylglucose anomers (Fig. 5B and fig. S16A). We hypothesize that these two sets of triacylglucoses, with an nC10 or iC10 chain, collectively explain the multiple peak signals of S. nigrum G3:16(2^R2,4^R3,10^R4) (Fig. 5B and fig. S3). This result is consistent with the abundance of nC10 and iC10 acyl chains on S. nigrum acylsugars (Fig. 1B). We also observed SnAGAT1 activity with the diacylglucose-rich S. nigrum extract fraction 2 as acceptor substrates and acetyl-CoA as donor, leading to accumulation of triacylglucoses that are indistinguishable to S. nigrum triacylglucoses (fig. S16C). These assays provide strong evidence that SnAGAT1 catalyzes the last step in S. nigrum triacylglucose production (Fig. 5C).

In vitro validation of the S. nigrum acylglucose biosynthetic pathway

We independently assessed the functions of the three BAHD enzymes on the reconstructed pathway by taking advantage of BAHD acyltransferase reversibility in the presence of free CoA. Using plant-derived acylglucoses as substrates for reverse enzyme assays, we verified that SnAGAT1 and free CoA convert S. nigrum extract fraction 1 triacylglucoses into diacylglucoses (Fig. 7A). The ion masses and mass spectra of these diacylglucoses are consistent with loss of an acetyl group (fig. S16B), supporting the designation of SnAGAT1 as an acetyltransferase (Figs. 5C and 7B).

Fig. 7. Independent validation of the reconstructed acylglucose biosynthetic pathway with in vitro reverse activities.

(A) LC-MS analysis of reverse enzyme assay products from SnAGAT1 activity (orange trace) shows SnAGAT1 deacetylating S. nigrum fraction 1 triacylglucoses in the presence of free CoA. Combined EIC under ES− showing telmisartan as internal standard (I.S.) (m/z 513.23) and formic adducts of G2:12, G2:13, G2:14, and G2:15. The corresponding m/z’s of these compounds are listed in table S1. (B) Schematic representation of the reconstructed S. nigrum triacylglucose biosynthetic pathway. Blue arrows indicate in vitro pathway reconstruction starting from sucrose, and green arrows indicate enzyme activities with plant-derived products. (C) SnASAT2 activities with the major (top), but not minor (second from the top), in vitro S2:14(iC4, nC10) products and the early- and late-eluting in vivo diacylsucrose S2:14(4,10) substrates from VIGS-SnASFF1 lines (second from the bottom and bottom, respectively) produce monoacylsucrose products that can further be deacylated by SnASAT1 activity (bottom right). The monoacylsucrose intermediates from the late-eluting in vivo S2:14(4,10) and in vitro S2:14(iC4, nC10) (top and third from the top) are chromatographically indistinguishable from S1:10(nC10^R4). Combined EIC under ES− showing formic adducts of S2:14(4,10) (m/z 611.29) and S1:10(10) (m/z 541.25). Photo credit: Yann-Ru Lou, Michigan State University.

SnASAT2 was tested for reverse activity with both early- and late-eluting in vivo S2:14(4,10) from VIGS-SnASFF1 lines and the major and minor SnASAT1 + SnASAT2 in vitro–generated S2:14(iC4,nC10) isomers (Fig. 7C). SnASAT2 converted the in vivo–derived and the major in vitro–produced diacylsucroses to monoacylsucroses S1:10(10) (Fig. 7C). The monoacylsucroses produced from both in vitro S2:14(iC4,nC10) and in vivo late-eluting S2:14(4,10) are chromatographically indistinguishable from the SnASAT1-forward reaction product S1:10(nC10^R4) (Fig. 7C). Addition of SnASAT1 to either of the SnASAT2 reverse reaction assays reduced the accumulation of S1:10(10) (Fig. 7C). In contrast, no reverse activity was detected using the minor in vitro diacylsucrose rearrangement product as substrate (Fig. 7C). Similarly, reverse assays using in vitro–generated monoacylsucroses as substrates revealed SnASAT1 enzymatic activity with S1:10(10^R4), but not with the S1:10(10^R6) rearrangement product (fig. S17). Together, these data provide independent evidence for the S. nigrum acylglucose biosynthetic pathway, with SnASAT1, SnASAT2, and SnASFF1 producing diacylglucoses and SnAGAT1 catalyzing the final step in triacylglucose biosynthesis (Fig. 7B).

VIGS validation of three acyltransferases in S. nigrum acylglucose biosynthesis

We deployed VIGS to test the hypothesis that SnASAT1, SnASAT2, and SnAGAT1 activities are necessary for in vivo acylglucose biosynthesis (Fig. 8). As expected for early steps in the core acylsugar biosynthetic pathway, silencing SnASAT1 and SnASAT2 each led to reduction of the four most abundant di- and triacylglucoses (Fig. 8, A and B; fig. S18, A and B; and tables S6 and S7). Silencing SnAGAT1 caused statistically significant increases in the ratio of diacylglucoses to corresponding triacylglucoses relative to the controls (Fig. 8C, fig. S18C, and table S8). These results provide strong support that di- and triacylglucoses are produced in vivo via the three- and four-step biosynthetic pathway demonstrated through in vitro biochemistry, respectively (Fig. 7B).

Fig. 8. In vivo validation of the reconstructed acylglucose biosynthetic pathway with VIGS.

(A) Comparison of acylglucose accumulation in SnASAT1-targeted and empty vector VIGS plants. Acylsugars were analyzed using LC-MS in ES+ mode. (B) Comparison of acylglucose accumulation in SnASAT2-targeted and empty vector VIGS plants. Acylsugars were analyzed using LC-MS in ES− mode. (C) Comparison of the ratio of diacylglucose G2:14 and triacylglucose G3:16 in SnAGAT1-targeted and empty vector VIGS plants. Diacylglucose quantities were measured by integrating peak areas of G2:12, G2:13, G2:14, and G2:15, whereas triacylglucose quantities were measured by integrating peak areas of G3:14, G3:15, G3:16, and G3:17. The integrated peak areas were normalized to the internal standard telmisartan and dry leaf weights. Peak areas of G2:14 and G3:16 were integrated under negative mode and normalized to the internal standard telmisartan and dry leaf weights to measure the ratio between tri- and diacylglucoses. The corresponding m/z’s of analyzed acylsugars are listed in table S1 unless otherwise specified. Significant levels are shown (*P < 0.05, **P < 0.01, and ***P < 0.001, Welch’s two sample t test).

DISCUSSION

While evolutionary diversification occurs by the deceptively simple mechanisms of gene duplication, changes in gene expression, and modification of gene product activities, the resulting phenotypic diversity is remarkable. In plant specialized metabolism, a matrix of a relatively small number of chemical feedstocks, generally from core metabolism, and varied enzyme classes lead to hundreds of thousands of metabolites of diverse structure and function. The sheer number and variety of these compounds present challenges for the elucidation of structure that is essential for rigorous metabolic pathway dissection. Conversely, the ability to use in vitro pathway reconstruction to test hypotheses regarding pathway evolution is a strength of evo-metabolism.

Acylsugar metabolism has become an exemplary system for understanding both the general principals and specific mechanisms by which enzymes are repurposed and combined to generate phenotypic diversity. Here, we describe three characteristics of S. nigrum acylsugar diversity and metabolism not previously reported for other species. First, although only acylhexoses were detected, they include a mixture of both inositol and glucose esters (Figs. 1 and 2). Second, an acylglucose acyltransferase is involved in making triacylglucoses (Fig. 5). Last, the acylglucose pathway intermediates are produced from a neofunctionalized invertase, which evolved independently from the recently reported wild tomato SpASFF1 triacylsucrose β-fructofuranosidase (Figs. 5 and 6). This work was facilitated by integration of LC-MS and GC-MS combined with 2D NMR methods, bypassing a requirement for metabolite purification (Fig. 1).

Integrative approaches to elucidate the S. nigrum acylglucose biosynthetic pathway

We present four lines of evidence that S. nigrum di- and triacylglucoses are synthesized by SnASFF1 and SnAGAT1 from the diacylsucrose intermediates produced by sequential reactions of SnASAT1 and SnASAT2. First, we reconstructed the biosynthetic pathway in vitro using purified enzymes, sucrose, and acyl-CoAs (Figs. 4 and 5). The acylglucoses generated are chromatographically indistinguishable from those observed in S. nigrum, while intermediates of the reconstructed pathway have acylation positions consistent with the in vivo final products. Second, genetic evidence supports the involvement of these enzymes in S. nigrum acylglucose biosynthesis (Figs. 3 and 8): VIGS silencing of the BAHD acyltransferases SnASAT1 and SnASAT2 reduces total acylglucoses; SnAGAT1 reduces total triacylglucoses, while SnASFF1 silencing causes accumulation of diacylsucrose pathway intermediates. Third, we cross-validated that the major in vitro diacylsucrose intermediates produced by the reconstructed pathway behave chromatographically and biochemically identically to in vivo diacylsucrose intermediates purified from VIGS-SnASFF1 lines (Figs. 4D, 5, and 7C). Notably, SnASFF1 converted both in vivo– and in vitro–synthesized diacylsucroses to diacylglucose products that can be converted to triacylglucoses by SnAGAT1. In addition, both the di- and triacylglucose anomers produced are indistinguishable from those extracted from S. nigrum. While this biochemical and genetic evidence provided strong support for the reconstructed pathway, observation of mono- and diacylated in vitro rearrangement products under pH >6 and elevated temperature (figs. S10 and S13) led us to seek a fourth line of evidence for pathway validation using reverse BAHD assays.

We compared reverse enzymatic products of in vitro and in vivo intermediates to verify the relevance of all the in vitro–produced intermediates. SnASAT2-reverse activity deacylates the in vivo, and major in vitro, diacylsucroses to monoacylsucroses that are indistinguishable from those obtained from SnASAT1 forward assays. All of these monoacylsucrose products are substrates for SnASAT1 reverse activities, supporting the hypothesis that SnASAT1 acylates at position 4. The lack of SnASAT2 reverse enzyme activity with in vitro minor diacylsucrose is consistent with the hypothesis that the rearranged product is not a biosynthetic intermediate. This result aligns with NMR analysis suggesting that the 6-position acylation of the medium-length chain on the minor monoacylsucrose products is not observed on S. nigrum acylglucoses. Hence, we conclude that the minor in vitro artifacts are not intermediates of the S. nigrum acylglucose biosynthetic pathway, despite their similar, but not identical, retention times and mass spectra.

The reverse activity also enabled us to facilitate pathway discovery. Using abundant plant-derived products, we found the first reported acylglucose acyltransferase, SnAGAT1, acting as the last step in S. nigrum triacylglucose biosynthesis (Fig. 7A). This acetyltransferase activity was later verified by forward enzyme assays with acetyl-CoA and in vitro diacylglucoses produced by the multistep, SnASAT1 + SnASAT2 + SnASFF1, reactions (Fig. 5B). SnAGAT1 is the fourth phylogenetically related acetyltransferase in acylsugar biosynthesis, falling in the same BAHD cluster with tomato SlASAT4, S. quitoense triacylinositol acetyltransferase (TAIAT), and S. sinuata SsASAT5 (Fig. 3A). Deploying BAHD reverse enzyme assays accelerated SnAGAT1 identification and allowed in vitro pathway reconstruction in the nonmodel species. To our knowledge, the S. nigrum acylglucose biosynthetic pathway is the first completely in vitro reconstructed acylhexose pathway in a nonmodel organism without the powerful genetic resources that enabled the discovery of S. pennellii acylglucose biosynthetic pathway.

Convergent evolution of acylsucrose hydrolyzing enzymes

Phylogenetic analysis of SnASFF1 and restricted accumulation of acylhexoses across the Solanaceae is consistent with the hypothesis that acylglucose production arose independently in the Old World black nightshade S. nigrum and New World wild tomato S. pennellii (Fig. 6). One strong line of evidence is that the two ASFF enzymes are members of different clades within the β-fructofuranosidase gene tree (Fig. 6). Second, SpASFF1 and synthesis of its P-type triacylsucrose substrates (15) are restricted within a subclade of the tomato Solanum group (12), consistent with this being a derived trait within the New World Solanum. The independent evolution of acylhexose biosynthesis raises the intriguing question of whether there are phenotypic advantages of producing acylsugars with a glucose core. Similarly, the observation of mixed acylsucroses and acylglucoses in S. pennellii and mixed acylglucose and acylinositols in S. nigrum begs the question of whether there are synergistic effects of acylglucoses with other acylsugars.

Despite co-option from distinct lineages, the two independently evolved ASFFs are both trichome-enriched, neofunctionalized GH32 β-fructofuranosidases, which lack the canonical amino acid sequence DDTK sucrose-binding pocket of Arabidopsis cell-wall invertase 1 (AT3G13790; fig. S4) (40–42). This is consistent with the observation that neither ASFF uses unacylated sucrose as substrate (fig. S12C) (15). In addition, both ASFFs are expressed in the same tip cells as their acylsucrose substrates (15). It will be interesting to learn whether invertase-like enzymes evolved independently in the acylglucose-producing Datura and Nicotiana species (23, 33). The diversity of apparent in vivo substrates for both ASFFs, diacylsucroses for SnASFF1 and triacylsucroses for SpASFF1, and the presence of acylsucrose intermediate accumulation in S. pennellii but not in S. nigrum also present opportunities to better understand invertase enzyme structure and function.

Streamlined analytical approaches for metabolite annotation

It is remarkable that we continue to discover acylsugar metabolic variations as new Solanaceae species are analyzed. Because biosynthetic pathway dissection requires knowledge of product and intermediate structures, the process of characterizing these molecules can be a bottleneck. We developed a structural analysis pipeline that reduces the cost and time involved in sample purification. The presence of three distinct acylhexose classes—diacylglucoses, triacylglucoses, and triacylinositols—in S. nigrum presented a challenging case study for this integrative approach. We designed the pipeline to leverage the substantial chemical shift library assembled from acylsugar studies in other solanaceous plants (10, 11, 13, 26, 29, 36).

While the pipeline performed as expected for the most abundant triacylsugars in S. nigrum, the many overlapping NMR signals in total trichome acylsugar extracts led us to perform partial purification using silica gel chromatography. We obtained clear signals of S. nigrum triacylglucoses from the less polar fraction, while 2D HSQC-TOCSY successfully separated proton correlation signals from diacylglucoses and triacylinositols in the high polarity fraction. With the simple partial purification, our NMR approach provided sufficient information to address key characteristics of the enzymes involved in acylsugar biosynthesis without the need to purify individual compounds: (i) recognition of different acylated sugar cores in mixtures and (ii) assignments of ester linkage positions on sugar cores. This approach extends the use of 2D ¹H-¹³C HSQC NMR spectra beyond previously published characterization of lignin structures (43). Limiting the need for purification is particularly beneficial for investigating anomers and isomers differing in acyl chain branching patterns, where overlapping retention times can frustrate attempts at large-scale purification to homogeneity. As acylsugars are sporadically found in plants of the Martyniaceae, Rosaceae, Geraniaceae, Caryophyllaceae, and Brassicaceae, our approach can further assist with large-scale screening for sugar esters beyond Solanaceae (44–48).

Cataloging the vast complexity of specialized metabolites in the plant kingdom will benefit from streamlined structure annotation strategies to facilitate biosynthetic pathway elucidation and natural product structure-function analysis. The integrated metabolite annotation approach described in this work has the potential to efficiently address this issue beyond acylsugars. The multidimensional NMR strategy exploits two acylsugar characteristics shared among other specialized metabolites: (i) a similar core structure within the group of structurally diverse molecules and (ii) the availability of a chemical shift database accumulated from past studies of related molecules. Development of pipelines for different classes of metabolites that use this approach to annotate mixtures of metabolites would accelerate the discovery of specialized metabolites.

In addition to illustrating principles by which the evolution of form occurs, our deepening understanding of specialized metabolic innovation has practical benefits. Metabolic engineering continues to be a trial-and-error enterprise (49). Understanding the evolutionary mechanisms leading to synthesis of biologically active metabolites provides new enzymes, transporters, and transcriptional regulators for the synthetic biology toolkit. Documenting recurring themes, such as the independent recruitment of invertases in multiple acylglucose-producing Solanum lineages, may reveal robust strategies for engineering synthesis of these protective compounds and other structurally similar compounds of economic or pharmaceutical importance.

MATERIALS AND METHODS

Plant material

Seeds of S. nigrum were obtained from the New York Botanical Garden (14). For germination, seeds were treated with half-strength bleach for 5 min and rinsed six times in deionized water before sowing on moist filter paper in petri dishes at 28°C. Seedlings were transferred to Jiffy-7 peat pellets (Jiffy Products of America, OH, USA) upon germination. Plants used for analysis were grown in a growth chamber at 22°C under a 16-hour photoperiod (photosynthetic photon flux density, 70 μmol m⁻² s⁻¹) with relative humidity set to 50%.

Acylsugar metabolite annotation

The acylsugar extraction protocol for LC-MS analysis is available in protocols.io at https://dx.doi.org/10.17504/protocols.io.xj2fkqe. As described previously by Leong et al. (15), leaf surface acylsugar extraction was carried out by gently agitating a single leaflet in 1 ml of acetonitrile:isopropanol:water (3:3:2 v/v with 0.1% formic acid and 1 μM telmisartan; isopropanol was obtained from J.T.Baker, Phillipsburg, NJ, USA; all others were obtained from Sigma-Aldrich, St. Louis, MO, USA) for 2 min. Telmisartan acts as an internal standard for high-performance LC (HPLC). The extraction solvent was collected and stored in 2-ml LC-MS vials at −20°C. All extracts were analyzed on LC-MS (Waters Corporation, MA, USA) using 7-, 30-, or 110-min LC gradients on an Ascentis Express C18 HPLC column (10 cm by 2.1 mm, 2.7 μm) (Sigma-Aldrich, St. Louis, MO, USA), which was maintained at 40°C. The 110-min method minimized chromatographic overlap in support of metabolite annotation. The HPLC-MS methods are described in protocols.io and table S10.

Acylsugars structures were inferred by positive and negative mode MS collision–induced dissociation as described previously (11, 26, 29, 34). In short, co-eluting fragments generated by collision-induced dissociation in negative ion mode were compared among three energy potentials to confirm acylsugar metabolites. Annotation strategy, thresholds, and confidence levels were described in detail in Supplementary Text.

Acylsugar acyl chain composition was determined by ethyl ester derivatization and subsequent GC-MS analysis. To create fatty acid ethyl esters, acylsugar samples were saponified and transesterified as previously described by Ning et al. (50) with some modification. In short, a leaflet was immersed in 1 ml of acetonitrile/isopropanol (1:1 v/v) for 2 min with gentle agitation. The extracts were transferred, evaporated to dryness under flowing air, and redissolved in ethanol with 300 μl of 21% (v/v) sodium ethoxide (Sigma-Aldrich, St. Louis, MO, USA). The reaction was gently vortexed every 5 min for 30 min in the fume hood. Four hundred microliters of hexane with tetradecane (55 μg/ml; internal standard; Sigma-Aldrich, St. Louis, MO, USA) was added to the reaction mixture for phase separation. The hexane layer was transferred and extracted three times using 500 μl of saturated aqueous sodium chloride each time. The final hexane phase (~100 μl) was transferred to autosampler vials with glass inserts and analyzed by capillary GC-MS on an Agilent J&W DB-5 column [10-m, 0.1-mm (inside diameter) fused silica column with a 0.34-μm-thick stationary phase; Agilent]. One microliter of each hexane extract was injected using splitless mode. The gas chromatography program is described in table S10. All compounds were analyzed using an Agilent 6890N gas chromatograph/Agilent 5975B single quadrupole MS using 70-eV electron ionization. Fatty acid ethyl esters were identified by library search against Agilent RTL library and compared with commercially available standards.

For total extract NMR profiling, acylsugars were extracted from 15 3- to 4-week-old S. nigrum plants by dipping aerial tissue into 500 ml of ethanol with 0.1% (v/v) formic acid with gentle agitation in a 1-liter beaker. The ethanol was evaporated under reduced pressure using a rotary evaporator with a warm water bath (40°C). The dried residue (~5 mg) of S. nigrum surface plant extracts was dissolved in CDCl₃ (99.8 atom % D, Sigma Aldrich, St. Louis, MO, USA) and transferred to solvent-matched (5 mm) NMR Shigemi tube (Shigemi Co. LTD., Tokyo, Japan) for analysis. ¹H, HSQC, HMBC, and TOCSY spectra were recorded using the Avance 900 MHz spectrometer (Bruker, Billerica, MA, USA) equipped with a TCI triple resonance probe at the Michigan State University Max T. Rogers NMR facility (see tables S2 to S4 for more details). All spectra were referenced to nondeuterated CDCl₃ solvent signals [δH = 7.26 (s) and δC = 77.2 (t) ppm].

For silica gel column chromatography, S. nigrum surface plant ethanol extract was concentrated in vacuo (~8 mg), dissolved in ethyl acetate/hexane (1 ml, 1:1 v/v) and loaded to the silica gel column (100 g, 200 to 425 mesh, 60 A; Jade Scientific Inc., Westland, MI, USA) packed with ethyl acetate/hexane (200 ml, 1:1 v/v) slurry. The compounds were eluted using mobile phase ethyl acetate/hexane (400 ml, 2:1 v/v in 0.02% acetic acid) with compressed air flash column chromatography. After thin layer chromatography analysis (4:1 ethyl acetate/hexane, p-anisaldehyde stain), fractions F_15–17 and F_24–28 were combined, concentrated in vacuo to give fraction 1 (~2 mg, Rf = 0.7) and fraction 2 (~2 mg, Rf = 0.3). Combined samples were analyzed on NMR as described above.

Gene identification and phylogenetic analysis

All transcript assemblies and expression data are by Moghe et al. (14) and were analyzed using Geneious R9.1.8 and R, respectively. To identify BAHD candidates, BLAST and TBLASTN searches were performed using ASAT sequences from tomato and S. sinuata. The TBLASTN hits were parsed to include only those with HXXXD motifs and DFGWG-like motif (one mismatch). The trichome-stem expression data of the remaining sequences were obtained from expression data available in the study by Moghe et al. (14), whereas a length of 400 to 500 amino acids and the relative positions of the two motifs on these sequences were confirmed manually (39). To identify ASFF candidates, 23 invertase-like sequences from tomato were used to BLAST and TBLASTN search S. nigrum transcriptome. The obtained sequences were checked for trichome enrichment and WXNDPNG, RDP, EC, and DXXK motifs as described above.

Phylogenetic reconstructions were performed using MEGA X (51). To obtain the BAHD tree, BAHD candidate sequences were aligned with several characterized ASATs (9–11, 14) and several other BAHD sequences by D’Auria (39) using the MUSCLE algorithm under default parameters. To obtain the invertase tree, invertase-like proteins from S. lycopersicum, S. pennellii, and S. nigrum were aligned with the same algorithm and settings. Maximum likelihood estimations were performed with Jones-Taylor-Thornton + G + I with five rate categories as the substitution model. One thousand bootstrap replicates were performed using partial deletion (30% gaps) for tree reconstruction.

Transient expression and purification of BAHD protein

All Sanger DNA sequencing confirmations in this study were performed with the indicated sequencing primers at the Research Technology Support Facility Genomics Core, Michigan State University, East Lansing, MI. All primer sequences are listed in table S9.

Recombinant BAHD proteins were generated using E. coli as the host for enzyme assays. The full-length open reading frames of c63608_g1, c65670_g1, c71009_g1, c53868_g2, c60145_g1, and c65306_g1 were amplified from S. nigrum young leaf and peduncle cDNA and cloned into pET28b(+) (EMD Millipore, MA, USA) using Bam HI and Xho I restriction sites and 2× Gibson Assembly Master Mix (NEB, Ipswich, MA, USA) according to the manufacturer’s instructions. The assembled constructs were transformed into BL21 Rosetta (DE3) cells (EMD Millipore, MA, USA) and submitted for Sanger sequencing.

Protein expression was carried out as described before (11, 16). In short, 1 liter of Luria-Bertani (LB) media with kanamycin (50 μg ml⁻¹) and chloramphenicol (33 μg ml⁻¹) was inoculated 500:1 with an overnight culture obtained from the bacterial strain with the desired construct. After incubating at 37°C and 225 rpm and reaching OD₆₀₀ (optical density at 600 nm) between 0.5 and 0.6, large-scale cultures were chilled on ice for 20 min before a final concentration of 50 μM isopropylthio-β-galactoside was added. Cultures were incubated at 16°C and 180 rpm for 16 hours before cell pellets were harvested by centrifuging at 4000g for 10 min under 4°C. The following protein extraction steps were also performed on ice or at 4°C. The cell pellets were resuspended in 25 ml of extraction buffer [50 mM NaPO₄, 300 mM NaCl, 20 mM imidazole, and 5 mM 2-mercaptoethanol (pH 8)] by vortexing and submitted to 8 cycles of sonication (30 s on ice with 30-s intervals for cooling). The obtained cellular extracts were centrifuged twice at 30,000g for 10 min to obtain clear supernatant. Nickel–nitrilotriacetic acid resin (QIAGEN, Venlo, Netherlands) was washed three times and resuspended in 1 ml of extraction buffer before being incubated with the centrifuged extracts at 4°C for 1 hour with nutation. After removing the supernatant by centrifuging the slurry at 3200g for 5 min, the resins were transferred to a gravity flow column (Bio-Rad Laboratories, Hercules, CA, USA) and washed with three column volumes of wash buffer [50 mM NaPO₄, 300 mM NaCl, 40 mM imidazole, and 5 mM 2-mercaptoethanol (pH 8)]. Two milliliters of elution buffer [50 mM NaPO₄, 300 mM NaCl, 3 M imidazole, and 5 mM 2-mercaptoethanol (pH 8)] was added to the column and allowed to incubate with the resin for 1 min. The elutes were then diluted into 15 ml of storage buffer (extraction buffer without imidazole), concentrated using 10-kDa centrifugal filter units (EMD Millipore, MA, USA) until 1000-fold dilution. A final volume of 40% (v/v) glycerol-elution solution was prepared and stored at −20°C. Immunoblot with the anti–His-antibody conjugated to peroxidase (BMG-His-1 monoclonal antibody; Roche, Basel, Switzerland) was used to confirm the presence of enzymes.

Transient expression and purification of SnASFF1 protein

ASFF1 protein expression and purification was carried out exactly as described before (15). SnASFF1 full-length open reading frame was amplified from S. nigrum young leaf and peduncle cDNA using c70979_Fw and c70979_Rv primer and cloned into the pEAQ-HT vector (52) using Nru I–HF and Sma I restriction sites and 2× Gibson Assembly Master Mix (NEB, Ipswich, MA) according to the manufacturer’s instructions. The completed vector was subsequently transformed into Agrobacterium tumefaciens LBA4404 cells. Single colonies from LB agar plates with rifampicin (50 μg ml⁻¹) and kanamycin (50 μg ml⁻¹) were used to inoculate 50 ml of 1% yeast extract, 1% peptone medium with the same antibiotics in the same concentration. After overnight incubation at 28°C and 300 rpm, culture was harvested and washed in 50 ml of buffer A [10 mM MES (Sigma-Aldrich, St. Louis, MO, USA) at pH 5.6 and 10 mM MgCl₂] before being resuspended to a final OD₆₀₀ = 1.0 with buffer A with 200 μM acetosyringone (Sigma-Aldrich, St. Louis, MO, USA). The suspension was incubated at room temperature with gentle rocking for 4 hours and then infiltrated into fully expanded leaves of 6-week-old N. benthamiana plants using a needleless 1-ml tuberculin syringe. Infiltrated leaves were harvested, deveined, and flash-frozen in liquid nitrogen after 7 to 8 days. Tissue was powdered under liquid nitrogen and added to 140 ml of ice-cold buffer B [25 mM 3-[4-(2-hydroxyethyl)piperazin-1-yl]propane-1-sulfonic acid at pH 8.0, 1.5 M NaCl, 1 mM EDTA with 2 mM dithiothreitol (DTT), 1 mM benzamidine, 0.1 mM phenylmethanesulfonyl fluoride, 10 μM trans-epoxysuccinyl-l-leucylamido(4-guanidino)butane (E-64), and 5% (w/v) polyvinylpolypyrrolidone; all reagents were obtained from Sigma-Aldrich (St. Louis, MO, USA) except DTT, which was obtained from Roche Diagnostics (Risch-Rotkreuz, Switzerland)]. The mixture was stirred for 4 hours at 4°C before filtered through six layers of Miracloth and centrifuged at 27,000g at 4°C for 30 min. After being passed through a 0.22-μm polyethersulfone filter (EMD Millipore, Billerica, MA), the supernatant was loaded onto a HisTrap HP 1-ml affinity column and eluted using a gradient of 10 to 500 mM imidazole in buffer B using an ÄKTA start fast protein liquid chromatography module (GE Healthcare, Uppsala, Sweden). Fractions were analyzed by SDS–polyacrylamide gel electrophoresis, and the presence of SnASFF1-HT was confirmed by immunoblot using the BMG–His-1 monoclonal antibody (Roche, Basel, Switzerland). Purified SnASFF1 proteins were stored in 100 mM sodium acetate (pH 4.5) with 40% (v/v) glycerol at −20°C.

Enzyme assays

Unless otherwise specified, all BAHD forward enzyme assays were performed by incubating purified recombinant proteins in 60 μl of 100 mM ammonium acetate (pH 6.0) buffer with 100 μM acyl-CoA and an acyl chain acceptor [unmodified sugar or purified or fractionated acylsugar acceptors in an ethanol:water mixture (1:1 v/v)] at 30°C for 30 min. After the incubation, 2 volumes of stop solution [acetonitrile:isopropanol (1:1) with 0.1% (v/v) formic acid and 1 μM telmisartan as internal standard] was added to the assays and mixed. Reactions were centrifuged at 17,000g for 5 min, and the supernatants were stored at −20°C after being transferred to LC-MS vials. SnASFF1 enzyme assays were carried out in a similar manner in the absence of acyl-CoAs. For BAHD reverse enzyme assays, 100 μM free CoA was used in place of acyl-CoAs. For negative controls, enzymes boiled at 95°C for 10 min were used in place of active enzymes. Methods used to determine the apparent K_m value for different substrates were performed as previously described (11). Tested sugars; free CoA; and nC10-, nC8-, iC5-, and iC4-CoAs were obtained from Sigma-Aldrich, St. Louis, MO, USA, whereas iC10-CoA and aiC5-CoA were produced following method described by Kawaguchi et al.(53).

Purification of S1:10 and S2:14

Purifications were performed using a Waters 2795 Separations Module (Waters Corporation) and an Acclaim 120 C18 HPLC column (4.6 mm by 150 mm, 5 μm; Thermo Fisher Scientific, Waltham, MA, USA) with a column oven temperature of 30°C and flow rate of 1 ml/min. For S1:10 purification, the mobile phase consisted of water with 0.1% (solvent A) and acetonitrile (solvent B). For in vivo S2:14 purification, two purification methods where solvent B consists of acetonitrile and methanol, respectively, were used in a sequential manner to enhance purity of compounds. The HPLC methods used for separating S1:10 and S2:14 are described in table S10. Fractions were collected using a 2211 Superrac Fraction Collector (LKB Bromma, Stockholm, Sweden).

VIGS analysis

VIGS target regions were selected to have a low chance of altering expression of trichome expressed nontarget genes (14). These fragments were amplified from S. nigrum young leaf and peduncle cDNA using primers listed in table S9 and were cloned into pTRV2-LIC as previously described (16). In short, the Pst I–HF–linearized pTRV2-LIC vector and all polymerase chain reaction (PCR) fragments were separately incubated in 5-μl T4 DNA polymerase reactions with 5 mM 2′-deoxyadenosine 5′-triphosphate or 3′-deoxythymidine 5′-triphosphate, respectively. The reactions were incubated at 22°C for 30 min and then 70°C for 20 min and then stored on ice. Two microliters of desired PCR reaction were mixed with 1 μl of vector reaction and incubated at 65°C for 2 min, followed by 22°C for 10 min. Sequenced constructs and pTRV1 were transformed into A. tumefaciens strain GV3101 using the protocol described previously (15).

The vacuum infiltration protocol was adapted from the work of Hartl et al. (54). Twenty-five milliliters of LB medium with kanamycin (50 μg ml⁻¹), rifampicin (50 μg ml⁻¹), and gentamicin (10 μg ml⁻¹) was inoculated with a single colony of the respective Agrobacterium strain from plate. After overnight incubation at 28°C and 225 rpm, cells were harvested by centrifugation at 3200g for 10 min and resuspended to a final OD₆₀₀ = 1 with the induction media [10 mM MES (pH 5.6) and 10 mM MgCl₂]. An equal volume of pTRV1 suspension was mixed with different pTRV2-LIC construct suspension. Acetosyringone (200 μM) was added to each mixture before the suspension was incubated in the dark at room temperature with gentle rocking for 3 hours. Seven- to 10-day-old S. nigrum seedlings were carefully transferred from the petri dishes to the bacteria solution. Vacuum was applied in a desiccator for 2 min, followed by slow release of the vacuum. Infiltrated seedlings were kept for 3 days under indirect light, then planted in Jiffy-7 peat pellets, and grown in a climate chamber as described above. Metabolite and tissue samples were harvested approximately 3 weeks after infiltration and phytoene desaturase silencing efficiency was monitored.

The QuanLynx function in MassLynx v4.1 (Waters Corporation, MA, USA) was used to integrate extracted ion chromatograms from untargeted LC-MS data as described previously (26). In short, all quantifications were performed using extracted ion chromatograms of the mass/charge ratio (m/z) value for the relevant [M + NH₄]⁺ or [M + HCOO]⁻ adduct ions using a mass window of m/z of 0.05. Isomeric forms of acylsugars (including anomers) were quantified in corresponding groups as shown in table S1. The retention time window was adjusted for each compound to include all isomers. Peak area of telmisartan was quantified and used as an internal reference. Then, acylsugar quantities (per milligram) were calculated by normalizing peak areas to the internal standard peak area and dry leaf weight, whereas acylsugar ratios were calculated by comparing peak areas over internal standard peak area of acylsugars of interest.

qPCR analysis

RNA was extracted with the RNeasy Plant Mini Kit including On-Column DNase Digestion (QIAGEN, Venlo, Netherlands) according to the manufacturer’s instructions. RNA was quantified with a NanoDrop 2000c instrument (Thermo Fisher Scientific, Waltham, MA, USA). cDNA was synthesized using 1 mg of the isolated RNA and SuperScript II Reverse Transcriptase (Invitrogen, Carlsbad, CA, USA). The cDNA samples were diluted 40-fold [10-fold initial dilution and 4-fold dilution into quantitative PCRs (qPCRs)]. qPCRs (10 μl) were created with SYBR Green PCR Master Mix (Thermo Fisher Scientific, Waltham, MA, USA), and primers were used at a final concentration of 200 nM. RT_SnASAT1_F/R, RT_SnASAT2_F/R, RT_SnASFF1_F/R, RT_SnAGAT1_F/R, RT_Actin_1_F/R, and RT_Actin_3_F/R primers were used to detect SnASAT1, SnASAT2, SnASFF1, SnAGAT1, ACTIN1, and ACTIN3 transcripts, respectively (table S9). Reactions were carried out with the QuantStudio 7 Flex Real-Time PCR System (Applied Biosystems) by the Michigan State University RTSF Genomics Core. The following temperature cycling conditions were applied: 50°C for 2 min, 95°C for 10 min, and 40 cycles of 95°C for 15 s and 60°C for 1 min. Relative expression of SnASAT1, SnASAT2, SnASFF1, and SnAGAT1 was calculated with the ΔΔCt method (55) and normalized to the geometric mean of ACTIN1 and ACTIN3 transcript levels. The mean expression values of the transcripts in the control plants were used for normalization. Three to four technical replicates were used for all the qPCRs.

Statistical analysis

All statistical analyses were performed using the “stats” R package (56). Welch two-sample t tests were executed on metabolites and transcript abundance data using the “t.test” command. The power of these analyses was determined using the “power.t.test” function.

Supplementary Materials

This PDF file includes:

Supplementary Text

Figs. S1 to S18

Other Supplementary Material for this manuscript includes the following:

Tables S1 to S11

View/request a protocol for this paper from Bio-protocol.